eles %

Vena ay

tins om

arSbyteerk iene dk

AWG. Metal edy frena on Meals As

AteWetls Bailte

656 coro es

Weitere ne Reed

eben tasteky 2 on,

HOA oe He Me he Mi,

iw ASV SAWS ADB Uh. ha Seay dyee

(WES ite <a hy ihe MOM Tce (urn, teal, 2s MOK Me Pt da Ma Gh fen

MMe eS Te Mtwto de = iat re he Corey

nALiile ili MeM OMe rte dea fea yh INAV ANN CADE Toledo IN

Aye EN he Itt mat MEMOS LEY Bouya, MEFS (aya

“Fe TEMG R Ge Ig I ay HU Gy,

Hele in MRD ras

AN ADE Di Neg sy) thew

Ma ae 6

+ sell Goth de Oe thks

fe EEN UM Mite

Nes

One Etter

» (thes Fy,

nile Cnet he tie WN EA eau i

inte hie Hdesihs Biattyare tN

edi

bar

te DNalp

ta ih HS Pe re

Nite Be ue ay 6

PLLA Thea YI lige

ON MOR ate, Aegis

site Teoh we

fa snnnee te

Deere

Te Boni de

lps the i

a betas 2, ors

Be Bee IN Mh ge

Se ee

Ae SEE se

“Ames Sar

OF ko he cts Fidel He hate ne Hhe tein

CAST ie eet diy dd ie

SPAN

PEO ely

in hte ie Tete

LSA He tay

Ak eR ea Kr S

ee Re terte INAS adte Oe De Madara

ae TEL Me vee By ee ltosdhy Ne ote, PFO YIN Peete

EA I Gots Weal itetteopere on

Mat Steatham A I5 Oaps de Boon ey ay

ALIN MAY Nek Oe Bremen,

Sp PL yenner

TEAL bnetuti tay) i

REDE TMNT Mt sy Coney

Pets Sat IN Ag

Matilda

COANE IX og

WOMEN TF go

Wort ey de

ce Rartighrt Wel i

toe BAN Fanaa

PEA A LENIN A Bie Beha

were Pye

Buordenter

Cee

Ie AIS Le SW PL Oe hee

STEM Nei le Se

Sierra,

Ha DN low Toth NOE eth trou

Pe Dede Wet diay FAN fst i\aN,

ena attinns LACT re 4 NUE

Dae te MN Le

KS Pci 7 SMe Rein sak am

He MNAN ANE Bons ter .

Bed Tele ath FOIE

Aes ees

MePnte heats

fet NON Tene

SIN recta Eh ce oe a PN Na Wn toe

SFO er i Nate

elt Uae

OW ANG DD

sree et

MWA OW AN the

NEB

Aya

Noth ere Our am,

HM Byler

serene

MEM tee ys gy

hte

SNe

1a fyi,

SAM

mem La,

Cre

A eee ene OT WS

CM Geti

aaatvnege

Dred eet

DEAN AES the

Phan givas

AE by Pinata sh

45 "

I POE ety

mes

PE Doody ty lore fuse

fre dNite

Nogthatte ony

a Cer Deas

car eeeiarer resins

. MOxy gies

PUI OMe

Ser cis

DAN 6

Fo Ate dN

ffir den

S Nanda

PUN itn

SWNT Geet

Tht pv tbe

Wentny SONI

iss

oth ener ens

DPW es Nn,

ASIN sth = hh ae

MAN gANSpai Ht Ne

ESM AY,

Bt oN.

ae bag

NAN NS ee I dN Sayh, te

Map ey TOW ea

Dogan iene erent

dot Th ed tye 8 wy . F <

oben,

han ge,

ite wh

red eae ne

AN gi AYE Dy,

Meee

Sees We Ye,

be deg

hee bad

SNMP netpet

NaF stella Me tage ath ut

What Ming

ree racer

Hides Arana

WS ant ml

Se i ay

sh aP et anh

SIs ty

~ UREN mein area Aon

1088 AY ah Noes tte

WVFoI Le we)t atnge

Nee

Co enn

DENTAL N Satie

beret ethic gt tt

here

ar home

dicen

OS eh

buy ou

eur een terrtas

eure ween

hath rk om

MONEE Heel Iy iba

RoeThnthil A sayy hs

hint

wis LEM Ming

Lind Me Walvis hate ae ahh

Bae wb

Fy ye My

Lat 2h eA AAd ot

He Mat ohh gi AL.

eM trict all 7

darigt ys

eae

aE bath

Mapes

tyke

gibt tthe

Pe Mebh aii ebay ir sch,

ME SMalt meat

Bln os S08 out

Sade

kes

eMail! iste

oeien es

Fab intemal) oe

SNe ane

RPS atari bop settee tly sie

thy

BPA hey}

aati

Sieh O98 Lara ern

Marhast a

Nesheve

Phat .

e erat

ainien wate

Riearees

. aN ny

NPE Dats A, a

FW Dio aay lb eve

PAE HT SAH AD eat ah

SENS sh Scecbw eye gd nt

Gree rey

ee ear

PMA dab w tages

Wa heh ait.

heh yeh

PN ot gb asthe

IN oN ahingn of So fo Alone

A Dal vale

hen age

Bante NIN sitaatt

Ean wr

Pe eth ad pre

reise

Woh aie get

Miele lib st

S phe do the nie

TN ede ia ot

Den Matin eat Sah ys?

PE ew ang

HPA tedeynbine gy

Neer

Pek

eT elle

wee

Carre res

YB GS Ut iathiat

AP ee Ue Muses ott

*S nid Nanna tstt a vey

ANC SP ahatwri eaten

pre

trl nt

Aeon

Khar

soles

PMMA hatte oF

Hehe aay

ways tte an

ewan ati

Sheth mttaer sZeee ae ree

Mirchibtha wi eet ee ciel Ae.

Se Mie slay AL ty aig. EEN y iM asl with wig it

WeDo allt aR Wt ee ay, TF ASM writ

Peat ra nal

VP aah

Nathan

a oa

Natl Perret!

ee RS

seta

Pie et

Soret as

Betcnging

Se esi “ oy

evry Medion

No typ eS ee ee

hah, pr teahe Tatar eee ie

. pian

AML ality te

Mathias he

MI as taal a

Ree ost

Ste IN ry,

boys

Madln tSce

Mota.

Arie

Peon

Coit dts ik vt

a nvie Lies

eae te

tall ats)

al ave ov Lang

este

Pre are

eo |. vt)

ANNALS OF THE ANNALE VAN DIE

SOUTH AFRICAN MUSEUM SUID-AFRIKAANSE MUSEUM

VOLUME 88 BAND 88

has.

- .

“

* =

’

ie P

i as

ony

5 J

= j

—.

i S.

* P

fo I ¢

«s

} »

x .

¥;

~ rs

- ry

f entender ae ;

. A he

Ss “| im

i j =

ae E

’ 7 %

i i

4 *

)

a Ff

; My

, a

ANNALS OF THE SOUTH AFRICAN MUSEUM

ANNALE VAN DIE SUID-AFRIKAANSE MUSEUM

VOLUME 88& BAND

4 S

“ouig wow WY

HE ERUSTEES OF THE DIE TRUSTEES VAN DIE

SOUTH AFRICAN MUSEUM SUID-AFRIKAANSE MUSEUM

CAPE TOWN KAAPSTAD

1982

SET, PRINTED AND BOUND IN THE REPUBLIC OF SOUTH AFRICA BY

THE RUSTICA PRESS (PTY.) LTD., WYNBERG, CAPE

DS4

LISTOF CONTENTS

Page

De VILuLiERS, H. & WILSON, M. L.

Human burials from Byneskranskop, Bredasdorp District, Cape Province, South

Arican(bublishcdwiunedOS2s)re aoa kas em ne Semmes eeix dee abe a oe 205

SCHWEITZER, F. R. & WILSON, M. L.

Byneskranskop 1. A Late Quaternary living site in the southern Cape Province,

SoutheAimicas (PublishedimnelOS2o\ ern loti: uc munatetee ova e ae FR eee awe ant 1

WILson, M. L. see DE VILLIERS, H.

WItson, M. L. see SCHWEITZER, F. R.

mes

—

Awe we

ni t

Tey ate

{ us

ane

= a

y ¥ u

> ! os

Ufo

} G

S a a =

’

rs a

4 /

’

fh ~

f ' bad

‘

P A.

ANNALS

OF THE SOUTH AFRICAN |

MUS

ISSN 0303-2515

KUM >

| CAPE TOWN

INSTRUCTIONS TO AUTHORS

1. MATERIAL should be original and not published elsewhere, in whole or in part.

2. LAYOUT should be as follows:

(a) Centred masthead to consist of

Title: informative but concise, without abbreviations and not including the names of new genera or species

Author’s(s’) name(s)

Address(es) of author(s) (institution where work was carried out)

Number of illustrations (figures, enumerated maps and tables, in this order)

(b) Abstract of not more than 200 words, intelligible to the reader without reference to the text

(d) Introduction

(e) Subject-matter of the paper, divided into sections to correspond with those given in table of contents

(f) Summary, if paper is lengthy

(g) Acknowledgements

(h) References

(i) Abbreviations, where these are numerous

3. MANUSCRIPT, to be submitted in triplicate, should be typewritten and neat, double spaced

with 2,5 cm margins all round. First lines of paragraphs should be indented. Tables and a list of

legends for illustrations should be typed separately, their positions indicated in the text. All

pages should be numbered consecutively.

Major headings of the paper are centred capitals; first subheadings are shouldered small

capitals; second subheadings are shouldered italics; third subheadings are indented, shouldered

italics. Further subdivisions should be avoided, as also enumeration (never roman numerals)

of headings and abbreviations.

Footnotes should be avoided unless they are short and essential.

Only generic and specific names should be underlined to indicate italics; all other marking

up should be left to editor and publisher.

4. ILLUSTRATIONS should be reducible to a size not exceeding 12 « 18 cm (19 cm including

legend); the reduction or enlargement required should be indicated; originals larger than

35 x 47 cm should not be submitted; photographs should be rectangular in shape and final

size. A metric scale should appear with all illustrations, otherwise magnification or reduction

should be given in the legend; if the latter, then the final reduction or enlargement should be

taken into consideration.

All illustrations, whether line drawings or photographs, should be termed figures (plates

are not printed; half-tones will appear in their proper place in the text) and numbered in a

single series. Items of composite figures should be designated by capital letters; lettering of

figures is not set in type and should be in lower-case letters.

The number of the figure should be lightly marked in pencil on the back of each illustration.

5. REFERENCES cited in text and synonymies should all be included in the list at the end of

the paper, using the Harvard System (ibid., idem, loc. cit., op. cit. are not acceptable):

(a) Author’s name and year of publication given in text, e.g.:

‘Smith (1969) describes...’

‘Smith (1969: 36, fig. 16) describes. . .’

“As described (Smith 1969a, 19695; Jones 1971)’

‘As described (Haughton & Broom 1927)...’

‘As described (Haughton et al. 1927)...’

Note: no comma separating name and year

Dagination indicated by colon, not p.

names of joint authors connected by ampersand

- et al. in text for more than two joint authors, but names of all authors given in list of references.

(b) Full references at the end of the paper, arranged alphabetically by names, chronologically

within each name, with suffixes a, b, etc. to the year for more than one paper by the same

author in that year, e.g. Smith (1969a, 19695) and not Smith (1969, 1969a).

For books give title in italics, edition, volume number, place of publication, publisher.

For journal article give title of article, title of journal in italics (abbreviated according to the World list o,

scientific periodicals. 4th ed. London: Butterworths, 1963), series in parentheses, volume number, part

number (only if independently paged) in parentheses, pagination (first and last pages of article).

Examples (note capitalization and punctuation)

BULLOUGH, W. S. 1960. Practical invertebrate anatomy. 2nd ed. London: Macmillan.

FISCHER, P.—H. 1948. Données sur la résistance et de le vitalité des mollusques. J. Conch., Paris 88: 100-140.

FiscHER, P.-H., DuvAL, M. & Rarry, A. 1933. Etudes sur les échanges respiratoires des littorines. Archs

Zool. exp. gén. 74: 627-634.

Koun, A. J. 1960a. Ecological notes on Conus (Mollusca: Gastropoda) in the Trincomalee region of Ceylon.

Ann. Maz. nat. Hist. (13) 2: 309-320.

Konn, A. J. 19606. Spawning behaviour, egg masses and larval development in Conus from the Indian Ocean.

Bull. Bingham oceanogr. Coll. 17 (4): 1-51.

THELE, J. 1910. Mollusca: B. Polyplacophora, Gastropoda marina, Bivalvia. In: SCHULTZE, L. Zoologische

und anthropologische Ergebnisse einer Forschungsreise im westlichen und zentralen Siid-Afrika 4: 269-270.

Jena: Fischer. Denkschr. med.-naturw. Ges. Jena 16: 269-270.

(continued inside back cover)

ANNALS OF THE SOUTH AFRICAN MUSEUM

ANNALE VAN DIE SUID-AFRIKAANSE MUSEUM

Volume 88 Band

June 1982 Junie

Part 1 Deel

“ 4 SN

@ VVVVWS

te NS

> S

ALi) NOVI lp

BYNESKRANSKOP, 1

A LATE QUATERNARY LIVING SITE

IN THE SOUTHERN CAPE PROVINCE,

SOUTH AFRICA

FRANZ R. SCHWEITZER

M. L. WILSON

Cape Town Kaapstad

The ANNALS OF THE SOUTH AFRICAN MUSEUM

are issued in parts at irregular intervals as material

becomes available

Obtainable from the South African Museum, P.O. Box 61, Cape Town 8000

Die ANNALE VAN DIE SUID-AFRIKAANSE MUSEUM

word uitgegee in dele op ongereelde tye na gelang van die

beskikbaarheid van stof

Verkrygbaar van die Suid-Afrikaanse Museum, Posbus 61, Kaapstad 8000

OUT OF PRINT/UIT DRUK

1 2(1=3\-528)) 3028 42508) taps) less 5s 7-9)

GC, b=pid)) 1-4), 82 S=2) aye 1013)

11(1-2, 5, 7, t-—p.i.), 15(4-5), 24(2), 27, 31(1-3), 32(5), 33, 36(2), 45(1)

EDITOR/REDAKTRISE

Ione Rudner

Kopieregnavrae aan die Suid-Afrikaanse Museum

ISBN 0 86813 021 4

Printed in South Africa by In Suid-Afrika gedruk deur

The Rustica Press, Pty., Ltd., Die Rustica-pers, Edms., Bpk.,

Court Road, Wynberg, Cape Courtweg, Wynberg, Kaap

BYNESKRANSKOP 1

A LATE QUATERNARY LIVING SITE

IN THE SOUTHERN CAPE PROVINCE, SOUTH AFRICA

FRANZ R. SCHWEITZER*

M. L. WILSON

South African Museum, Cape Town

(With 44 figures, 26 tables, and 1 appendix)

[MS accepted 23 December 1981]

ABSTRACT

The excavations at Byneskranskop represent the second phase in an investigation of the

ecology of the prehistoric inhabitants of the Agulhas region of the southern Cape.

The excavations revealed a possibly unbroken occupation sequence spanning the last

12 500 years, from the terminal Pleistocene to the early years of white settlement of the

sub-continent. They yielded quantities of artefactual and non-artefactual material that were in

most cases adequate to document clearly the changes that took place during the occupation of

the site.

As well as analysing the site data comprehensively, this report places an almost equal

importance on comparisons with data from other sites in the Cape coastal region and its

hinterland, since these are important for evaluating the chronology and nature of the changes

revealed by the site data.

It was possible to divide the occupation sequence into four main phases and two

intermediate ones, although the primary difference is between the content of the deposits

younger than about 6 000 B.p. and those older. These phases can be related to ‘Wilton’,

‘Albany’ and ‘Robberg’ industries, but the analyses show that the differences are quantitative

more than qualitative and that the changes in the various components of the excavated

assemblage were diachronous rather than synchronous. The inter-site comparisons also indicate

that the rate and nature of change were not consistent throughout the region during this period.

CONTENTS

PAGE

RntrOducitOnmteaar Ge eRe fee ata ee Serene = Ae aie cops oe 2 2

MMe Stead WS\EMVIROMMEM tc) 22 eres heer eee Ye cero dn. son set: ql

AV SIOCT ADI erate ce ein ses Se ae tie aaa ae gis ewe me es 9

SODA WELIS DET GES ch cera acelin ae arene ae 9

GCOLO UVM ras here antec hes at ese Meany Resa oe aG ew hoes 10

ID TPE IT EVES se 6d Bin ak tos Cle oe DES ie SRP oir ce a a ee ee 11

(CUNT EIS ig, Pe Se oo OO rn CE ERE IO RT are eee ee 11

WESC ta lO Mir er ett SG eee enh vn Emin so cis els ses 12

AUST ewes SON eens ede span Mess Pea hte s © id Gia twal Aiylasd) wus! ste 14

STOUTIVITEN RY ie ores bee Ge hte Wye ets Re oc ie ae Cn oe 15

* The principal author of this report died on 1 November 1979. The report has been

completed by Mr Schweitzer’s assistant, M. L. Wilson—Editor.

Ann. S. Afr. Mus. 88 (1), 1982: 1-203, 44 figs, 26 tables, 1 appendix.

2 ANNALS OF THE SOUTH AFRICAN MUSEUM

PAGE

Whe €XCavations 2502 226. 5. tence ed Se ee oe eee 16

Excavation: procedure: = 1). neal 2 eee ey er eee eee 16

Stratiorapmny =e i.0%. & csel ek EAR Ate ere Eee eS aa oe 16

Dating and: comelations. 4.3 cnon 0 ee ee ee eee ae Za)

Methods of analysis: 2123.0. octomiicue see eee rd ee ee are 24

SUMIMATY> (35 5:d.c akbar ee ene eee ote ee 24

The: artetact ‘assemblage aii 4a eran eee See eines cree ener Zr)

Stome-artefacts.:: . wie ae ce eee ane nee eee ne 75)

Bone artelacts, sn.n st panos nes hee oe eee ee ere eee 76

Marine shell-artefactsen 3 ts | eats cae oe are ee ea ue ee 82

Ostrichegs-shellrartetactsi eats oe Oe ca ee 89

POUR: Sooo. bho keg i ee ree ering, ear ere ae D2

Metall artefacts: 5 as oc: cies 3 le ieee ne eee ee a Sean un are 95

SUMIUMALY. 65,3 Geka di ooo epee lene SS Ie Cle ee an pone eee 95

umansantmaleanG sp lanitencniclini sree eee ae 97

Human Temains::%5., $4 ae, eee ee ees Mi aN oe ae 97

Miamatmals 5.2 )5.seteihd Rae cre Me ne oe ear ae 98

Shelf 2 eect. sate peters, ce Ma ens cao ameter ater tae Ae reece AE EN ce aa 106

| Gb ee Pe MCT ee NON ee tie eps tite Leth ey ere MMe i 116

ROCK. 1ODStER oh. ae Bes ete enc ets SRR Ace ee eee 118

AT GS 8 4sg hole tw aa Se a cee oe a era et ee na gee a ae 118

Reptiles! ity et hs gees ee caesar ate eh an Pac te Waren aces ee ee 118

Micromammals: palaeoenvironmental interpretation............. 120

Plant remains; bys: ovat cape weet ar sk cols ee ia een ee ane ee 122

SUIMMALY*4 6.57 Beptres) oe Reon ets wis eaiee te Sear ee 129

BNKwl inwelationsto other sitesm spate oe ae Se eee 130

Stonevartelactsie i) vat a tee Soo ee Gee ee Be eran Sih

BONS Artelacts rar sede eh ae ee els eles ates eee eerie 150

Miarnine:shellzantetacts:: <4 aa arrgs eer: ch cenieuers tea eee eee 151

Ostrichvese-shellantefacts:4.. 5 ss 5 see Ges eek ees 152

POUMCRY ih. oe 5h athe tele ta Bae as | ed ae ieee 152

Metaltartetactsnviews yt Ace te ae Ace oe eae ee ee tr ae 52

NMamunials «. 20:fs ue ne x Son eerie oer oe Ce or eee wees 153)

) 121) U1) ee ees eer ne ee Re MIS od ace Gianle Bare bio es 163

1S |e ee oe ee ere ge wa tone fo Mme ere en ne wa. aerators her Ty % c 170

Reptiles 4h sented sitet x pew Be ee Bras ro era er 174

General: summary: 2s jas? sevice orotate eet dae peat = eee 174

Acknowledgements: wade ac ee Ace t Wee a ee ee 183

References «cnstveaetics aah. Seater A ero acy gata eee 184

Appendix eR Germ ici renh Prete ina ea vue ae cave tae eae 189

Byneskranskop 1: mammals. Tables of the minimum numbers of

individuals represented by various skeletal parts.

INTRODUCTION

The Archaeology Department of the South African Museum has been

engaged since 1969 in systematic research in the Agulhas region, the southern-

most part of the African continent. The research programme commenced with

excavations in a cave at Die Kelders in the Hermanus district, on the eastern

shore of Walker Bay (Fig. 1). These excavations, which lasted until 1973,

exposed a late Pleistocene (Middle Stone Age) sequence separated by a deep

accumulation of archaeologically sterile sands from a relatively short late

Holocene (Late Stone Age) sequence that spanned about the first five centuries

of the present era.

BYNESKRANSKOP 1

UMOYS SI YoIyM ‘

“a SNYWHSNg

H1AdVZI14 1LdOd

TOM

WoogineyySW,

SIDNVYd “1S JdV>

<y SOOLWYD

[ODpseupuy

SING se od4) prog ur

BUG

Sad lawol

———OO———T CC)

0072 OS! 00! OS 0

Y ZLNOD

qInow SAD")

Torry Saiso]y) dg UOsSN Ava 1aSSOW,

TAY SSIOW rm

Ree ran

Sooo

ownts A

doysueiysoudg }da0xa poulpiapun oie sayis [eoIsojoaeysIY “1X9} oY} Ul poUONUsUI sant[eso] SuIMOYs oded UoYyNOs dy} Jo dey ‘T ‘SIy

of l

aks

SVHINOV 3dvD

NOISINYY INIOd YIDNVa

AVG Ya TW

d¥oasvasug Jepjey aq

QYOJNVISe dIINSINVH — INIOd 3d

dV)

NOG31V3@ poneucd

3 FSI fSUWOY

ADg sUOps0<) OP TOY

WVGN311aMS : re

YT APIDS =o

NMOL AdVD

"y SMno)

fo)

ono

v ~

Syn ea

-: ANNALS OF THE SOUTH AFRICAN MUSEUM

The Die Kelders excavations, of which only the Holocene sequence has as

yet been published (Schweitzer 1979), were important not only because they

yielded the then earliest dated evidence for the presence of pastoralists in

southern Africa, but also because of the evidence they provided of a much

earlier period of occupation. The deposits were, however, not entirely satisfac-

tory in view of the great discontinuity between the two phases of occupation

and the relative brevity of the Holocene occupation.

| Complementary research in the area was meanwhile being carried out on a

series of open sites by G. Avery (1976) of the Museum’s Archaeological Data

Recording Centre. Like the upper deposits at Die Kelders, these sites, located

from Hawston to Pearly Beach (Fig. 1), were all somewhat ephemeral late

Holocene accumulations.

Fig. 2. South-east aspect of Byneskranskop, showing the location of the main cave (1) and

other sites.

Since this work had made the archaeology of the coast fairly well known,

the next objective was to find a site in a different environmental zone, and

preferably one that contained a sequence of deposits sufficient to provide a link

between the M.S.A. sequence at Die Kelders and the L.S.A. occurrences at

that site and along the coast.

This second phase of the research programme was begun in 1973 when a

test excavation was carried out in a cave at Byneskranskop in the Bredasdorp

district (Fig. 1), some 10 km south-east of Die Kelders and situated at the

junction of the coastal plain and its hilly hinterland.

BYNESKRANSKOP 1 5

Interest in the site was stimulated by R. G. Klein, who had examined

faunal material from ‘Eyre’s Cave’ acquired in 1962 from the Medical School of

the University of Cape Town and now housed in the collections of the

Museum’s Cenozoic Palaeontology Department (accession nos SAM—PQ1860-

73). Included in the collection were the remains of two bovid species con-

sidered to have become extinct by the end of the Pleistocene, a giant buffalo,

Pelorovis sp., and a giant alcelaphine, Megalotragus sp. (Klein 1974: 273). The

presence of these animals suggested that a sequence of deposits might be found

in the area that would provide the link between the two phases of occupation at

Die Kelders.

The precise provenance of the material studied by Klein is not known, but

is believed to have been the lowest cave in Byneskranskop. This had been

cleared out for use as a cellar in unawareness of the palaeontological and

archaeological potential of the deposit removed. Several human skeletons were

also found during the clearance and this prompted an investigation of the cave

near the top of the hill. Trenches were dug around the walls of the upper cave

and, apparently, more human remains recovered. The fate of these remains is

uncertain: some are known to have been re-buried in the grounds of the

property below the cave, but cannot now be located. It was also subsequently

learnt (P. Swart 1980 pers. comm.) that, several decades ago, deposits from

caves at Byneskranskop were removed as ‘guano’, and that the local school-

master had been in the habit of digging in the caves to augment the collection

of skulls he used for teaching purposes.

It is the upper cave, designated Byneskranskop 1 (BNK 1), that is the

subject of this report. The reference BNK 2 was given to a lower overhang,

from the floor of which micromammalian remains were collected by D. M.

Avery of the Museum’s Archaeology Department as part of a comparative

collection against which to study the microfauna from BNK 1, Die Kelders and

other sites (D. M. Avery 1979). BNK 3 is the designation of a cave on the

lower slope of the hill, in which excavations were carried out in 1979-80 in the

hope that these would provide amplification of the somewhat scanty deposits of

the final occupation phases of BNK 1, and which will provide the basis for a

separate report. The location of the three sites is shown in Figure 2.

An initial examination of BNK 1 raised doubts as to the suitability of the

site for archaeological investigation on account of the earlier, unauthorized,

digging. Most of the area around the walls had been dug and much of the

debris thrown towards the middle of the cave. Since an important consideration

in any excavation was to try to establish a link with the coastal sites it was

feared that the disturbance of the upper deposit might have made the site

incapable of providing the requisite correlation. However, clearing of some of

the surface debris in the centre of the cave indicated that the disturbance there

was largely superficial, and when an auger sounding showed that the cave

contained quite a considerable depth of deposit, it was decided to go ahead

with a programme of excavation.

6 ANNALS OF THE SOUTH AFRICAN MUSEUM

Pe ed

SEES

pe eye stenesyest

| Wolke

: a Test pit

Excavated to bedrock

athe

YY Excavated short of bedrock

Excavated in 1976 to bedrock

Excavated in 1976 to¢.50cm

RIPLINE

“~,

Fig. 3. BNK 1. Plan of the cave, showing the excavated area.

In March—April 1973 a test pit was excavated. The original area of the pit

was 1 X 1 m but at a depth of about 0,7 m a human burial was partly exposed

and it became necessary to enlarge the excavation to 1,0 x 1,5 m. Bedrock was

reached at a depth of about 3 m, and laboratory examination of the material

recovered indicated that a larger excavation was merited.

Because of a lack of clarity in the stratigraphy, partly explicable in the

upper deposit as the result of disturbance due to the burial, much of the test pit

was excavated in spits of varying thickness. The actual stratigraphy was made

clearer by subsequent study of the walls of the test pit, but because of a lack of

precise correlation between the spits and the actual stratigraphy, material from

the test pit is not included in the analyses that follow.

In January-February 1974 a stratigraphically controlled excavation was

undertaken, extending from the test pit towards the mouth of the cave. Three

1-metre squares were taken down to bedrock and a fourth excavated to about

90 cm short of bedrock.

After analysis of the material from the 1974 excavation, which formed the

basis for a preliminary report (Schweitzer & Wilson 1978), it was decided to

extend the excavation in order to increase the size of the samples, particularly

the fauna. Accordingly, in November-December 1976 two 1-metre squares

adjacent to the 1974 trench were excavated to bedrock and a further four in

BYNESKRANSKOP 1 7

front of the enlarged trench were excavated to a depth of about 50 cm in an

attempt to resolve certain problems regarding the upper deposits.

The location of the excavation is shown in Figure 3. This also shows the

irregular outline of the cave, which is about 19 m at its widest and 15 m long.

At the back of the cave there is a debris-filled tunnel that, if cleared, would

allow access to the hillside above the cave. A sink-hole at the end of the tunnel,

visible from above, suggests that the cave originally had an inner chamber, as is

the case at BNK 3. In the roof of the cave are three quite large solution cavities

or ‘chimneys’ that open to the upper hillside. One of these is in the central area

of the cave, approximately above grid square O 28.

THE SITE AND ITS ENVIRONMENT

Byneskranskop (34°35’S 19°28’E) is a limestone hill at the junction of the

coastal plain and the hilly country that rises to the ranges of the Cape Folded

Mountain Belt some 60 km to the north (Fig. 4).

BNK 1 is situated near the top of the hill, some 60 m above mean sea-level

and at the top of a cliff that overlooks the valley of the Uilkraals River (Fig. 2).

The general aspect of the cave is south-east but protection against the prevail-

ing summer winds is afforded by a dense growth of Sideroxylon inerme trees

Fig. 4. Aerial view of Byneskranskop (centre), looking north. Beyond is the Swartkransberg,

followed by the Kleinriviersberge, with the Riviersonderendberge just visible on the horizon.

ANNALS OF THE SOUTH AFRICAN MUSEUM

‘doysuerysoudg punose evoie oy) Jo dew o1ydesZ0doy, ‘s “317

° ee

oote e

°e he

.--OE-

° 5009006,~c

° ° °

OO “eoccccee ef C:

obahed 0) eee =

282 cee

% . °

e ° e

Naisiti. 9h Ce ath 5,

Sapenls S °. 'e

5 3 ee genie 2

ool

00z

oot

)

ecce

eee,

'°

E61 SL 6l

SYNOLNOD IYOHS4IO Pa

FYOHS ANDOU nt

SJNNG

SauLaW NI SH1dd0/SLHSOISH

WwW» ote,

/SV,VE

d o° Peo

° ° °

@ 2°04, Qo

eeee, ye?

oP °

5 ° °°

Sooo eee

° oe e

° e008 7 % gt Pen 0 88 °

oo; oe oe ° «

Oo A, PO ee eacete ° x

e eo” See oe % a ©

° 6° eog eee 4

° .° d

e e

* °

oo :

on eccce e

oot face °

; : N

oe °

e 6 S

ee °

pe 2 °

z : .

. .

° ° 0

e ef cf

o a

OMe °

° ° e

& ° .

° & °

° a. °

° ° *

° . b

fe ° R

le ° ;

e . C

f °

° °

9 °

& pd A

5 °

. °

sid e

° °

. °

J e

°.

*e.

i Of VE

‘ 9

BYNESKRANSKOP 1 9

that rise from the slope below the cave and reach above the top of the cave

mouth. The situation of the cave also affords protection from the rain-bearing

north-westerly winds that blow during the winter months.

Since the environment in which the cave is situated is important in its

relevance to the ecology of the cave’s prehistoric inhabitants, its various aspects

are discussed in some detail.

PHYSIOGRAPHY

As indicated in Figures 4 and 5, the Agulhas region consists of a fairly wide

coastal plain separated by undulating hills from the ranges of the Cape Folded

Mountain Belt. Outliers of these mountains reach to within a kilometre of the

cost at the northern end of Walker Bay, but their main west—east spread is

some 60 km north of the Agulhas region.

Topographically, the area around BNK may be divided into two main

regions, the coastal plain and the hilly hinterland.

The coastal plain lies mostly below the 15,5 m (50 ft) contour line and in

the vicinity of BNK is about 7 km wide. Spies et al. (1963: 5-6) consider the

coastal plain to be an ancient marine terrace, standing about 6 m (20 ft) above

mean sea-level at the coast and rising to about 46 m (150 ft.) at its inner margin

which, in the vicinity of BNK, is a few hundred metres seaward of the

Uilenkraal survey beacon at the top of the hill. Spies et al. (1963: 20-21) cite

Du Toit (1954: 441, 464) as dating the terrace to the early upper Pleistocene,

and this is probably relevant to the formation of the limestone dune ridges of

which BNK is a part.

Much of the coastal plain between BNK and the sea consists of the valley

of the Uilkraals River, but westwards towards Danger Point the hills run closer

to the sea (Figs 2, 5).

The hills that lie between the coastal plain and the mountains are often

called ‘berge’ (mountains) but rarely rise above 300m above sea-level. In

contrast, the mountains of the Cape Folded Belt in the Riviersonderend range

to the north of BNK reach 1 465m (Fig. 4). The highest point of the

Franskraal se Berge, also known as the Duinfonteinberge, is only 381 m above

sea-level (Figs 2, 5), and the highest point in the vicinity of BNK is the

Swartkransberg, which rises to 515 m (Figs 4-5).

SEA-LEVEL CHANGES

Off-shore contours, showing depths of 10, 30 and 50m below present

mean sea-level, have been taken from the hydrographic survey map of the area

drawn up by the South African Navy (Republic of South Africa: South Coast;

Cape Hangklip to Cape Agulhas, 1970), and are shown in Figure 5.

On the basis of data contained in a paper on Cenozoic sea-level changes

(Tankard 1975, fig. 2), it is possible to suggest that between 13 000-9 000 B.P.

the sea-level was 50 m below the present, between 10 500-8 000 B.P. it was

30 m below, and between 7 000-6 000 B.P. it was 10 m below. The ranges of the

10 ANNALS OF THE SOUTH AFRICAN MUSEUM

pairs of dates are not intended to indicate the periods during which the

sea-level was at the depths mentioned but reflect a lack of concordance

between the authors on whose work Tankard drew.

A 50m drop in sea-level would add about 4-14 km to the width of the

present coastal plain between Danger Point and Dyer Island, a 30m drop

about 1-7 km and a 10 m drop about 0,5-6,0 km (Fig. 5). The higher figure in

each case covers the area around Dyer Island, which would be cut off from the

mainland by a sea-level as little as 5 m below the present. The distance added

to the mouth of the Uilkraals River, the closest point to BNK, would have been

the lower figure in each of the pairs.

Although the wide ranges of the dates given for the —50 m and —30 m

sea-levels indicate a need for caution in assumptions derived from these, it

would seem that for most of its occupational history BNK has been 10 km or

less from the sea. Consideration must, however, also be given to the possibility

of a marine transgression such as that postulated for Langebaan Lagoon by

Flemming (1977: iii, 81, 143-5). Flemming suggests that this transgression

commenced at about 6 500 B.p. and reached a maximum of 3 m above present

sea-level between 6 000-5 000 B.P., after which a regression continued until

about 2 000 B.P., when the present sea-level was again reached.

Morner (1978: 5-6) considers the period 10 000 to 6 000-5 000 B.P. as ‘the

rising towards and the reaching of the Holocene “climatic optimum’’. It also

corresponds to the postglacial sea-level rise and reaching of the highest

Holocene sea-levels (Some meters above the present) along most African

coasts.’ .

In the absence of similar evidence for the Agulhas area and because of the

lack of topographic data below the 15,5 m land contour for most of the coastal

plain in the BNK area, the effects of such a transgression can only be

speculated upon. Clearly, however, if such a transgression did commence at

about 6500 B.p. the sea-level at this time cannot have been at —10m, as

suggested by Tankard’s sources. If the sea-level at 6 500 B.P. was the same as at

present it would also need to have been higher in the preceding period. The

implication of Flemming’s and Morner’s suggestions is that for perhaps the last

seven to eight thousand years BNK has been less than 7 km from the sea, an

important consideration in view of the evidence that marine resources became

important to the occupants of BNK 1 only from about 6 000 B.P., as will be

shown in a later section.

GEOLOGY

The geology of the area between Bredasdorp and Gansbaai has been

studied by Spies et al. (1963). In the vicinity of BNK three main lithological

units may be recognized. These are principally important for their effect on the

vegetation they support, and secondarily as sources of some of the raw

materials from which the stone artefacts were made that were recovered from

the BNK 1 deposits.

BYNESKRANSKOP 1 1

The hills are formed of the quartzites and sandstones of the Table

Mountain Group of the Cape Supergroup. Exceptions to these are the lime-

stone ridges that flank the hills and are, lithologically, more properly part of the

coastal plain.

The coastal plain consists mainly of the Tertiary to Recent consolidated or

loose calcareous sands of the Bredasdorp Formation. These are mostly under-

lain by rocks of the Table Mountain Group which are generally exposed only

along the shoreline and are in parts capped by calcified dune sands of marine

and aeolian origin. The rock formed by calcification varies from crumbly

calcareous sandstone to hard crystalline limestone.

Exceptionally, the valley of the Uilkraals River does not consist of cal-

careous dune-sands but of sands derived from the weathering of the rocks of

the Table Mountain Group. These sands are, therefore, largely deficient in the

lime that characterizes the sands of the coastal plain and are lithologically part

of the hills unit.

The third lithological unit is the shales and sandy shales of the Bokkeveld

Group of the Cape Supergroup. Although the Geological Survey map (Depart-

ment of Mines 1963: 3419C, 3419D—Gansbaai) shows these as commencing

about a kilometre east of BNK, roadworks at the northern end of the hill show

that they extend at least as far as BNK, where they are shallowly overlain by

sandy soil. Road cutting through the foot of the hill along the river valley has

also exposed clays that are presumably derived from the weathering of these

shales. They may, however, be fluvial deposits from the shale areas up-river.

The BokKeveld shales extend eastwards for about 45 km in a band that

ranges from about 2 km wide in the west to 17 km wide in the east. For the

most part they are bordered on the north by the sheared shales and fine-grained

graywackes of the Malmesbury Group. These shales weather to form clay soils

that are variably mixed with quartzose sands.

DRAINAGE

The principal drainage of the BNK area is the Uilkraals River, which rises

in the Perdeberg some 20 km to the north-east and has its mouth about 6,5 km

south-west of BNK. The river is fed by three perennial tributaries, the Boes-

mans, Papies, and Baviaans rivers, but its flow is now evidently much reduced

by the drawing off of water for agricultural and domestic purposes. For the

most part the river’s sluggish flow is accentuated by the low-lying terrain of the

coastal plain, which gives rise to marshy areas that are flooded during periods

of heavy rainfall. The river-banks, particularly between BNK and the sea, are

now heavily infested by dense growths of exotic acacias (Figs 2, 4).

CLIMATE

Taylor (1961: 154-5) has described the climate of the area as ‘equable’,

with temperatures rarely exceeding 33 °C in summer or falling below 1 °C in

winter. Taylor gives the mean annual temperature recorded at Danger Point

1 ANNALS OF THE SOUTH AFRICAN MUSEUM

Lighthouse as 16 °C with a mean variation ranging from 18,9 °C in January to

13,3 °C in July.

The area is considered to lie within the winter rainfall region, although

during the years 1901-50 only 68 per cent of the mean annual rainfall of

543,8 mm recorded at Danger Point fell during the late autumn to spring

months (May—October) (Taylor 1961: 154), and farmers in the area say that

they receive rain throughout most of the year.

Additional moisture is provided by the sea mists that are not uncommon

during the late summer, and is also collected by the summer south-east winds as

they pass over the sea (Taylor 1961: 155); but the principal rainfall is con-

tributed by the north-west winds that are most prevalent during the winter

months.

Sunshine ranges from about 60 per cent of the possible duration in winter

to over 70 per cent in summer (Schulze 1965: 314). It was noted, however,

during the 1980 field season in the first fortnight of July, that the south-easterly

aspect of the hill deprives it of sunshine for most of the day, and the caves were

often unpleasantly cold and damp, even on sunny days.

VEGETATION

The natural vegetation of the area is the macchia-like growth known in

South Africa as fynbos. Kruger (in Day et al. 1979: 2) describes the main

physiographic features of the fynbos as a prevalent sclerophyllous shrub form, a

scarcity of trees, and the relatively minor importance of grasses and evergreen

shrubs.

Although more than a quarter of a century ago Acocks (1953: 129) pointed

out that the fynbos on the limestone of the Bredasdorp district, and that of the

Elim Flats, would have to be treated as separate veld types, no detailed floristic

survey or analysis of these two areas has yet been undertaken. It follows,

therefore, that general surveys of the fynbos are not wholly applicable to the

BNK area.

An early outline survey of the vegetation of the Bredasdorp and Caledon

districts was published by Jordaan (1946). Of this, the part (pp. 52-53) that

deals with the strandveld or coastal plain is principally relevant to this report.

The climax vegetation is described as tall shrubs 3-5 m high, predomi-

nantly Euclea racemosa and Pterocelastrus tricuspidatus, with dwarf trees such

as Tarconanthus camphoratus and Sideroxylon inerme, and an understorey of

other small shrubs and herbaceous plants.

At places such as Platbos at Uilkraal, the locality of BNK, and Grootbos, a

few kilometres to the north-west, remnant forest communities are found, which

Jordaan considered to be floristically much the same as the fynbos of the

coastal plain. The main difference is that what are shrubs on the coastal plain

are found in the forests as trees up to 12 m tall.

The ecology of these forests has been studied by Taylor (1961), who

reports (p. 154) that the geology of the forest areas consists mostly of Table

BYNESKRANSKOP 1 13

Mountain Sandstone except in the Tall Forest (Grootbos), where part grows on

granite, and in parts of the Short Forest (Platbos), where the Table Mountain

Sandstone is overlain by sandy limestone characteristic of the coastal plain.

Taylor (1961: 159-162) describes the Short Forest as essentially dominated by

Calvaria inermis (= Sideroxylon inerme) and Euclea racemosa, while the Tall

Forest is dominated by Celtis rhamnifolia (= C. africana), Olinia cymosa and

Apodytes dimidiata. An outlier of the Short Forest covers much of the cliff face

at BNK, as indicated in Figure 2. The trees on the lower slopes and in the

foreground, across the river, are predominantly Sideroxylon inerme but the

vegetation of the river valley, where not cleared for agriculture, now consists

mainly of Australian acacias (A. cyclops, A. saligna, and A. mearnsii) and the

European poplar (Populus sp.).

Jordaan (1946: 52) observed that the climax vegetation of the limestone

dunes differs floristically from that of the sandy soil, with Proteaceae predomi-

nant and Rhus spp. and reeds, notably Chondropetalum tectorum, more preva-

lent than on the coastal plain. He also pointed out that in the area where the

calcareous soils or limestone dunes have their greatest extent, between Cape

Agulhas and the Potberg, i.e. to the north-east of Cape Agulhas, parts are

ploughed for grain cultivation. J. Rourke (1980 pers. comm.) has also men-

tioned that the calcareous to neutral soils of the Bredasdorp district have

enabled the westward migration of sub-tropical genera such as Carissa, Cusso-

nia, Scadoxis, and Sideroxylon.

The climax vegetation of the clay soils around Elim is, according to

Jordaan (1946: 53), a fynbos mosaic in which, like that of the limestone areas,

Proteaceae predominate. Since the clay soils are good for grain cultivation the

climax vegetation is, however, seldom found today.

Another veld type mentioned by Jordaan (1946: 50) occurs on the hills to

the north of BNK (Fig. 4). This is locally known as ‘heiveld’ (heathland) and is

dominated by reeds, although Jordaan observed that it is also a mosaic

vegetation, which he considered resulted from the destruction of the shrubs of

the fynbos, which will regenerate if the land is left undisturbed. Jordaan

reported that farmers say that ‘heiveld’ is not suitable for grazing—it is said to

be ‘sour veld’—except where soft reeds predominate.

Acocks (1975: 87) considered that the climax vegetation of the coastal

fynbos is “a grassy, more or less open scrub, at least on the south coast belt and

on the Cape Flats’. Elsewhere, Acocks (1979: 681) divided the vegetation of

the ‘coastal plateau’ into two types, Renosterbosveld, dominated by Elytropap-

pus rhinocerotis on clay soils (cf. Acocks 1975: 86) and fynbos on sandy soils in

part overlying ‘calcareous tufa’. He suggested that a short closed forest may

have covered ‘all of the more or less sand-covered limestone ridges of the

Duineveld of Bredasdorp’ and areas to the east; that ‘this sort of short

forest rotated with tall Protea fynbos and Themeda-dominated grassveld’; and

that ‘the grassveld phase may have been shorter-lived than in the Renosterbos-

veld’.

14 ANNALS OF THE SOUTH AFRICAN MUSEUM

Taylor (1978: 204) also has two major subdivisions of the coastal fynbos,

but these are geographically greater than those of Acocks mentioned above.

Taylor’s divisions are one on marine sands north of Cape Town and the other

‘on limestone east of Danger Point to near Mossel Bay’. He considers that the

difference in the two vegetation types is more floristic than physiognomic, and

that both are ‘typical fynbos with restioid, ericoid and some proteoid forms, but

both have many more grass species and annuals than Mountain Fynbos’.

In contrast to the views of Acocks and Taylor, quoted above, that the

coastal fynbos can be described as grassy, are the observations of Spies et al.

(1963: 8), relevant to the research area, that perennial grasses are few and

consist mainly of lank, sour grass types that grow on the Table Mountain

Sandstone, but that during the winter months there is an abundance of annual

grasses. J. Rourke (1980 pers. comm.) has pointed out that the acid soils

derived from Table Mountain Sandstone are inimical to the development of

grasslands, and C. Boucher (1980 pers. comm.) has mentioned that many of the

grass species now present in the fynbos are probably not indigenous. It has

already been mentioned that the shale and clay soils to the north and east of

the BNK area are good grain-lands and therefore good grasslands, and it

should be pointed out that grain cultivation is a winter occupation in this area.

That grass can, and does, grow in the BNK area is not disputed. Local

farmers are able to maintain small herds of sheep and cattle on a year-round

basis, although these animals are not entirely dependent on grasses. It seems

unlikely, however, that the geology of the area would have permitted the

development of extensive grasslands on the coastal plain. As anyone who has

travelled in the area will know, the main areas of grain cultivation are away

from the coastal plain, and it seems reasonable to assume that in a region in

which the cultivation of grain is a primary agricultural industry, land suitable

for such a purpose would not be neglected.

The availability of adequate grazing as well as the overall nature of the

vegetation are important considerations in the light of the changes that took

place in the composition of the fauna represented in the BNK 1 deposits, and

which are discussed in a later section. Van Zinderen Bakker (1976: 181) has

suggested that ‘from a plant ecological view it seems more likely that the

evolution from open grassland to closed vegetation was caused by climatic

rather than sea-level changes’. This evolution was postulated for the Nelson

Bay Cave area by Klein (1972a, 1974), and also forms part of Acocks’ ‘rotating

vegetational pattern’ for the ‘coastal plateau’, as mentioned above; but an a

priori requirement for the development of grasslands, whatever the climatic

conditions, is the presence of suitable soils.

FAUNA

The terrestrial fauna of the area, especially when compared with the

remains of the animals recovered from the BNK 1 deposits, is nowadays

severely depleted. The smaller antelope such as grysbok (Raphicerus melano-

BYNESKRANSKOP 1 ES

tis), steenbok (R. campestris), and klipspringer (Oreotragus oreotragus) are still

to be found in the more undisturbed parts. Hares (Leporidae), porcupines

(Hystrix africae-australis) and some of the smaller carnivores, e.g. ratel and

otter (Mustelidae), genet and mongoose (Viverridae), and wild cat (Felis

libyca), are also still to be found, as are tortoises (Chersina angulata and

Pelomedusa subrufa) and a number of species of snake. The bontebok

(Damaliscus dorcas) is now preserved in the Bontebok National Park, about

110 km north-east of BNK, which is also the refuge for a number of other

antelope that previously ranged free in the area.

Marine resources, both of the shoreline and the deeper waters, are also

much depleted although most kinds of sheilfish are still relatively common in

some areas and the coastline still offers reasonable sport for anglers. Choice

seafoods such as the rock lobster (Jasus lalandii) and the perlemoen (Haliotis

midae) are, however, rarely to be found, even by skin-divers; and the local

trawling industry that operates out of Gansbaai and Hermanus has been

severely depressed for a number of years as a result of excessive competition.

Southern right whales (Eubalaena australis) are to be seen in Walker Bay

during the breeding and calving seasons.

SUMMARY

As will be evident from the foregoing, BNK 1 is well situated, in a climate

free from harsh extremes, close to permanent running water, at the junction of

two major environmental zones and not too distant from a third, the sea.

Access to the interior would have been facilitated by the cutting of the

Uilkraals River valley through the hills behind the coastal plain, and the

relative flatness of the plain would likewise provide few obstacles to travel.

Except in occasional instances of flooding, the river would not hinder

movement across the coastal plain.

A somewhat extensive examination of the published data on the vegetation

of the area has been provided, together with the differing views of various

researchers, since it is ultimately the vegetation of an area that determines the

suitability of that area for the existence of various life forms. It is of some

relevance to the human ecology of the area that among the flora of the remnant

forests listed by Taylor (1961) are at least twenty species known or said to be of

value as food or medicine (cf. Smith 1966), and local informants are able to add

to that number.

Although the local fauna, both terrestrial and marine, is nowadays much

reduced by comparison with what has been revealed by the faunal sample

from BNK 1 to have been present in the area in the past, it provides

parameters against which the effects of human activity and climatic change can

be gauged.

16 ANNALS OF THE SOUTH AFRICAN MUSEUM

THE EXCAVATIONS

EXCAVATION PROCEDURE

Excavation was carried out by the usual method of trowel and brush.

Excavated material was sieved at the site through a double rack of sieves, one

of 12,7 mm (3 in.) mesh above one of 3,2 mm (gin.) mesh. Separate spoil

heaps were used for the 1973-4 and 1976 excavations, and these were located

beyond the drip-line.

Preliminary sorting into raw material categories of stone, bone, and shell

was carried out at the site and the sorted material packed in bags marked with

the site reference, date, grid and stratigraphic references, and the raw material

category.

Carbon samples for radiocarbon dating purposes were removed directly

from the deposit, using a trowel or tweezers, and stored in labelled polythene

bags. During the 1974 excavation soil samples were also taken from the deposit

for palynological analysis, should this become practicable.

A daily site log was maintained in which records were kept of the squares

and stratigraphic units excavated and the number of buckets of deposit

removed. Observations regarding the nature of the deposit were also recorded,

and drawings made of the stratigraphic and other features such as the human

burial, hearths, and variations of the stratigraphy within grid squares. Although

much of the data recorded in the site logs is not included in this report, they

have formed the basis for the information provided. These site logs, one for the

1973-4 excavations and another for the 1976, are retained in the Archaeological

Data Recording Centre of the South African Museum.

In 1978, after it was decided not to extend the excavation further, the

walls of the main trench were protected by hoardings of treated wood, the

walls first having been lined with heavy polythene sheeting. The shallow

excavation in the front four squares was also lined with sheeting before being

back-filled.

Prior to the securing of the walls of the excavation, rain coming through

the central ‘chimney’ during an exceptionally heavy storm caused the collapse

of part of the upper deposit in grid square O 28. Corrugated iron sheets placed

at the edge of the excavation, under the ‘chimney’, now protect this from

further similar damage.

STRATIGRAPHY

The cave deposit consists mainly of fine, grey or brown soils of which the

stratigraphy is generally poorly defined. There was a certain amount of penetra-

tion of the deposit by roots of the Sideroxylon inerme trees that front the cave,

but never enough to hamper the excavation or disturb the deposit. The floor of

the cave dips towards the mouth, and there is a large depression in the

north-east part of the cave, possibly the result of scouring out of the deposit by

rain-water, or else of digging for ‘guano’.

BYNESKRANSKOP 1 17

In the 1973 test excavation of grid square P 26 and half of OQ 26, a total of

seventeen spits was excavated, bedrock being reached at a depth of

2,93-3,03 m below the surface. A human burial was encountered in Spit 6 of

square P 26, necessitating the extension of the test pit into square Q 26. The

presence of the burial obviously invalidated the stratigraphy of some of the

overlying deposit, but this was later determinable from the sections of deposit

exposed by the excavation.

In the 1974 excavation grid squares P 27-29 were excavated to bedrock,

and P 30 to some 90 cm short of bedrock. A tendency for the overburden from

the unauthorized diggings to fall into the trench necessitated the cutting back of

a shelf in squares OQ 29-30.

In the 1976 excavation squares O 29-30 were excavated to bedrock and

squares O-P 31-32 to a depth of about 50 cm.

In the 1974 excavation sixty-three individual stratigraphic units or features

were excavated, and sixty-one in 1976. Many of these were localized features

such as hearths or patches of differently coloured or textured soil that did not

extend over the whole of the excavated area or perhaps even a whole grid

square. Others were small lenses or somewhat arbitrary subdivisions of larger

stratigraphic units made to ensure precise stratigraphic control. After examina-

tion of the stratigraphy of the unexcavated deposit and laboratory analysis of

the material recovered, it was decided to combine the stratigraphic units into

twenty major units, here termed ‘layers’. The stratigraphy of these layers,

taken from the O-N line in squares 29-30, is indicated in Figure 6.

Layer +

This was a surficial deposit of fine, reddish-brown sandy soil that was

considered to be disturbed and possibly contaminated by debris from the

unauthorized diggings. A layer of leaf-mould approximately 2 cm thick was

considered to mark the base of this layer. Material recovered from this layer

has been retained but is not included in the analyses. The volume of material

excavated from layer + was not calculated.

Layer 1

The sand of the surficial deposits continues into this layer, where it

becomes more brown. In the front four squares, from which the overburden

from the unauthorized diggings was removed to a level approximately equal to

that of the rest of the surface, the deposit was considered to be undisturbed.

The middle sub-unit was distinguished by compacted ashy patches, and the

basal sub-unit was again a fine, brown sand. The volume excavated was

calculated as 2,6 m?.

Layer 2

The two sub-units of this layer are a sandy soil that is more grey than

brown, and a dark, humic soil. The volume excavated was calculated as

0,57 m?.

18 ANNALS OF THE SOUTH AFRICAN MUSEUM

layer 29 30

Leaf mould

(Disturbed)

Ash nodules

oF SA AO ei

eno tete ttt tee eee:

OOO) ees:

wna a 'e'e'9*,%,,0 0 2 =

=: Compacted ashy soil

Rock

Light ash

Dark ash

Compacted tortoise

A ~)’)”»»

WU yyy

Fig. 6. Stratigraphy of the main layers in grid squares 29 and 30, on the O-N line. Each square

is 1 m wide.

Layer 3

This is distinguished from the adjacent layers by a high density of marine

shell. Although it was not consistent over the whole of the excavated area, the

density of shell in this layer almost qualifies it to be called a shell-midden

deposit. Layer 3 was not recognized in squares P 27-28 and may thus be a

localized feature. The volume excavated was calculated as 0,25 m?.

BYNESKRANSKOP 1 19

Layer 4

The two sub-units in this layer are a fine, brown sandy soil of which the

base is marked by what appears to be a thin band of calcined shell. The layer

was characterized by a number of hearths in squares P 28-30, and in square

P 29 a patch of decayed vegetation was considered to be an area of bedding.

The excavated volume was calculated as 0,91 m?.

Layer 5

The number of sub-units in this layer varied from one to five in different

parts of the excavation. The layer as a whole was characterized by a relatively

high density of artefactual and faunal material in a greyish soil matrix. The

excavated volume was calculated as 1,24 m?.

Layer 6

This layer was distinguished by its high ash content, and included several

hearths. The excavated volume was calculated as 0,84 m?.

Layer 7

In places it was difficult to determine the base of layer 6, but on the whole

layer 7 was less compacted and less ashy than layer 6, although it also contained

a number of hearths. The excavated volume was calculated as 0,81 m°.

Layer &

This layer was not recognized in squares P 27-28, and in the other squares

its base was determined by the presence of a fine, grey ash band. The excavated

volume was calculated as 0,58 m2.

Layer 9

The upper sub-unit of this layer consisted of a dark-brown, gritty soil that

was mainly damp and appeared to contain relatively high frequencies of

microfauna and roof spalls. Although this sub-unit suggests a period of sparse

or no occupation, the relatively high artefactual and faunal content of the layer

as a whole suggests that occupation of the site during the period of deposition

of the lower sub-unit may have been fairly intensive. The two radiocarbon

dates for this layer (Table 1) do not indicate a hiatus. Part of the lower sub-unit

is characterized by numerous ash nodules. The excavated volume was calcu-

lated as 1,02 m?.

Layer 10

This is a compacted, ashy layer, yellow in parts and underlain by a black,

carbonized deposit. The excavated volume was calculated as 0,56 m’.

20 ANNALS OF THE SOUTH AFRICAN MUSEUM

Layer 11

This small layer was the in-fill between the carbonized deposit of the lower

part of layer 10 and an ash band, of which the base was taken to mark the top

of layer 12. The excavated volume was calculated as 0,28 m?.

Layer 12

Another compacted, carbonized deposit, easily recognizable, with poor

bone preservation in parts. The excavated volume was calculated as 0,56 m?.

Layer 13

This layer may be a facies of layer 12 as it contained much the same sort of

ashy or carbonized soil and was not clearly distinguishable from the overlying

deposit. Square P 30 was not excavated below this level. The excavated volume

was calculated as 0,45 m?°.

Layer 14

The upper part of this layer consists of a brownish soil in which, towards

the front of the trench, in squares O-P 29-30 a lens of gritty soil occurs. The

layer was remarkable for the dense concentration of tortoise remains that

comprise the basal deposit. This ‘tortoise midden’ is evidently a lens that begins

to thin out towards the front of the cave. The excavated volume was calculated

as 0,77 m?.

Layer 15

A loose brown soil without much material content, separated in places

from layer 16 by a band of darker soil. The excavated volume was calculated as

0,60 m?.

Layer 16

The soft sandy soil of this layer, brown in parts, grey in others, was not

always readily distinguishable from layers 15 and 17. Towards its base it

contained more stone and bone than the overlying soil. The excavated volume

was calculated as 0,62 m?.

Layer 17

A brown, sandy soil, distinguished from layer 18 by the presence of ashy

flecks or nodules. The volume excavated was calculated as 0,57 m?.

Layer 18

A light-brown soil with ashy patches. It was not always distinguishable

from parts of layer 19. The excavated volume was calculated as.0,65 m’.

BYNESKRANSKOP 1 21

Layer 19

The basal deposit is made up of a number of often poorly defined lenses

with varying characteristics, which overlie the bedrock and fill in the irregulari-

ties of its surface. In 1974 four sub-units were recognized, but in 1976 only two.

The excavated volume was calculated as 0,98 m’.

As mentioned, the stratigraphy was seldom clearly defined and, as was to

be expected, the various strata were not evenly distributed over the whole of

the excavated area. The only clearly definable units were the shell band of layer

3 and the concentration of tortoise remains in layer 14. Layer 6 was generally

more ashy than the other layers, but the presence of ash features, many of

them assumed to have been hearths, was characteristic of layers 4-9.

It is arguable that some of the smaller layers should have been combined,

and it is possible that some of the sub-units that have been combined should

have been kept separate. The poor stratigraphy also allows the possibility that

material from adjacent layers might have been incorrectly assigned in some of

the squares. The presence of seed cases of probably recent date, as far down as

layer 15 and ascribable to rodent activity (see section on plant remains),

represents another factor that cannot be ignored. However, since there was no

clear evidence of major disturbance of the deposits in the excavated area apart

from the surface deposits and the human burial in the test pit, the amount of

mixing of material is not likely to have been great. In the analyses of the

excavated material that follow it has accordingly been assumed that the

material has been recovered from its proper stratigraphic context. The analyses

indicate reasonable consistency within the layers, and this tends to support the

general correctness of the excavation procedure.

DATING AND CORRELATION

A total of twelve radiocarbon dates was obtained, which show the site to

have been occupied over a period of some 12 500 years. These dates are listed

in Table 1. Those with the reference Pta- were provided by the Natural

Isotopes Division of the Council for Scientific and Industrial Research, Pre-

toria. The second date for layer 9 was obtained by courtesy of R. G. Klein from

the radiocarbon laboratory of the University of Washington (ref. UW-), and

the date for the basal deposit was determined by Teledyne Isotopes, West-

wood, New Jersey (ref. I-).

The span of some 3 000 years covered by the four dates for layer 1,

obtained from carbon samples taken from some 35 cm of deposit, suggests a

much lower rate of deposition than that, for example, of layers 12-9. Here,

disregarding compaction, some 50-60 cm of deposit accumulated in a period of

about 1 400 years.

The youngest date for layer 1 does not mark the final occupation of the

cave but is associated with the only find of metal at the site, a small copper or

brass bead that came from the second of the sub-units of this layer that were

considered stratigraphically reliable. The second date for layer 1 was obtained

22 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 1

Radiocarbon dates (uncorrected) for occupation layers.

Layer Lab. no. Years B.P.

Ii(upper—O 315 1D) 3. Pe uscs =P ee ee Pta—1864 JD) 28, 5\)

(Uppers=P 32D). tulsa pete (hes hee eee Pta—1866 535) 2230)

(middle CBA) eb ronsceornatine eek sion etic Pta—1865 1 880 + 50

(bottom— EVA): ss Sao. ene Sue ae Pta—1631 3 220 + 45

Dye ge an oar eOue coe SCE Ses Ene ee Cee Pta—1569 3 400 + 55

Oe McA led ea Uo sere re Ss Me erat RRL inci 8 Gee Pta—1571 3 900 + 60

9 (upper = MOR), 4h). sian aa eae ee eee tas lg 6 370 + 90

(lower=-NEDL Yo 0.6 33 Ses ttre ticks eres ana rea UW-409 6 100 + 140

| RS ae pegmnarereminne ye MMs ence. 8 A Stee «dae ole dao ncunaaten AiouhS, p66 Pta—1905 6 540 +55

| OR rae eae are Re cr aan Arts at Siero So Pta—2347 T.A50 +90

La a a oA areca tntvysth ibaa s eaeareto Sea he a eerie gh sien Sine ira Pta—1587 9 760 + 85

Ne rane meee re ones came ROUTH), Ores Si ev shaty euce Ota 1-7948 12, 7302s

Note. All dates are based on the Libby half-life of 5 568 + 30 years.

from charcoal directly associated with potsherds in the same sub-unit as the

metal bead but a different square. The minimum difference of 180 years

between these two dates may indicate that layer 1 was more disturbed than was

appreciated during the excavations, or it may simply attest to the sporadic

nature of the occupation of the cave during this period.

The third date for layer 1 records the earliest occurrence of potsherds at

the site. About half a dozen were recovered from the lowest sub-unit but these

are considered to have been trampled in from the upper sub-unit. Klein’s

unpublished preliminary analysis of the mammalian fauna from the 1974

excavation shows that sheep remains were also found in the lowest sub-unit,

but it seems likely that Pta-1865 should also relate to the earliest occurrence of

domestic sheep in the BNK 1 deposits.

The oldest date for layer 1 is some 1 300 years older than the date for the

overlying sub-unit (CLA), and closer to that for layer 2. This suggests that the

basal sub-unit (EVA) and layer 2 should have been combined.

The dates for layers 2 and 5, relative to the depth of deposit, support the

suggestion of a lower rate of deposition for layer 1. They also pose problems

regarding the chronology of layers 6-8, again with regard to depth of deposit as

an indicator of rate of deposition, since the dates for layers 2 and 5 are only

some 500 years apart, whereas those for layer 9 are at least 2 000 years older

than that for layer 5.

As independent assessment of the date for layer 9 was sought as, although

the first series of dates for the site showed reasonable consistency, the date for

layer 9 was thought to be too young. This layer appeared to mark the beginning

of the ‘Wilton’ at the site, but dates for other sites such as Wilton (J. Deacon

1972: 36) and Melkhoutboom (H. J. Deacon 1976, table 2) indicated that, in

the eastern Cape at least, this occurred at a time closer to 8 000 B.p. In the

event, the University of Washington date confirmed the accuracy of the

BYNESKRANSKOP 1 23

C.S.1.R. (Pta-) date. Subsequently published dates for sites in the southern

Cape such as Nelson Bay Cave (J. Deacon 1978: 101; fig. 2) and Boomplaas

(H. J. Deacon et al. 1978, fig. 2) indicate that in this area the introduction of

the ‘Wilton’ occurred at about the same time as at BNK 1.

The closeness of the two dates for layer 9 does not indicate a hiatus

between the deposition of the two sub-units. The relatively high frequencies of

roof spalls and microfauna in the upper sub-unit may, however, indicate a

hiatus in the human occupation of the site between the lower part of layer 9

and layer 8. A limited analysis of scraper morphology (cf. Fig. 17) suggests that

layer 9 might have been intermediate between the industries of the underlying

and overlying layers, but in almost every other respect this layer has more in

common with the overlying layers than with those below it.

The date for layer 10 is consistent with the dates for layer 9 and, in terms

of depth of deposit, layer 12, but the difference of some 2 000 years between

the dates for layers 12 and 14 suggests either a period of reduced deposition or

an occupation hiatus. Analyses of the artefacts and, to a lesser extent, the

fauna indicate that this was a period of change in the ecology of the human

occupants of the site, but D. M. Avery’s analysis of the microfauna suggests an

environmental change at about the time layer 10 was deposited (D. M. Avery

1979: cf. Table 22 of the present report).

The approximately 3 000 years taken for the deposition of layers 19-15

indicate yet another period of low rate of deposition. This is to some extent

confirmed by the average density of stone artefacts in layers 17-13, which is the

lowest in the layers below layer 4 (Table 2), although the faunal samples

indicate a qualitative rather than a quantitative change where mammals are

concerned (Table 15).

Although the series of dates obtained for BNK 1 is adequate for establish-

ing the period of occupation of the site and indicates the chronology of the

major events, the uncertainties regarding the undated layers highlight the need

for greater numbers of radiocarbon dates. Some of the chronostratigraphic

problems are possibly related to the combining of individual stratigraphic units

into larger layer units, but present circumstances preclude the obtaining of

large numbers of dates from test excavations that would indicate the strategy to

be employed in excavation and subsequent analysis of the excavated material.

The possibility of breaks in the human occupation of the site cannot be

ruled out, but these are not clearly indicated by the sequence of dates, nor by

the artefactual and faunal components. With the possible exception of the

upper sub-unit of layer 9, or a part of it, there are no archaeologically sterile

deposits and this suggests that the cave was in more or less constant use from

about 12 500 years ago. Comparison of the depth of deposit in the various

layers, which is in part relative to the amount of debris left in the cave by its

human occupants, suggests that during the period 6 000—4 000 B.P. the cave was

relatively more intensely occupied than during the preceding or succeeding

periods. This observation is to a large extent supported by the artefact fre-

24 ANNALS OF THE SOUTH AFRICAN MUSEUM

quencies, although layers 11 and 12 have the highest average density of stone

artefacts of all the layers (Table 2).

METHODS OF ANALYSIS

All the material recovered from the excavations was washed in the labora-

tory before further analysis was undertaken. Material from the 1974 excavation

was accessioned under the reference AA8727 and that from the 1976 excava-

tion under AA8743. The artefactual material was sorted into three main

categories, stone, bone, and shell, and two minor ones, pottery and metal. The

non-artefactual material was sorted into mammals, shell-fish, fish, micromam-

mals, reptiles, birds, and plant remains.

The differences in the excavated volumes of the layers give rise to

problems in the comparison of the frequencies of the material from the layers.

Layer 1, for example, has almost ten times the excavated volume of layer 3 and

twice that of layer 5; and layers 14-19 were not excavated in square P 30.

In an attempt to remedy the distortion of data resulting from these unequal

volumes, an innovation in this report is the use, where appropriate, of what has

been termed the ‘average density’, abbreviated in the text to AD. This was

obtained by dividing the number of components of the various categories in

each layer by the amount of the excavated volume of that layer. This gives the

average density of components per cubic metre. Since only layers 1, 5, and 9

have an excavated volume in excess of a cubic metre, the calculated average

density is greater than the actual frequency in all but these three layers.

It is appreciated that ADs are not the ultimate solution to the problem of

comparing quantities of material from deposits of unequal volume, particularly

as in most cases they yield inflated frequencies and they do not take into

account compression of lower deposits; but it is none the less considered that

they provide a more equitable basis for comparison than one that ignores the

fact of unequal volume.

SUMMARY

The excavations, carried out in 1973-6, covered an area of 11,5 m’, of

which 4 m? were excavated to a depth of about 50 cm and another square metre

taken down to about 90cm short of bedrock, the remaining 6,5 m? being

excavated to bedrock which was reached at a depth of approximately 3 m from

the surface.

The deposit, of which the stratigraphy was seldom well defined, was

divided into twenty major stratigraphic units or layers. The uppermost layer,

layer +, was considered to have been disturbed and the material recovered was

not included in the analyses, nor was that from the test pit, grid squares

Eo 2G;

A series of twelve radiocarbon dates shows the site to have been occupied

from about 12 700 to 255 B.p. There are certain anomalies, particularly with

regard to layer 1, the sub-units of which span close on 3 000 years. The

BYNESKRANSKOP 1 DS

possibility cannot be excluded that there were breaks in the occupation of the

cave, between layers 14 and 12, 9 and 8, and in layer 1. However, on the

evidence of the low artefactual content of layers 17-13, the apparently anom-

alous groups of dates may indicate periods when the cave was only sporadically

occupied, and others when it was more intensively used, particularly during the

deposition of layers 9-5.

In an attempt to overcome the problems caused by the layers having

different excavated volumes, a system, ADs, has been introduced in which the

frequency of artefacts or faunal remains is given as a ratio of the excavated

volume of the layer.

THE ARTEFACT ASSEMBLAGE

STONE ARTEFACTS

Over 160 000 pieces of artefactual stone were recovered from the deposits.

Although frequencies of what may be termed diagnostic artefacts are low, as is

usual, sample sizes for individual layers in the various artefact classes are

generally adequate for meaningful analysis.

RAW MATERIALS

Figure 7 indicates the percentage frequencies of the major raw materials in

each layer. Chalcedony (0,2% of the site total) and shale (0,01%) are not

included because of their negligible frequencies.

Quartz

This accounts for 46,2 per cent of the site total and falls below 40 per cent

of the total of all raw materials in any layer only in layers 16 and 9-5.

Frequencies decline in layers 19-16, increase in layers 15-11, decline in layers

10-6 and increase again to layer 1, which has the highest frequency of quartz

(65,3 % of the layer total of all raw materials) after layer 19 (66,3 %).

Silcrete

This accounts for 22,0 per cent of the site total and shows a marked secular

trend. Frequencies are low below layer 12 but increase to a maximum occur-

rence in layers 9-5, ranging from 28,7 per cent of the layer total in layer 7 to

41,9 per cent in layer 5. From layer 4 up, frequencies decline again in favour of

quartz and quartzite.

Quartzite

This accounts for 26,5 per cent of the site total and shows a clear secular

trend in the lower layers, increasing from layer 19 to a peak (67,7% of the

layer total) in layer 16. Thereafter frequencies decline, reaching a low (18,6 %)

in layer 11 after which they are generally stable at around 25 per cent of layer

totals except in layer 6 (34,0%), layer 5 (19,9 %), and layer 1 (18,3 %).

26 ANNALS OF THE SOUTH AFRICAN MUSEUM

20-

TRESTONE ie

60-

40-

20-

QUARTZITE

60=

40-

20-

SILCRETE <1

60=

40-

20-

QUARTZ

LAYERS T2853 > 4° 95) 6) 87 89 OP 2) Si AS Oe 7a Caen

Fig. 7. Histograms of percentage frequencies, by layer, of stone artefact

raw materials. <1=less than 1 per cent of the layer total. Frequencies of

chalcedony and shale, each less than 1 per cent of the site total, are not

included.

BYNESKRANSKOP 1 Di

Limestone

This accounts for only 5,1 per cent of the site total and has low frequencies

in most layers, exceeding 10 per cent of the layer total only in layers 8 and 7.

The weak secular trends, with peaks in layers 16 and 8, are not considered

significant.

Other aspects of raw material use are discussed in the analyses that follow.

MAJOR ARTEFACT CATEGORIES

Following usual practice, the artefacts have been divided into three major

categories, although with changes of name. The inventory of these categories is

given in Table 2 and the individual categories are discussed below. Table 3

gives the frequencies of raw materials in the categories in each of the layers.

Although raw material frequencies form part of the following discussions, it is

worth noting from the outset that the secular variations in raw material

frequencies shown in Figure 7 are reflected mainly in the unmodified artefact

category only. Silcrete is the predominant raw material in both the utilized-

modified and retouched artefact categories. The only exceptions are, in the

utilized-modified artefact category, layer 15 in which the two artefacts are made

TABLE 2

Stone: inventory of major artefact categories.

utilized—

Category: unmodified modified retouched layer total AD/m?

Layer no. % no. % no. Jo no. Jo

ree A ah a 16 477 97,9 80 0,5 276 1,6 16 833 10,3 6 474

7h, OR Ae 2 032 93,9 24 sla 108 So) DACA eles 3 796

Bh ak cate ae een ae 1 902 93,9 23 el 100 4,9 D025 2 8 100

(hook ge I A 6 373 94,0 97 1,4 310 4,6 OmSO) Fel 7 450

See eso ee eee 14 015 94,0 146 1,0 740 a4) 14 901 OS ARON.

(GS), ie aE 9 840 94,5 126 2, 448 4,3 10414 6,4 12 398

UPR 10 262 95,9 92 0,9 351 55653 10705) G5 ) 13,216

Sree. ese 10 203 96,0 138 TPS; 291 2; LONG3229G as 331

DAR che 0) es mish 15 686 96,8 125 0,8 385 2,4 16196 9,9 15 878

Oe ee eee VANES, QU 68 0,9 103 1,4 7314 45 13 061

NOS on ee oe 7 364 99,0 23 0,3 54 0,7 Pee, Zo 7 ASSIS

Dees ct. 13 310 98,1 103 0,8 150 Hil 1B,563N ors eee 220

AIS ete ry erie Gs Re 3 979 98,8 20 0,5 29 OFZ. AN028) ee 8 951

114 Teste Ma eras ene 4 447 99,5 16 0,4 8 0,2 ATA ez ey 5 806

Se a he ae oc: 2 034 99,7 D 0,1 3 0,1 20389) 2 3 398

GRRE ARs. Pasec A. 3 292 99,6 8 0,2 5 0,1 3305 20 5 330

NORA Oi specs. on sea 4 064 99,6 8 0,2 1 0,2 ANNOYS). 2285) 7 156

TLCS eer sane inane 9 831 99,2 56 0,6 18 0,2 9905 6,0 15 238

NOR etre Ree ate 16 742 99,5 66 0,4 18 0,1 16 826 10,3 17 169

Sitetotales se. 4... 158 996 97,2 122i OW, 3 404 Maedl NGS CHA

Note. Percentages in the first three columns are of the layer total; those in the last column are

of the site total.

AD/m? = average density per cubic metre (see p. 24).

28 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 3

Stone: frequencies, by layer, of raw materials in the artefact categories.

Category: unmodified utilized—modified retouched

Raw material: QO. -S Ot L- O-F-Q S = OF Es 0-1-0 SO eae

Layer

IRS. kee ae 6651 651 18,5 951) OF | @1) G9), GS) == @) | 3058: 65:99 2-2 Ay

DIGS Secs tH, Pt 50,8 15,3. 25,5 7,9 0,4] G1) (G2) @) = = /|89 $32 =] 2s

SE ie Soe 47,2 20,6 28,4 3,6 0.22 | 92) 04) — — — ]19,0 77,0 30 — 1,0

A Ne Reese mec rane 4658) 2153) 2554 6;27 054 123) @0), 2) @ | 203 7ssi 0a

Dial cor ye Sages et oes 34,8) 3930) 21,1 -5,0052, 13:0) 754 832) 1A 1 9308925) 057 Oe

Geer Seer 195. 34,4 33,9 98 054 | 1d; 1o79 4254" We 1e6" 4:9 9254" 1057 eale tee On

Te ONS ce 32,3: 26,3 27,7 13,3 054 | (@)- 74). @ @) (194 86.9) 2:0 0 caaen

Seether 2154. 32,2653. 1359 1053511350) (8555) 057 —— 07 928 S529 Sales Osea

Qiao sd Sead ee 3533 35,2 25,8 354 0:2 16,8 8050 234 — 038 | 4:4 9353" 15>) 0aes

LORS ae eto 46,9 23,0 28,8 1,2 0,06] (11) (56) — — (1) | 8,7 874 39 — —

dill gas eee Peyton: 64,0 16,8 18,7 0,3 0,05} (4) (15) (@) — (2)/]6) 44 (4 — @Q)

| Oe reader Cat 57,2 21,6 20,9 0,2 0,01) 41,7 582 — — — /18,7 80,0 13 — —

1 i ah dt ek La 49.1 10,6 40,1 02 —J| (6) (42) (2) — — {4 @4 GQ —- —

ee ena 50,7 3,6 43,8 1,8 0,02} (4) (9) (G3) — —{]G3) 6) -—- —- =—

TD cc cis5 Was ee ee 46,6 0,6 500 27 —|Q@) — @) — —|]— @2 @ —- —

Gos Dee oe 24,6 1,5 67,9 5,9 0,03} — (6) (2) — —]Q2) B38) -—- — =

dD es ies arc roy: 41.1 9,1 44,8 4,9 0,05) (2) (6) @) — —/]@ @3 @B) - —

fe cae ae inararte 6359" 2,7°30/,4 2.8 032 | G7) (@s) -@) = —] 6) yy eo —

See ae ae eee 66,5 10,3 21,6 1,5 0,05} (40) (49) (7) — —|]— _ (8) (0) — —

Sitextotalis eas: A752 20;227,2 5,2 052° 22,0) 7153) 55> 033" O90 1129" S522 SO ees

Note. Raw materials: Q = quartz, S = silcrete, Ot = quartzite, L = limestone, O = other (chal-

cedony and shale).

Where category totals are 100 or more, percentage frequencies are given, but where totals are

less than 100 the actual frequencies are given in parentheses.

of quartz and quartzite, and layer 18 in which quartz predominates. In the

retouched artefact category, in layer 17 silcrete and quartzite have equal, if

low, frequencies, and in layer 19 there are two more quartzite artefacts than

there are silcrete.

UNMODIFIED ARTEFACTS

The name of this category has been changed from the more commonly

used ‘waste’ in recognition of the possibility that many of the artefacts included

in this category might have had a potential for use and because of the

probability that artefacts, particularly those in quartz and quartzite, have been

included that were, in fact, used but the evidence for which is not readily

detectable.

Table 4 provides an inventory of the frequencies in the classes that make

up this category, which are discussed below.

Chips

These are pieces of less than 10 mm maximum dimension, which comprise

the major component (50,9 %) of this category. This class includes small flakes,

but blades are included in a separate class, irrespective of their length.

BYNESKRANSKOP 1 29

TABLE 4

Stone: inventory of classes of unmodified artefacts.

Class: chips chunks flakes blades cores layer AD/m?

radial _ blade irreg. total total

Layer

1 9 212 3 611 3 574 743) 29 3 25 ay 16 477 6 337

2 616 687 702 15 8 — + 2 2 032 3 565

3 846 478 558 4 8 3 5 16 1 902 7 608

4 2 364 1 895 2 061 17 20 6 10 36 6 373 7 003

5 5 384 Sy79 5 250 146 24 5) 27 56 14015 11 302

6 3 957 2 412 3 340 76 21 1 33 55 9840 11714

7 4 417 1 652 4 053 69 43 6 22 ml 10262 12 669

8 4 253 3 123 2 690 88 16 4 29 49 1032037 17 591

9 7 114 3) IES) 5 165 144 18 11 39 68 15 686 =15 378

10 3 395 1 549 A) 53 10 8 9 27 TN435 2S

11 4 746 1 250 1 328 25 5 1 9 15 7364 26 300

12 8 955 2 416 1 863 42 14 2 18 34 13310 23 768

13 2 081 662 1 198 13 12 z 11 25 S979 8 842

14 2 234 1 256 948 2 3 2 2 a 4 447 5 762

15 811 658 550 i 2 6 14 2 034 3 390

16 1 604 1 081 590 5 D 3 it 12 3292 5 310

Uy 2 109 1 096 825 24 2 4 4 10 4 064 7 130

18 6 987 2 214 592 22 2 10 4 16 OSs lee ioe

19 9 815 8) OU 2 729 168 30 40 43 113 16742 17084

Site

total 80900 36331 40135 937 2s 113 307 693 158 996

Note. AD/m? = average density per cubic metre (see p. 24).

Quartz accounts for 63,1 per cent of the chips and is followed by quartzite

(18,8 %), silcrete (16,6%) and limestone (2,1%), with chalcedony and shale

making up the remaining 0,2 per cent of the site total.

Chunks

These are defined as pieces of irregular shape not showing a recognizable

platform or bulbar scar, and are considered to be pieces that have been broken

rather than deliberately shaped. The class probably includes a number of cores,

since these are apparently under-represented (see below) and might not have

been recognized as such.

Chunks comprise 22,8 per cent of the category total and are predominantly

quartzite (37,3 % of the class total) and quartz (36,1 %) with low frequencies of

silcrete (19,4 %) and limestone (7,0%).

Flakes

Flakes are artefacts with a discernible platform, bulbar scar and/or recog-

nizable dorsal and ventral surfaces, and which are not blades (see below). They

account for 25,9 per cent of the unmodified artefact category.

Quartzite is the most common raw material (37,4% of the class total),

followed by silcrete (26,5 %) and quartz (25,9 %), while limestone accounts for

only 10 per cent of the site total.

30 ANNALS OF THE SOUTH AFRICAN MUSEUM

Although counts were not made, it was noted during sorting that an

apparently high proportion of the quartzite flakes are split down their length,

from their point of impact. This may in part account for their slightly higher

frequency. Some indication of the frequency of broken flakes is given in Figure

8, where the number of unbroken quartzite flakes in each layer of the 1974

sample is given as a percentage of the total number of quartzite flakes in the

layer. The frequency of unbroken flakes ranges from 27-72 per cent of

individual layer totals, with a site mean of 54,4 per cent. This indicates that,

although the frequency varies from layer to layer, the overall proportion of

broken flakes is high.

The proportion of quartzite flakes to flakes of all other raw materials

ranges from 34~76 per cent of the layer total in the layers below layer 12 and

from 29-48 per cent in the layers from layer 12 up.

Hewitt (1931: 545) observed that in the layers underlying the ‘Wilton’

deposits at Melkhoutboom ‘were many large flakes and cores from which they

were made, all from boulders and pebbles of Witteberg sandstone’. Further in

the same report, Hewitt (1931: 547) commented that ‘this pebble industry

seems undoubtedly related to the typical Wilton and occurs to some extent in

nearly all levels: it may merely be a sandstone phase of the industry’.

Hewitt’s reference to ‘large flakes’ suggested that there might be differ-

ences in the size of the quartzite flakes from the ‘Wilton’ deposits and those

from the underlying deposits. It was therefore decided to measure a sample of

those from BNK 1, especially in view of the predominance of quartzite in some

of the lower layers. The length and width of a total of 4 490 unbroken flakes

from the 1974 excavation were measured and the results are shown diagram-

matically in Figure 8. Where length is concerned, the means for layers 19-9

vary erratically from layer to layer, while in layers 8-1 there is a trend of

increasing length in successive layers. Although there are differences if pairs of

layers are compared, the length of flakes in layers 19-9 is generally not

significantly different from that of flakes in layers 8-1, as indicated by the 2 x

standard error bars in the diagrams. In width, there is no significant variation in

layers 19-11, except in layer 17, while in layers 8-1 there is a generally

increasing trend that matches the trend for length. The total range of variation

of the means for all the layers is, however, only 10 mm in length and 7 mm in

width, and it cannot be concluded that there is any real difference in the size of

the quartzite flakes from the lower layers and those from the upper.

Blades

These are parallel-sided flakes with a length at least twice as great as the

width. From discussions with J. Deacon and J. Parkington it would seem that in

the analysis of the BNK 1 material this definition has perhaps been more rigidly

adhered to than by other workers. In the classifying of the BNK 1 material the

only exceptions that have been made to the criteria for blades are pieces with a

length : width ratio of less than 2 : 1 that are clearly broken blades.

BYNESKRANSKOP 1 31

LAYER LENGTH WIDTH no. %t

1 =—e=__— =—=—————_ 4 #8

2 = —-—__— ——e 12849

3 <— = —_ =: — 43. «49

Se =a - —_ po) 67

5 ———_—_— —==___—_ 602 42

6 =-+____ = =e 426 61

7 ne —j——_—__— 523 57

8 = ____ = at aa 244 58

. == —___ =—_—_—_—_ 45

10 = -_>s—___ ——_— 221 27

1 == =———_ 158 64

12 —f=__—_ === 189 72

13 <=-=-—__—___- {eo 192 36

14 = ——__ == 139 58

15 == = 194 64

16 =f 61 69

29 58

— +E as eee

We SS =——_—_— 374 52

20 40 60 80 100 20 40 60 80 100

Fig. 8. Modified Dice—Leraas diagrams of dimensions of unbroken unmodified quartzite flakes

from the 1974 sample. In each diagram the horizontal line indicates the absolute range, the

vertical line the arithmetic mean, the longer bar one standard deviation on each side of the

mean, the shorter bar two standard errors of the standard deviation (95 % confidence limits) on

each side of the mean. Frequencies in each layer sample are indicated on the right, together

with the percentage frequency that this represents of the total number of flakes in the layer, the

rest being broken.

32 ANNALS OF THE SOUTH AFRICAN MUSEUM

The blades are predominantly silcrete (69% of the class total), with low

frequencies of quartz (17,2 %), quartzite (8,1 %), limestone (5,5 %), and chal-

cedony (0,1%). The only secular trend discernible in raw material usage is a

decline from layer 9 to layer 6 in the frequency of quartz blades relative to

quartzite and limestone, a trend that parallels the changes in overall raw

material frequencies (Fig. 7). It is only in layer 1 that quartz replaces silcrete as

the dominant raw material in this class, but since the difference is of only one

blade (8:7) it cannot be considered significant.

The highest frequency of blades is in layer 19 (17,9% of the site total),

while frequencies in layers 9-5 (7,4-15,6%) are consistently higher than in the

remaining layers. The major difference between the blades from layer 19 and

those from the overlying layers is that the majority of the blades from layer 19

are ‘micro-blades’ 15 mm or less in length. That very few of these ‘micro-

blades’ were found with utilization or retouch suggests that if their production

had been deliberate they were used in an unmodified form, and possibly away

from the site more than at it. J. Deacon (1979 pers. comm.) has advised that

similar small blades were found in discrete clusters or ‘caches’ at Boomplaas.

Cores

Cores are actually modified pieces but have been included in the

unmodified category because their modification was incidental to the produc-

tion of flakes or blades and the cores themselves represent discarded material.

The analysis of the cores from BNK 1 follows J. Deacon (1978, table 1) in

dividing them into three sub-classes, radial, blade and irregular. This classifica-

tion was carried out for the BNK 1 cores without much confidence: there are

cores that do not fit readily into the two more ‘formal’ sub-classes and yet are

more than merely irregular; and it was, in general, easier to distinguish

between blade cores and other cores than it was to decide whether a core was

strictly radial or should be classed as irregular.

Radial cores are those from which flakes have been struck from the

circumference of a discoidal core towards its centre (Fig. 9A-B, E). The

frequency of these is generally relative to the overall frequency of stone

artefacts in the layers (Tables 2, 4) and no secular trends were discernible.

Blade cores are those that show parallel-sided flake scars that are at least

twice as long as they are wide. Cores with convergent-sided flake scars or scars

with a length : width ratio of less than 2 : 1 have not been included, although a

study of the blade cores from Nelson Bay Cave illustrated by J. Deacon (1978,

fig. 9) suggests that many of the cores in the BNK 1 sample that were excluded

from the blade core sub-class would have been included had they been found at

Nelson Bay Cave.

Blade cores have been divided into three size classes:

1. blade: flake scars more than 25 mm long (Fig. 9D);

2. bladelet: flake scars 15—25 mm long (Fig. 9F);

3. micro-blade: flake scars less than 15 mm long (Fig. 9C, G-I).

BYNESKRANSKOP 1 33

ee ee |

Fig. 9. Cores. A-B, E. Radial. D. Blade. F. ‘Bladelet’. C, G—I. ‘Micro-blade’. A. Lime-

stone, layer 1. B. Quartz, layer 1. C. Silcrete, layer 3. D. Silcrete, layer 9. E. Silcrete, layer

10. F. Silcrete, layer 5. G. Quartz, layer 18. H-I. Quartz, layer 19.

34 ANNALS OF THE SOUTH AFRICAN MUSEUM

Frequencies in each size class are given in Table 5, which also includes raw

material frequencies.

TABLE 5

Stone: blade cores; frequencies in size categories, in raw material classes, with core: blade

ratios.

Blade Bladelet Micro-blade |} Layer Core: blade

Flake scar length: >25 mm 25-15 mm <15 mm total ratio

Raw material: OorsS Ot S

Layer

Ns eRe RORY a nee —- — — 1 2 — 3 lee

DRS NE SN — — — —_— — — eis

See icn orem as —- — — 1 1 1 3 ia

A Bisse a a ane ae a —- — — — 3 3 6 iL 24) 2)

a ate Cie mA) Os: — i y) il 1 5 il 3 Sil

6 ey See ee — — —_—- — 1 1:80

TO ae GON ale ne SS 3 eS) 6 ibe ILS

Sree ar. Means —_—- — — 3 = 1 4 252

OSM Mel ae —- tt — 3 6 fl itil 1: 1454

LOPS Bc: ee 1 — 3 2) if 8 Ie FS)

1 pene a RR ee —- — — — — 1 1 125

2). ReMi a 3 Cerne —- — — — 1 1 2 1e22F5

LS) oe Cicer ac ome a Ane: i — —_— — DD 13655

ee eee —- — — — 1 1 2 al

Sars, ORO RE NENG Par oath —_—- — — —_ 2 — 2) 13 @5

NG ae acre cote bon Sea —_—- — 3 3 eat 7

DR co eee —- — — 4 4 1G

1 LC Ra ah aie cap es —- — — 9 10 boar

NOR H ck A. ABR —_— — il 34 40 1:4,8

SHS TOW svcd cocace D2 2 1 69 1113} ESO

Note. Raw materials: Q = quartz, S = silcrete, Qt = quartzite.

The core: blade ratios are based on the total of all unbroken blades, unmodified, utilized, and

retouched.

Blade cores are too few to warrant comment beyond pointing out that they

occur sporadically throughout the deposit. Most of the bladelet cores are from

layer 10 and above, with almost three-quarters of the site total in layers 10—5.

Micro-blade cores have the highest frequency of the three size classes (76,1 %

of the site total) and, although they were found in all but three of the layers,

almost half the total came from layer 19 alone. This layer also has the highest

frequency of unmodified blades, AD as well as actual (Table 4).

As will be seen from Table 5, almost all the micro-blade cores from layers

19 and 18 are quartz, and are of the type illustrated in Figure 9G-I. Initially

these were not considered to be cores, but the identification of similar quartz

artefacts from Nelson Bay Cave as cores (J. Deacon 1978, fig. 9: 16-18) has

resulted in those from BNK 1 being accepted as such. There appears to be

some similarity between these cores and the smaller scaled pieces (see below),

and it seems open to question as to whether some of the artefacts from BNK 1

BYNESKRANSKOP 1 35

classed as micro-blade cores are not in fact broken scaled pieces, or whether

some of those classed as scaled pieces are not actually cores.

Irregular cores are those that cannot be included in either of the preceding

sub-classes. As might be expected, they are the most common type of core

(44,3 % of the class total), although they are not much more common than

radial cores (Table 4).

Silcrete is the most common raw material for cores (53,2% of the class

total) and is the most common raw material for both radial (58,2% of the

sub-class total) and irregular cores (56,7%). Quartz, however, is the most

common raw material for the sub-class of blade cores (67,2 %), although in the

size classes there are more silcrete bladelet cores than quartz (Table 5).

No secular trends are discernible in the distribution of the three core types

through the layers. Excluding layer 19, which has the highest number of blade

and irregular cores (radial cores are most common in layer 7), the lower

frequency of cores in the lower layers is consistent with the generally lower

frequency of all artefactual material in these layers.

The ratio of blade cores to blades (in all categories), also shown in Table 5,

shows neither consistent patterning, nor any secular trends. The ratio is lowest

in layers 5-12 although these layers yielded about a third (33,5 %) of all the

blade cores.

Quartzite pebbles

These are not artefacts in the strict sense and have therefore not been

included in any of the inventories. They are, however, ‘manuports’ that are not

a component of the limestone rock and must have been brought into the cave.

They are round to oval in shape and up to about 2 cm greatest diameter, and

although they were not counted they occurred in sufficient frequency for their

presence to warrant mention. These pebbles are too small to allow their being

considered as potential cores and show no signs of having been used in any

way. The reason for their presence in the deposits can thus only be speculated

upon. Such speculations could include sling-shot or the equipment for a game.

Ochre

This was found throughout the deposit, in various colours ranging from

yellow to dark brown. No attempt was made at quantification, but it was

observed that some pieces showed wear consistent with their having been

rubbed, while others had striations that suggested that they had been scraped.

(See also Combination hammerstones-grindstones, Marine shell ornaments,

Pottery.)

UTILIZED AND MODIFIED ARTEFACTS

Artefacts showing damage as a result of use or with modification of their

shape by or for use (other than those classed as retouched pieces) account for

0,7 per cent of the site total of stone artefacts. They have been divided into

eight classes and an inventory of these is given in Table 6.

36 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 6

Stone: inventory of utilized—modified artefact category.

SE EE eee eee

heavy

Class: flakes blades scaled hammer- combin- ground/ edge- Layer AD/m?

pieces stones ations grooved flaked total

TTT ee eee nooo eee

Layer

1 27 — S/ WZ 1 — 3 80 a4

2 8 — Il) 1 — — — 24 42

3) 9 — 14 — — — — 23 o2

4 43 5 48 1 — = = 97 107

5 74 4 61 4 — Lia 1 146 118

6 62 3 56 1 — 2 Z, 126 150

7 38 2 51 1 — — — 92 114

8 84 Us 41 — = == = 138 238

9 71 9 45 — — — — 25 122

10 42 7 19 — — — — 68 1211

11 13) = 8 — 1 1 — 73) 82

12 52 3 48 — — — — 103 184

i138) IS) — S = = = — 20 44

14 10 — 5 — = if 16 21

iS = — 1 1 — — 2 8)

16 yy — 4 — 1 — 1 8 3)

17 4 2 Z — = = 8 14

18 7 10 38) — — =— = 56 86

19 Si 22 16 3 — — — iZ 67

Site

total Se 80 SS) 24 3 5 8 e277),

Note. a—1 bored stone fragment, 1 grooved stone; b—ground shale palette/pendant;

c—includes 1 milled-edge pebble.

Silcrete is the dominant raw material (71,3% of the category total),

followed by quartz (22,0%), quartzite (5,5%), shale (0,6%), limestone and

chalcedony (0,3 % each). Utilization is more easily observed on silcrete than on

the other raw materials, and quartz and quartzite are probably under-

represented in this category because of the difficulty in identifying utilization on

these raw materials.

Utilized flakes

These account for 48,2 per cent of the category total. They are classified on

the basis of relatively light chipping on the flake edge that is of sufficient extent

to suggest that this resulted from deliberate human activity. The degree of edge

damage, more often in the case of quartzite artefacts, caused problems in

deciding whether an artefact should be classified as utilized or retouched, and it

seems that there is no clear separation of these categories. Heavy utilization

resulting in the removal of relatively large flakes from the edge can produce a

form of modification of the flake that is similar to the deliberate, functional

modification termed retouch.

BYNESKRANSKOP 1 aii)

The utilized flakes are almost all silcrete (90,5 % of the class total), so that

frequencies of quartzite (5,2%) and quartz (3,2%) are low and those of

limestone (0,7%) and chalcedony (0,3%) minimal. Frequencies are low in

layers 18-13, 3 and 2 but there are otherwise no observable trends in the

distribution through the layers.

Utilized blades

These account for 6,5 per cent of the category total. Their distribution in

the layers follows that of utilized flakes more closely than it does those of

unmodified blades or blade cores (Table 4) although they are absent from eight

of the layers. Although layer 19 has a higher actual frequency than layer 8 the

AD for both layers is the same.

Forty-nine of the blades are broken. Of the remainder, using the size

classes defined for blade cores, 11 are blades, 9 bladelets, and 11 micro-blades.

There is no discernible patterning in the distribution of the size classes in the

layers.

Silcrete is the dominant raw material (97,5 % of the class total), with one

blade each of quartzite and chalcedony making up the total. It is of some

interest that, although quartz accounts for 67,2 per cent of the total of blade

cores and 87,5 per cent of those in layer 19 alone, not one utilized quartz blade

was found.

Scaled pieces

More commonly known as piéces esquillées, these are artefacts with chip-

ping or crushing of one or more edges, usually but not necessarily the shorter

edges of roughly rectangular pieces. As mentioned in the discussion of cores,

some doubt exists as to whether some of the artefacts in the BNK 1 stone

assemblage that have been classified as quartz micro-blade cores are not, in

fact, broken scaled pieces, or whether some of those classed as scaled pieces

are not actually cores. Some support for this observation is provided by the

comment by J. Deacon (1972: 14) on the crushed edges of the piéces esquillées

from Wilton: “This sort of modification has often been found on worked-out

bladelet cores and may possibly be due to attempts at removing further

bladelets.’

As will be seen from Table 6, scaled pieces were found throughout the

deposit, but with higher actual frequencies in layers 18, 12, 9-4, and 1.

Although layer 5 has the highest actual frequency, it is layer 12 that has the

highest AD, and the AD for layer 5 is, in fact, lower than those for layers 18,

12, 8-6 and 3. This provides a good example of the problems inherent in

comparing the actual frequencies of material from deposits of unequal

volume.

Although about 80 per cent of the micro-blade cores are quartz, the scaled

pieces are more or less evenly divided between silcrete (50,3% of the class

total) and quartz (48,9 %), with four of chalcedony completing the total.

38 ANNALS OF THE SOUTH AFRICAN MUSEUM

Hammerstones

In the BNK 1 sample, these are quartzite pebbles or cobbles with bruising

and pitting on their ends and/or sides. As indicated in the notes on Table 6, the

total for layer 19 includes a milled-edge pebble: a flat, circular pebble with

bruising round its circumference (Fig. 10C).

Hammerstones account for only 2 per cent of the category total. Because

they are distributed sporadically throughout the deposit, in only eight of the

layers, lower as well as upper, no significance can be attached to their absence

from any of the layers. That there are more in the upper layers than in the

lower is consistent with the distribution of most of the artefact classes, but the

fact that half the site total came from layer 1 is not entirely attributable to the

fact that this layer has the largest excavated volume.

Combination hammerstones—grindstones

These are quartzite cobbles that have the characteristics of both hammer-

stones and upper grindstones: bruising and pitting from hammering and the

flattening of end or side through grinding.

Only three of these artefacts were recovered, from layers 1, 11, and 16.

That from layer 1 has its grinding-surface heavily stained with brownish ochre.

That from layer 11 (Fig. 10A) is conventional in that it is an elongated cobble

with grinding-wear at one end and pitting from hammering at the other. There

is also a small amount of pitting on the ground surface. The third, from layer 16

(Fig. 10B), has grinding-wear on at least three of its six faces and pitting on its

ends, one of which is also one of the ground surfaces. As indicated in the

illustration, flakes have been struck from this piece so that it is technically also

a core, but it has not been included in the counts for that class.

Ground stone artefacts

These shale pieces consist of bored stone fragments, a grooved stone and a

broken palette or pendant. The bored stone fragments are from layers 5 and 6,

and the fragments from layer 6 are clearly pieces of two different artefacts. The

grooved stone, also from layer 5 (Fig. 11B) is somewhat unusual in being a worn

pebble with the groove on the edge instead of across one of the flat surfaces.

The broken palette or pendant, from layer 11 (Fig. 11A), is a finely ground

piece of shale whose surfaces have a metallic, pewter-like sheen. The perfora-

tion near the edge has been bored obliquely (i.e. grooved) from both sides.

Heavy edge-flaked pieces

These are split quartzite cobbles with evidence of large flakes having been

struck from the cortical surface of the split edge (Fig. 12). Artefacts of this type

were first described from Nelson Bay Cave (J. Deacon 1978: 91-92) and are

also included in the stone artefacts from Die Kelders (Schweitzer

1979: 177-178). As indicated in Table 6, their distribution in the BNK 1

assemblage is sporadic and with low frequencies.

BYNESKRANSKOP 1 39

Fig. 10. A-B. Combination hammerstones—upper grindstones, quartzite, layers 11 and 16.

C. Milled-edge pebble, quartzite, layer 9.

40 ANNALS OF THE SOUTH AFRICAN MUSEUM

—_—

= =

> oo

fe ire

> Rees

. (Ge

<> 5 x

4 Ae

y \

y 1

Fig. 11. A. Ground shale palette/pendant (broken), layer 11.

B. Grooved shale pebble, layer 5.

BYNESKRANSKOP 1 41

Fig. 12. Heavy edge-flaked pieces, quartzite. A. Layer 6. B. Layer 5.

RETOUCHED ARTEFACTS

The artefacts in this category are those that show deliberate modification

of the ‘blank’ (flake, blade, etc.) for the purpose of creating either a specific

shape, e.g. segments, or a particular type of working edge, e.g. scrapers. It also

includes a number of pieces that resist formal classification and are therefore

termed ‘miscellaneous retouch’. The modification of these artefacts appears to

have been deliberate and is generally distinct from the modification resulting

from casual utilization. Some sub-classes can be recognized within the class of

miscellaneous retouched artefacts, but the number of pieces is too low to

warrant their being placed in a separate class.

Classification of the retouched artefacts follows the conventional typology,

but there are problems that result from the degree to which the artefacts vary in

their attributes. In very few instances is it possible to draw an absolutely

42 ANNALS OF THE SOUTH AFRICAN MUSEUM

SCRAPERS BACKED

SCRAPERS

ont

o? @

UTILIZED MISCELLANEOUS Guce t

FLAKES RETOUCH SEEN ENS

BORERS BACKED

(POINTS) FLAKES

Fig. 13. Diagram indicating the intergrading of stone artefact types.

Broken lines indicate less common areas of intergrading.

clear-cut division between the classes or, as previously mentioned, between

artefacts classified as utilized-modified and those classed as retouched. Figure

13 gives some indication of the directions of intergrading but it would require a

multivariate attribute analysis to quantify the degree of intergrading in a

meaningful manner. It is for this reason that the more usual term ‘formal tools’

is not used in this report, and aspects of the problem are touched on in the

discussions of the various classes.

The inventory of the classes in the retouched artefact category is given in

Table 7 and the layer frequencies in Figure 14, in histogram form. In general,

the frequencies for the layers below layer 12 are too low to be amenable to

meaningful statistical analysis or even conversion to percentage frequencies.

Although this fact is clearly not without archaeological significance, it tends to

create a bias in any discussion in favour of the upper layers.

As is the case with the utilized and modified artefacts, the dominant raw

material for retouched artefacts is silcrete (85,2% of the category total),

followed by quartz (11,9%), quartzite (1,9%), and negligible frequencies of

limestone and chalcedony (0,5 % each). Quartz is not considered to be under-

represented in this category through lack of recognition of retouch since

modification of the shape makes recognition easier. There are secular trends in

the frequencies of raw materials in certain classes, and these are discussed

under the relevant heading.

Scrapers

These account for 52,1 per cent of the category total and, as indicated in

Figure 14, are the dominant retouched artefact in every layer below layer 3.

Examples of the more common types of scraper are illustrated in Figure 15, and

of the ‘extra large’ types in Figure 16.

The ADs for scrapers are less than 10 in layers 14-17 and 19, less than 100

in layers 1-2, 13 and 18, less than 200 in layers 3-4, and 10-11, and less than

300 in layers 5, 7, 9 and 12, while layers 6 and 8 have ADs in excess of 300.

There is thus no clear pattern of increase or decrease in the frequency of

scrapers beyond the fact that layers 19-14 have the lowest ADs and layers 12

and 9-5 the highest. When scrapers are compared with other artefacts in the

BYNESKRANSKOP 1 43

Layer total 276 108 100 310 AO 448 351 291 386 103 54 150 29 8 3 5 / 18 18

20-

10-

adscellapeers 2 2 2 3 8

10-

Borers

(drills) <| <I 1

10-

(points) (SSL eee ee

10-

Backed <I <I <l_<l 2

10-

Faqsee <1 2

Backed

segments

20-

10-

Backed

Scrapers <i <1

60-

50-

40-

30-

20-

10-

Adzes <i 2 ] 2

Scrapers

Layer ] 2 3 4 5 6 7 8 iF) 10 #11 12 13 14 15 16 17 18 19

Fig. 14. Histograms of percentage frequencies, by layer, of retouched stone artefact classes.

<1=less than 1 per cent of the layer total given at the top of the figure. Frequencies in layers

13-19 are too low for conversion to percentages and the actual frequencies are given instead.

Raw material frequencies are indicated for scrapers and segments only.

POP €

81 8I

8Z 81

ZI L

Ss 8 ¢

=) G €

2 OI 8

9 67

= 397 OST

a €61 1S

oO P81 €OL

—

2 LLE Se

a Z0S 162

Ms ee Ise

= EES lad

A 86S OrL

a Ore Ore

= 00r 001

= 681 801

em 901 9LZ

< 1210)

<wydv Jakey

“OSTUL

‘al <=! VS! SS!

OnArDAEe— NAN

soyey

peyojou

cor

(st[lIp)

s1010q

vil LI

= 6

I a

I =

I Sats!

8 S

8 a

OL =

OL I

oY ¢

8 €

I =

9 —.

vl I

(s1910q)

sjutod sapeiq

trl LC vl c6e CSC viv YL \I [e30}

StS)

= I = == aa ae 8 61

Cc = ae = = C Il 81

ca me = € = ms Cc LI

a — = I oom I € oT

= = are = = a I ST

ae a =< aad oa G 9 tI

a = — == a a 9C €l

I a = ¢ = 81 SIT ai

€ I = if a € 8e II

Cc oa — el C el 89 Ol

(&C if 9 CL € Ol 9CC 6

61 G I vv 6 Cc 98T 8

G6 = € Ip SI ¢ TE? L

61 6 I Le cL 6 09C 9

Le 9 € 801 Ol TS ele ¢

[f ¢ == eC ve 08 si v

C a 35 Cc aa 9S ve €

a = ao C I 8S ce G

9 © = Iv € vol 68 I

IokeT

soyey sanbijqo spiozoden syuaudas siadeios

saoaid payoeq sazpe siodeios S71)

*K10391e9 JORJOIIe Poyono}aI JO AIOJUDAUT :9U0}S

L alav ys

BYNESKRANSKOP 1 45

Fig. 15. Scrapers, general range. A. Silcrete, layer 10. B—C. Silcrete, layer 6. D, F. Silcrete,

layer 4. E. Silcrete, layer 19. G—I. Silcrete, layer 1. J-K. Silcrete, layer 6.

ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 16. ‘Extra large’ scrapers, both limestone. A. Layer 6. B. Layer 5.

BYNESKRANSKOP 1 47

category, however, there is a generally declining trend in the layer frequencies,

as indicated in Figure 14.

Silcrete is the dominant raw material (83,8 % of the class total), with only

low frequencies of quartz (13,2%), quartzite (2,0%), limestone (0,7%) and

chalcedony (0,3%). There are, however, discernible secular trends in the

frequencies of quartz scrapers (Fig. 14). These decline from layer 12 (19,5 % of

the layer total) to layer 6 (3,1 %) then increase sharply from layer 5 (14,1 %) to

layer 3, in which the frequency (47,0%) is almost equal to that of silcrete

(50,0 %). Lower frequencies of quartz occur in layers 2 and 1 (31,4% and

41,6%), but the ratio of quartz to silcrete is higher in these layers than in layer

5 and below. The quartz: silcrete ratio declines from 1:4 in layer 12 to 1:30 in

layer 6, layer 9 having an anomalous 1: 54. In layer 5 the ratio is 1: 6; in layers

4-2, between 1:1 and 1:2; and in layer 1, 1:6.

A limited morphological analysis of scrapers was undertaken, measure-

ments being taken of the three plane dimensions of all unbroken pieces. The

results of this analysis, together with the length:width ratio, are presented

graphically in Figure 17. What is immediately apparent is that secular variation

is virtually identical for all four attributes. Frequencies in the layers below layer

12 are low, and therefore statistically less reliable, and those in layers 15-19 are

too low for any meaningful observation. However, the trends may be sum-

marized as follows:

in layers 14~-11 size increases, shape changes from wide to long;

in layers 10-6 size decreases, shape changes from long to wide;

in layers 5—1 size increases, shape changes from wide to equilateral.

Layer 14 appears to mark the beginning of the trend that continues until

layer 10, while layers 15 and 17 in most respects reverse the trends indicated

by layers 19, 18 and 16. The changes in size and shape are most marked in

layers 12-10, with layer 9 intermediate between this group and layers 8-1. It

must be pointed out, however, that layer 11 has three ‘extra large’ scrapers,

47X45x13mm, 50*55x19mm, and 63xX77xX21mm, and although

scrapers of similar dimensions are found in most other layers the presence of

three in a sample as small as that of layer 11 (n =38) helps account for the

marked shift in the means for all the attributes indicated in Figure 17. In

layers with higher frequencies these ‘extra large’ scraper dimensions are offset

by the greater numbers of scrapers with dimensions within the normal range.

The overall variation from layer 8 up is not great and, as indicated in Figure

17 by the 2 X standard error bars, generally not statistically significant, so that

scrapers from these layers may be considered as morphologically homo-

geneous.

There are fifty-three scrapers in the BNK 1 sample that have, in addition

to the normal scraper retouch, step-flaking on their laterals of a kind usually

associated with adzes (see below). These were recovered from layers 14—1

except layers 7 and 2. Layer 12 has twenty-one of these artefacts and layer 9

has eleven, but in the other layers frequencies are low.

48 ANNALS OF THE SOUTH AFRICAN MUSEUM

Length Width Thickness Length:width ratio

--cie—

-a-- — + $s

-Sle—

Fig. 17. Modified Dice—Leraas diagrams of scraper dimensions and length: width ratios. Layer

sample frequencies are given on the right. For details of symbols, see Figure 8.

During the initial analyses it was considered that these artefacts repre-

sented a form transitional between the long, thick end-scrapers, which

appeared more common in layers 13-10, and adzes proper, which become more

common in the upper layers. In the preliminary report on BNK 1 (Schweitzer

& Wilson 1978: 138) it was speculated that ‘adzes and small scrapers may have

taken over most of the functions previously fulfilled by medium and large

scrapers’. However, these ‘scraper—adzes’ have their maximum frequency in

layers 12 and 9, while adzes do not become common before layer 4, and the

fact that this artefact type is found right up to layer 1 negates the possibility

that it might represent a transitional stage between the industries of the lower

layers and those of the upper. B. D. Malan has suggested (1980 pers. comm.)

that these might merely be scrapers made on large flakes, of which the width

was reduced in order to facilitate hafting.

BYNESKRANSKOP 1 49

A further sixty-one scrapers were found to show utilization of part of the

unretouched edge. Apart from one in layer 19, these came from layer 12 and

above but were not found in layers 11, 9 and 3, and the highest frequency

(twenty-six) came from layer 5. These are another example of the intergrading

of artefact types within and between categories and classes, and it is only the

acutely-angled ‘scraper’ retouch that distinguishes these pieces from backed

flakes (see below), many of which also show utilization of the unretouched

edge. This utilization on the scrapers possibly represents casual use of an

artefact primarily intended for another purpose. Although the number is few

(3,4% of the scraper total), the evidence of utilization on these scrapers

suggests that not all scrapers, or even all small scrapers, were mounted in

mastic (cf. Deacon & Deacon 1980).

Scrapers with indisputable traces of mastic were not found in the BNK 1

sample, although there is one broken artefact included with the ‘miscellaneous

retouch’ (see below and Fig. 24M) that is partly covered in mastic and may be a

broken scraper.

Adzes

These are artefacts that have step-flaking on one or both of their laterals.

They account for 12,2 per cent of the retouched artefact category but although

they are present in fifteen of the layers, from layer 18 up, they only become a

notable feature of the layer assemblages from layer 4 up. The ADs for these

artefacts, as well as the percentage frequencies (Fig. 14), indicate that adzes are

relatively most common in layer 3. They are almost all made of silcrete (98,5 %

of the class total), the remaining six being of quartz.

The adzes from BNK 1 are mostly oblong pieces made on thick flakes or

chunks. Some are unilaterally retouched, others bilaterally, and some of the

latter assume a convergent, more triangular shape. Attempts to classify adzes

according to shape, size, or position of retouch were not satisfactory owing to

the degree of variation within each layer sample. No typological or morphologi-

cal trends could be discerned, and the examples illustrated in Figure 18 are

representative of the site sample as a whole.

The dimensions of the adzes are generally consistent throughout the

sequence. Length ranges from 11 to 47 mm with a mean of 26,2 mm, and width

ranges from 6 to 37 mm with a mean of 17,3 mm. In all but three of the layer

samples the length does not vary from the site mean by more than 1,5 mm. The

exceptions are layers 7, 11 and 18 and the variation is 5,8-7,9 mm. The mean

width of the layer samples shows more variation, with only eight of the samples

varying from the site mean by 1,5 mm or less. In the other seven layers the

mean varies from the site mean by 1,9-10,7 mm. However, despite the fact that

the two adzes from layer 18 are larger than the rest (mean dimensions

32 xX 28 mm) (Fig. 18G, K), no patterning is evident in the morphological

variation.

50 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 18. Adzes, all silcrete except C, which is crystal quartz. A, F, M. Layer 1. B-C, E,

I. Layer 4. D, H, J, L. Layer 12. G, K. Layer 18. Stippled areas on A, C and F indicate

cortex.

BYNESKRANSKOP 1 51

Although, as indicated in Table 7 and Figure 14, there are adzes in fifteen

of the layers, below layer 12 there are only five (1,2% of the site total), in

layers 14, 16 and 18. The distribution in layers 12-10, although irregular, is

higher than in the layers immediately above or below; but, from a very low

frequency in layer 8 (n= 2), there is a progressive increase in frequency in the

overlying layers, which corresponds to a decline in scraper frequencies (Fig.

14). In layer 8 adzes account for only 0,7 per cent of the layer total of

retouched artefacts and scrapers for 63,9 per cent, but by layer 3 adzes have

increased to 56 per cent and scrapers have declined to 34 per cent. The decline

in the frequency of adzes in the overlying layers, to 37,7 per cent in layer 1, is

not matched by a corresponding increase in scraper frequencies, which reach

only 32,2 per cent of the category total in that layer. The adze: scraper ratio in

layer 1 is 1,2: 1, as against 1,7: 1 in layers 2 and 3.

Backed pieces

These are artefacts with abrupt retouch of one lateral, at right angles to the

plane of the flake or blade. That this definition does not always hold good, and

that it causes a division between the two types of borer are points that are

discussed later in the report; but by and large it holds good for the artefact

types included in this class.

Backed pieces account for 26,9 per cent of the retouched artefact category

total and, although found in all layers except layers 15-13, are most common in

layers 9-4, where individual layer frequencies exceed 24 per cent of the

category total for the layer and reach a maximum of 41,3 per cent in layer 5._

In common with the other classes in the category, backed pieces are

mostly. silcrete (82,3% of the class total). Differences in the relative fre-

quencies of the seven sub-classes listed in Table 7, as well as in raw material

frequencies, indicate, however, the profitability of individual analysis of the

sub-classes.

Backed scrapers (Fig 19) are the generally ellipsoidal artefacts otherwise

known as ‘double crescents’ or ‘double segments’. The name used in this

report, although in some ways more accurately descriptive than the other

terms, is not entirely satisfactory. While these artefacts almost invariably have

‘scraper retouch’ on one of the arcs, the retouch, or ‘backing’ on the other arc

varies considerably in its angle. Strictly, ‘backing’ should be at 90° to the plane,

but many of the backed scrapers have ‘backing’ at angles of 60° or less, which is

more like scraper retouch, and these pieces would perhaps be more correctly

termed ‘double scrapers’. Also, on some of the pieces the modification of the

arc opposite the backed arc is more like heavy utilization than deliberate

retouch. However, these artefact types, despite the differences mentioned, are

morphologically more like each other than they are like any of the other classes

or sub-classes, and there seems no good reason for separating them.

As the various names for these artefacts imply, backed scrapers are

typologically intermediate between segments and scrapers. They have the

2 ANNALS OF THE SOUTH AFRICAN MUSEUM

eet

Fig. 19. Backed scrapers, all silcrete except D, which is crystal quartz. A. Layer 1. B. Layer 2.

C-E. Layer 4. F-I. Layer 5. J-L. Layer 6. M. Layer 7. N-O. Layer 8. P. Layer 9. Q. Layer

10. Stippled areas on J and K indicate mastic traces.

BYNESKRANSKOP 1 53

backed arc of segments but not their opposing chord, having instead a convex

retouched edge like that of scrapers.

The fact that convex scrapers, segments, and backed scrapers all occur in

the same assemblages suggests functional differences between the three types.

Although metrical analyses were not undertaken, it is evident that backed

scrapers fall within the size ranges of both the smaller convex scrapers and

segments (cf. Figs 15, 19 and 20).

One of the pieces (Fig. 19N) is more like a ‘chunky segment’ (see Segments

below), but has retouch on the ventral surface, unlike the others in the ‘chunky

segment’ sub-class, which typically have retouch on the dorsal surface, as

indicated in Figure 21. Two pieces (Fig. 19J—K) appear to have traces of mastic,

as indicated by the stippling in the illustrations.

Sampson (1974, fig. 106) has suggested that backed scrapers are backed

flakes that were jettisoned when their utility was exhausted. In view of the

variability observed in the BNK 1 sample, this interpretation seems plausible.

There seems, however, not enough of the gradation between these two classes

of artefacts that might be expected in such a circumstance. It seems unlikely

that only lightly utilized backed flakes and heavily utilized backed scrapers

would remain on site. Also, as indicated in Table 7 and Figure 14, there seems

an insufficient correlation between the frequencies of each type in the indi-

vidual layers to provide support for Sampson’s hypothesis. This is, however, a

problem that could possibly be solved by replication studies.

Backed scrapers are a relatively small component of the retouched artefact

category (7,4% of the site total), although they account for almost a third of

the backed pieces. They are almost entirely (96,8%) made of silcrete, the

remainder being quartz, with one of chalcedony. They were not found below

layer 10 and are relatively common only in layers 6-4, in which 75 per cent of

the site total were found. There is a marked falling off in frequency in layers

3-1 which suggests that, at BNK 1 at least, the popularity or utility of this

artefact type was episodic. The ADs indicate the relative frequencies of these

artefacts: that for layer 7 is 18; those for layers 6 and 5, 89 each; and that for

layer 4, 37.

Segments (Figs 20-21), flakes or blades backed to form an arc opposed to a

sharp-edged, unretouched chord, are the most common backed artefact in the

assemblage (41,9% of the class total) and rank third in frequency in the

retouched artefact category (Table 7). Although these artefacts are known from

late Pleistocene assemblages (cf. Sampson 1974, table 32), there appear to be

both temporal and geographic gaps in their distribution in Holocene deposits.

In the BNK 1 sample, apart from four in layers 16 and 17, their occurrence is

restricted to layer 12 and above, i.e. the last 8 000 years or so.

The segments are mostly silcrete (71% of the sub-class total) and the rest

are quartz. There is some variation in the individual layers of the ratio of quartz

to silcrete that is not simply related to the numerical fluctuations (Fig. 14).

Ignoring layers 12, 11, 3, and 2, which contain five segments or fewer, the ratio

54 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 20. Segments. A-D, F. Silcrete, layer 5. E, I-J. Silcrete, iayer 9. H. Crystal quartz, layer

9. K-P, R-S. Silcrete, layer 6. Q. White quartz, layer 1.

BYNESKRANSKOP 1 55

Fig. 21. ‘Chunky’ segments. A-B. Silcrete, layer 4. C. Crystal quartz, layer 5. D-E, P. Sil-

crete, layer 5. F-H. Silcrete, layer 6. I-K. Silcrete, layer 7. L-M. Silcrete, layer 9. N-O. Sil-

crete, layer 10. Stippled areas on I indicate mastic traces.

56 ANNALS OF THE SOUTH AFRICAN MUSEUM

of quartz to silcrete rises from 0: 13 in layer 10 to 1: 1,7 in layer 7, declines in

layers 6 and 5 (1:6) and rises again from 1: 1,3 in layer 4 to 5,8: 1 in layer 1.

The trends in segment frequencies, regardless of raw material, are for an

increase from layer 12 to layer 9, a decrease until layer 6, with an apparently

anomalous increase in layer 5 followed by a decrease in layers 4 and 3 and

increases in layers 2 and 1. (Fig. 14). The ADs for layers 9-4 are: layer 9, 71;

layer 8, 76; layer 7, 51; layer 6, 44; layer 5, 87; layer 4, 25. This tends to

support the trends indicated by the actual frequencies, and shows that the

increase in layer 5 is not merely related to an increase in this layer’s excavated

volume.

Slightly fewer of the segments are made on blades than on flakes (44,7 %

of the site total), although in layers 10-7 segments made on blades are more

common (52-64 % of the layer totals).

A problem in the classification of segments is their tendency to grade into

backed flakes. There was little agreement among those who worked on or

examined the BNK 1 stone artefacts regarding marginal cases, and the alloca-

tion of these has been somewhat arbitrary. Another type that caused problems

was the informally named ‘chunky segment’ (Fig. 21). These artefacts have

coarse, broad retouch on the arc, and the possibility was considered that this

might not be retouch at all, but that the artefacts might be a type of core

rejuvenation flake struck obliquely from the end of a cylindrical (polar or

blade) core and the ‘retouch’ no more than flake scars. However, the type of

core from which such flakes might have been struck does not occur in the

assemblage and these artefacts were accordingly classified as segments or

backed flakes depending on the degree of resemblance to either type. It would

perhaps be closer to the truth to say that those that looked less like segments

were classed as backed flakes.

Intergrading between segments and backed scrapers was less of a problem

since, although a few segments have utilization of the chord that is heavy

enough to be called retouch, it does not extend the full length of the chord, nor

does it cause the chord to assume a convex shape. There was no evidence from

the site sample that prolonged utilization of segments might result in the double

arc of the backed scraper.

Figure 22 presents graphically the results of metrical analyses of the

dimensions of unbroken segments. These show that the mean size of quartz

segments in any layer is lower than that of silcrete segments. Although quartz

segments vary little in length in layers 9-7, those in silcrete show a diminishing

trend. In layer 6 the mean length of both quartz and silcrete segments is highest

for their respective raw materials, and from this layer up the segments tend to

become shorter, although the quartz segments from layer 1 are on average no

shorter than those from layers 3 and 4.

Secular trends are more evident in width. There is an increase in the mean

width of both quartz and silcrete segments from layer 10 up. This reaches a

maximum in layer 6, as is the case with length, so that the segments from this

BYNESKRANSKOP 1 Si,

Layer no

Saawe Length Width Length: width ratio no.

1@ -— —— fi —_ 31

1s ° : : 1

2 Q 1

RS 1

1o 20 30 40 5 10 15 20 io 20 30 40 50

Fig. 22. Modified Dice-Leraas diagrams of segment dimensions and length: width ratios.

Q = quartz, S =silcrete. Layer frequencies are given on the right. For details of symbols, see

Figure 8.

58 ANNALS OF THE SOUTH AFRICAN MUSEUM

layer are, on average, the largest in the sample. From layer 5 up the segments

tend to become narrower, although the quartz segments in layer 1 are on average

wider than those of layer 4 and about the same width as those of layer 5.

In shape, i.e. length: width ratio, there is a trend from a more or less

equidimensional shape in layer 10 to a somewhat longer shape in the silcrete

segments from layer 8, while the quartz segments from this layer indicate the

trend that continues in the upper layers, towards a wider form that reaches its

maximum in layer 6. Layer 5 has a somewhat anomalous reversion to a longer

shape in the silcrete segments and an equidimensional shape in the quartz, but

after this the trend towards a wider form continues.

These trends may be summarized as follows: quartz tends from a longer

shape (2:1 ratio) in layer 12 to an equilateral shape (1:1 ratio) in layer 6,

reverting to a somewhat longer shape (1,5: 1 ratio) in layers 5, 4 and 1; silcrete

tends from a longer shape (2:1 ratio) in layers 10-6 to a slightly longer shape

(nearer 2,5: 1 ratio) in layers 5 and 4.

Fifty-three segments (13,5% of the site total) appear to have traces of

mastic. As it was observed that most of the segments (376, being 95,7 % of the

site total) show utilization of the chord a more intensive analysis of these was

undertaken by D. Halkett.

Of the 17 segments that show no traces of utilization, 13 are silcrete, 3

white (‘milky’) quartz and 1 crystal quartz. Since it is only on white quartz that

utilization is difficult to detect, it is reasonable to assume that at least 14 of the

segments were not utilized. This is supported by the fact that 46 of the 111

utilized quartz segments were white quartz, showing that it is possible to

recognize utilization even on this raw material. Broken artefacts (87) were

excluded from the analysis as being incapable of providing complete informa-

tion regarding the extent and placement of utilization.

Results of Halkett’s analysis are given in Table 8. Two main groups,

unifacially and bifacially utilized, were recognized. Unifacially utilized seg-

ments account for almost 70 per cent of the sample total and could be divided

into four types, each subdivided according to whether the flake scars resulting

from utilization are on the dorsal or ventral surface. To counter the suggestion

that type 3 is merely an inverted version of type 1, it must be pointed out that

when dorsal and ventral surfaces are taken into account such an inversion is not

possible. To further obviate the possibility of error the segments were sorted

according to graphic convention, with the dorsal surface up and the arc facing

left.

The most common type of unifacial utilization is along the whole length of

the chord (type 4, 41,5 % of the group total), and in this type utilization on the

dorsal surface exceeds that on the ventral surface by a factor of 2:1. The

ranking of the other three types, in order of frequency, is type 1 (26%), type 3

(23,2%), and type 2 (9,2%). In each case the frequency of segments with

utilization on the dorsal surface exceeds that of utilization on the ventral

surface. (It must be stressed that it is the term ‘utilization on’ that is used, not

BYNESKRANSKOP 1 59

TABLE 8

Stone: analysis of utilization types of segment chords.

bifacial

de. S10) © 7

Class: unifacial

Type: 1 2 3 4

Surface: Di eaNes 7D oN;

Ce ee

oGb0SDOoOnbooooDoO oOo obooooKa ooo

Se ee

Ce ee

lo ee eres | eves | |

Description of types (position of utilization):

unifacial: type 1—upper only; type 2—middle only; type 3—lower only; type 4—whole length.

(D = dorsal surface; V = ventral surface)

bifacial: type 5a—upper dorsal and lower ventral; type 55>—lower dorsal and upper ventral;

type 6—whole dorsal and upper/lower ventral; type 7—whole length, dorsal and

ventral.

‘utilization of’. The use of the term indicates the surface on which the bulk of

the flake scars are visible.)

Of the 90 segments with bifacial utilization the most common are type 5b

(34,4% of the group total), while frequencies of the other three types are

similar (20,0-23,3 %). Types 5a and 5b are considered variants of one type

because when both surfaces show utilization it is less certain that one subtype is

not an inverted version of the other than is the case when the artefacts show

only unifacial utilization. The utilization patterns of types 5a and 5b could

result from reversal of the segment during the period of its use.

If the bifacially utilized segments are considered as artefacts that have been

used more than once and their utilization patterns are combined with those of

the unifacially utilized segments, two major utilization modes can be recog-

nized, and one minor:

mode 1: utilization of the end only (54,7% of the total), types 1, 3, 5 (a

and b), 6;

mode 2: utilization of the whole chord (39,3 %), types 4, 6, 7;

mode 3: utilization of the middle only (6,0%), type 2.

Type 6 is included in both modes 1 and 2 since the type of utilization on

each surface falls into a different mode.

60 ANNALS OF THE SOUTH AFRICAN MUSEUM

Halkett’s examination of the segments with traces of mastic did not allow

him to reach firm conclusions regarding the relationship between the area

apparently covered with mastic and the utilization of the chord. There were too

few artefacts on which the mastic traces were sufficiently well defined for the

area covered by the mastic to be determined with any confidence. The indica-

tions seem to be, however, that where the mastic was at an angle to the chord

and did not cover the whole of the arc, utilization was restricted to the end

(mode 1). Where the mastic was parallel to the chord and covered the arc,

utilization was usually of the whole chord (mode 2). In the seven cases where

the middle only was utilized (mode 3) the mastic appears to have been parallel

to the chord, as in mode 2. There was, however, one case where the mastic

appears to have covered only about a third of the length of the segment.

A segment from layer 6 (Fig. 20M) is somewhat anomalous in that it

combines more than one of the utilization modes described above. It has dorsal

utilization on the lower half of the chord and ventral utilization on the upper

half, which also includes a trimmed notch. Mastic traces on the lower ventral

surface extend on to the arc and suggest that this. was not completely covered

by the mastic. Wear of the transverse flake scars on the dorsal surface adjacent

to the utilized and notched part suggests that the utilization on the upper part

of the ventral surface of the chord resulted from a ‘push-plane’ type of action in

which the dorsal surface was abraded by contact with the artefact (?) the

segment was being used to work.

Although it would be necessary to carry out replication tests in order to

support any assumptions derived from the study of the utilization modes and

mastic traces where the latter exist, it seems clear from Halkett’s analysis that

segments were not all used in the same way. The implication of this is that they

were not all used for the same purpose, and possibly that some were used for

more than one task.

Clark (1959, fig. 29) and Phillipson (1976: 140) have provided examples of

how segments, amongst others, might have been hafted. The variety of the

positions and, in the case of Phillipson’s examples, the range of shapes, would go

some way towards explaining utilization modes 1 and 2 but not mode 3, but they do

not explain why utilization on the BNK 1 segments is more commonly unifacial.

If a cutting or scything action were employed, chipping on both ventral and

dorsal surface could be expected, but unifacial utilization scars seem to indicate

the application of pressure on one side of the chord, such as might be used in a

unidirectional scraping or paring technique in which the segment was held

parallel or at an acute angle to the surface of the object being worked. Halkett

(1979 pers. comm.) is of the opinion that utilization of the whole segment chord

would be consistent with hafting of the type in Phillipson’s examples 3-8, and

utilization of the ends only with Phillipson’s examples 1 and 2. Utilization that

produces only unifacial damage to the chord seems, however, contrary to

Phillipson’s suggestions of use as transverse arrowheads or ‘for pressure-cutting

rather than impact cutting or scraping’ (Phillipson 1976: 218).

BYNESKRANSKOP 1 61

Clarke (1979, figs 1-2) has illustrated twenty-two hafted composite tools

set with microliths and microblades. Seven of these are based on ethnographic

models and the others on archaeological finds in Mesolithic contexts. While not

all forms are appropriate to segments, they serve to illustrate the wide range of

hafts in which segments and similar artefacts could have been mounted, and

which would have produced different edge-wear patterns. They also serve to

remind that morphologically different artefact types need not necessarily be

restricted to different types of tools and functions.

The problem, in a southern African context, is that there is virtually no

evidence, archaeological or ethnographic, for the use of hafted implements of

the types illustrated by the authors previously mentioned. J. Parkington has

shown one of the authors (M.L.W.) an artefact from De Hangen, which is a

lump of mastic in which a segment has been mounted so that it projects at an

oblique angle in such a way as to suggest its use as a cutting or graving tool.

The mastic-mounted quartz artefact from Die Kelders layer 12 (Schweitzer

1979, fig. 37) was considered as possibly being one of the short and wide ‘petit

tranchet’ type of segments (Fig. 20H, Q), but X-ray photography of the artefact

failed to produce an image clear enough to substantiate this.

The seven glass-tipped arrowheads and an eighth tipped with either crystal

quartz or glass that were made in 1878 for Dr W. H. I. Bleek by the Bushman

Jantje, and of which an example was illustrated by Goodwin (1945, fig. 2) do

not have segment-shaped tips. It seems doubtful in any case whether the

segments from BNK 1, even if mounted in the manner illustrated by Goodwin,

would have been effective as arrowheads: they are not notable for having sharp

tips, and in many cases the junction of chord and arc hardly forms a point at all

(cf. Figs 20-21).

Three unusual types of segment must be mentioned, although they are too

few to warrant separate classification.

1. Notched. Seven segments, from layers 1, 5-6, 8-9, have a notch in the

centre of the chord (Fig. 20E-—G, M). The segment from layer 6 has already

been mentioned; those from layers 1 and 5 have type 5b utilization, that from

layer 8 and one from layer 9 have type 5a, and the remaining two, also from

layer 9, type 4. The notch is thus apparently not directly associated with a

particular type of utilization. It might have been accidental, although the notch

on the layer 6 segment is not a single flake removal but the result of repeated

chipping. The notch may have been made to facilitate securing the segment to

its haft by means of a thong.

2. Flanged. The five segments of this type are from layers 1, 5 and 12, and

are illustrated in Figure 20A—D, but most notably B and D; also Figure 21G.

The flange is not a fracture of the chord after the flake was struck from the core

since the whole of the chord, including the flange, has a sharp edge. Four of the

segments have type 4 utilization and the fifth, from layer 5, type 7. These

segments thus show a common pattern of utilization of the whole length of the

62 ANNALS OF THE SOUTH AFRICAN MUSEUM

chord, which suggests that the presence of the flange may have been irrelevant

to the artefact’s function.

3. Denticulate. Two very small segments, 7mm and 8 mm long, were

recovered from layers 9 and 12. That from layer 9 is illustrated in Figure 20H.

Both are made of crystal quartz and the notches in the chord are too deep and

too regular to be the result of utilization similar to that which caused the

edge-wear on the bulk of the sample. The regularity of the notching suggests

that the denticulation was a deliberate effect, but the function of such small

saw-edged artefacts can only be guessed at.

Trapezoids are parallel-sided flakes or blades with their ends backed

obliquely to the length of the piece to produce a trapezoidal shape (Fig.

23A-—M). They differ from the sub-class of obliquely backed pieces (see below)

in that the latter have only one end backed while trapezoids have both.

Whether this difference represents a ‘real-life’ difference or merely a typologi-

cal one is a matter that cannot be determined, since the function of these

artefact types is not known.

Trapezoids were found only in layers 9-5, with the highest frequency (six)

in layer 9. Four, possibly six, are made on blades, the rest on flakes. The four

made on blades have utilization flake scars on the dorsal surface of the longer

lateral and on the ventral surface of the shorter, but the utilization on the other

pieces cannot be so clearly determined, beyond the fact that it is always on the

longer lateral. One piece (Fig. 23C) has a notch as well as utilization, and on

three others the utilization is on both surfaces of the edge and appears to be

more consistent with the use of the artefacts for cutting rather than scraping,

although this is not altogether certain.

One trapezoid (Fig. 23E) is made of a fine-grained quartzite, and the rest

are all silcrete.

Obliquely backed pieces, abbreviated to ‘obliques’, are flakes (twenty-one)

or blades (six) that have one end truncated by backing at an angle oblique to

the length of the piece (Fig. 23N—P). Apart from one in layer 19, these artefacts

came from layer 11 and above.

All the pieces show utilization of the edge, 14 on the ventral surface, 8 on

the dorsal, and 3 on the dorsal surface of one edge and the ventral surface of

the other. On the remaining 2 pieces the utilization is restricted to the upper

part of the edge, adjacent to the truncation (Fig. 23P). This utilization is of the

type found on backed points (see below).

All the artefacts were made of silcrete except for one of crystal quartz and

one of chalcedony.

Backed flakes are a morphologically heterogeneous group, of which the

common factor is that all the pieces are wholly or partly backed along one

edge. As indicated in Figure 13, backed flakes intergrade with scrapers, backed

scrapers (to a lesser extent), segments, backed points (borers) and inevitably,

‘miscellaneous retouch’. In the same way that the class of ‘miscellaneous

retouch’ consists of pieces that cannot readily be included in any of the more

BYNESKRANSKOP | 63

(

Fig. 23. A-M. Trapezoids. N-P. Obliquely backed pieces. All silcrete except E, which is

quartzite. A-C. Layer 5. D. Layer 6. E-G. Layer 7. I-M. Layer 9. N. Layer 4. O. Layer 6.

P. Layer 8.

formally definable classes, so the sub-class of backed flakes contains those

artefacts that cannot be allocated to other classes, although it seems quite likely

that they do represent a validly distinct group in terms of function if the shape

of the more formally definable classes is related to function and not merely to

style. However, the fact that the artefact classes do intergrade provides the

principal reason for the avoidance in this report of the term ‘formal tools’.

64 ANNALS OF THE SOUTH AFRICAN MUSEUM

As mentioned previously in the discussion of backed scrapers, Sampson

(1974, fig. 106) has suggested that backed flakes, by the time their utility has

been exhausted, may take the form of backed scrapers. In the BNK 1 sample

backed scrapers outnumber backed flakes by about 2:1 and there is a lack of

intermediate forms—heavily utilized backed flakes—that would be expected if

Sampson’s suggestion were correct, since it seems unlikely that only moderately

utilized pieces and worn-out discards would be found. Also, very few of the

backed flakes in the BNK 1 sample have a backed lateral that is convex, a

prerequisite for backed scrapers.

Backed flakes account for 14,9 per cent of the backed artefact class and 4,2

per cent of the retouched artefact category. Apart from two in layer 18, their

distribution is restricted to layer 12 and above, with maximum frequencies in

layers 9-5, although these are never high (Fig. 14, Table 7). Silcrete is again the

most common raw material (88,5 % of the site total), with a low frequency of

quartz (10,8 %) and only one piece in quartzite.

The backing is variable, in some cases covering the whole of one lateral, in

others less than half. Apart from seventeen pieces, all probably broken, the

backing is opposed to a sharp edge that in all but four cases shows signs of

utilization. Among the utilized pieces are four that would have been classed as

backed points (borers) but for the fact that the utilization covers all, or almost

all, of the edge, a characteristic not consistent with the functional implication of

the name ‘borer’ and not to be found on the artefacts included in the latter

sub-class. In some cases the retouch is not at right angles to the plane of the

flake, and this caused some doubt as to whether these pieces should not be

classed as scrapers. They were, however, less like the bulk of the scrapers in

the sample, variable though these are, and were therefore included with the

backed flakes.

Backed blades, of which there were only seventeen, form a negligible part

of the backed artefact class (1,8 % of the site total) and the retouched artefact

category (0,5%). Apart from two in layer 19, their distribution was restricted

to layer 10 and above. Fifteen of the blades are silcrete and two, from layers 4

and 1, are quartz.

All but one of the blades are broken and it has not been possible to classify

them with any confidence in the way that H. J. Deacon (1976, fig. 61) has done

with the blades from Highlands Rock Shelter. However, the width of the blades

suggests that most would fall into the ‘bladelet’ size class (15-25 mm long).

Thirteen blades have utilization of the unretouched lateral. Six have

unifacial utilization, 2 with flake scars on the dorsal surface, 4 on the ventral.

The remaining 7 blades have bifacial utilization of a kind that suggests their

having been used for cutting rather than scraping. Two that could be consid-

ered to fall into Deacon’s class of backed points (H. J. Deacon 1976, fig. 61F

and G), but not into the BNK 1 class of backed points (borers) (see below),

have bifacial utilization, while a third shows utilization on the dorsal surface

only.

BYNESKRANSKOP 1 65

Borers

The artefacts in this class are of two types, here termed points and drills.

Points are included in the class of backed pieces and cover a range of forms

whose common attribute is backing of part or all of one, sometimes both, of the

laterals to form a sharp or tapered point (Fig. 24A—F). Drills, on the other

hand, are pieces that are often cylindrical in shape and have a blunt end (Fig.

24G-I). The retouch used in the making of drills cannot be called ‘backing’ in

the sense that has been applied to the backed artefacts in this report, and drills

have accordingly been excluded from the class of backed pieces.

There seems little doubt that drills do form a typologically distinct class

that might also be valid in terms of the artefacts’ function, whatever that might

have been. Points, on the other hand, like scrapers and backed flakes, are a

morphologically heterogeneous group united by a common type of retouch.

Microscopic examination of the backed points failed in most cases to reveal

clear evidence of utilization, but it appears from some of the artefacts,

particularly those made of crystal quartz, that there is utilization of the tip that

is consistent with the artefact’s presumed function as a boring tool. However, in

many cases the utilization appears to cover more of the unretouched lateral

than is consistent with boring unless holes were being bored that had diameters

equal to the greatest width of the artefact in the part utilized. It is, of course,

possible that these artefacts served more than one purpose, as seems to have

been the case with other artefact types.

Some of the backed points are similar to those classed by H. J. Deacon

(1976, fig. 61) as ‘backed points’ (types F and G) or ‘unsegmented blanks’ (type

K), as well as the unnamed types B and C. In the BNK 1 sample there were

three from layer 1 and eleven from layer 5. They are in the same size range as

those illustrated in Figure 24B-—C but are only 2-3 mm wide at the base.

The distribution of points and drills is similar (Fig. 14, Table 7), although

there are more drills than points in layer 4 and the opposite in layers 2 and 1.

Points account for 3,3 per cent of the retouched artefact category and drills for

3,1 per cent. Both types are numerically most common in layer 5, in which

points are also relatively most common (6,2% of the retouched artefact

category total) while drills are relatively most common in layer 4 (5,8 %).

Silcrete is the most common raw material, accounting for 80,7 per cent of

the total number of points and 91,3 per cent of the drills. Of the 19 quartz

points, 3 are in layer 2 and 9 in layer 1, which is consistent with the increasing

use of this raw material in layer 1 (Fig. 7, Table 3). There is only 1 quartz drill,

also from layer 1. There are 3 chalcedony points and 8 drills.

Notched flakes

These were found in sufficient quantity to warrant their separate classifica-

tion in Table 9. They were found throughout the sequence except layers 16 and

14, but are more common from layer 9 up, where they account for 84,6 per cent

of the site total. Silcrete is again the most common raw material (76,9 % of the

66 ANNALS OF THE SOUTH AFRICAN MUSEUM

raat ass

Fig. 24. A-E. Borers (backed points). F-I. Borers (drills). J-L. ‘Scraper rejuvenation flakes’.

M-P. Miscellaneous retouched pieces. A-B. Silcrete, layer 1. C-D. Crystal quartz, layer 1.

E-J. Silcrete, layer 5. K. Silcrete, layer 6. L. Silcrete, layer 9. M. Quartz, layer 13 (twice

natural size; stippled area indicates mastic traces). N-O. Silcrete, layer 7. P. Silcrete, layer 5.

BYNESKRANSKOP 1 67

TABLE 9

Stone: inventory of types of notched flakes.

PAV OTe S, focg tc Bho Aereota te aleye lb 23 4 5 © F 8B 9 10 WW WD WB Ww 15 1G iy Ws il Aner

Type 1: single, side .......... 2— 1 2 4 6 2 1 5—-—- ~—~ — — Pi Se et 26

De Sine, GGls as ancsucos a 7

3: double, large......... Die) Dee avd eels 6 le ee a SS ee 2 11

4: denticulate .......... 3 2— 3 4— 2— 4— 1— 1—~ ~~ — 1 > — 21

MNO Call ewee tee scars: rts heey reyes: Sie ear eens leee po e eI etl a ae eee eles te 3 65

Fig. 25. Notched flakes. A, C. Quartzite, layer 5. B. Quartzite, layer 17. D-E. Quartzite,

layer 19.

68 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 26. Notched flakes. A, C, G, I, K. Silcrete, layer 9. B, H. Quartzite, layer 9. E. Silcrete,

layer 11. F, J. Silcrete, layer 5. D, M. Silcrete, layer 1. L. Quartzite, layer 1. N. Silcrete,

layer 4. Stippled areas indicate cortex.

BYNESKRANSKOP 1 69

class total), followed by quartzite (16,9 %), the remainder being three of quartz

and one of limestone. As indicated in Table 9, the notched flakes have been

divided into four types, which are described below.

Type 1 (Fig. 26B-D, K) has a single, deep notch on the side. In most cases

the notch is on one side only, but 3 pieces have a notch on both sides. In 5

cases the notch results from a single flake removal but in the rest it is the result

of repeated flaking not unlike that on adzes; 18 are silcrete, 7 quartzite and 2

quartz.

Type 2 (Fig. 26H—J) has a large, wide notch at the end of the flakes rather

than on the side, and in all cases the notch is the result of step-flaking. All the

artefacts are silcrete.

Type 3 (Fig. 25A—-E, Fig 26L—N) has a large, wide notch on either the side

or the end; 6 pieces have two adjacent notches (e.g. Fig. 26M) but 4 smaller

pieces have only one; 6 are silcrete and 4 quartzite.

Type 4 (Fig. 26A, E-G) are denticulates, with two or more adjacent

notches. There was some doubt as to whether the double-notched pieces should

not be assigned to type 3, but the size of the notches is very different, and the

artefacts were sorted on the basis of the size of the notches rather than their

number; 16 of the pieces are silcrete, 2 quartzite and 1 quartz.

The division of the notched flakes into types is not absolute and is not

intended to do more than indicate the variety in the BNK 1 sample. As is the

case in most of the artefact classes, the division between types is not clear-cut

and classification is thus somewhat arbitrary. Frequencies of the four types are

too low for statistical analysis, and there are no evident secular trends, as

indicated in Table 9.

Miscellaneous retouch

As its name implies, this is an informal class of retouched artefacts that

cannot readily be assigned to any of the typologically more cohesive classes

dealt with above. As indicated in Figure 13, there is an intergrading between

the artefacts of this class and utilized flakes, scrapers, backed flakes and borers

(backed points); and it is probably true to say that this class consists in the main

of ‘rejects’ from the other classes.

There are, however, two types that can be classified separately but have

been included here because their frequencies are too low to warrant separate

classification. For purposes of this report these two types have been given

informal names, and these are used reluctantly because of the ease with which

informal terms tend to become formal. The frequencies of the types are given

in Table 10.

‘Planes’ (Fig. 27) are artefacts that have a retouched (or heavily utilized)

edge at the junction of a flat base and a relatively steeply angled upper surface.

In almost all cases the flaking has produced a straight or slightly convex edge

which is denticulated in a number of cases. This is, however, very different

from the well-defined notching of the denticulate notched pieces (Fig. 26).

70 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 27. Miscellaneous retouched stone artefacts: ‘planes’. A. Quartzite, layer 3. B. Lime-

stone, layer 9. C. Quartzite, layer 4. D. Quartzite, layer 18. E. Quartzite, layer 9. F. Quartz-

ite, layer 12. G. Silcrete, layer 10. H. Silcrete, layer 7. I. Quartzite, layer 15.

BYNESKRANSKOP 1 Wk

TABLE 10

Stone: inventory of sub-classes of miscellaneous retouched artefacts.

WAV CT. csr tine ter eA iin oa x Lt 2°33 4 5 © F 8 O WO tt WD WwW i i Io Wy WS I Woy

IAN ES) Sic aathomt sie otis. o destoyettet ee ed a Mel 2s 2 ae Se Sa Se See et 17

‘Cutters/scrapers’ ............ =— 2— 1-1 2 doe yHeHeeid S| ~— SS > — 3 10

(QUO Eis ene rid eee Oe eataieln Meera crore Ce eee oS 8 ly ot) hee ee tS le 62

Tio Galles Bea tyewssen tenants evcsa lacs Sis Sages ae] oe Cae ele TE SO LO 8s. BG Sh Sire yy Lien eengh li Die A 89

These artefacts are unlike th: ‘extra large’ scrapers (Fig. 16), especially in

having a flat base that suggests their having been held at 45° from the horizontal

plane, or more.

‘Planes’ were recovered from thirteen of the layers, but because fre-

quencies are so low their absence from the other layers, as well as their

relatively increased frequency in layers 12, 10 and 9 cannot be seen as

significant; 10 of these artefacts are quartzite, 6 silcrete and 1 limestone. The

silcrete pieces are from layers 12, 10, 7, and 1 and the limestone artefact is also

from layer 10.

‘Cutters/scrapers’ (Fig. 28) are in some ways similar to ‘planes’ except that

they are generally fairly flat pieces with a straight, retouched edge that suggests

the use of the artefact for cutting and/or scraping. They do not fit into the

scraper class (Figs 15-16), and only two or three could reasonably be called

‘knives’ (e.g. Fig. 28B). Seven are silcrete, three limestone and, as indicated in

Table 10, they are sporadically distributed through the sequence.

Other miscellaneous retouched pieces, which resist classification, make up

almost 70 per cent of the class total (Table 10). Some of them are probably

broken pieces and there are four that may be broken scrapers, although they

look as if the retouched edge of the scraper flake had been removed delib-

erately by striking the flake parallel to its worked edge (Fig. 24J—L). In Figure

24 the three illustrated are informally termed ‘scraper rejuvenation flakes’.

All but one of the miscellaneous retouched pieces are silcrete, and it is the

single quartz piece that is perhaps most worthy of mention. Illustrated in Figure

24M, it is a small (13 X 5 mm) ellipsoid that in section forms a quadrant. The

artefact is broken at one end and was probably originally 15 mm long. It may

also be a piece broken from a larger artefact such as a (convex) scraper or a

backed scraper, but whether the radii of the quadrant represent parts of a

break cannot be determined. Half the artefact is covered by mastic, but the

indications are that the whole of the surface forming the arc may be finely

flaked, rather in the manner of a ‘Still Bay’ or Solutrean point.

This artefact was recovered from layer 13 and a similar but larger artefact

in silcrete was recovered from layer 7 (Fig. 240). This piece has a flat ventral

surface and what appears to be a break at the bulbar end is, in fact, retouched.

Neither of these artefacts can readily be included in the scraper or backed

scraper classes.

Two other examples of miscellaneous retouched pieces are illustrated in

Figure 24N and P to indicate the varying nature of the artefacts in this class.

iz ANNALS OF THE SOUTH AFRICAN MUSEUM

Ne wl)

Fig. 28. Miscellaneous retouched stone artefacts: ‘cutters/scrapers’. A. Limestone, layer 10.

B. Quartzite, layer 7. C. Limestone, layer 2. D. Limestone, layer 8. E. Quartzite, layer 19.

F. Quartzite, layer 1.

BYNESKRANSKOP 1 73

Middle Stone Age artefacts

These were recovered from layers 1, 5—6, 9, 12, and 19, layers 1 and 5 each

yielding two pieces, the rest one. These are clearly objets trouvés and do not

call for special comment as it is not uncommon for M.S.A. artefacts to be

found in L.S.A. deposits, and there is no indication that the artefacts from the

BNK 1 deposits have been reworked, as is sometimes the case. It might be of

interest to compare these with the artefacts from the M.S.A. deposits at Die

Kelders (Schweitzer 1979: 106), which might have been their source.

DISCUSSION OF THE STONE ARTEFACTS

Perhaps the most obvious fact about the BNK 1 stone artefact sample is

that there is a marked difference between the contents of the lower layers and

the upper. However, as the various inventories indicate, this difference is more

quantitative than qualitative and there is no clear division, with changes taking

place throughout the sequence.

All the classes of unmodified as well as utilized or modified artefacts are

present in both groups of layers, and only one sub-class of the retouched

artefact category, backed trapezoids, is absent from the lower layers. This

seems to suggest a lower level of activity in the cave prior to about 6 000 B.P.,

possibly related to a lower intensity of occupation, rather than a subsequent

increase in the range of activities carried out in the cave.

Further study of the artefacts and inventories suggests that it may be

possible to isolate various stages or phases in the lithic technology of the cave

occupants. Layer 19 is characterized by a relatively high frequency of artefacts

(AD = 17 169), mostly unmodified, and with the highest frequency of all the

layers of unmodified blades and blade cores. This is an aspect characteristic of

the ‘Robberg industry’ as defined by J. Deacon (1978: 100), as is the high

frequency of quartz relative to other raw materials. Other characteristics of the

‘Robberg’ are said by Deacon to be the generally higher proportion of small

rather than large scrapers, and the presence of ‘backed tools’. In the BNK 1

sample, however, the few scrapers tend to be medium rather than small (Fig.

17), and backed artefacts are present in all the layers except 15-13 (Table 7).

Layer 19 is chronologically comparable with the uppermost unit of the ‘Rob-

berg’ at Nelson Bay Cave (J. Deacon 1978, figs 2-3) but younger than those at

Melkhoutboom (H. J. Deacon 1976, table 3).

From layer 19 to layer 15 there is a progressive decrease in the frequency

of stone artefacts in the successive layers, and though this increases again in

layers 14 and 13, it is only in layer 12 that the frequency again becomes

substantial and, relatively, exceeds that of layer 19 (Table 2). Layers 18-13

share with layer 19 low frequencies of utilized and modified as well as

retouched artefacts, although the frequency of utilized artefacts in layer 18 is

the highest until layer 10, and layer 13 has a relative (percentage) frequency of

retouched artefacts at least three times greater than those of any of the

underlying layers (Table 2). That these are mostly scrapers is not unusual:

74 ANNALS OF THE SOUTH AFRICAN MUSEUM

scrapers are the most common retouched artefacts in thirteen of the nineteen

layers (Table 7).

J. Deacon (1978: 100) has characterized the lithic aspects of the ‘Albany

industry’ at Nelson Bay Cave as being ‘the almost exclusive use of quartzite as a

raw material, the absence of Formal Tools other than scrapers and miscel-

laneous retouched pieces, the predominance of large scrapers over medium, the

absence of backed tools ...’. The ‘Albany’ layers at Nelson Bay Cave are

approximately contemporary with BNK 1 layers 18-13, there being a hiatus in

the Nelson Bay Cave sequence from about 8 000-6 000 B.p. There may also be

a hiatus in the BNK 1 sequence, somewhere between layers 14 and 12, as

indicated by the difference of about 2 000 years between the dates for these

layers (Table 1). If there is, the stone artefacts suggest that it occurred between

layers 14 and 13 rather than between layers 13 and 12. Although quartzite does

become increasingly common in the BNK 1 deposits and is the most common

raw material in layers 17-15, its use at this site cannot be said to have been

‘almost exclusive’ as it ranks second to quartz in almost half of the layers under

consideration. From layer 14 upwards silcrete can be seen to play an increas-

ingly important role (Fig. 7), and is generally the most common raw material

for those artefacts considered to have been used (Table 3).

Retouched artefacts other than scrapers and miscellaneous retouched

pieces are rare in layers 18-13, but they are present, except in layer 15 if it is

preferred to consider notched flakes as utilized rather than retouched artefacts

(cf. J. Deacon 1978, table 1), and backed pieces are absent only from layers

15-13. The paucity of scrapers in layers 18-13 makes it difficult to comment on

aspects of size beyond pointing to the evidence of Figure 17, which shows that

scrapers tend to become smaller from layer 18 to layer 16, then gradually

increase in size to layer 13; but in all cases except layer 17 the mean length and

width are indicative of medium to small scrapers, as defined by J. Deacon

(1978: 92), rather than large ones.

Because of the hiatus in the Nelson Bay Cave sequence, it appears not to

contain the sequence represented at BNK 1 by layers 12-10 and possibly 9,

which seems transitional between the industries of the lower and upper layers.

Layers 12-9 are characterized by the predominance of quartz as a raw material,

except in layer 9, and by the increase in the frequency of silcrete, which

becomes the dominant raw material in layer 9 (Fig. 7). In layer 12 there is a

marked quantitative increase in the number of artefacts, relative to layers

18-13, and layer 11 has the highest AD of all the layers (Table 2). From layer

12, too, the frequency of utilized and modified artefacts increases, as does that

of retouched artefacts, except in layer 11, which is more like layer 13 than the

other layers with which it is grouped stratigraphically. Layers 12-10 contain 9 of

the 12 classes of retouched artefacts and layer 9 all 12, compared with 2-5

classes in layers 19-13 (Table 7). It must be repeated, however, that the only

new additions to the range are backed points (borers) from layer 12 up, backed

scrapers from layer 10 up, and backed trapezoids from layer 9 up. There is a

BYNESKRANSKOP 1 1)

marked phasic trend in scraper morphology in layers 12-9 (Fig. 17), but

whether this trend really begins in layer 14 or below cannot be determined on

account of the small sample sizes from layer 13 down.

From layer 8 up the declining trend in the relative frequency of quartz

persists until layer 6, after which frequencies generally increase until the end of

the sequence in layer 1, where it again becomes the predominant raw material.

This gain is chiefly at the expense of silcrete which, however, still remains the

dominant raw material in the utilized and retouched artefact categories.

Scraper morphology remains largely unchanged in these layers. Scrapers

are mostly in the medium to small size ranges, with more in the medium range

than in the small, and shape tends more to the equilateral than to the longer

shape of the underlying layers (Fig. 17). The frequency of quartz scrapers

increases in layers 6-3 and remains more or less constant in layers 2 and 1, but

silcrete remains the most common raw material for this class (Fig. 14). The

frequency of scrapers declines from layer 8 up, mostly relative to that of adzes,

which show a continually increasing trend until layer 3, and even though

frequencies decline in layers 2 and 1, adzes are more common than scrapers

from layer 3 up. This suggests that the ‘Sandy Bay industry’ (Rudner & Rudner

1954), with its predominance of adzes, may represent a temporal phase of the

‘Wilton’ rather than a regional or environmental variant (cf. Sampson

1974: 414; Mazel & Parkington 1978). Comprehensive environmental data for

the BNK area for the past 3 500 years are lacking, however, and an environ-

mental interpretation is thus difficult to support.

It seems that while there may be a correlation between the frequency of

adzes and the presence of a wooded environment, as suggested by Mazel and

Parkington (1978: 382—cf. also Parkington 1980a and comments thereon), the

overall picture is not yet clear. Whether the correlation between the relative

frequencies of scrapers and adzes in different environments is more than merely

statistical also remains to be established on the basis of independent palaeo-

environmental data. At BNK 1 it is evident that the use of adzes is a feature

that antedates the onset of the ‘Wilton’ but reaches its peak in the latter part of

this ‘industry’. Since the location of the site has not changed through time, if

the frequency of adzes is correctly related to the availability of wood, then the

increase through time may suggest a changing environment. On the other hand,

it may simply reflect an increasing need for wooden artefacts in an environment

in which suitable wood had always been available.

In the layers from 8 up the incidence of backed scrapers can also be seen to

be episodic, with about 87 per cent of these artefacts in layers 6-4. Segments

are most common in layers 9-5, numerically as well as relatively: the AD for

layer 7 is 51, as against 16 for the numerically similar layer 1. The segments are

most commonly made of silcrete, except in layer 1, where quartz segments are

more common (Fig. 14). Segment lengths are largely constant in layers 9-6,

after which they become shorter, while widths, which increase in layers 9-6,

also decrease from layer 5 but increase again in layer 1. Segment shapes tend

76 ANNALS OF THE SOUTH AFRICAN MUSEUM

towards a wider shape in layers 9-6, a longer shape in layers 5 and 4, but by

layer 1 are again as wide as they were in layer 6 (Fig. 22).

Sampson (1974: 365-366), referring to the aptitude of the ‘Howiesonspoort

groups’ in making a range of backed tools that included trapeze forms and

obliquely backed pieces, comments that ‘all knowledge of these techniques had

completely disappeared by about 16 000 B.c. and cannot be detected in this

area again until their reintroduction (in diminutive forms) at about 6 000 B.c.

At this time the Oakhurst tool types are seen to dwindle via the early phase of

the Wilton as microlithic types begin to increase. The origins of the coastal

Wilton are not to be found in the Howiesonspoort sites’. Although the evidence

of the BNK 1 lithic assemblage shows that backed artefacts are absent only in

layers 15-13, say between about 10 000-8 000 B.p., it tends to support Samp-

son’s statement about their reintroduction or, more correctly, increased used

after about 6 000 B.p. (rather than B.c.).

It is difficult to know whether or where to separate layers 8-1 into different

‘industrial’ or small ‘phase’ groups since there is so much overlapping of the

trends in the various components of the lithic assemblage. It seems, too, that

such divisions should not be based on the data from one site alone but should

be held in abeyance until a more comprehensive synthesis of information, on at

least a regional basis, can be obtained. For the present, therefore, it seems

advisable to keep layers 8-1 in the general category of the ‘Wilton industrial

complex’ (cf. J. Deacon 1980: 91) although it is evident that there are differ-

ences within this group as great as those that distinguish it from the underlying

‘Albany—Oakhurst-pre-Wilton’ deposits. The data from BNK 1 suggest that

further work is needed before satisfactory divisions within, as well as between,

the ‘Wilton Complex’ and the ‘Oakhurst Complex’ can be made.

BONE ARTEFACTS

A total of 155 modified or utilized bone artefacts was recovered from the

deposits and an inventory of these is given in Table 11. From this it will be seen

that almost 91 per cent of the site total came from layers 9-1 and that seven of

the nine classes are restricted to these layers.

TABLE 11

Bone: inventory of modified and utilized artefacts.

WAY Clicrars. tern coe Sisrsieras aeisieeiens 2s 4 > © Foo Deno i We iy es sy WO iy We 1D “ial

Class

AWIS 63 since wicys <div n ore ero 4h il SI Te ee i 1 i = 2 59

POI tS) a oye tee es hoa 2 She) Ab) pa ee ee ee ye ee ee B

IeinkshattSieyt ees oem cere — — — 3 5 SaaS ia si si SS SS SS SS 8

Spatulacrereere scenes rr 2—- — — ~— — jt) a EE ae ee eee 3

Bishi Gorges! a jose Se ase —- — — — — 2 4 2 ~ ~ ~—~ — 8

Ocnamenise wiser: Jb 2S OS SS SS f—_ = = = => = 12

MUbES:, 82h gee eect thn ere, a an een ee ee a ee ee ee 11

Broken:

HPS) es 5 abe cet: She ocean 2— — 1 4 3—- —- ~—~ ~— — — 10

OUhE? oss tee sarees e JS—=— 260 2S] —]— =] = 15

Wtilized? WRG. rater sn ee es 20 — 2 — 2 = 2 = = 26

bayer totale et ae sp | 2 Wf iby Sil 5:52. Do Wiese SB 4 2 a 2 els

BYNESKRANSKOP 1 JET

Awls (Fig. 29)

These are the largest single class, accounting for 38,3 per cent of the site

total, and have the widest distribution through the layers. Layers 3 and 5 each

have an example of the slender artefacts termed ‘needle awls’ (Fig. 29A-B).

Seven of the awls, from layers 2, and 4-6, are made of bird bone, the rest being

of mammal bone.

Points

Artefacts in this class, considered to be projectile heads (e.g. Schweitzer

1979: 129-30), came from layers 1 and 5. The three artefacts are all broken and

are not illustrated (cf. Schweitzer 1979, fig. 10). Parts of broken points are

probably included among the unclassifiable broken artefacts (see below), so

that undue importance should perhaps not be attached to their apparently

restricted distribution.

Linkshafts (Fig. 30A—-H).

These pointed artefacts are also considered to be parts of composite

projectiles. In the BNK 1 sample they are represented by eight whole or almost

whole pieces from layers 4 and 5. The artefact illustrated in Figure 30H is

unusual in having a V-shaped notch at the lower end. This raises some doubt as

to its correct identification as a linkshaft as it may be the nock end of a

composite arrow. As far as is known, however, such parts made of bone have

not been recorded from other archaeological deposits. E. M. Shaw (1980 pers.

comm.) has pointed out that this piece, as well as one other (Fig. 30G), have

split tips, and has suggested that these might have been slots to take the

arrowhead. This would have been a type different from the points described

above, and the slots would apparently not have been more than 1-2 mm wide

and could thus only have accommodated a thin sliver of bone, stone or wood.

None of the linkshafts is decorated, as was the case with some from Die

Kelders (Schweitzer 1979, fig. 11).

Spatulae (Fig. 30I-K)

These were recovered only from layers 1 and 7. Those from layer 1 are

made from mammal ribs while that from layer 7 is made from hippo ivory,

probably a maxillary incisor or canine (R. G. Klein 1980 pers. comm.). The

cancellous bone of the complete spatula is worn smooth for about half the

length of the artefact and, opposed to this, the cortex bears a number of

indentations, visible in Figure 301. This wear might have resulted from the use

of these tools, inter alia, for prising Haliotis midae from the rocks. This

suggestion has been dismissed by various other workers on the grounds that the

bone would be too brittle to withstand the strain of the upward lifting

movement necessary for detaching the shellfish. Avery & Siegfried (1980: 33)

have, however, also suggested a similar use, and this hypothesis could be tested

by replication, using ‘green’ bone, since this is as likely to have been available

ANNALS OF THE SOUTH AFRICAN MUSEUM

he.

santiiiliilliy A I =,

isi eee.

ee POPE

OOD

ee LC

Bone awls. A, D, K, P, R-S. Layer 5. B—C. Layer

Fig. 29

aE:

J, N-O. Layer 1. I,

Me cavern Oo; layer:

Q. Layer 4. F. Layer 19. G-H,

ie,

BYNESKRANSKOP 1

MOM

Fig. 30. Bone artefacts. A-H. Linkshafts. I-K. Spatulae. A, C-D,

F-H. Layer 5. B, E. Layer 4. I, K. Layer 1. J. Layer 7.

80 ANNALS OF THE SOUTH AFRICAN MUSEUM

as dry bone. Similar wear and indentations are evident on some of the spatulae

from Die Kelders.

‘Fish gorges’ (Fig. 31)

These artefacts, of which eight were found, in layers 13-15, are small, flat

pieces of bone tapered to sharp points at both ends. What might be the

fragment of a ninth was recovered from layer 18 but has been included with the

unclassifiable broken artefacts since the fragment represents not more than a

third of what would have been its full length if it were a ‘fish gorge’. It is worth

noting at this point, in view of the assumed function of these artefacts, that only

four fish are recorded from layers 13-15 and none for layer 18. This leaves open

the option of deciding whether these were or were not fishing aids, whether

they were largely ineffectual, or whether the bulk of the fish caught with them

were eaten away from the site.

(C D E F

Fig. 31. Bone ‘fish gorges’. A, C-D. Layer 14. B. Layer 15. E-F. Layer 13.

Largest bone = 2,7 cm.

Ornaments

Bone artefacts considered to be ornaments were recovered from five of the

layers. They are mostly broken, and most are illustrated in Figure 32G—K and

M-Q. Those from layer 1 consist of half a small bead (Fig. 321), part of a ring

(Fig. 32K) and two fragments with drilled apertures (Fig. 32N—O), of which

BYNESKRANSKOP 1 81

dad Op

Fig. 32. Bone artefacts. A. Articular end of grooved and snapped bird

bone. C-E. Bird bone tubes (D and E broken). G—Q. Ornaments. A,

C. Layer 5. B, D, F. Layer 4.-E, I, K-L, N-O. Layer 1. G. Layer 7. H,

MeiEayer 6-35 Layer, 13. Layers. O} Layer 2:

only the second appears to have reached any degree of completion, having a

finely ground edge. The fifth piece from this layer, not illustrated, is doubtfully

ascribed to this class: it is a small bone tube, 3 mm long with an outside

diameter of 5 mm and a wall thickness of about 1 mm.

From layer 2 came a partly perforated fragment which has been broken

into a roughly circular shape (Fig. 32Q), and layer 3 yielded a better-finished

version of the layer 2 artefact, except that this is squared (Fig. 32P). This layer

also yielded a small piece of bone, 8 X 4 mm, that has incised grooves on one

side and five partial perforations on the other (not illustrated). Layer 6 yielded

a broken fragment of limb bone that shows cut marks along its edges and has

been pared down from its original dimensions to a wall thickness of 1-1,5 mm

(Fig. 32H), and a roughly circular disc, finely ground down to 1 mm thickness

(Fig. 32M). From layer 7 came the tooth of a Great White shark (Carcharodon

carcharias) with an ‘hour-glass’ perforation drilled from both sides (Fig. 32G);

and from layer 13 a bead similar to that from layer 1 but larger (Fig. 32J).

Tubes

Pieces of bird bone cut into tubes were found in layers 1, 4 and 5 and are

illustrated in Figure 32A-F. Two of the pieces, from layers 1 and 5, are

82 ANNALS OF THE SOUTH AFRICAN MUSEUM

articular ends that are probably discards after the tube had been detached from

the shaft by ring-grooving and snapping (Fig. 32A). The pieces illustrated in

Figure 32B and E, from layers 4 and 1, have linear decoration across the tube

while the others are plain. These artefacts might have been ornaments, or they

might have been links used in composite arrows. Their restriction to layer 5 and

above is consistent with the distribution of points and linkshafts in the deposit.

Unclassifiable broken artefacts

Twenty-five broken pieces have been listed in Table 11 under the headings

of tips and ‘other’, i.e. end or body parts. They are small pieces, usually not

more than 10 mm long, of which not enough is present to identify securely the

artefact type from which they came. The tips may be those of awls or points.

Two, both from layer 5, are unusual in having the first millimetre or so thinner

than the rest of the piece, and appear to have been used as drills. This feature

cannot be observed on any of the awls or points.

The other broken pieces are probably mostly fragments of points and

linkshafts although some of them that are more oval than round may be parts

of awls. The fact that both types are restricted to the same set of layers 1, and

4-6, may be coincidental, but it may also indicate that points and foreshafts

were in use before layer 5 was deposited.

Utilized pieces

These pieces, mostly from layer 1, are a miscellany that show marks of

damage or utilization of a type that suggests their informal use. A range of

these is illustrated in Figure 32L and Figure 33. Two horn-cores from layer 1, of

which one is illustrated (Fig. 33D), have marks from chopping or cutting, as

does the small polished piece shown in Figure 32L. Figure 33H-I illustrates two

bovid scapulae, probably sheep (R. G. Klein 1980 pers. comm.), from layer 1

and Figure 33J a comparative specimen, from a young adult sheep, to show the

extent of the wear on the archaeological specimens, of which the edges

opposite the posterior border are considerably reduced and worn smooth,

suggesting that they were used for some sort of scraping activity. The other

pieces shown in Figure 33 show varying degrees of wear or grinding and in

some cases (e.g. A and E) striations.

MARINE SHELL ARTEFACTS

Artefacts made of marine shell were recovered from layer 16 up and fall

into two major categories, ornaments and edge-damaged pieces. The inventory

of these is given in Table 12.

Ornaments

These have been divided into two classes, whole shells that are perforated

but have no other modification, and a range of modified shell fragments with or

without perforations, here informally termed ‘pendants’.

BYNESKRANSKOP 1

Fig. 33. Utilized or damaged bone, all from layer 1, with a modern sheep scapula

(J) to show extent of wear on H and I.

84 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 12

Marine shell: inventory of ornaments and edge-damaged artefacts.

Teayerie.. wetos sooo cathe eee 1 23 4 5e 6) 7S SO O12 1841S el Gre Motal

Perforated whole shells

Glycimeris queketti............... 126 —— 127

INCRE TT RSALT 550060088000500¢ 40 — 5 1—-— 71 11> —> — — 2 — — 50

WricoliaiKOchix Nees se. okey yal t = SS = eS ee 29

CONnUSISP(D) secre cea ose LoS ie Si SSS i Se D

Cypracasp(p) aes cee fe eee So = ea ee ee Se = ee ee 6

Bulligisppe xs v2cnti ce sone 1 Le ee Se ee 21

Donax serra

Whole- a8. eee a eee = =] =| =]—] 2 3 2S SS — Sa 8

brokenie. sees. 555 see — = = = 29 68 33 25 27 32> — 3:-=— 2) tf 19

Other cajensa stewee nee emo. 2 -2Y 23, 25h 1 03 SSS = SSS = 14

DOtall oP seen tea teerdte on hel SORT 191 «2. (3-189 54. 73) 36932" 32, Nae 1 Se oe eda

Pendants

perforated

Plain) 3 nets ese 4 2 3 4 6 4 4— 1—> ~ ~ ~ ~ ~ — 28

decorated sree caps ocbaces eee ers 122 2 62—. 11 >—-—>— => — — — — 17

unperforated

jo) EVID Were cree er achur BESERPRCNES cious 5-55 4—- 2— 4 1 — 2 — a 13

decorated). 52sec eee —- — — — 2 — ijl (Se 8

Totalwee has eee oe > ne eeee OF 4S TOS 8rd a 4 3 Se ee 66

Edge-damaged Donax serra

WHOLE) 3.2). NP cictse carat are sa eee eee as eS ee ee SSS = 5

frasMents 2 2. iuaeace eeugaties sock 97 — — 3 1 1 3-— 6 4 3. 6— —.—-— 124

* Frequencies estimated from the mass of broken fragments showing signs of perforation.

Perforated whole shells (Fig. 34) are the simplest form of shell ornament,

requiring only the making of a hole to allow for stringing or attachment to the

person in some other way. The most common of these are the valves of Donax

serra (not illustrated) of which most in the BNK 1 sample are broken. With

very few exceptions, the perforation was made by punching out a piece of the

shell wall, almost invariably from the inside of the shell and regardless of

whether it was a gastropod or bivalve. This may well account for the high

number of broken Donax serra valves, since the holes punched in these shells

are usually large, up to 20 mm diameter.

Five Glycimeris queketti shells have apertures made by grinding away the

umbo (Fig. 34G-H) while the third shell illustrated (Fig. 341) has a punched

hole. Two Conus sp. shells (Fig. 34L—M) have V-shaped notches filed into the

outer wall opposite the aperture of the shell. At Die Kelders, this method of

making perforations was restricted to Conus shells (Schweitzer 1979: 144), and

the presence of a similarly perforated shell in BNK 1 layer 12 attests to the

continuity of the tradition over a period of at least 6 000 years.

A single, broken Perna perna valve from layer 8, not illustrated and listed

in Table 12 under ‘other’, is the only instance of a shell ornament not a

‘pendant’ that has a drilled perforation. This was drilled near the umbo and

from the inside only and has the inward-sloping sides characteristic of drilled

apertures.

Four of the Cypraea shells are broken, with the section of wall opposite the

aperture missing almost to the columella. It has not been possible to determine

whether this damage is natural or not: similarly damaged shells are to be found

BYNESKRANSKOP 1 85

Fig. 34. Marine shell ornaments: perforated whole shells. A. Layer 8. B—C, E-I, L. Layer 1.

D, J-K. Layer 4. M. Layer 12.

on the sea shore. It is possible, however, that these shells may have been

ground down, as seems to have been the case with the Nassa kraussiana shells

(cf. Fig. 34F). One of the Cypraea shells, from layer 1, is filled with some kind

of ochre-stained deposit, possibly organic. This suggests that it might have been

used as an ornament, since similarly stained ostrich egg-shell beads are often

found.

Layers 9-1 yielded 97,3 per cent of the total of perforated whole shells, of

which 42,6 per cent are in layer 1 alone. After the estimated total of Donax

serra valves (42,6% of the category total), the most common species are

Glycimeris queketti (28,3 %) and Nassa kraussiana (11,2 %).

‘Pendants’ made of worked shell fragments are entirely restricted to layers

9-1, and are most common in layers 5 and 6. In Table 12 they are divided into

two classes, perforated and unperforated, each with two sub-classes, plain and

decorated. Examples of all the types are illustrated in Figures 35 and 36. A

more comprehensive analysis, based on attributes such as shape, size, number

of perforations, and type of edge decoration was found not to be useful since

the degree of individualism is so great that the analysis would ultimately have

had almost as many divisions as there are artefacts, and it must suffice to say

that the greatest variety occurs in layer 5.

86 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 35. Marine shell ornaments: ‘pendants’ with plain edges. A, D-E, J—L. Layer 5. B,

G-H. Layer 6. C, I. Layer 4. F. Layer 7. M, O. Layer 1. N. Layer 2. P. Layer 8.

All but three of the pendants are made of Turbo sarmaticus. They consist

mostly of the nacreous inner shell, the periostracum having fallen away or been

removed, and in every case the perforations have been drilled from the inner

surface of the shell. The three not made of Turbo sarmaticus are two of Haliotis

midae from layers 1 and 4 (Fig. 36J—K), and one from layer 6 made from the

apex of a Patella species (Fig. 36V).

Whether or not the unperforated pendants are merely unfinished pieces is

hard to determine. A particular type with denticulate edges (Fig. 36N—Q), of

which there were 2 in layer 4, 4 in layer 5 and 1 in layer 9, has no perforations,

nor are there any like it in the sample that do. This suggests that they were

possibly not intended to be perforated. Kolb (1738: 197) observed that the

BYNESKRANSKOP 1 87

Fig. 36. Marine shell ornaments: ‘pendants’ with decorated and plain edges; perforated and

unperforated. A, C-D, F-G, N, P, S, W. Layer 5. B, J. Layer 4. E, O, Q, V. Layer 6. H,

Ky Layer 2 Re Layer Le) ayer 92M, i—U\ Layer 3:

88 ANNALS OF THE SOUTH AFRICAN MUSEUM

Hottentots were very fond of ornaments for the head, which they fastened to

their hair. These denticulate discs could perhaps have been tucked into the hair

but would not have been very secure. With reference to ‘pendants’ in general,

Kolb (1738: 198) mentions that the Hottentots pierced their ears with what, in

archaeological parlance, would be termed bird-bone awls and then passed brass

wire rings through the perforations. ‘To these Rings the Wealthy and Eminent

hang Bits of Mother of Pearl, to which they have the Art of giving a very

curious Shape and Polishing.’ It seems unlikely, though not impossible since we

know so little of the social systems of the Khoi(san) in the early historic period,

that such ornaments would have been restricted to those of high status, and

their presence in the BNK 1 deposits as early as layer 9 certainly antedates the

arrival of the pastoralist Khoi in the southern Cape by several thousand years.

Although the evidence of the BNK 1 deposits indicates the manufacture of

marine shell ornaments for about the last 10 000 years, it is not until layer 9,

about 6 000 years ago, that they become common and increase in sophistica-

tion. The possibility that the more fragile pendants might not have survived

cannot be discounted, but the fact that these first appear at a time when there is

evidence of change in other artefact traditions as well as a major increase in the

frequency of marine shell representing food debris cannot simply be dismissed

as coincidental. The low frequencies of shell ornaments in the layers below

layer 9 are, however, also consistent with an interpretation based on a lower

intensity of occupation prior to about 6 000 years ago, while the presence of

ornaments from layer 16 up testifies to the exploitation of marine resources

from about 10 000 years ago.

Edge-damaged Donax serra valves, mostly broken, were recovered from all

but three of the layers from layer 12 up. In view of the generally low

frequencies in the individual layers the absence of these artefacts from layers 2,

3, and 8 is probably fortuitous. On the other hand, the frequency of these

artefacts in layer 1 (78% of the site total) is higher than can be explained by

this layer having the greatest excavated volume. In this respect layer 1 re-

sembles the Holocene deposits at Die Kelders, which fall within the time range

of this layer and also contained a high frequency of these artefacts (Schweitzer

1979 stablew0)»

Despite certain initial reservations as to the validity of classifying these

pieces as artefacts, let alone ‘scrapers’, evidence from Die Kelders suggests that

they are artefacts, even if their function can only be presumed. In the tidal

wash below Die Kelders Cave, valves of Tivela compressa are commonly found,

almost invariably with damage to the periostracum of the ventral margin. This

damage takes the form of chipping of the periostracum, almost as if small flakes

had been removed in a regular and deliberate manner from the margin. This

edge damage is, however, less pronounced than, and different from that found

on the Donax serra fragments from the deposits. It is noteworthy that no

Donax serra was recovered from the deposits with edge damage similar to that

on the Tivela compressa valves from the beach, and that Tivela compressa, a

BYNESKRANSKOP 1 89

bivalve almost as big as Donax serra, was not found in the deposit. It is,

therefore, concluded that the edge damage on the Donax serra valves and

fragments recovered from the archaeological deposits is not natural and that

such damaged pieces are thus correctly called artefacts.

That Donax serra ‘scrapers’ appear in the BNK 1 deposits only from layer

12 up is somewhat problematic. If these artefacts were indeed used as scrapers

it is difficult to understand why this innovation should have occurred at a time

when there is a marked increase in the frequency of stone scrapers (Table 7). It

is certainly not explicable in terms of a shortage of suitable stone, and can only

be seen as a technological innovation for which there seems, at present at least,

no reasonable explanation. In layers 4-1 the frequency of stone scrapers

decreases but there is not, except in layer 1, any increase in the frequency of

the shell artefacts.

OSTRICH EGG-SHELL ARTEFACTS

Ostrich egg-shell, in the form of beads, ‘pendants’ and fragments, plain

and decorated, was recovered from every layer of the deposits. Table 13

provides an inventory of the various classes, and it should be noted that,

whereas frequencies for all other classes are numerical, those for plain frag-

ments are the mass, in grammes.

Beads

Beads, whole and broken as well as partly made, are the most common of

the worked material. Frequencies in the lower layers are low, with ADs of less

than 100 for layers 19-10. From layer 9 frequencies increase until layer 6, after

Fig. 37. Fragments of decorated or worked ostrich egg-shell. A, G, K—-M. Layer 5.

B-C. Layer 9. D, F, N-O. Layer 4. E. Layer 16. H. Layer 1. I-J. Layer 6.

ANNALS OF THE SOUTH AFRICAN MUSEUM

Sts 9

89L CE eSL

c0S 0s LCE

68S IEG Nets

1L9 v9 9IP

OLE 87C 981

9¢1 oT SOL

8L 0 St

vec 07 lel

6c1 90 6£

SLI Oy 86

109 b6 el9

O9L eo) Ivy

879 SL 60S

6cV os 09

LIS 86 1v9

009 £8 OVS

809 ve LSI

88e ve ICC

67 [6 Na")

(3) [P}0}

WwW/dV os % (3) sseul

sJUSUIBRIy UTP]

syuepuod

SJUSUIBRAY

punoip

LI LO

I ae OS

— I BE

ae ae £

a G Ol

2 == L

= =e 9]

t i C

= = SC

= — 6£

= = OL

I WG 6SP

G Il C9S I

aa 61 IS9 |

é Si vL9 |

€ Ov S8L

€ 97 16L

1 I OLY

I 9 o9¢

ee 6 90S

Wav

ssutuodo sJUSUIdeIy

SPI, poye1099q

‘speoq opeur Ayjied se [Jam se UdyYOIG SopnypoUl ,d}0[dwIOoUT, “ajo

£0

£00

80°0

600

(all)

10°0

611

CLI

C8!

8 CI

Sail

[8101 ONS %

SO) fe

)

6£

897

906

Eas, |!

90r I

tL6

OCL

6l1

807

OTe I

{B10}

sprog

ILC SVC Co eee nee 4 [R10 L

9z CZ teense 61

C] Of) epueeucsac or 8]

Zz at. | RP 5 decane serie HI

Zz Athy pena dDe rns 9]

I Peace ep boaer en: CI]

¢ ee eae bl

; a ee ee ree ice E]

Zz qi) Poaeren pone nen: ZI

C O) eo nackte II

IZ BE 0 tttttttesse 01

977 ht 6

z69 poZ tenes 9g

616 BIp tte L

zg peg ttt 9

Sop cog tte ¢

LOP b

z/ if ee mepon sone ron: ¢

101 7) Zz

Z6b p7Q 0 ee |

9)9,dur09ul a,oyM ToAe'T

‘sasseyo JO AIOJUDAUT :]JOYS-d3d9 YOLNSC

C] ATAV

BYNESKRANSKOP 1 91

which they decline again, more noticeably in terms of ADs than of actual

numbers.

No attempt was made to determine whether there was any change in bead

size through time. Visual observation suggests that most beads fall within a size

range of 4-8 mm diameter, and in view of the low frequencies in the lower

layers any changes in size might be more statistical than real. Schweitzer

(1979: 150) has pointed out that abut 2 000 beads are required for a multi-

stranded necklace and 8 000 for an apron, and the BNK 1 site total of 3 345

completed beads is thus not impressive. It is perhaps interesting only for the

evidence it provides of the persistence of the tradition and manufacturing

technique over a period of some 12 500 years.

Decorated shell fragments

Apart from three pieces in layers 16 and 18, decorated shell fragments are

restricted to the upper layers, from layer 9 up. At least twenty-five different

types of decoration could be identified, and some of these are illustrated in

Figure 37.

Most of the decoration consists of linear incisions, but a few pieces have

decoration in which the surface of the shell has been pecked or abraded.

Quantification of the different types of decoration is not considered particularly

useful, since the number of fragments is not directly proportional to the

number of whole shells, nor is it likely that each type of decoration represents

an individual shell. A whole shell in the Museum’s collections (AA5863), taken

from a ‘Bushman’ grave at Boegoeberg, in the Gordonia district of the

northern Cape, is richly decorated over most of its surface, and at least ten

different types of decoration can be recognized (cf. J. Rudner 1971, fig. 1).

It is perhaps worth noting that layers 7-5 have the greatest variety of types

of decoration.

‘Flask’ openings

Fragments of shell with parts of ground circular apertures (Fig. 371) are,

apart from one in layer 19, restricted to layer 9 and above. As with the

decorated shell fragments from layers 16 and 18, the recovery of a ‘flask’

opening from layer 19 was cause for a certain amount of concern, if only

because worked artefacts are so rare in the lower layers.

Ground fragments

Pieces with ground edges and in some cases smoothing of the outer surface

of the shell were found in layers 5—1. These generally small fragments may be

parts of ‘pendants’ or discs, but complete examples of the particular type(s)

represented by these fragments were not found.

‘Pendants’

Fragments of four artefacts, informally termed ‘pendants’, were found in

layers 1 and 4-5. The most complete of these, from layer 1, is illustrated in

92 ANNALS OF THE SOUTH AFRICAN MUSEUM

Figure 37H, as well as a fragment with two holes drilled through it (Fig. 37G)

that bears some resemblance to the perforated marine shell ‘buttons’ (Fig.

35G-L).

Plain fragments

Unmodified fragments of shell are not distributed throughout the layers in

proportion to the frequencies of beads. Layers 19-16, 9-7, and 5-3 all have

mass ADs of more than 500 g, while layer 6, which has the highest frequency of

beads, actual as well as AD, has one of the lowest mass ADs of the upper

layers. Layer 19, on the other hand, has the highest frequency of beads in the

lower layers as well as the highest mass of plain fragments of all the layers.

To summarize, about 98 per cent of the site total of beads are from the

upper layers, as are 99 per cent of the decorated fragments, all but one of the

‘flask’ openings, all the ground fragments and the ‘pendants’, and almost 63 per

cent of the total mass of plain fragments. Whether the ground fragments and

‘pendants’ can be regarded as innovations during the last 4 000 years or so is

not clear from the sparse evidence provided by the BNK 1 data, but the

presence of the other classes, even if sporadically, in the lower layers suggests a

continuity in the traditions of artefact manufacture in ostrich egg-shell over a

period of at least 10 000 years. The presence of a ‘flask’ opening in layer 19 and

decorated fragments in layers 18 and 16 may be seen as problematic, since

these artefact types do not otherwise occur in the deposits before layer 9; but it

seems not improbable that people who had the technical ability to make ostrich

egg-shell beads would also have been capable of grinding smooth the openings

of perforated shells and of the simple, linear incisions required to decorate

these shell containers.

The probability cannot be discounted that some of the plain fragments may

represent food waste rather than artefactual material in the strict sense, but it

seems equally likely that even the shell used for the artefacts had its contents

extracted for food before the shell was used. This is probably true of most of

the classes of organic material and merely shows that it was not necessarily only

the edible parts of animals, plants and birds that had economic value.

POTTERY

A total of 426 potsherds was recovered from layer 1. Because of uncer-

tainty regarding the degree of disturbance of the upper sub-units of this layer,

because the front four squares were excavated in a greater number of sub-units

than the squares in the main trench, and because the number of sherds is so

small, no attempt has been made in this analysis to retain the sub-unit divisions

employed in the excavation, which might otherwise have provided information

regarding changes in style and technique. Twenty-six sherds were recovered

from the lowest sub-unit of layer 1 but are presumed to relate to the overlying

sub-unit in view of the date of 3 220+45 years B.p. obtained for the lowest

sub-unit (Table 1).

BYNESKRANSKOP 1 93

TABLE 14

Pottery: inventory of sherds.

no. % of working

total

ONC ANALG CLO EA er re ena rn. Mra a aw, cence Ms sols SuSE un 426

TONING 5 4 0 obs i eregsisia.d oa ewa tea onto etek ieee Maer ees 2a eer et ae aan ae 19

\AVORESITINS (TOG Aa cet gree an eee ee ik 2 ne a meee em eee 407

Rims:

CSS EEN WME TM Cre ie a ee ee om 20 4,9

RDO. oo a BUS BSIGUS OMe RR co Aa at ras Seeman aL 8 Rene eae an lear val 0,5

Decorated

TGA eT Te a AO oe hn Oe a hs AU garg el Pe 7 JEG

MCC KOO eee epee ee ON ee eM: A cute cade as che 2 0,5

| BY GUPRUIGI VEG! «a A Ts: GARR x i ar ok ia ne a 134 32,9

UMS He Cl gpI STOCHTC ees aesetes eevee coe tc cath Mook Showa weiss aD 5,4

WinbmGniShedimlUSOCMiC.,.. a2 cbs oie Sho <p eh Soars Sey it wre a whee 45 11,0

‘SOLE IORTTIVETITIS) vo Saati cS Spr ene antacid Oe ne 37) 14,0

Working total for thickness (working total less split fragments) ...... 350 86,0

Thickness

ANS CREE ICR ge Me RN oR be Leen, worn Utell Po iNel a gid 4 x alae oa cng 2-10 mm

DELI) oi sel 0 ' SeRee CE eta, RINE Sst See ARR! et cot <i a |e ne ee, 5,4 mm

SliGl, CEA SPs, 8 dts Ok & Mee aia a oy oe are Pe +1,3 mm

Table 14 contains an analysis of the relevant sherd data. From this it wiil

be seen that it was possible to join only 19 pairs of sherds, giving a working

total of 407 sherds. Of these, 22 are rim sherds, of which possibly 3 are spout

rim sherds (Fig. 38P—R). The rest are neck or body sherds and include part of a

decorated boss (Fig. 38T). A small fragment of a base was also found.

Burnish was evident on 156 sherds (38,3 % of the working total), of which

22 (5,4% of the working total) also showed evidence of red ochre staining.

Ochre was found on a total of 67 sherds (16,4 % of the working total). Burnish

appears to have been more common on the sherds of the lower sub-units, but

such an observation must be treated with caution, not only because of the

possible disturbance of the deposit but also because the number of sherds is not

a direct indication of the number of vessels represented; and it must also be

borne in mind that the evidence of burnish can be destroyed by use during the

lifetime of the vessel as well as by post-depositional factors. Carbonized matter,

presumably organic (cf. Schweitzer 1979: 165-7), was noted adhering to some

of the sherds.

A morphological analysis of rim sherds was not considered to serve any

useful purpose because of the high degree of fragmentation of the sherds and

because study of complete vessels has shown that there is often quite consider-

able variation in the shape of the rim of one vessel, to the extent that sherds

that could be classified as rounded or squared could form part of the rim of the

same vessel. A study of the rim sherds indicates, however, that at least nine

different vessels are represented, of which two, possibly three, were spouted

and one had a boss or bosses.

94 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fig. 38. Potsherds, all from layer 1.

Replication studies (Wilson & Halkett, 1981) have shown that the decora-

tion on the rims of the sherds illustrated in Figure 38A, R and (minimally) Q

was probably produced by impressing the ventral margin of a Donax serra valve

across the rim, and the same ‘tool’ may have been used to decorate the boss

(Fig. 38U). The decoration on the rims illustrated in Figure 38D—E might have

been produced either by pinching the rim between closed fingers, with the

thumb on the inside, or by pressing a finger, stick or other rounded implement

into the clay. The third type of decoration (Fig. 380, T) is a series of impressed

bands, for which the ‘die’ may have been a piece of reed or grass.

BYNESKRANSKOP 1 95

The sherds range in thickness from 2,0 to 10,0 mm, with a mean thickness

and standard deviation of 5,4 + 1,3 mm.

The bulk of the sherds contain a medium to low density of quartz grit,

mostly in the 0,5-1,0 mm size range but also including larger pieces. Some of

the sherds include shale or siltstone fragments, but these occur in low frequency

in a matrix of quartz grit and the sherds cannot therefore be classed as

‘shale-tempered’, as were some from De Hangen (Wilson 1974: 57; table 1) and

Die Kelders (Schweitzer 1979: 158-160). Crushed quartz chips were also

observed in sherds from the upper sub-units but these are included in a matrix

of fairly large rolled quartz grains up to 3,5 mm diameter. This suggests that

the chips might have resulted from the grinding or pounding of dry clay, or the

fracturing of larger quartz grains during firing. A change from the use of quartz

grit to crushed quartz chips, observed in sherds from the upper sub-units of the

DGL member at Boomplaas, was regarded as a technical innovation (Wilson

1976: 16; H. J. Deacon et al. 1978: 54). Shepard (1954: 27) has pointed out that

rounded grains weaken a clay body more than sharp, irregular ones.

Fifteen sherds contain carbonized organic material mixed into the clay, but

whether this was deliberately introduced is doubtful, since the sherds also

contain a ‘normal’ amount of quartz grit. The number of sherds with organic

inclusions may be higher than fifteen, since these inclusions are difficult to

detect in a dark-coloured clay body.

Eight sherds contain a medium density of fine shell grit about 0,5 mm or

less in maximum dimension. This grit is rolled, and the fact that it is mixed with

an equally fine quartz grit suggests that it is a natural admixture, either in the

clay or in the grit, if this was added by the potter. In either case, a source not

far from the sea is indicated.

METAL ARTEFACTS

A small bead of cuprous metal, copper or brass, was recovered from the

second sub-unit from the top of layer 1, in one of the front four squares, O 31.

Charcoal from the same sub-unit in this square yielded the youngest date for

the site, 255 + 50 years B.p. (Table 1).

The bead is 3,5 mm long and has an external diameter of 9,5 mm, with a

wall thickness of 1,5 mm. It was evidently made by rolling a strip of metal into

a circular form. The bead has not been cleaned to remove encrustation but it

appears that the two ends of the strip are merely joined, not soldered or

welded.

SUMMARY

Stone artefacts make up the bulk of the artefactual content of the deposit.

Quartz is generally the most common raw material, though silcrete shows an

increasing trend in layers 16-5 and is generally the most common raw material

for artefacts in the utilized and retouched categories. Quartzite is more com-

mon in layers 19-13 than in the overlying layers, and limestone, although in low

96 ANNALS OF THE SOUTH AFRICAN MUSEUM

frequencies in most layers, is more common in layers 9-1 than in the underlying

layers. Chalcedony and shale are both minimally represented in the sample.

As is usual, unmodified (‘waste’) pieces comprise the major part of the

stone artefact sample, with frequencies ranging from 93,9 to 99,7 per cent of

individual layer totals. Utilized or modified pieces are a constant if minor

component, with individual layer frequencies ranging from 0,1 to 1,4 per cent

of the layer total. Retouched artefacts are few in the lower layers, with

individual layer frequencies ranging from 0,1 to 1,1 per cent of layer totals in

layers 19-11. Frequencies in the upper layers increase, however, to the range of

1,4 to 5,0 per cent of individual layer totals.

Utilized flakes, 90,5 per cent of them silcrete, are the most common class

in the utilized and modified artefact category (48,5 % of the category total), and

are more common in layers 12-4 than in the other layers. Scaled pieces rank

second in frequency (41,7% of the category total) and are more or less evenly

divided between quartz and silcrete. Apart from layers 18 and 12, they are

more common in layers 9-4 than in the other layers.

Scrapers are the most common artefacts in the retouched artefact category,

accounting for 52,1 per cent of the category total. They are most common in

layers 12-4 but in layers 3-1 their predominance is superseded by that of adzes,

which account for only 12,2 per cent of the category total and assume relatively

significant frequencies only from layer 5. A limited morphological analysis

shows a difference between the scrapers in layers 13-10 and 8-1, those of layer

9 being intermediate between the two groups.

Backed scrapers (7,4% of the retouched artefact category total) are most

common in layers 7-4 and segments (11,5% of the category total) are most

common in layers 10-4 and again in layer 1. In layer 1 the segments are mostly

quartz, while in the other layers they are mostly silcrete.

Layers 19-13 yielded only 2,6 per cent of the total of retouched artefacts,

14,4 per cent of the utilized-modified artefacts and 27,9 per cent of the

unmodified artefacts, of which more than a third came from layer 19 alone.

The stone artefact assemblage suggests the following grouping: layers

19-18, 17-13, 12-5 and 4-1, although there are various aspects, particularly in

the retouched artefact category, that either cut across these groups or form

subdivisions within them.

Bone artefacts are a relatively minor component of the artefact assem-

blage. They are numerically most common in layer 1 but the ADs show higher

frequencies relative to excavated volume for layers 5 and 3. Frequencies, actual

or AD, are moderately high only in these three layers, and suggest that bone

was not an important component of the artefact assemblage in layers 19-6.

Marine shell artefacts consist of ornaments and edge-damaged Donax serra

‘scrapers’. Whole shells with perforations are relatively common only in layers

9-5 and 1 but are present from layer 16 while the more elaborately worked

‘pendants’, mostly made from Turbo sarmaticus, were not found below layer 9

and are most common in layers 6 and 5. Donax serra ‘scrapers’ are present in

BYNESKRANSKOP 1 97

most layers from layer 12 up but are common only in layer 1, in which stone

scraper frequencies are low compared to those for layers 9-4.

Ostrich egg-shell artefacts, mostly beads, are present in all the layers but

with individual layer frequencies of less than 1 per cent of the site total in layers

19-10. Beads are relatively most common, on the basis of ADs, in layers 8-6,

while fragments of decorated shell are numerically most common in layers 9-4,

being found only in layers 16 and 18 of the layers below layer 9. Apart from

one in layer 19, ‘flask’ openings were found only from layer 9 up, and ground

shell fragments, possibly intended or used as ornaments, only from layer 6 up.

‘Pendants’ were found only in layers 5-4 and 1. The mass of unworked

fragments varies from layer to layer, but frequencies are generally higher in the

upper layers, 9 and above, although layer 19 has the highest frequency of all

the layers.

Pottery fragments were recovered from layer 1 only, and the small number

of highly fragmented sherds provides little information of value.

Metal artefacts are represented by one small bead of copper or brass from

near the top of layer 1. This artefact, like the potsherds, is valuable chiefly for

its presence in the deposits rather than for any information it provides concern-

ing the ecology of the cave occupants.

HUMAN, ANIMAL, AND PLANT REMAINS

HUMAN REMAINS

As mentioned previously, during the excavation of the test pit in 1973, a

human burial was encountered at a depth of approximately 0,7 m from the

surface. These remains have been studied by H. de Villiers and the full analysis

is included in the report on the human burials from BNK, most of which came

from BNK 3 (De Villiers & Wilson 1982).

For the purposes of the present report it must suffice to quote De Villiers’s

conclusions that the remains are those of a child between the ages of 8-9 years

and that the small size and pentagonoid form of the cranial vault are indicative

of a Khoisan child probably assignable to the San (Bushman) group.

The skeleton has been added to the Museum’s physical anthropology

collection (SAM-—AP6053). It was found lying on its left side in the usual

contracted (foetal) position, with the skull facing approximately due north.

Ornaments or grave goods were not found in association. Because of the fragile

condition of the skull, it was treated in situ with a soluble adhesive.

It was not possible to establish from which layer the burial, found in what

is now layer 6, was made and correlation with the artefactual content of the

deposits will be possible only if a radiocarbon date is obtained. By analogy with

a child burial from BNK 3, however, it may be suggested that the burial was

made from between layers 2 and 4 rather than from layer 1.

98 ANNALS OF THE SOUTH AFRICAN MUSEUM

Fragmentary human remains were also recovered from layers 1, 5-6, 10,

and 14 (Table 15). As indicated in Klein’s lists of body parts (Table 8 of

Appendix 1), these are isolated teeth and/or postcranial material, mostly the

smaller limb bones. Because of their lack of association with more complete

burials these remains do little more than indicate the likelihood of disturbance

of the deposits, probably during the site’s prehistoric past as well as more

recently.

MAMMALS

The remains of some 600 individual mammals recovered from the deposits

were analysed and identified by R. G. Klein and these are listed in Table 15.

Klein (1981a) has reported on the mammalian fauna from the 1974 excavation,

which in general differs only quantitatively from that of the combined 1974 and

1976 samples.

Due to comminution of the bone or absence of diagnostic parts, identifica-

tion in some cases was possible only at genus or even family level. In this

connection it should be noted that in Table 15 the numbers for genera or

families include the individuals listed separately under species. Thus, for

example, the frequencies for Raphicerus melanotis are included in those for

Raphicerus spp. as well as in those for ‘Bovidae general—small’.

From Klein’s tables of skeletal parts (Appendix 1) it may be observed that

in general it is only the smaller animals that are represented by a more or less

complete range of bones and teeth. Four species, Atilax paludinosus, Equus ctf.

capensis, Phacochoerus aethiopicus, and Hippopotamus amphibius, are repre-

sented only by elements of the skull, and two, Ictonyx striatus and Panthera

pardus, only by post-cranial bones. In the case of the large bovids, individuals

in layers 19-10 are represented only by cranial and limb parts, while in layers

9-1 other parts are also present. This is also true for Oreotragus oreotragus and

small medium bovids, but in the case of Raphicerus spp. and large medium

bovids all skeletal parts tend to be represented where these animals occur. This

tends to dispel any suggestion of differential preservation in the lower and

upper layers.

Although allowance must be made for agencies other than human being

responsible for the accumulation of some of the bone, also that some of it

might have been brought into the cave solely for artefactual purposes, it seems

reasonable to assume that the bulk of the bone represents human food refuse.

In the analysis and discusssion that follow no attempt has been made to exclude

those individuals whose remains might not have contributed to the diet of the

prehistoric human inhabitants of the cave, with the exception of Homo sapiens,

microfauna and lizard, the last-named being too few to be of any consequence

(Table 21).

Excluding Homo sapiens, and allowing that the totals for genera and

families include the species or genera listed separately, there are thirty-six taxa,

mainly identified to species level, represented. No layer contains all the taxa,

BYNESKRANSKOP 1 99

and frequencies of these range from 9 to 21 in layers 19-10 and 7 to 26 in layers

9-1. Five species are restricted to layers 19-9: Atilax paludinosus, Equus cf.

quagga, E. cf. capensis, Phacochoerus aethiopicus, and Damaliscus dorcas,

while six were found only in layers 9-1: Ictonyx striatus, Aonyx capensis, Felis

cf. caracal, Raphicerus campestris, and Ovis aries.

Almost two-thirds (64,4%) of the individuals are in layers 9-1, but this

increase over the lower layers may in part be attributable to the different

excavated volumes, the upper layers having a total excavated volume 18,6 per

cent greater than the lower. However, no proportionate reduction of the

frequencies in the upper layers has been made to compensate for this, on the

assumption that the lower layers are likely to be more compressed than the

upper. The AD totals for layers 19-10 and 9-1 are 368 and 413 respectively,

indicating that the relative increase in the upper layers is only of the order of 12

per cent rather than 50 per cent. The increase in actual frequencies is, however,

81 per cent.

Frequencies in the individual taxa are generally too low to allow a

meaningful layer-by-layer analysis. In consequence, comparisons are mostly

restricted to the two groups of layers, and the frequencies have been combined

further by grouping the taxa into approximate size classes. Since the size and

mass of an animal are also relative to its age and sex, it has been difficult to

form valid size groups on the basis of the limited data available from the

remains. Table 16 summarizes such data as are available from Klein’s tables

(Appendix 1), in which juveniles and adults are indicated, where this could be

determined, on the basis of unfused or fused epiphyses. The table includes

some 90 per cent of the individuals listed in Table 15 but, as the third column in

Table 16 indicates, about 40 per cent of the total are listed as ‘indeterminate’.

This means that the age group divisions in fact cover only about two-thirds of

the total sample. Klein (19815) has arrived at differing frequencies in the age

groups of selected bovids, based on a study of dental wear.

The problem of whether the presence of limited body parts is merely a

chance factor partly dependent on preservation, or whether, and to what

extent, it represents the realities of prehistoric butchering and dietary practices

tends to make it almost impossible to form meaningful size groups, and it

would give a specious value to more precise attempts to interpret the dietary

habits of the site’s prehistoric inhabitants by using calculations based on the

flesh mass when these are based on such scanty data.

Given these problems, the data contained in Table 17 and Figure 39 have

limited value, and at best serve only to give a general idea of the distribution of

size classes represented in the faunal sample and to provide a means for

comparing changes in their relative frequencies. The six size classes used in

Table 17 and Figure 39 are based mainly on data derived from Dorst &

Dandelot (1972) and do not take into account the juvenile and adult frequen-

cies indicated in Table 16. Identified juveniles account for only 20,5 per cent of

the sample total and almost half of these are Bathyergus suillus, which falls into

ANNALS OF THE SOUTH AFRICAN MUSEUM

100

— if il il (das ee sno10d SnddOYIOUDIOd

I eed pa ee Bat oy Song es ee 5 tie ey fee I | Gr oro bo ciata od O-n1n 60 sisuadvo ‘Jo snnbq

Z = % I ia I 7 i I ae I ets, 28, BS 4 ees eet = 0 tO aoon ano ood GO v3sonb ‘Jo snnbq

eye Sepl ET Pies he! WO ee OT |e deleeen|| | i a Se ar | ee ‘ds ja “uad sepnoroooulyy

ee ye ep ST ee Sat aes aed hoe SUE corey ca & s1us091g so4a21q

(oa! I | = £ T i oa I IL I Bl ae sag GC ie eS eee sisuadvo vIAvIOAT

eR Mtn ry he a. I I I ea ele Re LK eet Gem ee en I I ee oe. = 600000 > Goud 600-0 puvoia{o DIUOPOXOT

oe te ba Pa ek: I ll I I <ar i C if I ae Ce a ee ee snqpisnd snjoydas0j04Vy

a, ee | ere fae ae ee SARE Ce ee Me. A: Se ee Ae es I ae a4 Se a Re a ea snpaod playjuvg

SME Se Rares Be ee ee: Ty Seer ae me eee ceee }DODADI “JD SHAH

10) I -% I p22 ae — | rch! TE. seats hood Beet pee oe I I I Tp) I I Ta ES ae ae ee pokqy sag

1) I tert te. = = I] ee ee, ee = = == 7 ——_ = = oe? Pa ey ae pauunig puavdy]

—- — — —_— — lL ---lcCOr —- — — —- — —_—_ — Ses US Me ee aa ake snsouipnyod xojuy

—- —- — —_— — Te —- —- —- — th I if Zz. ol lg — Te ae See ae snquajnaaajnd sajsadiay]

—- —- — —_— — ns Oe if I —_— — Ik: Sg SSS ee uowunauys1 sajsadiay]

2 a ee ln, Se eae Dt eit ce ok. oe ae I 2a palin = Se: (ene eer oS sisuadpo xduoy

eee teers me Meee a 2 en Te 8 I z=. a Oe a oe ee ey ee Th. he Raa ee en sisuadv9 DAOAIJOW

fae Sey eee Se Boy es Ce ae ee ain. Gee oe I Ba Se [OR ae ae snqv14js xAuojo]

sea Sew ee Ger ee eae Clee ahaa eo! be at [ee eee oe ee oe ae SDJQWIOSIUA “JO SIUD)

eS et Ss wen Fels I Soe SS ae es I tia} oe ae 1 oe ae suaidos OwoyH

Ese. ae a at eo! I = ne, if a I . if I I if as pee if G0ndDoG UO oon HOO Od On oo HO snuisan o1dvg

@IL tell EM Ol SI vi fel @ Wl OL © Se ile a9 Se~ al SG ll ‘OAR

‘S[ENPIAIPUI JO SIOqUINU WNUITUTW JO AJOJUDAUT :s[eUUTe/|

Cl ATaV

101

BYNESKRANSKOP 1

‘soroods

opun payst] Apeosye spenpraiput apnjour sjejo) Ayruey/snuayg ‘ulapy “OD “Y Aq pofiduioo sem ‘sq pur sjei0} sy} Jo UONdaoxa oY) YIM ‘a[qQe} OY], ‘AON

N FS Beh eae NGC

O€

| —4

i! 6A)

Foe mo mm OO CO

Doo Ro hoo 6 6 emamdun cemumadaddarr

Doon oGd oo OD oO (suaidps ‘H "[0x9) [R10 L

6406 FTO G00oOnDdOo moO DO snqjins snssadywog

Se Sar Nig Syvajsnv-av14f{v x14 FY

ououdDooKO0D O00 ‘ds 10 ‘ua3 sepiiodaT

Soo oooONvODOOoO do oOt oO Oooo oOo OO o31e']

coor ogdyoonadooo0d06 60 uINnIpoul o31e7]

Seles fos cee Le wunIpew [jews

[e1oues ‘oeplaog

Ms OR a eo Be a Salud S1A4Q

Saeco aia ae ceoEc oo oma Wa) ond af {vo sn4aouks

Se ee eee XAKAO SNSDAJOAND J,

Bond 500n00 snydvja2] Vy /sajavyIOuUn)

Seca. Ghost CACC RONG SvILOp SNISIDUDG

Aeon) MeO! DIDI D ec KIeO ROR OREO dds sn3p4oddiyy

dAG 000000000008 pinfnsoajnt pounpay

bos ae dod do Oop winulpundD pounpay

6 a)0.0. oo. Ove Geneon GO. nd dre Odo snjoasdvo Dajad

Aware o oeoRee ora sn8p4110a0 SN8DAJOIAC)

Eo Sart Raed aL E Ro Uiee eT oiioueeee dds snaan1ydvy

T= srasadummo snuaaiydoyy

PE Ne CeOeR ane none syouvjaU snaaa1ydvy

;

)

ay ;

‘

i any

. ¢

; |

. \

4 ’

t F 2

I t

Ms a

yh iy

“4 J

r ct y

, ;

; ~

~ ; }

, 5

i \

4 \ Par,

t i

) 4

t 7 VW

' ,

~ 7 e

; ¥

{ y)

= e

3 ) si

}

~ \ ’

4 . ‘ R 1

‘ an i

2 ban eee

ve a As

‘ a a :

\ Sn e

1 wet y

<4 4 n

? . * . :

i ee L big

) - f 4}

a \ {

oe t :

r i \

: 4 5 r) Ss)

* \ 4 At x

' ee

4 ys

iS 5 hi a, ¢ %

, } x d

, ' % z shy \

ayer

2 iged

pie Rates ath, = i SS SRT |

s\ i

7 fies 4

i 4 - #

M ai: i ; ' '

‘ 4

i ;

5 : u } wale hee

1 7 aul

\ By

On / a

) Nive wad if

1 f | i s

‘ ¢ ; : ;

‘%

1 b e

‘ ;

a, a . t

7 Phin

' * w ¥

< ¥

) 2

Nay)

oH is ay \ f h

i , th 4

]

‘ Hoh n

y ’

{

* ‘ :

}

, i

\ t

‘ 4

Ua er

x ‘ }

ey: ' J

i {

' “ f ; i jek

es A ") i

Yi uy

julie .

+ W) y

ea enees

- ‘ ’

}

. PhS,

ij U ¢ oly

a ,

‘ rb ‘ ; Apa

ey) 4 Oy ip ‘

net bal Pi =

=f / =) '

Gees * it I

x .

, p : 5 es

Z pe c Ey A8 "t

a ) é

i i i rr I

i eto - b I

y f } \ ; y

a \ A , Y

‘1 J \ ¥ lee Say

x vil} y i ‘ h Nie by ; 4

md - ee

f bina ea

= ate!

os 1

I tery

r i i)

} F

TABLE 15

Mammals: inventory of minimum numbers of individuals.

los)

Layer:

Papio urSinus.........0..00. 00. 0s

Homo sapiens .........0.0.02020.5

Canis cf. mesomelas ...............

A CIOWYX: SITIGIUS = sate cvesucs arotet « we set

Mellivora capensis.................

Aonyx CapensiS ..........6..0 000.

Herpestes ichneumon ..............

Herpestes pulverulentus ............

Atilax paludinosus ................

Hyaena brunnea .................. cf. — — cf.

EUS DY Cae. 222 Riels duvvxternns aia «tke

eUS\GhiGarQCaliae ee eco siecc sone

Panthera pardus ..................

Arctocephalus pusillus .............

Loxodonta africana ...............

Procavia capensis ...............-.

Diceros bicornis ..................

Rhinocerotidae gen. et sp. .........

Equus cf. quagga .................

Equus cf. capensis.................

Potamochoerus porcus......

ie}

-h

We Pr esha erste inact ol aly cdl nes

[pr Ss AS aire a hilar ee aN

w |

Peels hee |

Epil analog 4

Phacochoerus aethiopicus .

i Suidae, general

Hippopotamus amphibius ..........

Raphicerus melanotis ..............

Raphicerus campestris .:...........

Raphicerus spp......-..0+0+.e000e+

Oreotragus oreotragus .........+..+.

Pelea capreolus ..........0--+0+055

Redunca arundinum ...............

Redunca fulvorufula...............

Hippotragus spp........+-.0+-+0005

Damaliscus dorcas ..............+. _—

Connochaetes/Alcelaphus .......... 3

Taurotragus OrYX. 0.0... cece ee eee —

Syncerus caffer ....... ccc cece eee 6

OVS GIES ay scictete ie ests vate ib ins woe se 11

lelenwl] J

| TOSS Gee

ee ea eee Wee Seale das eva eee li ewpsievieredees vac

Wesel Guo

= |

Ae eet Seth Te eesetlfinse nls

| elas

STEM. oon ob masog dub eas po Bee 12

Lange MeGUm.... esse veces es 3

DAL DO teva) Weheponttarete sscuore ete riteares: 6

Leporidae gen. et sp............... 1

HK wWreeNH

Vas Clea resaletenl unloliomtatietllonss!'|) weset| ly Ay

once li rollins

ER el | it ft

eral N+ Reece afi fool

fh iy acest oF ogee? ex

i ogt be Tak fal

Poe ole WN a oh |

Ss ees |

BO cco | stil pogo: Lo! “Msi,

NNNND

ePrPpRew

\rex cone en

Hystrix africae-australis ............ 1

Bathyergus suillus ............0.... 37

Total (excl. H. sapiens) ............ 90

ADD / ini ae. ra Sarge aiccabe 6 35

5 19 49 IZ 5; 4

3925) 1211

72 47 85 46 30 36

N@ODKP eee PP

Note. The table, with the exception of the totals and ADs, was compiled by R. G. Klein. Genus/family totals include individuals alréady listed under

species.

| Sere ness

Wry

ANA

ONY WUNNHHEDWD

—

any

Pree taalarss cn Sk.

001

WAASAW NVOIdAV HLNOS AHL JO STVNNV

1 dONSNVUNSANAG

TOT

102 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 16

Mammals: minimum numbers of individuals in adult and juvenile age groups, as represented by

fused/unfused epiphyses (see Appendix 1).

Adult Juvenile Indet. Total

ICLONVIGSULALUS ren ee ee 2 — = 2(0)

Melliv OnGNCGDERSISH anne na eee 1 — 1(1) XCD)

AON COPCNSISi, nin earns one ee if — 1 2(0)

Ja AADOUA VC MIEDWOR o0cc000000000¢008 y — 3(1) 5(1)

Jaks | QUUSGANAOWS 4 sab 95480bo0¢408 soos 3 1 6(2) 10(2)

RCLSHLIDY COS 5 on c-n ale san 024 OER Oe 1 — 8(3) 9(3)

HIN CIACAT ACO See ye nt a a tt 2 — 5 5(0)

ArclocepialusspusilitiSi nr ere eee 3 3(1) 9(6) 15(7)

[POCAVIG CADENSIST.. aoa ee a ee 4(2) 3(1) 11(6) 18(9)

Suidae—-penerall ea. oe ee — 1 22(10) 23(10)

IRADRICETUS SPDi ae nin Aen ce eee 47(10) 29(7) 20(4) 96(21)

DOU HADS OQNGOUAELS ogo050000006000¢ 3(1) 4(1) 3(1) 10(3)

Bovidae—small medium .............. 18(7) 10(4) 4 SZ (Gin)

SVS WHSCMONIN 6 02020 aacaas 19(9) 7(4) 40(37) 66(50)

Planes he eda be. cen ines eee 18(9) 5 28(9) 51(18)

Wepondae—ceneraliay tay ee 4(3) — 13(8) 17(11)

Fel SENEXAIICAC-AUSULGI IS pee ee — 1 4(2) 52)

IO OARGIS SULTS 5 cocacccvocec0secoes 75(11) 61(12) 38(9) 174(32)

MRO ballet cy oye este eee ae ha vey See ane 203(52) 125(30) 214(99) 542(181)

Note. Numbers in parentheses indicate the numbers of individuals in layers 10-19.

TABLE 17

Mammals: comparison of frequencies in size classes in upper and lower layer groups.

Layers: 1-9 10-19

no. % no. %

Very small Ictonyx, Herpestes, Atilax, Felis (libyca),

(<10 kg) Procavia, Leporidae, Bathyergus............. Wy 255) 59 27,4

Small Canis, Mellivora, Aonyx, Felis (caracal),

(10-25 kg) Raphicerus, Oreotragus, Hystrix ............. MQ) 2)! 33 Io3)

Small medium Papio, Hyaena, Panthera, Arctocephalus, Pota-

(25-100 kg) mochoerus/Suidae (excl. Phacochoerus), Pelea,

REGUNCACOVISIS ES Secsctciee ee CIR ee ee: SL ORe il 14,4

Large medium Equus (cf. quagga), Phacochoerus, Hippo-

(> 100 kg) tragus, Damaliscus, Connochaetes/Alcelaphus 17 4,4 63 ZN9).3)

Large Equus (cf. capensis), Taurotragus, Syncerus ... 33 Oro) 18 8,4

(>500 kg)

Very large Loxodonta, Diceros/Rhinocerotidae, Hippo-

> OLAIUS. si i ek ee ak eee ce eae eos) Cee : ;

(>1000kg) p 1? § 22:80). ty Baom

the smallest size class. It is perhaps worth noting that the ratio of identified

adults to juveniles is almost the same in the upper and lower layers: 1,6: 1 in

the lower layers, 1,7: 1 in the upper. There is thus less than a two per cent

increase in the relative frequency of identified juveniles in the upper layers.

The size classes for ‘Bovidae, general’ in Table 16 follow Brain (1974, table 2)

and these are the same as those used in Table 17, Ovis aries being included in

the small medium class on the basis of comparative modern data.

BYNESKRANSKOP 1 103

oi “QL WT YW ize

Small medium

50-

40- dL

5 oy

30- a

20-

ayn

Large medium

2 YY Died - :

10- 6

ce DD ee OY Porto tS TLL.

Very large

Layer eto MAS Ow Wa Onn so a tOn ail 12) Se J4e5; 6 17 18, 19

Fig. 39. Histograms of percentage frequencies, by layer, of mammal size classes. Hatched areas

indicate bovids, unhatched areas other mammals. Numbers above each column are actual

frequencies.

Arctocephalus pusillus has also been included in the small medium size

class although only three of the fifteen individuals have been identified as

juveniles. The assumption that few, if any, fully-grown individuals are repre-

sented is subjective but receives some tenuous support from the data from

Elands Bay Cave (Parkington 1972: 239-241; 1976, figs 3-4, 8). In any event,

since frequencies in the upper and lower layers are about the same (8:7), the

range of error for each group may also be about the same.

104 ANNALS OF THE SOUTH AFRICAN MUSEUM

The most notable difference between the upper and lower layers is in the

large medium size class, which drops from 29,3 per cent of the layer group total

in the lower layers to 4,4 per cent in the upper. Making allowance for the fact

that both Equus species are now extinct and that Hippotragus and Conno-

chaetes/ Alcelaphus have not been identified as to species, the animals in this

size class may be described broadly as gregarious grazers with a preference for

open or lightly wooded grassland (Dorst & Dandelot 1972: 159-165, 174,

204-205, 221-222, 227-228, 230-232 for comparable species, where applicable).

Equus cf. capensis, Phacochoerus aethiopicus, and Damaliscus dorcas are

restricted to the lower layers, Equus cf. quagga is not present after layer 9, and,

although Hippotragus spp. and Connochaetes/Alcelaphus are present in the

upper layers, their frequencies are markedly reduced by comparison with the

lower layers (Table 15).

There is no significant change in the relative frequency of the large size

class, but that of the very large size class is reduced in the upper layers to about

55 per cent of the relative frequency in the lower layers, even though the actual

counts are the same. There is no significant change in the relative frequencies

of the small medium class (a 7% reduction in the upper layers) but, whereas

the combined frequency of the small and very small size classes in the lower

layers is 42,7 per cent of the total for these layers, in the upper layers the very

small size class alone accounts for 45,5 per cent of the total and the two size

classes combined make up 70,9 per cent.

The indications are, then, that in the upper layers there was a marked shift

away from the procurement patterns of the lower layers, with more animals in

the smaller size classes and fewer in the larger being obtained in the upper

layers than was the case in the lower. An approximate calculation of the total

live mass of the individuals, excluding the very large size class, suggests that

there was a marked reduction in the total mass procured in the upper layers,

especially when comparisons are made on an AD basis. The estimated total live

mass for the upper layers, 31 972 kg, is only about 10 per cent higher than that

for the lower layers, 29 154 kg, but the AD for layers 9-1 is 3 625 while that for

layers 19-10 is 4 827, 33 per cent higher than the AD for the upper layers.

Allowance should also be made for a higher proportion of inedible residue in

small animals than in large, relative to their total mass.

Although no great confidence can be attached to such calculations, based

as they are on approximations and imponderables, they do allow the suggestion

that the increase in the number of animals procured during the second half of

the occupation of the cave was not sufficient to provide the same amount of

meat as had been procured during the first half.

Although most of the layer totals are so small as to make it of dubious

value to convert them to percentage frequencies, especially when they are

divided among the six size classes, to do so can have some use in indicating

change on a smaller scale than the bipartite division into lower and upper

layers. Figure 39 presents in histogram form the percentage frequencies of each

BYNESKRANSKOP 1 105

of the six size classes in the individual layers and, while this shows that there is

some justification for the bipartite division, it also shows that there are

variations that transcend this division. As is general in this report, however, the

emphasis is on the study of trends rather than of layer-by-layer variation.

In the very small size class two major trends are discernible, a decline in

relative frequencies from layer 19 to layer 10 and an increase in the overlying

layers, at least up to layer 4. The decline in layer 3, followed by increases in

layers 2 and 1, is not easy to interpret on the basis of the mammalian fauna

alone. It may in some way be related to the marked increase in the frequency

of marine shell in this layer, which is discussed in the following section.

The small size class shows a generally increasing trend from layer 19 to

layer 5, followed by relatively similar frequencies in layers 4-2 and a decline in

layer 1.

The pattern of variation in the small medium size class is more complex

than those of the two preceding size classes. It is possibly most simply

interpreted as showing lower relative frequencies in layers 19-15 than in layers

14-10, an erratically declining trend in layers 9-5 and an increasing trend in

layers 4-1.

The large medium size class also varies erratically in the lower layers but

the trends may be interpreted broadly as increasing to layer 13 then decreasing

to layer 8, after which frequencies are stable but low. The cyclic trend in layers

19-9 is a factor that would have been overlooked had the simple bipartite

division been the only method of comparing the content of the layers.

The large size class, which consists entirely of bovids, shows a mildly

declining trend in layers 19-13, an increasing trend to layer 9 and thereafter

minor fluctuations to the end of the sequence.

The very large size class has very low frequencies, both actual and relative,

and such fluctuations as are evident in Figure 39 are not considered significant.

Perhaps the most important trends are the increase in the relative fre-

quency of large medium animals up to layer 13, that is, until the early

Holocene, after which they decline until by layer 9, some 6 000 years ago, they

are no longer an important factor. The consistent increase in the relative

frequency of small animals is one that transcends the bipartite division. This

and the declining trend in the very small size class up to layer 10 serve to warn

against perhaps too simplistic interpretations based on broad groups of layers.

The ratio of bovids to other animals is equal in layer 19, higher in layers

18, 14-9, and 7 but lower in layers 17-15, 8, and 6-1. This indicates that there

was only one main period, from about 10 000-6 000 B.p., when bovids were

more common than all other animals brought back to the site. With the

exception of Oreotragus oreotragus and Taurotragus oryx, which are browsers,

and Raphicerus spp. and Redunca fulvorufula, which are mixed grazers—brow-

sers, the bulk of the indigenous bovids are primarily grazers (Dorst & Dandelot

1972). Including the equids, there are more grazers in layers 19-9 than

browsers or mixed feeders, except in layer 16, in which the ratio is equal,

106 ANNALS OF THE SOUTH AFRICAN MUSEUM

as is the case in layer 8. In layers 7-1 there are more browsers and mixed

feeders than grazers.

SHELLFISH

Remains of close on 25 000 identifiable individual shellfish were recovered.

As the inventory in Table 18 indicates, shell was recovered from every layer,

but the increase in frequency from layer 9 up is in the range of at least one

order of magnitude. As is usual in this report, the marked difference in the

frequencies of the lower and upper layers causes them to be treated as separate

units, with the major emphasis inevitably given to the upper layers. In the case

of the shellfish, layers 9-1 contain 98,6 per cent of the total of identifiable

individuals.

As mentioned in the discussion of the stratigraphy, layer 3 is virtually a

shell midden deposit and was, by virtue of the high density of shell, one of the

few clearly definable stratigraphic units. The AD of 12 120 for this layer is

more than three times that for layer 2, which has the next highest AD, and ten

times that for layer 9. Apart from the anomaly of layer 3, however, the ADs

given in Table 18 indicate a gradual increase in the frequency of shell through-

out the deposit.

The numerical frequencies given in Table 18 represent the minimum

numbers of identifiable individuals. The bulk of the shell is fragmented and

counts were based on diagnostic features such as head or tail valves for

Dinoplax gigas, umbones for bivalves, and apices for gastropods, as well as

opercula in the case of Turbo spp. Where there was more than one diagnostic

part, counts were based on the higher frequency of either part.

In several cases counts were made only at genus level, either because of

difficulty of identification at species level, or because some genera are repre-

sented by a principal species, with only few individuals of other species. The

principal Haliotis species is H. midae, and difficulty was experienced in identify-

ing some of the smaller individuals, which may be either H. (sanguineum)

spadicea or immature H. midae. The Patella species are detailed in Table 19

and are discussed later in this section. At least two species of Burnupena could

be identified, B. delalandii and B. (papyracea) cincta, but identification of most

individuals was not possible at species level due to poor preservation. Because

of fragmentation or loss of specific diagnostic features, about three per cent of

the mussels, Choromytilus meridionalis and Perna perna, and about five per

cent of the limpets, Patella spp. could not be identified at genus and species

level respectively.

Turbo sarmaticus, with 31,4 per cent of the site total, is the most com-

monly represented shellfish at the site and, except for layer 3, dominant in all

the upper layers. As indicated in Figure 40, there is a clear secular trend, with

frequencies increasing in layers 9-7, then decreasing to layer 3. This is comp-

lemented by a reverse trend in the frequencies of Oxystele sinensis which

decrease in layers 9-7 then increase to layer 3 in which the frequency, 38,8 per

107

BYNESKRANSKOP 1

OCL VC

COTE

SIL P

6c0 IT

SSLL

VLE €

109 ¢

C9C

L09

88C C

Gee

168 I

LEV

mM Oo D

Sl iS

atc

0¢ cy (G6

cl (G3 Or

I — —

I = if

I — ——

=" C €

= if G

ra I I

C = T

ee a I

v 16 e

ie 6 I

oa — I

€ 6 a

ST v1 el

OCT

ILI

(Qi GL WS Gsl SSSe GBC WISE OC SE wBe¢ “Vere seers seconees”™ wav

OH Gre ALi Get GIG. OSE BWC CME WIG Gye_ --mrrrerreeoeeerees [e10} Toke]

oe ea ams ae a at es ae G _ pesesauasesecosouudwcsecven: TO

€ gal OIL 78 rani Ter L9Z TSP ZOE OG ES ae St ‘dds puadnuing

€ ZOL 9€ 61 89 Liv 06S a — Silty OSSM 5 an ae ee eae sisuauls a]aisXxO

LS Svs 979 LI8 Met OL VOL GO GOL GG “rr eeeeeeomeees je1oy “dds oquny

ep S87 29 019 S56 007 T $99 669 LOL OLOGa Re le ee Rees SNINOULADS * J

rl = 097 POL LOT 60% €9S 6rE OEE 7ZE Cyo me ae ee ara SLUDP19 OGAN J

L LST 061 P61 6€€ ces 7SZ 00€ 617 GE ae ae a ae eal ‘dds vyjaivd

if 61 IZ IZ 6S 78 CZ 9 p CZ CGonDoonoVoonoOoDPHAD OO ODO ‘dds SyOND EY

8ST 69 Iv 6S OL SP 9 S € Ve a ee ea padas XDUOG

Gi LOU 76 Or L8 797 801 a 6r QUGs[e > RT ae 1210} DUsag—SnjNAUOLOYD

v sii 9 6 ¢ oT = = == i eae 2 ‘JOpuUl DUsag/snjyAwos1oyy

I cE cE ll €T IS a7 €€ O€ OL ohletiah\<clelisiivallshi=tlsit sixalleitel (eM alieitelfell «Melisiis vudsad Dusad

Oe Ln IS 61 69 Lol 19 II 61 Sica ee ge SYDUOIPLIAU Sn]NAUOLOY)

Gi el Iz cI a 87 9¢ LI LI oe ie Grae eg s0818 xvjdouiq

OL 6 8 L 9 ¢ p € Z I Toke]

‘Blouas/satoods Jo AIOJUSAUT :YSYT[OYS

QT ATaV LE

108

Genus/species

Dinoplax

Choromytilus

Perna

Donax

Haliotis

Patella

Turbo cidaris

Turbo sarmaticus

Oxystele

Burnupena

Layer

ANNALS OF THE SOUTH AFRICAN MUSEUM

10°

<1 <1 <1 <1

20=

10=

<1 <

102

<] <) <]

<] <] <l

Fig. 40. Histograms of percentage frequencies, by layer, of major shellfish genera/

species in layers 1-9. <1=less than 1 per cent of the layer total.

BYNESKRANSKOP 1 109

cent of the layer total, exceeds that of Turbo sarmaticus (23,1%). Oxystele

sinensis is much smaller than Turbo sarmaticus and nearer the size of T. cidaris

so that the increase in the frequency of Oxystele sinensis, measured against the

decrease in that of Turbo sarmaticus, cannot be seen simply as the replacement

of one food resource by another.

The frequencies of Patella spp., Turbo cidaris, and Burnupena spp. are all

under 20 per cent of individual layer totals. Patella frequencies are stable in

layers 8-5 but decline thereafter, while those for Turbo cidaris increase in

layers 8-6 then decrease to layer 3, and those for Burnupena spp. decline in

layers 9-6 and increase to layer 3.

Frequencies of the other genera are all low and do not call for any special

comment except in the case of Choromytilus meridionalis which, with fre-

quencies ranging from less than 10 per cent of the layer total in layers 9-4 and

less than one per cent in layers 3 and 2, increases quite markedly in layer 1 to

16,9 per cent of the layer total.

As mentioned in the preliminary report on BNK 1 (Schweitzer & Wilson

1978: 137) Donax serra is not considered to have been collected by the cave

occupants primarily for its food value. D. serra artefacts in the form of

perforated shells used as ornaments and edge-damaged fragments (‘scrapers’)

have already been discussed, as has the use of these shells for decorating

pottery, but since the use of this shellfish for food prior to the use of the shell

for ornaments or tools cannot entirely be discounted, frequencies of unmodified

valves are included in Table 18. It is of some interest that, although frequencies

of all shell in the lower layers are so low as to be statistically negligible,

D. serra is the most common shell in these layers, as is also the case with

artefactual shell (Table 12). a

The category ‘other’ consists of 9 valves of the surf-clam Scissodesma

spengleri from layers 1, 9 and 17-19, 5 valves of the freshwater mussel Cafferia

caffra from layers 1 and 9, and two unidentifiable fragments, probably of oyster

Ostrea sp(p). from layers 1 and 15. Although no actual localities are known in

the vicinity of BNK 1 where Cafferia caffra is to be found today, it seems likely

that it was obtained from the Uilkraals River. Fragments of the razor-shell or

pencil-bait Solen capensis were recovered from layers 2, 5-6, and 9-10 but

because of the high degree of fragmentation and a general lack of diagnostic

parts counts were not made. There were probably not more than two or three

in any layer.

Interest has been shown by archaeological researchers working on the

Cape west coast (e.g. Parkington 1976; Robertshaw 1977, 1978, 1979; Bucha-

nan et al. 1978) in changes in shellfish sizes as possible indicators of the effect of

human predation on shellfish populations. It has been considered that a

diminution in the size of genera such as Choromytilus and Patella in successive

layers of midden deposits is an indication of the ‘farming down’ of shellfish

populations, with continuous collecting preventing the shellfish from reaching

maximum growth.

Za)

fo)

110

B10)

sotoads

| esis

val

Gil

vOv

6te

‘saroads vyjajvg JO A1OJUSAUI :YSYyTTOYS

6] ATV]

0 wyav

166 tt ttt [e10} 1ake7q

[Gt tt tet eeeeeees ypu]

Totten s1upjnqnl “dq

Gee oo eee smno0 “dq

Zo ttt eens DIDIUIU “¢

CE] AAT ease REE cha MRC LCCING COMMON oo DISOI18UO] fel

EG Sbubanorctano won stavjnuvss *¢q

ep 92 289A OB TO Ro one. pUNDUDAS “gq

ee pssasduiod ‘gq

Sp bebo b een ac anu neess 409]YI09 “d

fo URE EGGcc aan tones DADGADG “dd

(eh Ee kee 10]]1AuasAD ‘d

, Take]

BYNESKRANSKOP 1 111

The Choromytilus meridionalis samples from BNK 1 are too fragmented to

provide numbers of unbroken shells adequate for metrical analysis. The same is

true of the Patella species, though to a lesser extent. An inventory of the

species is provided in Table 19 and percentage frequencies in histogram form in

Figure 41. All the unbroken shells from layers 1 to 9 were measured and the

length data are presented diagrammatically in Figure 42. The numerical data

for the sample sizes given in this figure indicate that the frequency of unbroken

shells in any layer ranges from 38 to 57 per cent of the total of identifiable

Patella in the layer. It was not possible to estimate the lengths of the broken

shells and it is thus not certain whether there is any bias in the sample, perhaps

towards the preservation of larger shells.

Branch (1971, 1974a, 1974b) has published details of his research into the

ecology of modern Patella populations on the coasts of the Cape Peninsula, and

these provide useful parameters for the analysis of samples from archaeological

deposits. Since Branch studied populations on both the warmer False Bay coast

and the colder Atlantic coast, it would seem legitimate to use his data for

comparison with the samples from BNK 1, since the coast from which the

shellfish were probably collected lies within what Stephenson (1944: 297) has

termed the ‘western overlap’, an area in which the marine life is affected, to

varying degrees in different localities, by both the warm waters of the Agulhas

Current and the cold waters of the Benguela Current.

Branch (19745, figs 13-14, 16-17) has given the population structure of the

Patella species with which Figure 41 is concerned, except P. argenvillei which,

in any case, is inadequately represented in the BNK 1 sample. The data

contained in Branch’s histograms have been used to provide a basis for

comparison with the archaeological samples and are included in Figure 42. The

frequencies of most of the BNK 1 layer samples of the five species are too low

for them to be amenable to valid testing of statistical significance and the data

are generally useful only for comparison at a broad level.

P. argenvillei, for which no comparative data are given by Branch, is

present in low frequencies. These indicate little change in mean length in layers

9-5 but a decline in layers 4—2, layer 2 having the lowest mean length of all the

layers. The mean length for layer 1 is closer to the means for layers 9-5 than to

those for the intervening layers.

P. cochlear has mean lengths in layers 9-4 that are close to the mean

length of the total modern population, while those in layers 2 and 1 are closer

to the mean for the population of individuals 20 mm or longer (hereafter called

the larger modern population). There is a mildly increasing size trend in layers

8—4 but layer 3 has the lowest range and mean of all the layers, and the mean is

also lower than that for the total modern population.

P. granatina has mean lengths in the archaeological samples that are

mostly well beyond those for the modern population. The mean length

increases in layers 9-7, decreases in layers 6-3 and increases again in layers 2

WI ANNALS OF THE SOUTH AFRICAN MUSEUM

Species %

20=

102

P argenvillei (sar, wee: ery

10=

R barbara SS ee SS SS OS

202

102 3 —_

P cochlear

102

30°

20°

10°

PR granatina

102

60=

50°

40=

30°

20°

R longicosta

102

P miniata <I <!

302

20=

P oculus

10°

102

ee Pie ee eee

Layer 1 22 3 4 5 6 7/ 8 9

Fig. 41. Histograms of percentage frequencies, by layer, of Patella species in layers 1-9.

<1=less than 1 per cent of the layer total.

113

BYNESKRANSKOP 1

998 ‘sjoquids Jo uorjeuejdxa 10,4 ‘afdures yoeo ut S[ENPIAIpPUI Jo Joquinu

Joy (q) pue uonejndod sjoym ay} 10y (e) are uoneindod ulapour oy

2y} Woy suonefndod usspour 10} eyep sayeredutoo YUM 6-] Ss1oAe] WIOI

1&4

£7

TAS

‘ou

—

~£:

REED

eff —

snjnoo g

827

Lvs

8€

£6

621

‘ou

DysorBuo] q

€7

‘ou

"g ons

puijoun16 4

(a At

907

€7

4Da]Yyr0> 4

os 09

(S40p ou)

(240p ou)

i or

n --

EER

cee

rat Ae

a SBesh—

ou Jajjauabin g

OY) SBdIPUI WeISeIpP YORd JO IYSII 9Y} 0} sIoquINNy “193u0] Jo WU Oz sjenprarput

1 Joy sweseIp oy, “(LI-OT ‘pI-€l s3y ‘gpL61 yoursg oye) ejnsutuag odeD

J STPYS YJJaIVq USYO1quN Jo syysud] Jo sweIdeIp seelay—soIq peyIpoyw ‘Zp “SL

(ww) y46ua7

(q) wapow

(2) wiapow

4ahvq

114 ANNALS OF THE SOUTH AFRICAN MUSEUM

and 1. Layer 3 again has the lowest mean of all the layer samples though this is

still somewhat higher than that for the larger modern population.

P. longicosta is in general the only species with individual layer samples

that are large enough to be statistically reliable. The mean length in layers 9-5

is closer to that of the larger modern population but in layers 4-1 it is

intermediate between this and the mean for the total modern population. The

mean length declines in layers 8-3 but increases again in layers 2 and 1, and it is

again layer 3 that has the lowest mean of all the layers.

P. oculus has mean lengths greater than those of the modern population.

These show a gradually declining trend in layers 8—2 while the mean length for

layer 1 is the same as that for layer 3. This species, lik P. argenvillei, has the

lowest mean in layer 2 rather than in layer 3.

Branch’s population data indicate that almost 70 per cent of the P. coch-

lear population is less than 30 mm long. The P. granatina population distribu-

tion, on the other hand, is bimodal, with a distinct falling-off in frequency

between about 35-55 mm. About 58 per cent of the population is smaller than

35 mm and about 32 per cent larger than 55 mm. In the case of P. longicosta

almost half the population is less than 20 mm long and only 25 per cent is

longer than 50 mm. Less than 15 per cent of the P. oculus population is less

than 30 mm long and just over 60 per cent is in the 40-60 mm range.

The indications of the BNK 1 samples are, therefore, that collecting was

generally directed at the larger-sized individuals, although in the case of

P. cochlear it appears that, except in layers 2 and 1, it was the whole

population that was being exploited. That layer 3, which has the highest

relative frequency of shell, also has the lowest mean lengths for three of the five

species may be indicative of heavy human predation, although the low frequen-

cies of all the species except P. longicosta suggest that such an inference be

treated with caution.

Unless preservation factors are involved, the virtual absence of individuals

under 20 mm long can probably be explained by their being too small to be

worth the effort of collecting. Moreover, Branch (1971) points out that the

juveniles of several species make their homes on the shells of larger individuals

of the same or other species, also on Oxystele sinensis or Pyura (red-bait). The

small individuals drop off almost immediately if the host is inverted. A

probable explanation of the scarcity of the largest individuals is that they are

comparatively scarce in the population structure (cf. Branch 19745, figs 13-17).

In many cases, too, the larger individuals tend to occupy zones of deeper water,

though this is species-dependent (Branch 1971). In this connection, it is worth

noting that the species most common in the BNK 1 samples, P. granatina,

P. longicosta, and P. oculus, are those that occupy the higher littoral zones

(Branch 1971, figs 3-4). The virtual absence of P. granularis, described by

Branch 1971: 7) as ‘probably the most widespread as regards vertical and

horizontal distribution’, can possibly be explained by the fact that the bulk of

the modern population is less than 30 mm long (Branch 19745, fig. 15) and

BYNESKRANSKOP 1 138)

might not have been worth the effort of collecting in an area in which larger

individuals of other species were readily available. The bulk of the P. cochlear

population is, however, also mostly under 30 mm long and is much less

accessible than P. granularis, which occupies the highest littoral zone of all the

Patella species (Branch 1971, figs 3-4).

Branch (19746: 182) names P. longicosta as one of several Patella species

in which the size decreases as the mean sea temperature increases. Given the

relatively long period of time, almost 2 500 years, spanned by layers 8-3, it is

possible that the decreasing length of this species in these layers may reflect a

gradually increasing sea temperature rather than the effects of human preda-

tion. On the other hand, in layers 6-4 the minimum length of the samples

measured increases while the maximum length remains constant. This appears

contrary to any indication that the overall length of the population was, for

whatever reason, diminishing.

Stephenson (1944: 316) and Branch (1971: 11) describe P. oculus as the

warm-water ecological counterpart of P. granatina, which has the eastern limit

of its habitat in the Agulhas region. Although the P. granatina frequencies in

the BNK 1 samples are variable, up to layer 4 they are consistently higher than

those of P. oculus, which then becomes the more common species (Fig. 41,

Table 19). If the decreasing mean length of P. longicosta in layers 8-3 is taken

as indicative of an increasing sea temperature, the increasing frequencies in

layers 9-5 of P. granatina, which has a preference for cold water, cannot be

taken to indicate a decreasing sea temperature during the same period. It may,

however, be of some relevance that the P. granatina: P. oculus ratio drops

from about 4,4: 1 in layer 7 to 1:2 in layer 3 and the decline continues in layers

2 and 1. However, although the relative frequencies of these two species do

show an inversion, and the frequency of P. oculus also increases constantly in

layers 7-1 relative to all other species (Fig. 41), the mean and maximum

lengths show a more or less constant decline in layers 8-2 (Fig. 42), whereas in

an increasing sea temperature they should be expected to increase.

It would seem, then, that the data derived from the measurement of the

length of the BNK 1 Patella samples do not lend themselves to the extrapola-

tion of inferences regarding past changes in sea temperatures. This is not

surprising, considering the number of variables involved, of which perhaps one

of the most important is the probable collecting area. There is an upwelling of

cold water at Danger Point (Branch 1979 pers. comm.) and changes in the

littoral ecology could result from shifts in the flow of this water without it being

necessary to invoke an overall change in mean sea temperature. Branch (1979

pers. comm.) has pointed out that a localized predominance of P. longicosta or

P. granatina is a common occurrence on the False Bay coast, depending on

where there are up-wellings of cold water.

Layer 3, with the highest AD of all the layers, for Patella as well as other

genera, generally has the smallest sizes of all the Patella species other than

P. argenvillei and P. oculus, for which the lowest means and maxima are in

116 ANNALS OF THE SOUTH AFRICAN MUSEUM

layer 2. By contrast, layer 1, which shares with layer 9 the lowest AD for

Patella, tends to have higher means and ranges for most of the species. While it

might be expected that more intensive predation, as indicated by the layer 3

frequencies, would result in a lowering of the mean size of the shellfish

collected, less intensive predation, as suggested by the layer 1 frequencies,

could be expected to result in greater selectivity, with fewer small individuals

but, except for P. cochlear, this does not seem to have been the case. A

possible interpretation is that predation was as intensive as in layer 3 times but

with more, and possibly smaller, human groups competing for the same set of

resources. This could have had the effect of fewer Patella being brought back to

BNK 1 but these covering a wider, less selective, range of sizes.

In addition to the fact that the layer frequencies of the five Patella species

are, apart from P. longicosta, too low to provide reliable data, it is felt that the

chronology of the upper layers is too extended and contains an insufficient

number of dates for any indications of changes in shell length to be capable of

being interpreted as evidence of the effects of human predation. The life-span

of the Patella species ranges from 2 to 3 years for P. oculus to 15 to 25 years for

P. cochlear (Branch 19745, table 3). The upper layers of BNK 1 span some

6 000 years, of which about half is covered by layer 1 alone, and even the span

of 200 years or so between the oldest date for layer 1 and that for layer 2 covers

from several to many lifetimes of the various species.

It must be concluded, therefore, that although the shellfish remains from

BNK 1 provide evidence of an increasing reliance on this food resource, from

at least layer 9 on the layer samples, together with their chronology, do not

lend themselves to the extrapolation of inferences regarding changes in sea

temperatures or the effects of human predation on shellfish populations.

FISH

Fish remains were recovered from all layers except 16 and 18. These were

identified by C. E. Poggenpoel and are listed in Table 20.

As indicated in the inventory, there is a marked increase in frequencies

from layer 9 up, although these are still low. The ADs indicate rather more

clearly than the actual frequencies the progressive increase up to layer 5, after

which frequencies decline. It is of interest to note that there is an approximate

correlation between the number of genera and the number of individuals, with

an increase in the number of individuals corresponding with an increase in the

number of genera.

Pachymetopon blochii, with 37,4 per cent of the site total, is the most

commonly represented fish, followed by the combined totals for Lithognathus

spp. (16,5%) and Rhabdosargus globiceps (15,8%). Pachymetopon blochii

generally lives among kelp (Ecklonia maxima, Laminaria pallida) growing in

deeper water off rocky shores, while the other two genera favour sandy areas

and may be found off estuaries.

BYNESKRANSKOP 1

bLE

PCI

8ST

£0

[®10}

aus %

[8}0}

satsads

Z i 5 2 ¢ i He Ce fe @) “ea Gro () . 7eseuyppcopecb canon Puc snanmunn a: way

¢'0 v0 L0 b0 v0 Cit 09 Cr 6Z P91 9'6E £9 9°72 2 gE ODO UOMO O DODO OAD DOD RO COO 6 #10} ouIS Y

I i Z I I p Ol ZI IZ 8b SOI TZ h Il OL avaiialieltalisiie fo abieiienei (sil) (s.ie)is\iolin\iels/tvltvilel'=jielalir/isie ye101 JoAeT

, ey 3 Lae se | Se ae ae Sein ee Gentes poynuepiug

Se et ee ae ag le ee ee ee ae a Eo AE GA ACER RE oe Iuospavyou vz]

== = — —_ == = — sen oa if = — — — See Be chs ei a Oo DUOZOAKBAD DUOZOLKBIY

Seal See, SS SO be Cy 10. Tye re eC eners, AC eee a se ee 1yoo)q uodoraudyoog

—_ ee, = ane ee = a € Z I I Naked ee. if —— —o SHbnH ODDO ooo ROoDOdODD DODD dds snyjousoynTy

saan ooh ans or aa xe, pes eo, ne. = € sae as. == —— © O0GODDCDHD DO ODoOnO dD OOnD ODO oO SnAKWAOUL aT

Sig Sp Ee em ily, Ce Ve OC ae CoC ee eee ate aa snyjoudoyin snywusoysT]

wes = ier _ ie I € es ae I I eds = i: ==. *6ono0coon0anoooD0b ato0d m0 sn3Suvs snpojdiq

Se ae al ie ae te Oj “Ce EC eS Ol, are a I a ee ee sdao1qo]3 sndavsopqoyy

as = I as oat. es Sead I Z Il € I 2 ae ce - d.00800uHo00G000000000000 sisuadvo snu1g0109,

ee a Se I = I eS CO OT ae snjopidajojoy sniuosoi1ksay

a. sie We ag a BS a ee ee ae eee fe a eg ee enncens eee oe Jowwyvs snuowuog

ST v1 €l cl Il Ol 6 8 L 9 S v € C I qokeT]

‘saroods Jo AIOJUDAUT ‘YS

VWAESCOAD

118 ANNALS OF THE SOUTH AFRICAN MUSEUM

Local informants (M. & J. Kemp 1979 pers. comm.) have advised that

there are areas between Die Kelders and Pearly Beach where certain genera

are more commonly caught. Apparently this does not depend on a simple

distinction between rocky shore and sandy beach, and may also have to do with

off-shore currents and localized up-wellings of cold water. Pachymetopon

blochii is more common on the Walker Bay coast between Die Kelders and

Danger Point, Diplodus sargus at Die Kelders, Coracinus capensis, Argyroso-

mus hololepidotus and Lithognathus spp. at Die Gruis, a rocky outcrop fronting

a sandy shore between Uilkraalsmond and Pearly Beach, while Mugilidae are

best caught at Romansbaai between Gansbaai and Danger Point, and at

Uilkraalsmond, which is also a favourable locality for Argyrosomus hololepido-

tus and Lithognathus spp.

From measurements of dentaries, Poggenpoel (1979, fig. 8) has extra-

polated the lengths of thirty Rhabdosargus globiceps in the BNK 1 sample.

Poggenpoel’s calculations show that the range of this sample, 170-380 mm,

exceeds that of the numerically larger sample from Elands Bay Cave

(150-330 mm) as well as those of four modern marine samples (130-350 mm),

and is far greater than those of four samples from estuarine environments

(25-150 mm). The BNK 1 size range indicates that the fish were taken from the

open sea rather than from the Uilkraals estuary, but the maximum length of the

sample is well below the largest recorded size of approximately 510 mm (Smith

& Smith 1949: 47).

ROCK LOBSTER

Remains of the rock lobster Jasus lalandii were not recovered from the

deposits, nor were they recorded from Die Kelders (Schweitzer 1979). They are

to be found in the area today although they are extremely scarce, except in

Gansbaai harbour, as the result of excessive human predation. The possibility

of a preservation factor cannot be excluded, although this is unlikely to have

been the case at Die Kelders as well as at BNK 1, and the remains are

preserved in open midden sites.

BIRDS

Bird remains from the site are being analysed by G. Avery, but the

analysis is not yet complete and will doubtless be published elsewhere.

REPTILES

Tortoise counts based on humeri were undertaken by L. Lawrence. The

frequencies given in Table 21 represent the minimum numbers of individuals in

each layer, based on the higher humerus count, whether left or right. Two

species were recognized, the land tortoise Chersina angulata, and the water

tortoise Pelomedusa subrufa, both of which are common in the area today.

A remarkable feature of the deposit was the dense concentration of

tortoise remains in layer 14. Like the dense concentration of marine shell in

BYNESKRANSKOP 1 119

TABLE 21

Reptiles: inventory of species.

Tortoise Snake* Lizard*

Layer Chersina Pelomedusa AD/m? Pseudaspis Bitis indet. Cordylus Agama

angulata subrufa (both spp.) cana arietans cordylus atra (?)

1 261 21 108 5 1 4 — —

2 52 5 100 10 1 1 — —

3 62 6 272 1 2 — — —

4 65 6 78 1(+) — 2 — 1

5 184 10 156 5(+) — 1 5 —

6 150 g 189 4 — 1 — —

7 57 3 74 1 — — — —

8 52 3 95 1 — 1 — —

9 168 8 168 1 — 1 — —

10 82 5 155 1 — — — —

11 45 1 164 1 — — — —

12 136 5 252 1 — — — —

13 65 2 149 1 — — — —

14 234 6 304 — — — — —

15 61 — 102 — — — — —

16 63 — 102 — — — — —

17) 32 1 58 — — — — —

18 30 2 49 — — oo 1 —

19 75 4 81 — — — — —

Total 1 874 92 33(+) 4 11 6 1

* Snake and lizard frequencies based on counts from 1974 excavation only.

layer 3, the tortoise remains in layer 14 made it one of the few clearly

identifiable stratigraphic units in the deposit and in the 1976 excavation these

were removed as a separate sub-unit of layer 14. The concentration is

apparently a mound or lens for, although it was found throughout the exca-

vated area, it begins to thin out in square O 30 and probably ends in row 31 or

32. In squares O 29 and O 30, from which the tortoise ‘midden’ was excavated

as a separate sub-unit, the AD is 430 as against 304 for the layer as a whole.

Although the layer AD is not markedly higher than those for layers 12 or 3, it

is still clear that layer 14 does contain a very high frequency of tortoise remains.

Although tortoise remains are too common in southern African archaeolo-

gical deposits for their presence at BNK 1 to warrant much comment, the

‘midden’ in layer 14 does call for some discussion.

In addition to the high frequency of tortoise, layer 14 contains a high

number of mammal taxa (Table 15): 21 for a total of 29 individuals, as against

10 taxa for 18 individuals in layer 15 and 15 taxa for 24 individuals in layer 13.

R. Rau (1979 pers. comm.) has suggested that an abundance of tortoises is

more likely to be indicative of dry conditions than wet. The dense concentra-

tion of tortoise remains in layer 14, coupled with the apparent diversification of

mammal predation suggested by the high number of taxa that mostly contain

only one individual, may be indicative of a period of environmental stress. This

is difficult to substantiate, particularly when the relatively high frequency of

large medium (grazing) bovids in this layer is taken into account. These are

animals that would probably have moved out of the area in a time of drought,

although the river area may have provided a refuge. It must also be borne in

mind that layer 14 is a composite unit with a depositional history of unknown

120 ANNALS OF THE SOUTH AFRICAN MUSEUM

length and the tortoise ‘midden’ is probably a single phase representing an

accumulation during one year.

Although the total of tortoise remains in the upper layers is some 28 per

cent higher than that in the lower layers, the actual frequencies as well as the

ADs do not indicate a pattern of increase through time. Layers 19-17 have

lower ADs than most of the overlying layers and those for layers 14-9 are

consistently high but in the rest of the upper layers they fluctuate from layer to

layer.

Snake remains from the 1974 excavation were identified by G. A. McLach-

lan. Counts were based mainly on jaws and the fact that these are mostly

fragmented accounts for the uncertainty as to the actual numbers of individuals

in layers 4 and 5, which are probably higher than the totals given in Table 21.

The molesnake Pseudaspis cana is a large non-poisonous snake common in

the BNK area today and although it is very fast-moving its lack of venom

possibly accounts for its being more common in the deposits than the other

identified species, the puff-adder Bitis arietans arietans. However, Visser

(1979: 9) describes the molesnake as ‘a vicious snake, biting fiercely and

capable of inflicting serious wounds’. The puff-adder is extremely venomous

and possibly even more dangerous because its deliberate movements create a

false impression of sluggishness (Visser & Chapman 1978: 35).

That snake remains occur in the BNK 1 deposits only from layer 13 up may

be of some significance, indicating yet another aspect of the changing pattern of

resource exploitation from that of the lower layers to that of the upper. It may

also be part of this changing pattern that resulted in layers 14-9 having

consistently high frequencies of tortoise and provides a further instance of the

arbitrariness of the division of the deposit into lower and upper layer groups

because of the generally higher frequencies of most components of the artefacts

and fauna in the upper layers.

The two species of lizard occur so infrequently that beyond recording their

presence and observing that this need not be related to human activity in the

cave, no further comment seems necessary.

MICROMAMMALS: PALAEOENVIRONMENTAL INTERPRETATION

The micromammalian fauna from BNK 1 were analysed by D. M. Avery

(1979: 123-127, 174-177, 197-199). Modern comparative samples were col-

lected from the overhang below BNK 1, designated BNK 2 by Avery (Fig. 2

herein), and from an owl roost near Stanford (Fig. 1 herein).

On the basis of her analysis Avery was able to divide the BNK 1 deposits

into three major palaeoenvironmental units. The data are summarized in her

table 26, reproduced here as Table 22 with minor amendments to the radio-

carbon dates and omitting the ‘cultural units’. Microfauna from the 1976

excavation were not studied by Avery and the dates from the upper sub-units of

layer 1 are thus strictly not applicable, so have been omitted from Table 22.

BYNESKRANSKOP 1 WAI

TABLE 22

Palaeoenvironmental interpretation of BNK 1 sequence, based on analysis of

micromammalian fauna (after D. M. Avery 1979, table 26).

Dates B.P. Layer Vegetation General

1880+50 1 (lower) | UNIT 1 mildest

3 220+45 intermediate; as unit 3 but more scrub mild

3 400+55 2 and less restioid vegetation on hills

UNIT 2

3 900+ 60 >) extensive (more open ?) grass on flats; | warmer

6 rather less restioid/‘grassy’ vegetation on

7 hills and more scrub; less dense vege- moderate

8 tation near river and on lower slopes.

6370+90 9 (upper) | (relatively dry ?) harsh

6100+140 9 (lower)

6 540+ 55 10 —_—_—_—_—— changing a -

11 UNIT 3 mild

7 750+90 12 extensive dense vegetation near river and mildest

13 on lower hill slopes; extensive (more rather

9 760+85 14 closed ?) grass on flats with pans and | general harsher

155 some low scrub; restioid/‘grassy’ (and | increase in | moderate

16 proteoid ?) element on hillsides (peak | temperature

My restioid element in layer 13) harsh

ISOS 19 (relatively wet ?) cooler moderate

Perhaps the most immediately important aspect of Avery’s study has been

her recognition of a relatively marked change in the vegetation, from that of

Unit 3 to that of Unit 2, at around 6 500 B.p. As the analyses of the artefactual

and other non-artefactual components of the deposit indicate, it is at about this

time that the continuum of change appears to undergo a marked fluctuation.

That the division into lower and upper layer groups at layer 9 is somewhat

arbitrary has been stressed elsewhere in this report. The division is nowhere

precise and it is probably closer to the truth to suggest that the period during

the deposition of layers 14-9, from about 9 800—6 100 B.P., was one of greater

flux in the ecology of the human occupants of the cave than the preceding or

succeeding periods, possibly because of greater environmental change. It is to

be expected that the suggested changes in the environment were not all

simultaneous and instantaneous and that their impact would therefore be

reflected in different ways and at different times in the cave deposits.

The changes in the composition of the micromammalian fauna in Avery’s

sample that led her to separate layers 2 and 1 into a discrete vegetation unit

covering the period from about 3 400 B.p. until the final occupation of the cave,

are apparently not reflected in the artefactual and other faunal material. Such

changes as there are appear to take place in layers 4 and 3 and are of a lesser

nature than those in layers 14-9.

122 ANNALS OF THE SOUTH AFRICAN MUSEUM

The suggestion, queried by Avery herself, that the grassland of Unit 3 was

more closed than that of Unit 2, contrasts with the evidence of the larger

mammals (Table 15). It is in the lower layers that Hippotragus sp. and

Connochaetes/Alcelaphus are more common; Damaliscus dorcas is not present

in the deposits after layer 11 and Equus cf. quagga after layer 9. These animals

may be characterized as gregarious grazers with a preference for open or lightly

wooded grassland, an environment that differs from Avery’s suggestion of

‘extensive (? more closed) grassland on the flats’. There is, of course, the

problem of habitats: those of the grazers need not have been the same as those

of the micromammals that formed the prey of the raptorial birds and thus

found their way into the deposit—some of them, at least. It has been suggested

elsewhere in this report that the coastal plain between BNK and the sea was

possibly never an area of grasslands owing to the unsuitability of the soil and

that a more likely locality for extensive grasslands would have been on the

shale soils to the east and north-east of the site. G. A. McLachlan (1980 pers.

comm.) has advised that most owls are intensely territorial and are not likely to

range over a radius of more than a couple of kilometres in search of their prey.

This would bring them to the edge of the shale areas. It must be borne in mind,

however, that the BNK area is a mosaic of contrasting environments that

provide habitats for a varying range of plants and animals, and the habitats of

micromammals are relatively small.

Avery’s general climatic inferences for unit 3, indicating relatively moder-

ate to the mildest climatic conditions, except for the layer 13 period, may

provide support for the previously mentioned suggestion that the relative

paucity of artefacts in the lower layers may point to less intensive occupation of

the cave during the earlier periods of its use by humans consequent on a milder

climate being more conducive to life in the open. Against this must be placed

Avery’s tentative suggestion that it might have been relatively wet during this

period while during the period of Unit 2, when the cave appears to have been

most intensively occupied, Avery has suggested that it might have been

relatively dry. It would seem, however, that warmer and wetter conditions in

the Unit 3 period would have been more favourable for the development of

forests than of grasslands.

A point that cannot be overlooked is that two different selective agencies

contributed to the depositional process, man and the raptorial birds, the

predatory behaviour of each of which would sample different aspects of the

same ecosystem. This may to some extent explain some of the differences of

observation and interpretation.

PLANT REMAINS

A large quantity of plant remains was recovered from the deposits but for

want of a competent analyst most of it remains unsorted. The bulk of the

material is carbonized wood, which might be useful for deriving inferences

about environmental change, as has been done with material from Boomplaas

BYNESKRANSKOP 1 Ds

(H. J. Deacon 1979: 252-3). Other material such as seeds and corm cases of

geophytes could provide information about the vegetable component of the diet

of the prehistoric human inhabitants of the cave.

A certain amount of plant material remained with the non-organic and

other organic material after this had been washed. It has been analysed by

L. Horwitz and is listed in Table 23. Quantification has not been attempted

since the sample analysed is only a small portion of the total.

A problem in the analysis of plant remains from archaeological deposits is

to ascertain what proportion of the total was brought to the site by its human

occupants. The presence of three solution cavities or ‘chimneys’ in the roof of

the cave has undoubtedly been a contributory factor, and others are wind,

birds, bats, and a variety of quadrupeds as well as man.

Proteaceae grow above the cave and Sideroxylon inerme in front of it and

these might have been introduced by wind action although the woody parts

could have been used for firewood. The seeds of the exotic Acacia cyclops

found in layer 1 are probably the contribution of birds or rodents since this

plant is not yet found in the immediate vicinity of the cave or above it, although

it grows on the lower slopes of the hill and there are extensive stands of it in the

river valley and on the coastal plain (cf. Figs 2, 4).

Restio spp. and Willdenovia argentea are also components of the fynbos

above the cave and natural agencies could also account for their presence in the

deposits. Some of the Restio stalks are charred and they might have been

brought into the cave for fire-making or for bedding.

Three species of Diospyros are indigenous to the area, D. dichrophylla, D.

glabra and D. whyteana (Coates Palgrave 1977: 745-746, 753-754). The fruit of

D. dichrophylla is said to be poisonous: a common name is ‘poison-peach’. No

information is given regarding the fruit of D. glabra (cf. Smith 1966: 116) and

despite one of the common names of D. whyteana being ‘Hottentot’s cherry’

(Smith 1966: 252) information regarding the edibility of this fruit is not given.

Euclea racemosa is apparently the only species of Euclea indigenous to the area

(Jordaan 1946: 52, Coates Palgrave 1977: 739) but again there is no mention as

to the edibility of the fruit. Members of the genus Euclea are commonly known

as ‘guarri’ or ‘ghwarrie’ and the most widespread species, E. undulata, has

edible fruits. Extracts of the bark, leaves and roots of this and other species are

used for medicinal purposes (Watt & Breyer-Brandwijk 1962: 391, Smith

1966: 238-9, Coates Palgrave 1977: 741) but according to Coates Palgrave the

distribution of E. undulata does not extend west of the Breede River area, at

the eastern end of the Agulhas region (Fig. 1). The question of the introduction

of the fruits of the Ebenaceae thus remains unanswered.

The fruit of Sideroxylon inerme are said to be edible (Smith 1966: 336) but

their abundance on the trees in the BNK area and in the Cape Peninsula

suggests that either this fact is not well known or that the taste is not suited to

modern palates. The presence of the trees at the mouth of the cave could be a

contributory factor to the presence of the seeds in the deposit without requiring

the aid of man.

ANNALS OF THE SOUTH AFRICAN MUSEUM

124

(ASAG]T © UL DUUDSIIG SIJEIIPUL K AJON

x x xX x «x XGern eer ee Spoos viafijiuoU sapiowayuvskiy)

_ (aVLISOdNOO =) AVAOVAALSV

Ve ee a Eg A ch Seger eoney spas sniouD] “Jo snynantg

avaoviidunono

xt EEX x x x Kx X Xe theo Se Spoos DSOWAIDA “JO DaIINT

we Pap oh ob cb ap < spoos ‘ds soddsoig

aVaOVNadd

x x x Xe Xe EX Ke OE SX Gre aa Manes i a spoos aUldaul UOJ[AXOAAPIS

AVAOVLOdVS

x ee SOABI] pons snyy

AVAOVIGUVOVNV

x x x x Cx Sig pone Spoos psouids vypuvj xn

AVAOVIVDATOd

SE leg ca ak spoos (on0xa) sdojako ninvay

(AVSONINNDAT =) AVAOVEVA

x x x ome cee eee tes Spoos wnssaiduod uoodjoyD

AVAOVIVINVS

rina saaars aseq 1aMOoy ‘(d)ds vajo1g

5 dS ae ame ae spoos ‘(d)ds wnwuadsoonaT

x OSE a ees spoos ‘(d)ds uoipuapvonaT

AVaOVaALOUd

x Sean oe eee sare weak Spoos DIJOfIPAOD DIUAPW

AVAOVOAW

Stee the ee A SS I a LE ee PE ee ee kaa Sased WIOS ‘ds vaviow

ee ee i Bea SdSBd WIOS DIOfIsZUo] SIJO]sDXAP]

x x x SC mee fae Sased WI0 ‘Joput ‘ds 19 ‘uaH

aVaOvalal

x x oe aX Ge ea Scere Spoos DajuasIéD DIAOUAP]IIM

x ee Bin Se a ee sy]eIS -(d)ds ONSay

AVAOVNOILSAY

x Xerxes xX Kiny eennee oe sae SOABO] sisuadv) D4a]SOZ

AVAOVNOLADOWNVLOd

VLAHdOLVWaddS

x ‘'¢¢ sjuowdely ods DUIXDUL DIUO]YIT

AVaOVIAVIV

AVOTV

Claes eee Oe ier Vik eh Ch il Ole ors 8. bee OS. Ve ee oc Toke]

Sin be SS ee eee a ee eee eee eee ee

‘syisodap oy} Ul payUosoidol sUTeUIOI Jue

€7 AlavL

BYNESKRANSKOP 1 125

There can be no question, however, that the presence of Ecklonia maxima

and Zostera capensis is the result of human agency, but their function is

problematic. There are, as far as is known, no early ethnographic records of the

use of kelp (E. maxima) by the Khoisan for either food or medicinal purposes.

Budack (1977: 37) reports, however, on the modern use by the Topnaar

Hottentots of South West Africa of powdered EF. maxima and Laminaria

schinzii on wounds, noting that both are rich in iodine. In the entry for 6 March

1654 in the Journal of Jan van Riebeeck there is a reference to the use by the

Hottentots of ‘trombas’ for the storage of train-oil rendered from the blubber

of a dead whale (Moodie 1838: 46, Van Riebeeck 1651-5: 205, Thom ed.

1952: 218). “Irombas’ (trumpets) is the old Portuguese name for E. maxima.

The long, hollow and distally swollen stipe could also have been used for

carrying water, and a modern use as a ‘pressure-cooker’ for cooking Haliotis

midae in the coals of a barbecue pit (A. E. Louw 1980 pers. comm.) could

possibly represent the continuation into modern times of a tradition derived

from the Khoisan. C. E. Labuschagne (1980 pers. comm.) has advised that the

sliced stipes of E. maxima make a palatable if glutinous foodstuff when boiled

and can be used as a thickener for stews. Such a use would require a cooking

vessel, but it cannot be assumed that the recovery of remains of E. maxima

only from layer 1 is related to the introduction of pottery into the area as

preservation factors may also be involved. In the collection from Rooiels Cave

near Cape Hangklip which was excavated in 1921-2 by A. D. Divine (South

African Museum collection AA4029-80) is a knotted piece of kelp stipe.

Although intrinsically interesting this artefact throws little light on the possible

uses of this marine plant.

The estuarine grass Zostera capensis is present in small quantities, and

probably in more of the layers than indicated in Table 23. It occurs in shreds or

small wads not more than about 5 x 5 x 0,5 cm greatest dimension and never in

quantities sufficient to indicate its having been used as a bedding material. Such

a use was suggested by the presence of apparently larger quantities at Oak-

hurst, Glentyre and Matjes River (Sampson 1974: 308; cf. Goodwin 1938: 236,

238 re Oakhurst). There are no known records, early or modern, of the use of

this plant for food or medicine and its presence in the BNK 1 deposits in such

small quantities cannot be satisfactorily explained.

The waxberry Myrica cordifolia is a shrub of the coastal dunes rather than

of the coastal plain. Its wax-covered berries are eaten by starlings (Sturni-

idae—P. Fairall 1980 pers. comm.) which nest in and near BNK 1 and could

therefore be considered as possible contributors of the seeds. Thunberg

(1795: 167), observing that the wax was removed by boiling the berries,

commented that ‘the farmers use it for candles, but the Hottentots eat it like a

piece of bread, either with or without meat’. The berries contain, inter alia, an

unsaturated oil (Watt & Breyer-Brandwijk 1962: 785) and the fruit may thus

have had a dietary value beyond that of just the wax. Coates Palgrave

(1977: 95) mentions that the ‘wax’ coating of M. serrata is, in fact, not a wax

but a true fat rich in unsaturates; and the same may be true of M. cordifolia.

126 ANNALS OF THE SOUTH AFRICAN MUSEUM

All the other plants listed in Table 23 that have not already been discussed

have edible fruits or, in the case of the Iridaceae, corms. It seems probable that

the Iridaceae were brought to the site by primates, baboons if not man. Not all

species of Moraea are edible (Watt & Breyer-Brandwijk 1962: 505 ff., Smith

1966: 470) but it may be assumed that the species present in the BNK 1 sample

are edible as there is no record of the toxic substances in the inedible species

having been used as a poison by the Khoisan. A limited survey of the area in

the vicinity of the cave showed that Iridaceae are common but not abundant on

the Table Mountain Sandstone hill slopes adjacent to BNK. Three genera could

be identified in the field: Chasmanthe (cf. aethiopica: Maytham Kidd

1973: 19.5), Gladiolus (cf. alatus: Maytham Kidd 1973: 59.1), and Aristaea (cf.

macrocarpa: Maytham Kidd 1973: 79.2). Melasphaerula ramosa (Maytham

Kidd 1973: 62.7) grows on rocky outcrops in the shade of the trees in front of

the cave.

All the berry-bearing plants listed in Table 23 are to be found in the

vicinity of the cave and there are others with hard seeds, such as Carissa

bispinosa and Olea capensis that grow near the cave but have not been

identified in the samples. Rhus glauca, of which only leaves are present in layer

1, has edible drupes (Smith 1966: 308). Smith also records that several species

of Rhus (korentebos) have ‘strong resilient branches which were formerly used

by Bushmen and Hottentots for making their bows’.

An unusual feature was the recovery from layer 15 of sixteen well-

preserved seed skins of a member of the family Cucurbitaceae. As indicated in

Figure 43, these all show marks of having been gnawed. T. Arnold (in litt. 7

November 1980) has advised that the seed skins have been identified as being

from a species of Citrullus, all the modern cultivars of which are presently

referred to C. lJanatus. Arnold is of the opinion that the skins cannot be

contemporary with the deposit in layer 15 and points out that microscopic

examination of the skins indicates that they have been gnawed by a rodent to

Fig. 43. Cucurbit (Citrullus sp.) seed skins from layer 15 showing evidence of having been

gnawed by rodents.

BYNESKRANSKOP 1 127

obtain the endosperm. It would seem advisable in the circumstances to regard

these seed skins as intrusive into layer 15.

Although the plant remains represented in Table 23 may be considered a

random sample in that they represent non-deliberate remnants in the excavated

material after it had been washed, because the sample is small in relation to the

total amount of plant remains present in the deposit, it is not appropriate here

to attempt to derive any environmental or human ecological inferences from

the remains studied, save on a very general level.

Coates Palgrave (1977: 722) describes Sideroxylon inerme as ‘almost

always a tree of coastal woodland and littoral forest’ and evidence of its

presence from at least layer 17 up suggests that there has been no major

environmental change in the last 10 000 years or so that has affected the tree’s

viability in the area.

One of the common names for Nylandtia (Mundia) spinosa is ‘duinebessie’

(dune berry: Smith 1966: 205) and the presence of the seeds of this plant, also

from layer 17 up, attests to the relatively unchanged sandy nature of the coastal

plain. It should be mentioned, however, that while N. spinosa is abundant on

the coastal plain and in the river valley, it generally grows. there as a small

shrub up to 1 m in height, while on the limestone ridge of BNK, where it forms

a not insignificant part of the fynbos, it grows to a height of over 2 m.

Euclea racemosa is another plant of the coastal zone, found as a shrub in

the sandy soil but growing to tree height (6 m) in coastal forests (Jordaan

1946: 52, Coates Palgrave 1977: 739).

Chrysanthemoides monilifera is adapted to a wide range of habitats, from

coastal dunes to rocky hillsides and the verges of evergreen forests (Coates

Palgrave 1977: 913) and therefore cannot be used as a sensitive environmental

indicator.

In the Cape Peninsula Myrica cordifolia flowers in the winter (April—July),

Nylandtia spinosa in the winter and spring (April-October), Rhus glauca is

spring-flowering (August-September), Chrysanthemoides monilifera flowers

throughout the year but mainly from spring to early winter (October—May),

and Colpoon compressum from summer to winter (December—June) (Maytham

Kidd 1973). Even allowing for localized variation in the flowering and thus

fruiting season, the presence of the seeds of these plants offers no clear

indication of the season(s) during which the cave was occupied, even if it is

assumed that all the seeds were brought into the cave by its human occupants.

A more precise, quantitative analysis may, however, reveal that there were

changes in the seasons during which the cave was occupied in different times

during its occupation history.

Iridaceae are generally spring to summer flowering and their corms may be

assumed to reach their maximum growth during the period of the winter rains

and before flowering commences. In the BNK area, as has been mentioned,

Iridaceae are common but not abundant. Moreover, the corms of these plants

seem mostly to be small and firmly wedged into crevices in the rock. It was

128 ANNALS OF THE SOUTH AFRICAN MUSEUM

found almost impossible to remove these intact, even using a metal screw-

driver. This suggests that a thicker and more easily broken implement such as a

wooden digging stick would have been ineffectual, although the presence of

corm cases in the BNK 1 deposits testifies to a certain success in collecting

them. However, the lack of abundance of these geophytes as well as the

difficulty of getting them out of the ground suggests that they might not have

been an important part of the diet of the human occupants of BNK 1, or not to

the extent that has been suggested for the occupants of Melkhoutboom (H. J.

Deacon 1976: 162, 174).

Although Olea africana or O. capensis have not been identified in the

BNK 1 sample, a phenomenon observed during the 1980 field season at BNK 3

during the first two weeks of July is worth recording. On the Die Kelders coast,

which faces north, low shrubs of O. capensis 1,0-1,5 m in height were found to

be heavily laden with ripe fruit, especially where the shrubs were protected

from the seasonally prevailing north-west wind. At BNK, however, where O.

capensis grows into sturdy trees 3 m or more in height, most of the trees had

not even begun to come into flower and only one small shrub was found, near

the top of the hill, that had a few ripening fruit.

While the Die Kelders coast receives full exposure to the available sun-

shine, at this time of year (mid-winter) the southerly aspect of BNK receives

only an hour or two a day, in the early morning or late afternoon, and this is

undoubtedly the reason for the differences in the seasonal development of this

species in the two areas. In view of the edibility of this fruit (Coates Palgrave

1977: 759), and assuming that it is not a recent immigrant into the area, it is

somewhat surprising that its seeds were not found in the deposits at Die

Kelders. There was a conspicuous absence of plant remains from the Die

Kelders deposits (Schweitzer 1979: 206-7), and the presence of seeds of O.

capensis, evidently available during the winter, would have provided supportive

evidence for the hypothesis that the site was occupied during the winter months

(Schweitzer 1979: 219). On the other hand, the presence of seeds of O.

capensis in the BNK 1 deposits, should they be identified when the bulk of the

plant remains are analysed, need not indicate spring-summer occupation to

coincide with the ripening of the fruit in the vicinity of the cave as the fruit

could have been collected from the Walker Bay coast during the winter as well

as from BNK later in the year.

In an early account of the Hottentots of the Cape, Grevenbroek

(1695: 185) recorded that it was the work of the men to prepare the winter’s

supply of plant food. Grevenbroek mentions ‘wild almonds’ (Brabeium stellati-

folium (growing in the Bredasdorp district according to Coates Palgrave

(1977: 123) but not observed in the BNK area)), the ‘larger arum’ (probably

the rhizome of Zantedeschia aethiopica, growing in the Uilkraals River valley)

and ‘various bulbs’ (probably including Iridaecae). These were exposed to the

sun, roasted on a small fire and then stored ‘in ditches and caves’. This

observation carries several interesting implications regarding seasonality, espe-

BYNESKRANSKOP 1 129

cially with regard to the occupation of caves, but has no direct relevance to

BNK 1 since, apart from the corm cases of ‘various small bulbs’, no other

indications were found of the preparations described by Grevenbroek, nor were

any storage pits found at the site such as those discovered at Boomplaas (H. J.

Deacon et al. 1978: 55). It seems likely, however, that most of the edible fruits

whose remains are recorded from the BNK 1 deposits would have been eaten

at the time they were collected since most, if not all, would not have been

amenable to any process of preservation or storage.

SUMMARY

Mammals, excluding Homo sapiens and micromammals, were divided into

six size classes based on their live mass. Perhaps the most significant changes to

be observed are in the large medium and very small size classes. The large

medium size class, which consists mostly of bovids, shows a gradually increasing

trend from layer 19 to layer 13, in which the frequency reaches almost 46 per

cent of the layer total. From layer 12 up frequencies decline, accounting for less

than 10 per cent of the layer total in layer 9, and in layers 8-1 the frequency

exceeds 5 per cent only in layers 3 and 2. |

The very small size class, in which there are no bovids, shows a generally

declining trend in layers 17-10, after which frequencies increase, reaching a

peak of 53,5 per cent of the layer 4 total. In layers 8-1 the frequency falls below

40 per cent of the layer total only in layers 7 and 3, whereas in layers 19-9 it

reaches this frequency only in layer 17.

It is, however, only in layers 15 and 13 that the three larger size classes

account for more than 50 per cent of the layer total, and in layers 6—1 the three

smaller size classes account for over 80 per cent of layer totals.

There is one main period only, that from layer 14 to layer 9, in which

bovids are numerically more common than all other animals. Because there has

been no separation of the sub-units of layer 1 and because this layer includes a

basal sub-unit in which there are no sheep remains, it has not been possible to

determine the effect of the introduction into the area of domestic animals on

the economy of the cave occupants beyond observing that layer 1 has the

highest frequency of small medium animals of all the layers, and that almost

two-thirds of these are Ovis aries.

Shellfish frequencies are generally very low in layers 19-10 and account for

only 1,3 per cent of the site total. Relative frequencies (ADs) show a generally

increasing trend from layer 12, however, up to layer 2 with a slight decline in

layer 1. Layer 3 has an anomalously high frequency, with an AD a hundred

times that of layer 12, more than ten times that of layer 9, and three to four

times those of layers 6-4 and 2-1.

In layers 16-11 Donax serra is the most commonly represented species but

Turbo sarmaticus is always the most common species in the upper layers in

terms of flesh mass, although it declines numerically in layers 6-3, in which

Oxystele sinensis becomes more common.

130 ANNALS OF THE SOUTH AFRICAN MUSEUM

An analysis of the lengths of five species of Patella and comparison of four of

these with data for modern populations shows that it was generally the

larger-sized individuals that were being collected, except in the case of P.

cochlear, which was collected mainly in the middle of its population size range. It

was concluded that despite evidence for changes in the mean lengths of some

species in successive layers, the periods of time involved, compared to the

longevity of the species, do not allow any conclusions to be reached regarding the

effect of human predation on the shellfish populations; nor was it considered

possible to extrapolate inferences regarding changes in mean sea temperature.

Fish are sparsely represented in the lower layers, with some 95 per cent of

the site total in layers 9-1. The ADs indicate a gradually increasing trend in

layers 9-5 and a decline thereafter.

Reptiles are represented by two species each of tortoise, snake and lizard.

Layer 14 contained what may be termed a ‘tortoise midden’ but the ADs reveal

no clear trends of increase or decrease, suggesting that the collection of these

animals was not in any way affected by the changes in the representation of the

mammal size classes. Remains of snake are present only from layer 13 up (in

the sample from the 1974 excavation) and while this may have some human

ecological significance, frequencies are too low to allow more than the sugges-

tion of such a possibility.

Micromammals from the deposit were analysed by D. M. Avery (1979),

who has used the data derived to suggest palaeoenvironmental changes

throughout the period of the occupation of the cave, with three major vegeta-

tion units represented, those of layers 19-10, 9-3, and 2-1. Avery’s interpreta-

tions contrast to some extent with those that can be derived from the analysis of

the artefacts and the other fauna, but this may simply be a reflection of the fact

that two very different types of predator are involved.

Plant remains have for the most part not been examined. The small sample

that has been studied shows it to consist of species present in the area today but

does not allow the deriving of environmental or human ecological inferences.

In general, the faunal evidence indicates a change in exploitation patterns

that began about 7 500 years ago. Although there is evidence of exploitation of

marine resources from the initial occupation of the cave, this only became

marked from about 6 000 B.P. and represents a substantial addition to the diet

of the cave occupants at a time when the pattern of procurement of land game

indicates a decrease in the amount of flesh being brought back to the cave. This

suggests a change in the ecology of the human occupants of BNK 1, but

whether the cause was demographic, cultural, or environmental, or a combina-

tion of these, is not directly deducible from the site data alone.

BNK 1 IN RELATION TO OTHER SITES

The artefactual and faunal samples from BNK 1 provide a useful body of

information that may be compared with data from sites of comparable age in

the adjacent regions of the Cape. In this way the BNK 1 sequence can be

BYNESKRANSKOP 1 WA

‘placed’ in the industrial sequence already established for the area, and a

contribution made towards the assessment of variability within the Late Stone

Age.

Of particular importance is Nelson Bay Cave, since the material from this

site has provided the basis for the first comprehensive description, and conse-

quently definition, of the artefactual components of the ‘pre-Wilton’ industries

that have been termed ‘Albany’ and ‘Robberg’ at Nelson Bay Cave (Klein

1974, J. Deacon 1978), and Melkhoutboom (H. J. Deacon 1976).

Problems in the comparisons arise from the fact that sites such as Die

Kelders and Wilton contain only part of the post-Pleistocene sequence while

others, such as Nelson Bay Cave, Melkhoutboom, Elands Bay Cave and

probably also Wilton, have sequences that are interrupted by periods of

non-occupation. The Nelson Bay Cave sequence studied by J. Deacon (1978)

also lacks the final occupation sequence, from about 5 000 B.P., excavated by

R. R. Inskeep and yet to be published. In some cases, too, the series of

radiocarbon dates, for BNK 1 as well as other sites, are insufficient to allow

precise correlation of stratigraphic units and their contents. Comparisons are

also limited by differences in data presentation or because comprehensive site

reports have yet to be published.

For convenience, artefactual and faunal categories and classes are dis-

cussed in the same order as in the analysis of the BNK 1 material. These

divisions are, of course, somewhat arbitrary and the overall character of the

assemblages is also important.

STONE ARTEFACTS

Raw materials

In the BNK 1 stone artefact sample certain trends in raw material fre-

quencies may be observed that allow the layers to be grouped on this basis

(Fig. 7, Table 26). The trends, in which quartz, quartzite and silcrete are

variously the predominant raw material, are chiefly influenced by the

unmodified artefact category and even within this category there are differences

in the predominance of the raw materials for classes such as unmodified flakes

(Table 26). Silcrete is the most common raw material in the utilized and

retouched artefact categories, except in a few cases in the lower layers in which

artefacts in these categories occur only in low or minimal frequencies. Although

the overall character of the artefact sample is useful for comparative purposes,

the bias introduced by classes such as chips and chunks, which will have had

little or no potential for use, suggests that it is probably more profitable to

compare raw material frequencies in specific categories, e.g. utilized—modified

and retouched, or classes, e.g. unmodified flakes and blades, cores and the two

categories of artefacts that can be shown to have been used.

BNK 1 layer 19, and possibly also layer 18, falls within the time range of

Nelson Bay Cave layer BSL, the uppermost of the three “Robberg’ layers,

layers YSL and YGL being in the time range of 16 000-18 000 B.p. (Klein

£52 ANNALS OF THE SOUTH AFRICAN MUSEUM

1972a: 202; J. Deacon 1978: 89-90, fig. 3). In the BNK 1 layers quartz is the

predominant raw material, as it is in Nelson Bay Cave layers YGL and YSL,

while quartzite is predominant in BSL and in all the overlying layers (J. Deacon

1978, table 2). There is an increase in the frequency of quartz in the ‘Wilton’

layers but this does not reach the frequencies of the ‘Robberg’ layers. Fre-

quencies of silcrete are low in all layers although in layer YSL it is marginally

(1,1%) more common than quartzite. Quartzite predominates only in BNK 1

layers 17-15, while silcrete is most common in layers 9, 8, 6, and 5, and quartz

predominates in all the other layers.

In the retouched artefact category silcrete is absent from all the ‘Albany’

layers at Nelson Bay Cave and does not reach 25 per cent of the category total

in any of the other layers. Quartzite is generally the most common raw material

in the ‘Albany’ layers (the two artefacts in layer GSL are of quartzite and

chalcedony) and quartz or chalcedony in the ‘Wilton’.

The ‘Robberg’ industry of the Basal Unit at Melkhoutboom is probably at

least 1 000 years older than BNK 1 layer 19, and is separated from the ‘Albany’

deposits by a hiatus covering the period 14 000-10 500 B.p. (H. J. Deacon 1976,

table 3). Deacon (tables 8-9, 12) provides raw material frequencies for specific

classes only. Combination of the data in his tables 8 and 9, for chips, chunks

and unmodified flakes, indicates that quartzite is the most common raw

material except in the lower ‘Wilton’ units WBM and M in which chalcedony

predominates, and in the uppermost ‘Wilton’ unit OMB in which silcrete

predominates. Quartzite is the most common raw material for unmodified

flakes in all the layers except OMB in which silcrete is most common. For the

four classes of retouched artefacts in Deacon’s table 12 chalcedony is the most

common raw material in the four lower ‘Wilton’ units WBM-—MB, and silcrete

in the two upper, CAF and OMB. There were apparently no ‘formal tools’ in

the ‘Albany’ and ‘Robberg’ deposits (H. J. Deacon 1976, tables 10-12).

The Wilton Large Rock Shelter has yielded dates obtained from charcoal

in layers 2B and 3F that show these deposits to span the period from about

2 270-4 860 B.p. A third date of about 8 620 B.P. was obtained from bone from

a human burial excavated by Hewitt in 1921 and considered by J. Deacon

(1972: 14) to relate to the base of layer 3. This date is problematic when related

to the stratigraphy (J. Deacon 1972, fig. 3) and suggests a hiatus in the

deposition, by inference between layers 3F and 3J, but more probably between

3J and 4A. Thus, while it is possible to relate the Wilton deposits down to layer

3F to those from about the middle of BNK 1 layer 1 to somewhere between

layers 5 and 9, the Wilton layers below 3F cannot be satisfactorily correlated

with the BNK 1 sequence, although H. J. Deacon (1976, table 3) has assigned

layers 3G-—3I to the ‘Formative Wilton’, dated to about 6 000-8 000 B.p. and

layers 4A and 4B to the ‘Albany’, dated to about 8 000-11 000 B.P.

J. Deacon (1972, figs 4-5) provides frequency diagrams for raw materials

in selected artefact classes only. The indications are, however, that in the

‘waste’ classes there were higher frequencies of quartzite and quartz than other

BYNESKRANSKOP 1 133

raw materials in layers 4B and 4A. Silcrete appears to have been increasingly

used throughout the rest of the sequence, as well as being the most commonly

used raw material for retouched artefacts, although more scrapers in layers

4B-3H were chalcedony than silcrete.

The raw material frequencies for unmodified flakes from Buffelskloof

(Opperman 1978, table 4b) show quartz to have been the most common raw

material for this class throughout the whole sequence. Chalcedony ranks

second in the upper part of the sequence, layers BOL1—CH2, quartzite in the

middle, layers CH3—ZJ2, and ‘other’ (predominantly hornfels) at the bottom,

layers ZJ3—HE2. Quartz and chalcedony are generally the raw materials most

commonly used for retouched artefacts (Opperman 1978, fig. 6). The Buffels-

kloof sequence covers much the same period of time as BNK 1 layers 19-5,

although layer HE2 has a basal date in excess of 22 000 B.Pp. (Opperman

1978: 21).

Unmodified artefacts

Flakes from BNK 1 account for 21,2 per cent of the unmodified artefact

category (Table 4), while in the Nelson Bay Cave sample (J. Deacon 1978,

table 1) frequencies range from about 86 per cent in the ‘Robberg’ to almost 92

per cent in the ‘Wilton’. A partial explanation of the difference is that Deacon

(1978: 91) has included the class of chips in the untrimmed flake class. How-

ever, even if chips are added to the unmodified flakes in the BNK 1 sample

they still account for only 76,1 per cent of the unmodified artefact category

total, well below the total for any of the Nelson Bay Cave industrial units. A

possible reason for these differences is the high frequency of quartzite at

Nelson Bay Cave, a material that is predominant at BNK 1 in layers 17-15

only.

Blades, in recent analyses of stone artefacts (e.g. J. Deacon 1972, 1978; H.

J. Deacon 1976; H. J. Deacon et al. 1978; Opperman 1978), have not been

separated from flakes in the unmodified artefact category, so that it is not

possible to determine whether the BNK 1 distribution is similar to that at other

sites. This is unfortunate since one of the characteristics of the “Robberg

industry’ is a high frequency of quartz micro-blade cores and it would have

been useful to have been able to compare core: blade ratios, especially in

samples from the ‘Robberg’.

Cores in the Nelson Bay Cave sample (J. Deacon 1978, table 1) were more

common in the ‘Albany’ layers than in the ‘Robberg’ or ‘Wilton’ and bladelet

cores more common in the ‘Robberg’ than in other layers (Deacon’s term

‘bladelet’ equates the use in this report of ‘micro-blade’). The first instance is

unlike the situation at BNK 1, where 63,2 per cent of the cores came from

layers 1-9, but the second is similar, allowing that Nelson Bay Cave layer BSL

and BNK 1 layer 19 are approximately of the same age. Blade cores were

recovered only from the upper part of the Buffelskloof sequence (Opperman

1978, table 3).

134 ANNALS OF THE SOUTH AFRICAN MUSEUM

Utilized and modified artefacts

Utilized flakes in the Nelson Bay Cave sample (J. Deacon 1978, table 1)

include notched flakes, which are included in the retouched artefact category in

the BNK 1 analysis. Even when notched flakes are excluded, there are more

than twice as many utilized flakes in the Nelson Bay Cave sample than in that

from BNK 1. Using totals that exclude notched flakes, utilized flakes account

for 58,6 per cent of the utilized artefact category and are relatively most

common in the ‘Robberg’, the percentage frequency decreasing by about 10 per

cent in the ‘Albany’ and again in the ‘Wilton’. In the BNK 1 sample, utilized

flakes account for only 42,8 per cent of the utilized and modified artefact

category and are relatively more common in layers 17-13 (57,4% of the

category total) than in layers 19-18 (29,7 %) or layers 9-1 (48,8 %, or 54,4 % in

layers 9-7, which may be the layers chronologically comparable with that part

of the Nelson Bay Cave ‘Wilton’ studied by Deacon).

In the Melkhoutboom sample (H. J. Deacon 1976, table 9) ‘retouched

flakes’ are most common in the W and M units of the ‘Wilton’, numerically as

well as on an AD basis.

‘Trimmed flakes’ were included in the ‘waste’ category in the Wilton

inventory (J. Deacon 1972, table 1). The total of these, 1 367, is considerably

higher than the BNK 1 total of 592 utilized flakes. The highest frequencies are

in Wilton layers 3B-3G and 4A and in BNK 1 layers 9-5 (except layer 7), so

that the situation at both sites is broadly similar.

Opperman (1978, table 2) has also included ‘trimmed flakes’ in the ‘waste’

category in the Buffelskloof inventory. Frequencies are generally low through-

out the sequence and even the highest frequencies, in layers BOL2 and CH3,

seem merely to reflect the fact that these layers have the highest frequencies of

stone artefacts. Layer BOL2 is younger than BNK 1 layer 9, and layer CH3

younger than layer 12.

Utilized blades have not been listed separately in the inventories for any of

the sites used for the inter-site comparisons. It may be that, as in the case with

the BNK 1 sample, they are relatively few.

Scaled pieces in the Nelson Bay Cave inventory (J. Deacon 1978, table 1)

occur in very low frequencies in layers BSL—RA, and 83 of the site total of 97

were recovered from the uppermost ‘Wilton’ layers BSC and IC. There is thus

a general similarity in the distribution of these artefacts at Nelson Bay Cave

and BNK 1, although layer 18 has a higher frequency (percentage as well as

actual) than layer BSL.

These artefacts are also uncommon in the Melkhoutboom stone artefact

sample (H. J. Deacon 1976, table 7). There were none in the ‘Robberg’ and

‘Albany’ deposits, and the highest frequency came from the W unit, which is

about the same age as BNK 1 layer 9.

In the Wilton inventory (J. Deacon 1972, table 1) piéces esquillées were

included with ‘other tools’, there being no category of utilized artefacts. They

are not common in any of the layers and the highest frequencies (12 and 13)

BYNESKRANSKOP 1 135

were in layers 3D and 3E, which may be somewhat older than BNK 1 layers

7-5, in which the latter site’s highest frequencies, occurred.

There were apparently no scaled pieces found at Buffelskloof (Opperman

1978, table 2).

Grindstones are surprisingly rare in the BNK 1 stone assemblage and the

three that were recovered are classified as combination hammerstones—upper

grindstones (Table 6). Die Kelders yielded 87 (Schweitzer 1979, table 17) and

Nelson Bay Cave 44 (J. Deacon 1978, table 1), although these sites have

smaller total stone artefact samples than BNK 1. Low frequencies of grind-

stones are also recorded from Wilton (J. Deacon 1972, table 1) and Melkhout-

boom (H. J. Deacon 1976, table 13), and none at all appear to have been found

at Buffelskloof (Opperman 1978, table 2).

In a study of sites in the Clanwilliam district, Mazel (1978, Appendix 3)

records variable but usually low or nil frequencies in the western Cape sandveld

or coastal plain, one exception being the Verlore site, which is closest to the

coast of all the sites studied by Mazel. BNK 1 is, however, almost as close to

the coast as Verlore, and the recovery of 17 upper grindstones (‘rubbers’) from

the late Holocene DGL member at Boomplaas (H. J. Deacon et al. 1978, table

2) tends to discount the suggestion that the use of these artefacts might have

been more common at sites at or near the coast.

Bored stones, even in the fragmentary form in which they were found in

the BNK 1 deposits, by their presence in a site on the edge of the coastal plain

are not consistent with the suggestion by Mazel (1978: 79) that such artefacts

are not to be found on the coastal plain because weighted digging sticks are

unimportant away from the mountains. What seems more likely, especially

where surface exposures are concerned, is that these artefacts have proved

attractive to casual collectors. Reference to the monograph by Goodwin (1947)

on the bored stones of South Africa will reveal that these artefacts are not—or

have not been—absent from sites at the coast or on the coastal plain.

Heavy edge-flaked pieces were sporadically distributed through the BNK 1

deposit, whereas at Nelson Bay Cave they were found in almost every level

(J. Deacon 1978, table 1). At BNK 1 they account for only 0,7 per cent of the

utilized and modified artefact category total, while at Die Kelders they com-

prise 2,7 per cent of the utilized artefact category total. At Nelson Bay Cave

they account for 1,2 per cent of the total for the utilized artefact category in the

‘Robberg’ layers but increase to 23,3 per cent of the ‘Albany’ total and 24,6 per

cent of the ‘Wilton’ total. Whatever their function, heavy edge-flaked pieces

were clearly a more important component of the Nelson Bay Cave artefact

assemblage than of those of BNK 1 or Die Kelders.

Retouched artefacts

Scrapers were of particular interest in the analysis of the BNK 1 stone

artefact assemblage. This interest was stimulated by the evident difference of

the scrapers from the middle layers from those of the upper. On the basis of the

136 ANNALS OF THE SOUTH AFRICAN MUSEUM

analysis of the material from the 1974 excavation it had been assumed that

layer 9 marked the beginning of the ‘Wilton’ tradition at the site, and it was

recalled that Goodwin (1938) had stated that, underlying the ‘Wilton’ deposits

at Oakhurst, there were deposits containing ‘Smithfield’ artefacts.

According to Goodwin (1938: 313, 314) these deposits contained ‘vast

numbers of typical Smithfield scrapers’, said to have exceeded 5 000 in the

‘Smithfield C’ levels alone. The fate of the bulk of this large collection is

unknown. A small, representative collection is retained by the Archaeology

Department of the University of Cape Town but the major part of the

University’s holding was donated in 1961 to the South African Museum

(AA6990) by R. R. Inskeep on behalf of the University.

In view of the evident differences in the size and shape of the scrapers

from the layers above and below layer 9 at BNK 1 (Fig. 17) it was decided to

analyse the Oakhurst ‘Smithfield’ scrapers in the Museum’s collection to see

how these compared with those from BNK 1 layers 9-13, those from layers

14-19 being considered too few to warrant inclusion.

The Oakhurst material had previously been analysed by Fagan (1960) and

Schrire (1962). The differences of opinion or interpretation that exist as a result

of these analyses are outside the scope of the present report except in so far as

they concern the reliability of the Museum’s collection as being representative

of the Oakhurst scraper sample as it originally existed. The material is still

stored in the bags in which it was received from the University and these bear

the excavation level references but, since the artefacts themselves have not

been marked, it is no longer possible to determine whether they are still

correctly bagged.

Fagan (1960, table 3) worked on a ‘Smithfield C’ scraper sample of 720 and

a ‘Smithfield B’ sample of 274. Schrire (1962: 181, table 1), who claimed to

have had access to material not seen by Fagan, recorded a total of only 496

‘Smithfield C’ scrapers but 354 of ‘Smithfield B’. Schrire’s reportedly aug-

mented sample was thus actually some 17 per cent smaller than Fagan’s; her

‘Smithfield C’ sample was about 31 per cent smaller than Fagan’s, while her

‘Smithfield B’ sample was 29 per cent larger. This raises the question of

whether the material was mixed subsequent to Fagan’s analysis and prior to

Schrire’s.

There are also problems regarding the distribution of scraper sizes in the

Oakhurst samples. These may result from mixing or they may relate to

different methods of measurement used by Fagan and Schrire and for the

present analysis. For the present analysis, determination of the plane dimen-

sions followed the method of J. Deacon (1972, fig. 6) used for the measuring of

scrapers from Wilton. Length is the dimension along an axis taken from the

centre of the retouched edge, and width the dimension along the axis at right

angles to the length. Thickness is the greatest vertical distance between the

dorsal and ventral surfaces of the scraper.

Fagan (1960, table 3) recorded no ‘Smithfield C’ scrapers less than 15 mm

or more than 40 mm long and Schrire (1962, table 1) recorded no ‘Smithfield B’

BYNESKRANSKOP 1 137

scrapers less than 3in. (approx. 13 mm) long, but in the present analysis

scrapers in these length ranges were found (Fig. 44).

It was found that the scrapers from the top and bottom of Terrace 3, the

‘Smithfield B’ deposit, have been kept separately bagged, and as these

appeared to be different in size they were treated as discrete samples. There

were only 42 scrapers from the top of Terrace 3 and 24 from the bottom, and it

was therefore decided that only one bag of scrapers, numbering 101, would

suffice for the ‘Smithfield C’ sample from Terrace 2. Examination of other

scrapers from this level indicated that the sample chosen was representative.

The results of the metrical analysis of the Oakhurst ‘Smithfield’ scrapers

are presented diagrammatically in Figure 44. These indicate a continuous

decrease in size from the bottom of Terrace 3 (‘Smithfield B’) to Terrace 2

(‘Smithfield C’) but no change in the mean shape (length : width ratio).

Statistical comparison of the samples from BNK 1 layers 9-13 with those

from the Oakhurst ‘Smithfield’ was by means of the Kolmogorov—Smirnov

two-tailed test (Siegel 1956: 127-136) carried out on each pair of samples and

for each of the four attributes. Three of the samples, those from BNK 1 layers

11 and 13 and that from the bottom of Oakhurst Terrace 3, have frequencies

below the minimum of forty specified for the test. However, application of the

one-tailed test, for which no minimum is stipulated, in most cases gave the

same results as the two-tailed tests. Where there were differences, these were

of the order of one level of significance, in the direction of greater significant

difference.

Because of the generally low frequencies in each sample, especially when

these are distributed over the whole range, frequencies in each of the attribute

classes were grouped. Class intervals of 5 were used for length and width and 3

for thickness and length : width ratio.

The results of the two-tailed tests are given in Table 24. It is immediately

apparent that the layer 9 sample is sufficiently different from all the other

samples, except that from layer 13, to warrant its exclusion from further

consideration at this point. The layer 10 sample is not significantly different

from the ‘Smithfield C’ sample, and the layer 11 sample differs significantly

from the ‘Smithfield B’ samples only in width. The layer 12 sample is more like

that from the top of the ‘Smithfield B’, also differing significantly only in width.

The layer 13 sample is not significantly different from the ‘Smithfield C’ and

differs from the sample from the top of the ‘Smithfield B’ only at a low level of

significance, again in width. The difference in width is, however, very highly

significant when the layer 13 sample is compared with that from the bottom of

the ‘Smithfield B’. In general the ‘Smithfield B’ scrapers are significantly wider

than those from BNK 1 layers 10-13, a fact that is borne out by the Dice-

Leraas diagrams in Figures 17 and 44. That the Oakhurst scrapers are predomi-

nantly made of quartz while those from BNK 1 are silcrete cannot be accepted

as a Satisfactory single explanation, and since the relevant deposits are believed

to be approximately coeval, another explanation, possibly cultural, must be

sought.

ANNALS OF THE SOUTH AFRICAN MUSEUM

138

10l

“Ou

olyo4 = Yyypim:4yBuay

"2 oINSI 99S ‘sjoquiks Jo UOTeULIdxd 104 “dZIS [dues dy) 9}edIPUT 1YsII 9Y} UO SIOqUINN]

‘POUIQUIOD Udy) ‘A][ENpPIAIpUL UMOYS JsIY o1e ¢ dRIIO], JO WO}0q pue doy oy} WoT safdures oY} OJ IVP OUL “.G PlOYYIWS, =¢ sPLAL © O Ply

-YJIUS, = Z dOVIOT, “ISINYYLO 1” S]oAg] ,P[EyYIWS, oy) Wosy s1odesos Jo sone YIPIM : YISUI] Puk SUOISUSWIP Jo SWILISLIP SeeIAT—VIC POYIPON ‘py “SIA

SSUHxIIY |

(HIP)

(wo44oq)

(ee

aoD149)

€

BYNESKRANSKOP 1 139

TABLE 24

Stone: matrix of results of Kolmogorov—Smirnov tests on paired samples of scrapers from

BNK 1 and Oakhurst.

Layer

10 VHS HS

VHS O

11 VHS HS HS O

VHS VHS O LS

12 VHS VHS S O

VHS VHS O O

13 VHS O O O

S O O O

SC ae - O O

O O

SB(t) VHS se e ae HS 3 wee LS LS

VHS IES O S O

SB(b)

VHS VHS |} VHS’ VHS

VHS O S O

Note. The numbered layers are from BNK 1, those with letters are from Oakhurst. SC =

‘Smithfield C’ (Terrace 2); SB(t) = ‘Smithfield B’ (top of Terrace 3); SB(b) = ‘Smithfield B’

(bottom of Terrace 3).

In each block, upper left = length; upper right = width; lower left = thickness; lower right =

length : width ratio.

VHS = very highly significant difference at 0,001 level; HS = highly significant difference at

0,005 level; S = significant difference at 0,01 level; LS = little significant difference at 0,05 level;

O=no significant difference at 0,1 level.

ee ae ee we VHS VHS O O

VHS O O O

SC SB(t)

It seems reasonable to assume that despite the inference that can be drawn

from the comparison of the BNK 1 and Oakhurst ‘Smithfield’ scrapers, a

‘Smithfield C’ phase did not, at BNK 1 or elsewhere, precede as well as succeed

a ‘Smithfield B’ phase, as is implied by the indication in Table 24 that the

BNK 1 layer 13 scrapers are not significantly different from the Oakhurst

‘Smithfield C’ scrapers, nor from those in BNK 1 layers 12-10 and that layer 13

is the only layer in which the scrapers have any resemblance to those of layer 9.

It would be justifiable, in fact, to regard such an inference as no more than a

‘statistical artefact’ resulting from the comparison of small samples on the basis

of an inadequate number of attributes with too great a common range.

Goodwin (1938: 318) observed that ‘there is no great difference to be observed

between the general types of implements to be found between this floor and

those found above. The variation is one of size or, more accurately, a change in

the distribution of sizes’. The ‘floor’ referred to is the ‘carbon floor’, which

Goodwin (1938: 239) considered to separate the ‘Smithfield C’ deposit from the

underlying ‘Smithfield B’. This has been dated to 7 910 = 70 years B.P. (Pta-377:

J. Deacon 1979: 31), which makes it somewhat older than BNK 1 layer 12

(Table 1).

140 ANNALS OF THE SOUTH AFRICAN MUSEUM

Where the BNK 1 scrapers are concerned, however, a change in size tends

to be accompanied by a change in shape and, as indicated in Figures 17 and 44,

the scrapers from BNK 1 layers 10-13 tend to be longer than they are wide,

while the Oakhurst ‘Smithfield’ scrapers are equilateral or, more precisely,

circular. B. D. Malan (1979 pers. comm.) who assisted Goodwin at Oakhurst,

has said that he told Goodwin at the time that he did not think the scrapers

could be considered typical of the ‘Smithfield B’.

Van Riet Lowe (1929: 179), writing of the ‘Smithfield B’ industry in the

Orange Free State, observed that the ‘duckbill end-scraper’ was by far the most

common implement. In the scrapers from BNK 1 layers 10-13 those that can be

called ‘duckbill’, even though they may be end-scrapers, are few (cf. Fig. 15B;

although this example is from layer 6 it is typical of the ‘duckbill’ scrapers in the

BNK 1 assemblage). They account for 15 per cent of the scraper total in layer

10, 19 per cent in both layers 12 and 13, and 21 per cent in layer 11. The

frequencies have been quoted as percentages despite the small size of some of

the samples in order to emphasize the relative paucity of their occurrence.

Schrire’s analysis of the Oakhurst tool categories (Schrire 1962, tables 3-4)

also shows low frequencies of this type of scraper (type 5) although there is

some discrepancy between the statistics in her two tables. Table 3 gives a total

of 16, whereas conversion of the percentages in Table 4 gives a total of only 8.

Fagan (1960: 84) used a different typology for the scrapers from Glentyre

shelter, but his class of ‘long scrapers’ numbers only 2, or 1,2 per cent of the

scraper total. The analysis of scrapers from Matjes River Shelter layer D by

Sampson (1974, table 42) gives a frequency of 16,1 per cent for ‘frontal

scrapers’ (cf. his fig. 91.24) which is comparable with the low frequencies in

BNK 1 layers 10-13.

The indications are, therefore, that ‘duckbill end-scrapers’ are not the

predominant type in those industries in the southern Cape that may be

considered analogous to the so-called ‘Smithfield’ at Oakhurst, which Sampson

(1974: 263-270) has termed the ‘Oakhurst industry’. Whether this ‘Smithfield

episode’ occurred at Nelson Bay Cave is doubtful in view of the hiatus between

about 8 000-6 000 B.p. (J. Deacon 1978: 100).

J. Deacon (1978: 92) has divided the scrapers from Nelson Bay Cave into

three size groups based on maximum horizontal dimension. The sequence

analysed by Deacon includes only the lower ‘Wilton’ deposits and, as indicated

in her table 1, the frequencies in nine of the eleven layers are extremely low.

Deacon’s figure 4 indicates, however, that small scrapers less than 15 mm long

are most common in the ‘Robberg’, large scrapers greater than 30 mm are most

common in the ‘Albany’, and small are again most common in the ‘Wilton’

followed by medium (15-30 mm). Small scrapers are present only in the

uppermost level, RB, of the ‘Albany’, while medium scrapers are present only

in RB and the underlying layer J.

BNK 1, with almost seven times the number of scrapers recovered from

the Nelson Bay Cave sequence studied by Deacon, shows marked differences in

the relative frequencies of Deacon’s three size classes:

BYNESKRANSKOP 1 141

BNK 1: large 6,2% medium 61,1 % small 32,7 %

NBC: 14,2 % 19,2% 66,5 %

Although BNK 1 has a longer sequence referable to the ‘Wilton’ than that

part of the Nelson Bay Cave sequence studied by Deacon, and although the

mode for the length of the scrapers from layer 9 up is at or near 15 mm (cf. the

means in Fig. 17), the size class divisions used by Deacon do not seem

appropriate for the BNK 1 scrapers. The frequencies in layers 19-14 are too

low to be useful, those from layer 5 up are certainly beyond the time range of

Deacon’s Nelson Bay Cave sample, and those for layers 13-6 may be sum-

marized as follows:

layers 13-10: large 13-17% medium 75-83 % small 0-13 %

9-6: 5-9 % 56-60 % 37-42 %

In both groups medium scrapers are dominant, but in the upper layer

group (9-6) small scrapers increase in frequency at the expense of both medium

and large. In the Nelson Bay Cave sequence small scrapers are not present

before the uppermost layer of the ‘Albany’ and become the dominant size class

in the ‘Wilton’, although it should be borne in mind that this site lacks the

equivalent sequence of BNK 1 layers 12-10. In the BNK 1 sequence it is only

in layer 3 that small scrapers are the dominant size class so that although it is

true that small scrapers increase in frequency in the upper layers of BNK 1 it

cannot be said that they are the dominant size class, as is the case in the Nelson

Bay Cave ‘Wilton’ layers studied by Deacon.

H. J. Deacon (1976, table 15) has given dimensional data for scrapers from

Melkhoutboom. There were apparently no scrapers from the ‘Albany’ units

MBS and RF (cf. tables 10-11) so that discussion is limited to the ‘Wilton’ units

at Melkhoutboom and the upper layers at BNK 1.

The means for length and thickness of the Melkhoutboom scrapers are

similar to those of BNK 1 layer 8 and above. The Melkhoutboom scrapers are

generally narrower than those from BNK 1 although those from the two CAF

units are similar to those from BNK 1 layers 8, 5 and 4. The CAF units are,

however, younger than the BNK 1 layers mentioned, falling within the time

range of layer 1.

Comparison of the means for the three dimensions and the length : width

ratios of the scrapers from Wilton (J. Deacon 1972, Appendix A, tables 1-4)

reveals very little similarity between these and those for the BNK 1 scrapers.

(For purposes of comparison, Deacon’s width:length ratios have been con-

verted to length: width ratios by dividing them into 1.) Although the Wilton

scrapers show decrease-increase trends that may be considered analogous to

those of BNK 1 layers 10-6 and 5-1, there are few points at which the scrapers

can be said to be morphologically similar (cf. also J. Deacon 1972, figs. 9, 11,

13, 15). The distribution of scrapers in Wilton layer 4A, in low frequencies

across a wide range, is broadly similar to the distribution in BNK 1 layer 10

but, whereas about a third of the Wilton scrapers are longer than 35 mm

(J. Deacon 1972, fig. 8), only three of the sixty-eight in BNK 1 layer 10 are.

142 ANNALS OF THE SOUTH AFRICAN MUSEUM

The metrical analyses of the scrapers from Buffelskloof (Opperman

1978, tables 5—8) show trends from layer ZJ3 up that are broadly similar to

those from BNK 1 layer 13 up. The Buffelskloof length: width ratios, how-

ever, show a trend from equilateral to wider shapes that starts at the bottom

of the sequence in layer HE2 but does not move into longer shapes as in the

case in BNK 1 layers 13-11. The scrapers from BNK 1 layers 13-5 are,

however, metrically more like those from Buffelskloof layers ZJ3—BOL1

than they are like those from Melkhoutboom or Wilton. Layer ZJ3 has not

been dated, but the inferred date for this layer, and the date for layer BOL1

(Opperman 1978: 21), suggest a broad contemporaneity with BNK 1 layers

13-5.

Adzes, termed ‘slugs’, were included in artefact collections from ten sites

in the southern to western Cape that contained an industry of which these were

considered by Rudner & Rudner (1954: 103) to be the type implement. Nine of

these sites are coastal and are located from Sandy Bay on the west coast of the

Cape Peninsula to Arniston (Waenhuiskrans) east of Cape Agulhas (Fig. 1).

The tenth site, Het Kruis, is in the western Cape, about 30 km south-west of

Clanwilliam and about 40 km from the coast.

In the samples from all sites except the most easterly, Hawston and

Arniston, which are closest to BNK 1, and the single inland site, Het Kruis,

‘slugs’ or adzes outnumber scrapers in frequencies that range from 52,9-80,1

per cent of the artefact total. In the three samples in which scrapers outnumber

adzes, frequencies for the latter range from 16,7—23,1 per cent of the artefact

total. Because of the apparent association of adzes with ‘Smithfield B’ artefacts

at Het Kruis and because of the resemblance of the ‘slugs’ to artefacts from

‘Smithfield N’ assemblages in Natal, Rudner & Rudner (1954: 106) ascribed the

‘Sandy Bay industry’ to the ‘Smithfield Culture-complex’ and suggested

(p. 107), on the evidence of the Cape Hangklip sites, that this industry

appeared to antedate the ‘Wilton’.

The Rudners’ inventories include only ‘slugs’, scrapers, bored stones and

grooved stones, although two segments and a borer were reported (p. 105)

from the Hout Bay midden. Potsherds were reported from Sandy Bay (p. 103),

Hout Bay, and Hangklip West (p. 105), also Hawston (p. 106). In a later paper

the Rudners (1956: 80) listed only the ‘type site’, Sandy Bay 1, and Hout Bay

as locally representative of ‘the Sandy Bay Variation of the Smithfield Culture’,

Noordhoek being described (p. 79) as ‘Wilton’.

Sampson (1974: 414) comments that while ‘worked-out’ adzes are present

in ‘Wilton’ deposits, ‘more work will be needed to determine whether the

Sandy Bay group represents a discrete cultural tradition, a later industry, or

merely an activity (or regional) variant of the Wilton in this area’.

Another view is that of Mazel & Parkington (1978: 382) who, after

comparing the Rudners’ scraper: adze ratios with those of other coastal and

inland sites in the western Cape, suggested that ‘in the case of the adze-

dominated assemblages this relates not to coastal settlement but to increased

BYNESKRANSKOP 1 143

woodworking activities, probably associated with underground plant food

gathering and the availability of suitable wood resources’.

Adzes were not found at Nelson Bay Cave, either in that part of the

sequence studied by J. Deacon or in the upper part of the ‘Wilton’ deposit

excavated by Inskeep (J. Deacon 1978; 1981 pers. comm.).

Adzes are common only in the CAF and OMB units at Melkhoutboom

(H. J. Deacon 1976, table 10) but are outnumbered by scrapers by a ratio of

about 3: 1. These layers fall within the time range of BNK 1 layer 1, in which

adzes are more numerous than scrapers (Table 7).

At Wilton (J. Deacon 1972, table 1) adzes were recovered from layers

3G-3A in frequencies lower than 10 per cent of the tool category total. Their

frequency never exceeds that of scrapers and their distribution does not seem

similar to that in the upper layers of BNK 1.

Low frequencies of adzes are recorded for the DGL member at Boomplaas

(H. J. Deacon et al. 1978, table 2), but frequencies are relatively high in the

BLD units of this member, particularly in BLD 2a, which yielded 21 of the site

total of 38. The frequency of adzes never exceeds that of scrapers as is the case

in BNK 1 layer 1, within the time range of which the DGL member falls.

Opperman (1978, table 2) has recorded low frequencies of adzes from three

of the upper layers of Buffelskloof, BOL1, BOL2 and CH2, as well as one from

the basal layer, HE2, but the frequency of these never exceeds that of scrapers.

The upper layers are probably within the time range of BNK 1 layers 4-11.

Backed scrapers were not recorded from Die Kelders (Schweitzer 1979,

table 17), nor from Nelson Bay Cave, (J. Deacon 1978, table 1). Considering

their distribution in the BNK 1 deposits, their absence from the Die Kelders

sample is perhaps fortuitous, taking into account the very low number of

retouched artefacts recovered; and it is possible that they might have been

found in the upper deposit at Nelson Bay Cave excavated by Inskeep.

Goodwin (1938: 307) recorded rare ‘double-edged crescents’ in the upper-

most Oakhurst deposits, and commented that ‘on the Riversdale and Cape

Peninsula coasts the single-edged type is the rarer’. Unless Goodwin was

referring to very localized distributions in the specific areas mentioned, BNK 1

and Die Kelders would appear to be exceptions to Goodwin’s observation.

Schrire’s analysis of the Oakhurst material (Schrire 1962, fig. 6, tables 3—4)

records the presence of ‘double crescents’ (type 11) in two of the ‘Smithfield C’

spits and in two of the ‘Wilton’. Again, however, conversion of the percentage

frequencies in Schrire’s table 4 does not yield the numerical frequencies given

in table 3. It is thus not clear whether there were more of these artefacts in the

‘Wilton’ than in the ‘Smithfield C’ (table 3) or the opposite (table 4). In any

event, the total (13 or 7) is so low as to do little more than indicate the

presence of these artefacts in contexts chronologically broadly comparable with

those at BNK 1 in which they also occur.

Backed scrapers are not recorded from Melkhoutboom (H. J. Deacon

1976: 59) and, although none are recorded from Wilton (J. Deacon 1972,

144 ANNALS OF THE SOUTH AFRICAN MUSEUM

table 1), J. Deacon has advised (1979 pers. comm.) that these artefacts were

recovered and were included with ‘double scrapers’. ‘Double scrapers’ occur in

Wilton layers 4A-3A, with maximum frequencies in layers 3F—3E. Layer 3F is

probably somewhat older than BNK 1 layer 6 and about 1 000 years older than

layer 5. The temporal distribution at the two sites may, however, be seen as

being broadly similar.

There were apparently no backed scrapers in the DGL member at Boom-

plaas (H. J. Deacon et al. 1978, table 2) but at Buffelskloof Opperman (1978,

table 2) recorded a total of 19 ‘double scrapers’ from layers ZJ2—BOL1, 12 of

them in BOL2 alone. This layer is probably as much as 1 000 years older than

BNK 1 layer 6, but there may well be typological as well as terminological

differences between the double scrapers at Buffelskloof and the backed

scrapers at BNK 1 (cf., however, Opperman 1978, fig. 7.4).

In the ten assemblages from western Cape sites studied by Mazel (1978)

there appears to be no clear correlation between the frequency of backed

scrapers and the environment or locality of the sites. Four of the five sites in the

fynbos region (the term used by Mazel to distinguish the vegetation of the

inland areas from that of the Strandveld, the coastal plain and coast region)

have frequencies of backed scrapers ranging from 0 to 4 per cent of the ‘formal

tool’ category while the fifth, Koopmanskraal 2, has a frequency of 21 per cent

(Mazel 1978, Appendix 3). These sites on the inland edge of the coastal plain

are probably located the most analogously to BNK 1 although the climatic

regimes of the two areas are very different. The five sites situated in the

Strandveld have frequencies that range from 0 to 14 per cent. This suggests that

BNK 1, in terms of backed scraper frequencies, has more in common with sites

on the western Cape coastal plain than with those on its inland margin, which is

generally further from the coast than in the BNK 1 area. The sites studied by

Mazel possibly fall within the range of dates for BNK 1 layers 2 and 1 only

(J. Parkington 1979 pers. comm.). Both these layers have minimal frequencies

of backed scrapers (Table 7).

Segments are surprisingly abundant at Die Kelders, considering the largely

informal nature of the stone artefact assemblage. They outnumber scrapers by

almost 5:1 and account for 46,8 per cent of the total of ‘formal tools’

(Schweitzer 1979, table 15). Quartz is the predominant raw material (Schweit-

zer 1979, table 14), which accords with the chronologically comparable situa-

tion in BNK 1 layer 1 (cf. Fig. 14), although in this layer scrapers outnumber

segments by rather more than 2:1 (Table 7).

Goodwin (1938: 304-17) recorded the recovery of upwards of forty stone

‘crescents’ from the ‘Wilton’ layers at Oakhurst, as well as half a dozen from

the deposit immediately underlying his 36-inch (91,4 cm) test spit. These

Goodwin also associated with the ‘Wilton’. Another dozen segments were

recovered from the ‘Smithfield C’ deposits but it is not clear whether the six

included in Goodwin’s inventory for the 40-45 inch (102-114 cm) spit are the

same as those referred to above.

BYNESKRANSKOP 1 145

Segments were most abundant at the ‘Smithfield C—Wilton’ interface, with

the next to highest frequencies at the top of the ‘Wilton’ deposit. This deposit

was also remarkable for the great number of ‘crescents’ made of the shell of the

black mussel Choromytilus meridionalis (Mytilus edulis in Goodwin’s text).

Only about a dozen of these were recovered from the ‘Smithfield C’ levels but

Goodwin (1938: 322) considered that they, as well as the stone segments,

should be ignored as intrusive material brought into the lower deposit by the

human burials. Goodwin (1938: 304) also considered that in the ‘Developed

Wilton’ segments made of shell largely replaced those made of stone.

Schrire (1962, table 3) recorded 23 segments (type 10) from the ‘Smithfield

C’ levels and 28 from the ‘Wilton’, which conflicts somewhat with the pre-

viously mentioned observations by Goodwin. However, conversion of the

percentage frequencies in Schrire’s table 4 gives a total of 18 segments in the

‘Smithfield C’ levels (36-60 inches) and 9 in the ‘Wilton’ (0-36 inches). There is

thus little clarity regarding the distribution of segments in the two units.

In the BNK 1 sample segments are most common in layers 9-5, layer 5

having the highest frequency, AD as well as actual. If layers 12-10 are

considered the equivalent of the “Smithfield C’ at Oakhurst and layer 5 to the

‘Developed Wilton’, then the correspondence between the two sites, where

segments are concerned, is not close. Shell ‘crescents’ were not recovered from

the deposits at BNK 1, although they are recorded from Die Kelders (Schweit-

Zeno 9> 53,459):

Two segments were recovered from the ‘Robberg’ layers at Nelson Bay

Cave, none from the ‘Albany’ and 32 from the ‘Wilton’, of which 23 came from

the uppermost layer, IC, excavated by Klein (J. Deacon 1978, table 1). In the

BNK 1 sequence segments are present only in layers 17 and 16 of the layers

chronologically comparable with the ‘Albany’ of Nelson Bay Cave, but present

in every layer from 12 up. The increasing frequency of segments in BNK 1

layers 9-5 and in the ‘Wilton’ of Nelson Bay Cave is similar, as is their relative

frequency in the ‘formal tool’ or retouched artefact category: 12,6 per cent of

the ‘Wilton’ total at Nelson Bay Cave and 13,6 per cent of the total for BNK 1

layers 9-5. However, whereas segments account for 76,2 per cent of the backed

tool class in the ‘Wilton’ sample from Nelson Bay Cave, they account for only

39,8 per cent of the class total in BNK 1 layers 9-5. The BNK 1 class of backed

pieces includes relatively high frequencies of backed scrapers in these layers, an

artefact type that is absent from the Nelson Bay Cave ‘Wilton’ sample.

H. J. Deacon (1976: 59) has observed that all the segments from Melk-

houtboom were made from flakes, and his table 12 shows that chalcedony is the

predominant raw material, with 20 per cent or less in each layer made of quartz

and/or other raw materials. The segments are restricted to the ‘Wilton’ layers

and Deacon’s table 25 indicates that they were more common in the lower

layers M—MB than in the upper CAF-S, a situation analogous to that at BNK 1

although the M—MB layers may be somewhat older than BNK 1 layers 9-5. The

mean length of the Melkhoutboom segments is within the range of variation of

146 ANNALS OF THE SOUTH AFRICAN MUSEUM

the quartz segments from BNK 1 but the mean width is less than that of the

quartz and silcrete segments in layers 9-5. The mean length: width ratios

(converted from Deacon’s width:length ratios) are comparable with those of

BNK 1 layers 10, 9 and 5-1 and indicate a more elongate shape than those in

layers 8-6.

Segments from Wilton were relatively fewer than from BNK 1, accounting

for only 4,0 per cent of the tool category and 32,6 per cent of the backed tool

class (J. Deacon 1972, table 1, including broken segments). Segments account

for 11,6 per cent of the retouched artefact category in the BNK 1 sample, and

40,9 per cent of the backed artefact class. They are most common in Wilton

layer 3F, which is somewhat older than BNK 1 layer 5, which has the latter

site’s highest layer frequency. The Wilton segments are predominantly silcrete

and their mean length and width (J. Deacon 1972, fig. 5; Appendix A, table 8)

are within the range of those from BNK 1 layers 6-4. The average shape of the

Wilton segments is one in which the length is about three times the width while

in the BNK 1 segments the length is only 1,5—2,5 times the width.

Only two segments were recovered from the DGL member at Boomplaas

(H. J. Deacon et al. 1978, table 2). The BLD-BLD 1 and BLD 2 units from

which these came are within the upper time range of BNK 1 layer 1 which has a

relatively high frequency of segments: 14,8 per cent of the retouched artefact

category and 80,4 per cent of the backed artefact class.

At Buffelskloof Opperman (1978, table 2) recovered 104 segments, includ-

ing broken pieces. None were found below layer CH3 and slightly more than

haif the total came from layer BOL2 alone. This layer is about 1 000 years

younger than BNK 1 layer 9 and about 1 200 years older than layer 5. The

range of variation in the relative frequencies in the Buffelskloof layers is,

however, not great and is much the same as that in BNK 1 layers 10, 8 and 7,

all of which probably fall within the time range of the Buffelskloof layers.

Layer BOL1, which is approximately coeval with BNK 1 layer 5, has a low

frequency of segments, which account for only 5,9 per cent of the tool category,

compared with 14,6 per cent for BNK 1 layer 5. The mean length of the

Buffelskloof segments (Opperman 1978, table 16) is similar to that of the

silcrete segments from BNK 1 layers 9-7 and those in quartz from layer 6. The

mean width decreases from layer CH3-MDA and increases slightly in BOL1.

The segments in layer CH3 are similar to the quartz and silcrete segments in

BNK 1 layer 6, which are the widest of all the BNK 1 segments (Fig. 22), while

those from layers MDA and BOLI are similar to the quartz segments from

BNK 1 layers 5 and 4. In shape the Buffelskloof segments appear to follow a

reverse trend to that of the BNK 1 segments, being wider in layer CH3 than in

the overlying layers while in the BNK 1 segments the trend is for an incr2ase in

width from layer 9 to layer 6 and, after the anomalous reversion to a longer

shape in layer 5, the wider shape continues in the upper layers. A denticulate

segment similar to those from BNK 1 layers 9 and 12 (Fig. 20H) came from one

of the approximately contemporary Buffelskloof layers, CH1—CH3, although it

BYNESKRANSKOP 1 147

is somewhat larger than the BNK 1 segments, being approximately 15 mm long

(Opperman 1978, fig. 8.9).

Mazel (1978: 87) does not separate segments from his general class of

backed pieces, though they are listed separately in the inventories in his

Appendix 3. Segments are numerically most common at the two ‘lacustrine’

sites, Uithoek and Verlore. Five of the other eight sites have ‘formal tool’ totals

ranging from 13 to 67 which makes comparison of the relative frequencies in

the various classes with those from BNK 1 of doubtful value. It seems, though,

that although segment frequencies are low, they are higher at sites on the

coastal plain that at sites in or near the mountains.

J. Deacon (1974: 17) has observed that ‘assemblages from southern

Rhodesia and South Africa with more than 10 per cent of segments relative to

scrapers are likely to date to the middle of the Holocene sequence, while those

with more than 50 per cent segments in relation to other backed tools are likely

to be older than 4 000 B.p. It seems reasonable to propose that segments can be

regarded as a time-controlled feature or a “Temporal Type” in the later Stone

Age sequence in southern Africa and that they can be expected to occur in

quantity in horizons dated to between 7 000 and 3 000 B.p.’. Although part of

BNK 1 layer 1 falls within Deacon’s 7 000-3 000 B.P. time bracket, segments

account for 31,5 per cent of the scraper—-segment total and for 61,2 per cent of

the total of backed artefacts (Table 7). The Die Kelders Holocene deposits, all

of which are younger than 3 000 B.P., yielded 29 segments but only 6 scrapers

and 7 other backed artefacts (Schweitzer 1979, table 15) and thus also presents

a deviation from Deacon’s suggestions regarding the temporal distribution of

segments.

Backed flakes in the Die Kelders assemblage were included in the class of

‘other backed tools’, of which there were only six (Schweitzer 1979, table 15).

There were also few of these artefacts in BNK 1 layer 1.

There were apparently no backed flakes from Oakhurst (Schrire 1962, fig.

6; tables 3-4). They might have been included with scrapers, although Goodwin

(1938: 314), in commenting on the ‘Smithfield C’ deposit from which over 5 000

white quartz scrapers were recovered, mentioned that ‘in addition there are

some two hundred flakes of brown chalcedony, carefully used and showing

perfect cleavage, yet only in a dozen instances has any attempt been made to

trim these into implements’. This observation may also be taken to indicate that

there were no recognizable backed flakes in the overlying ‘Wilton’ deposits

since Goodwin (1938: 307, 310, 313) does not mention them.

There were only 5 backed flakes and bladelets from Nelson Bay Cave, 2 in

the ‘Robberg’ levels and 3 in the ‘Wilton’ (J. Deacon 1978, table 1).

Backed flakes are not listed separately in the inventories of artefacts from

Melkhoutboom (H. J. Deacon 1976) although Deacon’s table 9, which is

concerned with raw material frequencies, has two categories of flakes, un-

modified and retouched. Frequencies of retouched flakes in layers M and W are

particularly high, accounting for 33,2 and 42,9 per cent of the site total, and the

148 ANNALS OF THE SOUTH AFRICAN MUSEUM

total for layer M alone (146) exceeds the BNK 1 site total for backed flakes.

There were more retouched flakes in the ‘Robberg? layer B than in the ‘Albany’

layers RF and MBS, but 93,2 per cent of the retouched flakes are in the

‘Wilton’ layers WBM-—OMB, a situation analogous to that of BNK 1 where 94,4

per cent of the backed flakes are in layers 9-1.

There are only 21 backed flakes in the Wilton stone artefact inventory

(J. Deacon 1972, table 1). Their maximum occurrence is in layers 3F and 3E,

which may be approximately coeval with BNK 1 layers 6 and 5. These layers

mark the end of the period of relatively high frequency that began in layer 9.

Only three backed flakes and blades are recorded from the DGL member

at Boomplaas (H. J. Deacon et al. 1978, table 2) and these come from the

lowest part of the member, layers BLD 2 and 2a.

Of the twenty backed flakes from Buffelskloof (Opperman 1978, table 2),

twelve came from layer BOL2 which 1s possibly about the same age as BNK 1

layers 8-7.

Mazel (1978, Appendix III) does not have a separate class of backed flakes

and if there were any recovered they are probably included in his class of

miscellaneous backed pieces.

Backed blades were rare at Oakhurst, only two being recorded from the

lowest of the ‘Wilton’ levels (Schrire 1962, tables 3-4).

Although bladelet cores are considered a diagnostic feature of the ‘Rob-

berg’ industry at Nelson Bay Cave and ‘a number’ of unworked bladelets were

recovered, a total of only five backed flakes and bladelets is recorded from the

site (J. Deacon 1978: 88, 100; table 1; fig. 4). Two were from the ‘Robberg’

levels and three from the ‘Wilton’. This suggests that if blades were being

produced in any quantity they were not being backed or were used and

abandoned away from the site more than at it.

At Melkhoutboom (H. J. Deacon 1976, tables 10-11) backed bladelets

were recovered in low frequencies (maximum 6) from all the ‘Wilton’ layers

except the lowest, WBM. In contrast a total of 122 backed bladelets was

recovered from Highlands Rock Shelter in the Cradock district (H. J. Deacon

1976, table 44). These came from the Upper Member, for which a maximum

date of 4 500 + 60 years B.P. was obtained. Deacon (1976: 129) has interpreted

the virtual lack of segments at Highlands (only one was found) and the

abundance of backed blades, compared with the reverse situation at Melkhout-

boom, as indicating ‘a significant difference in formal hafted tool design and

preference’ reflecting “some measure of social distance’.

Over 100 backed blades and snapped backed blades were recovered from

Wilton (J. Deacon 1972, table 1). These were most common in layers 3F (31)

and 3C (27), which are possibly contemporary with BNK 1 layers 6—4, from

which just over half the site total of backed blades came. The frequencies of

these artefacts from Wilton are, however, very different from all the sites

discussed so far except Highlands, and it is of some interest that the name site

of the “Wilton industry’ should, even in one respect, have more in common

BYNESKRANSKOP 1 149

with a ‘Smithfield’ site (H. J. Deacon 1976: 168-169) than with other sites

containing ‘Wilton’ deposits.

At Boomplaas only three backed ‘flakes/blades’ were recovered, from the

basal units of the DGL member (H. J. Deacon et al. 1978, table 2) while

Opperman (1978, table 2) recorded twenty-six backed blades from Buffels-

kloof. These came from the uppermost layers, MDA-BOL1, with BOL2 alone

contributing twenty-three. These layers may be the same age as BNK 1 layers

7-4, from which almost half the site’s backed blades were recovered.

In the samples from the western Cape sites studied by Mazel (1978,

Appendix III), frequencies of backed blades are generally low, ranging from 0

to 5 per cent of the ‘formal tools’; but at three sites, Kookfontein, Koopmans-

kraal and Verlore, they account for 6 to 10 per cent. Mazel (1978: 87) has

observed that backed pieces (as a class, including backed blades) have higher

frequencies at sites in the Sandveld (sic) than at those in the mountains, but

there is no clear correlation between the frequency of backed blades and

locality. For example, the Verlore and Uithoek sites are only about 10 km

apart and both are on the coastal plain, although Verlore is rather closer to the

sea and to the Verlore Vlei lagoon. Backed blades account for 9,7 per cent of

the Verlore ‘formal tools’, while at Uithoek they account for only 1,5 per cent.

Conversely, at Koopmanskraal and Kookfontein backed blades account for 6

per cent of the category total at each site, though the former is in the interior of

the coastal plain and the latter near the mountains.

Borers were minimally represented at Nelson Bay Cave by two in the

lower ‘Wilton’ layers RA and BSC (J. Deacon 1978, table 1). These layers are

about the same age as, or slightly younger than BNK 1 layers 10 and 9.

Borers were similarly restricted to the ‘Wilton’ layers at Melkhoutboom

(H. J. Deacon 1976, tables 10-11) with maximum frequencies of 10-12 in the M

and MB units, which are contemporary with BNK 1 layer 10 and some of the

overlying layers.

There were also few borers in the Wilton sample (J. Deacon 1972, table 1).

They have their maximum frequency (8) in layer 3F, which is about 1 000 years

older than BNK 1 layer 5.

One borer is recorded from the basal unit, BLD 2a, of the DGL member

at Boomplaas (H. J. Deacon et al. 1978, table 2) while eight were found at

Buffelskloof, sporadically distributed through layers HE1-BOL2 (Opperman

1978, table 2). These layers span the probable time range of BNK 1 layers

13-7.

Borers were recorded from only 5 of Mazel’s 10 western Cape sites (Mazel

1978, Appendix III), with only Koopmanskraal and Verlore, each in a different

environment, yielding more than 10, Verlore having a high total of 54, although

these account for only 3,9 per cent of the ‘formal tools’, a frequency similar to

that for BNK 1.

Notched flakes are something of a typological or terminological problem in

the inter-site comparisons. Schrire (1962: 191) mentions the presence of ‘small

150 ANNALS OF THE SOUTH AFRICAN MUSEUM

saws’ in the ‘Smithfield B’ assemblage from Oakhurst, but her inventories in

figure 6 and tables 3 and 4 list a separate category of ‘denticulates’ (type 7).

These are most common in the ‘Smithfield C’ but are also present in the

‘Wilton’ and possibly also in the ‘Smithfield B’: two are listed as from these

deposits in Schrire’s table 3 but do not appear in figure 6 or table 4. In the

BNK 1 assemblage denticulates are more common in the upper layers than in

the lower, a situation different from that at Oakhurst.

Sampson (1974, table 71) has listed the presence of ‘notched scrapers’ in

assemblages from five southern Cape sites, from Nelson Bay Cave to Klasies

River Mouth and AndrieskraalI. J. Deacon (1978, table 1) has, however,

included the notched pieces from Nelson Bay Cave in the category of utilized

artefacts rather than with the ‘formal tools’. Over 500 notched flakes were

recovered, a situation very different from that at BNK 1, and they were

recovered from every layer, showing, as does the BNK 1 sample, that this is an

artefact type that transcends ‘industrial’ divisions. Only seven ‘notched

scrapers’ were recorded from Andrieskraal I (J. Deacon 1965, table 1), and the

Klasies River Mouth data are still to be published.

At least three of the ‘slugs’ from Sandy Bay (Rudner & Rudner 1954: 104

figures 4-6) would fit into the class of notched flakes as defined for the BNK 1

assemblage except that one, no. 6, has scraper retouch as well. No. 5 is

described in the caption as a ‘serrated slug’ and Sampson (1974, fig. 153.3)

reproducing the same illustration, has called it a ‘denticulate piece’.

BONE ARTEFACTS

Although the sample of bone artefacts from BNK 1 is much smaller than

those from Die Kelders (Schweitzer 1979, table 2) and Nelson Bay Cave

(J. Deacon 1978, table 4) it covers much the same range. Where BNK 1 differs

significantly from Nelson Bay Cave, however, is that bone artefacts were

sufficiently abundant in the ‘Albany’ layers to allow J. Deacon (1978: 104) to

postulate that ‘the reaction of the people in the southern Cape to the environ-

mental changes coincident with a rising sea level and warmer conditions at the

end of the Pleistocene . . . was [inter alia] . . . to make a wide range of bone

rather than stone Formal Tools’. In BNK 1 layers 18-10, even though

retouched stone artefacts are relatively rare, particularly in layers 18-13, they

outnumber bone artefacts by 31:1, or 23:1 if scrapers are excluded. In the

‘Albany’ layers of Nelson Bay Cave bone artefacts exceed retouched stone by

2:1 or almost 28:1 if scrapers are excluded. The BNK 1 evidence for the lower

layers points to a lower degree of occupation rather than a major change in the

traditions of artefact manufacture. Bone artefacts seem to be more common at

coastal sites than at inland ones (cf. J. Deacon 1972: 14 re Wilton; H. J.

Deacon 1976: 49 re Melkhoutboom, table 39 re Highlands; H. J. Deacon et al.

1978, table 6 re Boomplaas; Opperman 1978: 21 re Buffelskloof) and BNK 1

seems intermediate between the two groups.

BYNESKRANSKOP 1 tot

Fish gorges have a much greater temporal restriction at BNK 1 than at

Nelson Bay Cave, where they were found in every layer from the final

‘Robberg’ (BSL) to the first ‘Wilton’ (RA) thus spanning the period from about

12 000 to 6 000 B.P. (J. Deacon 1978, fig. 3, table 4). In the BNK 1 sample they

are present only in layers 15-13 which perhaps span the period 10 000-8 000

B.P. Parkington (19806: 317) reports the presence of about 800 of these arte-

facts at Elands Bay Cave, mostly in layer 12, which is dated to around

9 600 B.P.

Ornaments from BNK 1 cover a wider range than the sample from Die

Kelders (Schweitzer 1979: 137, fig. 18D) but not unlike that from Nelson Bay

Cave, where they were found in most layers but chiefly in the ‘Albany’ and

‘Wilton’ (J. Deacon 1978, table 4).

MARINE SHELL ARTEFACTS

Ornaments similar to those from BNK 1, apart from those made from

Donax serra, were also found at Die Kelders (Schweitzer 1979, table 4)

although the relative frequencies of the different types at the two sites differ.

Nassa kraussiana beads as well as perforated Donax serra were recovered at

Melkhoutboom (H. J. Deacon 1976, figs 28.2, 29) and Nassa kraussiana beads

are also recorded from the DGL member at Boomplaas (H. J. Deacon et al.

1978, table 10).

Pendants from BNK 1 have shapes in common with those from Die

Kelders (Schweitzer 1979, fig. 21) although the latter lack the triangular shape

(Fig. 35A, 36D-G), nor do any of the Die Kelders pendants have decorated

edges.

The illustrations of the ornaments from Oakhurst (Goodwin 1938, figs

18-62) do not always receive comment in the text but it is evident that shell

ornaments, including pendants, were recovered from the ‘Wilton’ layers. Good-

win (1938: 316) does mention that nacre ornaments were found in the ‘Smithfield

C’ deposits though these were rare and came from the upper 9 in. (23 cm).

Shell ornaments are not recorded from the lower ‘Wilton’ deposits at

Nelson Bay Cave excavated by Klein (J. Deacon 1978) but were recovered

from the upper deposits excavated by Inskeep and include designs similar to

that from BNK 1 shown in Figure 36I (Inskeep 1978, fig. 15).

Pendants made of Turbo sp. shell are recorded from the W unit at

Melkhoutboom which is dated to between 6 980-5 900 B.p. (H. J. Deacon 1976,

fig. 28; table 2). This makes their occurrence approximately contemporaneous

with their first appearance at BNK 1. The Melkhoutboom pendants are mostly

circular to oval and are unperforated as well as perforated, and four of the six

illustrated have notched or denticulate edges. This type of decoration, occur-

ring at BNK 1, Nelson Bay Cave and Melkhoutboom, appears to have been

widespread.

Edge-damaged Donax serra valves or fragments were recovered in fairly

high frequencies from Die Kelders (Schweitzer 1979, fig. 23, table 10). They

iS ANNALS OF THE SOUTH AFRICAN MUSEUM

are also recorded, in small numbers, from Oakhurst (Goodwin 1938: 307) and

apparently came only from the ‘Wilton’ layers. At Melkhoutboom, Donax serra

is present throughout the entire sequence (H. J. Deacon 1976, table 4a), but it

is not recorded when edge-damaged pieces first appear. H. J. Deacon (1976,

fig. 29) has suggested that the edge modification could have resulted from

utilization ‘although a very similar edge results in [sic] removing the central part

of the valve’. At Elands Bay Cave, Donax ‘scrapers’ were recovered from

layers dated at about 8 000 B.p. (Parkington 1979: 11), at about the same time

as they first appear at BNK 1, but apparently earlier than at Oakhurst.

OSTRICH EGG-SHELL ARTEFACTS

Decorated ostrich egg-shell was not found at Die Kelders, nor were any

pendants (Schweitzer 1979: 149-50). Fragments of ‘flask’ openings were found,

however, and this suggests that their absence from BNK 1 layer 1 is probably

fortuitous.

The presence of decorated ostrich egg-shell in the ‘Smithfield C’ deposits at

Oakhurst (Goodwin 1938: 317; fig. 60) supports the evidence from BNK 1 that

the practice of decorating ostrich egg-shell was not confined to the ‘Wilton’,

and the presence of fragments decorated with the ‘ladder pattern’ (Fig. 37B—C)

at Boomplaas (H. J. Deacon et al. 1978, fig. 12.6-7) indicates that this type of

decoration is not locally restricted.

POTTERY.

There is little that can be said about the pottery from BNK 1 except that

the sherds point to a relationship with the pottery from the upper layers of Die

Kelders (Schweitzer 1979: 162-169) and fall within the general range of coastal

pottery described by Rudner (1968). The Donax-impressed decoration (Fig.

38A-B; Schweitzer 1979, figs 30, 31A) cannot be claimed as part of a distinctive

tradition restricted to the Agulhas region since similarly decorated sherds have

been recovered from Diepkloof, near Elands Bay Cave (Wilson 1974, fig.

17.3-5), and possibly elsewhere (cf. Rudner 1968, figs 6.1, 10.6, 31.51). In

passing, it is perhaps worth mentioning that in the only known painted site in

the Agulhas region, a cave in the Kleinriviersberge some kilometres east of

Stanford, are hand-prints with semicircular patterns similar to those at Elands

Bay Cave (W. van Ryssen 1981, pers. comm.).

Spouted pots have been found at coastal sites from the Saldanha area to

East London (Rudner 1968, table 6) and Rudner (1968: 455) mentions that

they have also been found at inland sites.

METAL ARTEFACTS

Metal artefacts are rare in Late Stone Age deposits in the southern Cape,

possibly because metal in its raw form was not readily available and most of the

Khoisan peoples of the area might have lacked the knowledge necessary for its

BYNESKRANSKOP 1 153

extraction and processing. If metal artefacts were scarce they would have had

added value as objects of status, or for barter, and they might have been

considered too valuable to be discarded even when worn out or broken, or to

be buried with their owners.

Goodwin (1956: 50) is unlikely to be correct in his assertion that ‘the Cape

Hottentots had no metals and no knowledge of metal-working in 1650’. In the

earliest record of contact between voyagers from Europe and the local inhabi-

tants of the Cape, which took place at St. Helena Bay in November 1497, the

writer of the journal of Vasco da Gama’s voyage recorded that the Portuguese

traded the artefacts of the inhabitants for small copper coins and observed:

‘From this it seemed to us that they prized copper; and they also wore small

beads of it in their ears’ (Raven-Hart 1967: 4). St. Helena Bay is about 130 km

distant from Table Bay (Fig. 1) and it seems unlikely that the inhabitants of the

Cape Peninsula would not have had access to copper artefacts, especially when

the Cochoqua or ‘Saldanhars’ are considered, whose annual pastoral migration

included the whole of the coastal area between the two bays. If, as seems

likely, the source of the metal was the north-western Cape, movement of the

metal from tribe to tribe quite possibly extended along the southern Cape as

well. In any event, extracts from the journals of visitors to the Cape prior to

1652 (Raven-Hart 1967) provide abundant testimony of an awareness of the

uses of metals, particularly copper and brass, to the extent that, as Goodwin

(1956: 47) himself observed, they were ‘willing to trade their beloved cattle to

obtain it’.

The Namaqua, whose territory lay in the north-western Cape, certainly

had the ability to extract and work the copper ores that are still mined in the

vicinity of Springbok, as testified to in the records of the expeditions by

Meerhof in 1661 (Moodie 1838: 400-412) and Simon van der Stel in 1685-6

(Waterhouse ed. 1932). There seems little reason to doubt that the copper

plates and beads produced by the Namaqua would have been traded with the

tribes to the south, even though the mechanics of trade among the early

Khoisan are still poorly understood.

In any event, it is quite clear from the early records that by the end of the

seventeenth century, the date indicated for the bead from BNK 1, a consider-

able quantity of metal had been in circulation among the Khoisan for a century

or more. It would be interesting to know what became of it all.

MAMMALS

The mammalian fauna from Die Kelders (Schweitzer 1979, table 27) covers

as wide a range as that from BNK 1 layer 1. There are thirty-two taxa in the

Die Kelders sample as against twenty-seven in BNK layer 1. Nine taxa in the

Die Kelders sample are not present in the BNK 1 layer 1 sample: ?Canis

familiaris, Genetta sp(p)., Felis cf. serval, Mirounga leonina, Loxodonta afri-

cana, ?Bos taurus, Tragelaphus scriptus, Pelea capreolus and Cetacea. Of these,

Loxodonta africana and Pelea capreolus are both present in BNK 1 layer 4 and

i154) ANNALS OF THE SOUTH AFRICAN MUSEUM

below, suggesting a chance element in their absence from layer 1. Present in

the BNK 1 layer 1 sample but absent from the Die Kelders sample are Aonyx

capensis, Herpestes ichneumon, Hyaena brunnea, and Felis cf. caracal. It is

possible that the felines from the two sites may be the same species, and none

of the others are so habitat-specific that they would be restricted to one side of

the Franskraal Mountains or, in the case of Muirounga leonina and the

cetaceans, to Walker Bay, although this is a favoured area for whales in the

breeding and calving seasons. There is one positively identified Damaliscus

dorcas in the Die Kelders sample as well as another six probable identifications.

This raises problems with regard to the ascription of the absence of this species

from BNK 1 after layer 11 to environmental factors: unless these antelope are

not predominantly grazers, the eastern side of Walker Bay is, and probably has

been for the past few thousand years, a less likely habitat than the area to the

east of BNK 1 (cf. Schweitzer 1979: 113-114, and the section on vegetation in

the present report).

The Die Kelders sample (n = 2 017) is considerably larger than that from

BNK 1 layer 1 (n= 90) but, while both are numerically dominated by Bathyer-

gus Suillus, at Die Kelders this species accounts for 71,5 per cent of the total

whereas in BNK 1 layer 1 it constitutes only 41,1 per cent of the total. The size

class distribution of all animals in both samples is largely similar, with frequen-

cies generally ranked inversely to size, i.e. very small animals are most

common, very large least so.

After Bathyergus suillus the most common species at Die Kelders are

Raphicerus spp. and Arctocephalus pusillus, followed at some distance by Ovis

aries. In the BNK 1 layer 1 sample Raphicerus spp. also rank second but Ovis

aries ranks third and is followed by Syncerus caffer, which ranks only sixth at

Die Kelders. As at BNK 1, identified Raphicerus melanotis is more common

at Die Kelders than R. campestris. Arctocephalus pusillus accounts for only

5,0 per cent of the Die Kelders site total, but this is higher than at BNK 1,

where it accounts for 2,2 per cent of the layer 1 total and 2,5 per cent of the

site total.

Comparison of the BNK 1 mammals with those from Nelson Bay Cave on

the basis of comparable chronology is restricted to Nelson Bay Cave layers

BSL-IC and BNK 1 layer 19 to somewhere between layers 9 and 5. Account

must also be taken of the hiatus in the Nelson Bay Cave sequence between

layers RA and RB (J. Deacon 1978: 100-101), which covers the period of

deposition of BNK 1 layers 13/12-10.

The Nelson Bay Cave mammals have been analysed by Klein (1972a,

1972b, 1974) and a later paper (Klein 1976a) has frequency diagrams that

include data from the upper levels excavated by Inskeep. The most complete

numerical data are, however, in Klein’s third paper and it is these that have

been used for the present comparison. In this paper Klein (1974, table 4) has

grouped the frequencies into ‘Robberg’, ‘Albany’ and ‘Wilton’ units which may

be considered as broadly analogous to BNK 1 layers 19-18, 17—13/12 and 9-6.

BYNESKRANSKOP 1 155

However, for the purpose of the present comparison the BNK 1 sample is

simply divided into the lower layers, 19-10, and the upper, 9-1.

The Nelson Bay Cave sample is some 30 per cent larger than that from

BNK 1 and contains forty taxa, of which sixteen are bovids. More than half

(52,7 %) of the individuals are in the ‘Wilton’ deposits, with 28,9 per cent in the

‘Albany’ and 18,3 per cent in the ‘Robberg’. In the BNK 1 sample 64,4 per cent

of the site total came from layers 1-9, 27,2 per cent from layers 10-17 and 8,4

per cent from layers 18 and 19.

There are 27 species in the ‘Robberg’ of Nelson Bay Cave, 27 in the

‘Albany’ and 25 in the ‘Wilton’. This indicates that, as was the case at BNK 1,

although there was some shift in the species hunted, the change was more in

the frequency with which certain species were procured rather than in a

broadening of the range.

Taxa common to both sites but restricted to the “Robberg’ at Nelson Bay

Cave are Hyaena brunnea, Equus cf. quagga, Redunca arundinum, Damaliscus

sp., and possibly Connochaetes sp. and Lepus cf. capensis. Of these, Hyaena

brunnea is found at BNK 1 up to layer 1, Equus cf. quagga up to layer 9,

Redunca arundinum up to layer 6, Damaliscus dorcas up to layer 11, and

Connochaetes/Alcelaphus and Leporidae up to layer 1.

Species restricted to the “Robberg’ and ‘Albany’ at Nelson Bay Cave

include Phacochoerus aethiopicus, Oreotragus oreotragus and Taurotragus oryx.

Phacochoerus aethiopicus was found only in layer 13 at BNK 1, Oreotragus

oreotragus sporadically up to layer 1, and Taurotragus oryx in layers 19-16 and

7. There are no species common to both sites that are restricted to the ‘Albany’

at Nelson Bay Cave and are not also found in the lower layers of BNK 1.

Species restricted to the ‘Wilton’ at Nelson Bay Cave are Aonyx capensis,

Atilax paludinosus, and Felis libyca. Aonyx capensis was found at BNK 1 only

in layers 4 and 1, and Atilax paludinosus only in layer 14 while Felis libyca was

present sporadically in layers 19-13, then in every layer from 6 up. Eleven

species, seven of them bovid, as well as Delphinidae, are not present in the

BNK 1 sample. Species present in the BNK 1 sample but not in that from

Nelson Bay Cave are Ictonyx striatus, Loxodonta africana, Diceros

bicornis/Rhinocerotidae, Raphicerus campestris, Ovis aries, and Bathyergus

suillus.

Arctocephalus pusillus, of which only one doubtfully identified individual

was present in the ‘Robberg’, is otherwise the most common species at Nelson

Bay Cave and accounts for about 20 per cent of the site total, whereas at

BNK 1 it ranks sixth in order of frequency and accounts for only 2,5 per cent of

the site total. The bovid species account for about 42 per cent of the site total at

both sites. Of these, 24,8 per cent are in the ‘Robberg’ at Nelson Bay Cave,

29,4 per cent in the ‘Albany’ and 45,9 per cent in the ‘Wilton’ while at BNK 1

40,9 per cent are in the lower layers and 59,1 per cent in the upper. If

consideration is given to the fact that Nelson Bay has a longer ‘Robberg’

sequence than the comparable layers (19-18) at BNK 1, lacks the final ‘Albany’

156 ANNALS OF THE SOUTH AFRICAN MUSEUM

(or initial “Wilton’) and final “Wilton’ sequences, the differences in the frequen-

cles given above may have a reduced significance.

In terms of frequencies in the size classes used for the analysis of the

BNK 1 mammals, Neison Bay Cave provides the following rankings, from

highest to lowest (in the two cases where the classes are joined by ‘and’ this

indicates that they are of equal rank):

‘Robberg’: very small, small medium and large medium, large, small, very

large;

‘Albany’: small medium, small, very small and large medium, large, very

large;

‘Wilton’: small medium, small, very small, large, large medium, very

large.

It will thus be seen that the general trend is for the larger size classes to

become less frequent through time, although in the ‘Robberg’ it is the smallest

size class that predominates and in the ‘Albany’ and ‘Wilton’ there is only a

difference in the ranking of the large and large medium size classes and in each

case the three smaller size classes have the highest ranking.

The lower layers of BNK 1 may be divided into two groups on the basis

of size class ranking. Layers 19-16 have the very small and large medium size

classes in the first two rankings, with small, small medium or large ranking

third. Layers 15-11 have large medium ranking first and two of the three

smaller size classes ranking second and third. The major difference from the

‘Albany’ of Nelson Bay Cave is that there it is the small medium size class

that ranks highest. Where the upper layers of BNK 1 and the ‘Wilton’ of

Nelson Bay Cave are concerned the only difference is in the ranking of the

three smaller size classes, all of which rank higher than the three larger size

classes.

Klein (1974: 273-276) has suggested that the presence of Equus cf. quagga,

Damaliscus dorcas, Connochaetes cf. gnou, Antidorcas cf. marsupialis, and

possibly also the extinct Pelorovis sp. and Megalotragus sp. ‘clearly imply more

Open vegetation near Nelson Bay in Robberg times’. Although the ‘Albany’

deposits lack the predominantly grazing animals of the “Robberg’, Klein con-

siders that the presence of Phacochoerus aethiopicus and Taurotragus oryx

suggests ‘more open vegetation than the historic evergreen forest’. He also

observes that faunal assemblages from sites as far afield from Nelson Bay Cave

as Wilton and Elands Bay Cave reveal ‘a common emphasis on smaller

terrestrial animals, especially very small bovids’ in ‘Wilton’ times. Klein sees

this as being ‘in marked contrast to the preceding Albany and may reflect

end-Pleistocene/early Holocene environmental change leading to reduction in

suitable grazing for many large bovids over much of the Southern Cape’. The

increased frequency of small bovids the size of Raphicerus is evident in the

upper deposits of both sites, but appears more marked at BNK 1 because of the

dominance of Arctocephalus pusillus in both the ‘Albany’ and ‘Wilton’ of

Nelson Bay Cave.

BYNESKRANSKOP 1 ey

While Bathyergus suillus is the dominant small animal at both BNK 1 and

Die Kelders, at Nelson Bay Cave it is Procavia capensis, a fact no doubt

reflecting the lack of a suitable sandveld environment for Bathyergus suillus in

the vicinity of that site. Species such as Tragelaphus scriptus, Sylvicapra

grimmia, and Ourebia ourebi reflect a variety of environments in the vicinity of

the cave, but it is probably only Tragelaphus scriptus that might not have found

a suitable habitat in the area around BNK 1, although one individual is

recorded from layer 12 of Die Kelders (Schweitzer 1979, table 27). Dorst &

Dandelot (1972: 223) indicate that the present distribution of this species is

only east of Cape Agulhas.

The fauna from Klasies River Mouth (Klein 1976a) is mostly from M.S.A.

deposits and therefore not directly comparable with that from BNK 1, although

it is worth noting that the bulk of the bovids were from these levels. The

samples from the L.S.A. deposits in Caves 1, 1D and 5 (Klein 1976a, tables 1,

4) are too small to allow for more than the observation that the faunal list is

much the same as those for Nelson Bay Cave and BNK 1. Arctocephalus

pusillus and Syncerus caffer are the most common species in the L.S.A. levels

of Cave 1, followed by Alcelaphus buselaphus and Raphicerus melanotis, and

Arctocephalus pusillus is also the most common species in Caves 1D and 5.

The faunal sample from Melkhoutboom (H. J. Deacon 1976, table 35) is

much the same size as that from BNK 1. The ‘Robberg’ deposits (B unit) are

beyond the time range of BNK 1, and show a predominance of large grazers,

while the small bovid species are absent.

In the ‘Albany’ deposits RF and MBS the large grazers still predominate,

although Hippotragus and Alcelaphus are present only in the lower unit.

Damaliscus cf. dorcas is present only in this unit, in which several of the smaller

species first occur. The three highest rankings in these units in the size classes

for all mammals are: large medium, small medium, ard small, a situation

different from that for the lower layers of BNK 1.

In the “Wilton’ units at Melkhoutboom Equus cf. quagga is present only in

the lowest two units and is absent after about 7 000 B.p., about 1 000 years

before it disappears from the BNK 1 deposits. Raphicerus sp., probably

R. melanotis on the basis of Klein’s observations (see discussion at the end of

this section), is the most common species, followed by Tragelaphus scriptus and

T. strepsiceros. Syncerus caffer first appears in the lowest unit, WBM, and

Pelea capreolus in the overlying M unit whereas in the BNK 1 sample Syncerus

caffer is present from layer 19 and Pelea capreolus from layer 14. The three

highest rankings in the size classes are small, small medium, and very small.

The frequencies of the first two are almost equal and the distribution is thus

closer to that of the ‘Wilton’ layers from Nelson Bay Cave than to that from

BNK 1 layers 9-1.

Excluding Homo sapiens, the ‘undetermined antelope’ from the M unit

and Otomys sp. there are 11 species in the ‘Robberg’ at Melkhoutboom, and 21

each in the ‘Albany’ and ‘Wilton’. Of the total of 32 species, 22 (or the same

158 ANNALS OF THE SOUTH AFRICAN MUSEUM

genera, at least) are also present in the BNK 1 sample. Some of the species in

the Melkhoutboom sample, but not in that from BNK 1, have a preference for

a more closed habitat, e.g. Tragelaphus scriptus, T. strepsiceros, and Cephalo-

phus monticola, while others such as Sylvicapra grimmia and Ourebia ourebi

prefer more open habitats, suggesting a mosaic of vegetation types in the area.

The absence of any of the species from the various units at Melkhoutboom,

particularly in the upper units, may be attributable to chance: as at BNK 1 the

layers with the highest totals tend also to have the greatest variety of species. It

may, alternatively, reflect changes in the vegetation; but by and large the

presence of the species mentioned suggests an environment different from that

around BNK 1.

The faunal list from Wilton (J. Deacon 1972, table 4) reveals the exploita-

tion of a much smaller range than the sites hitherto discussed. Excluding the

micromammals, there are 2—7 taxa in the individual layers, compared with 26 in

BNK 1 layer or 23 in layer 5, and a total of 15 taxa are represented, compared

with 36 at BNK 1, 32 at Die Kelders, 40 at Nelson Bay Cave and 32 at

Melkhoutboom.

The Wilton faunal sample does not show a predominance of the very small

size class. The ranking of the size classes, from highest to lowest, is: small,

small medium, very small, and large medium. The large and very large size

classes are not represented, but otherwise the predominance of the smaller size

classes is similar to the pattern for the upper layers of BNK 1, the Holocene

deposits at Die Kelders, and the ‘Wilton’ deposits at Nelson Bay Cave and

Melkhoutboom. The Wilton faunal sample consists principally of small and

small medium bovids, in terms of Klein’s and Brain’s size classes and as such is

markedly different from that of the upper layers of BNK 1 in which, although

small bovids are the largest individual bovid size class, there is a higher

frequency of large medium and large bovids than of small medium, except in

layer 1; and the very small non-bovid species are also more common at BNK 1

than at Wilton.

The preliminary analysis of the fauna from Boomplaas (Klein 19785,

table 1) suggests much the same range of species as is present in the BNK 1

sample. Among the more marked differences between the two sites is the

persistence at Boomplaas of Equus zebra/quagga through to the terminal

deposit while at BNK 1 E. cf. quagga is not present after layer 9, some 4 000

years earlier. Potamochoerus porcus and/or Suidae-general occur only sporadi-

cally at Boomplaas until the lower ‘herder’ levels while at BNK 1 they are

present throughout the sequence, except in layer 11. Damaliscus dorcas/niro is

present at Boomplaas only in the lowest but one of the ‘Albany’ layers, while at

BNK 1 D. dorcas is absent only after layer 11, perhaps 2 000-3 000 years later

than at Boomplaas.

In the ‘Robberg’ levels at Boomplaas large medium bovids are the most

common of the bovid size classes, followed by large, and with small and small

medium ranking equal third. In the remaining levels small bovids are most

BYNESKRANSKOP 1 159

common, but in the ‘Albany’ large medium rank second and small medium

third, while in the ‘Wilton’ and ‘Herder’ levels this order is reversed. The

Boomplaas data seem to suggest that the introduction of domestic sheep into

the area affected the hunting, not of small bovids as might be expected but of

large medium and large ones, although these are not common, even in the

preceding ‘Wilton’ levels. BNK 1 layer 1 contains a sub-unit without Ovis aries

but the pattern for this layer, compared with the underlying layers, is for a

decline in the frequency of small and large medium bovids while the frequency

of small medium, in which Ovis aries is included, and large both increase.

When all species are considered, the Boomplaas sample indicates changes

in the three top ranking size class frequencies:

‘Robberg’: large medium, large, very small;

‘Albany’: small, large medium, small medium;

“Wilton’: very small, small, small medium;

‘Herder’: small, small medium, very small.

This shows that while the pattern of ranking for the ‘Wilton’ of Boomplaas

is identical to that of the upper layers of BNK 1 there are differences between

the two sites where the lower layers are concerned, since the ranking for the

lower layers of BNK 1 is large medium, very small, and small.

Klein (1978a, table 1) has also analysed the fauna from Buffelskloof.

Although the sample is small, maximally 146 individual mammals in twenty-five

taxa, the pattern of ranking of the size classes in the ‘Wilton’ levels is much the

same as at Boomplaas and BNK 1, although there is a relatively higher

frequency of large medium animals. These account for 16,0 per cent of the

‘Wilton’ total at Buffelskloof, 9,0 per cent at Boomplaas and only 4,4 per cent

of the total for the upper layers at BNK 1. This gain is largely at the expense of

the very small size class, which is some 12,0 per cent higher at Boomplaas and

BNK 1 than it is at Buffelskloof.

As is the case in the lower layers of BNK 1 the large medium size class

ranks highest in the ‘Albany’ layers at Buffelskloof. However, while at Buffels-

kloof the very small size class ranks second and the small third, in the lower

layers of BNK 1 the second and third rankings vary, from large medium

generally second in layers 19-16 and small or small medium third, while in

layers 15-11 the second and third rankings are generally filled from the three

smaller size classes.

In terms of relative chronology the fauna from Elands Bay Cave (Parking-

ton 1980b, table 1), also analysed by Klein, may be divided into two units,

layers 16-10 relating to BNK 1 layers 19-14/13, and layers 9-1 relating to BNK

1 layers 4/3-1. The Elands Bay Cave faunal list does not include Arctocephalus

pusillus, which is listed separately (table 3) but is included in the comparisons

that follow.

The three highest ranking size class frequencies for Elands Bay Cave are:

layers 16-10: small, very small, small medium;

layers 9-1: small medium, very small, small.

160 ANNALS OF THE SOUTH AFRICAN MUSEUM

In the selected BNK 1 layers the rankings are:

layers 19-13: large medium, small medium, small;

layers 4-1: small, small medium, very small.

This shows that while the only difference in the samples from the upper

layers of both sites is the order in which the three smaller size classes rank, and

the same is also true for both groups of the Elands Bay Cave layers, whereas in

the lower layers of BNK 1 the large medium size class is highest ranked but

does not appear in the Elands Bay Cave top rankings. There is, in fact, less

difference between the samples from the upper and lower layers of Elands Bay

Cave than between those of BNK 1.

The three smaller size classes account for 86,5 per cent of the total for the

lower layers at Elands Bay Cave but only 42,9 per cent of the total for BNK 1

layers 19-13. The large medium size class accounts for only 3,8 per cent of the

total for the lower layers at Elands Bay Cave but 37,5 per cent of the total for

those at BNK 1. In the upper layers the three smaller size classes account for

91,4 per cent of the Elands Bay Cave total and 79,6 per cent of the BNK 1

total.

Raphicerus spp. are the most common mammals at Elands Bay Cave,

followed by Bathyergus suillus and Arctocephalus pusillus. While Raphicerus

melanotis is the more common of the two species at BNK 1 and R. campestris

apparently only a later arrival some 4 000 years ago, R. melanotis is the more

common species in the lower layers of Elands Bay Cave, but after the hiatus,

i.e. from about 4 000 years ago, R. campestris is more common. Bathyergus

suillus is more common in the lower layers of Elands Bay Cave than in the

upper, with 72 per cent of the site total in layers 14-10 alone, while in the BNK

1 sample almost 82 per cent of the site total comes from layers 9-1. Elands Bay

Cave layer 12 yielded half the site total of Arctocephalus pusillus but apart from

this frequencies in the upper layers are higher than in the lower. This species

does not, however, assume the importance at Elands Bay Cave that it does in

the ‘Albany’ and ‘Wilton’ deposits at Nelson Bay Cave.

DISCUSSION

On the basis of the radiocarbon dates for BNK 1, layer 19 and probably

some of the overlying layers, up to but not including layer 14, fall within the

conventional dating of the terminal Pleistocene, while the remaining layers are

within the time range of the Holocene. The restriction of the single Equus ctf.

capensis to layer 19 is thus not unexpected since this animal is listed among

those considered to have become extinct by the end of the Pleistocene (Klein

1974, table 3; 1980: 265). At the other end of the chronological scale, the

restriction of Ovis aries to layer 1 is likewise not surprising, since domestic

sheep have not been recorded from archaeological deposits in the southern

Cape before about 2 000 B.p. (Schweitzer & Scott 1973, G. Avery 1974, H. J.

Deacon et al. 1978, Schweitzer 1979). Sheep remains found in the lowest

sub-unit of layer 1 are considered to be intrusive and the date of 1 880 + 50

BYNESKRANSKOP 1 161

years B.P. probably marks the earliest occurrence of these animals in the BNK 1

deposits.

Although the evidence of a single occurrence cannot be regarded as

conclusive, the restriction of Phacochoerus aethiopicus to layer 13 may be

related to the disappearance from the deposits of Damaliscus dorcas after layer

11 and Equus cf. quagga after layer 9. These three species are characterized as

having a preference for open country, grassland in the case of the last two

(Dorst & Dandelot 1972: 173, 228, 164 (for Equus burchelli); Klein

1974: 273-276). Their absence from the upper layers may thus reflect a change

in the vegetation of the area by about 6 000 B.p. Against this, however, must be

considered the identification of Damaliscus dorcas in the late Holocene deposits

at Die Kelders and the continued presence in the BNK 1 deposits, albeit in

reduced frequencies, up to layer 1 of the alcelaphine Connochaetes/Alcelaphus

and of Hippotragus spp. up to layer 4. These bovids are all characterized as

gregarious grazers with a preference for open or lightly wooded grassland, or

mixed bush and grassland in the case of one of the hippotragines (H. niger)

(Dorst & Dandelot 1972: 230-232, 218-222, 204-206). Hippotragus leucophaeus

and Alcelaphus buselaphus have been tentatively identified at Die Kelders

(Schweitzer 1979, table 27) and the problem therefore seems one of establish-

ing whether there has been a major vegetational change in the general Die

Kelders—BNK 1 area in the past 2 000 years or so (Avery’s (1979) vegetation

unit 1?) or where the prehistoric hunters hunted these grazers.

The apparent absence of Raphicerus campestris before layer 6, i.e. about

4 000 years ago, is of interest although it is not inconceivable that this species

may be represented in the otherwise unidentifiable Raphicerus spp. in the

underlying layers, since only four of the forty-one individuals in these layers

could be identified as to species. Pienaar (1974: 187) lists R. campestris as one

of the three most successful antelope species in southern Africa, ‘being able to

exploit a wide range of ecological situations, even those that have become

severely degraded by the activities of man’. Tinley (1969, fig. 3) shows R. cam-

pestris as occupying a much wider range of habitats than R. melanotis, and

Dorst & Dandelot (1972: 269, 275) show R. campestris as endemic to virtually

the whole of southern Africa, while R. melanotis is restricted to the southern

Cape.

Klein (1976b: 171) has noted that in modern populations R. campestris is

more common than R. melanotis in the western part of the southern Cape (St.

Helena Bay to False Bay), while in the southern and eastern parts (False Bay

to Cape St. Francis) R. melanotis is the more common species, and he has

elsewhere observed (Klein 1972a: 196) that R. melanotis is the only species now

to be found in the vicinity of Nelson Bay Cave. On the basis of archaeological

data, Klein (1976b: 181) has postulated that the present distribution and

relative frequencies of the two species had become established at least by the

late Holocene. The evidence from Elands Bay Cave indicates that both species

were present in the area by the end of the Pleistocene and that by about 3 500

162 ANNALS OF THE SOUTH AFRICAN MUSEUM

years ago R. campestris had become the more common species, while the

evidence from BNK 1 suggests that by about 4 000 years ago R. campestris may

have moved into an area apparently previously occupied only by R. melanotis.

The two species have somewhat different habitat preferences (Dorst & Dande-

lot 1972: 264, 266) and the indications from Elands Bay Cave and BNK 1 are

therefore that by about 4 000 years ago an environmental change had taken

place in the vicinity of both sites, and consequently possibly also elsewhere,

that created the more open habitat preferred by R. campestris. Whether this

change was the result of a degree of aridification or of human activity, e.g.

veld-burning, or a combination of these and perhaps other factors, is a matter

that cannot at present be determined.

That most of the species listed in Table 15 are present throughout the

BNK 1 sequence suggests, in view of the changes in their relative frequencies, a

change in the balance of the methods of procurement rather than in the

methods themselves. There are few tools in the assemblage, whether of stone

or bone, that can be associated specifically with hunting, and the restriction of

bone points and, probably, linkshafts to layer 6 and above is not consistent with

the changes in the faunal patterns indicated in Table 15 and Figure 39.

Similarly, changes in the relative frequencies of the stone artefacts, especially

backed pieces, do not correspond to the changes in the faunal patterns. Most of

the smaller animals are crepuscular or nocturnal, and non-gregarious, and it is

possible that these animals were caught in traps or snares. The larger animals

would perhaps have been hunted with bow or spear or, in the case of the very

large animals, trapped in pitfalls. The digging of pitfalls, and possibly also the

hunting of the larger animals, would have required communal effort while the

setting of traps and snares would not, and it may therefore be suggested that

the changes in the faunal patterns indicate changes in social patterns, in that

they indicate an increase in individualistic behaviour during the Holocene.

Whether this change in social behaviour can confidently be linked to changes in

the balance of resources as a result of environmental change is a matter that

must await external evidence in the form of independently derived palaeo-

environmental data.

Because Ovis aries is restricted to layer 1 at BNK 1 and frequencies in the

individual sub-units of this layer are so low, it is not possible to assess the effect

of the introduction of domestic animals into the area on the procurement

patterns of the cave occupants. The 2,3 per cent increase in the relative

frequency of animals in the small medium size class in layer 1, compared with

the frequencies for layers 2 and 3, cannot be considered significant, and

perhaps the most that can be inferred from the BNK 1 data in this respect is

that after about 1 900 B.P. an animal that was possibly more easily procured was

added to the range of those already available.

When the data for the BNK 1 mammal fauna are compared with those for

other sites in the southern and western Cape they show quite marked differ-

ences in the nature and chronology of the patterns of change, except in the

BYNESKRANSKOP 1 163

later Holocene, when there appears to have been a common emphasis on the

procurement of a greater number of smaller animals than was previously the

case. The probable explanation of these differences is that the environments of

the various sites have never been precisely similar. The ecological balance in

the various areas would therefore have been different and, even given an

overall climatic change, the rate and nature of the effects of that change would

have varied from region to region. The reason for the general similarity of

procurement patterns in the later Holocene cannot, however, be ascribed to

climatic change inducing a general homogeneity in the ecology of the regions in

which the various sites are located. It seems that there must be an overriding

factor that is probably cultural, possibly demographic, but such an assumption

cannot be tested until a greater body of information is available, from more

than single-site observations, on patterns of human distribution and land-use.

SHELLFISH

The differences between the shellfish samples from Die Kelders (Schweit-

zer 1979: 186-194) and BNK 1 are marked, although the two sites are only

some 10 km apart. The chronology of the Holocene deposits at Die Kelders is

short and approximates only that of the last sub-unit but one from the bottom

of BNK 1 layer 1, so that comparability is somewhat limited. However, even

when the layer 1 sample is compared with that from Die Kelders the difference

in the relative frequencies of the genera is great, as Table 25 shows.

TABLE 25

Shellfish: ranked frequencies of genera from BNK 1 layer 1 and Die Kelders.

BNK 1 % DK 1 %

VO ee ean ah 35,0 Choromytilus-Perna:......:.0..52. 66,9

OR SIA O Sle da ae 24,5 BUTRUDCN Os Were eons. sabia: ES

Choromytlus=Perna 3.23 ase ne SS 17,9. Bate llarsey erst ah nck Meike ese eke 653

[EITC DAU ee ere es es 11,4 On SICl Og aan oS ossiety ss ae e A ct (Osi

I ERTQUPS, ke eee ere hace ek ei ee ae Se? TUDO te ete swe eee Dali

Dim ODiGreee: orton Neb es Aches ee te So PIQUOUSCI ahr. awe ate Mee Cee 0,8

UD OMAN Wis lis oh Pond icte dikes ew wack 1,9 DORGGRE os Ve SA eetsetn i Hs SUS eS 0,3

[EILEDIS § Ree Re eR ee 0,3 ESTADO 5 hs OE a ee ee 0,2

DD OPIN ee cite ee AE oath et 0,04

(Die Kelders (DK 1) details after Schweitzer 1979, table 19.)

Human preferences apart, the most reasonable explanation of these differ-

ences seems to be the differing littoral topography which, in the vicinity of Die

Kelders, appears to offer a more favourable habitat for Choromytilus meridio-

nalis than does the coast east of Danger Point. With the exception of Bullia,

which scavenges in the surf zone of sandy beaches, all the gastropods listed in

Tables 18 and 25 inhabit rocky crevices or pools, many of them in the lower

tidal zones (Day 1969), and it seems likely that the steeper rocky shores in the

164 ANNALS OF THE SOUTH AFRICAN MUSEUM

vicinity of Die Kelders would have made these shellfish less accessible than on

the coast east of Danger Point, which is less steep and contains gullies and tidal

pools (cf. Fig. 5).

Although Day (1969: 1) states that Walker Bay is an area of warm water in

which there are more south coast species (than those of the colder waters of the

west coast), the composition of the Patella species from Die Kelders (Schweit-

zer 1979, table 21) is more representative of the west coast distribution, while

that from BNK 1 is closer to the south coast distribution although containing

aspects of the cold-water distribution (cf. Branch 1971, figs 3-5); (see also

Schweitzer 1979: 111). P. granatina dominates the Die Kelders Patella sample

and the P. granatina: P. oculus ratio is 6:1. The BNK 1 layer 1 Patella sample

is dominated by P. longicosta, which is absent from the Die Kelders sample,

and the P. granatina: P. oculus ratio 1s less than 1: 2.

G. Avery (1976: 125-126), who has studied shell middens at Pearly Beach,

south-east of BNK 1, and at Hawston to the north-west, has preferred to

characterize middens on the basis of the dominant shellfish represented in

terms of ‘meatmass’ rather than simple numerical frequency. Not surprisingly,

the middens are consequently primarily characterized as Haliotis or Turbo

middens, or a combination of, or variations on, these. Haliotis midae and

Turbo sarmaticus are among the largest shellfish to be found on the South

African coast and have an average flesh mass 4-60 times greater than that of

genera such as Oxystele and most of the Patella species (cf. G. Avery 1976,

table 5). Additionally, however, Avery (1974: 112; 1976: 111 ff.) has character-

ized some of the middens as Oxystele—Patella-Turbo which apparently reflects

the numerical ranking of the species rather than ‘meatmass’ dominance since

Haliotis and/or Turbo are always the dominant genera in terms of flesh mass

(cf. G. Avery 1976, table 7).

With the exception of layer 3, in which Oxystele is numerically dominant

but exceeds Turbo sarmaticus by a ratio of less than 2:1, all the shell samples

in the upper layers of BNK 1 can be considered as Turbo dominated. This

Avery (1976, table 7) found to be unequivocally the case in only one of the

eight sites he studied, although one (PB 1) showed a change from Turbo

domination in the lower part of the midden to Haliotis domination in the

upper.

Although no quantitative analyses were carried out of the shellfish from

Oakhurst (the material was, in fact, not taken from the site), Goodwin

(1938: 322-323) mentioned that the ‘Smithfield B’ deposit contained mainly

oyster and razor shells (Ostrea and Solen spp.), while in the ‘Smithfield C

deposits a change was noted from Donax serra in the lower levels to Choromy-

tilus meridionalis in the upper (but see also p. 314, where Goodwin remarks

that although there were C. meridionalis shells in the 36-45 inch (91-114 cm)

level, they form a negligible part of the whole deposit’). With regard to the

‘Wilton’ Goodwin (1938: 323) observed only that ‘throughout the Developed

Wilton the basis of subsistence is mainly shellfish, fish, and animals’. Unless a

———————

BYNESKRANSKOP 1 165

ranking of the frequency with which the three taxa occurred is implied (and the

omission of plant foods from the list may or may not have any significance), the

observation has little value. Elsewhere, Goodwin (1938: 305) lists the molluscs

found in the upper 9 inches (23 cm) of the deposit. These are not quantified,

and the most that can be said is that the list contains most of the genera found

at BNK 1. Donax serra is numerically dominant in BNK 1 layers 14-11, which

may be relatable to the ‘Smithfield’ deposits at Oakhurst, but the frequencies of

all the genera in the lower BNK 1 layers are too low for this to be considered

meaningful. In layer 10 Turbo sarmaticus assumes its dominant role, suggesting

a marked difference from the Oakhurst ‘Wilton’.

Klein (1972a, fig. 4) indicates that at Nelson Bay Cave marine shell was not

found in the deposits below the base of the ‘Albany’, but J. Deacon (1978: 89)

has reported its presence in small quantities in layer BSL, the uppermost of the

‘Robberg’ layers, from which frequencies increase through time. Although this is

similar to the pattern at BNK 1, there is a great difference between the times

when this occurred at the two sites. Nelson Bay Cave layer BSL is dated at

11 950 + 150 years B.p. (Klein 1972a: 202), which approximates the date of BNK

1 layer 19 and, while marine shell is present in layer 19 and thereafter, it is only in

layer 10, some 5 500 years later, that the occurrence of marine shell can be said

to indicate an increasingly important food resource.

The Nelson Bay Cave shellfish sample is dominated by Perna perna

(Choromytilus meridionalis in the lower layers), with Patella ranking second.

This mussel-dominated sample is more like that of Die Kelders than that of

BNK 1, although at Die Kelders Patella spp. rank after Choromytilus—Perna

and Burnupena spp.

J. Deacon (1979: 78) reports that while there was no quantification of

marine shell from the Matjes River Shelter, one of the largest shell midden

deposits in the southern Cape, a change was noted, from Donax serra being

more common in layers D and C to Choromytilus meridionalis in layers B and

A. The top of layer C is dated to 5 400 + 250 years B.p. and the B—A interface

to 3 555 + 35 years B.P. (J. Deacon 1979: 74) so that the Matjes River deposits

probably cover the same span of time as the upper layers of BNK 1, although

Sampson (1974: 263-269) has indicated that layer D, at least, is ‘pre-Wilton’

and has included the lithic component in his ‘Oakhurst industry’. The sparse

information available suggests that the shellfish component of Matjes River

Shelter may be more similar to that of Die Kelders, Nelson Bay Cave, and

Oakhurst than to that of BNK 1.

Voigt (1975, table 1) has shown diagrammatically the shellfish frequencies

for the ‘pre-pottery L.S.A.’ and ‘pottery L.S.A.’ levels of the Klasies River

Mouth cave sites. Although the sites have a record of shellfish exploitation

extending back to the Middle Stone Age, the levels mentioned are the only

ones chronologically comparable with BNK 1. These levels are approximately

dated to between 4 800-2 200 B.p. (Singer & Wymer 1969; Butzer

1978: 146-147) and are thus contemporary with that part of the BNK 1 deposit

166 ANNALS OF THE SOUTH AFRICAN MUSEUM

from the lower part of layer 1 to somewhere between layers 5 and 9. Although it

is not clear on what evidence the superpositioning of the sequences in the various

caves in Voigt’s figure 1 is based, the indications are that the L.S.A. levels are

generally dominated by Patella, principally P. longicosta, but there is a marked

increase in the frequency of Oxystele in the uppermost, pottery-bearing layers of

Cave 1D. Patella longicosta is always the predominant Patella species in the

upper layers of BNK 1, but the genus is always subordinate to Turbo and, from

layer 4 up, Oxystele. In terms of flesh mass, however, Turbo sarmaticus is always

the dominant species in the Klasies River Mouth L.S.A. levels, as it is in the

relevant BNK 1 layers. In general, the Klasies River Mouth L.S.A. shellfish

samples show a greater similarity to those from the upper layers of BNK 1 than

do either of these to the samples from the other sites discussed.

The Bonteberg Shelter on the west coast of the Cape Peninsula (Fig. 1)

represents the earliest local attempt at systematic analysis and characterization

of shell midden deposits (Maggs & Speed 1967). The deposits from this shelter

were dated from marine shell samples to about 4 500-2 000 B.p. (Grindley et al.

1970) but since radiocarbon dates from marine shell tend to be older than those

from charcoal (Klein 1972a: 202-203; Vogel & Visser (1981: 43) suggest about

400 years), the Bonteberg deposits may be considered approximately coeval

with those from BNK 1 layer 5 to the lower part of layer 1.

The Bonteberg shellfish sample, although small, is essentially dominated

by Patella spp. but shows a shift in the top layer, 2b, towards Oxystele and

Burnupena. The totals for Patella granatina and P. oculus have been combined

but these are by far the most common species in the samples (69-84 % of the

layer totals for Patella) suggesting that a higher littoral zone was being

exploited than at BNK 1, where P. longicosta is the dominant species (cf.

Branch 1971, figs 3-4). Choromytilus meridionalis accounts for about 2-10 per

cent of layer totals, but increases almost threefold in layer 2b, a situation

similar to that in BNK 1 layer 1.

Buchanan (1977) carried out a rescue operation in a small shelter at Hout

Bay (Fig. 1). Dates of 1 840 + 50 years B.p. for the top of the bottom layer and

1 460+50 years B.p. for the middle of the top layer (Buchanan 1977: 14)

suggest that the whole of the occupation of the site falls within the time range

of BNK 1 layer 1.

As Buchanan’s table 3 indicates, Choromytilus meridionalis is dominant in

each of the five layers, with frequencies of 45-56 per cent of individual layer

totals. Patella spp. (19-24 %) rank second in layers 5, 4 and 2, while Burnupena

frequencies (16-26%) rank second in layers 3 and 1. Patella granatina is the

most common Patella species, followed by P. granularis in layers 5-3, and P.

cochlear in layers 2 and 1. The species representation suggests exploitation of

littoral zones higher than those indicated by the BNK 1 sample, and not unlike

those indicated by the Bonteberg sample. Deeper water species such as Turbo

sarmaticus, Haliotis midae, Patella argenvillei, and Argobuccinum argus are

rare in, or absent from, the layer samples.

BYNESKRANSKOP 1 167

Buchanan’s table 5 gives means and standard deviations for lengths of

Patella granatina, P. granularis and P. cochlear from the site. Sample sizes are

small and in most cases the standard deviation exceeds the difference between

the highest and lowest means. It would seem unwise, in the circumstances, to

use such data as the basis for speculations regarding the fluctuating mean

lengths as being indicative of the ‘farming down’ of the local shellfish popula-

tions through human predation.

Buchanan’s table 6 shows that the mean lengths of Patella granatina and P.

granularis from archaeological sites are 25-36 per cent smaller than those of

modern samples. This has led Buchanan (1977: 28) to support the suggestion by

Parkington (1976: 134-135) that intensive predation reduced the average life-

span and thus the average size of the shellfish without, however, endangering

species survival. Buchanan continues by observing that since the size reduction

is evident in the first occupation layer ‘the predation/survival equilibrium must

have been reached prior to the first occupation’.

Although the Hout Bay Shelter samples are chronologically comparable

with those from only some of the sub-units of BNK 1 layer 1, the fact that there

is very little change in the mean length of most Patella species from BNK 1

layer 9 up can similarly be used to suggest that a ‘predation: survival equi-

librium’ had been reached in the BNK area from the earliest stages of intensive

shellfish collecting. This in turn prompts the hypothesis that the marked

increase in shellfish representation in the upper layers of BNK 1 is relevant to

the site only and that, in fact, intensive predation of the local shellfish

populations had been in progress prior to about 6 500 B.P. but that the bulk of

the ‘catch’ was not being brought back to the site. Support for such a

hypothesis would be provided by the discovery of shell middens or other sites in

the area with high frequencies of marine shell that could be dated to earlier

than 6 500 B.P., but such sites have not yet been found. A marine transgression

commencing at about 6 500 B.P. (see p. 10) could have destroyed earlier coastal

middens but ought to have left others on the edge of the then coastline(s),

which might have been covered by the seaward advance of the dunes and

vegetation during the subsequent regression, but it seems unlikely that evidence

of these would not have been found before now.

Robertshaw (1977, 1978, 1979) excavated shell middens at Paternoster,

Stofbergsfontein, and Duiker Eiland on the Cape west coast (Fig. 1). The dates

obtained for the sites are all within the range of BNK 1 layer 1. While there

was a marked increase in the frequency of Choromytilus meridionalis in the

uppermost layer of an otherwise Patella-dominated deposit at Paternoster

(Robertshaw 1977, table 4), at Stofbergsfontein the main unit was dominated

by Patella species but in the subsidiary unit Choromytilus was dominant

(Robertshaw 1978, table 4), and at Duiker Eiland Patella was always dominant

(Robertshaw 1979, table 5).

In the sample from Paternoster layer 1 the mean lengths of Patella

granularis and P. granatina are 2-3 mm and 7-11 mm greater than those of the

168 ANNALS OF THE SOUTH AFRICAN MUSEUM

underlying layers (Robertshaw 1977, table 5). From this, Robertshaw

(pp. 67-68) has reasoned that the increase in the sizes of the two species in

layer 1 might have resulted from a lower rate of predation of Patella species

consequent on the increased exploitation of Choromytilus meridionalis in this

layer. The two units of the Stofbergsfontein deposit could not be related

stratigraphically (Robertshaw 1978: 142) so that the differences in the mean

lengths of the two Patella species (less than 1 mm and 3,5 mm respectively)

cannot be interpreted. For the three layers of the Duiker Eiland sample

(Robertshaw 1979, table 6) Patella granularis has a maximum variation in mean

length of 2 mm and P. granatina of just over 4 mm. Although the mean lengths

for layers 1 and 2 are about 2-4 mm higher than those for layer 3 the small

variation suggests little change in the structure of the populations being

exploited, and the same is true for at least P. granularis in the Paternoster

samples.

Robertshaw (1977: 67, 1979: 9) has suggested that the mean lengths of

these two species are small when compared with modern population data given

by Branch (1974) and Buchanan et al. (1978), and elsewhere (Robertshaw

1978: 143) he observes that ‘the mean sizes ... seem to reflect a rate of

predation high enough to have had the effect of farming down the local

populations of these species . . .’. It must be pointed out, however, that Branch

(1974a or 19746) does not give mean lengths for the modern Patella populations

he studied but provides a series of histograms showing the frequency distribu-

tion of the population over its total size and age range. Calculation of the mean

length of each of the two species under discussion from Branch’s histograms

(Branch 19745, figs 14-15, see also Fig. 42 herein re P. granatina) yields a

mean length for P. granularis of 13,9 mm for the whole population or 26,1 mm

for those individuals 20 mm or longer, and for P. granatina 37,0 mm for the

whole population or 44,4 mm for the larger individuals. The higher mean for

P. granularis is in each case lower than those for the archaeological samples:

10-13 mm in the case of Paternoster, 14-15 mm in the case of Stofbergsfontein,

and 10-12 mm in the case of Duiker Eiland. Where P. granatina is concerned,

the mean for the larger modern population is again /ower than those of the

archaeological samples: 2-13 mm lower than Paternoster, 9-13 mm lower than

Stofbergsfontein, and 10-13 mm lower than Duiker Ejiland. In the BNK 1

P. granatina samples it is only that from layer 3 that approaches the mean

length of the larger modern population, the means from the other layers being

5-10 mm higher (Fig. 42).

The mean lengths of the modern samples given by Buchanan et al. (1978,

fig. 3) are stated to be those of the largest available individuals that could be

measured in ten minutes, and the same is probably true of the samples

measured by Olivier (1977, cited by Buchanan 1977, table 6). The motivation

for these sampling procedures is (partly) understandable but they do not

provide an accurate reflection of the total population structure and therefore do

not, any more than Branch’s data, support Robertshaw’s suggestion that the

BYNESKRANSKOP 1 169

mean lengths of the archaeological samples are small when compared with

those of modern populations, nor can they be said to indicate ‘farming down’.

What they do indicate, rather, is that prehistoric shellfish collectors had a

preference for the larger-sized individuals but that they were not averse to

taking individuals of whatever size were available and, presumably, that they

considered worth the effort of prising off the rocks.

The available data on shellfish from Elands Bay Cave (Parkington 1979,

table 5) are sparse, but indicate that while Patella spp. make up the bulk of the

shellfish in most of the lower layers there was a marked increase in the

frequency of Choromytilus meridionalis in the layers after the hiatus, also of

‘whelks’ in layers 1 and 2. Layer 12 is anomalous, as in several other instances

previously mentioned, in having a higher frequency of whelks than of mussels,

while in layer 1 it appears that there was a more equal distribution of the three

main taxa. Although in the BNK 1 sample Turbo sarmaticus remains the

dominant species in layer 1, there is a marked increase in the frequencies of

Choromytilus meridionalis, Oxystele sinensis, and Burnupena spp. relative to

the lower layers, and this parallels to a certain extent the situation at Elands

Bay Cave.

DISCUSSION

Although it is evident from Table 18 that shellfish were brought to BNK 1

from the initial occupation of the site some 12 500 years ago, it is not until

about 6 000 years later that there is any clear evidence of the exploitation of

shellfish as an increasingly important food resource. That the systematic exploi-

tation of shellfish appears to have begun towards the end of the Pleistocene at

sites like Nelson Bay Cave and Elands Bay Cave may be related to the

increasing proximity of the sea to those sites from this time on. It does not

seem, however, that this explanation can be applied to the apparent lateness

with which this practice started at BNK 1 since, as shown during the discussion

of the topography of the BNK area, the shore is not likely to have been more

than 10 km from the cave at any time during the occupation of the site.

Possible explanations are that prior to about 6 500 B.P. the availability of land

game was sufficient for the human population of the area without the need to

exploit possibly less favoured resources; or that from about 6 500 B.p. changes

in demographic patterns resulted in the more frequent use of the cave, either as

a base for stays of longer duration or as a transit camp en route from this part

of the coast. However, although the lack of precise data leaves much to be

desired, the indications from sites such as Oakhurst and Matjes River suggest

that there was a general increase in the exploitation of shellfish at about the

same time as this occurred at BNK 1.

The frequency with which the various species are represented at the

different sites is undoubtedly primarily related to the ecology of the littoral

zones in the vicinity of these sites, although factors such as accessibility and

preference cannot be ignored. The relative scarcity of Haliotis, particularly

170 ANNALS OF THE SOUTH AFRICAN MUSEUM

H. midae, at BNK 1 and Die Kelders is not easy to explain. This shellfish has

the highest ratio of flesh mass to total mass of all the genera represented at

these sites and occupies the same littoral zone as Turbo sarmaticus, and there is

abundant evidence all along the coast from Pearly Beach to Die Kelders that

Haliotis midae was heavily exploited. The fact that there are shells in the

deposit at BNK 1 tends to diminish the possibility that the flesh was removed

and the shells left at the coast. A. B. Smith (1981 pers. comm.) has suggested

that there may be a temporal factor involved, but this could only be tested by

obtaining a good range of dates from concentrations of H. midae.

The low frequency of Donax serra is also not readily explained. The white

mussel is fairly easily obtainable by digging in the sand at low tide and has a

greater flesh mass in proportion to its size than either the black mussel

Choromytilus meridionalis or the brown, Perna perna. These species are,

however, also not well represented in the BNK 1 samples, and it may simply be

that along the coast east of Danger Point the collecting of Turbo sarmaticus was

more profitable, and Choromytilus-Perna in the Die Kelders area.

The increase in the frequencies of Oxystele sinensis and Burnupena spp.

may be seen as reflecting the need to augment the food supply as increasing

demands on the larger species thinned them out, although there is not the

proportionate increase in the frequencies of these smaller species that might be

expected in such a circumstance. It seems, though, more than mere coincidence

that similar patterns are evident in the shellfish samples from sites as far afield

from BNK 1 as Klasies River Mouth and Elands Bay Cave. As indicated in the

discussion of Patella lengths, the archaeological data do not appear to lend

themselves to the deriving of inferences regarding the effects of human

predation on the ecology of local shellfish populations or regarding changes in

past sea temperatures although the potential, based on a more systematic

approach on an interdisciplinary level, should not be underestimated. .

FISH

As at BNK 1, Pachymetopon blochii is the most commonly represented

fish at Die Kelders (Schweitzer 1979, table 23) but whereas at BNK 1 it

accounts for only 37,4 per cent of the site total or 33,3 per cent of the layer 1

total, it accounts for 96,4 per cent of the Die Kelders total. Lithognathus

lithognathus and Rhabdosargus spp. are next in order of frequency but their

ranking is reversed in the Die Kelders sample and their frequencies much lower

than at BNK 1. The differing frequencies probably reflect the differing off-

shore environments in the vicinity of the two sites, with the coastline near BNK

1 offering a wider range of habitats than that around Die Kelders. The Die

Kelders sample, however, contains minimally seven genera, only two less than

the BNK 1 sample.

Goodwin (1938: 323) noted that at Oakhurst ‘at the Wilton level begins a

marked increase in fish-bone, suggesting that efficient means of catching fish

had been evolved’. If the date of 7 910 + 70 years B.P. (Pta—377) obtained from

BYNESKRANSKOP 1 171

just above the ‘carbon floor’ is ‘consistent with the dating of the base of the

Wilton at other Cape sites’ (J. Deacon 1979: 36), then the effective exploitation

of fish began at Oakhurst at a time closer to the date of BNK 1 layer 12, some

1 500 years earlier than layer 9, in which the relatively large increase in fish

remains first becomes evident.

Poggenpoel (1979) has recently re-analysed the fish remains from the

1964-71 excavations at Nelson Bay Cave by Inskeep and Klein. The five upper

layers excavated by Inskeep that provided the data for Poggenpoel’s table 2 are

all dated to less than 3 000 B.p. (Inskeep in J. Deacon 1979: 64), while those

excavated by Klein that provided the data for Poggenpoel’s table 3 are dated to

between about 5 000 to 18 000 B.p. (Klein 1972a: 202; J. Deacon 1978: 84,

fig. 2). The upper group of layers all fall within the time range of BNK 1 layer 1

and the rest, up to layer BSL, are within the range of dates for BNK 1 layers

5-19.

Perhaps the major difference between the two sites where fish remains are

concerned is in frequencies: over 15 000 individuals have been identified from

Nelson Bay Cave as against only 265 from BNK 1. Fish remains first appear in

the Nelson Bay Cave sequence in layer BSL, the uppermost of the ‘Robberg’

units. The “‘Robberg’ yielded 0,1 per cent of the site total, the ‘Albany’ about

9,0 per cent and the remaining 91,0 per cent came from the ‘Wilton’. BNK 1

layer 19 yielded 0,4 per cent of the site total (there were no fish remains in

layer 18), layers 17-10 4,5 per cent, and layers 9-1 the remaining 95,0 per cent.

Although there were more fish in the ‘Albany’ layers than in BNK 1 layers

17-10 the chronological distribution at both sites is similar.

Another major difference between the two sites is in the number of genera

represented. In the BNK 1 sample only 9 genera have been identified, while in

the Nelson Bay Cave sample there are at least 17, taking the Clinidae/Blennii-

dae to represent only one genus each although they are actually families

containing 16 and 7 genera respectively (Smith 1949: 342-346, 350-358). At

BNK 1 the number of genera ranges from 1 to 3 per layer in the lower layers

and 3 to 7 in the upper, indicating a broadening of the resource base. At

Nelson Bay Cave the number of genera increases from 3 in the ‘Robberg’ to

3-12 in the ‘Albany’ and 10-17 in the ‘Wilton’. In passing, it should be noted

that layer Rice B (Poggenpoel 1979) or RB (J. Deacon 1978) contains 12

genera and is thus more typical of the ‘Wilton’ layers than of the ‘Albany’ in

which Deacon has included it, since the other layers in this unit contain only

3-7 genera per layer.

A further point of difference between the two sites is that while BNK 1

shows a clear pattern of increase only from layer 10 to layer 5, after which

frequencies decrease until the end of the sequence, the Nelson Bay Cave

sample shows a broad pattern of increase throughout the sequence. The lower

‘Wilton’ layers contain more than twice as many individuals as the ‘Albany’ and

the upper ‘Wilton’ more than three times as many as the lower ‘Wilton’ and

almost eight times as many as the ‘Albany’. The possibility of different

GA ANNALS OF THE SOUTH AFRICAN MUSEUM

excavated volumes cannot be discounted but it seems likely that the broad

pattern of increase would override the quantitative bias caused by these.

Also, whereas at BNK 1, possibly because of the low layer frequencies,

there is no discernible change in the relative frequencies of the principal

genera, changes are evident throughout the Nelson Bay Cave sequence. Sparo-

don durbanensis is generally most common in the ‘Albany’, Diplodus sargus in

the lower ‘Wilton’ and Clinidae/Blenniidae in the upper.

The indications are, therefore, that at Nelson Bay Cave fishing became

increasingly important through time and although frequencies relative to exca-

vated volume (ADs) would provide a more accurate picture of the importance

of fish in the dietary economy of the cave inhabitants, on the present evidence

it seems that fishing was a major industry for the people of Nelson Bay Cave,

which does not seem to have been the case for the people of BNK 1.

Poggenpoel’s paper also contains an analysis of the fish remains from

Elands Bay Cave. These first appear in the second layer (19), estimated to date

at some 15 000-17 000 years B.P. Poggenpoel’s table 4 gives the total number of

identified individuals as 1 306, almost five times as many as the total from BNK 1.

Elands Bay Cave layer 12 is again anomalous in that it contained 63 per cent

of the site total, but otherwise individual layer frequencies are low, ranging

from 0,8 to 5,4 per cent of the site total (with 20 layers the mean frequency

based on an even distribution would be 5%). Beyond the fact the frequencies

in the layers above layer 15 are higher than those in the layers below, there is

no evident pattern of increase through time the way there is at Nelson Bay

Cave. Elands Bay Cave layer 15 falls within the time range of BNK 1 layers

19-15.

Lithognathus lithognathus is generally the most common species in the

Elands Bay Cave sample, followed by Rhabdosargus globiceps but in layer 13

Liza richardsoni accounts for sixty of the seventy-one individuals, and this is

the only species present in layers 16-19. Pachymetopon blochii, the most

common fish in the BNK 1 sample, accounts for less than one per cent of the

Elands Bay Cave sample.

Again counting the Clinidae and Blenniidae as representing only one genus

each, there are 14 genera in the Elands Bay Cave sample as against 9 in that

from BNK 1. Individual layer frequencies range from 1-3 genera in the lower

layers and 3-7 in the upper. Although there are temporal differences between

the layer groups designated ‘lower’ and ‘upper’ at the two sites, there is a

greater similarity in the restricted numbers of genera at Elands Bay Cave and

BNK 1 than there is between these two sites and Nelson Bay Cave.

DISCUSSION

Fish do not appear to have been as important a factor in the diet of the

occupants of BNK 1 as they were to the inhabitants of Nelson Bay Cave, where

fish can be considered a primary food resource that became more important in

the course of time. Fish seem to have been important to the people of Elands

BYNESKRANSKOP 1 173

Bay Cave to a greater degree than to those of BNK 1, to perhaps the same

degree as to those of Die Kelders, but to a much lesser degree in all three cases

than to the people of Nelson Bay Cave. Because of the hiatus in the Elands

Bay Cave sequence it is not possible to determine when fish became more

important than they were in the late Pleistocene, but it is evident that at BNK

1, Nelson Bay Cave, Elands Bay Cave, and possibly Oakhurst, fish were more

important in those parts of the deposits referable to the ‘Wilton’ than in the

preceding period. Why fish should have become less important to the occupants

of BNK 1 after about 4 000 B.p. and more particularly in layers 2 and 1 is a

problem to which there is no ready solution.

Another problem is how the fish were caught. If the bone artefacts from

BNK 1 layers 15-13 that have been termed ‘fish gorges’ (Fig. 31) really were

such, they do not seem to have been very effective since these layers yielded a

total of only four fish. The absence of these artefacts from the upper layers of

BNK 1 (though not from those of Nelson Bay Cave), in which fish remains

become more frequent, suggests that other, more efficient fishing methods had

been devised. Poggenpoel (1979) has suggested that terminal Pleistocene—early

Holocene environmental changes may have been responsible for at least some

of the changes in the fish sample from Nelson Bay Cave while in the later

Holocene the causal factor may have been technological.

Goodwin (1946), in an article on prehistoric fishing methods in South

Africa, apparently concluded that the bulk of the fish caught, in ‘Wilton’ times

at least, came from tidal fish traps. However, he quoted (pp. 139-140) the

statement of a correspondent that a local fisherman at the Gouritz River mouth

(Fig. 1) had found ‘stuck away in a hole a fishing line made of a certain wild

vine of fibrous texture. This had been shredded and turned into fishing line,

and the hook was a bone tied in the middle and sharpened on each side’.

Goodwin follows this statement by commenting that no such bones had yet

been observed from midden deposits. This is perhaps not surprising in view of

the apparent restriction of ‘fish gorges’ to deposits dated to the early ‘Wilton’ or

earlier, although Goodwin might thus have found them at Oakhurst.

G. Avery has recorded the presence of a tidal fish trap at Franskraal,

between Uilkraalsmond and Danger Point, and his details of catches from

another trap, at Die Dam, a few kilometres east of Pearly Beach, include

virtually all the genera recorded from BNK 1 (G. Avery 1975, fig. 1, tables

1-2). However, Mugilidae comprise about 70 per cent of the catches listed in

Avery’s table 2, while in the BNK 1 sample this family is represented by only

three Liza richardsoni, and the most common fish in the BNK 1 sample,

Pachymetopon blochii, is absent from the Die Dam catches. This seems to

argue against the use of fish traps as the principal means of catching the fish

represented in the BNK 1 sample. It may be that Pachymetopon blochii was

caught on lines, using hooks other than ‘fish gorges’, and Lithognathus spp. are

historically recorded as having been speared in the surf at False Bay by the

Hottentots (Moodie 1838: 93).

174 ANNALS OF THE SOUTH AFRICAN MUSEUM

Day (1970: 219), writing of the marine biology of False Bay, observed that

Johnius (now Argyrosomus) hololepidotus and Lithognathus lithognathus con-

gregate on the shallow, sandy shores of the bay during the summer. Liza

richardsoni is among their prey, hunted in the surf. Pomatomus saltator, a

warm-water fish, arrives in the bay by September and forms dense shoals by

midsummer. Lithognathus lithognathus is the only species of the four that is

present in the BNK 1 samples in anything approaching a relatively high

frequency, but individual layer frequencies, even in layer 5, are low. The

topography of the coast east of Danger Point is very different from that of False

Bay and the total fish sample from BNK 1 is so small that it would seem unwise

to use information such as that given by Day in an attempt to deduce the

season during which BNK 1 might have been occupied.

REPTILES

In the Die Kelders faunal sample (Schweitzer 1979, tables 26-27) the dune

mole-rat Bathyergus suillus outnumbers tortoise by a ratio of 4:1, while in the

BNK 1 sample the position is reversed, with tortoise outnumbering dune

mole-rat by almost 11:1. In the case of BNK 1 layer 1, within the time span of

which the Die Kelders Holocene deposits fall, the ratio is almost 8:1. This is

probably a reflection of the differing environments of the two sites, and the

higher frequency of tortoise at BNK 1 suggests the greater ease with which

these can be procured, as dune mole-rats are abundant in the area today.

The differing frequencies of tortoise at BNK 1 and Die Kelders may also

reflect differential scheduling of activities: the people of Die Kelders might

have had more time to devote to the trapping of dune mole-rats; they might

have been there at a time of year when dune mole-rats were more plentiful

than tortoises (winter: Schweitzer 1979: 206); or they might have had a greater

need of, or use for, the pelts of these animals as well as their flesh.

The tortoise remains from Elands Bay Cave have been quantified only in

terms of mass per cubic metre (Parkington 1979, table 4), so that no direct

comparison with BNK 1 is possible. It appears, however, that the highest

frequencies are in Elands Bay Cave layers 16-12, suggesting a greater exploita-

tion of tortoise in the period 12 500—9 600 B.P., whereas at BNK 1 the period of

relatively greater exploitation was 9 800-6 100 B.pP.

GENERAL SUMMARY

In the preceding analyses and inter-site comparisons the various com-

ponents of the BNK 1 assemblage have been dealt with individually and sum-

marized on the basis of the major components. This general summary attempts

to draw these components together and to group the layers in a chronological

sequence according to what appear to be their common characteristics.

Table 26 presents those aspects of the assemblage considered to be most

relevant to this summary. It will be immediately apparent that there is no point

BYNESKRANSKOP 1 175

at which an absolute division can be made between any of the layers or groups

of layers across the whole range of components and that the division of the

layers into ‘phases’ is thus arbitrary. It is also self-evident that not all the

components have equal importance in the evaluation of change and that

changes in frequency may be less important than changes in shape or style, or

even presence or absence. The changes in the frequencies of stone raw

materials, for example, have reduced significance when it is recalled that 71 per

cent of the utilized artefacts and 85 per cent of the retouched pieces are

silcrete. Similarly, the ranking of the mammal size classes becomes less impor-

tant when it is remembered that these are based on numerical frequency rather

than on total live mass, even though the size classes themselves are based on

live mass. As an example, the eight individuals in the top-ranked very small

size class in layer 16 have an estimated total live mass of less than 80 kg while

the four individuals in the second-ranked large medium size class have an

estimated total live mass in excess of 400 kg.

Phase 1

Layers 19 and 18 are characterized by a predominance of quartz as a raw

material and higher frequencies of stone artefacts than the overlying layers.

Layer 19 has the most unmodified and utilized blades (mostly silcrete) and

blade cores (mostly quartz ‘micro-blade’) and yielded two backed blades. Layer

18 has few unmodified blades and blade cores, more utilized blades than flakes

but no backed blades and a very high frequency of scaled pieces. Utilized stone

artefacts: scaled pieces, flakes and blades, are more common than retouched

pieces, which are mostly scrapers. There are very few bone artefacts, none of

marine shell and very low frequencies of ostrich egg-shell beads. Layer 19 has

part of an ostrich egg-shell ‘flask’ opening and layer 18 a fragment of decorated

shell.

These layers share with layers 17 and 16 a similar ranking of the mammal

size classes with the very small size class numerically most common and the

large medium ranked second. The large size class ranks third in layer 19 but

shares this position in layer 18 with the other two smaller size classes, suggest-

ing a transition to the rank pattern of layers 17 and 16, in which this size class is

absent from the top three rankings. They share with layers 17-13 the lowest

frequencies of marine shell, with layers 17-11 the lowest frequencies of fish,

and with layers 17-10 a predominance of grazing ungulates over browsers or

‘mixed feeders’.

On the basis of the stone artefact frequencies it may be suggested that

during phase 1 the cave was relatively more intensely occupied than during

phase 2.

Phase 2

Layers 17-13 are characterized by a low density of stone artefacts in all

three categories, although layer 13 has a higher frequency of retouched pieces

TABLE

176

ANNALS OF THE SOUTH AFRICAN MUSEUM

Summary of artefactual and faunal data from BNK 1.

Phase] Date B.pP. | Layer

Raw-material

ranking AD/m? Unmodified

255-3 220 1 Q@aLls <10° | 1-3: flakes: QtQS

3 400 2 QasL 1-4: few blades and

4 blade cores

3 Q2@asL

yee cl OP ele SITS RE OUT Relies Th Wi hla secgeeks

3/4 4 QsaL flakes: QQtS

3 900 5 SQaL >10° | 5-6: flakes S Qt Q

5-9: more blades,

6 SQQL few blade cores

3 7 QsaQaL flakes: QtS Q

8 Sea L flakes: S Qt Q

6 370

6 100 9 SQeaL

sao Rees ie wie ee, eh eae ear ne cae as, cad

6 540 10 Q2@asL flakes: Qt QS

2/3 11 QaSsL |>2~x 10% flakes: QQtS

fewer blades and

blade cores

7750 12 QsaL

13 Q@asL <10* | few blades and blade

cores

9 760 14 Q2@asL 13-18: flakes Qt QS

oe eal Sl

2 15) QatQLs 14-17: very few blades

and blade cores

16 OUOOEES A ooo Pole ile Neda Teale main Ad

17 QaQqsL few blades and blade

---- 4 cores

18 QasL > 103

] 12 730 19 © OF DS IL flakes: Q QtS

Note. Raw materials: Q = quartz; Qt = quartzite: S = silcrete; L = limestone. Letters in parentheses are also)

materials. |

most blades (S) and

blade cores (Q ‘micro’

AD/m? = average density per cubic metre (see p. 24).

Mammals: VS = very small; S = small; SM = small medium; LM = large medium; L = large;

equal ranking.

OES = ostrich egg-shell.

STONE ARTEFACTS

Utilized

Retouched

scaled pieces and flakes,

no blades

scaled pieces and flakes,

few blades

highest frequencies of

scaled pieces and flakes,

few blades

as 13-16, but with few

blades

mostly scaled pieces,

more blades than flakes

flakes, most blades (S),

scaled pieces

more segments (mostly |i!

1-3: more adzes than I

scrapers

2-3: few segments or

backed scrapers

4-6: backed scrapers mip

common |

5—9: segments and baciiii

flakes most common; |i

trapezoids in these layel|

only.

Scrapers generally smal}ifi

and nearer to equilate|}

(circular)

first backed scrapers;

scraper size decreases |

larger, longer scrapers )Mi

first backed points (bo)

scraper size increases; |§

first borers (drills)

10-18: no backed blad/

13-19: very few retouc|

pieces, mostly scrapers)

16-17: segments in thei

layers only

backed blades (

+ indi¢

BYNESKRANSKOP 1 177

BONE MARINE SHELL OTHER

ARTEFACTS ARTEFACTS ARTEFACTS SHADE BIE lPIoE! es

Size class ranking AD/m*

yost and most Donax scrapers pottery, metal bead, VS SMS first sheep more Choro- <a)

dest range OES pendants mytilus-Perna

ry few no Donax scrapers ripe OES pendants VS S SM more <20

browsers/ ——_—

mixed feeders

(bovids)

VS S SM >12000} more Oxystele| >20

low eee pe eS Se SS Se Se ee Se ee eee eee — — — 4 than Turbo

1W 1-9: shell pendants VS S SM 1-9: Turbo

and perforated whole sarmaticus

shell ornaments generally ----

most common

species

M (?) points first OES pendants VS S SM >50

id linkshafts

gore artefacts than | Donax scrapers first OES discs VS S SM first (?) Raphi-

nove or below cerus

—— campestris =-=-=

ADs for OES beads VS S SM > 20

>400; most decorated

OES

ry few, mostly VS S SM

vis last Equus cf. —_———

VS S SM quagga <20

——— eS oH ———

ADs for OES beads Donax serra

i <100; other OES S SM+LM VS generally <10

P= ees So5 artefacts rare or absent }—-——-——_—____——_ most common | ———

| species

yone LM VS S+SM last <5

Damaliscus

dorcas

b first Donax scrapers LM S VS+S

15: ‘fish gorges’ perforated whole shell LM VS SM more grazers

Ornaments only (bovids and

(no pendants) equids)

‘i LM VS+SM §S

LM VS S

tt a

try few oe VEY 5 Sai VS LM S+SM

no artefacts or VS LM S+SM

ornaments

VS+LMS+SM+L

VS LM L Equus cf.

capensis

{

EY j

= vat

}

_ v

‘

j i

‘ ;

\ =” -

| =) :

i z

i ay

;

i i

j A

} _

i i}

ns 4

Mi ed

, \

‘

Recs

f >

i Y

5 y

‘ \

; By

‘

§ tam

; cat (i ;

k .

a ~

y te A

2 1

. ier

} (

f ". f

\ : ia

1 . ‘1

{

: J

2 3 ra r

ete

rent /

j see

176

TABLE 26

ANNALS OF THE SOUTH AFRICAN MUSEUM

Summary of artefactual and faunal data from BNK 1.

BYNESKRANSKOP 1

177

| BONE MARINE SHELL OTHER

SON SUMS ARTEFACTS ARTEFACTS ARTEFACTS MAMMALS SHELL-FISH FISH

Phase| Date B.P. | Layer

Raw material 5 on

ranking AD/m! Unmodified AD/m* Utilized Retouched Size class ranking AD/m* AD/m*

255-3 220) 1 QaLs <10? | 1-3: flakes: QtQS <50 | scaled pieces and flakes, more segments (mostly Q)| most and most Donax scrapers pottery, metal bead, VS SM S s .

no blades | widest range OES pendants SSESveEc 2.000 | more Choro- <s

eS eee LS SS Ee SAS a Se mytilus-Perna

3 400 2 QasL 1-4: few blades and 1-3; more adzes than very few no Donax scrapers no OES pendants VS S SM more

4 blade cores Scrapers browsers/ pee =e

2-3: few segments or PS

a Da OCXCL ESSER aaa | ame alae <100 > 300 Bickediscrapes = Ee Bie see VS S SM -12000| more Oxysrele 20

3/4 4] QSQL flakes: Q Qt S 5100 | scaled pieces and flakes, 4-6: backed scrapers most| few 1-9: shell pendants ‘|vs s sM Ih sean cae

few blades common and perforated whole = ene,

See 0 i ee ee ee ee eer shell ornaments SARMaAliciis

generally “<<

| most common

3900 5 SQQL >10* | 5-6: flakes S QtQ highest frequencies of 5-9: segments and backed) first (?) points first OES pendants VS S SM Baars ~50

scaled pieces and flakes, flakes most common; { and linkshafts ¥

few blades trapezoids in these layers |

5-9: more blades, only.

6 SQQL few blade cores Scrapers generally smaller| more artefacts than | Donax scrapers first OES discs VS S SM first (2) Raphi-

and nearer to equilateral | above or below certs

er a a a IN (circular) | campestris ern ele

3 7 Q@saL flakes: Qt SQ ADs for OES beads VS S SM ~ 1000 > 20

j------- 0. f------------4 | >400; most decorated

OES

8 SQaQaL flakes: S QtQ very few, mostly VS S SM

6 370 awls last Equus ef. ——

6 100 9 SQaQaL VS S SM quagga <20

Seis | ————+- - -- -------- ee Ee ee et pe LS

first backed scrapers; | ADs for OES beads Donax serra

6540 10 Q@asL flakes: Qt QS <200 | scraper size decreases | <100; other OES S SM+LM VS <200 | generally <10

ie oS SSS SS aa Pe ae ae er toe ey i ame me ee a Ne ea artefacts rare or absent /}-— ———_—_——_4 most common | ———

species

2/3 11 QarSL />2x 10) fakes: QQtS <100] as 13-17 larger, longer scrapers none LM VS S+SM last <5

fewer bladesand = J- - - 4+ -- - - - - - -- - - -+-H | Damaliscus

blade cores | dorcas

7750 12 QsaL >100} as 5-10 <300] first backed points a first Donax scrapers LM S VS+S

ee ———————————— SoS eso e eae Se)=- == 256 S55 Play ce

13 QasL <10° | few blades and blade <50 | very few scaled pieces and} <100) scraper size increases; 13-15: ‘fish gorges’ perforated whole shell LM VS SM more grazers = 50

cores flakes, no blades first borers (drills) | ornaments only (bovids and

I- —-- -------- Sit | (no pendants) equids)

9 760 14 Q@aQasL 13-18: flakes QtQS <50 | 10-18: no backed blades LM VS+SM S

2 15 QaQLs 14-17: very few blades| 13-19: very few retouched LM VS S

and blade cores pieces, mostly scrapers

16 COP OMES SOL a NE NR ore SGP hata ieee Care eee 16-17: segments in these | very few Lp iy 2) Bie Vs LM S+SM

as 13-16, but with few layers only

17 QaQsL few blades and blade blades no artefacts or VS LM S+SM

| i} cores ornaments

18 e@asL > 10° <100} mostly scaled pieces, VS+LM S+SM+L

more blades than flakes

1 12730 19 QQasL flakes: Q QtS flakes, most blades (S), backed blades VS LM L Equus cf.

most blades (S) and scaled pieces capensis

blade cores (Q ‘micro’

Note. Raw materials: Q = quartz; Qt = quartzite; S = silcrete; L = limestone. Letters in parentheses are also ra

materials.

AD/m® = average density per cubic metre (see p. 24).

Mammals: VS = very small; S = small; SM = small medium; LM = large medium; L = large; + indicat

equal ranking.

OES = ostrich egg-shell.

178 ANNALS OF THE SOUTH AFRICAN MUSEUM

than the other layers. Quartzite is the predominant raw material in layers 17-15

and quartz in layers 14 and 13. Unmodified flakes are, however, mostly quartz

and there are few unmodified blades and blade cores, very few in layers 16-14.

Utilized blades occur in layer 17 but not in layers 16-13, and there are no

backed blades in any of the layers. There are segments in layers 17 and 16 but

the few retouched pieces are mostly scrapers. A change in scraper morphology

appears to begin in layer 13, that continues into phase 2/3. The first borers

(drills) were recovered from layer 13.

As in phase 1, there are very few bone artefacts in the layers of phase 2,

and ‘fish gorges’ are restricted to layers 15-13. The first shell ornaments, simple

perforated shells, occur in layer 16 but phase 2 shares with phase 1 a paucity of

marine shell, mostly Donax serra as well as very low ADs for beads and other

artefacts of ostrich egg-shell.

Layers 17 and 16 form a group with those of phase 1 in the ranking of the

mammal size classes, with the very small and large medium size classes in the

first two rankings. Layer 15-13, on the other hand, share with layers 12 and 11

a predominance of the large medium size class, with the very small size class

generally ranked second. These layers are the only ones in which the large

medium size class is numerically predominant but the ungulates are, like those

of phases 1 ind 2/3 mostly grazers. As in phase 1, fish are sparsely represented.

Apart from the fact that there is no evident decline in the relative

frequency of mammals, phase 2 appears to indicate a period of less intensive

occupation of the cave than during phase 1. The ‘fish gorges’ in layers 15-13,

the first occurrence of marine shell ornaments in layer 16 and of borers (drills)

in layer 13 as well as the beginning (?) of a change in scraper morphology in

layer 13 all indicate a degree of technological difference from phase 1, and the

changes in the mammal size class rankings suggest a degree of economic

change.

Phase 2/3

Layers 12-10 share with the upper layers of phase 2 a predominance of

quartz, although layer 12 is anomalous in having silcrete ranked second

whereas the other layers have quartzite. Layers 12 and 11 have the highest ADs

for stone artefacts of all the layers, while layer 10 belongs to phase 3 in this

respect. Unmodified flakes are mostly quartz in layers 12 and 11 but quartzite

in layer 10, and all three layers share with phase 3 an increase in unmodified

blades relative to phase 2, but there are still very few blade cores. Layers 12

and 10 are more like phase 3 in the frequency of utilized artefacts while layer 11

is more like phase 2. Layer 12 has a higher frequency of retouched artefacts

than layers 11 and 10 and has the first backed points (borers), while the first

backed scrapers came from layer 10. In layer 11 the change in scraper

morphology first observed in layer 13 reaches its peak (but see p. 47), while

layer 10 sees the beginning of a reversion to a smaller, more equilateral

(rounder) type. There are no bone artefacts in layers 12 and 11 and very few in

BYNESKRANSKOP 1 179

layer 10. The first Donax serra ‘scrapers’ came from layer 12 but marine shell

ornaments continue to be the simple perforated shells first recorded in layer 16.

These layers share with phases 1 and 2 very low frequencies of ostrich egg-shell

artefacts.

Layers 12 and 11 are like the upper part of phase 2 (layers 15-13) in the

ranking of the mammal size classes while layer 10 suggests a pattern intermedi-

ate between that of the upper part of phase 2 and that of phases 3 and 4. There

are still more grazers than browsers or mixed feeders but the last Damaliscus

dorcas recorded at the site came from layer 11. Shellfish frequencies are still

low, but higher than in phases 1 and 2, and Donax serra continues to be the

most common species in layers 12 and 11 while in layer 10 Turbo sarmaticus

assumes the dominant position it holds in most of the upper layers. Fish

frequencies are higher in layer 10 than in layers 19-11, but still low.

As the numbering indicates, phase 2/3 is seen as transitional between

phases 2 and 3. While in many respects continuing the pattern of phase 2 it

presages that of phase 3 and in general suggests an increase in the intensity of

the occupation of the cave as well as some changes in the technology and

economy of the cave occupants. Layer 10, it may be recalled, marks the

division between D. M. Avery’s vegetation units 3 and 2, although Avery has

suggested that the climate of phase 2/3 was milder than that of phase 2 (cf.

Table 22).

Phase 3

Layers 9-5 are marked by a predominance of silcrete as a raw material,

except in layer 7 in which quartz predominates. The ADs for unmodified

artefacts are lower than for phase 2/3 but higher than for phases 2 or 4 and in

the same range as for phase 1. Unmodified flakes are more commonly quartz in

layers 7 and 6 but silcrete in the other layers. There are generally more

unmodified blades than in phases 2 and 2/3 but still few blade cores. The ADs

for utilized artefacts are the highest except for layer 12 and scaled pieces and

flakes continue to be the most common types. Layer 8 has the highest

frequency of utilized blades after layer 19, but frequencies of these are

generally low throughout the sequence. The highest frequencies of retouched

artefacts are in layers 9-3 and it is generally characteristic of phase 3 that the

layers contain the fullest range of retouched artefact types, which are only

sporadically present in the underlying layers. Scrapers are still the most

common retouched artefacts but while those of layer 9 are morphologically

intermediate between those of layers 13-10 and 8-1, those of layers 8-5 show a

greater degree of homogeneity and are generally in the small to medium size

class (< 30 mm). Apart from layer 1, segments have their highest frequencies in

phase 3, as do backed flakes. Trapezoids occur only in phase 3 and backed

scrapers are most common in layers 6-4.

Bone artefacts are still few although there are more in layers 7-5 than in

the underlying layers. The first positively identifiable points and linkshafts

180 ANNALS OF THE SOUTH AFRICAN MUSEUM

occur in layer 5 but may be present from layer 6. Marine shell ‘pendants’ first

occur in layer 9 while the simple perforated shell ornaments present from layer

16 continue to be found. After layer 1, the layers of phase 3 have the highest

frequencies and widest range of these artefacts. Ostrich egg-shell beads and

decorated ostrich egg-shell are also most common in phase 3.

The ranking of the mammal size classes is virtually identical in layers 9-1,

with the three smaller size classes predominant. Layer 9 contains the last Equus

cf. quagga recorded from the site and marks a change from a predominance of

grazing ungulates in the lower layers to a predominance of browsers and mixed

feeders. The first positively identified Raphicerus campestris occurs in layer 6.

The frequency of shellfish increases markedly in layer 9 and frequencies

generally continue to increase throughout this phase, with Turbo sarmaticus the

most common species. Fish frequencies, although still low, also increase in

phase 3 and the site maximum is reached in layer 5.

Phase 3 suggests the highest intensity of occupation of the cave of the

whole BNK 1 sequence. Where artefacts are concerned there is an increase in

frequency rather than in range, although there are some technological innova-

tions and changes. Change is more marked in the faunal component, with a

decline in the general size range of the mammals brought back to the site and a

marked increase in the amount of shellfish as well as a less marked increase in

fish frequencies.

Phase 3/4

Layer 4 is intermediate between the layers of phase 3 and those of phase 4.

It shares with phase 4 a predominance of quartz and lower ADs for unmodified

artefacts than phase 3 and few blades and blade cores, but the flakes are mostly

quartz whereas in phase 4 they are mostly quartzite. It shares with phase 3 a

higher frequency of utilized artefacts than phase 4 and contains utilized blades,

which are absent from phase 4. The retouched artefact category also has more

in common with phase 3 although it contains a higher frequency of adzes. In

layer 5 the scraper: adze ratio is about 6:1, in layer 4 it is 6:4. There are few

bone artefacts and the layer contains Donax serra ‘scrapers’ and ostrich

egg-shell ‘pendants’, both of which are, perhaps fortuitously, absent from layers

3 and 2 but present in layer 1.

The ranking of the mammal size classes is the same as for phases 3 and 4

and shellfish ADs are similar to those for the upper layers of these two phases,

but fish frequencies are lower than in phase 3.

Phase 3/4 appears to mark the beginning of a period of reduced intensity of

occupation of the cave but provides few indications of technological or eco-

nomic change.

Phase 4

The final occupation phase of BNK 1 has quartz as its predominant raw

material, a lower frequency of unmodified artefacts than phase 3, a predomi-

BYNESKRANSKOP 1 181

nance of quartzite flakes and few blades and blade cores. There is also a

reduction in the frequency of utilized artefacts and this category contains no

blades in these layers. In layer 3 the AD for retouched pieces is the same as for

phase 3 but it drops in layers 2 and 1. The most notable difference of phase 4

from phase 3 is that adzes are more common than scrapers. Backed scrapers

and segments are few, except in layer 1, in which the frequency of segments is

as high as in phase 3, though in layer 1 they are mostly quartz.

Layers 3 and 2 have very few bone artefacts but layer 1 has the highest

frequency and the fullest range of all the layers. Layer 1 also has the most

Donax serra ‘scrapers’, while layers 2 and 3 have none. Other artefacts

restricted to layer 1 are potsherds and the site’s single metal artefact.

Where the fauna is concerned, phase 4 is not markedly different from

phase 3, with the notable exception of Ovis aries in layer 1. Layer 3 has a very

high density of marine shell and Oxystele sinensis is more common than Turbo

sarmaticus which, however, remains the dominant species in terms of flesh

mass. Shellfish and fish frequencies decline in layers 2 and 1 and in layer 1 there

is a marked increase in the frequency of Choromytilus meridionalis relative to

the frequencies in the underlying layers. Domestic sheep are present only in

layer 1, and this causes a change in the ranking of the small medium size class

which in this layer ranks second whereas it ranked third in the underlying

layers.

Phase 4 appears to mark a period of less intensive occupation of the cave

than in phase 3. This is particularly noticeable in the wide range of dates for

layer 1 relative to the depth of deposit. Technological change is indicated by

the predominance of adzes and an overall decline in the frequency of scrapers,

while in layer 1 bone artefacts and Donax serra ‘scrapers’ reach their highest

frequencies and segments are as common as in phase 3. Economic change is

indicated by the decline in the frequency of marine resources and, in layer 1, by

the addition of domestic sheep to the range of wild game already available.

DISCUSSION

By and large, it would seem that in the BNK 1 assemblage technological

change is not concomitant with economic change. The site data indicate for the

lower layers a general emphasis on large medium mammals, mostly bovids,

little reliance on marine resources and very little use of artefacts, in the cave at

least. From layer 12, or perhaps layer 13, technological changes are evident in

the addition of new types of retouched artefacts, although these represent only

a third of the total range of artefact types in this category. There is a change in

scraper morphology evident in layers 13-9 which might have proved less

significant had larger samples been available for layers 19-13 and 11; and the

replacement of the dominance of scrapers in layers 19-4 by that of adzes in

layers 3-1 is undoubtedly also not without significance, although the functional

relationship between these two artefact types has yet to be demonstrated. The

addition, from layer 9 up, of the more elaborate shell pendants to the range of

182 ANNALS OF THE SOUTH AFRICAN MUSEUM

ornaments already available suggests an increase in the leisure (?) time of at

least some of the cave occupants—leisure, that is, from the necessary aspects of

day-to-day life although the shift in resources from the larger animals of the

lower layers to a range of smaller fauna that includes shellfish and fish suggests

a change to a way of life that was more labour-intensive. In this connection it is

worth reiterating the social change implicit in the change of the economic base

from that of the lower layers to that of the upper. It seems axiomatic, in simple

societies at least, that the greater the dependence on communal effort in

resource procurement, the greater the obligation for those resources to be

shared and therefore the greater the sense of community. The analogy of the

'Kung may not be directly applicable to the prehistoric and early historic

hunter-gatherers of the southern Cape but the evidence here (Marshall

1976: 295-303) is that, while sharing is an essential need of the community, the

obligation to share does not extend to smaller game, plant foods and the like.

The change in the BNK 1 fauna from that of the lower layers to that of the

upper is primarily one of a change from the procurement of large game to that

of smaller game, shellfish and fish (and presumably plant foods), resources that

would require more individual effort to procure and thus a diminution of the

need or obligation to share those resources. Although this need not necessarily

result in a weakening of the sense of community it does indicate an increase in

the degree of individualistic behaviour within that community and therefore

some measure of social change.

On the basis of general similarities, phase 1 can be related to the ‘Robberg

industry’ and phase 2 to the ‘Albany’ (H. J. Deacon 1976: 76-78, 117-122; J.

Deacon 1978: 100-109), although it is possible that phase 1 may mark the

transition from the ‘Robberg’ to the ‘Albany’, in the same way as layer BSL at

Nelson Bay Cave seems to do. Phase 3 would unquestionably be termed

‘Wilton’ and, by analogy, phase 2/3 ‘Formative Wilton’ and phases 3/4 and 4

‘Post-climax Wilton’ (cf. H. J. Deacon 1976, table 3). The problem of the

nomenclature and content of the ‘industries’ of the Late Stone Age of southern

Africa has recently again been raised by Parkington and discussed by various

commentators on his paper (Parkington 1980a and comments thereon). It is not

considered appropriate to this report (especially since in its published form it

lacks the conclusions of the principal author) to enter into this debate which, it

is felt, is only likely to be resolved by the publication of detailed reports from a

greater number of sites in a greater variety of localities.

The BNK 1 data indicate that at one level there is some justification for

viewing the sequence as consisting of two major units, one from the initial

occupation of the cave some 12 700 years ago until about 6 000 B.p. and the

second from the latter date until the final occupation of the cave. They also

indicate reasonable validity for the division into ‘phases’ based on a certain

degree of homogeneity in technology and economy. What they indicate above

all, however, is the arbitrariness of these divisions and they emphasize that

change, to greater or lesser degree but none the less change, was continual

BYNESKRANSKOP 1 183

throughout the occupation of the site, and that it was diachronous rather than

synchronous. The relevance of this to the ecology of the human inhabitants of

the region cannot be determined from the site data alone and requires substan-

tiation from other sites in the area as well as from independently-derived

environmental data. The interaction between man and his environment, viewed

from the perspective of archaeological data alone, is at best one-sided since it

presents aspects of man’s use of his environment without the correlate of that

environment’s effect on man. In this regard, it is regrettable that the sparseness

of the final deposits precludes more detailed observation of what may have

been the most significant effect of man on the environment of the Cape: the

introduction of pastoralism.

ACKNOWLEDGEMENTS

The excavations were carried out under permit from the National Monu-

ments Council.

Funds for this research, in respect of subsistence and transport, were

provided by the Human Sciences Research Council, which is not to be regarded

as responsible for or concurring with the opinions expressed or conclusions

reached. The cost of equipment, the provision of vehicles, and other related

expenses were borne by the South African Museum.

Professional assistance was given by: Mr T. Arnold, Botanical Research

Institute, Pretoria (identification of Citrullus sp. seed cases); Dr C. Boucher,

Botanical Survey, Stellenbosch (botanical information); Professor G. Branch,

Department of Zoology, University of Cape Town (marine biological informa-

tion); Professor H. de Villiers, Department of General Anatomy, School of

Dentistry, University of the Witwatersrand, Johannesburg (analysis of the

human remains); Mrs P. Fairall and Mrs C. E. Labuschagne, National Botanic

Gardens, Kirstenbosch, Cape Town (botanical information); Mr and Mrs M. C.

Giles, Dr P. A. Hulley, and Miss A. E. Louw, all of the Marine Biology

Department, South African Museum (marine biological information); Mr D.

Halkett, University of Cape Town (artefact analyses and most of the technical

drawings); Dr Q. B. Hendey, Cenozoic Palaeontology Department, South

African Museum (faunal identification and information); Miss L. Horwitz, now

at the Hebrew University, Jerusalem (plant remains); Professor R. G. Klein,

Department of Anthropology, University of Chicago (analysis of the mamma-

lian fauna and provision of a radiocarbon date); Dr G. A. McLachlan,

Herpetologist, South African Museum (identification of reptiles and informa-

tion on birds); Mr C. E. Poggenpoel, Department of Archaeology, University

of Cape Town (analysis of fish remains); Mr R. Rau, Taxidermy Department,

South African Museum (information on tortoises); Mrs T. Reyneke, Mrs I.

Brunke, Miss B. Mann and Mr I. Bendie, all of the South African Museum

Library (library matters); Dr J. Rourke, Curator of the Compton Herbarium,

National Botanic Gardens, Kirstenbosch (botanical information and arranging

184 ANNALS OF THE SOUTH AFRICAN MUSEUM

for the identification of the Citrullus sp. seed cases); Miss E. M. Shaw,

Ethnology Department, South African Museum (ethnological information);

and Dr J. C. Vogel, Head of the Natural Isotopes Laboratory, Council for

Scientific and Industrial Research, Pretoria (provision of radiocarbon dates).

Mr V. Branco, Entomology Department, South African Museum, drew Figure

3; Cdr A. Fawthrop, Directorate of Hydrography, South African Navy, made

available off-shore survey charts; Mr L. Lawrence, Archaeology Department,

South African Museum, took most of the photographs, drew the statistical

figures, and assisted in most aspects of the work; and Mr A. Schweitzer

provided air photographs of the site and surroundings. Thanks are also due to

the many people who assisted in the field and in the laboratory.

Assistance in various forms was also provided by Dr D. M. Avery and Mr

G. Avery, Archaeology Department, South African Museum; Professor H. J.

Deacon, Mrs J. Deacon and Mrs M. Brooker Leslie, Department of Archae-

ology, University of Stellenbosch, also Mr J. N. F. Binneman, now at the

Albany Museum, Grahamstown; Mr B. D. Malan; Mr A. D. Mazel, Ethno-

archaeology Department, Natal Museum, Pietermaritzburg; Professor J. Par-

kington, Department of Archaeology, University of Cape Town; and Dr T. P.

Volman, Department of Archaeology, University of the Witwatersrand, Johan-

nesburg (now at Cornell University).

Mr P. van D. Swart, ‘Uilenkraal’, Strandskloof, gave permission for the

excavations to be carried out and Mrs M. Eyre, ‘Skuilplek’, Strandskloof,

permitted access to the site through her property. To them and to Mrs Swart,

thanks are also due for much hospitality.

Special thanks are due to Dr O. Pollak Schweitzer, Dr Q. B. Hendey, and

Dr M. Hall, Head of the Archaeology Department, South African Museum,

for their unfailing support; also to the Director and Trustees of the South

African Museum for enabling the research to be carried out and published, and

to Mrs I. Rudner, Editor of the Annals, for preparing this report for the press.

Thanks are also due to Mrs Deacon, Dr Hall and Professor Klein for helpful

comments on the manuscript.

REFERENCES

Acocks, J. P. H. 1953. Veld types of South Africa. Mem. Bot. Survey S. Afr. 28: 1-192 (Ast

edn.).

Acocks, J. P. H. 1975. Veld types of South Africa. Mem. Bot. Survey S. Afr. 40: 1-128 (2nd

edn.).

Acocks, J. P. H. 1979. The flora that matched the fauna. Bothalia 12: 673-709.

Avery, D. M. 1979. Upper Pleistocene and Holocene palaeoenvironments in the southern

Cape: the micromammalian evidence from archaeological sites. D. Phil. thesis, University

of Stellenbosch.

Avery, G. 1974. Open station shell midden sites and associated features from the Pearly Beach

area, south-western Cape. S. Afr. archaeol. Bull. 29: 104-114.

Avery, G. 1975. Discussion on the age and use of tidal fish-traps (visvywers). S. Afr. archaeol.

Bull. 30: 105-113

Avery, G. 1976. A systematic investigation of open station shell midden sites along the

southwestern Cape coast. Unpublished M.A. thesis, University of Cape Town.

BYNESKRANSKOP 1 185

Avery, G. & SIEGFRIED, W. R. 1980. 150 000 year tradition: food gatherers along South

Africa’s seashore. Oceans 13: 32-37.

BRAIN, C. K. 1974. Some suggested procedures in the analysis of bone accumulations from

southern African Quaternary sites. Ann. Transvaal Mus. 29: 1-8.

Brancu, G. M. 1971. The ecology of Patella Linnaeus from the Cape Peninsula, South Africa.

1. Zonation, movements and feeding. Zool. Africana 6: 1-38.

BrancH, G. M. 1974a. The ecology of Patella Linnaeus from the Cape Peninsula, South Africa.

2. Reproductive cycles. Trans. R. Soc. S. Afr. 41: 111-160.

Brancu, G. M. 1974b. The ecology of Patella Linnaeus from the Cape Peninsula. 3. Growth

rates. Trans. R. Soc. S. Afr. 41: 161-193.

BUCHANAN, W. F. 1977. Rescue dig at a Late Stone Age cave, Hout Bay, Cape Province.

Unpublished Archaeology Additional Project, University of Cape Town.

BUCHANAN, W. F., HALL, S. L. HENDERSON, J., OLIVIER, A., PETTIGREW, J. M., PARKINGTON,

J. E. & RosBertsHAw, P. T. 1978. Coastal shell middens in the Paternoster area,

south-western Cape. S. Afr. archaeol. Bull. 33: 88-93.

Bupback, K. F. R. 1977. The #Aonin or Topnaar of the lower !Khuiseb Valley and the sea.

Khoisan Ling. Std. 3: 1-42 (Johannesburg: University of the Witwatersrand, African

Studies Institute (A. S. I. Communication no. 6)).

BuTzer, K. W. 1978. Sediment stratigraphy of the Middle Stone Age sequence at Klasies River

Mouth, Tsitsikama coast, South Africa. S. Afr. archaeol. Bull. 33: 141-151.

CiarK, J. D. 1959. The prehistory of southern Africa. Harmondsworth: Penguin Books.

CLARKE, D. L. 1979 Mesolithic Europe: the economic basis. Jn: CLARKE, D. L. Analytical

archaeologist: 207-262. London: Academic Press.

CoATES PALGRAVE, K. 1977. Trees of southern Africa. Cape Town and Johannesburg: Struik.

Day, J. H. 1969. A guide to marine life on South African shores. Cape Town: Balkema.

Day, J. H. 1970. The biology of False Bay. Trans. R. Soc. S. Afr. 39: 211-221.

Day, J., SIEGFRIED, W. R., Louw, G. N., & JARMAN, M. L., eds. 1979. Fynbos ecology: a

preliminary synthesis. South African National Scientific Programmes report no. 40. Pre-

toria: Co-operative Scientific Programmes, C.S.I.R.

Deacon, H. J. 1976. Where hunters gathered. Claremont: South African Archaeological

Society, monograph series no. 1.

Deacon, H. J. 1979 Excavations at Boomplaas Cave—a sequence through the Upper Pleis-

tocene and Holocene in South Africa. Wild Archaeol. 10: 241-257.

Deacon, H. J. & DEAcON, J. 1980. The hafting, function and distribution of small convex

scrapers with an example from Boomplaas Cave. S. Afr. archaeol. Bull. 35: 31-37.

Deacon, H. J., DEACON, J., BROOKER, M. & WILson, M. L. 1978. The evidence for herding at

Boomplaas Cave in the southern Cape, South Africa. §. Afr. archaeol. Bull. 33: 39-65.

Deacon, J. 1965. Part I: cultural material from the Gamtoos Valley (Andrieskraal I). S. Afr.

archaeol Bull. 20: 193-200.

DEAcoNn, J. 1972. Wilton: an assessment after fifty years. S. Afr. archaeol. Bull. 27: 10-48.

DEACON, J. 1974. Patterning in the radiocarbon dates for the Wilton/Smithfield Complex in

southern Africa. S. Afr. archaeol. Bull. 29: 3-18.

DEACON, J. 1978. Changing patterns in the late Pleistocene/early Holocene prehistory of

southern Africa as seen from the Nelson Bay Cave stone artifact sequence. Quatern. Res.

10: 84-111.

DEACON, J. 1979. Guide to archaeological sites in the southern Cape. Occasional publication of

the Department of Archaeology, University of Stellenbosch. (Duplicated.)

DEACON, J. 1980. Comment on: PARKINGTON, J. 1980. Time and place. S. Afr. archaeol. Bull.

35: 89-93.

De ViILuierS, H. & Witson, M. L. 1982. Human burials from Byneskranskop, Bredasdorp

district, Cape Province, South Africa. Ann. S. Afr. Mus. 88(2).

Dorst, J. & DANDELOT, P. 1972. A field guide to the larger mammals of Africa. London:

Collins.

Du Tor, A. L. 1954. The geology of South Africa. Edinburgh: Oliver & Boyd.

FaGAaNn, B. M. 1960. The Glentyre Shelter and Oakhurst re-examined. S. Afr. archaeol. Bull.

15: 80-94. (See also Corrigendum, 1962. S. Afr. archaeol. Bull. 17: 195.)

FLEMMING, B. W. 1977. Depositional processes in Saldanha Bay and Langebaan Lagoon. Joint

Geological Survey/University of Cape Town Marine Geoscience Group Bulletin no. 8 (NRO

Professional series no. 2) Department of Geology, University of Cape Town.

186 ANNALS OF THE SOUTH AFRICAN MUSEUM

Goopwin, A. J. H., 1938. Archaeology of the Oakhurst Shelter, George. Parts I, VI, VII.

Trans. R. Soc. S. Afr. 25: 229-245, 303-320.

Goopwin, A. J. H. 1945. Some historical Bushman arrows. S. Afr. J. Sci. 41: 429-443.

Goopwin, A. J. H. 1946. Prehistoric fishing methods in South Africa. Antiquity 20: 134-141.

Goopwin, A. J. H. 1947. The bored stones of South Africa. Ann. S. Afr. Mus. 37: 1-210.

Goopwin, A. J. H. 1956. Metal working among the early Hottentots. S. Afr. archaeol. Bull.

11: 46-51.

GREVENBROEK, J. G. de 1695. An elegant and accurate account of the African race living round

the Cape of Good Hope commonly called Hottentots. In: ScHAPERA, I., ed. 1933. The early

Cape Hottentots: 172-299. Cape Town: The Van Riebeeck Society. (First series, no. 14.)

GRINDLEY, J. R., SPEED, E., & Maccs, T. 1970. The age of the Bonteberg shelter deposits,

Cape Peninsula. S. Afr. archaeol. Bull. 25: 24.

Hewitt, J. 1931. Artefacts from Melkhoutboom. S. Afr. J. Sci. 28: 540-548.

INSKEEP, R. R. 1978. The peopling of southern Africa. Cape Town and London: David Philip.

JORDAAN, P. G. 1946. Die plantegroei van die distrikte Bredasdorp en Caledon. Tydskr.

Wetensk. en Kuns 6: 47-58.

KEIN, R. G. 1972a. Preliminary report on the July through September 1970 excavations at

Nelson Bay Cave, Plettenberg Bay (Cape Province, South Africa). Palaeoecol. Afr. 6:

177-208.

KLEIN, R. G. 1972b. The Late Quaternary mammalian fauna of Nelson Bay Cave (Cape

Province, South Africa): its implications for megafaunal extinctions and environmental and

cultural change. Quatern. Res. 2: 135-142.

KLEIN, R. G. 1974. Environment and subsistence of prehistoric man in the southern Cape

Province, South Africa. Wild Archaeol. 5: 249-284.

KLEIN, R. G. 1976a. The mammalian fauna of the Klasies River Mouth sites, southern Cape

Province, South Africa. S. Afr. archaeol. Bull. 31: 75-98.

KEIN, R. G. 1976b. The fossil history of Raphicerus H. Smith, 1827 (Bovidae, Mammalia) in

the Cape Biotic Zone. Ann. S. Afr. Mus. 71: 169-191.

KLEIN, R. G. 1978a. The vertebrate fauna from the Buffelskloof rock shelter. S. Afr. archaeol.

Bull. 33: 35-38.

KLEIN, R. G. 19785. A preliminary report on the larger mammals from the Boomplaas Stone

Age cave site, Cango Valley, Oudtshoorn district, South Africa. S. Afr. archaeol. Bull. 33:

66-75.

KLEIN, R. G. 1980. Environmental and ecological implications of large mammals from Upper

Pleistocene and Holocene sites in southern Africa. Ann. S. Afr. Mus. 81: 223-283.

KLEIN, R. G. 1981a. Later Stone Age subsistence at Byneskranskop Cave, South Africa. In:

HarpinGa, R. S. O. & TELEKI, G., eds Omnivorous primates: gathering and hunting in

human evolution: 166-190. New York: Columbia University Press.

KLEIN, R. G. 1981b. Stone Age predation on small bovids. S. Afr. archaeol. Bull. 36: 55-65.

KoLB(EN), P. 1738. The present state of the Cape of Good Hope. 1 2nd edn. Trans. by

G. Medley. London: Innis & Manby.

Macs, T & SPEED, E. 1967. Bonteberg Shelter. S. Afr. archaeol. Bull. 27: 80-93.

MARSHALL, L. 1976. The !Kung of Nyae Nyae. Cambridge, Mass. and London: Harvard

University Press.

MAyTHAM Kipp, M. 1973. Wild flowers of the Cape Peninsula. Cape Town: Books of Africa.

MazeL, A. D. 1978. Stories in stones: aspects of artefact distribution in the Clanwilliam

district, south-western Cape. Unpublished B.A. (Hons) project, University of Cape Town.

MAZEL, A. & PARKINGTON, J. 1978. Sandy Bay revisited: variability among Late Stone Age

tools. S. Afr. J. Sci. 74: 381-382.

Mooprg, D. ed. 1838. The Record. Amsterdam & Cape Town: Balkema. (Photostatic reprint

1960.)

Morner, N-A. 1978. Palaeogeoid changes and palaeoecological changes in Africa with respect

to real and apparent palaeoclimatic changes. Palaeoecol. Afr. 10: 1-12.

Oxivier, A. M. 1977. The archaeological survey of the Cape Peninsula. Unpublished Archae-

ology Additional Project, University of Cape Town.

OpPERMAN, H. 1978. Excavations in the Buffelskloof rock shelter near Calitzdorp, southern

Cape. S. Afr. archaeol. Bull. 33: 18-34.

PARKINGTON, J. 1972. Seasonal mobility in the Late Stone Age. Afr. Stud. 32: 223-243.

BYNESKRANSKOP 1 187

PARKINGTON, J. 1976. Coastal settlement between the mouths of the Berg and Olifants rivers,

Cape Province. S. Afr. archaeol. Bull. 31: 127-140.

PARKINGTON, J. 1979. Report on research in the Verlore Vlei 1976-9. In: Soaqua: reports on

research into the Late Stone Age of the Western Cape: 2-34. Archaeology Department,

University of Cape Town. (Duplicated.)

PARKINGTON, J. 1980a. Time and place: some observations on spatial and temporal patterning in

the Later Stone Age. S. Afr. archaeol. Bull. 35: 73-83, 107-109.

PARKINGTON, J. 1980b. The Elands Bay Cave sequence: cultural stratigraphy and subsistence

strategies. Proc. pan Afr. Congr. Prehist. Quatern. Stud. 8: 314-320.

PARKINGTON, J. & ROBERTSHAW, P. T. 1978. Coastal shell middens in the Paternoster area,

south-western Cape. S. Afr. archaeol. Bull. 33: 89-93.

PHILLIPSON, D. W. 1976. The prehistory of eastern Zambia. Nairobi: British Institute in Eastern

Africa.

PIENAAR, U. DE V. 1974. Habitat preference in South African antelope species and its

significance in natural and artificial distribution patterns. Koedoe 17: 185-195.

POGGENPOEL, C. E. 1979. Fish remains, technological change and environments in the terminal

Pleistocene and Holocene of the Cape. Unpublished paper read at the biennial meeting of

the Southern African Association of Archaeologists, University of Cape Town, 23-26 June

1979. (Duplicated.)

RAVEN-HART, R. 1967. Before Van Riebeeck. Cape Town: Struik.

ROBERTSHAW, P. T. 1977. Excavations at Paternoster, south-western Cape. S. Afr. archaeol.

Bull. 32: 63-73.

ROBERTSHAW, P. T. 1978. Archaeological investigations at Langebaan Lagoon, Cape Province.

Palaeoecol. Afr. 10: 139-148.

ROBERTSHAW, P. T. 1979. Excavations at Duiker Eiland, Vredenburg district, Cape Province.

Ann. Cape Prov. Mus. (human sci.) 1 (1): 1-26.

RuDNER, I & RuDNER, J. 1954. A local Late Stone Age development. S. Afr. archaeol. Bull. 9:

103-107.

Rupner, I. & RupDNER, J. 1956. Excavation of the Logies Rock cave, Llandudno. S. Afr.

archaeol. Bull. 11: 77-80.

RUDNER, J. 1968. Strandloper pottery from South and South West Africa. Ann. S. Afr. Mus.

49 (2): 441-663.

Rupne_er, J. 1971. Ostrich egg-shell flasks and soapstone objects from the Gordonia district,

north-western Cape. S. Afr. archaeol. Bull. 26: 139-142.

Sampson, C. G. 1974. The Stone Age archaeology of southern Africa. New York & London:

Academic Press.

SCHRIRE, C. 1962. Oakhurst: a re-examination and vindication. S. Afr. archaeol. Bull. 17:

181-195.

SCHULZE, B. R. 1965. Climate of South Africa. Part 8: General Survey. Pretoria: Government

Printer (WB 28, reprinted 1974).

SCHWEITZER, F. R. 1979. Excavations at Die Kelders, Cape Province, South Africa: the

Holocene deposits. Ann. S. Afr. Mus. 78 (10): 101-233.

SCHWEITZER, F. R. & Scott, K. J. 1973. Early occurrence of domestic sheep in sub-Saharan

Africa. Nature, Lond. 241: 547.

SCHWEITZER, F. R. & Witson, M. L. 1978. A preliminary report on excavations at Byneskrans-

kop, Bredasdorp district, Cape. S. Afr. archaeol. Bull. 33: 134-140.

SHEPARD, A. O. 1954. Ceramics for the archaeologist. Washington D.C.: Carnegie Institution of

Washington. (Publication 609).

SIEGEL, S. 1956. Non-parametric statistics for the behavioral sciences. Tokyo: McGraw-Hill

Kogakusha Ltd.

SINGER, R. & Wymer, J. J. 1969. Radiocarbon dates for two painted stones from a coastal cave

in South Africa. Nature, Lond. 224: 508-510.

SMITH, C. A. (ed. E. P. Puituips) 1966. Common names of South African plants. Mem. Bot.

Survey 35: 1-642.

SMITH, J. L. B. & Smit, M. M. 1966. Fishes of the Tsitsikama Coastal National Park. Pretoria:

National Parks Board of Trustees of the Republic of South Africa.

SPIES, J. J., ENGELBRECHT, L. N. J., MALHERBE, S. J. & VILIOEN, J. J. 1963. Die geologie van

die gebied tussen Bredasdorp en Gansbaai. Pretoria: Government Printer (Dept. of Mines,

Geological Survey).

188 ANNALS OF THE SOUTH AFRICAN MUSEUM

STEFHENSON, T. A. 1944. The constitution of the intertidal fauna and flora of South Africa. Part

2. Ann. Natal Mus. 10 (3): 261-358.

TANKARD, A. J. 1975. Cenozoic sea-level changes: a discussion. Ann. S. Afr. Mus. 71: 1-17.

TayLor, H. C. 1961. Ecological account of a remnant coastal forest near Stanford, Cape

Province. Jl. S. Afr. Botany 27 (3): 153-165.

Taytor, H. C. 1978. Capensis. Jn: WERGER, M. J. A. ed. Biogeography and ecology of southern

Africa: 173-229. The Hague: Junk.

Tuo, H. B. ed. 1952. Journal of Jan van Riebeeck 1. Cape Town & Amsterdam: Balkema.

THUNBERG, C. P. 1795. Travels in Europe, Africa and Asia, performed between the years 1770

and 1779 1. London: Richardson & Egerton.

TINLEY, K.L. 1969. Dikdik Madoqua kirki in South West Africa: notes on distribution, ecology

and behaviour. Madoqua 1: 7-33.

Van RieBEECK, J. A. 1651-5. Daghregister gehouden by den Oppercoopman Jan Anthoniz van

Riebeeck 1 (1651-1655). Cape Town & Amsterdam: Balkema, 1952.

VaN Riet Lowe, C. 1929. The Smithfield Industry in the Orange Free State. In: Goopwin,

A. J. H. & VAN Riet Lowe, C. The Stone Age Cultures of South Africa. Ann. S. Afr.

Mus. 27: 151-206.

VAN ZINDEREN BAKKER, E. M. 1976. The evolution of late-Quaternary palaeoclimates of

southern Africa. Palaeoecol. Afr. 9: 160-202.

VISSER, J. 1979. Common snakes of South Africa. Cape Town: Purnell.

VISSER, J. & CHAPMAN, D. S. 1978. Snakes and snakebite. Cape Town: Purnell.

VoGEL, J. C. & VisseER, E. 1981. Pretoria radiocarbon dates II. Radiocarbon 23 (1): 43-80.

VoicT, E. A. 1975. Studies of marine mollusca from archaeological sites: dietary preferences,

environmental reconstructions and ethnological parallels. In: CLason, A., ed. Archaeozoo-

logical Studies: 87-98. (Papers of the 1974 Archaeozoological Conference, Groningen,

Holland.) Amsterdam: North Holland Publishing Co.

WATERHOUSE, G., ed. 1932. Simon van der Stel’s journal of his expedition to Namaqualand,

1685-6. Dublin: Hodges, Figgis & Co.

Watt, J. M. & BREYER-BRANDWUK, M. G. 1962. The medicinal and poisonous plants of

southern and eastern Africa. 2nd edn. Edinburgh & London: E. & S. Livingstone.

Witson. M. L. 1974. Pottery from three cave sites in the south-western Cape. Unpublished

Archaeology Additional project, University of Cape Town.

Witson, M. L. 1976. Report on the pottery from the 1974-1976 excavations at Boomplaas

Cave, Cango Valley. Unpublished B.A. (Hons) project, University of Stellenbosch.

Witson, M. L. & HALkKeTr, D. 1981. The use of marine shell for decorating Cape coastal

(Khoisan) pottery. S. Afr. archaeol. Bull. 36: 43-44.

APPENDIX 1

BYNESKRANSKOP 1: MAMMALS. TABLES OF THE MINIMUM

NUMBERS OF INDIVIDUALS REPRESENTED BY VARIOUS

SKELETAL PARTS

R. G. KLEIN

University of Chicago

(Compiled January 1978)

ANNALS OF THE SOUTH AFRICAN MUSEUM

“soskydide pasny pue pesnjun Aq pojuoseida s[enpIrAIpu! Jo siaquinu 9Y} 9}¥dIPU! sesoyJUeIed UT sIoquINU s2[qe} SUIMOT[OJ By} PUL SITY) UT ‘asoNy

190

I I z ws ee z ra , ae ae fe Dobbpeonenonh splosy

[@ISIP 9}JVUIWIIO}OpUT

pe == —_ == = = — — — — _— —_— I = = = IT °°° [eisip—

oe = s em I a I Si Lf I I I (4 if I — Z = yewtxoid—sprourseg

ae es = = a = — Gero a a = sre GrOe (roe Or CaO (Gri 9° [eSp=

I = I I I = = I ¢ I I € ¢ Z I I € [ewlxoid—|esiejejp

= = I = I = = Ss I I I v 8 Z I = Z °° °° proqns-opnoraenN

= as = I I = =F = I G = = L I = T € “°° snpooljeu [e19}e 7

a oat a. poss fe a its ac ats bas oe , 1 I = ae G ctr z

i? = =e = == = = I = I Zz C Ol 6 I Z Ca ea ] wiojsiounD

= I = G (6 I = I ¢ G € v II € z I Gere snjesensy

= = = I I = = = = I € v II € (6 z ES ee See uinours[ed

is = Orr WT Garoie Oar We = = = re Care Care Cars Car 08 @ron Mare (iskO Gry -es20°8* [e1sIp—

aE ae Se oe s = ee T ae = Ory = CarDs Ord Ory — Ate °° yeurxoid—eiqry

= = xP Oru — Orr DL Orr = Csr ie re a Ware Gras Gare Oxi Gare Carae °°°°°° fe3sIp—

== (iar @)ii —= (see Osr@yi mg sed = Ware UsrOl Use De Careo Garger Gee — +E U+p)s °*° [ewIxoid—inwe;

3 soil I I io a a 1 z 1 ag z 6 ed , je A cheonngane can slang

as ae nee se ae 1 : me z a a2 I 7 Zz Aca 1 G) geben sadeon wmnigosy

oe = a — ae z st ae ¢ 1 eee I or 1 1 ok: po eee wanT

= I = == I I I I I I I v 8 € Z C 1 ee ees pig —

= (ar@yil — (+0)I G+)Dz — (@+D~2 +o) GA+dDz GA+HYz7 (+01 (€+Dr (G4+0L G4+D2Z G4+HNz G@+dDz @+0r 7°" puz—

(i+0)I (+0O)I — (+01 G+D~% +01 G+Dz2 (+01 (+01 G+or G4+Nz @+YDe G+DL G4+Hz W4+Ye G+Yz (€4+d0s °°: Js] — sosurvyeyg

= = I I I = = = G = I € € I I € € jewrxoid—yjedivovioyy

ae = med = es a = . I = z z 6 Z a nae Gott tees wIO}OUQ

= ES i = = = ee, ce i ; é ; 5 > | = Q reenact ayeuny

= = T I = I I == € G € 9 TI € 1 == Pipe a a wdojsounD

= I = = I as = == = I € 9 9 I = = Cr gee proydesg

aa == = age = I == I Tt I G 9 Ol € a =a Cg i ee uinuse yy

ie <- — (+01 aa = a — @+DE O+DI G+OI C+YE &+ee G+oO1 A+D~2 ODI G+vdE °°: [eqsIp —

— (ise xe (EO = — (+01 = — (b+0)r 1+0)I (+07 +Ds (€Cl+H)LI G+01 G+NzZ O+DI &+e)L °°°° TeIsIP—

—— —, os iG I a Ao mS + == I ty I I G == Z Z 9 OO COO GO OOO 50 eindess

— = —= == = == = = —S —— — —_ Vv — T I € Nene aes SBIQGIIOA [BIDNVS

ae = =) SFONE (Oar ii = = = (are OF OskWn (see Ge Di2 (isk We aaa — (1+0)I °° ° 9B1qaqI9A IequINT

= ae — (I+0)I = — (+0)I (+01 G+Dz7 G4+Nz7 OFDI G4+Nz7 @+Hs G+HYz7Z G4+Yz7Z G+HN7 (T4+VE °° 9v1QA1I9A O1IORIOYL

= = Saat )e- OD == (srOl ClarOM Gisr ke OFF CEO Cee Wako — (+0)1 (+O0)E (T+DE L-€ 281Q919A [BOTAIOD

= = I I = = = G G v € € €l € I = leas eee ee SIXW

ais I ae pa , a ae ae ; i es Zz Ol eo. i was Poco sen

I Zz I € G € I € € € ¢ L €7 8 € € 6 uoT UEP Ie[NqipuLyy

ia == I i = = = a I = € ¢ LI Z = = € ° aApuoo Iejnqrpuey|

I = G € = G omar = v v v E 07 € Z v CO ea eee BIXe A

I T = == —- — = I G I € v 6 Z Z = p °° afApuoos jeqdI999

= G == I T —= == I = = I — v = I I Ds Ne nae Ja]UOI

LI 91 cI vl €l TI IT Ol 6 8 L 9 ¢ v € Z I

‘dds snaaa1ydvy

] dav

TABLE 2

Oreotragus oreotragus.

BYNESKRANSKOP 1 191

Pd a Pe

Ly AS RTS Se Ret WES TS Syl

Ve ITI IS UTI ae a Se ea

PPD SS DT see es Get Sse [oT

10+1) —

PED Te) a de aes a

PE Ta ee a aS Sista es eee

Pe ee th st

10+1) —

Se Stee en OP MEen acetal eweires cin! a tous ens cole uted ue)! ce, fei tties Nei eohte le! nell vet et. jet 8) fel y deh, cell Mes Gee jel) jely eh sai “fet se

. . . Neat Rea ove es ek eR sanep ome Cee vey (eC iiefe lle) Crete iw fe) of Si) cee fe) Tee ler) .e)%,~ te eh ie” es Jechlore

ion

SE OCG) econo CMe Cees uel. Giey (ise iat Wels cole oguiteM 5) cer wel dey! jel“ ej riev- cashes jel fei. cet teh ey oe) eh fe! ine.

. . -_ . .

oO meena rer | ie ee eRe Ae ys co ebmeoe MCSY Xen UM alain lgs velit ey = ve

Ole! TO. Sto oe = Os

eit wh ace ter), ie) :8)

° = . . . . . .

ea ey cel ie) gy. weet ee) eg a ie re it) afer OMY ei tet Pre hite a" 8

—distal ..

—distal .

Indeterminate distal

—2nd

—3rd

Sesamoids — proximal .

nS 6 9 6 BA

—distal ..

—distal ...

Tibia —proximal ,

metapodial ........

Metatarsal —proxima

Hyoid ...

Cervical vertebrae 3—7

Thoracic vertebrae.

Phalanges—Ist .

Femur — proximal

Calcaneum .

Astragalus. .

Cuneiform 1

Lateral malleolus.

Naviculo-cuboid .

Unciform .

Patella ..

Magnum .

Scaphoid .

Cuneiform

Lunate ...

Sacral vertebrae .. .

Scapulatesne ee oe

Lumbar vertebrae .

DNS AS. Gini Cosco Ee

Mandibular denti

Atlas ..

Occipital condyle... .

Maxillary. 5 8 octes os

Frontlet ..

TABLE 3

Small medium bovids.

192 ANNALS OF THE SOUTH AFRICAN MUSEUM

SS 0 a al el esate (ime a ea Giclee Pa lee eel wifes es cde th fe

ie ae bd dk kl

SE CEE Te els ee

eet hei Soe ett hl Lia!

Se Te TTS eT AT aN Sta IST se Teter S ests Tis a

; =

| = ee Peon Siete IT i lida

ae 2

aes cae Sata et tr t i tilda!

x iq

Teese ose ae ai Lehi pees] | la

Sec ee reer y ri pit titi ide

ee aan Sata t ta ttl lad il

ee ik :

Saloon Shree Se i esp it. fsa.

EST Sere as

See eee eee IV IPSS tle biidi

aa a

eee Pee VPI SSSI SlhTiieltiliiat

S S) =) S

a5 = B

Pam ae ie eae pole eel eee eae a) epee Se Sia als Soe fe foal = |

eee WA Se ES ey ar eas ee

ea A LS ee Sea ape

Set Peery TAS St Ph arti ted i!

FES E8RE2 aac ES S

| FL i BoE ae

SS] see] ==e =, 2 =

IAT en Sh ferent es Se Aire tay | los Pleo aacee Sas

eB i ob ce ee eee

Qe 22 ON BB des vee eit: ae Sore ee ase eee oe eee

Ie cee Oe as cic yale es a See fe oO BR aS

See econ ie eo ml eee ee Sat |

gisiS i igggied y' FESS .EF SB el TT ese ash =

SSEEES 2252585 5 ESEZEGS § EEZ2 £ BEEZ SES sg

LOSS > 2 eUeoaoe nt @ SSSOooDS a | Seam a anomie eee

11+0) 10+0) 10+1) 10+1) —

metapodial ........

Indeterminate distal

FIV OIG!) cA eS Pee

193

BYNESKRANSKOP 1

= +

1 8c 8eenmcon ane plosy

]@1SIP oJeUuIUIOJOpuyT

G6 99° * (eto

€ ° [ewixo1id—sprowrseg

(I+O)I °° [easip—

I I

= I J ‘Jewrxoid—|]esieiepy

(0+ IDI

i seas plogns-ojnotAeyy

— “"** snjoayjeur [e191e7

1 I

I I

— I as == = Z SonGcooaad I wiojlaunD

— —— if a= Ss I SO0000nD gH snyesensy

Lf I ms

== —— (7+ 0)Z rd jeisip—

je1sip —

ae — (O+]))I °° °° Jewrxoid—inwe7

pull cae [occttttttttssees siqng

ae <a fe B ane onon DE wINnIYys]

as rar ise sae le en We wnt]

(fe ln sc san i 2) Sl ey

(O+ IDI ==

1 I I I I I

(I1+0)1 — (I[+0)I = — (I+ D7 (+01 G4+Nz (1+0)1 (+0)1 (+o0)l G+pN~2 +01 (+01 +01 G+o0)1 G4+)z °°": puz—

(1+0)1 — (+0) G+0)1 G+0)1 G+o01 G+01 (+0)! +01 +01 G+o1 G+o01 G+oLr Gt+or1 (@+Dz G+o1 G+ol Ato (+ DE c°°*: 1S] — sosurjeyg

ag wa a (O+ IDI — — (aro = — — (I+ Nz °° TeIs1p—

— if I — if [ [ewrxoid—sjediesej0py

I — —— eee Ve QTE CA wail oc ojeun]

— —_— I — = = Toe eee wuojreund

— if — a — Pe | ie ed CeO proydeog

— — I (Ghee DOO wnuseyy

i rs sar = CSG. aa jewxoid — eur)

ae ae = Ses i = RS MONO 6 jeisip —

= == (Oar Di = — — (I+ )z °° Jeurxoid—snipey

= (aro = a = — Se Sars yeisip —

= == — ve — ~-* jewrxoid—snisuny

= oot ss a uy tee eee ejndeog

mae == = = == foneas 2P1Q9119A [e1DeS

a = = = — ‘7? 9vIqeVI9A IequINT

== = = — (I+NDZ °° °*9vI1Q9z10A D1IDvIOY TL

a = — (I1+0)I — */-€ 9BIQ91IOA [BOTAIAD

to ee eee sIXy

ae ae = ee = Zo. aude asians sew

(1+0)I com @ Gaex()) | = — (1+0)1 — (srt (era

(I+0)I oe = a = =e os

I °° ofApuoo ie[ngipueyy

Toccrcttttsees RI[IXeA]

a = = — BO B10" a[Apuod ][eIIdISDO

an ie ae bie Hoon joyyUOIT

1 -» I Z * UONNuUep 1ejnqipueyy

lela hil I ti

“SPIAOG WNIpow o31e 7]

p FTav

13 14 15 16 17 18 19

TABLE 5

Large bovids.

ANNALS OF THE SOUTH AFRICAN MUSEUM

—

| |

I el he ta ddl db all

oi See ee ee ee

ee ee Pe ee el

fs 1 pees me geiks Le sab S24 pate

a Ba = a a x 1 = =

1( & 10-+1) 10+1) 10+1) 1(0+1) 10+1) 10+1) —

21+1) 1€0+1) 10+1) 104-1) — — — 10+1) 1041) —

1 — 1 1

ih ee ee

1. a a a3

fen seen erie

lie tleadaiea

21+1) 10+1) 10+0) —

aohl lity tia tlalaa

2ai+1) 1€@+1) 10+0) 10@+1) —

=i ile stilt bhiddetdeadl-a)=-=

Peel IT lt bh beee tl Sy ri ttl ttt yl la

10+1) —

Pa a a ee he ei rte fs) at)

21+1) 10+1) 10+1) 10+1) 20 +1) 1 +1) 10+1) 10+1) 20+1) 10+1) —

Mott Li doaesas | |—-l—!]la=taw

. . . . . . Cotta To eb OG ao) 0) rg OO, fo. SO tO

ai rym Ay csty Sel, Eloi mneD ues rely tere PNURIMIS Ga; Gn. ie: tei’ cept ks

sc ONO ow. 8

MeO IT OW ae 6 OW cr 0

—distal ...

—distal .

Indeterminate distal

—distal ..

Sesamoids — proximal

Mandibular dentition

—distal ..

Tibia — proximal

Mandibular condyle .

AGIA Wee ieee kos Serer

Occipital condyle....

IMMA 56 8o6000bo0e

Brontletnc ayes + ars

Cervical vertebrae 3-7

Thoracic vertebrae.

Femur — proximal

Lateral malleolus

Naviculo-cuboid .

Magnum .

Scaphoid .

Cuneiform

Lunate ...

Unciform .

Phalanges—Ist .

Patella ......

Calcaneum ...

Astragalus....

Cuneiform 1 ..

Sacral vertebrae ...

Scapular:

Lumbar vertebrae .

PASISE ee ato Soros

(@e ib) —

Oi) ==

1(1+0) 10+1) 1@+1) 1@+1) —

21+1) —

metapodial ........

FLY 1h. 4 5).ccx: wee

195

BYNESKRANSKOP 1

(O+€)€

(7+)

(7+ DE

(+ Dz

(1+0)1

G+ )Nz

lon be

Ane ham

ey)

([+0)I

(+01 G+0)1

(O+ DI

(O+DI

(+6)r ©O+DI

(1+0)I

O+DI

(1+0)1

(O+DI

(O+ DI

(+ DI

(1+0)I1

(0+ DI

(O+ DI

(0+ 7z

(1+0)I

OO+DI

(O+ DI

(1+O)E

O+ DI

(O+DI

(O+ DI

(+0)1

WISP IDE

laacal [lias

Ae AN

uouowre

(9+0)9 (ZI +¥)9I

(1+9)L (T+ LD6I

(1+ $)9 (€+¥1)9I

9 €l

9 LI

9 91

€ 9

€ ¢

€ ft

(0+0zt (+601

(i+6)r &+5)6

(€+Dr (+8)7I

(S+O)s (€Z7+0)ET

(T+ )9 (T+SDLI

WANED TFTANNMNTAN

oa)

AMNMAAREOANANT HM

= 9

I Ol

I v

= ¢

(T+0)% (O1+ H) FI

(O+7c (1+8)6

(T+£)S (€+6)7ZI

(1+) (S+0DS1

I 6

iG L

I 9

I /l

I il

(0+%¢ (t+¥)9

(I+ ©)r G+IDSI

(T+DE (7+DE

(+ve (8+D6

(y+0)r (07+ DIZ

(0+ 02 (C+ ODI

€

m=ANO KR KN AN la

isa)

eee ee ee

eee ee eww we ew eee

eee ee

je1sip—

Doon oo Oo Gon DO OD

Gon oo ocr oO sjeipodeiofy

nC Oey boo snjeseisy

[eIsIp—

SONOS 60.0 bOOX siqng

eee eee ee ew oe

éndnddoceocd DON winyy

ceoeee

[eIsIp —

Sqr

[estp—

jeisip—

** yewIxoid — sniswinpy

ee) eyndesg

aries Bee 9P1Q9}19A [BIDS

“*** QBvIQGOLIDA IVQUINT

**** QRIGS}IOA SIDVIOY YT,

L-€ S[BdIAIOD

SIX

sey

VI[IXeA

* UOIII]USp Ieynqipueyy

‘+ g[Apuood Iepnqipueyy

REDRESS. agfApuods [e1dI999

“SnjjIns sns1iadywg

9 ATaV

196 ANNALS OF THE SOUTH AFRICAN MUSEUM

PIS TPP UP LEP 11t ae |e

PPE Pd Py ty 1 1 1 See

ei tlh ttt tt kay.

Pleblilhelitii i

Cee its tha hh ye

S |

+ |

PRUGRURMGCDURO enh itis) iad

SMC CUREPORPMmmemrrivi rl iitild

TABLE 7

Various species (1).

Peet hhh tia

load Ii bli ld Lee |

lilealeaetall!

Mebhri tli ll i i.e

slallolealta

CC en ne en ey nr any ae Cn SM COmeties Oo > 'O Ce oO 0 8 OG OO Re

ee eC en ee en en MTC mene ss O CF OF; OO OF FO GF AO OO Bei

a an eT ae en en ee ye Te Cmte eo OO & OF OO GC GC 0 GO Be!

CnC et ee nr ey nn ns Let yr eet oO te O TO CG PO (OO 0 DO me

en ee ee ey en ee oy en ee Pe ee COP Ow Oo 8 FON DP OG O OF 8

le ee ee ee ean a eC ee mt Met Ose OO; OO OO DO

Ce ee ee ee en en ee er ee OMe eet Gare Ol 1 OG me OM! OO 0

Ce en ne en ey er er ee ee en re MOCO MS OOF OO | OOO sO

ne ne Sat en Ye en Ya en ey Sn een COSCON ON On 0) OO HO! Oe he

er ae en er rn ee en er ee er ere meOn cd OO OO oO! oO Oo CL OD Be)

ee nn ee eT er an i ee ae remem mmm OV ONO OC GC Oo 9 GO 88

ee Ef Sf kk

rT erm no oo Gl PO UG \

Fe ne en ey i me OTe UNECE Ene OD OOF 6 8 Om GO {

Qo bee pt Pa oe Or

= ’

S| if i: Dre ieee eo ees

Blog: + 2th: REI tn 2) 2 Se ees fe

R125 Ree LE ee Woeaante AS eis Ao cia Pegg Ele cs& casa ee |

° . ° . 2 3 ° — Chae ie

e|¢eo ae BS oe ee Se KS.c=k%S -: - 2 de

SO re eee ae i ae oS c2e80n 2, 3

SoS IO il by | eS ele Sos eae eee ee

So) ee Gop ote oA 7 ere) . Seite | oS See .

SS) Bo 5g 6 BIDS Ss nw ” .€. So CR:

Olas Se Ole Siais os Sis Bese Pesos See

= l/gpeeaeerbekakg go & (SGI S 3 SSE

S| SegSe5esee 2 8 Fes S Ge oie oe

(3) jonny pos! or i Sj

CWIOBAAK40R AGAR 4 DPD OUS4AE B&B & BOO 4m

197

BYNESKRANSKOP 1

I CeCe CnC CCC Cm) jewixoid — SPIOUIeSasg

I Cee sosurleyd

te eee tte tee ew ee ete ewe eee sjesiey,

Me eee ak cia ct ae ec ero uonnused

=i = = a = = = = I

I I I I I Te I I re

61 LI oT SI al CI IT 6 8

SI KPI

oT A!

aa aoa = aay a =e a ae aE I

I Ey a ag = I L ara sae I

= ae x a i = = (rol a =

= a aa ae se = = CFO = =

(eC OCL ee i een ee ee = a

a = I - atl = wa ee roa I

= a I I I I sy =a, C I

= ai = = =" = = = I I

arse es 7 ar —_ cr r aT ee cr i — nT _< ae,

en ele ssi i=!

J See SoroegsesonEvabono0sne uonnueqd

9 S1U4091G SOsDIIG

ee et ee ee ew we ee ee ee ee ee eee sqry

= dgnsc0noobcDDD ODD ODO OODONDD sosuryeyg

I silaiicikclicitcWetefehicnstislisiialislKolicieliololsliatre’ s]eIpodeioypy

tt te te we te ew ee ee ew [ewIxo.1d — eiqry,

(I +0)1 com@oooOdddoaoboo00GgGDGODD yeIsIp—

61 +0)I PaO CnC CnC mC CnC mim jewrxoid — INUW9,J

SS copodcons do oboosobGGoD0UNN0 sjedieg

— «(Con Goo DODO poo Ob OKod Oo GOOGS [eisIp— eulp

tt tt te et ew ee ee ee ee [eIsIp— snIpey

a +0)1 ee eee twee eee wees jewrxoid — sniowinyy

Tie, edhe hee recess koe ee yyndeos

So Go0boesoCG obo odouGG 9vIGO}IOA IEquINT]

— oo Od DOHDDOHOOON DD SVIGOIIOA DIOKIONU T,

wt et te ee ee tee eee ee L-€ S[edIAI9D

WS boon coco oo eo cro moo mo SG oom G a]qIpueyl

ee ee OC Ca BTIXe A

I “anu “ds 19 ‘ues sepIIodeT

TABLE 7

Various species (1).

Arctocephalus pusillus 1 4 5 6 7 9 10 11 12 13 14 16 17

Occipital condyle .................-. 1 —_— — — — = — =

Mandibular condyle . . = — = 1 = ES ak = a =¥ A ma =

Dentition ........... 1 1 1 2 1 — — — 1

Adlasi teas. = = 1 = me = oe ef e = =

IBSIS cite. seis = = == ae = a3 ae Aas 7 & = =a. A

Cervicals 3-7 . 1 = 1 = ee = = = = = = a: =

Thoracic ..... _— — 1 — — — = = — = ae =e =

MEUM DAD state sferdtsi ets Wisiaiaraeis(cisie wd 1 — — — — = = = = = aa =

Sacral .. pastel estes cin oreteion — —_— 1 — — — — = — = me - =

Scapula ........... 1 — — — =— — = = 1 a a es =

Humerus—proximal . 1(1+0) — _ = = — — = = = = re me

—distal .... = = = 1(0+1) = es a Ee az

Radius—proximal .................4 11+0) — 1(1+0) — = = = = Es a me a Es

Sadlistalyacne stim scaatreee cies — = = 10+1) — ae = = s.

Ulna—proximal .. 21+1) — 1+0) — = = = = = = es, i.

—distal...... Ee = 11+0) — = = a = aa i 1 we a

Carpals ..... 1 _— 1 — 1 = = = = Ls = =

Minas eis igisteace sianete scaievelo ioe — = = — ced = = =

Led vo ia gon breesenode — — = =

LSS ee pie a a = = pa = as = we Pea ma a <a

Femur—proximal . — — = 1(1+0) = rest == ee = ca as, <i =

—distal..... = = == AGI) — os a = ae 5S

Tibia—proximal .. — = 1(1+0) o

—distal ..... — — — == =, aay se a

Fibula —proximal. . = = = = a = nas a sS ~ a

—distal .... = = 11 =e = = = ae =

LECH EY Sees 1 — x i = = = ae a) —s = a ee = =

Calcaneum .. — — —_ 1 _ = = — = — az = =

Astragalus ... — — = 1 = pass <e ae 1 — ee et =

Other tarsals . 1 1 1 — — —_ _ — os = ae

Metapodials .. 1 = 1 1 1 = = = 1 = a ce =

RPM An esas sce eek on tenns so aye 1 1 1 1 1 1 1 1 1 oe 1 1

Leporidae gen. et sp. indet. 1 Z 6 8 9 10 11 13 14 15 16 17 18 19

WUC RTTE Seog eicheo Deen COE IO ernte — _ _ 1 1 = = = = = = —= — =

Mandible .... — _ 1 2 —_ —_ 1 1 1 1 _— 1

Cervicals 3-7 ..... — — a = = = = = = = = aes me itd

Thoracic vertebrae .. == 1 = = = = a? a = aa cat a= ie =

Lumbar vertebrae ... = — v3 | 1 = — _— _— = — 1 ae = 1

Scapula ..........-. 1 — — — —_ 1 = = — — = 3 = ee

Humerus—proximal . 1(0+1) — — — — — = — 1(00+1)

Radius—distal ...... — —_— _— _ — — — — — — = 1(0+1) — 1(0+1)

Ulna—distal . . . _— — — — — 10+1) — = = = = = = =

Carpals .......... aa = = = = Ss = =e a ae oe , =e 1

Femur—proximal . 10+1) — = = = 10+1) — cl = = = = = ‘Sep

—distal..... 10+) — — = = = =e = Ara)

Tibia—proximal . = = = = = =< re a 10+1)

Metapodials .. . 1 _ _— 1 == = 1 1 = = = = 1 1

Phalanges ... bie = = — — = Trl a a = 1 1 -

RID Sihenininiyeiara Ack feleieta aleteiabe wetekpeles, ve = = = 1 — cs — ad al mp Get = t= Pa

Diceros bicornis 6 9 11 14 16

MenteGns com as -isibinttle mm arstacasaias ve 1 I 1 Icf. Icf.

Rhinocerotidae 1 a 6 8 9 11 12 14 15 16 17 19

Men tition eer teaceeen oeee ames 1 = 1 = 1 1 1 1 1 1 1 1

Tarsals ... _— 1 — — _ —_ — st — = oe =

Phalanges ........... nee! —_— — 1 — —

Sesamoids—proximal ............--- 1 _ = — — = —

961

WOgSAW NVOIMAV HLNOS AHL JO SIVNNVY

1 dONSNVUNSANAT

L61

198 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 8

Various species (2).

Hippopotamus 1 5 6 7 13

DENtitiony s2.8 seis 1 1 1 1 1

Atilax 14

Miandibles assoc aoe aca 1

Loxodonta 4 5 14 15 16

IDSNtitlOMN. ele sases a es) eoevene cre — — 1 1 1

JONES 5 oo0a000ccc000000 1 1 — 1 —

Hystrix 1 2 5 15 17

IDENtitION: Jess aenme Boe 1 1 1 — 1

Tibia—proximal .......... — — 14+0) — —

== GS tall eevee. eccre iets cians — — 1(1+0) — —

IMIGTAOCKIS ooooccocuc0000 — — 1 — —

VAISS on ccccod00000000 — — — 1 —

Panthera pardus 1 5

Carpal. ee oes cusiemctely sees — 1

Phalanges! pence deeeioric 1 —

Mellivora 1 10

Mandible: re ucsnae ceerine — 1

IA XISIRE seni otiede ie cette mace 1 —

Wina—proximeallipeeneeneieee 11+0) —

= GiStalit pelea e 11+0) —

Herpestes pulverulentus 1 3 4 5 6 7 8 13 14

Occipitalicondylese na. soa. — — — — 1 — — —

Miaxullaeec ote tcausesiseetctet — — — 2 1 — 71 a =

Mandiblemn sane eee — — 1 1 — as

Lumbar vertebrae ......... — — — — — — — 1 —

Humerus—proximal ....... — — — 11+0) — — — — 1(00+1)

== diStall name — — — 100+1) — — = — =

Wina— proximal cer — — oe 1(1+0) — — — = =

==(GiS tallbi so Gereteee tere rene gerey — — 10+1) — — — = — =

AUT Eee caste act ote.cretetiee — — 1 — — — == = =

[sehium 3 ak schoo — — 1 — — — — = =

Pubisi tsa nicer ee re — — 1 — — — = — —

Femur—proximal.......... 100+1) — — — — — = = =

= GiStald epestcactnarece 10+1) — — — — — = — =

Tibia—proximal .......... — — 10+1) — — — — —

= distally ei cwasts eee — — — 10+1) — — = = =

Eibula—distallaanaeeeenoee — — — 104+1) — — = = ==

IAStragalusmemmeeeioee ieee — ?1 — — — — — —

MetapodialSsannncneeencen 1 — — — — — — — 1

Phalangesman cern onienae 1 — — — — — — 1

Herpestes ichneumon 1 4 5 8 14

Miaixallatescps cy cteiene coe ra hewn: — — 1 — 1

Miandibles vce nates cee 1 — — — 1

Eumbanverteoracmameennerers 1 — 1 — —

Mumerus—distalin.. asec — — — 10+1) —

cf.

Ulna—proximal........... — — 10+1) — —

— distal no. anna mere — — 10+1) — —

Aiibia— proximally. aces — —_ 1(00+1) — —

= GiStal- cies veer ans — — 100+1) — —

Fibula—proximal ......... — — 10+1) — —

= distally Lisas ocean bees — — 10+1) — —

€gleaneumi nee 1 — — — —

Metapodialsies aces acl — 1 il = =

(continued opposite)

BYNESKRANSKOP 1

Papio ursinus 1 4 5 6 7

IMUaxallay sete ievaren isahe steno ces 1 — — — —

IManGhollS Goococecovs00c0s 1 1 1 1 1

Gl AV ICIS. o aisicte law saeco teens 1 — — — —

Canpalsiisie skeet eases 1 1 — 1 —

MG ATS AUS ceed atsnaseancherat- tes, ode teats — — — — —

Metapodials .............. — — — 1 —

IRHAIAN GES) 5 keke cease — 1 — —

Felis cf. caracal 1 4 5 6 9

DEntitiONy 2s. Wok wares dees 1 — — — —

Ulna—proximal........... 10+1) — — — —

= distal cc ces eee ees — 1(0+1) — —

Wanpalsh saison soe ees Wea — — — —

Tibia—distal ............. 10+1) — — —

Metapodials .............. 1 — — 1 71

Rhalanges: peda ce shows etns 1 1 — 1 1

Felis libyca 1 2 3 4 5

IDENTILIONY “an cow sc ee ome me 1 — — 1 —

Genvicals) 3=7 56st au ww os — — Icf. — 1

Lumbar vertebrae.......... — — — — icf.

Radius—proximal ......... — — — —

@anpals! sanstemes ae accuse gers — 71 — — —

UD wan Feats cepre oroneh Ocoee Oise — — — 1

LSC F hres cise ae naresec — — — — 1

RalbiStienraa crores cveroetcrara ene — — — — 1

Tibia—proximal .......... — 10+1) — — —

Metapodials .............. 1 — — — 1

IPhalanees™ js jcc cscs cae eee 1 — — 1 1

Canis cf. mesomelas 1 3 4 5 6

Occipital condyle .......... 1 — — — —

IN Taxa lai ers traitiiece stvig arse ero 1 — 1 — —

IM [enayelil NOS Sesto ge om arco orm 1 — — ?1 1

Canal sirens og seesianouaters8 — — — — 1

Calcane une % coeeses alee — — — — —

Mietapodials) sae sce eeecc os 1 — — — —

Phalanigesie.. ciaeave. os moos 1 1 — 1 —

Homo sapiens 1 5) 6 10 14

DEMON rics Benes Sa ais 1 1 — 1 1

PUL EUS tctis, resis efficacies nye teclavs 1 — — — —

Tibia—proximal .......... 10+1) — — — —

“TEESE Ge acatal Secon ce eRe eon eee — — — 1 —

phalanges! Gacnss+ oer acess 1 1 1 — —

Aonyx capensis 1 5

Mandible: is,s.00¢ecen0 «sme — 1

INSGIS 16 Binde OGIO DEPOT Seon — 1

Thoracic vertebrae ........ — 1

Sacralivertebrae armen eee — 1

Radius—proximal ......... — 10+1)

—=GIMIEY! oosop00c0000 — 100+1)

Metapodials .............. 1 1

elrallanigesicvaen. a cisiaosswanoueoae 1 1

Hyaenid 1 5 14 19

Den tition Aeris be eh — —_ 1 1

Phalanmges! ~<a oe Seen 1 1 — —

Ictonyx striatus 1 5

Humerus—distal .......... 10+1) —

Ulna—proximal........... — 1(0+1)

9 11

1 =

aie 1

s 1

1 a

6 13

1 —_

aS 1

= 1

10 14

21 1

= i

199

10+ 1)cf.

200

TABLE 9

Various species (3).

ANNALS OF THE SOUTH AFRICAN MUSEUM

Procavia capensis 1 4) So Gee a 28 > t- 12 18) ts 16 17 19

Mia xa a recess dra teeneie a acre ars 1 — 1 1 1 — 1 1 1 1 — 2

Mandible: <:ccsdseenmannas i — — Po es a = I 1 1

PALL als tes eirrtesetena torsuetie tes aseueoure obees — = = = = —& es ee es 2

UNERAS sas eels ee cater waroe oiehs ae eTeue 1 — SS eS =, oS) 2 = eee eS =

Cervical vertebrae 3-7 ..... 1 es, Fe Ses fey Se ee Meco Sere Mca) ee ice rag eee Cs

Thoracic vertebrae ........ 1 — — 1l— — 220 See SS ee SSS aoe

Lumbar vertebrae ......... — — — 1i—_—— Se eae ES See =

Scapulatcirsmerce seem: 1 SS eS eS ee NEN Bec es oy ee =

Humerus—proximal ....... QO) ss See fh peer eee Pe aN | yh "be

SOS sascoconce AM sei) Se | ee ee (Ossi)

Radius—proximal ......... 10+1) — — — — 1041) — — — — ~—~-— — =

= distal sive asiceneee = en er ie a a) a:

Ulna—proximal........... 10+1) — — — — — = =| S| SS Sl 1(1 +0)

JUwia nee Aer avaeediaco csoaiateoorcinte 1 See ee fe egy eg eens ees, ys, os

TSCM 50.5 eee een 1 ee 2 pee Se es ee a

Bub isi aceon ence 1 SS) ee Ee ees ES ee oe ee ee =

Femur—proximal.......... 1@s+1) => #« os = = S SS |S SS —S — 2(1+1)

— distal. vt gesmee. — aS eS So es we ee a ea eee aa

Tibia—proximal .......... 1a+0) — — — — 10+0) — — — — — —- — 1(1 +0)

iS tale ioy.cccre le eae = =S =a = i400 ] jas =] = = ae

Ratelliay sencrc cuvette ane = a ee ge ee, ee 1

Galcaneumte oan. cance ae: — — = Ss = = = es ee ey ee Lz

Astragalltis” (csnvceucceta cee — —- —- —- — — 1 = = t —- — — —

IMIeeNoXeGhAUIS 5 5500000000006 1 —- l—_—-— — LS Se —

iDentitionaa seen 1 2D 1 1 1 1 — — 2 — 1 1 —- — —

UhA=prowimel ooo = = = = IG4Q0 —] a= ]eHeieaee See Se SS S| —S |

= CISA 600066 ee ee) SS ee ee a SS SS eS SS

CAMOBUIS Soo000c00c — | 1 — — — | — ~ ~ ~ ~—~ 1 — — 1 — =

iEarsalsy! Batt in veces —a => = = oS wae Owes ee a ge ee eee

Metapodials ...... 1 1 1 1 eS a a)

Rivalangeseeeeere 1 1 1 1 1 1 1 1 1 — 1 1 — 1 =— | 1 1

Potamochoerus porcus 1 DANS “4 ein ay) sees 9 I@ il Ww We WI ils IG 7 iB 9

IDOSNOMOM oaoccacccccs Ue DY De D's aps ae lo fl Ll — =| ff =i 1 —-— — —

Phacochoerus aethiopicus 1 2 Se |S) 6 IS 8 109 2 Ss AIS Ger omen.

IDentition Ener ee be ek! See” WE ee ee ed ee ee

Equus cf. quagga 9 10 11 12 13 14 15 16 1 19

IMiaxdll aii earnest eeeaene — — 1 1 1 1 — 1 — — =

IMEVAVIONS sooococcoonsoces 1 — — = 1 1 Dp 1 2 — 2

Marsal Swe etenrvs nity oe te poke — — = = — = = = 1 == —

MetapodialSsaenennene reer: — — — 1 = ats — a8 = oe =

Phalanges— lst see none one — — — — 1 — — — — — —

Sesamoids—proximal ...... = = — — — 1 — 1 — — 1

CGA ogoocnoes — — = — 1 = = é== —= = =

Equus cf. capensis 19

IMAVENIDID. soovococccoccece 1

Alcelaphus|

Connochaetes

dP4UW

* indicates fragment

BYNESKRANSKOP 1

TABLE 10

Bovid dentition—various species (1).

1 ee

— 1

1 —

me A

M2UW M3UW

U L U L

| pa! | Geen.

—_- — gh

aig pean Goat

eee one |

hae ae cee ee

Se nis ge a)

DI =e

—- — Po Ot

a | oe

Nee eae |

— 1

201

Totals

men |

ANH | NwuUwe | = | New

Wee NYRR RPK ONWNABRNKH WN

ANKE KP RWUWNDN RK KH New

Ree NRF RP Re Re AWWW ARNK OD

Be Ne

202 ANNALS OF THE SOUTH AFRICAN MUSEUM

TABLE 11

Bovid dentition— various

species (2).

Raphicerus spp. dP4UW MiIUW M2UW M3UW P4UW P4x/+

MM FEF Wo IL Wf IE, Ur Wie Ue E We It

2 2 1 2 4 — 1 — 1 1 a DD

— 1 2 1 1 t — — — a ih al

1 — 3 i al —- — — — == Beh.» (eae EAS)

= 4 to 2 tye oe —- — — 2 — 1 1, 32

2 2 5 SS — 2 4 3 2 6 i — Sin) OS)

ees == 6 ee | os | = 2 —_- — D 2

1 = aT a —— — 1 De ee —- — We 2

— 8 Bo —_—- — 2 — 1 —- — 1 1

— 9 lil —_—- — —_- — 2 — a 1D

— 1 10 — | —- — —_— — —_— — — | — |

— 11 — 1 —_— — —- — —- — — = —

— == 2 taal —- — —_- — — 1 = = =

1 — 13 — 1 —_- — a ——— i ees |

1 = 14 eS al —_— — —- — — 1 — = cya |

— 15 —- — —_— — 1 — a = = |

1 1 16 etl = — —— et ey =e Se

— 17 — |i —- — —_- — = = Se ice

ee i ee ae Ree lar Sie Men 1 tem ee eee a8 eS

1 — 19 1” = —- — —- — — 1 —_— — —_ —

Raphicerus campestris

Sierras eerie epee 1 — 1 — = 1

De er roeeke ic tannn aeve — — —= _ = a

SI a at aioeesttnee tas — — — aes cow vase

Bison stoadele s snaihenetee — — = — 1

Seat a Rar EG 1 — — 1 = ae

Gis jena care aneaxt — — — 1 = a

Fp nseeea tates cro ware — — — = ae oe

See akoe slau deme — — a = = ae

OP rere nes) eins eke — = == — es sar

INO aeaenres cleyeconatecoro. — — — — = ae

Lees epeaeieeces santnlaaete — — — = — =

Dae ale losies tabate teyaeie — — = oe a wes

1S Roe Tees costabriay — — — =e i ae

De os ake eteeneseanairs — — = = ans ae

LS ire tocgee Sec tran ats — — == we as, aa

LNA arahectaen chatter — — _ — = saa

Tig Aa ee rt nonce ate — — — = =— ae

DS essa eer ope — = _— —_ ace =

ee Sena reac reat ne nee — — oe = ae ae

Raphicerus melanotis

Wore Sinacsete taser 2 1 — 1 == =

Dias fea teetanra env seh eats 1 1 — = == —

B tire arene tenes Sete — — = ae as 2

A ete sen Bietesis on shece 1 — 1 2. 1

SD) Ae hifeneteg uavoustenair 4 1 3 6 — 5

OU Bee nianenareeniwss — 1 1 — 1

Wostceenaiarea sires beiaheaage — — = == = 1

Sires wae mere — = —= a ae a)

Dip era eh aie ee cies — — = = = =

LLU ators orcha ee eto = = — = = =

1 hares neon ieee — == = a = —

Dy orpenctene ecard ae 1 — == = a= ==

IVS Feet oiettewe use cei coer ete — == = aa _ a=

ASA Stunt soon Goeters — — —= 1 a ey

Morn ateromhoncio rt 2 = a <= — a ae

HOM ereisis cau ano erate 1 — — — = mt

1s aay Ata Niene oc — — — — poe —

Ovis aries

dP4UW MIUW M2UW M3UW P4UW P4x/+

MM FF WW IL U L UW ok wy It Wi ok, Une

OP et | a) ON Cte) haere Lae oe) eS nee

| a | S| | PRRRANWNAUN

(=)

Totals

L U+L

pct foul Palette chee ol: [ict omens eee ipa Ban ee eee eek Bae ye Qe nee ee

wv | NNNANYRKWUUDCHWOARLO

el eget cme re

Pelea capreolus

BYNESKRANSKOP 1

TABLE 11 (continued)

MM FF dP4UW MIUW M2UW M3UW P4UW P4 x /+ P44+ + Totals

— — 4 = = ———— a a — —=- jl — — — | 1

— — 6 —- — —_—- — —_- — — I! — — —_— — —- — — 1 1

— — 8 —_- — — 1 — a — — = =— — — 1 1

1 — 14 — = — = a ————— = — 1 — —_—- — 1 — 1

Oreotragus oreotragus

dP4UW MIUW M2UW M3UW P4UW P4x/+ P44 + Totals

Nee casket cas tasty sae — —_- — — — i 1 — 1 — —- — 3 — 8

Se Mic hatsneins tn wiiee — 1 —- — —_ — —_— — —_- — | —_- — 2, @ 3

Oe ativan aah Crea een — —_- — —_- — —_—- — —_— — — | —_- — — jl 1

Pd i dace Pace cere ee 1 — ————— —— — —_- — —_- — —_- — ine

IIS) aes ae rOaose RNC ene —_—- — —_ — = = — a = il — — 1 1 2

Redunca fulvorufula

dP4UW MIUW M2UW M3UW P4UW P4x /+ P4+ + Totals

il eat eae eee —_—- — —— — ————— a 1 1 — — 1 1 1

113): Gcateeatacelaeceerieeanans —— —_- — —_—- — — 1 —- — — 1 — — —— 2 2

Aa as eet Osco — — a = «= — il — — —_- — — — 1 1

Redunca arundium

MM FF dP4UW MIUW M2UW M3UW P4UW P4x/+ P44 + Totals

iW Ib WU 2k WIL WIL W IL W Ib Wo IL, Wy 1G WEIL

— — 6 —_—- — —_- — — — — 1 ————— —_ — —_ — —1 1

— — 10 { — = = = = a — = a — = 1

— 12 — a a — — —_—- — — 1 —_ — iL al 2

1 — 14 — = ——— ——— ——— a — = =< — |* 1

— — 15 —_—_ — — a — i — = = — — — — 1 1

— — 16 —_—- — —_ — —_ — —_—- — = il —_ = = == jl 1

— — 19 —_ — a ——— a —= = = i —_- — — 1 1

*dental fragment

Taurotragus oryx

RAR cse cos: Seckits —_—- — — — = — 1 —_ — 1 — —_- — ia 2

Giese t acts olalscare — 1 — I —- — —_—- — —_- — —_- — —_— — —2 2

17/5 acre ee een — i a ———— —= = — — —- — 1 => i

1D) A aletoue aeeean ena e 1 1 —_- — 1 — — = — — — — — 2 il 2

Hippotragus

Aas Shearer oi an Seeeaeires —_ — = i ———— a —— a —- — — 1 1

5). Soa Gale sto eRe oan — — — 1 —_ — —_—- — —_— — 1 — — — ie 2)

Sierra ieataeset tee is — ——— = a a ae 5 pel n= ol

Oo catonepenaratc canon ae —_- — —- — —_— — —- — —_- — 1 — ———— ti — 1

OS cy aes tans eave rane i — — ———— 1 — —_—- — —_—- — — 1 —- — iol 2

Mice BEN ae cecescys — = = 2 ae ey ra Sey ee ila ye | (ee |

1B) irre ee ue re pee — 2 —- — — 1 — — —_- — — ij = — — 4 4

RA ea cect che oni Goan —_— — — = = —= = ————— 1 — —_- — a el

lO acatct Serer a eae = = —— — — — == — 1 — — — | => il

Fe ae eco cs fo 5 Be — 1 — —_ — —_ — —- — 1 1 1 — 3 il 4

203

: . : . F a,

; % ; x 5 a = 4 : " :

“ 1 \ #1 Nea = j . fd

6. SYSTEMATIC papers must conform to the International code of zoological nomenclature

(particularly Articles 22 and 51).

Names of new taxa, combinations, synonyms, etc., when used for the first time, must be

followed by the appropriate Latin (not English) abbreviation, e.g. gen. nov., sp. nov., comb.

nov., syn. nov., etc.

An author’s name when cited must follow the name of the taxon without intervening

punctuation and not be abbreviated; if the year is added, a comma must separate author’s

name and year. The author’s name (and date, if cited) must be placed in parentheses if a

species or subspecies is transferred from its original genus. The name of a subsequent user of

a scientific name must be separated from the scientific name by a colon.

Synonymy arrangement should be according to chronology of names, i.e. all published

scientific names by which the species previously has been designated are listed in chronological

order, with all references to that name following in chronological order, e.g.:

Family Nuculanidae

Nuculana (Lembulus) bicuspidata (Gould, 1845)

Figs 14-15A

Nucula (Leda) bicuspidata Gould, 1845: 37.

Leda plicifera A. Adams, 1856: 50.

Laeda bicuspidata Hanley, 1859: 118, pl. 228 (fig. 73). Sowerby, 1871: pl. 2 (fig. 8a—b).

Nucula largillierti Philippi, 1861: 87.

Leda bicuspidata: Nicklés, 1950: 163, fig. 301; 1955: 110. Barnard, 1964: 234, figs 8-9.

Note punctuation in the above example:

comma separates author’s name and year

“semicolon separates more than one reference by the same author

full stop separates references by different authors

figures of plates are enclosed in parentheses to distinguish them from text-figures

dash, not comma, separates consecutive numbers

Synonymy arrangement according to chronology of bibliographic references, whereby

the year is placed in front of each entry, and the synonym repeated in full for each entry, is

not acceptable.

In describing new species, One specimen must be designated as the holotype; other speci-

mens mentioned in the original description are to be designated paratypes; additional material

not regarded as paratypes should be listed separately. The complete data (registration number,

depository, description of specimen, locality, collector, date) of the holotype and paratypes

must be recorded, e.g.:

Holotype

SAM-—A13535 in the South African Museum, Cape Town. Adult female from mid- tide region, King’s Beach

Port Elizabeth (33°51’S 25°39’E), collected by A. ‘Smith, 15 January 1973.

Note standard form of writing South African Museum registration numbers and date.

7. SPECIAL HOUSE RULES

Capital initial letters

(a) The Figures, Maps and Tables of the paper when referred to in the text

e.g. °... the Figure depicting C. namacolus ...’; ‘. . . in C. namacolus (Fig. 10)...’

_(b) The prefixes of prefixed surnames in all languages, when used in the text, if not preceded

by initials or full names

e.g. Du Toit but A.L.du Toit; Von Huene but . F. von Huene

e.g. Therocephalia, but therocephalian

Punctuation should be loose, omitting all not strictly necessary

Reference to the author should be expressed in the third person

Roman numerals should be converted to arabic, except when forming part of the title of a

book or article, such as

‘Revision of the Crustacea. Part VIII. The Amphipoda.’

Specific name must not stand alone, but be preceded by the generic name or its abbreviation

to initial capital letter, provided the same generic name is used consecutively.

Name of new genus or species is not to be included in the title: it should be included in the

abstract, counter to Recommendation 23 of the Code, to meet the requirements of

Biological Abstracts.

FRANZ R. SCHWEITZER

M. L. WILSON

BYNESKRANSKOP 1

A LATE QUATERNARY LIVING SITE

IN THE SOUTHERN CAPE PROVINCE,

SOUTH AFRICA

m2 JUNE 1982 | ISSN 0303-2515

yy,

OF THE SOUTH AFRICAN

INSTRUCTIONS TO AUTHORS

1. MATERIAL should be original and not published elsewhere, in whole or in part.

2. LAYOUT should be as follows:

(a) Centred masthead to consist of

Title: informative but concise, without abbreviations and not including the names of new genera or species

Author’s(s’) name(s)

Address(es) of author(s) (institution where work was carried out)

Number of illustrations (figures, enumerated maps and tables, in this order)

(b) Abstract of not more than 200 words, intelligible to the reader without reference to the text

(d) Introduction

(e) Subject-matter of the paper, divided into sections to correspond with those given in table of contents

(f) Summary, if paper is lengthy

(g) Acknowledgements

(h) References

(i) Abbreviations, where these are numerous

3. MANUSCRIPT, to be submitted in triplicate, should be typewritten and neat, double spaced

with 2,5 cm margins all round. First lines of paragraphs should be indented. Tables and a list of

legends for illustrations should be typed separately, their positions indicated in the text. All

pages should be numbered consecutively.

Major headings of the paper are centred capitals; first subheadings are shouldered small

capitals; second subheadings are shouldered italics; third subheadings are indented, shouldered

italics. Further subdivisions should be avoided, as also enumeration (never roman numerals)

of headings and abbreviations.

Footnotes should be avoided unless they are short and essential.

Only generic and specific names should be underlined to indicate italics; all other marking

up should be left to editor and publisher.

4. ILLUSTRATIONS should be reducible to a size not exceeding 12 « 18 cm (19 cm including

legend); the reduction or enlargement required should be indicated; originals larger than

35 <x 47 cm should not be submitted; photographs should be rectangular in shape and final