Dean Falk*, John C.Redmond, Jr andJohn GuyerDepartment of Anthropology,University at Albany, SUNY,Albany, NY 12222, U.S.A.E-mail: D.Falk@Albany.edu;Redmond@empireone.net;Jguyer@capital.net

Glenn C. ConroyDepartments of Anatomy andNeurobiology/Anthropology,Washington University Schoolof Medicine, St Louis,MO 63110, U.S.A. E-mail:Conroyg@thalamus.wustl.edu

Wolfgang RecheisDepartment of Radiology II,University of Innsbruck,Anichstr. 35, A-6020,Innsbruck, Austria. E-mail:Wolfgang.recheis@uibk.ac.at

Gerhard W. Weberand Horst SeidlerInstitute of Human Biology,University of Vienna,Althanstrasse 14, A1091,Vienna, Austria. E-mail:Gerhard.weber@univie.ac.at;Horst.seidler@univie.ac.at

Received 6 May 1999Revision received 2 August1999 and accepted5 November 1999

Keywords: Australopithecus,endocasts, frontal lobe,paleoneurology,Paranthropus, phylogeny,temporal lobe.

Early hominid brain evolution: a new lookat old endocasts

Early hominid brain morphology is reassessed from endocasts ofAustralopithecus africanus and three species of Paranthropus, and newendocast reconstructions and cranial capacities are reported for fourkey specimens from the Paranthropus clade. The brain morphology ofAustralopithecus africanus appears more human like than that ofParanthropus in terms of overall frontal and temporal lobe shape.These new data do not support the proposal that increased encephali-zation is a shared feature between Paranthropus and early Homo. Ourfindings are consistent with the hypothesis that Australopithecus afri-canus could have been ancestral to Homo, and have implications forassessing the tempo and mode of early hominid neurological andcognitive evolution.

� 2000 Academic Press

Journal of Human Evolution (2000) 38, 695–717doi:10.1006/jhev.1999.0378Available online at http://www.idealibrary.com on

Introduction

Much of what is known about hominid brainevolution has been learned from endocranial

0047–2484/00/050695+23$35.00/0

casts (endocasts) that reproduce details ofthe external morphology of the brain fromthe internal surface of the braincase.Because these endocasts are usually fromfragmentary pieces of fossilized skulls,their missing parts must be reconstructed.

*To whom correspondence should be addressed.

� 2000 Academic Press

696 . ET AL.

Discoveries such as KNM-WT 17000 (P.aethiopicus) and KNM-WT 17400 (P. boisei)(Leakey & Walker, 1988; Walker et al.,1986; Brown et al., 1993) provide evidenceof previously unknown parts of the Paran-thropus brain. Prior to these discoveries,Paranthropus endocasts were sometimesreconstructed using endocasts of A. africa-nus (e.g., Sts 5) as a model (see below). Inthis study we compare endocasts of Paran-thropus (including P. robustus, P. aethiopicus,P. boisei) with those of Australopithecus afri-canus and identify, quantify, and interpretpreviously unknown differences in the fron-tal and temporal lobe morphology betweenthese genera. In addition, we provide revisedestimates for the mean cranial capacity ofParanthropus.

Materials and methods

Previously unknown parts of the cerebralcortex in Paranthropus were observed andmeasured on both the endocast ofKNM-WT 17000 (P. aethiopicus) and on asilicone endocast prepared from a cast ofKNM-WT 17400 (P. boisei). Correspondingobservations and measurements for Austra-lopithecus africanus were obtained from sili-cone endocasts prepared from museumcasts of Sts 5 (Mrs. Ples) and Stw 505 (Mr.Ples), as well as from a copy of the naturalendocast of Sts 60. Other endocasts used forcomparative purposes included KNM-ER23000 (P. boisei), Sts 19 (A. africanus) andthe Sterkfontein Number 2 natural endo-cast (A. africanus). Comparative endocastmeasurements were taken (by JG) from tengorillas (G. gorilla), nine chimpanzees (P.troglodytes), nine bonobos (P. paniscus), andten modern humans. Associated cranialcapacities were obtained with mustard seedfor the gorilla and chimpanzee sample andfrom the literature for the human, bonobo,and early hominid sample.

As a preliminary step, GWW and DFvalidated the size of the silicone endocasts

for two of the specimens (Sts 5 and Stw505) by comparing several measurementsobtained using calipers with measurementstaken on their corresponding virtual endo-casts that had been acquired with 3D-CTtechnology from the original skulls (Conroyet al., 1998). The maximum length, height,and width obtained by measuring the virtualendocast of Sts 5 on the computer screenwere 0·98, 1·00, and 1·00 of the respectivemeasurements obtained with calipers fromthe silicone endocast. The length of thefragmentary Stw 505 virtual endocast andthe distance between its left frontal andtemporal poles were each 0·98 of the respec-tive measurements obtained from the sili-cone endocast. A third measurement on thevirtual endocast of Stw 505 (between itshighest point on the dorsal surface and itslowest point at the anterior end of the tem-poral lobe) measured 0·99 of the compar-able measurement of the silicone endocast.Thus, as detailed elsewhere (Weber et al.,1998), endocasts prepared from museumquality casts of skulls reproduce measure-ments obtained with 3D-CT technologyfrom the braincases of the originalspecimens with a high degree of accuracy.

Eight measurements (described below)were obtained with calipers from basal viewsof endocasts and projected onto the basalplane. The procedure for orienting an endo-cast in basal view is to first determine themaximum antero-posterior diameter of theendocast in left lateral view (using the righthemisphere, if left is not present) that con-nects the frontal and occipital poles asdescribed and illustrated by Connolly(1950:124–125). The endocast is thenturned upside down and secured so that themaximum anterior-posterior diameter is inthe horizontal or basal plane and the mid-sagittal plane is vertical. In cases of partialendocasts (e.g., Sts 60, Stw 505; KNM-WT17400), basal orientations were estimatedby aligning them next to correctly orientedfull endocasts from the same genus (e.g., Sts

697

5; KNM-WT 17000). The fossil hominidmeasurements were from undistorted andunreconstructed portions of endocasts. Inorder to reduce potential observer error orbias, measurements were taken together bythree observers (DF, JG, and JCR) on twodifferent occasions and the results averaged.JG took the measurements with sliding cali-pers, while JR and DF confirmed his selec-tion of landmarks and readings from thecalipers, and made sure the calipersremained oriented so that measurementswere projected onto the basal plane.

In order to quantify remeasurement error,the three workers together repeated all of themeasurements on the fossil hominids oneyear after the first measurements wereobtained and then compared their resultswith the earlier ones. For each of the eightmeasurements, remeasurement error wascalculated as the mean of the absolute dif-ferences (determined for each specimen)between the first and second sets ofmeasurements. Remeasurement error wasthen expressed as a percentage of the aver-age length for each measurement. Theresults were 1% for measurements 1, 2, 6,and 8; and ranged from 2–8% for the otherfour measurements. The highest remeasure-ment errors expressed as percentages ofmean lengths were for the shortest lengths.Similarly, JG remeasured 12 endocasts(three each from humans, bonobos, chim-panzees, and gorillas) one year after takingthe initial measurements from these speci-mens. The results ranged from 1–8%, withthe greatest relative remeasurement errorassociated with the shortest lengths.

The measurements included (Figure 1):(1) bat–bat, the distance between the mostanterior points of the temporal lobes in basalview; (2) mat–mat, maximum width of thefrontal lobes at the level of bat; (3) mbat–rof(tan), the shortest distance between themiddle of the line connecting the two batsand the tangent to the most rostral point onthe orbital surfaces of the frontal lobes (note

that rof should not be confused with thefrontal pole); (4) mcp–mbat, the shortestdistance between the middle clinoid process(or anterior border of the sella turcica) andmbat; (5) mcp–rof(tan), the shortest dis-tance between the middle clinoid processand rof(tan); (6) cob–rof(tan), the short-est distance between the caudal boundary ofthe olfactory bulbs (cribriform plate) androf(tan); (7) rob–rof(tan), the shortest dis-tance between the rostral boundary of theolfactory bulbs and rof(tan); (8) rof(tan)–bpc(tan), the shortest distance betweenrof(tan) and the tangent to the most

2

1

rof

rob

cob

5mat

bat

mcp

4

mbat

36

7

batmat

8

bpc

Figure 1. Measurements obtained from basal views andprojected onto the horizontal (basal) plane from endo-casts of australopithecines, apes, and humans (see textfor details and Table 1 for data). Landmarks: bat, mostanterior point on temporal lobe from basal view; mat,most lateral point on endocast at the level of bat inbasal plane; mbat; middle of the line connecting thetwo bats; rof, the most rostral point on the orbitalsurfaces of the frontal lobes; mcp, middle clinoidprocess; cob, caudal boundary of olfactory bulbs (cri-briform plate) in midline; rob, rostral boundary ofolfactory bulbs in midline; bpc, most posterior point oncerebella in basal view.

698 . ET AL.

posterior point on the cerebella in basal view(bpc).

Measurements 3, 6, 7, and 8 were used tocalculate three additional lengths for eachspecimen: [3–6] the length between mbatand cob; [6–7], the length of the olfactorybulb (cribriform plate); and [8–3], thelength of the basal aspect of the endocastcaudal to mbat. Indices that express eachmeasure as a percentage of endocast lengthin basal view were calculated by dividingthese three lengths as well as measurements1–7 by measurement 8 for the great ape andhuman samples, and for the two hominidendocasts for which measurement 8 wasavailable (KNM-WT 17000 and Sts 5).

Descriptive statistics were provided forthe lengths (Table 1) and indices (Table 2)obtained from endocasts of Homo, Gorilla,Pan, Paranthropus and Australopithecus.These data were first compared in livinghominoids in order to establish a compara-tive basis for assessing endocasts represent-ing Australopithecus and Paranthropus. For allof the comparisons in this study, there weresignificant differences between groups whichwere determined by post-hoc analyses ofselected contrast within the general linearmodel (GLM) of SPSS (version 8.0). Thealpha level was preset to P�0·05 aftercorrection with Bonferroni’s method formultiple comparisons where the indicatedP-values had been adjusted (Tables 2 & 3).Differences in the mean lengths and indiceswere also tested for statistical significancebetween the two species of Pan. The onlytwo measurements that were found to differsignificantly between P. troglodytes and P.paniscus were variables 4 and 4/8. Conse-quently, for these two variables, results arereported separately for these two species.Endocasts of Paranthropus and Australo-pithecus were compared to each other andto endocasts from Pan, Gorilla, and Homo(Table 3) by computing mean differences,standard errors, and P-values from the dataprovided in Table 1. Finally, the above

observations and statistically significantresults were synthesized and the key featuressummarized for endocasts from apes, earlyhominids, and humans (Table 4).

Additionally, because previously un-known parts of Paranthropus endocasts arenow available, new endocast reconstructionswere made for Paranthropus specimensSK 1585 (P. robustus), OH 5 (P. boisei),KNM-ER 732 (P. boisei), and KNM-ER407 (P. boisei), using appropriate unrecon-structed parts of Paranthropus endocasts asmodels. Endocast reconstruction methodsand cranial capacity determinations aredetailed in the Appendix.

Results

Gorilla, Pan, and HomoMean measurements from basal views ofendocasts are presented in Table 1, andmeans of indices generated by dividing vari-ables 1–7 and [3–6], [6–7] and [8–3] byendocast lengths are provided in Table 2.Not surprisingly, the means for larger-brained Homo are significantly greater thanthe means for smaller-brained Gorilla andPan for variables 1–8, [6–7], and [8–3].All P-values are <0·001 except for theHomo-Gorilla comparison for variable 3(P=0·021). Variable [3–6], on the otherhand, is significantly shorter in Homo than inGorilla or Pan (P<0·001), which corre-sponds with an increased mean length ofvariable 4 in Homo (indicating a greaterextent of temporal pole projection, seebelow). Gorilla is significantly longer thanPan for variables 8 (endocast length) and[8–3] (P<0·001 for both comparisons). Forvariable 4, on the other hand, P. troglodytes issignificantly longer than Gorilla (P<0·001)and P. paniscus (P<0·01), which do notdiffer significantly from each other(P=0·82).

The mean indices (Table 2) that expressvariables as percentages of endocast lengths

699

Tab

le1

En

doc

ast

mea

sure

men

ts

bat–

bat

1m

at–m

at2

mba

t–ro

f(ta

n)3

mcp

–m

bat

4

mcp

–ro

f(ta

n)5

cob–

rof(

tan)

6

rob–

rof(

tan)

7

rof(

tan)

bpc(

tan)

8

Cra

nial

capa

city

(cm

3)

9

Len

gth

mba

t–co

b[3

–6]

Len

gth

olf.

bulb

[6–7

]

Bas

e–ca

udal

toba

t[8

–3]

Hom

o(n

=10

)M

ean

69·7

010

8·90

38·9

016

·00

53·8

930

·30

9·00

150·

1013

508·

6021

·30

111·

20S

.D.

6·27

7·42

3·76

2·31

5·88

4·74

2·11

6·71

175·

503·

924·

526·

61R

ange

(58·

0–80

·0)

(98·

0–12

1·0)

(34·

0–43

·0)

(12·

0–20

·0)

(46·

0–64

·0)

(22·

0–37

·0)

(6·0

–12·

0)(1

40·0

–164

·0)

—(1

·0–1

3·0)

(14·

0–27

·0)

(101

·0–1

22·0

)G

orill

a(n

=10

)M

ean

49·0

077

·60

34·0

05·

7040

·90

16·6

04·

6012

0·90

483·

5017

·40

12·0

086

·90

S.D

.2·

494·

622·

911·

773·

251·

901·

515·

9079

·33

2·27

0·67

5·26

Ran

ge(4

5·0–

52·0

)(7

1·0–

86·0

)(3

1·0–

39·0

)(3

·0–8

·0)

(37·

0–46

·0)

(13·

0–19

·0)

(2·0

–7·0

)(1

14·0

–129

·0)

(375

·0–5

85·0

)(1

3·0–

20·0

)(1

1·0–

13·0

)(8

0·0–

96·0

)P

an(n

=18

)M

ean

47·3

074

·94

31·4

18·

5640

·29

16·4

44·

7910

7·56

392·

6714

·76

11·2

676

·33

S.D

.3·

665·

223·

172·

233·

992·

931·

353·

8825

·00

2·70

3·54

4·68

Ran

ge(4

0·0–

53·0

)(6

4·0–

83·0

)(2

6·0–

36·0

)(5

·0–1

2·0)

(32·

0–45

·0)

(10·

0–21

·0)

(2·0

–8·0

)(9

7·0–

114·

0)(3

60·0

–445

·0)

(10·

0–22

·0)

(1·0

–17·

0)(6

6·0–

84·0

)P

aran

thro

pus

WT

1700

050

·00

74·0

034

·00

5·00

39·0

020

·00

7·00

114·

0041

0·00

14·0

013

·00

80·0

0W

T17

400

45·0

068

·00

29·0

04·

0033

·00

14·0

07·

00—

390·

0015

·00

7·00

—M

ean

47·5

071

·00

31·5

04·

5036

·00

17·0

07·

00—

400·

0014

·50

10·0

080

·00

S.D

.3·

544·

243·

540·

714·

244·

240·

00—

14·1

40·

714·

24—

Ran

ge(4

5·0–

50·0

)(6

8·0–

74·0

)(2

9·0–

34·0

)(4

·0–5

·0)

(33·

0–39

·0)

(14·

0–20

·0)

(7·0

–7·0

)—

(390

·0–4

10·0

)(1

4·0–

15·0

)(7

·0–1

3·0)

—A

ustr

alop

ithec

usS

ts5

56·0

085

·00

39·0

014

·00

53·0

030

·00

14·0

011

8·00

485·

009·

0016

·00

79·0

0S

tw50

5—

—33

·00

11·0

044

·00

29·0

011

·00

—51

5·00

4·00

18·0

0—

Sts

6062

·00

—31

·00

11·0

042

·00

——

—42

8·00

——

—M

ean

59·0

0—

34·3

312

·00

46·3

329

·50

12·5

0—

476·

006·

5017

·00

—S

.D.

4·24

—4·

161·

735·

860·

712·

12—

44·1

92·

501·

00—

Ran

ge(5

6·0–

62·0

)—

(31·

0–39

·0)

(11·

0–14

·0)

(42·

0–53

·0)

(29·

0–30

·0)

(11·

0–14

·0)

—(4

28·0

–515

·0)

(4·0

–9·0

)(1

6·0–

18·0

)—

Not

e:D

escr

ipti

vest

atis

tics

for

mea

sure

men

ts(s

eeF

ig.1

)ta

ken

from

endo

cast

sof

aust

ralo

pith

ecin

es,a

pes,

and

hum

ans.

Cra

nial

capa

citi

esfo

rgo

rilla

san

dP

anw

ere

obta

ined

from

skul

ls,w

hile

thos

efo

rhu

man

sar

efr

omP

akke

nber

g&

Gun

ders

en,1

997.

Alt

houg

hth

em

ean

cran

ialc

apac

ity

liste

dfo

rP

anis

for

P.t

rogl

odyt

eson

ly,

am

ean

cran

ial

capa

city

of35

0cm

3ha

sbe

enpu

blis

hed

else

whe

refo

rP

.pa

nisc

us(C

ram

er,

1977

).F

orS

tw50

5,m

bat

was

dete

rmin

edby

the

inte

rsec

tion

ofth

eta

ngen

tto

the

left

bat

wit

hth

ein

tact

mid

sagi

ttal

plan

e.

700 . ET AL.

Tab

le2

Des

crip

tive

stat

isti

csan

dP

-val

ues

for

end

ocas

tin

dic

es

1/8

2/8

3/8

4/8

5/8

6/8

7/8

[3–6

]/8

[6–7

]/8

[8–3

]/8

Hom

o(n

=10

)M

ean

0·46

*0·

73*

0·26

†0·

11*†

0·36

0·20

*†0·

06*†

0·06

*†0·

14*†

0·74

†S

.D.

0·04

0·03

0·02

0·02

0·04

0·03

0·01

0·03

0·03

0·02

Ran

ge(0

·38–

0·50

)(0

·66–

0·76

)(0

·22–

0·30

)(0

·08–

0·13

)(0

·31–

0·42

)(0

·16–

0·25

)(0

·04–

0·08

)(0

·01–

0·09

)(0

·10–

0·18

)(0

·70–

0·78

)G

orill

a(n

=10

)M

ean

0·41

*‡0·

64*‡

0·28

0·05

*‡0·

34‡

0·14

*0·

04*

0·14

*0·

10*

0·72

S.D

.0·

020·

030·

020·

010·

020·

020·

010·

020·

008

0·02

Ran

ge(0

·37–

0·44

)(0

·57–

0·69

)(0

·25–

0·31

)(0

·03–

0·06

)(0

·30–

0·60

)(0

·10–

0·16

)(0

·02–

0·06

)(0

·11–

0·17

)(0

·09–

0·11

)(0

·69–

0·75

)P

an(n

=18

)M

ean

0·44

‡0·

69‡

0·29

†0·

08†‡

0·37

‡0·

15†

0·04

†0·

14†

0·10

†0·

71†

S.D

.0·

030·

040·

030·

020·

040·

030·

010·

020·

030·

03R

ange

(0·3

8–0·

49)

(0·6

2–0·

77)

(0·2

5–0·

34)

(0·0

5–0·

11)

(0·3

0–0·

42)

(0·0

9–0·

20)

(0·0

2–0·

08)

(0·0

1–0·

20)

(0·0

1–0·

16)

(0·6

6–0·

75)

*P<

0·00

1*P

<0·

001

†P=

0·02

2*P

<0·

001

‡P=

0·04

*P<

0·00

1*P

=0·

001

*P<

0·00

1*P

=0·

004

†P=

0·18

‡P=

0·02

6‡P

=0·

003

†P=

0·00

1†P

<0·

001

†P=

0·00

8†P

<0·

001

†P=

0·00

5‡P

<0·

001

Par

anth

ropu

sK

NM

-WT

1700

00·

440·

650·

300·

040·

340·

180·

060·

120·

110·

70A

ustr

alop

ithec

usS

ts5

0·47

0·72

0·33

0·12

0·45

0·25

0·12

0·08

0·14

0·67

Not

es:

Sym

bols

repr

esen

ta

sign

ifica

ntdi

ffer

ence

betw

een

grou

ps:

*Hom

o–G

orill

a;†H

omo–

Pan

;‡G

orill

a–P

an.

Indi

ces

for

endo

cast

sin

basa

lvie

ww

ere

obta

ined

bydi

vidi

ngth

ele

ngth

sof

the

vari

able

spr

ovid

edin

Tab

le1

bym

easu

rem

ent

8.S

igni

fican

cete

sts

wer

eca

lcul

ated

wit

hpo

st-h

ocm

ulti

ple

com

pari

sons

corr

ecte

dw

ith

Bon

ferr

oni’s

met

hod

wit

hin

the

gene

ral

linea

rm

odel

(typ

eII

I)of

SP

SS

.In

dice

sw

ere

avai

labl

efo

ron

lyon

eP

aran

thro

pus

(KN

M-W

T17

000)

and

one

Aus

tral

opith

ecus

(Sts

5)en

doca

st.

The

sear

epr

ovid

edfo

rco

mpa

rati

vepu

rpos

es,

alth

ough

they

wer

eno

tco

mpa

red

stat

isti

cally

wit

hth

ose

ofap

esan

dhu

man

s.

701

in basal view indicate that Gorilla differssignificantly from Homo and Pan in havinggenerally narrower endocasts at the level ofthe temporal poles (variable 2/8, P<0·001and P=0·003 respectively), temporal polesthat are relatively closer together (variable1/8, P<0·001 and P=0·026), and temporalpoles that do not project as far forwardrelative to sella (variable 4/8, P<0·001 forboth comparisons). The mean relativelengths of the orbital surfaces of the frontallobes of Gorilla (variable 5/8) are signifi-cantly shorter than those of Pan (P=0·04),while the mean relative lengths of the twomost anterior regions of the frontal lobes(variables 6/8 and 7/8, P�0·001 for bothcomparisons) and olfactory bulbs (variable[6–7]/8, P=0·004) are significantly shorterthan those of Homo.

Endocasts of Pan, on the other hand, aresimilar to those of Homo and differ signifi-cantly from those of Gorilla in mean relativewidth at the level of the temporal poles(variable 2/8), mean relative distancebetween the temporal poles (variable 1/8),and mean relative length of the frontal lobes(variable 5/8) (P=0·003, 0·026, 0·04respectively). The mean relative length ofthe portion of the frontal lobes that isanterior to the temporal poles (variable 3/8),however, is significantly longer in Pan thanHomo (P=0·022), but not Gorilla. Corre-sponding to this, the mean relative length ofthe posterior portion of the endocast (vari-able [8–3]/8) is significantly shorter in Panthan in Homo (P=0·018). The mean relativeprojection of the temporal poles (variable4/8) does not differ from that for Homo orP. troglodytes (P=0·63), but is significantlyshorter in P. paniscus than for both Homo(P<0·001) and P. troglodytes (P<0·03). Thisvariable is significantly greater in P. paniscus(P=0·03) and P. troglodytes (P<0·001) thanit is for Gorilla. As is the case for Gorilla, themean relative lengths of the two mostanterior regions of the frontal lobes (vari-ables 6/8 and 7/8) and olfactory bulbs (vari-

able [6–7]/8) of Pan are significantly shorterthan those of Homo (P<0·001, 0·008 and0·005 respectively).

Although the mean relative length of theentire frontal lobe in basal view (variable5/8) does not differ significantly betweenHomo and either of the two apes, the meanrelative lengths of certain subregions withinthe frontal lobe (variables 6/8, 7/8, and[6–7]/8) are significantly longer in Homothan they are in Gorilla and Pan (see Table 2for P-values). The mean relative length ofvariable 4/8 is also significantly longer inHomo than in Gorilla (P<0·001) and P.paniscus (P<0·001), but not P. troglodytes.On the other hand, the relative lengthbetween the anterior end of the temporalpoles and the posterior end of the olfactorybulbs (variable [3–6]/8) is significantlyshorter in Homo than in either ape (P<0·001for both comparisons). As detailed in thediscussion section, these findings sup-port other comparative studies on actualbrains of apes and humans, which showthat the frontal lobes of Homo are re-organized compared to those of Gorilla andPan.

To summarize the main findings regard-ing the relative proportion of endocasts fromliving hominoids: endocasts of gorillas aregenerally longer and narrower than those ofPan (variables 8, [8–3], 1/8, and 2/8) andnarrower than those of Homo (variables 1/8and 2/8), while endocasts from P. troglodytes(but not P. paniscus or Gorilla) furtherresemble those of humans in having rela-tively projecting temporal poles (variable4/8). Although the overall length of thehuman frontal lobe (variable 5/8) does notdiffer significantly from those of Pan orGorilla, the proportions of areas withinhuman frontal lobes are dramatically differ-ent from those of apes (variables 6/8, 7/8,[6–7]/8, [3–6]/8 and, except for Pan troglo-dytes, 4/8). In particular, the most anteriorregions (variables 6 and 7) of the frontallobe are relatively longer in humans.

702 . ET AL.

Tab

le3

Mea

nd

iffer

ence

,st

and

ard

erro

rsan

dP

-val

ues

for

com

par

ison

sw

ith

Par

anth

rop

us

and

Au

stra

lop

ith

ecu

s

Com

pari

son

grou

psM

ean

diff

eren

ceS

tand

ard

erro

rP

-val

ues

Com

pari

son

grou

psM

ean

diff

eren

ceS

tand

ard

erro

rP

-val

ues

Gro

up1

Gro

up2

Gro

up1

Gro

up2

Var

iab

le1:

bat

–bat

Var

iab

le1:

bat

–bat

Par

anth

ropu

sP

an0·

203·

131·

000

Aus

tral

opith

ecus

Pan

11·3

72·

620·

001*

Gor

illa

�1·

503·

251·

000

Gor

illa

9·67

2·76

0·01

2*A

ustr

alop

ithec

us�

11·1

73·

830·

059

Hom

o�

11·0

32·

760·

003*

Hom

o�

22·2

03·

25<

0·00

1*V

aria

ble

2:m

at–m

atV

aria

ble

2:m

at–m

atP

aran

thro

pus

Pan

�3·

944·

251·

000

Aus

tral

opith

ecus

Pan

——

—G

orill

a�

6·60

4·41

0·86

1G

orill

a—

——

Hom

o�

37·9

04·

41<

0·00

1*H

omo

——

—V

aria

ble

3:m

bat

–rof

(tan

)V

aria

ble

3:m

bat

–rof

(tan

)P

aran

thro

pus

Pan

0·09

2·48

1·00

0A

ustr

alop

ithec

usP

an2·

922·

081·

000

Gor

illa

�2·

502·

581·

000

Gor

illa

0·33

2·19

1·00

0A

ustr

alop

ithec

us�

2·83

3·04

1·00

0H

omo

�4·

572·

190·

438

Hom

o�

7·40

2·58

0·06

6V

aria

ble

4:m

cp–m

bat

Var

iab

le4:

mcp

–mb

atP

aran

thro

pus

Pan

�4·

061·

560·

134

Aus

tral

opith

ecus

Pan

3·44

1·31

0·12

1G

orill

a�

1·20

1·62

1·00

0G

orill

a6·

301·

380·

001*

Aus

tral

opith

ecus

�7·

501·

910·

004*

Hom

o�

4·00

1·38

0·06

2H

omo

�11

·50

1·62

<0·

001*

Var

iab

le5:

mcp

–rof

(tan

)V

aria

ble

5:m

cp–r

of(t

an)

Par

anth

ropu

sP

an�

4·29

3·34

1·00

0A

ustr

alop

ithec

usP

an6·

052·

800·

368

Gor

illa

�4·

903·

471·

000

Gor

illa

5·43

2·95

0·73

3A

ustr

alop

ithec

us�

10·3

34·

090·

158

Hom

o�

7·56

2·95

0·14

5H

omo

�17

·89

3·47

<0·

001*

703

Tab

le3

(Con

tinue

d)

Com

pari

son

grou

psM

ean

diff

eren

ceS

tand

ard

erro

rP

-val

ues

Com

pari

son

grou

psM

ean

diff

eren

ceS

tand

ard

erro

rP

-val

ues

Gro

up1

Gro

up2

Gro

up1

Gro

up2

Var

iab

le6:

cob

–rof

(tan

)V

aria

ble

6:co

b–r

of(t

an)

Par

anth

ropu

sP

an0·

562·

421·

000

Aus

tral

opith

ecus

Pan

13·0

62·

02<

0·00

1*G

orill

a0·

402·

511·

000

Gor

illa

12·9

02·

13<

0·00

1*A

ustr

alop

ithec

us�

12·5

02·

960·

001*

Hom

o�

0·80

2·13

1·00

0H

omo

�13

·30

2·51

<0·

001*

Var

iab

le7:

rob

–rof

(tan

)V

aria

ble

7:ro

b–r

of(t

an)

Par

anth

ropu

sP

an2·

211·

180·

696

Aus

tral

opith

ecus

Pan

7·71

0·99

<0·

001*

Gor

illa

2·40

1·23

0·58

3G

orill

a7·

901·

05<

0·00

1*A

ustr

alop

ithec

us�

5·50

1·45

0·00

5*H

omo

3·50

1·05

0·01

8*H

omo

�2·

001·

231·

000

Var

iab

le[3

–6]:

[mb

at–c

ob]

Var

iab

le[3

–6]:

[mb

at–c

ob]

Par

anth

ropu

sP

an�

0·26

2·17

1·00

0A

ustr

alop

ithec

usP

an�

8·26

1·81

0·00

1*G

orill

a�

2·90

2·25

1·00

0G

orill

a�

10·9

01·

91<

0·00

1*A

ustr

alop

ithec

us8·

002·

660·

046*

Hom

o�

2·10

1·91

1·00

0H

omo

5·90

2·25

0·12

6V

aria

ble

[6–7

]:[c

ob–r

of(t

an)–

rob

–rof

(tan

)]V

aria

ble

[6–7

]:[c

ob–r

of(t

an)–

rob

–rof

(tan

)]P

aran

thro

pus

Pan

�1·

262·

451·

000

Aus

tral

opith

ecus

Pan

5·74

2·05

0·08

0G

orill

a�

2·00

2·55

1·00

0G

orill

a5·

002·

160·

264

Aus

tral

opith

ecus

�7·

003·

000·

251

Hom

o�

4·30

2·16

0·54

1H

omo

�11

·30

2·55

0·00

1*

End

ocas

tsfo

rva

riab

les

liste

din

Tab

le1.

(Pan

incl

udes

both

P.t

rogl

odyt

esan

dP

.pan

iscu

s.)

Com

pari

sons

wer

edo

neon

lyfo

rgr

oups

that

cont

aine

dtw

oor

mor

esp

ecim

ens

for

each

vari

able

,an

dsi

gnifi

canc

ete

sts

wer

eca

lcul

ated

wit

hpo

st-h

ocm

ulti

ple

com

pari

sons

corr

ecte

dw

ith

Bon

ferr

oni’s

met

hod

wit

hin

the

gene

ral

linea

rm

odel

(typ

eII

I)of

SP

SS

(*in

dica

tes

sign

ifica

ntdi

ffer

ence

s).

704 . ET AL.

Paranthropus and AustralopithecusTable 3 presents mean differences, standarderrors and P-values for comparisons ofParanthropus and Australopithecus with Pan,Gorilla, Homo, and each other for the vari-ables in Table 1. As shown by the P-values,Paranthropus does not differ significantlyfrom either Pan or Gorilla for any variablelisted in Table 3. On the other hand, Paran-thropus endocasts are significantly smallerthan those of Homo for the means of vari-ables 1, 2, 4, 5, 6, and [6–7] (P�0·001 forall six comparisons), which is not surprisinggiven the much larger cranial capacity of thelatter (Table 1). Paranthropus endocasts arealso smaller than those of Homo for variables3 and 7, although these comparisons do notreach statistical significance (P=0·066 and1·0 respectively). Finally, Paranthropus, likeboth apes, is greater than Homo for variable[3–6], but not significantly so (P=0·126).These statistics reveal that endocasts ofParanthropus are entirely ape-like in theabsolute variables that reflect the gross mor-phology of the frontal and temporal lobes of

the brain. [It should be noted, however (seebelow), that KNM-WT 17000 differs fromapes for indices 6/8 and 7/8.]

In lateral view, the ape like variables of thefrontal lobes of Paranthropus are manifestedin orbital surfaces that have a beaked-shapedprofile similar to that of chimpanzees andgorillas (Figure 2), and unlike the moreflattened orbital rostrum of humans. Vieweddorsally (Figure 3), the rostral portions ofthe frontal lobes in Paranthropus specimensKNM-WT 17000 and KNM-WT 17400 arerelatively pointed (Holloway, 1988b), beingcomparable to the unreconstructed portionsof OH 5 (P. boisei) and KNM-ER 23000(P. boisei). These specimens show that theape like variables for the frontal lobes thatare reproduced from endocasts in ourParanthropus sample are manifested in anoverall teardrop shape when viewed dorsally[Figure 3(b)]. Compared to endocastsfrom Homo and Australopithecus (seebelow), the ape like variables for the tem-poral lobes of Paranthropus are manifested inrounded temporal poles [KNM-WT 17000

Figure 2. Basal (left) and lateral (right) views of endocasts from A. africanus specimens Sts 5 and Stw 505,and Paranthropus specimens KNM-WT 17000 and KNM-WT 17400. The lateral views are positionedfrontal-lobe-to-frontal-lobe, and include a gorilla and chimpanzee for comparative purposes. Note therelatively expanded orbital surfaces of the frontal lobes of Sts 5 and Stw 505.

705

(P. aethiopicus), KNM-WT 17400 (P.boisei), SK 1585 (P. robustus)], shorter for-ward projections of the poles beyond theanterior borders of sella turcica (variable 4,Table 3; Figure 2), and shorter distancesbetween the temporal poles when seenin basal view (measurement 1, Table 3;Figure 2).

The picture for endocasts of Australo-pithecus is quite different. Despite the factthat mean cranial capacity of the Australo-pithecus specimens listed in Table 1(476 cm3) is between that of Pan (393 cm3)and Gorilla (484 cm3), a number of meanvariables for Australopithecus are significantlylarger than they are for either Pan or Gorilla(Table 3). These significant differencesinclude variables 1 (P=0·001 and 0·012respectively), 6 (P<0·001 for both compari-sons), and 7 (P<0·001 for both compari-sons). For variable 4, Australopithecus issignificantly larger than Gorilla and P.paniscus (P=0·001 and 0·003 respectively),but not P. troglodytes (P=1·0). On the other

hand, Australopithecus, like Homo, is signifi-cantly smaller than both apes for variable[3–6] (P�0·001 for all four comparisons).In contrast to Paranthropus, and despite itssmall cranial capacity compared to Homo,endocasts of Australopithecus do not differsignificantly from those of Homo for vari-ables 4, 5, 6, and [6–7] (Table 3). Further-more, variables 4, 6, and 7 are significantlylarger in Australopithecus than in Paranthro-pus endocasts (P=0·004, 0·001, and 0·005respectively), while variable [3–6] issignificantly smaller (P<0·05).

The similarities between endocasts ofAustralopithecus and Homo are manifested inexpanded and blunted, rather than beak-shaped, orbital rostra compared to apes andParanthropus (variable 6, Tables 1 and 3;Figure 2), as well as cribriform plates (olfac-tory bulbs) that are longer on average (vari-able [6–7]). In Australopithecus, expansionof the frontal lobes in the region directlylateral to rof also produces a wider rostralend of the frontal lobe when viewed dorsally

(a) (b)

Sts 5KNM-ER 23000

Sterk. No. 2 KNM-WT 17000

Sts 60

KNM-WT 17400

Stw 505OH 5

Figure 3. Outlines of dorsal views of endocasts from (a) A. africanus (Sts 5, Stw 505, Sts 60, No. 2specimen from Sterkfontein), and (b) Paranthropus (OH 5, KNM-WT 17000, KNM-ER 23000,KNM-WT 17400). Endocasts of Paranthropus appear more pointed, while those of A. africanus are widerat the rostral ends of the frontal lobes.

706 . ET AL.

compared to that seen in Paranthropus(Figure 3). Compared to Paranthropus,endocasts of Australopithecus, like those fromHomo, also have temporal poles that projectmore in an anterior and lateral directionrelative to sella turcica (variables 4 and 1,Tables 1 and 3; Figure 2).

Because endocast indices can be obtainedfor only one Paranthropus (KNM-WT17000) and one Australopithecus endocast(Sts 5) (Table 2), comparisons of theseindices for the two genera cannot be statisti-cally analyzed as was done for the variablespresented in Table 1. One may note, how-ever, that the Paranthropus endocast falls onor nearer the ape means while the Australo-pithecus endocast falls on or nearer the Homomeans for relative variables 1/8, 2/8, 4/8,[3–6]/8, and [6–7]/8 (Table 2). Interest-ingly, the Australopithecus endocast is notice-ably larger than those for Paranthropus,Homo, and both apes for relative vari-ables 3/8, 5/8, 6/8, and 7/8; and notice-ably smaller for relative variable [8–3]/8(Table 2).

Summary of key features on endocasts fromGorilla, Pan, Paranthropus,Australopithecus and HomoThe key findings regarding absolute andrelative variables described above for endo-casts of Gorilla, Pan, Paranthropus, Australo-pithecus, and Homo are summarized in Table4. Except for the indices for Paranthropusand Australopithecus (and unless otherwisestated in the legend), table entries for Homoindicate that its mean differs significantlyfrom the means for apes, while entries forapes indicate that their means differ signifi-cantly from those for the apes that are notmarked. This table is conservative. Forexample, although the mean [6–7] for Aus-tralopithecus is intermediate between that ofapes and Homo (Table 1), this variable is notmarked because the differences between themean for Australopithecus and those forapes did not achieve statistical significance

(Table 3). Conversely, the indices forParanthropus and Australopithecus, could notbe compared statistically with means fromother groups because they are availablefor only one specimen each (i.e., KNM-WT 17000 and Sts 5). In these cases,entries indicate whether the index measuredfrom one representative is closer to thatfor apes (G and/or P) or humans (H)(Table 2).

The observations in Table 4 are organizedinto four complexes, each of which containsinterrelated (dependent) features. Withregards to general size and shape of endo-casts, the long (variable 8) and narrow (1/8and 2/8) endocasts of Gorilla differ markedlycompared to those of Pan. Endocasts ofHomo, however, are associated with largercranial capacities than those of apes, as wellas longer relative lengths of the posteriorportion of the brain [8–3]/8 compared toPan. Two indices for Australopithecus (5/8,[8–3]/8) differ noticeably from those of apesbecause variables 3 and 5 of Sts 5 areincreased greatly compared to apes while itsoverall length (variable 8) is not. Theseindices are smaller in Homo than Australo-pithecus, on the other hand, because Homoequals Sts 5 in the mean dimensions ofvariables 3 and 5, but has a much longermean overall length.

The second set of variables in Table 4pertain to subdivisions on the orbital sur-faces of the frontal lobes and reveal thatAustralopithecus and Homo differ similarly ina number of key features compared to thoseof apes and Paranthropus. Again, the notice-ably larger indices for the frontal lobes ofAustralopithecus compared to Homo (Table2: 3/8, 6/8, and 7/8) are largely the result ofa shorter variable 8 for Sts 5 than for Homo.These observations are consistent with theinterpretation that, compared to endocastsfrom apes and Paranthropus, endocastsof Australopithecus are characterized bydifferentially lengthened frontal lobes andsubdivisions thereof, concomitantly with

707

relatively shortened posterior portions ([8–3]/8). This accounts for the comparativelyexpanded and squared-off appearance of theorbital rostra in Australopithecus endocasts(Figures 2 and 3).

Because olfactory regions represent a phy-logenetically older part of the brain thanneocortical areas (Finlay & Darlington,1995), the two measurements pertaining to

the olfactory bulbs are placed in a thirdcomplex (Table 4). As detailed in the dis-cussion section, Australopithecus appearsmore like humans than apes in the size andshape of its olfactory bulbs. This is also truefor the temporal poles described by thefourth complex of variables (Table 4),as confirmed by visual observation ofendocasts from Australopithecus and Homo

Table 4 Summary of key endocast features for Gorilla, Pan, Paranthropus, Australopithecus andHomo

Description Variable Gorilla Pan Paranthropus Australopithecus Homo

General size and shapeCranial capacity cm3 +Absolute length endocast 8 + +Relative width bat–bat 1/8 � P H +G

Relative width mat–mat 2/8 � G H +G

Relative length mcp–rof 5/8 + G >G, P, HAbsolute length mcp–rof 5 +Relative length bpc–mbat [8–3]/8 P <G, P, H +Pan

Absolute length bpc–mbat [8–3] + +Frontal lobes

Relative length mbat–rof 3/8 P >G, P, H �Pan

Absolute length mbat–rof 3 +Relative length mbat–cob [3–6]/8 G, P H �Absolute length mbat–cob [3–6] � �Relative length cob–rof 6/8 H >G, P, H +Absolute length cob–rof 6 + +Relative length rob–rof 7/8 H >G, P, H +Absolute length rob–rof 7 + +

Olfactory bulbsRelative length cob–rob [6–7]/8 G, P H +Absolute length cob–rob [6–7] +

Temporal polesRelative length mcp–mbat 4/8 +(P. troglodytes) G H +G, Pp

Absolute length mcp–mbat 4 +(P. troglodytes) �Pt +G, Pp +Absolute width bat–bat 1 + +

Based on statistical analyses of measurements obtained from basal views and illustrated in Figure 1. Symbols: +under Homo or Australopithecus, the mean for that variable is significantly larger than the means for apes; + underGorilla or Pan, the mean is significantly larger than those for the unmarked apes; + with superscripts (e.g., +G, Pp),the mean is significantly larger than the means for only the apes indicated by the superscript (superscripts: G,Gorilla; Pan, chimpanzees and bonobos; Pp, Pan paniscus; Pt, Pan troglodytes); + (P. troglodytes), the mean variableis significantly longer in P. troglodytes than either P. paniscus or Gorilla; � with and without superscripts, the sameconventions as above, except that the means are significantly smaller than those for apes. Because indices could notbe compared statistically with means from other groups for Paranthropus (KNM-WT 17000) and Australopithecus(Sts 5), G and/or P, or H indicate that a particular index is closer to the mean for Gorilla and/or Pan or Homo (Table2); >G, P, H indicates that the index for Sts 5 is noticeably greater than the indices for Gorilla, Pan, and Homo,while <G, P, H means that the index for Sts 5 is noticeably smaller. Note that Paranthropus is similar to Homo foronly two indices, and that none of its mean absolute variables differ significantly from those of apes in the samedirection (+ or �) as Homo. Australopithecus, on the other hand, is similar to Homo for five indices, and the meansof four of its absolute variables differ significantly from those of all apes in the same direction as Homo. These datahave implications for understanding the sequence in which cortical reorganization occurred during hominid brainevolution.

708 . ET AL.

that share a forward (4/8, 4) and lateral (1)projection of the temporal poles (Figure 2).

In sum, Table 4 reveals that endocasts ofParanthropus are similar to those of Homo foronly two indices, while none of its meanabsolute variables differ significantly fromthose of apes in the same direction (+ or �)as the means from endocasts of Homo. Incontrast, endocasts of Australopithecus aresimilar to Homo endocasts for five indices,and the means of four of its absolute vari-ables differ significantly from those of allapes in the same direction as the means fromHomo. The implications of these data forunderstanding the sequence in whichcortical reorganization occurred duringhominid brain evolution are explored in thediscussion section.

New endocranial capacities

Because much of the above information wasnot available when endocasts of Paranthro-pus specimens SK 1585, OH 5, KNM-ER407, and KNM-ER 732 were reconstructed(apparently with the endocast of Sts 5 asa frequent model), we reconstructed theendocasts of these specimens (Figure 4),

using the unreconstructed parts of otherParanthropus endocasts as models and re-calculated their endocranial capacities (seeAppendix for detailed descriptions of eachreconstruction) (Table 5).

Five water displacements of the newlyreconstructed endocast of SK 1585 (P.robustus) resulted in a mean of 476 cm3

(470–484 cm3), 54 cm3 less than thecurrently accepted estimate of 530 cm3

(Holloway, 1972) (Figure 4).Five cranial capacity estimates of the

newly reconstructed endocast of OH 5 (P.boisei) resulted in a mean of 500 cm3 (498–502 cm3), 30 cm3 less than the currently ac-cepted estimate of 530 cm3 (Tobias, 1967)(Figure 4). This loss is due mostly to reduc-tion in the orbital olfactory region comparedto the earlier reconstruction, which did notbenefit from reference to Paranthropus speci-mens that were discovered subsequent to itsreconstruction. Our reconstruction differsfrom the earlier one in having a smaller,beaked-shaped rostral orbital region, andsomewhat less anteriorly extended temporalpoles (Holloway, 1972, 1975).

Five cranial capacity estimates for thenewly reconstructed endocast of KNM-ER

Figure 4. Newly reconstructed endocasts from four Paranthropus specimens. The reconstructed regions ofSK 1585 are dark; those of KNM-ER 407, KNM-ER 732, and OH 5 are white. This endocast of SK 1585contains matrix between the inferior border of the temporal lobe and the cerebellum that was removed ina subsequent procedure described in the Appendix. These reconstructions reproduce the beak-shapedrostral orbital area that is found in Paranthropus, but not Australopithecus.

709

732 (P. boisei) result in a mean of 466 cm3

(460–472 cm3), 34 cm3 less than thecurrently accepted estimate of 500 cm3

(Holloway, 1988a) (Figure 4). Our recon-struction differs from the earlier one in hav-ing a smaller, beaked-shaped rostral orbitalregion.

Five cranial capacity estimates for thenewly reconstructed endocast of KNM-ER407 (P. boisei) result in a mean of 438 cm3

(430–446 cm3), 72 cm3 less than thecurrently accepted estimate of 510 cm3

(Holloway, 1988a) and 68 cm3 less than anearlier estimate of 506 cm3 (Falk & Kasinga,

Table 5 Cranial capacities for Paranthropus and Australopithecus specimens. Those for A. africanusare from the literature

SpeciesDating(Ma) Specimen

Cranialcapacity(cm3) Reference

Thisstudy Method Eval.

P. robustus �1·8–1·7 SK 1585 530 1 476 B 1P. aethiopicus �2·5 KNM-WT 17000 410 2 (410)P. boisei �2·4 Omo L338y-6 427 3 (427)

�1·88 KNM-ER 13750 450–480or 500

45

—

�1·9 KNM-ER 23000 491 5 (491)�1·8 KNM-WT 17400 390–400

or 50045

(390)

�1·8 OH 5 530 6 500 A 1�1·7 KNM-ER 406 525 7 —�1·85 KNM-ER 407 510 7 438 B 1–2�1·7 KNM-ER 732 500 7 466 B 1�2·2 Omo 323 490 5 —

Mean 479·4or 492·1

449·8

A. africanus �3·0 MLD 37/38 425 8�3·0–2·5 Sts 60 428 7

Sts 71 428 9Sts 5 485 7Sts 19 436 7

�2·8–2·6 Stw 505 515 10�2·5–1·0? Taung 440 7

Mean 451

Revised cranial capacities for Paranthropus are in bold; those that are accepted from the literature are inparentheses. The first mean for Paranthropus includes estimates for KNM-ER 13750 and KNM-WT 17400 fromHolloway (1988b); the second mean uses estimates for these two specimens from Brown et al. (1993). Ouracceptance of the estimate for Omo L338y-6 is tentative pending an opportunity to do our own reconstruction.KNM-ER 13750 is excluded from the present study because of the disparity in estimates between Holloway(1988b) and Brown et al. (1993) and the fact that we do not have a copy of this specimen from which to make ourown judgment. We accept the lower estimate for KNM-WT 17400 from Holloway (1988b) after comparing thisspecimen with a large number of ape and australopithecine endocasts in our collection. KNM-ER 406 is excludedbecause its capacity is based on external skull measurements and calculated from a formula that incorporates afactor (f) that is based on erroneous cranial capacity estimates for OH 5 and SK 1585 (Holloway, 1973). Omo 323is excluded because it is too fragmentary to yield an accurate estimate. References: 1, Holloway (1972); 2, Walkeret al. (1986); 3, Holloway (1981); 4, Holloway (1988b); 5, Brown et al. (1993); 6, Tobias (1967); 7, Holloway(1988a); 8, Conroy et al. (1990); 9, Conroy et al. (2000); 10, Conroy et al. (1998). Methods: A, waterdisplacement of a full or hemi-endocast (times two) reconstructed in silicone with minimal distortion; B, volumeof water contained by mold of hemi-endocast times two. Evaluations of confidence in cranial capacity estimatesdue to completeness of original specimens: 1, highest confidence; 2, high confidence. See Appendix for details ofthe new endocast reconstructions.

710 . ET AL.

1983) (Figure 4). Our reconstruction differsfrom earlier ones in that most of the frontallobe and temporal pole required reconstruc-tion using the appropriate Paranthropusmodels.

Table 5 compares currently acceptedendocranial capacities of Paranthropuswith our revised values (estimates of otherspecimens in parentheses are consideredacceptable). Our new estimates for thesefour Paranthropus specimens are all lowerthan earlier estimates, and the new mean of450 cm3 for eight specimens is significantly(P<0·05, two-tailed) lower than one of thecurrently accepted means of 492 cm3 (Table5). However, the new mean does not differsignificantly from that of 451 cm3 forAustralopithecus (P�0·95), which contra-dicts the commonly held view thatParanthropus and early Homo had, onaverage, significantly larger brains thanAustralopithecus (Holloway, 1973).

Discussion

The more human like cortical morphologyreproduced on Australopithecus endocasts isnot due to allometric scaling because (1) themean endocranial volume of the three Aus-tralopithecus specimens measured in thisstudy is less than that for both gorillas andhumans (Table 1), and (2) because themean endocranial volume of a wider sampleof seven Australopithecus specimens does notdiffer significantly from that of eight Paran-thropus specimens (P�0·95, Table 5). Fur-thermore, it is unlikely that the beak-shapedorbital rostra of endocasts from apes andParanthropus are due to a high degree ofpostorbital constriction in their skulls, sinceskulls of Australopithecus that are also char-acterized by a high degree of postorbitalconstriction produce endocasts with orbitalsurfaces that are expanded and wide, ratherthan pointed (beak-shaped) and narrow atthe very front (Figure 2). Thus, as othershave suggested (Dean, 1988), endocranial

aspects of the cranial base, while greatlyinfluenced by the morphology of the brain,appear to be relatively independent fromaspects of the masticatory system. It is alsoimportant to note that, although the cranialbase of the skull has been shown to beaffected by intentional deformation of thecranial vault (for cultural reasons) in nativeAmericans, the effect is indirect via thealtered cranial vault’s effects on braingrowth (Cheverud et al., 1992; Kohn et al.,1993). These studies show that the cranialbase responds directly to changes in braingrowth.

An extensive literature based largelyon comparative studies of actual brainsindicates that the enlarged brain of Homosapiens is derived compared to the brains ofextant apes (Connolly, 1950; Holloway,1988b; Falk, 1992; Semendeferi, 1994;Deacon, 1997; Tobias, 1997; Passingham,1998; Semendeferi & Damasio, 2000).Within this context, frontal lobes havetraditionally been of special interest topaleoneurologists because of their knownfunctions with respect to language, abstractthought, planning, and execution of motoractivities. For example, comparative studieson actual brains led both Deacon (1997)and Semendeferi (1994) to conclude thatprefrontal regions of the frontal lobes areenlarged and derived in humans as a resultof cortical reorganization that occurred dur-ing the evolution of their early hominidancestors. Passingham (1998) arrived at thesame conclusion regarding the inferiorfrontal cortex and temporal lobe. It is alsoimportant to note that brains need not beenlarged to be derived, i.e., that neurologicalevolution may entail cortical reorganizationor redistribution of cortical tissues withoutan increase in brain size (Holloway, 1988b).The developmental mechanisms that arelikely to have operated during the courseof brain expansion and reorganization inmammals, including humans, have recentlybeen elucidated within a framework that

711

accommodates both allometric scaling andthe evolution of neurological specializations(Finlay & Darlington, 1995).

Holloway’s (1975, 1988b) longheld beliefthat cortical reorganization may alreadyhave been underway in australopithecinesprior to the increase in brain size thatoccurred subsequently in Homo is supportedby our observations for Australopithecus, butnot Paranthropus. As detailed below, ourmorphological findings for the orbitalsurfaces of Australopithecus endocasts corre-spond with the reorganized cortical mor-phology that Semendeferi (1994) earlierhypothesized would have existed in thehominid ancestors of Homo and that wouldhave been derived relative to the moreprimitive ape like morphology. Specifically,our analysis of endocasts has shown that theorbital surfaces of the frontal lobes of Aus-tralopithecus were expanded and the relativelengths of subareas rearranged (reorganized)compared to Paranthropus, which appearsmore ape like and less human like thanAustralopithecus.

Some phylogenetic speculationsUsing Australopithecus as a hypotheticalmodel for the ancestral Homo condition, it ispossible to hypothesize about the sequencein which certain neurological featuresreorganized during the course of hominidevolution. It thus appears that the frontallobes and temporal poles may haveincreased in size early on (i.e., in the austra-lopithecine ancestors of Homo), followed bysubsequent (additional) enlargement of pos-terior regions during the course of brainevolution in Homo. In addition to anincrease in overall size of the frontal lobes(as indicated by length), the subregionswithin the orbital surfaces of the frontallobes appear to have become reorganizedwith respect to one another in a sequentialmanner. For example, although humanolfactory bulbs are estimated to be roughly1/2 to 1/3 the volume of those of P. troglo-

dytes and Gorilla gorilla (Stephan et al.,1981), inspection of endocasts shows thatthe shape of the human olfactory bulb islong and flattened compared to the shorter,more protuberant bulbs of apes. In keepingwith this, the olfactory bulb measurement[6–7] of 21 mm is longer in humans than inapes and Paranthropus (Table 1). Measure-ment 6 (the length of the olfactory bulb plusmeasurement 7) averages 30 mm in bothAustralopithecus and humans. However, themean length of the olfactory bulb in Austra-lopithecus (17 mm) is 4 mm shorter than themean for humans, while that of measure-ment 7 is 3·5 mm longer. These differ-ences would disappear if the olfactory bulbsincreased their length rostrally by 4 mm—i.e., to the human length while maintainingthe overall length of measurement 6. Thesedata are consistent with the hypothesis thatthe orbital surface of the frontal lobes wasexpanded in the region of rof in conjunctionwith some lengthening and flattening of theolfactory bulbs in Australopithecus comparedto Paranthropus, and that the olfactory bulbscontinued to lengthen in a rostral directionsubsequent to this (i.e., in descendants ofAustralopithecus that may have given rise toHomo).

Our findings have wider implications forthe evolution of cognition in early hominids.Both the blunt-shaped, relatively enlargedportions of the orbital surfaces of the frontallobes and the anteriorly expanded, laterallypointed temporal poles of Australopithecusappear more human like compared toParanthropus and African apes. The area thatis expanded near rof in the frontal lobes ofAustralopithecus corresponds to Brodmann’sarea 10 in both apes and humans, which hasbeen shown experimentally to be involved inabstract thinking, planning of future actions,and undertaking initiatives (Semendeferi,1994). Because the relative size of humanarea 10 is twice that of both bonobos andchimpanzees, Semendeferi (1994) suggestedthat this area of the cerebral cortex increased

712 . ET AL.

in relative size at some point along the linefrom the first hominids to the early repre-sentatives of the genus Homo. Our resultssupport her suggestion, and further suggestthat area 10 had begun to increase in size inAustralopithecus.

Significantly, the temporal poles of chim-panzees (area TG) receive fibers from theorbital surface of the frontal lobe (area FF)(Bailey et al., 1950). In humans, the tem-poral poles (Brodmann’s area 38) connectwith the frontal lobes, limbic structures, and

‘‘through their interconnections with visualand auditory association cortex, an elaborate

association complex is built up in this anteriorend of the temporal lobe’’

(Crosby et al., 1962:472). Interestingly,the anterior lateral regions of the temporalpoles of humans are activated during therecognition and naming of familiar humanfaces (Damasio et al., 1996).

Until now, received wisdom has been thatbrain size began to increase rapidly in thegenus Homo around 2·0 Ma (Falk, 1992).Our findings that Paranthropus had smalleraverage cranial capacities than previouslybelieved (Conroy et al., 1998; Falk, 1998),and that reorganization of the frontal and

0.0

1600

0–3.5

Millions of years ago

Cra

nia

l cap

acit

y (c

m3 )

–2.5

600

1400

1200

1000

800

400

200

–3.0 –2.0 –1.5 –1.0 –0.5

AustralopithecusParanthropusHomo (individuals)Homo (grouped)

T

Figure 5. Cranial capacities from Australopithecus and Paranthropus plotted against time (Table 5). Forcomparative purposes, cranial capacities are also provided for a number of key representatives of Homo(both individuals and groups; see Falk, 1987 for data). The relative dates for the South African specimensare indicated by error bars; the East African specimens (all Paranthropus) are associated with more preciseradiometric dates. The age estimates for Taung are from Partridge (1986) and McKee (1993). The errorbars for the dates for Homo provide the general range of dates that have been suggested by various workersfor individuals or groups (Falk, 1987; Swisher et al., 1994). The four arrows illustrate the magnitude of thedecrease in new cranial capacities reported on here compared to earlier estimates. T indicates the adultprojection for Taung, which is the only hominid from its site and, although it is the type specimen forA. africanus, manifests a number of Paranthropus-like characteristics in its skull, teeth, and endocast (Falket al., 1995). This graph and the morphological data pertaining to endocasts (see text) suggest that brainsize may have begun to increase in australopithecine ancestors of Homo between 2·5 and 3·0 Ma.

713

temporal lobes appears to have been under-way in Australopithecus well before 2·0 Ma,suggest that the trends for increased brainsize and cortical reorganization may havebegun one million years earlier than pre-viously believed—i.e., in the Australopithecusancestors of Homo (Figure 5). Evidencepertaining to cranial blood flow, on theother hand, suggests that earlier hominidspecies, like A. afarensis from Hadar,Ethiopia (now placed in the genusPraeanthropus by Strait et al., 1997) couldhave been ancestral to Paranthropus, butnot to the Australopithecus–Homo lineage(Falk & Conroy, 1983; Falk et al., 1995;Falk & Gage, 1998). These hypotheses areconsistent with the recent findings of otherworkers based on analyses of postcrania(Berger, 1998; McHenry & Berger,1998). As additional fossil hominids cometo light, we look forward to learningmore about wider areas of the cerebralcortex in early hominids, and to future testsof the ideas and hypotheses presented inthis paper.

Acknowledgements

This research is supported by NSF grantSBR-9729796 (DF), and was made possibleby the cooperation of curators at theTransvaal Museum in Pretoria (C. K. Brainand E. Vrba), the Department of Anatomyat the University of Witwatersrand inJohannesburg (P. V. Tobias), and theNational Museums of Kenya (R. Leakey)who generously provided DF with casts andendocasts from original specimens. Wethank Tim Gage and Bruce Dudek for stat-istical advice, Dieter zur Nedden for makinghis CT unit available to us, Alan Walker forhelping us to obtain copies of hominid endo-casts and for providing a copy of the occipi-tal endocast of Omo L338y-6, R. MacPheeand F. Brady for permitting us to makeendocasts from ape skulls at the AmericanMuseum of Natural History, and Terry

Harrison for helpful suggestions regardingthe manuscript.

References

Bailey, P., Bonin, G. von & McCulloch, W. S. (1950).The Isocortex of the Chimpanzee. Urbana: University ofIllinois Press.

Berger, L. R. (1998). The dawn of humans, redrawingour family tree? Nat. Geog. 194, 90–99.

Brown, B., Walker, A., Ward, C. V. & Leakey, R. E.(1993). New Australopithecus boisei calvaria from eastLake Turkana. Am. J. phys. Anthrop. 91, 137–159.

Cheverud, J., Kohn, L. A. P., Konigsberg, L. W. &Leigh, S. R. (1992). Effects of fronto-occipital artifi-cial cranial vault modification on the cranial base andface. Am. J. phys. Anthrop. 88, 323–345.

Connolly, J. C. (1950). External Morphology ofthe Primate Brain. Springfield, Illinois: Charles C.Thomas.

Conroy, G. C., Falk, D., Guyer, J., Weber, G., Seidler,H. & Recheis, W. (2000). Endocranial capacity in Sts71 (Australopithecus africanus). Anat. Rec. in press.

Conroy, G. C., Vannier, M. W. & Tobias, P. V. (1990).Endocranial features of Australopithecus africanusrevealed by 2- and 3-D computed tomography.Science 247, 838–841.

Conroy, G. C., Weber, G. W., Seidler, H., Tobias,P. V., Kane, A. & Brunsden, B. (1998). Endo-cranial capacity in an early hominid cranium fromSterkfontein, South Africa. Science 280, 1730–1731.

Cramer, D. L. (1977). Craniofacial morphology of P.paniscus: a morphometric and evolutionary appraisal.Contrib. Primatol. 10, 1–64.

Crosby, E. C., Humphrey, T. & Lauer, E. W. (1962).Correlative Anatomy of the Nervous System. New York:Macmillan.

Damasio, H., Grabowski, T. J., Tranel, D., Hichwa,R. D. & Damasio, A. R. (1996). A neural basisfor lexical retrieval. Nature 380, 499–505.

Deacon, T. W. (1997). The Symbolic Species. NewYork: W. W. Norton.

Dean, M. C. (1988). Growth processes in the cranialbase of hominoids and their bearing on morphologi-cal similarities that exist in the cranial base of Homoand Paranthropus. In (F. Grine, Ed.) EvolutionaryHistory of the ‘‘Robust’’ Australopithecines, pp. 107–112. New York: Aldine de Gruyter.

Falk, D. (1983). Cerebral cortices of East African earlyhominids. Science 221, 1072–1074.

Falk, D. (1987). Hominid paleoneurology. Ann. Rev.Anthrop. 16, 13–30.

Falk, D. (1992). Evolution of the Brain and Cognitionin Hominids (James Arthur Lecture). New York:American Museum of Natural History.

Falk, D. (1998). Hominid brain evolution: looks can bedeceiving. Science 280, 1714.

Falk, D. & Conroy, G. C. (1983). The cranial venoussinus system in Australopithecus afarensis. Nature 306,779–781.

714 . ET AL.

Falk, D. & Gage, T. (1998). Radiators are cool: aresponse to Braga & Boesch’s published paper andreply. J. hum. Evol. 35, 307–312.

Falk, D. & Kasinga, S. (1983). Cranial capacity of afemale robust australopithecine (KNM-ER 407)from Kenya. J. hum. Evol. 12, 515–518.

Falk, D., Gage, T. B., Dudek, B. & Olson, T. R.(1995). Did more than one species of hominid co-exist before 3·0 ma?: evidence from blood and teeth.J. hum. Evol. 29, 591–600.

Finlay, B. L. & Darlington, R. B. (1995). Linkedregularities in the development and evolution ofmammalian brains. Science 268, 1578–1584.

Holloway, R. L. (1972). New australopithecine endo-cast, SK 1585, from Swartkrans, South Africa. Am.J. phys. Anthrop. 37, 173–186.

Holloway, R. L. (1973). New endocranial values for theEast African early hominids. Nature 243, 97–99.

Holloway, R. L. (1975). Early hominid endocasts:volumes, morphology and significance for hominidevolution. In (R. H. Tuttle, Ed.) Primate FunctionalMorphology and Evolution, pp. 391–415. The Hague:Mouton.

Holloway, R. L. (1981). The endocast of the Omojuvenile L338y-6 hominid specimen. Am. J. phys.Anthrop. 54, 109–118.

Holloway, R. L. (1983). Human paleontological evi-dence relevant to language behavior. Hum. Neurobiol.2, 105–114.

Holloway, R. L. (1988a). Brain. In (I. Tattersall, E.Delson & J. van Couvering, Eds) Encyclopedia ofHuman Evolution and Prehistory, pp. 98–105. NewYork: Garland.

Holloway, R. L. (1988b). ‘‘Robust’’ australopithecinebrain endocasts: some preliminary observations. In(F. Grine, Ed.) Evolutionary History of the ‘‘Robust’’Australopithecines, pp. 97–105. New York: Aldine deGruyter.

Kohn, L. A. P., Leigh, S., Jacobs, S. & Cheverud, J.(1993). Effects of annular cranial vault modificationon the cranial base and face. Am. J. phys. Anthrop.90, 147–168.

Leakey, R. E. F. & Walker, A. (1988). New Australo-pithecus boisei specimens from east and west LakeTurkana. Am. J. phys. Anthrop. 76, 1–24.

McHenry, H. M. & Berger, L. R. (1998). Body pro-portions in Australopithecus afarensis and A. africanusand the origin of the genus Homo. J. hum. Evol. 35,1–22.

McKee, J. K. (1993). Faunal dating of the Taunghominid fossil deposit. J. hum. Evol. 25, 363–376.

Pakkenberg, B. & Gundersen, J. G. (1997). Neocorticalneuron number in humans: effect of sex and age.J. Comp. Neurol. 384, 312–320.

Partridge, T. (1986). Paleoecology of the Pliocene andlower Pleistocene hominids of southern Africa: howgood is the chronological and paleoenvironmentalevidence? S. Afr. J. Sci. 82, 80–83.

Passingham, R. E. (1998). The specializations of thehuman neocortex. In (A. D. Milner, Ed.) Compara-tive Neuropsychology, pp. 271–298. Oxford: OxfordUniversity Press.

Semendeferi, K. (1994). Evolution of the hominoidprefrontal cortex: a quantitative and image analysis ofarea 13 and 10. Ph.D. Dissertation, University ofIowa.

Semendeferi, K. & Damasio, H. (2000). Brain size inliving hominoids. J. hum. Evol. in press.

Stephan, H., Frahm, H. & Baron, G. (1981). New andrevised data on volumes of brain structures ininsectivores and primates. Folia primatol. 35, 1–29.

Strait, D. S., Grine, F. E. & Moniz, M. A. (1997). Areappraisal of early hominid phylogeny. J. hum. Evol.32, 17–82.

Swisher III, C. C., Curtis, G. H., Jacob, T., Getty,A. G., Suprijo, A. & Widiasmoro (1994). Age ofthe earliest hominids in Java, Indonesia. Science 263,1118–1121.

Tobias, P. V. (1967). Olduvai Gorge, Volume 2 TheCranium and Maxillary Dentition of Australopithecus(Zinjanthropus) boisei. Cambridge: CambridgeUniversity Press.

Tobias, P. V. (1997). Evolution of brain size, morpho-logical restructuring and longevity in early hominids.In (S. U. Dani, A. Hori & G. F. Walter, Eds)Principles of Neural Aging, pp. 153–174. Amsterdam:Elsevier.

Walker, A. C., Leakey, R. E., Harris, J. M. & Brown,F. H. (1986). 2·5-Myr Australopithecus boisei fromwest of Lake Turkana, Kenya. Nature 322, 517–522.

Weber, G. W., Recheis, W., Scholze, T. & Seidler, H.(1998). Virtual anthropology (VA): methodologicalaspects of linear and volume measurements—firstresults. Coll. Antropol. 22, 575–583.

Appendix

Endocasts were reconstructed for fourParanthropus specimens and their cranialcapacities determined by JG in consultationwith DF. The procedures for reconstructingeach specimen and determining its cranialcapacity are described below.

SK 1585The natural SK 1585 endocast of P. robustusfrom Swartkrans, South Africa, reproducesmost of the right hemisphere, except for asmall portion at the rostral end of the frontallobe (Figure 4). DF studied the originalspecimen, made a replica of it, and verifiedthat the replica corresponded metrically tothe original. Available aspects of the SK1585 endocast reveal features of both theoccipital sinus and the superior sagittal sinusextending without interruption to within6 mm of bregma. The right temporal pole is

715

intact and resembles the temporal poles ofKNM-WT 17000 and KNM-WT 17400,differing dramatically from the temporalpoles of Sts 5, Sts 19, Sts 60, and Stw 505.Extension of the midline provides a reliablesite along which to reconstruct the pole ofthe frontal lobe which is otherwise largelyintact including the orbital surface. Wereconstructed the missing portion of thefrontal lobe directly on this replica, usingunreconstructed portions of endocasts fromParanthropus specimens OH 5, KNM-WT17400, and KNM-WT 17000 as models forthe shape. The latter two specimens (Leakey& Walker, 1988) were not available in 1972when SK 1585 was first reconstructed(Holloway, 1972). Although OH 5 (whichwas available) reproduces the dorsal surfaceof rostral frontal lobe, it lacks the orbitalsurface. We molded our reconstruction to fitthe natural contours of the available por-tions of SK 1585’s frontal lobe. As describedfor the 1972 reconstruction (Holloway,1972), we reconstructed the hypophysialregion and the inferior surface of themedulla to the border of the foramen mag-num (referring especially to OH 5 andKNM-WT 17000), and filled in a few minorpits and depressions (Figure 4).

After the general shape of the right hemi-sphere of the SK 1585 endocast was recon-structed, we made a silicone mold of it tothe midline. Although the cleft between theinferior border of the temporal lobe and thecerebellum (i.e., the space normally occu-pied by the petrous pyramid) is filled medi-ally with matrix, the lateral portion of thiscleft reproduces the brain nicely (Figure 4).We measured the surface dimensions of theextraneous matrix, the depth of its lateralextent, and the length of the intact portionof the cleft with calipers. These measure-ments, together with observation of theintact (left) side of the comparable region inOH 5, were used to reconstruct the shape ofthe cleft in plasticine, which was then fittedinto the appropriate place within the mold of

our reconstruction, effectively reproducingthe missing portion of the petrous pyramidwhile subtracting the extraneous matrix.This mold was used to produce a finalsilicone copy. To obtain a cranial capacity,we filled this mold with water and measuredits volume in a graduated cylinder anddoubled the measurement to representboth sides of the reconstructed endocast.

OH 5We first compared our copy of an endocast(copied by DF at the National Museums ofKenya) from the original reconstructed skullof OH 5 to silicone impressions preparedfrom the interior surfaces of Wenner-Grencasts of five cranial fragments that, whenreassembled, reproduce much of the brain-case of the skull. The fragments include therostral end of the dorsal surface of the fron-tal bone on both sides, a large fragment thatreproduces both parietals, two fragmentsrepresenting the petrous temporals, and theoccipital bone. After we verified that theimpressions from the individual fragmentsmetrically and visually reproduced thedetails revealed on the endocast fromNairobi, we coated the internal surfaces(i.e., the braincase) of each of the five cranialfragments with a silicone rubber to which weadded russet pigment. We then recon-structed the cranium from the coated cranialfragments using plasticine to compensatefor missing portions that would bereconstructed in subsequent steps. In mod-eling our reconstruction of the skull, wefollowed the natural contours of the avail-able cranial fragments. We then cast theentire braincase using unpigmented silicone.After the silicone cured, we disarticulatedthe cranium. The resulting endocast waspigmented russet in areas that reproducedthe internal surfaces of cranial fragmentsand white in other reconstructed regions.We proceeded to refine the white areas ofthe endocast, while remaining faithful to theoverall size and contours, by referring to the

716 . ET AL.

unreconstructed portions of Paranthropusendocasts SK 1585, KNM-WT 17000,KNM-WT 17400, KNM-ER 732, andKNM-ER 23000. (We also compared thegeneral shape of our reconstructed endocastwith that seen in photographs of the endo-cast of KNM-ER 13750, Leakey & Walker,1988).

The orbital surface in the region of thenasal rostrum required sculpting with sili-cone sealant in order to reflect the typicalbeaked-shape and relative size of KNM-WT17400 and KNM-WT 17000. Dorsal viewsof these two endocasts, KNM-ER 23000,and a photograph of KNM-ER 13750 wereused to reconstruct the area between therostral region of the frontal lobe and theposterior two-thirds of the braincase [Figure3(b)]. The anterior portions of the temporallobes (Figure 4) were reconstructed basedon the morphology of SK 1585, KNM-WT17000, and KNM-WT 17400. After wewere satisfied that our reconstructed endo-cast was as true to the known morphology ofother Paranthropus endocasts as possible, wecompared it with measurements and draw-ings provided for an earlier reconstruction ofthe endocast (Tobias, 1967). Our overallmeasurements and those provided earlier(Tobias, 1967:92) varied by less than1%. Finally, a master mold was cast inplaster and used to pour a final siliconeendocast. A cranial capacity was determinedby filling our reconstructed endocastwith latex foam and water displacing it.Due to the completeness of the originalcalvarium, we have the highest confidence inour endocast reconstruction and cranialcapacity.

KNM-ER 732We coated the portions of the interior of acast of this specimen (acquired from theNational Museums of Kenya) that repro-duce morphology of the cerebral cortex withpigmented silicone. A clearly visible frontal

crest arcs posteriorly into the sulcus thatlodges the superior sagittal sinus, whichtogether comprise an arc of approximately5·7 cm. The midline of the endocast wasdetermined by projecting straight back fromthe caudal end of the visible portion of thesulcus. The partial right hemicranium wasthen cast to the midline with white silicone,and the resulting hemi-endocast was color-coded as described for OH 5. This specimenwas then compared metrically and visuallyto a copy of an endocast that was preparedfrom the original fossil (acquired by D.F. inNairobi). Once we were assured that thesilicone endocast maintained the integrity ofthe original specimen, we proceeded toreconstruct the missing portions. Theseincluded most of the orbital surface and partof the lateral surface of the frontal lobe,anterior end of the temporal lobe, largeportions of the parietal lobe, and virtually allof the occipital lobe and cerebellum (Figure4). Reconstructions of the frontal and tem-poral lobes were based on the same com-parative specimens used to reconstruct theseregions in OH 5. The remaining portionswere reconstructed by referring to theunreconstructed portions of SK 1585,KNM-WT 17000, KNM-WT 17400, andKNM-ER 23000. The completed recon-struction was filled with latex foam for rigid-ity and then cast in plaster, and that moldwas used to yield a final silicone copy. Thecranial capacity estimate was obtained byfilling the plaster mold with water whichwas measured in a graduated cylinder, anddoubled.

KNM-ER 407The approach to this reconstructioninvolved the alignment of four cranial frag-ments that, in addition to providing a signifi-cant length of the sulcus for the superiorsagittal sinus, also provide bilateral featuresof the coronal suture that could be projectedto intersect the midline at bregma. It is alsofortunate that aspects of the lambdoidal

717

suture are apparent for both hemispheres inaddition to the groove for the left occipitalsinus as these features inform a reconstruc-tion with confidence. We coated the internalsurfaces of casts of fragments of the craniumobtained from the National Museums ofKenya with pigmented silicone, articulatedthe fragments using plasticine to compen-sate for the missing portions, and then castthe entire braincase with white silicone. Themidline for the anterior portion of this speci-men was determined in the same manner asit was for SK 1585, using the same availablelandmarks. We confirmed that the pig-mented portions of the resulting endocastmaintained the integrity of the originalspecimen by comparing it with the compar-

able aspects on an endocast that was recon-structed using the original fossil (Falk &Kasinga, 1983). We then proceeded toreconstruct the missing portions on the leftside which included the anterior end of thetemporal, and the rostral end of the frontallobe, compensating for fractures and distor-tions reproduced from the reconstructedcranium. For reference, we used appropriateunreconstructed portions of endocasts fromKNM-WT 17000, KNM-WT 17400, OH5, and SK 1585 (right hemisphere). Thecompleted hemicast (Figure 4) was then castin plaster, and that mold was used to yield afinal silicone copy. A cranial capacity wasobtained by filling the mold with water anddoubling its measured volume.