Canine sexing and species number in the Paşalar large hominid sample

News and Views Canine sexing and species number in thePasalar large hominoid sample

Jay Kelley Department of Oral Biology (m/c 690), College of Dentistry, Universityof Illinois at Chicago, 801 South Pauline, Chicago, IL 60612, U.S.A.Email: [email protected]

Berna Alpagut Dil ve Tarih-Cografya, Facultesi, Ankara University, Sihhiye, Ankara,Turkey

Journal of Human Evolution (1999) 36, 335–341Article No. jhev.1998.0283Available online at http://www.idealibrary.com on

? 1999 Academic Press

The ability to sex individual fossil specimensis useful for addressing a variety of issues inhominoid paleobiology, among them tax-onomy. Two methods for sexing fossilhominoid canines have recently been pro-posed and applied to the hominoid samplefrom the middle Miocene site of Pasalar,Turkey (Kelley, 1995a; Waddle et al.,1995). The two analyses differed in the sexassignment of certain canines, with im-portant implications for the species-leveltaxonomy of the Pasalar sample.

Both the upper and lower Pasalar caninessegregate into two discrete size clustersbased on canine height (Kelley, 1995a; seealso Table 3 for more current data). Kelley(1995a) found that all mandibular caninesidentified as male were in the high-crownedcluster and all those identified as femalewere in the low-crowned cluster (Figure 1).While taxonomy was not specifically dis-cussed, the implication of this result wasthat, based on the canines, there is no reasonto presume that there is more than onespecies in the sample. Waddle et al. (1995)on the other hand, identified two groups offemales among the mandibular canines, onecomprising the low-crowned canines and theother a subset of the high-crowned canines(Table 1; Figure 1). They concluded thatthe presence of two distinct size groupsamong the canines identified as female was

0047–2484/99/030335+07$30.00/0

clear evidence for the presence of twospecies in the sample.

Based upon re-examination of the entirePasalar canine sample, we contend that theanalysis of Waddle et al. resulted in sexingerrors, specifically, that all of the high-crowned canines identified as female are infact male. Our purpose here is to detail ourreasons for thinking so, and to present anupdated sexing analysis of the now muchlarger samples of both upper and lowercanines from Pasalar.

Sexing methods

The canine sexing methods of Kelley(1995a,b) and Waddle et al. (1995) have anumber of similarities. Kelley calculated twosimple shape indices for both upper andlower canines. For the lower canines, theseare canine height divided by maximummesio-distal length, and mesial ridge lengthdivided by height. The two indices are usedto spatially segregate male and femalecanines within a bivariate plot. Waddle et al.used a rather elaborate procedure to calcu-late discriminant functions from a largenumber of initial variables. In the end, theirdiscriminant functions for upper and lowercanines also utilized two shape indices each,one of which—canine height divided bymaximum mesio-distal length—is identical

336 . .

0.9

2.0

1.00.3

R/HT

HT

/L

0.4

1.8

1.6

1.4

1.2

0.5 0.6 0.7 0.8

High crowned

Low crowned

High crowned

Canine size

Figure 1. Sexing of Pasalar mandibular canines reproduced from Kelley (1995a). Three of the fourhigh-crowned canines identified as female by Waddle et al. (1995) were in this sample and are indicated bystars; all have low values for relative canine height. Outlined areas are male (upper right) and female (lowerleft) canine shape territories defined by a large mixed-species sample of extant great apes (see Kelley,1995b). R/HT: mesial ridge length/buccal height; HT/L: buccal height/maximum mesiodistal length.

Table 1 Pasalar mandibular canine data fromWaddle et al. (1995)*

SpecimenCrownheight

Height/length

Sexassignment

Low-crowned:C175 9·7 1·11 FemaleC171 10·0 1·04 FemaleD115 10·8 1·24 FemaleHigh-crowned:BP58 15·1 1·12 Female†C203 15·1 1·16 Female†BP55 15·6 1·25 Female†BP57 16·0 1·25 Female†C176 16·3 1·37 MaleBP56 17·0 1·39 MaleC173 17·1 1·39 Male

*Specimens ordered according to canine height,measured in mm.

†Considered here to be sex assignment errors.

to one of Kelley’s indices. The other iscrown cross-sectional area divided by rootarea, both measured at the cervix. Bothmethods also used data from extant greatapes to characterize male and femalecanines, and to evaluate the reliability of themethod for accurately sexing individualcanines.

The two methods differed sharply in theaccuracy of sex assignments of extant apecanines. Although calculated differently,Waddle et al.’s error rates were betweentwo and seven times higher than thoseof Kelley when the methods were appliedto mixed-species samples of known sex.While error rates alone caution againstuncritical acceptance of Waddle et al.’ssuspect assignments, we feel there are

337

Table 2 Comparison of canine height measure-ments of the six Pasalar canines common toWaddle et al. (1995) and Martin & Andrews (1993)

Specimen Waddle et al. Martin & Andrews

D115 10·8 11·1C203 15·1 16·4BP55 15·6 17·2BP57 16·0 17·0BP56 17·0 17·1C173 17·1 19·1

Measurements in mm.

other, more substantive reasons forquestioning their results.

Canine height and wear

Given that one of the two indices in bothKelley’s and Waddle et al.’s methods is ameasure of relative canine height, relativecanine height is critically important in theperformance of the two methods. In fact, wesuggest that it is primarily this measure thatwas responsible for what we believe to be thesexing errors in the analysis of Waddle et al.The specific problem appears to have beenthe failure to take into account the influenceof canine wear. While this problem alsoplagued Kelley’s analysis to some degree, itapparently exerted a stronger influence onresults in Waddle et al.’s method than inKelley’s.

Kelley used the canine measurementsprovided by Martin & Andrews (1993).Re-examination of the Pasalar caninesample revealed that the height measure-ments given by Martin & Andrews wereinconsistently corrected for wear and thatmany of these canines are in fact heavilyworn. Three of the four high-crownedcanines identified as female by Waddle et al.are in this group and all are heavily worn(see Figure 1). The fourth, BP 58 (errone-ously identified as BP 38 by Waddle et al.),was not listed by Martin & Andrews, but ittoo is heavily worn. All of the canine heightmeasurements given by Waddle et al. appearto be actual heights, uncorrected for wear.Consequently, their height measurementsare mostly substantially lower than thosegiven by Martin & Andrews for each of thesix canines common to the two lists (Table2). Further, the four high-crowned caninesclassified as female by Waddle et al. areuniformly lower crowned than the threeclassified as male. As a result, those classi-fied as female also have height/length valuesthat are correspondingly much lower thanthose of the three classified as male, and

that cluster with the height/length values ofthe three low-crowned Pasalar canines, allclassified as female (see Table 1).

The influence of using height measure-ments that are uncorrected for wear can beseen by comparing the measurements inWaddle et al. with those of 18 unworn (13)or very minimally worn (5) high-crownedmandibular canines presently available inthe Pasalar sample (with the heights of thefive minimally worn canines corrected forwear). The reported heights of the sevenhigh-crowned canines listed by Waddle et al.(see Table 1) are uniformly lower than thelowest-crowned specimen of the 18 unworn/minimally worn canines (Table 3, malecanines). The average reported height of justthe four high-crowned canines identified asfemale by Waddle et al. is only 15·5 mm, asopposed to an average height of 18·8 mm forthe 18 unworn/minimally worn canines. Ifwe assume an average unworn height of18·8 mm for the four canines, then theyhave lost, on average, more than 17% oftheir original unworn crown height whilecrown length has remained virtuallyunchanged. Therefore, the values for theheight/length variable in Waddle et al.’sanalysis are artificially low and must beconsidered artifactual.

Since Kelley used the canine measure-ments given by Martin & Andrews(1993), artefacts introduced by caninewear were present in his analysis as well.

338 . .

Table 3 Pasalar canine heights and shape indices

Upper canineheight (mm)

Upper canineheight/length

Root length/crown length

Lower canineheight (mm)

Lower canineheight/length

Mesial ridge length/crown height

Male:x 19·8 (12) 1·25 (12) 1·00 (12) 18·8 (18) 1·43 (18) 0·74 (18)Range 17·9–21·3 1·15–1·40 0·93–1·03 17·7–20·8 1·24–1·70 0·70–0·85

Female:x 11·5 (15) 1·15 (15) 0·90 (15) 12·1 (12) 1·27 (12) 0·56 (12)Range 10·6–12·4 1·05–1·26 0·87–0·93 11·3–12·6 1·14–1·35 0·51–0·62

Values in parentheses are sample sizes. See Kelley (1995b) for precise definitions of canine indices.

Figure 2. Sexing of Pasalar unworn or minimally worn mandibular (above) and maxillary (below) canines.Outlined areas are male (upper right) and female (lower left) canine shape territories defined by a largemixed-species sample of extant great apes (see Kelley, 1995b). R/HT: mesial ridge length/buccalheight; HT/L: buccal height/maximum mesiodistal length; RT/CR: root mesiodistal length at cervix/crown maximum mesio-distal length (root length measured along the same axis as crown maximumlength).

Misclassification of canines was apparentlyavoided because Martin & Andrews’ heightmeasurements were still greater than thoseof Waddle et al. Perhaps more importantly,Kelley’s second index, relative mesial ridgelength, compensated for these artefacts to anextent that Waddle et al.’s seemingly did not(presumably because both measurements ofKelley’s second index are equally affected bywear).

Sexing the Pasalar canines

Figure 2 shows the unworn or minimallyworn mandibular and maxillary caninespresently in the Pasalar sample, plotted onthe male and female shape territoriesdefined by the large sample of extant greatape canines used by Kelley (1995b). All ofthe mandibular canines of this expandedsample plot are either unambiguously maleor unambiguously female (none in the zoneof overlap), with sex assignments beingconcordant with canine height. None of thehigh-crowned mandibular canines evalu-ated by Waddle et al. is included in thisplot since all were considered too heavilyworn to reliably estimate unworn crown

height and could not, therefore, be sexedaccording to shape. However, a plot ofcanine basal crown dimensions reveals thatthe four high-crowned canines identified asfemale by Waddle et al. cluster withcanines determined to be male accordingto the shape indices (Figure 3). Table 3gives updated heights and shape indices forthe Pasalar canine sample using only theunworn or minimally worn canines, thelatter corrected for wear (see Kelley,1995b).

Figure 2 and Table 3 together demon-strate that, with the elimination of heavilyworn specimens from the sample, thePasalar mandibular canines are not as rela-tively low-crowned as reported in Kelley(1995a) and as shown in Figure 1, especiallythe male canines. All now fall within theranges of values expressed among extantgreat apes for this index, although two indi-viduals, one male and one female, fall justoutside the male and female territoriesdefined by the two indices together.

The maxillary canines on the otherhand—which were not analyzed by Kelley(1995a)—are, on average, extremely rela-tively low-crowned, with many males having

339

0.9

2.0

1.00.3

R/HT

HT

/L

0.4

1.8

1.6

1.4

1.2

0.5 0.6 0.7 0.8

Low crowned

High crowned

Canine size

1.1

2.0

1.00.7

HT

/L

0.8

1.8

1.6

1.4

1.2

0.9 1.0

Low crowned

High crowned

Canine size

RT/CR

340 . .

16

12

68

Length

Bre

adth

13

7

11

10

9

8

9 10 11 12 14 15

Females

Males

Waddle et al. females

Sex unknown

Sex

Figure 3. Pasalar mandibular canines plotted according to crown maximum mesio-distal length andperpendicular breadth. Sexed specimens are those from Figure 2(a). Sex unknown specimens were eithertoo worn or broken to calculate shape indices. Stars are the four high-crowned canines identified asfemales by Waddle et al. (1995).

values for the height/length index that arebelow the minimum values expressedamong extant great apes. As a result, most ofthe males lie outside the male territory asdefined by the extant great apes. That thesecanines are in fact male is confirmed by thesecond upper canine index, root length/crown length, which is the single mostrobust index of the four upper and lowercanine indices (Kelley, 1995b). Given thesmall number of extant great ape species, itshould not be surprising that there would befossil species that extend the boundaries ofthe extant great ape canine shape territories,particularly with respect to the height/lengthindex since most Miocene catarrhines havecanines that, on average, are relatively lower

crowned than those of extant great apes(Kelley, 1995a).

Conclusions

While it is often difficult to establish objec-tively that one method has given moreaccurate results than another, we feel thatwe have provided sufficient evidence torefute Waddle et al.’s (1995) claim thatthere are both high and low-crowned femalecanines in the Pasalar sample. We maintainthat the high-crowned canines identified asfemales are instead clear cases of sexassignment errors. It follows, therefore,that canine morphology cannot be used tosupport the presence of two species in

341

Acknowledgements

We thank Peter Andrews for discussion, andboth Andrews and an anonymous reviewerfor comments on the manuscript.

References

Kelley, J. (1995a). Sex determination in Miocene catar-rhine primates. Am. J. phys. Anthrop. 96, 391–411.

Kelley, J. (1995b). Sexual dimorphism in canine shapeamong extant great apes. Am. J. phys. Anthrop. 96,365–389.

Kelley, J. & Plavcan, J. M. (1998). A simulation test ofhominoid species number at Lufeng, China: Implica-tions for the use of the coefficient of variation inpaleotaxonomy. J. hum. Evol. 35, 577–596.

Martin, L. B. & Andrews, P. (1993). Species recogni-tion in middle Miocene hominoids. In (W. H.Kimbel & L. B. Martin, Eds) Species, Species Conceptsand Primate Evolution, pp. 393–427. New York:Plenum Press.

Waddle, D. M., Martin, L. B. & Stock, D. A. (1995).Sexing isolated hominoid canines with special refer-ence to the middle Miocene specimens from Pasalar,Turkey. J. hum. Evol. 28, 385–403.

the Pasalar sample as claimed by Waddleet al.

This does not necessarily mean, however,that there is not more than one largehominoid species at Pasalar. Other attributesof the sample, such as overall sample metricvariation (but see Kelley & Plavcan, 1998)and particularly upper incisor morphology,may well be indicative of the presence oftwo species (Martin & Andrews, 1993).Canine morphology, however, should not beconsidered as corroborating evidence of thistaxonomic hypothesis nor as a legitimaterefutation of a single-species hypothesis.

Our rejection of the results of Waddleet al. (1995) should not be construed as animplied rejection of the method itself. Theartefacts introduced by canine wear inWaddle et al.’s data mean that this methodhas not yet been applied in an unbiasedfashion. Reported error rates notwith-standing, their method may well produceacceptably accurate results when applied toappropriate data. It is desirable to have morethan one useful method for all paleobiologi-

cal inferences and to this end we welcome avalid application of Waddle et al.’s methodfor sexing the canines of fossil species.