variate. The most likely peculiarity in combined-sex samples is the sex 

 composition of the samples: for instance, the gorilla sample of 668 

 capacities comprises 414 males and only 254 females. 



Related to this factor, we have already examined one possible charac- 

 teristic of the variate itself that may play a part, namely the existence 

 of differing degrees of sexual dimorphism in cranial capacity among the 

 various hominoid taxa, such as distinguish gorilla from chimpanzee capac- 

 ities. This factor, coupled with differing proportions of the 2 sexes in 

 the various samples, may explain some part of the differences in distribu- 

 tion pattern among the hominoids cited. The age distribution of the 

 samples is another possibly correlated variable. Although all the samples 

 quoted here are of adults, we do not know if the cranial capacity decreases 

 with age. We know that brain weight and volume diminish with old age 

 (Mettler 1955), but, to the best of my knowledge, no corresponding 

 diminution in cranial capacity has been demonstrated, as was mentioned 

 earlier. Hence, it remains uncertain whether age distribution of the 

 samples has contributed to the discrepancies of skewing. 



In a letter to me, Dr. T. Bielicki raised the possibility that the 

 inclusion of different geograpliical variants may be a further factor con- 

 tributing to the differences among the samples. Some of the earlier 

 workers (for example, Selenka 1898, 1899; Gaul 1933) were at pains to 

 distinguish such variants, recognizing larger- and smaller-brained races as 

 represented within available samples. However, it would be most difficult, if 

 not impossible, to obtain accurate details on the geographical provenience 

 of all the pongid data in the literature, as authors generally have not given 

 this information. 



Simpson, Roe, and Lewontin have pointed out an interesting statis- 

 tical reason why samples of normally distributed variates should show 

 a small positive skew: 



It may be noted that the assumption on which the use of V and the study of 

 variability in general are based involves a constant tendency for such characters 

 to show a small positive skew. This assumption is that the dispersion is propor- 

 tionate to the absolute value of the variate. If this is true as between different 

 samples and different variates, it should be true also within a single distribution. 

 The absolute dispersion should tend to be, or for an average of many different 

 homologous distributions should be, greater for higher values of the variate than 

 for lower and should increase steadily from the left-hand end of the graphic dis- 

 tribution through to the right-hand end. Since the values thus tend to be spread 



^ 58 



