Myer M. Orenstben 
75 
Figures 1 to 6 give the calculated frequency curves. The lengths of the 
ordinates represent the frequency per thousand of observations in a corresponding 
group where the ordinate considered is the mean for the unit of grouping. The 
dotted curves have been drawn from the observed frequencies reduced to the 
same base. A glance at these figures is sufficient to show that normal curves fit 
pretty well our statistics. The quartiles have been drawn with the aid of an 
integraph. 
6. Coefficients of Correlation. With the same grouping of the data I have 
constructed correlation tables. These are given in Tables VIII to XIV. 
The coefficient of correlation for the length and the breadth of the head was 
found to be r = + 0-244 ± 0-020. This coefficient is as we should have expected 
lower than the results from the remaining pairs of characters. Attention may be 
drawn to the fact that this particular coefficient varies very much from race to 
race. A comparison with similar results for different races may be of interest here : 
6 
Smith Sound Eskimos ... r 
Ainos ... ... ... ... r 
Naqada Race ... ... ... r 
Germans ... ... ... r 
Bavarian Peasants ... ... r 
Cairo Natives ... ... ... r 
French Peasants ... ... r 
British Columbian Indians ... r 
= +0-47 ±0-08 
= +0-43±0-06 
= +0-34±0-05 
= +0-29±0-06 
= +0-28±0-06 
= +0-24±0-02 
= +013±0-09 
= +0-08+ ? 
Taking into consideration the great variation of this coefficient from race to 
race, the writer is greatly inclined to believe that a correlation between the length 
and the breadth of the head really exists, but that its variation in value is very 
probably due to special factors, probably working independently and tending to 
modify the real measure of relationship. This is being further investigated, and 
a closer study of such factors is being undertaken. 
The results of the computation of Tables VIII to XIV are grouped in Table V. 
From this table we may deduce the following conclusions : Were all human bodies 
perfectly similar all the coefficients of correlation between the different organs 
would be unity; if on the contrary there should exist a complete independency 
of the organs, the coefficients would be all nil. We may therefore conclude from 
Table V that we are to expect a tall man to have more often long feet than long 
fingers and short men will more often have long fingers than long arms. Further, 
men with long arms will very rarely have short fingers, and such people will more 
often be men with greater stature than men with long feet, and so on. 
Table VI gives the probable error for each coefficient of correlation shown in 
Table V. They are in all cases very small, which proves our results to be very 
reliable. 
