3o8 DISCOVERY REPORTS 



We may consider first the degree of variation shown by each measurement. In this 

 respect the actual spread of the curve is not to be rehed on, for the range of percentage 

 measurement values (shown as abscissae) is not on the same scale in all the charts. 

 Perhaps the best idea of the amount of variation is to be had by dividing the largest 

 reading by the smallest. Thus, in No. 21, "Greatest width of skull," except for two 

 "outsize" measurements, the largest reading is only 1-3 times as great as the smallest, 

 indicating a narrow range of variation, while in No. 14, "Vertical height of dorsal fin," 

 the largest reading is nine times as great as the smallest, indicating a very wide range 

 of variation. It should be mentioned, however, that No. 14 (like Nos. 12 and 15) is 

 a bad measurement in the sense that it is difficult to take it each time in a uniform 

 manner, and the wide range of the readings may be due, to a limited extent, to this 

 cause. 



It will be unnecessary to discuss the range of variation of each measurement, as this 

 can be seen at a glance from the tables, and the information may be supplemented by 

 reference to the charts on pp. 298-307. 



Normal variation should give normal frequency curves from data treated in this 

 way, and with one or two exceptions the curves are undoubtedly of this type. Minor 

 irregularities can be attributed to the comparatively limited data from which they 

 are constructed. 



Of the curves which do not conform to the normal frequency type measurement 

 No. 4 (males) may be quickly disposed of since its lack of shape is simply due to paucity 

 of data. Measurement No. 8 is somewhat erratic, but would probably resolve itself into 

 a normal curve with more data. The curves which need more careful examination are 

 those of measurements Nos. 13, 15, 17 and 20, for either in one or both sexes these 

 curves show a tendency to resolve themselves into two peaks. One object of studying 

 the variations of the external characters is to find whether by any chance more than 

 one race is to be distinguished among the whales examined, and if any measurement 

 constituted a distinguishing feature between two such races the probable effect would 

 be two separate maxima in the frequency curve. On the other hand, two peaks in this 

 curve, if not very marked, might be due to chance (unless constructed from a great 

 number of readings) or to faults in the actual taking of the measurements, for there 

 are one or two cases in which there is always some doubt as to the exact point to which 

 the measurement should be taken. 



Although measurement No. 20 increases relatively as the whale length becomes 

 greater, the variations of measurements Nos. 13, 15 and 17 are independent of the 

 length of the whale. Thus for these three measurements we may compare the curves 

 already obtained with curves constructed from the whole of the data relating to Blue 

 whales instead of only those between certain lengths. In other words, we may see 

 whether these double-peaked cur\'es can be shown to be in reality single when a larger 

 amount of data is used. In measurement No. 20 this irregularity appears only in the 

 curve for females. Now the length of the head is very little affected by the length of 

 the body between about 23 and 28 m. (see table on p. 282), so that we may at least draw 



