362 



FISHERY BULLETIN OF THE FISH AND WILDLIFE SERVICE 



TOTAL LENGTH mm 



Figure 4. — Relations between length of longest dorsal 

 spine and total length. Open circles and fine line repre- 

 sent Costa Rican data; solid circles and heavy line 

 represent Hawaiian data. 



which Schaefer (1948) had fitted a linear regression 

 to the original data. Again, the resulting regres- 

 sions, plotted in figure 5, when subjected to 

 covariance analysis, indicate a small, constant 

 average percentage difference between finlet 

 lengths of the two populations, the Hawaiian fish 

 having the longer finlets. 



Head length and distances from snout to fin 

 insei'tions 



As mentioned earlier, Godsil (1948) has pub- 

 lished the measurements of total length, head 

 length, and distances from tip of snout to the 

 insertions of first dorsal, second dorsal, anal, and 

 ventral fins for nearly 2,000 specimens of yellowfin 

 tuna from the American west coast between Cape 

 San Lucas and Panama. The original measure- 

 ments were published with his analyses of them, 

 so we are able to compare tbese extensive data 

 both with the Costa Rican data published by 

 Schaefer (1948) and with the Hawaiian data pre- 

 sented herein. In figures 6 to 10 have been 

 plotted head length and distances from snout to 

 fin insertions against total length, which is taken 

 in each case as the independent variable. For 

 each of the three groups of data (Godsil's, Costa 



Rican, Hawaiian) have been plotted tbe mean 

 values of the two variables in each grapb for each 

 10 cm. of total length. To the pooled west-coast 

 data (Godsil's plus my Costa Rican) have been 

 fitted and plotted linear regressions. Also plotted 

 are the curvilinear regressions computed by Godsd 

 (1948, p. 13) forhisdata, of thetype?/=a+62;-|-c/x. 

 On the same graphs have been plotted also the 

 linear-regression line best fitting the Hawaiian 

 data and the best-fitting curvilinear regression of 

 the type selected by GodsU. 



For the Hawaiian data, except in one case 

 (snout to insertion of second dorsal of Hawaiian 

 fish), the cxu-vUinear regressions provide a slight 

 improvement in fit over the linear regressions. 

 Inspection of the figiu-es, however, reveals that 

 the differences between the linear and curvUinear 

 regressions are small in comparison with the 

 differences between west-coast and Hawaiian 

 samples. The reduction of the variance about 

 the regression line also is very small in comparison 

 with the difference between the two regions 

 when a cm-vilinear rather than a linear equation 

 is employed. In consequence, the linear-re- 

 gression equations will be employed below in 

 considering the application of analysis of covari- 

 ance to the comparison of samples. 



400 600 800 1000 1500 2000 



TOTAL LENGTH MM 

 Figure 5. — Relations between length of longest dorsal 

 finlet and total length. Open circles and fine line 

 represent Costa Rican data; solid circles and heavy line 

 represent Hawaiian data. 



