r.jvi/>/i\ FisiniRii.s i:\iT.niTioy\ loi'rio 



147 



6. Comparison vith the samples from Magdalen Islands. — The following table 32 

 shows total averafTt's for the nine samples from Xewfoundlaiid, taken together; for the 

 two Magdalen Islands samples taken together; the differences between gp>\vth 

 dimensions of the same character with the corresponding standard error, and the 



fraction -j-. 



a 



Tabf^e 32. — Growth of Newfoundland herring compared with that of herring from 

 Magdalen Islands by help of total averages for year-class 1904 and 1903 res- 

 pectively. 



.\cea of sanijiles 



Newfoundland 



Magdalen Tsland.-s . . 



Difference ( D) 



D _„ 



Standard error of difference — r 

 ti 



071 

 73 

 002 



10 



It will be noticed that all differences except /2 ' and tio must be regarded a3 

 significant, being greater than seven times their error. Judging from the values of 



the fraction ;y the growth dimensions ^i, t-,, tr,, t-, and ^s would seem to differ most 



in the two sets of samples, h being greatest in the sample from the Magdalen islands, 

 the remaining dimensions being the smallest. On comparing each of the tw^^ samples 

 from Magdalen islands with each of the nine samples from Xewfoinidland, exactly 

 similar results were obtained, even in the case of the 1915 sample from the former 

 locality, where the number of specimens was so very small (seventeen). This will be 



seen from the following table, showing the value of the fraction -^ for each growth 



dimension up to t.^, all values exceeding four, however, being for the sake of conveni- 

 ence placed in one group. 



Table 33. — Showing distribution of the values of -, arising by comparison of either 



a 



of the samples from Magdalen Islands with either of the samples from Xew- 



foundland. 



It will be seen that save for ^, all the growth dimensions exhibit distinct differ- 

 ences between the Newfoundland samples on the one hand, and those from the Mag- 

 dalen islands on the other; most of the differences, moreover, are of a very consider- 

 able magnitude in pro{X)rtion to their errors, and the impression produced by this 



