THE VARIATION, RACES AND MIGRATIONS OF THE MACKEREL. 261 



and Kerry the frequencies of the highest and lowest values are 

 approximately the same (8 °/o) ; while for the Xorth Sea and Channel 

 the lowest values are twice or thrice as frequent as the highest values 

 (9:5 and 12:4). 



So far, therefore, as concerns the main conclusions drawn from the 

 variation of the entire series of bars, the variation of the partial series 

 •furnishes a fairly satisfactory confirmation. These conclusions were 

 the existence of three main races of mackerel in American and British 

 waters, viz. (1) an American, (2) an Irish, and (3) a race common 

 to the North Sea and the Channel. 



The only respect in which the results are at variance concerns the 

 affinities of the fish from Brest and Scilly, which were clearly with 

 the Channel and North Sea fish for the entire series of bars, and 

 with the Irish fish for the partial series. Nothing but the examination 

 of a larger number of fish from this region will solve the difficulty. 

 Additional facts, however, are adduced in the sequel which tend to 

 show that the Brest and Scilly fish are intimately related to those of 

 the Channel and North Sea, but show, under each character examined, 

 a slight approximation towards the Irish race. 



VII. Shape of Transverse Bars, 



The cause which determines the general shape of the transverse bars 

 has already (p. 256) been discussed, and has been found to be the 

 deposit of pigment along the external surface of the myotomes. As 

 the myotomes in the dorsal half of the body have a marked > shaped, 

 or geniculate, curvature, the bars consequently show a marked tendency 

 to assume a corresponding shape. The breaks and anastomoses, to 

 which the bars are subject at the time of their formation, show, however, 

 such diversity and complexity of form that they defied all my earlier 

 efforts to discover a suitable system of classification by which the 

 vagaries in the shape of the bars might be recorded for subsequent 

 comparison. 



I began by noting whether the general arrangement of the bars was 

 regular or irregular, adopting several grades of regularity and irregularity 

 to cover the intermediate conditions ; but, although I have a complete 

 set of data in these terms, and have tabulated the results for comparison, 

 the examination which I have made of them shows them to be 

 practically worthless for exact conclusions. • This is largely due, as 

 I now know, to the fact that regularity "in the bars, i.e., parallelism, 

 with an absence of breaks and bifurcations, may be of two • very 

 distinct kinds, which may be termed, for comparison, primary and 

 secondary. Primary regularity is due to the bars having retained their 



