290 



fined. Furthermore, Coffoi/nsfer has Iteen taken only in the 

 larger streams or their immediate neighborhood, as is shown 

 by the distribution map for that species; while Diplesion is 

 limited to the smaller rivers and creeks. 



With respect to Cntfof/dsfer sluiinanli and EtJicostoma crpru- 

 leniii (1436 and 1477), the case is a little less clear, and it is 

 quite possible that with a larger number of collections contain- 

 ing the former species, the two might have been found in com- 

 pany. It is true that only 4 per cent, of our collections of Efhe- 

 osfovia ('(eriih'ioii have come from the larger rivers and from 

 stagnant waters to which Coffogaster is confined. On the other 

 hand, a concurrence of the locality marks on the maps of dis- 

 tribution of these species (Maps V. and XII.) shows that the 

 two were taken from the same locality — although not in the 

 same collections — in three out of nine possible cases. 



The lack of any coincident occurrence of Coffoi/at^fer .s/in- 

 mardi and Efhcdsfonia zonale (1436 and 1461) is explained by 

 a glance at the maps (V. and X.). as due, not to a difference of 

 geographical distribution, which is approximately identical for 

 the two, but to that of local preference, the former species oc- 

 curring only in or near the largest streams, and the latter be- 

 ing limited to the smaller rivers and creeks. Indeed, the two 

 species were not taken by us from even the same locality at 

 any time. 



Nearly the same niaj' be said of Diphsion hlcinili)ides and 

 Boleosnina rainundii (1443 and 1448), which have come from the 

 same locality but once, although in general distribution they 

 are not mutually exclusive. B/citiiioidcs, as may be seen from 

 Table VI., is a species of more indefinite preferences than 

 camuruni, and occurs in various situations from which the lat- 

 ter is excluded. 



I take up next five pairs of species, representatives of 

 which have been occasionally taken together by us, but the co- 

 efficients of whose association are nevertheless very small. 



Ef/ieostoDia zoiiale and E. Jcfisia' (1461 and 1474), for ex- 

 ample, with an associative coefficient of only .37, show a pre- 



