BERYCOID FISH FAMILY POLYMIXIIDAE LACHNER 203 



1 and 2). On the basis of my limited material, counts of the number 

 of soft dorsal fin rays reveal that 83 percent of all specimens of nobilis 

 &ndjaponica are "identifiable" (portions of the frequency distributions 

 of each species showing no overlap), whereas only 43 percent of the 

 specimens of lowei and japonica are similarly identifiable. Using the 

 total number of gill rakers as a criterion to distinguish these species, 

 only 25 percent of all specimens of nobilis and japonica can be segre- 

 gated, and 97 percent of lowei a,nd japonica. A slightlj' higher percent- 

 age of specimens are distinguishable by using a character index (table 

 3) formed by subtracting the total number of gill rakers for each indi- 

 vidual specimen from the number of soft dorsal fin rays. With this 

 method 85 percent of all specimens of nobilis and japonica are separable 

 and 97 percent of lowei and japonica are separable. If we consider 

 differentiation among these forms on the basis of these percentages 

 only, nobilis and japonica might be considered subspecifically distinct, 

 and lowei and japonica as having reached a specific level of differen- 

 tiation. Thus, these data suggest that nobilis and japonica are more 

 closely related than are lowei and japonica. This may be an erroneous 

 assumption, for the close relationship of japonica with each of these 

 species involves different sets of characters. Also, if two extreme 

 variant specimens (table 3) were removed (one representing a speci- 

 men of nobilis and one a specimen of japonica) the three forms would 

 be completely separable. 



P. japonica occasionally has a sixth dorsal spine, the total number 

 ranging from V to VI as in lowei, but nobilis occasionally has one less 

 spine, the total number ranging from IV to V. 



This measure of strong divergence in japonica leads one to conclude 

 that the principal problem 3^et to be solved is the determination of 

 the exact nature of divergence among the Pacific and Indian oceanic 

 populations. This opinion is supported by the following factors: 

 (1) lovjei and nobilis definitely react as species where they have been 

 found together in Cuban waters, (2) japonica is geographically isolated 

 from the Atlantic forms, and (3) populations of japonica show partial 

 differentiation in some erratic frequency distributions of meristic 

 characters, being not nearly as stable as Atlantic populations. How- 

 ever, this determination must await additional exploration and study. 



The Philippine specimens of japonica show some differentiation 

 on a racial level from the Japanese specimens in the number of dorsal 

 fin rays, lateral line pores, and vertical scale rows and in the length 

 of the barbel. These differences require much verification before 

 any conclusions can be formed regarding their real significance. 



I find that the primary characters listed hy Gilbert (1905, p. 616) 

 distinguishing berndti from japonica (longer maxillary, larger scales, 



