78 
BULLETIN OF THE BUREAU OF FISHERIES 
2 are graphic representations of the frequency distribution of the gill rakers in the 
group of larger specimens. In 69 smaller specimens of arenarius of 30 to 60 milli- 
meters the mode for the total number of gill rakers falls at 15, with the class 16 a 
very close second. Comparing these smaller specimens of arenarius with the larger 
regalis, it may be seen that the modes are even then distinct, although the overlapping 
is quite considerable. I do not have a sufficient number of smaller regalis to estab- 
lish the frequency distribution, but from the few specimens counted it seems evident 
that the difference between the larger and smaller members of this species is not so 
marked as it is in arenarius. 
A precaution to be taken when the number of gill rakers is used as a distinguishing 
character may be mentioned. As may be conjectured, it is sometimes difficult to 
NUMBER OF DORSAL RAYS. 
Figuee 3. — Frequency distribution of soft articulate dorsal fin rays of Cynoscion arenarius and 
C. regalis. The first short ray, which remains unjointed in the largest specimens, was not in- 
cluded in the count, and the last two rays, which are apparently united at their base, have been 
counted as one 
decide as to what constitutes a gill raker and what is a mere tubercle at each end of 
the gill arch. It is hard to describe in so many words where a line may be drawn. 
The number counted will vary somewhat with the observer, but when the same 
investigator makes the counts in both species the numbers are comparable. In the 
present study all gill rakers were included that were big enough to be manipulated 
with a dissecting needle, or about 1 millimeter long as a rough estimate. If the gill 
rakers on the lower limb only are counted, the number of doubtful cases will be 
surprisingly small. When the total number of gill rakers is enumerated, the diver- 
gence in the two species is emphasized because there is usually a difference of one or 
two gill rakers also on the upper limb. However, there is greater Affiance for error 
