QUAST; Srbaitrs vmirsatus SP. N. 



Table 4. — Upper and lower 95% confidence limits for 

 single future observations on proportions (percent of 

 SL), based on the material examined.' 



' Based on confidence belts for individual predicted measurements 

 (Soliial and Rohlf, 1969:422) from the parameters for regression (Table 3). 

 Dota were bacl<-transformed to orjthmetic values for computation of pro- 

 portrons. If limits ore desired in the form of measurements, the percent- 

 ages moy be multiplied by the appropriate SL. 



obtained from dn'r confidence limits for fits to 

 regression (Table 3) back-transformed to the 

 original variates and converted to percent of 

 standard length. Marr (1955) discussed the 

 disadvantages of using proportions in syste- 

 matics, particularly when size-specific changes 

 are not identified, but neglected several advant- 

 ages. Size-specific changes are here identified, 

 and the presentation of data in the form of pro- 

 portions facilitates rapid comparisons between 

 specimens and allows data on measurements to 

 be presented more economically than in graphs. 

 Proportions vary much less with length than 

 original measurements, sometimes surprisingly 

 little when the numerous factors that influence 

 their variation are considered (Table 4), and 

 therefore allow easier interpolation between 

 tabulated reference points than representations 

 of original variates. Also, the continued use of 

 pi-oportions in fish systematics, often without 

 indication of size-specific changes, attests to a 

 prevailing opinion that proportions are useful 

 despite their occasional misuse and other draw- 

 backs. 



Plots of the limits for proijortions (Table 4) 

 disclose that they are slightly curvilinear and 

 asymmetrical from left to right. These char- 

 acteristics reflect the y-intercept effect dis- 

 cussed by Marr (1955) combined with the ef- 

 fects of normal divergence of confidence belts 

 in regression with distance from the combined 

 mean, allometry between the original variates, 

 and distortion caused by back-transformation 

 from the logarithm of a function. Size-specific 

 confidence limits for proportions whose allomet- 

 ric coefticients did not differ significantly from 

 unity (Table 3) are included in Table 4 because 

 the calculated value is a better estimate of the 

 exponent than is arbitrarily assumed isometry. 



AXIAL SKELETON 



X-rays of types show 27 vertebrae, including 

 urostyle (Table 1). Pterygiophores of spinous 

 dorsal fin single in spaces between neural spines 

 except between neural spines 2 and 3, which 

 contains pterygiophores of dorsal fin spines 2 

 and 3. Pterygiophores of soft dorsal fin usually 

 doubled in interneural spaces except single pre- 



395 



