FISHERY BULLETIN: VOL. 69. NO. 2 



to 21 mm from a 50-cm K. pelamis. Even the 

 longest raker of a 125-cm C. hippuriis was only 

 16 mm — shorter than that of a 50-cm K. pelamis. 



SIZE AND SPECIES COMPARISONS 



METHODS 



Linear regressions relating gill I'aker gap to 

 fork length and log filtering area to log fork 

 length were computed for each species. Re- 

 gressions were computed once about the mean, 

 and a second time, were forced to pass through 

 the origin. The latter procedure was used be- 

 cause the ranges of fork lengths of some species 

 were not sufficient to obtain reasonable equations 

 (Figure 2). 



Both K. pelamis and T. albacares were rep- 

 resented by large samples that included small 

 and large specimens. Their regressions of gill 

 raker gap on foi-k length passed close to the 

 origin even when not forced to do so; the y-in- 

 tercept was 0.00 mm for K. pelamis and 0.19 mm 

 for T. albacares (Figure 2) . In contrast, kawa- 

 kawa, Eiithynnus af finis (Cantor), and T. ala- 

 lunga were represented by small samples that 

 did not include small specimens. Regressions 



extrapolate outside the size ranges represented 

 in our samples, the regressions forced to pass 

 through the origin were used for all computa- 

 tions of gill raker gap. 



The same reasoning was used for the relations 

 between log filtering area and log fork length. 

 In this case, the zero-zero intercept was equi- 

 valent to 1 cm fork length and 1 mm- filtering 

 area rather than zero fork length and zero filter- 

 ing area. Since most comparisons made later 

 were for fish at least 35 cm long with filtering 

 areas near 100 mm', errors owing to the posi- 

 tion of the intercept were believed negligible. 



SIZE AND SPECIES COMPARISONS 



Linear regressions passing through the origin 

 that relate gill raker gaj) to fork length and log 

 filtering area to log fork length are presented 

 in Table 1 along with the numbers and lengths 

 of fishes measured. 



Mean gill raker gap increased with fork length 

 and was equal to 1.4 and 6.6 '"r of fork length 

 for frigate mackerel, Auxis rochei (Risso), and 

 C. hippuriis, respectively. Gill raker gap in the 

 middle of the lower branch was usually 1.0 to 

 1.2 times the mean gill raker gap except for 



Table !.■ — Linear regressions passing through the origin that relate mean gill raker gap to forl< length and log 

 filtering area to log fork length and the number and length of fish measured. 



Species 



Number 



of (ish 



measured 



Regressions 

 of gap (G) 

 and fork 

 length (/) 



Fork length 



Mean 

 (cm) 



Range 

 (cm) 



G 



(mm) 



Standard 



error of 



estimate 



for G 



(mm) 



Regressions of log 



filtering area (log A) 



and log fork 



length (log /) 



Log / 

 (cm) 



Log A 

 (mm-) 



Standard 



error of 



estimate 



for Log A 



(mm-) 



of gill raker gap on fork length for these two 

 species did not closely approach the origin (Fig- 

 ure 2) ; we believe these equations would also 

 have had ^/-intercepts near 0.0 mm if lengths 

 of our specimens had been more evenly distrib- 

 uted. Since some comparisons were made that 



A. rochei (1.3) and K. pelamis (1.4). Mean 

 gap increased in direct proportion to fish length; 

 i.e., if length doubled, gap also doubled. 



Filtering area increased as the 1.4 to 1.8 pow- 

 er of fork length. When these regressions were 

 not forced to pass through the origin, the filter- 



364 



