482 



Fishery Bulletin 98(3) 



This procedure allowed us to determine objectively outliers 

 in the values of N estimated for fish in groups with similar 

 times at liberty. Fish with the same values of liberty, L, and 

 number of complete increment cycles. A', lay on a line of 

 slope UL in the plot of cycle frequency ( V^) against the sum of 

 the initial and final fractions ilF+FF) in Figure 8. Fish with 

 the same value of L and an estimate of N larger or smaller 

 by one complete increment cycle were offset from this line 

 by VL cycles per year. 



In addition, this procedure allowed us to exclude 18 fish 

 from the "direct method" of estimating periodicity along 

 either reading axis, under three criteria. First, fish with 

 anomalies in opaque zone formation, such as "false annuli" 

 (e.g. L. erythropterus sl991,L.johnii sl947,L. sebae sl982), 

 or with other errors in otolith interpretation were included 

 only in analyses exploring the possible relationships 

 between formation of false annuli and environmental 

 conditions. Second, fish with FF>1.0 (e.g. L. erythropterus 

 sl958, sl964 in Fig. 5 ) were excluded because they violated 

 our second working approximation of the equal spacing of 

 outer zones. Third, other fish were excluded because the 

 OTC marks were absent along one axis (L. munostigma ) or 

 were visible outside the outermost opaque zones (L. sebae 

 sl980, L. gibbus). 



When these outliers were removed, the "direct method" 

 produced a normal distribution of cycle frequency along 

 the ventral axis about a mean of 0.96 ±0.32 cycles/yr 

 (n=66 Shapiro-Wilk a'=0.98 P<u>=0.71i for all species 

 pooled. Data from the sulcal axis produced a mean of 



1.02 cycles/yr (/2=64 u'=0.95 P<u-0.04), but this was not 

 normally distributed. 



The estimated cycle frequencies were grouped loosely 

 about one cycle/yr for L. erythropterus. Ljohnii,L. rivulatus, 

 and L. sebae, with L. malabaricus, L. argentimaculatus, and 

 L. bohar having a lower estimate and L. carponotatus and 

 L. vitta having a higher frequency (Fig. 9). All means lay 

 within the range 0.78-1.21 cycles/yr: apparently a lower 

 cycle frequency was estimated when radii measured along 

 the ventral axis (Table 4i were used. Analysis of variance 

 showed no significant differences in cycle frequencies 

 among these species and between reading axes, but a 

 nonparametric Wilcoxon signed-rank test for paired data 

 found significant differences between axes <?)=61 S=-363.5 

 prob>ISI=0.008). 



Simple linear regression of the number of whole and 

 partial cycles formed after OTC marking (liberty fraction 

 LFion time at liberty showed significant linear relationships 

 along both the ventral axis {n=66 ndf=l ddf=64 F=739.37 

 P>F=0.0001) (Fig. 10) and the sulcal axis (n=64 ndf=l 

 ddf=62 F=9 13.36 P>F=0.0001). The slopes were j3=1.01 



