origin (i.e., a0= 0). There are SKJ as small as 10 cm in the commer-cial catches, and this corresponds to an age of about 40 days (0.1years) assuming a0= 0. If a0was negative, then these fish wouldbe estimated to be even younger (and less than zero for a0< −0.1).Furthermore, a0= 0 gives a mean length of 49 cm for fish of age 1,which is consistent with otolith readings for small SKJ in the easternIndian Ocean (Kayama et al., 2004).The VB log k growth curve appears to fit the data reasonablywell for all species, but this can be better evaluated through resid-ual plots (Fig. 5). The standardized release and recapture lengthresiduals plotted against estimated age show that the majorityof residuals fall within the desired ±2 standard deviations for allspecies; however, some interesting patterns are apparent. For YFT,there is a cloud of negative values in the release length residualsbetween age 2 and 3. This is the age at which the transition betweengrowth phases is estimated to occur, and because growth becomesvery rapid after this point, the residuals are sensitive to its value.The recapture length residuals are positively biased for the oldestages (>6 years) and suggest that L∞may be underestimated. ForBET, the standardized release and recapture length residuals do notshow any patterns of concern, except some indication of unequalvariance across ages (Fig. 5). Finally, for SKJ, there is high variabil-ity in the release length residuals around the age of transition (0.8years), but otherwise the model fits the data well (Fig. 5).Sex information exists for a relatively small number of YFT andBET recaptures (Table 1), all of which were recaptured as large fish(105–160 cm) after being at liberty for several (2–7) years. Unfor-tunately these data are insufficient for estimating all parametersof the VB log k growth model freely, especially since they containvery little information about growth during the transition betweengrowth phases (Fig. 6). Instead, we were able to fit sex-specificmodels by fixing several parameters at the maximum likelihoodestimates