88 
Fishery Bulletin 119(1) 
LU T T 
500 750 1000 
800 1200 
L.(mm TL) 
T = T 
1000 2000 3000 4000 
L..(mm TL) 
Figure 6 
The 95% confidence intervals for probability distributions of the asymptotic length (L,.) and 
length at which 50% of individuals are mature (L,,,,) from use of the stepwise stochastic sim- 
ulation approach (solid circular lines) and from use of the FishLife approach at both the genus 
(dotted circular lines) and family (dashed circular lines) levels for the 4 selected test species: 
(A) javelin grunter (Pomadasys kaakan), (B) redbreasted wrasse (Cheilinus fasciatus), (C) cam- 
ouflage grouper (Epinephelus polyphekadion), and (D) blacktip reef shark (Carcharhinidae 
melanopterus). Published values from life history studies (black circles) are also presented for 
each species, with line segments connecting the median of the distributions to the study value. 
TL=total length. 
this issue by removing growth studies with highly nega- 
tive A, from the analysis (an indication of a lack of data 
points for juveniles needed to anchor the growth curve 
near the origin). Again, we believe removing studies with 
highly negative A, simply adds to the overall variability 
and does not introduce any specific bias. 
Another caveat is that we used the SST at the locations 
of the life history studies, and those temperatures may 
not reflect the lifelong SST exposure of highly migratory 
species. In the study described here, the difference in 
temperatures is mainly a concern for certain shark spe- 
cies that may migrate between tropical and temperate 
waters. Therefore, our SST limit of 20°C may not exclude 
species that spend most of their time in SST <20°C if 
the study took place in a warmer climate. A difference in 
temperature may introduce a slight bias in M estimates, 
but it is unlikely that the tropical or temperate species 
that occur in SST above 20°C spend much time in very 
cold SST. 
Further, we used the L,,,, value from NOAA diver 
surveys to distance our test from the L,,,, values in the 
original data set and to reduce the potential for confound- 
ing test results. The only input needed for the stepwise 
approach is L,,,,, and we wanted this value to come from 
an independent source. We could have used L,,,, from the 
original life history study, as we did for the javelin grunter, 
but we felt this was a more independent test. 
Finally, it would have been more appropriate to test our 
stepwise approach on species not included in the regres- 
sion models we used to build our approach. However, we 
did not want to exclude the data from our 4 test species in 
these regression models given our low sample sizes. It is 
important to note that parameter estimates for the 4 test 
species came from the same database (FishBase) as those 
used in the FishLife multivariate models; therefore, their 
inclusion does not bias our approach comparisons in favor 
of the stepwise approach. 
Sources of variability in estimates of life history parameters 
There are multiple sources of variability included in the 
distributions of life history parameters from the step- 
wise simulation. Some sources are directly modeled, oth- 
ers are likely indirectly included as part of the overall 
residual error of each model, and some minor sources 
of uncertainty were not included. The main source of 
