214 
Fishery Bulletin 96(2), 1998 
tember through March occurred three times during 
the study period, whereas the period April through 
August occurred only twice), the magnitude of the 
imbalance during the study period was not large. 
Data were too few to examine any annual effects. 
We used maximum-likelihood estimates of model 
parameters from the best-fit models to estimate ex- 
pected discard per set for each set type. We calcu- 
lated estimates for individual geographic areas where 
model fits indicated significant areal effects on pa- 
rameter values (i.e. for school and log sets), then cal- 
culated a “pooled” estimate of expected discard per 
set from the area-specific estimates. Where no areal 
stratification was appropriate (i.e. for dolphin sets), 
we calculated only one “pooled” estimate using the 
maximum-likelihood estimates of parameter values 
for the unstratified model. 
We estimated standard errors for expected discard 
per set by bootstrapping rather than by using ana- 
lytic formulae because only bootstrap methods could 
be efficiently applied to all three set types. Analytic 
methods could be easily applied only in the case 
where no areal stratification was appropriate (i.e. 
dolphin sets, where we fitted a single model for all 
areas) and in the case where complete areal stratifi- 
cation was appropriate (i.e. log sets, where we fit sepa- 
rate models in each of the geographic areas). 
We estimated annual discard and standard errors 
of those estimates for the U.S. fleet as observed dur- 
ing the study period by multiplying the estimated 
overall discard rate for each set type by the corre- 
sponding average annual number of sets. 
To investigate the effect of entirely eliminating 
dolphin sets, we predicted potential discard after 
redirecting all dolphin sets to school or log sets (or 
both). We used three hypothetical redistributions: 1) 
all dolphin sets were reassigned to school sets; 2) half 
of dolphin sets were reassigned to school sets, the 
remaining half were reassigned to log sets; and 3) 
all dolphin sets were reassigned to log sets. Sets were 
reassigned in proportion to the observed proportions 
of set types by the U.S. fleet in each geographic area; 
therefore the pooled estimates of expected discard 
per set could be used for each set type. 7 
We predicted total redistributed discard during the 
study period by multiplying pooled estimates of ex- 
pected discard per set for school and log sets by the 
7 The geographic distribution of set types by U.S. vessels tends to 
differ from the geographic distribution with international fleet 
as a whole because U.S. vessels are largely restricted from the 
nearshore Exclusive Economic Zones (EEZ’s) of coastal nations. 
If data for the geographic distribution of effort by the interna- 
tional fleet had been available, it would have been more appro- 
priate to use that. However, the general results of the present 
study would likely remain unchanged (see “Caveats” in “Dis- 
cussion” section). 
redistributed number of sets for each type. This pro- 
cedure assumes that the redirected effort did not 
change the fishing habits, geographic patterns of the 
fleet, or the discard per set realized during the redi- 
rected effort. 
Results 
Patterns in set-type distributions 
Several patterns are evident in the number of sets of 
each type. During the study period, about half the 
sets by U.S. boats were made on dolphins, about one 
quarter were made on school sets, and one quarter 
on log sets (Table 1). Sets were most common in the 
relatively nearshore area 1, rare in the offshore equa- 
torial area 2, and moderate in the remaining area 3. 
Only dolphin sets were made with any frequency in 
area 2 (Table 1; Fig. 2). All three set types were com- 
mon in areas 1 and 3. 
About 70% of school sets were observed in area 3 
(Table 1: Fig. 2). Of those sets, most occurred during 
April and May (59% of sets), between the hours of 
8 am and 4 pm local time (68% of sets). Observed 
school sets in area 1 also occurred primarily between 
8 am and 4 pm (76% of sets) but were distributed 
more homogeneously throughout the year and were 
much less common than in area 3. No school sets 
were observed in area 2. 
Log sets were also observed primarily in area 3 (58% 
of sets, Table 1), although fewer sets were observed 
here during the period June— September (Fig. 2). Ob- 
served log sets decreased during summer and win- 
ter months in area 1. Log sets usually occurred early 
in the day in all areas (80% before noon, 56% before 
8 am, Fig. 2). Less than 1% of observed log-set effort 
occurred in area 2. 
Contrary to school-set and log-set effort, dolphin 
sets were observed primarily in area 1 (66% of sets, 
Table 1 ) with a small but important number of sets 
observed in area 2 (17% of sets). Effort in area 2 ob- 
viously replaced effort in area 1 during the months 
of June and July (Fig. 2). Only 1.6% of effort in area 1 
but 78% of effort in area 2 occurred during these two 
months, and all effort in area 2 occurred between 
June and October. In all three areas, dolphin sets 
were observed primarily (79%) between the hours of 
8 am and 4 pm during which greater effort was evi- 
dent before noon. 
Discard patterns 
Estimated tuna discard from U.S. vessels was re- 
corded for 59% (Table 1) of observed dolphin sets, 
