424 
Fishery Bulletin 99(3) 
Figure 3 
Representation of a school of 50 Atlantic bluefin tuna. Individuals have been 
separated into depth intervals according to the mean values for all school 
sizes. Layers are arbitrarily separated in the figure to allow visualization of 
school structure. 
Predicting number of fish in school (NFS) from school 
dimensions School length, width, and depth in number of 
individual fish were recorded as indicators of school shape 
(Fig. 3). As in the prediction of NFS from surface counts, 
measurements were recorded in number of fish, not metric 
distance, because a precise spatial scale could not be estab- 
lished consistently. The lack of an accurate spatial scale 
also precluded the measurement of fine-scale school struc- 
ture such as interindividual distances. School dimension 
measures were recorded according to the movement axis 
of a school. Length in number of fish was measured along 
the axis of school motion (x); depth was measured verti- 
cally (y) and width ( 2 ) was measured perpendicular to x 
on the horizontal plane (Fig. 3). School length and width 
were analyzed in relation to school size by least-squares 
regression. Regression models employed each one or both 
dimensions (i.e. length, width, length and width) of school 
shape as independent variables, and N s as the dependent 
variable. Depth data were not analyzed in relation to N s 
because these data may not be collected practically from 
wild schools. The relat ionship between dimensions of each 
school and selected environmental variables was exam- 
ined by using least-squares regression. 
Results 
Behavior of the specimens 
The captive specimens used in our investigation exhibited 
a high degree of awareness of the walls of the enclosure, 
even during periods of excited behavior. No collision with 
or brushing of the net wall was observed from above the 
surface or in the analysis of diurnal activities from under- 
water video footage, and no evidence of nocturnal colli- 
sions was observed. After a brief period of acclimation, the 
tuna did not actively avoid divers in the enclosure; they 
reacted only to avoid collision. 
Visualizing the model 
A three-dimensional model of the typical structure of a 
school of ABT was constructed based on the mean charac- 
teristics of the schools analyzed and on qualitative obser- 
vations of school structure (Fig. 3). The proportionate 
distribution of individuals within school depth intervals 
varied little (see ANCOVA results below), suggesting that 
a single model adequately describes the mean vertical dis- 
tribution of individuals for schools of varying size. 
Number of schools and NFS 
When a single school comprised the entire group of 50 
tuna, less than approximately 20% of the enclosure volume 
was involved in containing such a school (senior author, 
pers. obs.). Fish swimming within such a school were 
observed to travel along a slowly arcing path around the 
entirety of the enclosure without making sharp turns. 
Smaller schools (up to 25 individuals) occupied only a very 
small portion of the volume of the enclosure. Single large 
schools separated into two or more smaller schools and 
joined back together with fluidity. When more than one 
school was observed simultaneously, each school exhibited 
movement independent of another. If two schools came 
close to one another in the enclosure, they would either 
pass by, move through the other group, or join together to 
