406 



Fishery Bulletin 104(3) 



For each location, in situ speed increased with size, 

 as expected, except, perhaps for the largest individual 

 (Table 1, Fig. 3). For the three locations with three or 

 more observations, in situ speed increased at 1.7-2.6 

 cm/s per each mm of growth depending on location 

 (Table 1), and the 95% confidence interval for the three 

 slopes overlapped broadly. However, because the height 



of the speed versus size lines varied (v intercept ranged 

 from -6 to -18), the actual speed at any size differed 

 among locations by as much as 10 cm/s. For the in situ 

 speed data, PA provided fits with r- values that were 

 very similar to those of SL. The r- for PA was 0.07 

 higher than that for SL for Nan Wan Bay North, equal 

 for Her Chen, and 0.15 lower for Wan Li Tong. 



Comparison between critical speed 

 and in situ speed 



For each 1-mm (SL) increment in size, we cal- 

 culated a mean speed for both critical speed (54 

 individuals) and in situ speed (19 individuals). 

 These size-specific speeds are summarized in 

 Table 2 (both measurements were not available for 

 every size increment). The ratio of the two size- 

 specific measures (/;; situlU^^^^) ranged from 0.35 

 to 0.75, and had a mean value of 0.52 ±0.15. The 

 size-specific speed measures were not significantly 

 correlated (in s(7w = 0.45f;„„-hl.45, 7-^=0.47, P=0.13, 

 ;; = 6). Scaled in situ speed (BL/s) was one-half 

 scaled f/^,.,^. 



Endurance 



Endurance data were available for 12 individuals 

 ranging in size from 9 to 14 mm SL, and for two 

 ages (21 and 24 days at the start of the endur- 

 ance run) measured in two batches of six. The 

 range in fish swimming times was 12.6 to 105 

 hours. Swimming endurance increased with size 

 from about 5 km in the smallest individuals to 

 about 40 km in the largest (Fig. 4). Again, a linear 

 model provided the best fit for this relationship 

 (£;«rf. = 5.2SL-41.5, P<0.001, r- = 0.80, n=\2). and 

 the difference between the linear model and the 

 others was small (r- for the other models ranged 

 from 0.73 to 0.79). In contrast to critical speed, 

 the fit of the relationship between size and perfor- 

 mance was not improved by using TA or PA instead 

 of SL (r2=0.57 and 0.62, respectively). Endurance 

 also increased with age, but endurance data were 

 available only for larvae of two ages (21 and 24 

 days). Size was a better predictor of performance 

 than was age {End.= 6.6age-128, P=0.001, r^=().6Q, 

 n=\2). 



Vertical distribution 



There was not a significant overall relationship 

 between mean swimming depth and size (P=0.29, 

 r'' = 0.05). In spite of this, larvae of different sizes 

 differed in their depth selection behavior. The 

 smallest larvae (8-10 and 10-12 mm) swam pri- 

 marily between 4 and 10-12 m (Fig. 5, A and B). 

 One larva of each of these smallest size groups did, 

 however, swim deep monotonically. If one ignore 

 these two deep-swimming larvae, both smallest size 

 groups reached an overall mean depth of about 9 m 



