THRESHER ET AL.: LARVAE OF GADOID, MACRURONVS NOVAEZELANDIAE 



ment of the water column off the west coast to 

 depths of at least 100 m, i.e., virtually the entire 

 depth range occupied by larval M. novaezelandiae. 

 Hence, it is likely that the direction and speed of 

 larval transport vary, though still being predomi- 

 nantly southwards. Such variability is indicated by 

 our drift card data. Drift cards released at the mid- 

 shelf station of transect 5 on 22 July 1985 were 

 recovered inshore and south of the release point; 

 cards released at the same location 19 days later, 

 however, were mostly recovered north of the release 

 point (Fig. 7). Given the depth of the wind-driven 

 effects, it is likely that larval fishes present at that 

 site on the two dates would also have been advected 

 either south or north, depending on temporary con- 

 ditions of wind and current. 



Indeed, some larvae apparently develop wholly off 

 the west coast. In both years of the study, the range 

 in sizes and ages of larvae at transect 5, just north 

 of the spawning grounds, was nearly as wide as 

 those at all other transects combined. On this basis, 

 we suspect that some oceanographic feature on the 

 mid-west coast of Tasmania results in significant 

 retention of larvae in that area. One possibility is 

 that, as larvae are most abundant near shore, some 

 are trapped in relatively static pockets of water near 

 the coast and not entrained in the general souther- 

 ly current stream. Another possible retention mech- 

 anism is a coastal gyre, as yet unreported, that 

 perhaps forms in the winter off the west coast. In- 

 deed, our sea-surface temperature data consistent- 

 ly show a westward bend of surface isotherms im- 

 mediately offshore of transect 5, which could in- 

 dicate such a gyre. 



Whatever the retention mechanism, a conse- 

 quence is that larvae vary widely in the location at 

 which they undergo planktonic development. Such 

 variability is not trivial in M. novaezelandiae. Ap- 

 parent rates of larval growth in the species vary 

 significantly both with time and location: faster in 

 1984 than in 1985, faster in some months than in 

 others, and faster off the west coast for young lar- 

 vae and off the east coast for older larvae. There 

 are two ways these differences can be interpreted: 

 either the differences are real and reflect variability 

 in conditions that promote growth of larvae, or they 

 are only apparent, deriving not from variations in 

 grovrth rates, but from grovrth-dependent mortality 

 that varies in intensity in time and space. 



Testing these hypotheses directly in the field is 

 difficult. They can be tested indirectly, however, by 

 examining the distributions of residuals around the 

 population-mean grovi1;h trajectories. Consider three 

 possibilities: first, local differences in growth are 



real and are determined wholly by food availability; 

 second, local variation is real, but upper and lower 

 limits to growth are determined by physiological 

 constraints inherent in the metabolism of the lar- 

 vae; and third, real growth rates do not vary local- 

 ly, but appear to differ due to variably intense selec- 

 tion (predation) against slower growing larvae. The 

 first hypothesis (unconstrained growth) implies nor- 

 mal distributions of growth rates around population 

 means for both fast and slow growing populations; 

 the variance may alter with the mean, but skewness 

 should not. The second hypothesis (constrained 

 growth), however, implies distributions of growth 

 residuals will vary with mean growth rate: the dis- 

 tribution will be negatively skewed (to the left) when 

 mean growth rate is high (more individuals near the 

 maximum growth rate) and positively skewed (to the 

 right) when mean growth rate is low (more indivi- 

 duals near the minimum growth rate). The third 

 hypothesis (growth-dependent mortality) also im- 

 plies a relationship between the distribution of 

 growth residuals and apparent mean rates of 

 growth, but the relationship is opposite that implied 

 by the constrained grovHh hypothesis. If predators 

 selectively remove slow growing larvae, such mor- 

 tality will skew distributions of growth residuals to 

 the right. The greater the intensity of growth- 

 dependent predation (= the higher the apparent 

 mean growth rate), the more positive the skew. 

 Hence, the growrth-dependent mortality hypothesis 

 implies that when apparent mean growth rate is low, 

 the distribution of residuals should be normal or only 

 weakly positively skewed; when apparent mean 

 growth rate is high, the distribution should be 

 skewed strongly to the right. 



These predictions can be applied to field data for 

 M. novaezelandiae. The mean growth rate of larvae 

 was higher in 1984 than in 1985 and, for older lar- 

 vae, was higher on the south and east coast than 

 on the west coast (too few young larvae were caught 

 on the east coast to warrant a comparison for that 

 age group). The distributions of residuals for 1984 

 and 1985 are depicted in Figure 13, and those for 

 west and southeast coast populations of larvae older 

 than 25 d postfirst-feeding are depicted in Figure 

 14. The data are throughout consistent with the 

 constrained-growth-rate hypothesis. As predicted by 

 this hypothesis, the distribution of growth residuals 

 is skewed negatively, albeit weakly, in 1984 (^2 = 

 -0.35, t = 1.58, P < 0.1), and skewed positively, 

 also weakly, in 1985 {kz = 0.45, t = 1.36, P < 0.1). 

 Similarly, growth residuals for the relatively fast- 

 growing larvae caught off the south and east coasts 

 are distributed normally (kz = 0.25, t = 0.61, NS), 



43 



