198 



Fishery Bulletin 88(1). 1990 



suggests that, in Bristol Bay, herring larvae are re- 

 tained in the nearshore zone during the larval period 

 to experience a temperature regime similar to that 

 observed in the basin. Herring larvae transported out 

 of coastal waters would develop in much colder water 

 and exhibit a reduced rate of growth. 



The mortality of larvae in the basin was high during 

 the first 2 weeks of the experiment, with daily mor- 

 tality estimated to be as high as 14%. This is about one- 

 half that estimated for a wild population observed over 

 a 20-day period in which daily mortality was 32% 

 (Stevenson 1962). The higher survival in the basin is 

 consistent with the results of Moksness and 0iestad 

 (1987), who estimated daily mortality to be higher (4%) 

 for the first 30 days and then much reduced (0.8%) to 

 day-100 for Atlantic herring in the same basin. 



The higher initial mortality of Pacific herring in the 

 basin may be due to predation. Hydromedusae known 

 to consume herring were abundant during the first 

 week Pacific herring were in the basin. In the experi- 

 ment with Atlantic herring, the occurrence of hydro- 

 medusae was much later, beyond day-40. Aral and Hay 

 (1982) estimated that hydromedusae predation in Bri- 

 tish Columbia could account for a daily larval loss 

 of 9%. 



Starvation may not have been a significant source 

 of mortality because the abundance of copepod larvae, 

 the preferred food of herring larvae, was much higher 

 at the time of first feeding of Pacific herring than was 

 estimated at the same point for Atlantic herring 

 (Moksness and (Diestad 1987). Ki0rboe et al. (1985) 

 report that herring are able to successfully initiate 

 feeding at prey densities much lower than usually pres- 

 ent in the sea at the time of first feeding. 



Cannibalism may have been an important source of 

 mortality in late-stage larvae. Blaxter and Hunter 

 (1982) report that cannibalism has been observed in 

 situations in which a size hierarchy has developed, 

 although they believe this is a rare occurrence based 

 on the scarcity of reports of cannibalism in aquarium 

 studies. However, the observation of attacks on smaller 

 herring in this study suggests that cannibalism was not 

 a rare event in the basin population. Cannibalism may 

 have been related to the reduced abundance of zoo- 

 plankton at the time it was observed. However, fish 

 larvae may be a normal component of the diet of post- 

 larval herring since capelin Mallotus villosua larvae 

 were rapidly consumed by newly metamorphosed her- 

 ring in an earlier basin study (Moksness and 0iestad 

 1987). Also, Hourston et al. (1981) reported that age-1 

 Pacific herring, 71-91 mm standard length, readily con- 

 sumed larval herring when placed together in a 197-L 

 tank. 



An interesting aspect of herring growth in the basin 

 was the development of three distinctive length modes 



Figure 8 



Length-frequency distribution of Pacific herring from Bristol Bay. 

 Alaska, in the 2000-m'' basin at Fl0(levigen Biological Station, Nor- 

 way, at hatching, week-3, and study termination on day-63. 



from an essentially unimodal population at hatching 

 (Fig. 8). The cause of distinct size modes in this experi- 

 ment is unknown, but it may be related to success at 

 first feeding or be genetically based, because size dif- 

 ferences developed early. 



Blaxter and Hunter (1982) report that strong size 

 hierarchy has been observed in some aquarium studies 

 of herring, which they speculate may involve crowding 

 or food competition; however, in a previous study of 

 Atlantic herring larvae in a 4400 nr'' basin a size hier- 

 archy did not develop (0iestad and Moksness 1981). 

 McGurk (1984) found that length and weight of larvae 

 of the same hatch differed at 30 days posthatch as a 

 result of withholding food from to 14 days. This sug- 

 gests that modes may be a reflection of feeding suc- 

 cess. Moksness et al. (1989) observed that in ocean 

 catfish differences in feeding success at first feeding 

 was expressed as differential growth that persisted 

 throughout the larval period. 



Although most of the literature suggests that differ- 

 ential growth is related to early feeding success genetic 

 factors may also influence larval growth and survival. 

 Christopher et al. (1988), in a study of capelin Multotus 

 viUosus larvae from known females crossed with a 

 single male, found that the amount of yolk varied 

 among the females and that growth and survival was 

 correlated with yolk quantity. 



Whatever the cause, one would tend to conclude that 

 the modes were distinct "cohorts" from different 

 hatchings if observed in the wild. The significance of 

 these results is that a method such as otolith daily ring 

 counts is required to validate larval age and the use 

 of length-frequency analysis to determine the occur- 

 rence of cohorts may lead to erroneous assumptions. 

 If differential growth among a brood is a common 



