132 
Fishery Bulletin 95 ( I ), 1997 
Fall 1984 
Spring 1985 
Fall 1985 
Spring 1986 
Length (cm) 
Figure 4 
Length-frequency distributions of the 1982 spring-hatched cohort from the fall of 1984 to the spring of 1986, showing the relative 
proportions of immature herring that eventually adopted either the spring- or the autumn-spawning season as determined by 
their maturity stages. 
same pattern is evident for the 1986 autumn-hatched 
cohort, which also showed a large bimodal length- 
frequency distribution at age 3 (Fig. 3C). 
The significance of these differences in length com- 
position is shown by following these cohorts as they 
became mature and began to spawn. We observed 
that when the length-frequency distribution of the 
juvenile spring-hatched herring was unimodal and 
had a relatively small variance (1982 cohort), the 
majority of them spawned in the spring of 1986 at 
age 4 (Fig. 4). Conversely, when the juvenile length- 
frequency distribution was bimodal and had a rela- 
tively large variance ( 1980 cohort), there was an al- 
most even split between those that became spring 
spawners and those that eventually became autumn 
spawners (Fig. 5). In addition, if one follows the 1980 
spring-hatched cohort after maturity, those that be- 
came spring spawners were significantly smaller la- 
test: P<0.0001, SAS Institute Inc., 1985) than those 
that became autumn spawners (Fig. 6A). This length 
difference was sustained throughout their early adult 
life, i.e. from age 3 through age 6. A similar pattern 
was also seen with the 1979 autumn-hatched cohort. 
Those that became spring-spawners were smaller at 
age (Fig. 6B), although the differences are not sig- 
nificant for certain ages owing to small sample sizes. 
Discussion 
Einarsson (1952 ) hypothesized that year-class par- 
allelism (twinning) in sympatric seasonal-spawning 
herring populations came about through a correla- 
tion between larval survival conditions in the fall of 
one year with conditions in the spring of the follow- 
ing year, assuming the larval stage to be the critical 
phase that determines year-class strength. However, 
given the ontogeny of the larvae of sympatric sea- 
sonal-spawning populations, it is difficult to conceive 
of a single mechanism by which both cohorts would 
experience similar survival conditions over a 10- 
month period, especially since twinning does not 
normally occur with two successive year classes 
within the same year. If one follows the development 
