Selection for birth date in North Sea haddock and its relation to maternal age


P. J. Wright, Fisheries Research Services, Marine Laboratory Aberdeen, PO Box 101, 375 Victoria Road, Aberdeen AB11 9DB, Scotland, UK. E-mail:


  • 1Birth date can be important to lifetime reproductive success. However, selection for birth date has rarely been addressed in fish, despite the opportunity provided by otolith microstructure.
  • 2This study examined the relationship between maternal age, spawning time and early survivorship in the North Sea haddock stock. Temporal changes in egg production were compared with the birth date distribution of progeny surviving to the demersal phase in 1994, 1996 and 1999, when the age structure of the spawning stock differed.
  • 3Estimates of intra-annual variation in stock egg production indicated that first-time spawning 2-year-olds began spawning much later than older age-classes.
  • 4The form and magnitude of selection on birth date varied between years, indicating that the production of multiple batches of eggs over an extended period has some adaptive significance to progeny survival.
  • 5Survivorship was consistently poor from the late spawning period when age 2 females contributed most to stock egg production. This persistent selection against late hatched offspring could reflect either low parental investment, as age 2 females produce smaller eggs, or the short length of the growing season prior to settlement.
  • 6Variability in birth date selection, particularly with respect to first vs. subsequent years of spawning, implies a strong selection pressure for a long reproductive lifespan. As such, reproductive potential in this and other exploited fish species with a similar reproductive trait may have been affected adversely by the general decline in repeat spawning females in recent years.


The early survival of many animals is often linked to the date on which they are born. Selection on birth date can be viewed as selection on parental breeding time, particularly in species that reproduce only once a year. The importance of birth or hatch date as a component of selection on breeding schedules is well documented in birds (Price, Kirkpatrick & Arnold 1988; Harris, Halley & Wanless 1992; Verhulst, van Balen & Tinbergen 1995) and mammals (Rutberg 1987; Green & Rothstein 1993), but less so in fish. While there is good evidence for temporal variability in mortality over the spawning season, with examples of selection for both early (Methot 1983; Cargnelli & Gross 1996) and late birth date (Crecco & Savoy 1987; Rice, Crowder & Holey 1987; Moksness & Fossum 1991; Wright & Bailey 1996; Pine & Allen 2001; Garvey, Herra & Leggett 2002) only one study has related birth date to lifetime fitness in a species of fish (Schultz 1993).

Spawning times in fish may be expected to effect progeny survival through synchrony with favourable environmental conditions on hatching (Cushing 1975; Cury & Roy 1989), as well as the period available to complete phases of development (Sinclair & Tremblay 1984). Hence, the duration of the spawning period may reflect how temporally predictable is the optimum environmental window for progeny survival (Cushing 1990). Uncertainty in the timing of optimal conditions may favour the production of many small batches over periods of time longer than the time scale of the environmental variation (Giesel 1976; Mertz & Myers 1994). Thus selection for protracted spawning can be seen as a trade-off between the mean and variance in offspring survival, because if individuals spawning larger and fewer egg batches also experience greater variance in reproductive success, then they may have lower fitness than individuals laying many smaller batches (Gillespie 1977; Stearns 1992). However, as protracted spawning in multiple spawning fish is a mechanism that enables fish to have a much higher fecundity than the restriction imposed by body cavity size (Garrod & Horwood 1984), it is not necessary to infer that the timing and duration of spawning has any significance to progeny survival.

Investigating selection on birth or hatch date in fish is possible because daily increments in their otoliths may allow accurate estimation of age from hatching (Campana 1992). In a study of egg production and larval survival of two year-classes of Coregonus hoyi, Rice et al. (1987) found that early spawned progeny experienced the highest mortality. Two studies following a single year-class of Lepomis macrochirus in the same population found different survival patterns, with either the early (Cargnelli & Gross 1996) or late (Garvey et al. 2002) hatched progeny having the lowest mortality. Schultz (1993) examined survival in one year-class of female dwarf perch, Micrometrus minimus, from birth through to reproductive success in their second breeding season. He found that early-born females were more successful in their first breeding season than late-born females (negative directional selection on birth date) but were less likely to survive the period between birth and first reproduction, relative to females born in the middle of the season. As a consequence of birth date and survival in the first year, overall selection on female birth date was stabilizing. Thus selection for birth date can be positive, negative or stabilizing. While these studies illustrate the potential for selection on birth date and hence reproductive timing, it is difficult to assess the long-term consequences to selection from the one or two year-classes studied. For example, studies of hatch date selection in birds have demonstrated that the form of selection can vary between years (Harris et al. 1992). Annual variation in birth date selection is likely to be important, as it would be difficult to reconcile any consistent discrete temporal selection with a protracted spawning period. Consequently, in order to address the effect of birth date fully it is desirable to examine survival patterns over a number of year-classes.

In several fish species, as individuals grow older they spawn progressively earlier and produce more batches over a longer period, thus extending their lifetime spawning duration (DeMartini & Fountain 1981; Parrish, Mallicoate & Klingbeil 1986; Lambert 1990). Consequently, the progeny of individuals spawning for the first time may therefore have fewer chances to coincide with favourable environmental conditions. At the population level, increases in the proportion of these first-time spawners could exacerbate the effect that short-term mortality episodes have on strength of the year-class. Correlations between the age composition of the spawning stock and recruitment lend circumstantial support to this view (Lambert 1990; Marteinsdottir & Thorarinsson 1998). Given the possible implications of such changes in spawning experience to recruitment potential and the increasing tendency for heavily exploited stocks to contain fewer and younger age-classes, there is now an increasing need to investigate the significance of age-related reproductive traits to year-class variability (Trippel, Kjesbu & Solemdal 1997).

The protracted spawning season of North Sea haddock, Melanogrammus aeglefinus (Linnaeus, 1758) is partially a consequence of repeat spawning of individuals (Hislop, Robb & Gauld 1978; Robb 1982). However, individuals are also likely to begin spawning at different times as the stock spawning period can extend for 12 weeks (Saville 1959; Hislop 1984). Female haddock can mature at age 2 and over the decade these young first-time spawning fish have tended to form a substantial component of the spawning stock (see ICES 2004).

In this investigation, we use the otolith birth date technique (Campana 1992) to explore the fitness consequences of reproductive timing in North Sea haddock. This is achieved by comparing the birth-date distributions of juvenile survivors following settlement with estimates of the temporal variation in egg production. For 3 years of contrasting stock age composition and year-class strength, otolith microstructure was used to estimate birth date of survivors while egg production was estimated by surveying the ovarian development of females collected at intervals over each annual spawning period. By this means we test the null hypothesis that survivors are a random subset of the annual egg production against the alternative, of selection for birth date. Finally we relate the temporal pattern of mortality with the age-specific differences in maternal spawning time in order to investigate whether maternal age has any effect on reproductive success.


temporal variation in egg production

For 3 years (1994, 1996 and 1999), adult haddock were obtained from commercial sources at 1–2-week intervals over the spawning season (February–May) from important spawning areas in the haddock's North Sea range (Saville 1959; Heath, Rankine & Cargill 1994; Fig. 1). Additional samples were obtained from the Scottish contribution to the ICES quarterly bottom trawl survey in the first and second quarters of the year. At least 150 adult haddock (> 17 cm) were sampled at 1–2-week intervals. Data were recorded on macroscopic maturity stage, total length (±1 cm), total and eviscerated wet weight (±1 g), age (from sagittal otoliths), liver weight (±1 g) and gonad weight (±1 g) were collected for all specimens (see Table 1). With the exception of samples containing mostly atretic oocytes, a minimum of 100 ovaries from each sample were examined histologically for estimation of fecundity, recent spawning and preovulatory atresia.

Figure 1.

Study area showing the main areas of commercial sampling (Scottish east coast; SEC and Northern North Sea; NNS) and the location of juvenile sample stations (▵ = 1994, ○ = 1996, + = 1999). Stippled grid = ICES demersal areas within the North Sea (see ICES 2004).

Table 1.  Summary of the number of adult maturity samples collected (length, age, weight, macroscopic stage), histological preparations and numbers per age-class by year upon which temporally resolved egg production estimates were based
YearCollectionsHistological preparationsFemales per age-class
199423 92957716822167574
199912 7454821196 395637

Within-season variations in egg production were estimated by first calculating annual egg production for the North Sea stock and then partitioning this estimate over time based on the field-derived estimates of spawning intensity. Annual production of eggs for an age-class (Ea) in the North Sea was estimated as:

image( eqn 1 )

where for a given age-class (a); N = number of fish, W = mean weight, M = proportion mature, F = relative fecundity (number of eggs × g × body weight) and R = proportion of females. Values for Na and Wa were obtained from the Working Group on the Assessment of the Demersal Stocks in the North Sea and Skagerrak (ICES 2004). A correction for total mortality, based on reported age-specific natural and fishing mortalities, was applied to the ICES assessed numbers at age to account for the difference in numbers between the assessment date (1 January) and spawning times. The values for Ma and Ra were derived from the field sampling programme together with data from the ICES first quarter Internatinal Bottom Trawl Surveys of the North Sea. Fa was estimated for 330, 1230 and 1848 females prior to the 1994, 1996 and 1999 spawning seasons, respectively, and corrected for the mean intensity of atresia during each spawning season. The stereological method of Emerson, GreerWalker & Witthames (1990) was used to derive estimates of relative fecundity and estimate preovulatory atresia.

In order to partition egg production over the breeding season temporal egg production per age-class (Ta) for a given time interval (i) in days was estimated from:

image( eqn 2 )

where for age-class a and time interval iBa,i = batch fecundity and Sa,i = proportion of females whose ovaries contained hydrated oocytes (i.e. spawning females) and n is the number of time intervals. Age-specific batch fecundity was determined using the stereological method of Emerson et al. (1990). Because there was no significant difference in age-specific batch fecundity between sample times (Kruskal–Wallis one-way analysis of variance P > 0·1) and to reduce the effect of outliers, a geometric mean value of B was used for each age-class. The proportion of females spawning on a given day was estimated using linear interpolation of values from consecutive sampling dates. Confidence limits (95%) for Sa on sample dates were derived using the proportions parameter test in s-plus 2000 (MathSoft). Values of Ta were summed for all ages to derive stock level egg production distributions.

validation of daily increments

Birth date analyses require that primary increments are formed daily from hatching over the life-stages of interest. Therefore the periodicity of primary increment formation was investigated in larvae reared from hatching and in field-caught pelagic juveniles held through to settlement. Larvae were reared in 20 L incubators while juveniles were kept in circular 2-m tanks. Two groups of larvae were sacrificed at 6 and 11 days after hatching. Marks were induced on the otoliths of juveniles by heating and cooling the water by 3 °C every 6 h over a 3-day period and then groups of fish were sacrificed at 16, 29 and 50 days post-mark. Experiments were conducted at ambient sea temperature and day-length, with ad libitum feeding.

birth-date distribution

Samples of demersal juvenile 0-group haddock were collected during the Scottish North Sea bottom trawl survey in August. This survey is so timed as to provide an early index for 0-group gadoids such as haddock (ICES 2004), as it coincides with the period shortly after settlement (Bailey 1975). Samples of between 150 and 400 fish were taken for age and size analysis from stations covering the main range of this species between latitudes 54°30′N−61°00′N and longitudes 2°00′W−5°00′E (Fig. 1). Fish were sampled in proportion to the ln total 0-group abundance of each haul. Lapilla otoliths were used to obtain increment counts using the method of Wright et al. (2002). Increment counts for 20% of juveniles were cross-checked between two experienced readers and between-reader differences were < 5% of counts. Hatch dates were determined from the number of otolith increments between the date of capture and the hatch check.

Data from different sample stations were pooled using a weighting factor based on the catch per unit effort relative to the total. The catch data were ln transformed to avoid excessive influence of one or two exceptional catches on the population composition. In order to compensate for the problem of cumulative mortality when estimating true hatch-date distributions a correction was applied to the total numbers of each age class collected (see Campana 1992). An annual mortality rate of 2·19, based on the MSVPA estimate of natural mortality, was used to correct for the effect of differential cumulative mortality (ICES 1997). The duration of egg development was estimated from a relationship for temperature specific egg development (Moksness & Selvik 1987) and monthly resolved mean sea surface temperature. Birth date distributions were then derived by subtracting the number of days of egg development from the hatch date. As within-year spatial differences in survival could confound stock level comparisons of egg production and survivorship, birth date compositions for different ICES demeral areas (see Fig. 1) within the study region were compared before pooling data.

Length at age relationships for demersal juveniles were compared between and within years for 20-day cohorts. Separate regressions were performed for these periods and the homogeneity of the regression lines (slopes and intercepts) was tested using analysis of covariance.

survivor analyses

The association between birth date and temporal egg production was described using a relative fitness functions f(z) estimated using the non-parametric technique of Schulter (1988), as modified by Anderson (1995), for cross-sectional data. The method uses cubic splines to estimate the form of selection acting on a quantitative trait and makes no assumptions of the underlying fitness function. The form of selection was calculated using the daily Ta to give the numbers in the ‘before-selection’ group (S1) and the birth-date compositions from the demersal survivors to give the numbers in the ‘after-selection’ group (S2).

Relative fitness (f) was estimated as:

image( eqn 3 )

where the conditional probability that a fish with birthdate z was caught in the sample of survivors, given that it was caught in either in the ‘before-selection’ or ‘after-selection’ group is:

image( eqn 4 )

Confidence bounds for f(z) were estimated with 100 bootstrapped replicates of the before and after selection samples (Anderson 1995). The relative fitness function has a mean of approximately one, weighted by the frequency of individuals in the sample taken ‘before’ selection. Positive and negative values of the index give an indication of whether a cohort was over- or under-represented relative to the cohort's egg production.

In order to consider absolute survival rates in relation to birth date, the number of 0-group juveniles at 1 July based on assessments (ICES 2004) were partitioned in time according to the annual birth-date composition of sampled juveniles. The proportion surviving was estimated from the ratio of the temporally resolved numbers of 0-group haddock and annual egg production. Survivorship over the spawning season was compared for periods prior to and during age 2 egg production.


Annual egg production was estimated to be 1·75 × 1013, 2·96 × 1013, 2·11 × 1013 in 1994, 1996 and 1999, respectively. Age-specific mortality in the 3 years were similar and reduced annual fecundity estimates to between 68·0 and 71·7% of the value excluding mortality. There were differences in the onset of spawning and significant interannual variation in temporal egg production (Ta) estimates between years (Kolmogorov–Smirnov two-sample test P < 0·001; Fig. 2). By far the most important feature of these distributions is the qualitative differences in the timing of egg production between first-time spawning 2-year-olds and older age-classes. Age 2 females began spawning 27–36 days later than older females, and as they comprised a significant proportion of the egg production in 1994 and 1996, they effected the temporal variation (Fig. 2). Hence while peak egg production occurred in March in 1999, most egg production occurred in April in 1994 and 1996. All age-classes finished spawning around the same time of year.

Figure 2.

Temporal changes in age-stratified egg production of North Sea haddock for the years 1994, 1996 and 1999.

Lapilli exhibited a distinct check on the day of hatching. The number of increments following either the hatch check or induced mark in juveniles was not significantly different from the number of days following marking (Table 2). Therefore, we can assume that primary increments in wild haddock otoliths will provide a reliable means of estimating hatch date.

Table 2.  Comparison between primary increment counts and known age for 0-group haddock validation experiments. P-values are based on one-sample t-tests, N = sample size. Days post-mark refers to day- from night-time hatching for larvae and days post-thermal mark for juveniles
StageDays post-markNumber of increments (mean + SD)NP
Larvae 6·5 6·69 (0·60)150·23
Larvae10·510·0 (0·71) 50·19
Juvenile1616·2 (0·67) 90·35
Juvenile2929·6 (1·95) 70·53
Juvenile5051 (1·91) 70·22

No significant differences were detected between the hatch date compositions of survivors between ICES demersal areas in North Sea in any of the 3 years of study (Kolmogorov–Smirnov two sample test P > 0·1). Hence, the estimated annual birth date distributions should be representative of the entire juvenile distribution. Due to the relatively low mortality rate, correction for cumulative mortality affected the resulting birth date distribution by only < 7%. As with egg production, the cumulative frequency distributions of population birth dates differed significantly between years, especially between 1999 and other years (Kolmogorov–Smirnov P < 0·001). In each year, the range in birth date distributions fell within the range of egg production (Fig. 3). However, while spawning extended into May in all years, no survivors originated from that month and the annual birth-date frequency compositions were significantly different from the egg production curves (Fig. 3; Kolmogorov–Smirnov < 0·001).

Figure 3.

Proportion frequency distribution of demersal juvenile birthdates from the 1994, 1996 and 1999 year-classes (black line). The corresponding proportion egg production for all age-classes of females is added for comparison (grey line; based on Fig. 2).

Estimates of the form and magnitude of birth date selection between the egg and demersal phases, based on relative fitness functions, are presented in Fig. 4. In all years, there were significant departures from a relative fitness of 1 during part of the spawning period, as indicated by the bootstrapped confidence intervals. In all years, the form of selection had both directional and optimizing components such that fitness was highest for juveniles with intermediate birth dates. However, low relative survival values in the initial part of the spawning season result from a very small number of observations and this is reflected in the wide confidence bounds (Fig. 4). In 1994, a peak in fitness was evident around the mid-point (day 90) of the spawning season. In 1996 and 1999 peaks in spawning did not correspond with peaks in birth-date selection and there was a positive skew to all but the earliest birth dates. Hence, with the exception of the initial 20 days of spawning in 1996 and 1999 the form of selection was negatively directional.

Figure 4.

Relative fitness ± 95% confidence band between egg production and birth date of demersal juveniles from the 1994, 1996 and 1999 year-classes. The confidence interval is based on 100 bootstrap replications and its width varies with the local density of data points.

The 1994, 1996 and 1999 year-classes examined in the present study, respectively, correspond to 1·4, 0·6 and 3·3 times the average size of recruitment for the period between 1970 and 2000 (ICES 2004). The highest and lowest proportion of 2-year-old females in the spawning stock corresponded to the lowest and highest survival rates in the 3 years examined (Table 3). Within years the contribution of age 2 to egg production increased over time and survival in the second half of age 2 egg production was much lower than annual survivorship in all 3 years. Further survivorship in the period prior to age 2 spawning in 1996 and 1999 was higher than during the first half of age 2 egg production but not in 1994. Uncertainty in the estimates of proportion of age 2 spawning did not affect this conclusion because there was still poor survival in the late spawning period even when the upper 95% confidence limits of spawning proportion were used to define the periods of age 2 spawning.

Table 3.  Survival estimates prior to and during the age 2 spawning period. The age 2 spawning period is divided into two halves and the percentage of eggs produced by this age-class relative to the total is given. Estimates of survival to 1 July are based on the abundance of 0-group juveniles as a proportion of the number of eggs produced for the corresponding time period. Values in parentheses refer to survival estimates recalculated for the dates of age 2 spawning based on upper and lower 95% confidence limits of proportion spawning
YearAge 2 contribution to egg productionProportion surviving to 1 July (× 10−3)
AnnualEarlyLateAnnualPre-age 2 spawningEarly age 2 spawningLate age 2 spawning
19944656613·03 2·18 (1·89–2·42)5·62 (5·21–6·05)1·12 (0·36–1·64)
19965458820·71 1·13 (0·64–1·14)0·82 (0·98–1·10)0·26 (0·05–0·14)
19992337515·8711·53 (9·97–22·84)1·74 (1·60–1·76)0·02 (0·01–0·16)

Linear regressions provided the best fit to the annual total length at age relationships. The average daily somatic growth determined from the regression of total length against increment number was 0·85, 0·81 and 0·89 mmday−1 for the 3 years, respectively (Fig. 5). The slopes did not differ significantly between years (ancovaF2,799 = 0·43; P = 0·65), although there was a very high residual variation in the length – age relationships. Similarly, there was no evidence for differences in the length–age relationship within years for successive 20-day cohorts of juveniles (ancova, intercepts: slopes: 1994: P = 0·99; 1996: P = 0·07; 1999: P = 0·29). As only demersal growth was considered, this finding does not preclude cohort specific differences in growth prior to settlement.

Figure 5.

Relationship between total length and age for demersal juveniles presented for the 1994, 1996 and 1999 year-classes. Symbols represent consecutive 20-day cohorts (◊ = 20–39; ▿ = 40–59; ○ = 60–79; = 80–99; □ = 100–119). Linear fits to data are also shown.


Survivorship to settlement in haddock was related to birth date and hence the null hypothesis that survivors are a random subset of annual egg production can be rejected. The presence of selection for birth date is consistent with the findings of four other studies comparing fish reproduction with the survivorship of progeny (Rice et al. 1987; Schultz 1993; Cargnelli & Gross 1996; Garvey et al. 2002). However, by examining the fate of three year-classes, the present study has demonstrated that the form and magnitude of selection for birth date can vary between years, as has been found in studies of birds (e.g. Harris et al. 1992). The evidence for stabilizing selection in all years is generally consistent with that found for a single year-class of M. minimus (Schultz 1993). However, the positive skew in 1996 and 1999, suggesting negative directional selection for all but the earlier spawned progeny, was similar to that reported for a single year-class of L. macrochirus (Cargnelli & Gross 1996). The difference in birth-date selection between two year-classes of the same population of L. macrochirus highlights that studies of a single year-class can lead to a false perception of long-term selection pressures (Cargnelli & Gross 1996; Garvey et al. 2002). The large intra- and interannual variation in offspring survival rates found in the present study indicate that the optimal conditions for survivorship are not predictable in time. This temporal variability in offspring survival would therefore be expected to produce a selection pressure for a high number of spawnings per lifetime, i.e. a longer reproductive lifespan (Stearns 1992). Consequently, the present study supports the view that the production of multiple batches of eggs over an extended period has some adaptive significance to progeny survival (Giesel 1976; Gillespie 1977; Mertz & Myers 1994) and is not just a mechanism to increase fecundity above the limitations imposed by the body cavity.

While there was positive selection for either early or mid- spawning dates in the 3 years there was a consistent negative selection for very late spawned progeny. As the majority of late egg production in the present study resulted from the contribution of age 2 females, poor survival during this period may be related to some characteristic of this age-class. The eggs produced by 2-year-old haddock are smaller than those produced by older age-classes (Hislop 1988) and larvae from large gadoid eggs and more experienced adults may be more viable (Solemdal, Kjesbu & Fonn 1995; Marteinsdottir & Steinarsson 1998). Hence, selection on late birth date may be the result of low offspring viability of age 2 females rather than an effect related to temporal fluctuations in environmental conditions. A similar relation between parental investment and birth-date selection has been indicated in M. minimus (Schultz et al. 1991) and has been suggested to explain the often persistent selection for early-hatched offspring found in many species of birds (Price et al. 1988). However, manipulation experiments that have attempted to distinguish the relative contribution of egg quality and environmental timing on reproductive success in birds have indicated that the later factor can play a more significant role (e.g. Verhulst et al. 1995). Clearly, without some means of identifying the progeny of age 2 haddock we are unable to consider the relative contribution of offspring viability vs. environment. Nevertheless, whatever the cause, it can be inferred that age 2 female haddock will have very low reproductive success in some years. Lower reproductive success in the first year of spawning implies that repeated spawning seasons as well the number of spawnings per lifetime and fecundity can be important to an individual's reproductive success.

The approach to assessing phenotypic selection used in this study relies on unbiased sampling of the same population at the egg stage and after settlement (Sogard 1997). Potentially the apparent selection for old individuals could have been due to age-related differences in either movement; distribution or settlement time as opposed to selective mortality on birth date. However, these potential sources of bias are unlikely in the present study as samples were taken throughout the distributional range of demersal juvenile haddock and birth-date distributions were similar over much of the area sampled. In addition to sampling bias, the estimation of relative fitness requires accurate estimates of total egg production and population birth date. The estimated distributions of egg production were largely a function of the difference in the onset of spawning between 2- and 3 + age females. Consequently, the temporal distribution of egg production will reflect closely the composition of these age-classes. Relative numbers at age in the stock are likely to be well estimated by recent assessments because the fishery accounts for the majority of mortality within the study range, and estimates of even the youngest parental year-class (1997) will have been made over several years (ICES 2004). Importantly, the pattern of selection is robust to extreme errors in the estimation of spawning proportions, as indicated from the comparison between survival estimates and 95% confidence limits of egg production.

While the selection for early- or mid-hatched juveniles found in the present study is similar to that for some other studies of juveniles (Methot 1983; Cargnelli & Gross 1996) most studies of temperate fish species sampled between the hatching and larval metamorphosis period have found selection for late-hatched individuals (Crecco & Savoy 1985; Rice et al. 1987; Moksness & Fossum 1991; Wright & Bailey 1996; Hare & Cowen 1997; Pine & Allen 2001). Schultz (1993) suggested that the different forms of selection reported by studies of birth date are related to the stage of development considered. In most spring-spawning temperate species, factors favouring the late production of offspring act early in life, while those favouring the early production of offspring act later in life (Schultz 1993; Anderson 1995; Miller 1997). For example, early-spawned cohorts will probably experience higher mortality rates than late-spawned fish due to unfavourable temperatures (Rutherford & Houde 1995; Secor & Houde 1995; Pine & Allen 2001), mismatch with plankton production (Cushing 1990) or reduced growth rates and an extended stage duration due to colder temperatures (Rice et al. 1987). However, differences in traits that are correlated phenotypically with birth date may contribute indirectly to total selection on birth date. For example, early hatching maximizes the time for growth (Folkvord, Øiestad & Kvenseth 1994; Cargnelli & Gross 1996), which may be important where there is a selective advantage for the early timing of a life-history event that is size-dependent, such as metamorphosis (Sinclair & Tremblay 1984), age at first maturity (Schultz 1993) or overwintering (Cargnelli & Gross 1996). Consequently, if attaining a specific size is important there must be a trade-off between size, growth rate and hatch date. In the present study, the larger-sized individuals sampled as demersal juveniles did tend to be early-spawned individuals, suggesting that a longer growing season may confer an advantage in some years. Young haddock may have been exposed to a number of potential size-selective pressures including size-dependent predation pressure and competition for space around the time of settlement (Cook & Armstrong 1986; Lough et al. 1989). Consequently, selection for older juveniles could have arisen because they settled first or were better able to compete for space than smaller conspecifics. Clearly, further work is needed to elucidate if and when size selective process act to indirectly select for birth date.

Current management advice regarding the reproductive potential of a stock is based only on the spawning stock biomass, a proxy for fecundity. However, the present study demonstrates that reproductive traits other than fecundity can also be important to the reproductive potential of a stock. As such, the present study provides empirical support for the paradigm that age and size structure of the spawning stock should be considered in fisheries management advice (Lambert 1990; Trippel et al. 1997). Management measures intended to increase the proportion of older, repeat spawners may help to reduce the probability of recruitment failure when the spawning stock is at low levels. This may also be relevant for many other commercial species because protracted and age-specific differences in spawning are relatively common traits (DeMartini & Fountain 1981; Parrish et al. 1986; Lambert 1990; Marteinsdottir & Thorarinsson 1998). Given the importance of age and size, future revisions of precautionary biomass limits would be advised to consider stock age composition when analysing stock-recruitment time series. Further, in setting precautionary limits based on historic data sets it may be important, first, to distinguish the contribution of first-time and repeat spawners.


This work was funded by The Scottish Executive Environment and Rural Affairs Department MF0460 and MF0462 and contracts FAIR-CT95-0084 and FAIR-CT98-4122 from the Commission of the European Communities. We would like to thank I. Gibb, D. Mennie, G. Strugnell, W. MacDonald and M. Bell for assistance in this project as well as the masters and crews of FRVs Scotia and Clupea, RRS Challenger and MFVs Sunbeam and Falcon for help in sample collection. Welcome improvements were suggested by F. Neat and an anonymous referee.