Spatio-temporal dynamics of alternative male phenotypes in coho salmon populations in response to ocean environment



    1. Field Science Center for Northern Biosphere, Hokkaido University, N9 W9, Sapporo 060–0809, Japan; and
    Search for more papers by this author

    1. Department of Fisheries and Wildlife and Coastal Oregon Marine Experimental Station, Oregon State University, Hatfield Marine Science Center, Newport, OR 97365, USA
    Search for more papers by this author
    • Current address: Ocean Sciences Centre, Memorial University of Newfoundland, St John's, NL, A1C 5S7, Canada.

and current address: Yusuke Koseki, Ocean Sciences Centre, Memorial University of Newfoundland, St John's, NL, A1C 5S7, Canada. E-mail:


  • 1The coexistence of alternative reproductive phenotypes will probably be shaped by spatial and temporal variability in the environment. However, the effects of such variability on coexistence and the scale at which it operates are seldom understood.
  • 2To quantify such effects, we examined spatial and temporal dynamics in the abundance and frequency of alternative phenotypes of male coho salmon, Oncorhynchus kisutch Walbaum, which mature as either large fighters (age-3 ‘hooknoses’) or small sneakers (age-2 ‘jacks’). Using over 20 years of data on coded-wire tagged fish released from nine Oregon hatcheries, we tested for the effects of ocean environment independent of those due to freshwater rearing.
  • 3Annual fluctuations of the abundance of jack and hooknose males within populations were correlated strongly by brood year (cohort) but not by return year (breeding group). This occurred independently of significant effects of release practice (i.e. the number of fish released, body size at release and date of release), indicating that a synchronized fluctuation in mortality during the first year at sea was the predominant cause. As a result, the annual frequency of the alternative phenotypes at breeding varied considerably within populations.
  • 4Spatial patterns in the annual fluctuations of the two phenotypes were similar (i.e. synchronous among populations), except that jacks showed local spatial structure (decreased synchrony with distance) not evident among hooknoses. This suggests that oceanic processes affecting the two phenotypes operate at different spatial scales. Despite effects on salmon abundance, the ocean environment had little influence through its effects on salmon growth on the relative frequencies of the alternative phenotypes within and among populations.
  • 5The results provide insight into the evolutionary dynamics of alternative phenotypes, including an intragenerational time lag that increases annual variability in phenotype frequencies at breeding (return years) and the significance of local freshwater processes, rather than oceanic processes, on phenotype expression. Freshwater processes, such as juvenile growth, timing of migration and breeding competition, operating at evolutionary and intragenerational time-scales, are probably the predominant forces affecting phenotype frequency.


The maintenance of discontinuous phenotypical variation within populations is a focus of considerable interest in evolutionary ecology (Gadgil 1972; Gross 1996; Shuster & Wade 2003). It has also recently attracted the attention of applied fields, such as fisheries management and conservation, where its role in population sustainability is being recognized (Gross 1991; Hutchings 2002; Young 2004). Such polymorphisms are associated frequently with alternative reproductive behaviours, particularly among males, and represent alternative solutions for attaining reproductive success (e.g. sneaking vs. fighting). Alternative reproductive phenotypes have been reported in most major taxa, and have evolved primarily in species where the variance in male mating success is high and thus sexual selection is strong (Gadgil 1972; Gross 1996; Shuster & Wade 2003). Theoretical models have been developed to explain the genetic and ecological mechanisms for the coexistence of alternative phenotypes (Lively 1986; Hazel, Smock & Johnson 1990; Hazel, Smock & Lively 2004) and empirical data have accumulated in line with the theoretical frameworks (for review, see Gross 1996; Roff 1996; Brockman 2001; Shuster & Wade 2003).

Despite the theoretical possibility that alternative phenotypes may be genetically determined strategies with equal fitnesses maintained by frequency dependent selection, this situation appears to be rare (Gross 1996). The predominant view holds that alternative phenotypes are, in most cases, the product of adaptive phenotypical plasticity (i.e. conditional strategy), where depending on the environmental and demographic conditions (and ultimately competitive status), individuals adopt a phenotype that yields the greatest fitness (Gross 1996; Roff 1996; Hazel et al. 2004). Recent empirical studies have focused on demonstrating differing fitness functions for alternative phenotypes along a gradient of competitive status (e.g. body size; Hunt & Simmons 2001; Forslund 2003). Moreover, attempts have been made to associate spatial variation in the frequency distribution of phenotypes among populations with environmental and demographic variables, such as population size and density (Moczek 2003; Tomkins & Brown 2004). For example, Tomkins & Brown (2004) showed that among-population variation in forceps dimorphism of the European earwig, Forficula auricularia, was related to population density, with the proportion of males with long forceps increasing with density. These studies suggest that the population state and perhaps temporal dynamics of alternative phenotypes are influenced strongly by environmental and demographic conditions, as formalized by theoretical studies (Lively 1986; Hazel et al. 1990, 2004). Moreover, recent game models show that environmental and demographic stochasticity can promote the coexistence of alternative phenotypes (Clark & Yoshimura 1993; Calsbeek et al. 2002). However, no empirical study, to our knowledge, has investigated temporal variation of alternative phenotypes within populations or processes underlying it.

Like many other salmonid species, the semelparous coho salmon, Oncorhynchus kisutch Walbaum, exhibit two alternative male reproductive phenotypes, hooknoses and jacks (reviewed in Fleming 1998). Hooknose males have well-developed secondary sexual characters, including elongated jaws that they use in fighting for access to females on the freshwater breeding grounds. Like females, these males mature typically at 3 years of age, rearing for 18 months in freshwater then migrating to the ocean for the vast majority of their body growth, before returning to their natal rivers to breed. By contrast, jack males, which mature at 2 years of age (i.e. after only 6 months in the ocean) and show little development of secondary sexual characters, use their considerably smaller body sizes to sneak access to breeding opportunities. Such precocious maturation in salmonids is adopted by juvenile males that appear to have the highest status (e.g. large body size, rapid growth and large lipid stores; Utoh 1976; Rowe, Thorpe & Shanks 1991; Silverstein et al. 1998), to which both genetic and environmental factors may contribute (Iwamoto, Alexander & Hershberger 1984; Glebe & Saunders 1986; Heath et al. 1994).

It has been suggested that the frequency of alternative phenotypes in salmonids should be sensitive to changes in habitat conditions (Gross 1991; Hutchings & Myers 1994; but see Koseki et al. 2002). Demonstrating the dynamics of alternative phenotypes associated with environmental variation, however, has received less empirical attention (but see Young 1999; Healey, Henderson & Burgetz 2000). Using spawning survey data, Young (1999) showed that the proportion of O. kisutch jacks varied within and among streams from two neighbouring river basins, and that the among-stream variation was related to variation in local demographic (i.e. density) and environmental conditions (stream gradient, elevation and distance to the bay). However, the dynamics of alternative phenotypes, particularly as affected by the ocean environment, have not been characterized in time, nor have the spatial patterns of temporal dynamics been described at large scales.

Here, we examined temporal fluctuations in the frequency of alternative phenotypes of O. kisutch and its spatial patterning. Our focus was on patterns induced by oceanic processes rather than on the underlying factors affecting aspects such as marine survival, which has been well documented (e.g. Mantua et al. 1997; Coronado & Hilborn 1998; Hobday & Boehlert 2001; Koslow, Hobday & Boehlert 2002). For this purpose, we analysed hatchery-released fish (i.e. fish reared artificially during their freshwater life and released to undertake their ocean life naturally), where records of release practice including body size at release and date of release were available to allow us to separate statistically freshwater and marine influences. Moreover, because of their hatchery origin, it was reasonable to assume that selective forces targeting alternative phenotypes (e.g. fisheries and hatchery breeding techniques) has either remained relatively unchanged or changed in a similar way across populations during the temporal period of the study. Our main objectives were to test whether (1) temporal fluctuations in the abundance of jack and hooknose returns as driven by ocean conditions are related within populations, (2) fluctuations of the two phenotypes result in temporal patterning in the ratio of alternative phenotypes, (3) synchrony exists among populations in the temporal abundance and ratio of alternative phenotypes and (4) phenotype frequencies are affected by early marine growth.

Materials and methods


The data were derived from the coded-wire tag (CWT) database of hatchery fish, available from the Pacific States Marine Fisheries Commission ( Only data from non-experimental releases of yearling fish were used. For each tag code (release group) from hatcheries in Oregon over the brood years 1976–97, we extracted data on brood year and number of fish released, as well as release date (mean Julian day) and body size at release (mean weight, g; Table 1), which are known to affect returns of jacks and age-3 fish (hooknose males and females) (Bilton, Alderdice & Schnute 1982; Mathews & Ishida 1989). The returns (catch plus escapement to hatchery) of jacks and age-3 fish were calculated for each release group, using the numbers of fish estimated from equations that account for different sampling efficiencies among fisheries (for details see Johnson 1990). Because age-3 returns were not separated by sex in most catches, the catch of hooknose males was calculated using the sex ratio in the corresponding hatchery escapement. Early marine growth (first 6 months) was calculated as the mean length of jacks at return minus that of juveniles at release. Because Oregon hatcheries measure weight at release only, length was calculated using an empirical function derived from CWT data for Washington hatchery fish in brood years 1984–97 [log(weight) = −8·6 +2·4log(length), n = 952, adjusted R2 = 0·75, P < 0·001]. In the end, the data set used to examine spatio-temporal dynamics consisted of 347 release groups from the nine hatcheries having the most complete time-series of data (Fig. 1).

Table 1.  Summary of numerical data for coded-wire tagged (CWT) O. kisutch releases (n = 347) from nine Oregon hatcheries in the brood years 1976–97
DataMedianQuartile range
  • a

    n = 339 (eight releases had no returns).

Release date (Julian day)118101–126
Mean weight at release (g)3431–40
Number of males released25 91824 682–27 001
Jack returns
 Number of fish51–20
 % returns0·0230·0037–0·076
Hooknose return
 Number of fish11738–317
 % returns0·490·17–1·16
Jack : hooknose ratioa0·0420·005–0·155
Figure 1.

Location of the nine populations of O. kisutch in Oregon.

models for the effects of release programme

Variability in the jack and hooknose returns in relation to release practice was examined using a generalized linear model (GLM) with hatchery and brood year as nominal predictor variables, release date and body size as numerical predictors and the natural logarithm of the number of juvenile males released (i.e. half the total number of juveniles assuming an equal sex ratio; see Heath et al. 1994) as an offset (i.e. control for the size of the release; for a full explanation of this and other GLM terminology below, see McCullagh & Nelder 1989). Preliminary analysis of the Poisson GLM with a logarithm link function, which is a standard model for count data, indicated strong overdispersion (i.e. an extremely high residual deviance compared to the degrees of freedom). To adjust for this, a quasi-Poisson approach was applied and F-tests rather than χ2 tests used for model selection (McCullagh & Nelder 1989). Forward stepwise selection, where explanatory variables were added sequentially (significance level of 0·05), was used to find a parsimonious subset of variables to be included in a model. A similar approach was used in the analysis of the ratio of alternative phenotypes (jack : hooknose ratio), except the offset was excluded. The GLM for the jack : hooknose ratio was a quasi-binomial model with a logit link function:


where µ is the proportion of jacks and therefore 1 − µ the proportion of hooknoses in a release group. Eight release groups having no male returns were excluded from the analysis.

spatial and temporal analyses of dynamics

The deviance residual from the GLM, which accounts for size at, and date of release (see Results), was used in the analyses of spatio-temporal dynamics unless stated otherwise because our focus was on variation due to oceanic processes independent of freshwater processes. It was averaged for each brood year and hatchery combination, resulting in a set of nine population series over 22 brood years, with 2–7 missing data (years) in each population due to the absence of releases those years (summarized in Table 2).

Table 2.  Population series of the return abundance and ratio of O. kisutch male alternative phenotypes for nine Oregon hatchery populations during the brood years 1976–97. For each phenotype and ratio, series are based on deviance residuals from the GLM, which accounts for release practices (see Table 3). Positive (negative) mean values indicate that, on average, the abundance or ratio of phenotypes within that series was higher (lower) than expected, given the conditions at release
HatcheryYearsJackHooknoseJack : hooknose ratio
  • *

    For the ratio series, one year is missing because of the occurrence of zero returns.

Eagle Creek15−1·342·230·9416·43−0·742·30
Big Creek19/18*−1·002·91−3·2514·56−0·132·01
Salmon River200·718·5910·01 8·093·495·51
Fall Creek15−3·663·05−0·5718·02−3·293·11
Cole Rivers162·278·681·8616·772·348·22

A third-order polynomial function was used to describe the long-term fluctuations in jack and hooknose returns and jack : hooknose ratio for each population, taking into account apparent non-linear trends shown by visual inspection of the series. The relation between temporal fluctuations in jack and hooknose returns within a population was tested by cross-correlation using Pearson's correlation coefficient. To assess whether such correlations resulted from year-to-year fluctuations independent of long-term trends, the series were detrended by taking residuals from the third-order polynomials and re-examining the cross-correlation. This was conducted independently of the significance of the polynomials to ensure that potential trends did not obscure patterns. The cross-correlation analyses were also performed with and without a lag of 1 year to account for the additional year hooknoses spend in the ocean compared to jacks. That is, hooknoses males of a given brood year (cohort) experience ocean environments in common with jacks of the same cohort (in the first ocean year) as well as those of the subsequent cohort (in return year).

For each series of returns and ratios, analysis of spatial structure in temporal variation was performed on the original data (i.e. a series of annual means of deviance residuals), as well as the trend (predicted values from the third-order polynomial) and detrended data (the original minus trend). This was performed to highlight the different patterns of long- and short-term fluctuations. The mean correlation coefficient among the series was calculated to quantify the spatial synchrony across the region (lower Columbia River–coastal Oregon):


where n is the total number of populations (i.e. n = 9) and ρi,  j is the Pearson's correlation coefficient between the series for population i and j. Because of non-independence of the correlation coefficients (i.e. spatial autocorrelation), a 95% bootstrap confidence interval for the mean correlation was estimated using 1000 iterations of sampling with replacement among the populations and recomputing of the mean correlation (Bjørnstad, Ims & Lambin 1999).

For jack and hooknose returns, which showed region-wide synchrony (see Results), a Mantel test was performed to assess whether the degree of synchrony between populations was related to the geographical distance (Koenig 1999). The test involves measuring the similarity (i.e. Mantel correlation) between two matrices (Manly 1997). In our analysis, the first matrix consisted of pairwise geographical distances (river/ocean km) among hatcheries and the second, pairwise correlation coefficients among the time-series. Again, because of spatial autocorrelation, statistical significance for the Mantel correlation was evaluated by a randomization test, where the observed correlation was compared to the empirical distribution based on 10 000 permutations of the geographical matrix and recomputations of the Mantel correlation.


the effects of release programme

The abundance of returning jacks was best explained by a model that included the positive effect of body size at release (t = 7·73, P < 0·001) and negative effect of release date (i.e. earlier release in the year was favoured; t = −5·12, P < 0·001; Table 3). Inclusion of the interaction term or quadratic terms did not improve the model significantly. Similarly, a model containing body size at release (t = 2·40, P = 0·017) and release date (t = 6·39, P < 0·001) explained hooknose abundance most effectively. However, in contrast to jacks, releases late in the year were favoured. Consistent with the above patterns, the jack : hooknose males ratio was explained best by large size at release (t = 8·13, P < 0·001) and release earlier in the year (t = −12·2, P < 0·001).

Table 3.  Generalized linear model (GLM) parameter estimates for the effects of body size at release and release date on the return abundance and ratio of O. kisutch male alternative phenotypes
Predicted variableBody size effectRelease date effectIntercept
  • *

    P < 0·05,

  • ***

    P < 0·001.

Jack returns0·102***0·013−0·018***0·004−8·9550·504
Hooknose returns0·020*0·0080·015***0·002−7·2140·402
Jack : hooknose ratio0·091***0·011−0·037***0·003−1·5590·430

spatial and temporal patterns of dynamics

Jack abundance exhibited variability in time and space (Fig. 2; see also Table 2). Close inspection revealed that long-term trends, represented by the third-order polynomials, were similar among populations describing a decreasing trend from the early 1980s to the early or mid-1990s. Hooknose abundance was more variable than that of jacks and the corresponding declines during the 1980s and early 1990s were more pronounced (Fig. 3; note the larger scale of the ordinate compared to Fig. 2). By contrast, temporal fluctuations in the ratio of jack to hooknose males were dampened, showing no consistent trend among populations (Fig. 4). Variation in the jack : hooknose ratio was not explained by early marine growth (linear regression, slope = 0·027, n = 92, adjusted R2 = 0·02, P = 0·08).

Figure 2.

Temporal fluctuations in abundance of O. kisutch jacks returning to nine Oregon hatchery populations (see Fig. 1) over the brood years 1976–97. The relative abundance data (points connected by dotted lines) represent the annual means of deviance residuals from a GLM accounting for release practices (i.e. release date, size at release and number released). The trends are fitted by third-order polynomial curves (solid lines). The adjusted R2 values for the trends are shown with the significance: *P < 0·05, ***P < 0·001.

Figure 3.

Temporal fluctuations in abundance of O. kisutch hooknose males returning to nine Oregon hatchery populations (see Fig. 1) over the brood years 1976–97. The relative abundance data (points connected by dotted lines) represent the annual means of deviance residuals from a GLM accounting for release practices (i.e. release date, size at release and number released). The trends are fitted by third-order polynomial curves (solid lines). The adjusted R2 values for the trends are shown with the significance: *P < 0·05, **P < 0·01, ***P < 0·001.

Figure 4.

Temporal fluctuations in the ratio of jack to hooknose males in O. kisutch for nine Oregon hatchery populations (see Fig. 1) over the brood years 1976–97. The relative bias data (points connected by dotted lines) represent the annual means of deviance residuals from a GLM accounting for release practices (i.e. release date, size at release and number released). The trends are fitted by third-order polynomial curves (solid line). The adjusted R2 values for the trends are shown with the significance: *P < 0·05, **P < 0·01.

Cross-correlation analysis indicated a strong, positive relationship between the temporal fluctuations of jack and hooknose returns within a cohort (lag 0; P < 0·05 for eight of nine populations; Table 4). The pattern was similar for the detrended series, as all correlations were positive and six of the nine were significant. When comparing returns by return year (lag 1), however, correlations were significant after adjustment for multiple comparisons for only two of the nine populations in the original data series and none of the populations in the detrended series (Table 4). This suggests that the ratio of alternative phenotypes might be more variable by return year than brood year. An examination of the jack : hooknose males ratio confirmed this in eight of the nine populations (exact binomial test, P = 0·04; Fig. 5).

Table 4.  Pearson's correlation coefficients between jack and hooknose abundance for the respective populations, computed by brood year (lag 0) and by return year (lag 1) using the original data and detrended series
HatcheryLag 0Lag 1
  • *

    P < 0·05,

  • **

    P < 0·01,

  • ***

    P < 0·001. All P-values except those denoted by an asterisk in parentheses are significant at the 0·05 level following adjustment for multiple comparisons using the false discovery rate procedure (Benjamini & Hochberg 1995).

Eagle Creek150·67**0·79***110·10−0·49
Big Creek190·71***0·67**170·65**0·50(*)
Salmon River200·89***0·54*170·47−0·09
Fall Creek150·77***0·44120·29−0·10
Cole Rivers160·210·27110·74**0·42
Figure 5.

Variability (coefficient of variation, CV) in the ratio of jack to hooknose males by brood year (open circle) and return year (solid circle) for the nine O. kisutch populations in Oregon (see Fig. 1).

Both jack and hooknose males showed similar spatial patterns in the dynamics and the trend components. Mean correlations for the original data series of jacks and hooknoses were positive, as were the 95% bootstrap confidence intervals (i.e. the lower confidence limit was above zero; Table 5). This indicates that the marine survivals of the two phenotypes were synchronized significantly across the region. Furthermore, positive mean correlations and confidence intervals for the trend series indicate that this region-wide synchrony was related to long-term trends (Table 5). However, the detrended series, which address year-to-year fluctuations independent of trends, showed that synchrony at this scale differed spatially between jacks and hooknoses. In jacks, the mean correlation for the detrended series was positive and the confidence interval suggested marginal significance, indicating a tendency for region-wide synchrony (Table 5). This pattern was reinforced and became significant when Cole Rivers, the outlier in terms of distance from the other populations (580–1020 km away; Fig. 1), was excluded from the analysis. By contrast, the year-to-year fluctuations of hooknoses did not show region-wide synchrony, even with the exclusion of Cole Rivers. Finally, there was no evidence of significant spatial synchrony in the jack : hooknose ratio for any of the data series. This was not a simple function of reduced variability or low statistical power of the series, because these were similar to those of the jack series (Table 2).

Table 5.  The degree of spatial synchrony across the region, as measured by the mean correlation coefficients among the nine series using the original data, trend and detrended series of return abundance and ratio of O. kisutch male alternative phenotypes. 95% bootstrap confidence intervals are shown in parentheses. The values in square brackets represent the correlation coefficients recomputed excluding Cole Rivers, the population distant from all others
SeriesJackHooknoseJack : hooknose ratio
Original0·26 (0·03, 0·48)0·40 (0·17, 0·58)0·03 (−0·20, 0·30)
[0·37 (0·22, 0·54)][0·48 (0·32, 0·64)][−0·03 (−0·28, 0·22)]
Trend0·41 (0·10, 0·74)0·64 (0·41, 0·82)0·09 (−0·29, 0·70)
[0·51 (0·20, 0·79)][0·62 (0·37, 0·81)][0·01 (−0·33, 0·65)]
Detrended0·17 (−0·01, 0·34)0·15 (−0·09, 0·36)0·01 (−0·13, 0·13)
[0·24 (0·07, 0·40)][0·19 (−0·07, 0·40)][−0·01 (−0·17, 0·16)]

Jack and hooknose males also showed similar spatial structure by distance within the region. In jacks, the Mantel correlation was not significantly different from zero for either original data or trend series (Table 6), indicating that there was no strong local-scale structure in the dynamics or the trend components (Fig. 6). In hooknose males, the Mantel correlation for the original data series suggested that there was local structure when Cole Rivers was included (Table 6). However, when Cole Rivers was excluded the relation became non-significant, indicating that the previous result was inflated due to this single population. As with jacks, the trend series showed no local structure. For the detrended series, the two phenotypes exhibited differing patterns. The Mantel correlation for jacks was significantly negative, indicating that year-to-year fluctuations were structured spatially, with synchrony declining with distance. This pattern did not depend on whether Cole Rivers was included or excluded (Table 6; Fig. 6). By contrast, there was no such a structure in the year-to-year fluctuations of hooknose males.

Table 6.  The degree of decline in spatial synchrony with geographical distance (Mantel correlation coefficients) for the original data, trend and detrended series of O. kisutch male alternative phenotypes. In parentheses are the coefficients recomputed excluding Cole Rivers. Significance was evaluated by a randomization test
  • *

    P < 0·05,

  • **

    P < 0·01.

Original−0·57 (−0·17)−0·55* (−0·01)
Trend−0·22 (0·09)−0·05 (−0·10)
Detrended−0·55** (−0·42*)−0·27 (−0·07)
Figure 6.

Spatial synchrony (pairwise correlation coefficient) between populations in relation to geographical distance for original, trend and detrended data series of jack and hooknose males. The curves (correlograms) were fitted using local polynomial regression, or the ‘loess’ function implemented in r (Ihaka & Gentleman 1996) with a smoothing parameter α = 1. Curves including (dashed line) and excluding (solid line) the distant population, Cole Rivers (data denoted by open circle), are shown. The horizontal dotted lines represent no correlation.


Fluctuations in the abundance of alternative male phenotypes in O. kisutch showed spatial and temporal structure across the regional scale we examined. Ocean conditions were an important determinant of these fluctuations. This effect was independent of freshwater influences, such as the number of fish released, fish size at release and release date. The predominant trend for jack and hooknose males was one of a decrease in abundance during the 1980s and early 1990s. This is a pattern consistent with other time-series of adult survival in the same region (Coronado & Hilborn 1998; Koslow et al. 2002; Logerwell et al. 2003) and indicative of broad-scale ocean climatic shifts (Mantua et al. 1997; Hare, Mantua & Francis 1999; Beamish et al. 2000; Mantua & Hare 2002). The oceanic effect on jack and hooknose abundance was manifested primarily through effects on survival common to both phenotypes during the first year at sea or, more specifically, during the six months prior to the spawning migration of jacks back to their natal rivers. Indicative of this was the positive and frequently significant correlations between jack and hooknose abundance based on brood year (cohort), but not based on return year. Moreover, a similar pattern was apparent in the detrended series where the influence of autocorrelation was reduced. While a correlation between jack and hooknose abundance is well recognized (Pearcy 1992; Cole 2000; Logerwell et al. 2003), our results elucidate that this is more a function of highly synchronous year-to-year fluctuations in survival than common long-term trends.

Dynamics of the alternative phenotypes showed similar spatial patterns, with the annual fluctuations and the long-term trends being synchronized across the region. The year-to-year fluctuations, however, identified differences in regional synchrony between the two phenotypes. Synchrony among populations in the year-to-year fluctuations of jacks decreased with distance (i.e. spatial structure at a local scale), whereas the fluctuations of hooknoses showed weaker, non-significant local structuring. This is indicative of some environmental processes operating at different spatial scales on the two phenotypes, possibly reflecting differences in their migratory ranges/ocean distributions. Post-smolt O. kisutch reside frequently in local coastal waters or even near the mouths of their natal rivers during their first summer after ocean entry (Pearcy & Fisher 1988). Jacks, which mature during that first autumn and return to their natal rivers to spawn, thus do not undertake long ocean migrations and are subject to coastal-water conditions that may vary at local scales. By contrast, hooknoses, which frequently migrate greater distances and show relatively broad ocean distributions (Weitkamp & Neely 2002), are likely to be affected more by regional or larger-scale oceanographic processes. Thus, our results from the spatial analysis indicate that different patterns of ocean distribution result in the two phenotypes experiencing different oceanic processes, which in turn results in different spatial structuring in survival.

While ocean environmental variation appeared to influence fish survival, it did not influence the proportion of individuals within a cohort adopting one phenotype or the other. Jack : hooknose ratio showed neither temporal nor spatial patterns. Moreover, the phenotype ratio within a cohort was independent of early marine growth while controlling for the effects of size at release, release date and number of males released. This indicates that early marine growth had little effect on early maturation, despite the potential for mediating freshwater effects (see below). Vøllestad, Peterson & Quinn (2004) found a similar pattern for Chinook salmon, O. tshawytscha and a negative effect of marine growth on jack maturation for O. kisutch from the University of Washington hatchery, although this latter effect was complicated by a temporal trend of decreasing growth rates. Our results extend those of Vøllestad et al. (2004) to a broader scale, confirming that the triggers for early maturation occur primarily during early life in freshwater. The determination of life history prior to ocean entry may influence subsequent behaviour in the ocean (e.g. migratory and foraging patterns). This is likely to be of particular importance in populations where fish destined to mature as hooknoses migrate considerable distances to ocean feeding grounds and those destined to mature early as jacks must remain closer to their natal rivers.

Freshwater environments apparently have a strong impact on the dynamics of alternative phenotypes. The observed effects of juvenile size and release date on jack and hooknose abundance and their ratio are generally consistent with previous studies across a variety of regions (e.g. Bilton et al. 1982; Mathews & Ishida 1989; Hobday & Boehlert 2001; Vøllestad et al. 2004; but see Holtby, Andersen & Kadowaki 1990), indicating that the effects are common over the range of the species’ distribution. Several processes are probably involved in the size and date effects. First, large size at release contributes to increased survival for both male phenotypes when entering the ocean, due to factors such as size-dependent predation (Holtby et al. 1990; Willette 2001) and susceptibility to starvation (Kieffer & Tufts 1998). Secondly, rapid growth (i.e. size at release controlled for release date) leads usually to early maturation (Alm 1959; Day & Rowe 2002), which is also the case in Pacific salmonids (Heath et al. 1994).

Thirdly, jack returns decreased and hooknose returns increased with release date, which may relate to a growth spurt associated with arrival in estuarine or nearshore waters. It may be that the earlier the growth spurt the greater is the potential for it to influence maturation schedules. While an increase in the proportion of males maturing as jacks will influence hooknose male abundance negatively, its effect will be diminished by an approximate order of magnitude greater abundance of hooknoses (Table 1; see also Bilton et al. 1982; Hobday & Boehlert 2001). Juveniles destined to become hooknoses also probably incurred higher mortality early in the migratory season due to risks associated with predation, feeding conditions and physiological readiness for seawater adaptation (Fisher & Pearcy 1988; Mathews & Ishida 1989; Pearcy 1992). Our results above suggest that size at and date of downstream migration in the wild, both of which are governed by various freshwater processes including temperature (e.g. Holtby 1988; Nagata et al. 1998), discharge (Jonsson et al. 2001; Smith et al. 2002) and stream productivity or food availability (Bilby, Fransen & Bisson 1996), have significant effects for the dynamics of alternative phenotypes.

One of the most important evolutionary insights from this study is the considerable annual variability in the jack : hooknose males ratio on the breeding grounds. The 1-year difference in age at maturation between the two phenotypes necessarily decouples their synchronous fluctuation patterns within cohorts due to common environmental effects, resulting in increased annual variability in the phenotype frequency within breeding populations. This, together with the frequency dependent nature of the fitness of the two phenotypes (Gross 1985, 1996; Hutchings & Myers 1994), causes the intensity of sexual selection acting on each of the two phenotypes to fluctuate/alternate annually. The implications are that an evolutionary stable state may be more difficult to reach, and that in any given year it is unlikely that the alternative tactics are in evolutionary equilibrium (sensuLively 1986; Hazel et al. 1990, 2004). This implies that studies attempting to test for such an evolutionary equilibrium in O. kisutch will need to incorporate the dynamics generated by the annual fluctuations in phenotype frequency.

In this context, hatchery rearing of fish may exacerbate the situation because freshwater rearing has become artificial and decoupled from the natural environment to which the plasticity of the phenotype has evolved to respond. Moreover, the releases of hatchery fish can result in large numbers of strays to neighbouring streams, altering the phenotype frequency on the breeding grounds and ultimately affecting the evolutionary dynamics of the alternative phenotypes in the natural populations.

In conclusion, the spatial synchrony in jack and hooknose returns was indicative of regional scale oceanographic effects, while the lack of such a synchrony in their relative frequency was indicative of population- or basin-specific effects. In other words, the frequency dynamics of jack and hooknose males within and among populations are influenced largely by freshwater processes varying at evolutionary and intragenerational time-scales, such as juvenile growth, migration and breeding competition.


We thank David Schneider, Peter Lawson, Ryan Calsbeek and an anonymous reviewer for providing valuable comments on earlier versions of the manuscript and Peter Lawson and William Pearcy for helpful discussions. Mark Lewis provided useful information about the CWT hatchery fish database. Financial support was provided by the Japan Society for the Promotion of Science (JSPS) Postdoctoral Fellowships for Research Abroad to YK and the Coastal Oregon Marine Experimental Station and the Natural Sciences and Engineering Council of Canada to IAF.