Individual diet specialization, niche width and population dynamics: implications for trophic polymorphisms



    Corresponding author
    1. Department of Ecology and Environmental Science, Umeå University, SE-901 87 Umeå, Sweden
    2. Department of Limnology, Evolutionary Biology Centre, Uppsala University, Norbyv. 20, SE-752 36 Uppsala, Sweden
      Richard Svanbäck, Department of Limnology, Evolutionary Biology Centre, Uppsala University, Norbyv. 20, SE-752 36 Uppsala, Sweden. Tel. +46 18 471 2704; Fax: +46 18 531134; E-mail:
    Search for more papers by this author

    1. Department of Ecology and Environmental Science, Umeå University, SE-901 87 Umeå, Sweden
    Search for more papers by this author

Richard Svanbäck, Department of Limnology, Evolutionary Biology Centre, Uppsala University, Norbyv. 20, SE-752 36 Uppsala, Sweden. Tel. +46 18 471 2704; Fax: +46 18 531134; E-mail:


  • 1We studied a perch Perca fluviatilis L. population that during a 9-year period switched between a phase of dominance of adult perch and a phase dominated by juvenile perch driven by cannibalism and intercohort competition. We investigated the effects of these population fluctuations on individual diet specialization and the mechanisms behind this specialization.
  • 2Due to cannibalism, the survival of young-of-the-year (YOY) perch was much lower when adult perch density was high than when adult perch density was low.
  • 3Both the individual niche breadth (if weighed for resource encounter) and the population niche breadth were highest when adult population density was high and, consequently, individual specialization was highest at high adult perch densities.
  • 4When adult perch density was low, the abundances of benthic invertebrate and YOY perch were high and dominated the diet of adult perch, whereas the density of zooplankton was low due to predation from YOY perch. At high perch densities, benthic invertebrate abundance was lower and zooplankton level was higher and some perch switched to feed on zooplankton.
  • 5Our results show that individual specialization may fluctuate with population density through feedback mechanisms via resource levels. Such fluctuations may have profound implications on the evolution of resource polymorphisms.


Many animal populations have generalized diets but each individual function as a specialist with a restricted diet (Bolnick et al. 2003). Individual diet specialization, an interindividual niche variation, has been thought to reduce intraspecific competition (Van Valen 1965; Roughgarden 1972; Polis 1984; Smith 1990), and in some cases, diet specialization among individuals may exceed differences between conventional species (e.g. Werner & Sherry 1987; Ehlinger & Wilson 1988; Meyer 1989). Within-population niche differentiation can occur in a number of ways, including sexual dimorphism (Shine 1989, 1991), ontogenetic niches (e.g. Polis 1984; Werner & Gilliam 1984), discrete polymorphisms (Skúlason & Smith 1995; Smith & Skúlason 1996) or by individual-level variation (Bolnick et al. 2003).

Even though individual diet specialization is widespread, we largely lack the knowledge about the mechanisms behind this pattern. Present evidence suggests that intraspecific competition may be important as it may induce rapid evolution of a wider population niche breadth (Bolnick 2001). Theoretically, expansion of the population niche width and adaptation to novel resources have been shown to be one of the causes of resource polymorphisms and ultimately sympatric speciation (Smith & Skúlason 1996; Dieckmann & Doebeli 1999; Doebeli & Dieckmann 2000). The expansion of the population niche width is thought to be due to increased intraspecific competition or decreased interspecific competition (e.g. Grant 1972; Robinson & Wilson 1994; Robinson & Schluter 2000). Whether the population niche expansion is achieved by increased individual niche width or increased interindividual variation is a question that has been raised theoretically and repeatedly over past decades (e.g. Van Valen 1965; Roughgarden 1972; Taper & Case 1985; Bolnick et al. 2003), but has received little empirical attention.

Although individual variation in resource use thus has attracted substantial theoretical interest, it is symptomatic that the theoretical analyses of diet choice, polymorphism and sympatric speciation have been restricted to consider populations at numerical equilibrium. This situation contrasts with the fluctuation in density over time of many populations as a result of variation in resource levels (Grant 1986; Mittelbach et al. 1995; Smith et al. 1999; Grant & Grant 2002; Klemola et al. 2002; Persson et al. 2003). These fluctuations in resource levels have been related to both external environmental factors (e.g. Grant & Grant 2002) and intrinsically driven dynamics (e.g. Persson et al. 2003). In both cases, population fluctuations will influence diet and habitat choices and thus the selective forces on the population (Wilson & Turelli 1986). Existing models predict that generalist populations of individual specialists will occur under conditions of high food predictability (little or no seasonality) and high food availability and diversity (Van Valen 1965; Roughgarden 1974, 1979). However, little is known of how intraspecific variation in resource use among individuals changes in relation to intrinsically driven density-dependent processes in fluctuating populations. Therefore, studying numerically fluctuating populations provides an opportunity to test clear predictions about the role of intraspecific competition on individual resource use.

The purpose of our study was to analyse the effect of intraspecific competition on individual diet specialization in a population where the population is driven by size-dependent competition and cannibalism (Persson, Byström & Wahlström 2000; Persson et al. 2003). The system consisted of a population of perch (Perca fluviatilis L.) that during a 9-year period experienced an intrinsic cycling due to size-dependent intercohort competition and cannibalism involving a more than 20-fold change in the density of adult individuals (Persson et al. 2000, 2003). We focus on two main issues behind individual specialization. First, we investigate the effect of population density on diet choice in perch including feedbacks of perch consumption on resource levels. Secondly, we investigate the mechanisms (individual and population diet breadth and resource availability) that may give rise to differences in individual specialization. Finally we discuss the potential implications of population dynamics on evolutionary dynamics.


lake description

The study was carried out in Lake Abborrtjärn 3, a small (9·3 ha) oligotrophic lake in central Sweden (64°29′-N, 19°26′-E) (Persson et al. 1996). The maximum depth of the lake is 12 m and the mean depth is 6·3 m. The lake has sparse vegetation and 69% of the shoreline is covered with trees that have fallen into the water. Perch was the only fish species in the lake during 1992–96. In late autumn 1996 and in spring 1997, roach (Rutilus rutilus L.) and artificial vegetation were added to the lake as a part of a whole-lake experiment. However, the roach population is still very small and the fish community totally dominated by perch (Lennart Persson, unpublished data). For more detailed information about the surrounding area and lake chemistry, see Persson et al. (1996, 2000).

field sampling

Fish were sampled with cylindrical plastic traps and fyke nets. In 1992, 60 traps (40 × 100 cm, mesh size 10 and 20 mm, 30 of each mesh size), and in 1993–2000 90 traps (besides the 60 traps used in 1992, 30 traps 25 × 50 cm, mesh size 5 mm) and 10 fyke nets (mesh size 6 mm) were used. Traps with the two largest mesh sizes mainly captured perch ≥ 2 years old, whereas the traps with the smallest mesh size and the fyke nets captured perch ≥ 1 year and also young-of-the-year perch in August and September. In this lake perch ≥ 2 years old are > 100 mm, 1 year old perch are 60–100 mm, and young-of-the-year perch in August and September are 30–60 mm (Byström, Persson & Wahlström 1998; Persson et al. 2000, 2004).

Traps were set on four sampling dates each year (end of May and first weeks in July, August and September, hereafter May, July, August and September, respectively). The traps were set in groups of six (groups of four in 1992), including two traps of each mesh size at 15 different stations. Ten of these stations were situated inshore at a depth of 0·5–2 m. The other five stations were set offshore along two vertical profiles, one with two stations and the other with three stations and with both profiles including the maximum depth of that basin. One fyke net was set perpendicular to the shore at every shore station. All sampling gear used were set at 13.00–14.00 h and raised at 09.00–10.00 h the following day. All perch captured were measured (to nearest millimetre) and weighed (to nearest 0·1 g). Length–weight regression at each sampling date was used to estimate the average weight of a perch of size 150 and 200 mm. On every sampling date stomachs of perch were flushed for dietary analyses. The stomach contents were frozen for later laboratory analyses.

In addition to trap-catches, in August every year perch were sampled with gill nets of standard survey link type (1·5 m high, 36 m long, mesh sizes 5, 8, 10, 12·5, 16, 19·5, 24, 29, 35, 43, 55 mm). Pelagic nets were set at surface (1 m) and the benthic nets were set at a depth of 5 m. The nets were set at 10 am and were lifted at 10 am the following day.

Perch population size

A large trapping effort was carried out every spring for 3 weeks (end of May and beginning of June) to estimate population size of perch ≥ 2 years old by multiple mark–recapture methods (Youngs & Robson 1978). Perch were marked with blue dye injected with a Pan Jet injector (Hart & Pitcher 1969). In spring 1-year-old perch were electrofished from a boat along the shore where they were concentrated. In years when 1-year-old perch abundance was too low to allow population estimates based on mark–recapture (1992, 1993, 1994), spring electrofishing and the total number of 1-year-old perch captured in traps and fyke nets over the whole sampling season were used as abundance indices. The abundance of 1-year-old perch also yielded an estimate of the survival of young-of-the-year (YOY) perch from the date when they had moved to the shore habitat to an age of 1 year old. All data on perch population censuses represent the spring situation.

Zooplankton and macroinvertebrate resources

Pelagic zooplankton was sampled seven to eight times every year during the growing season at three pelagic stations. Zooplankton was collected with a 100 µm-mesh net (diameter 25 cm). One tow was made at each pelagic station from the thermocline (estimated with a thermistor) to the surface at an approximate speed of 0·5 m s−1. Zooplankton levels fluctuate during the growing season, but the average biomass during the part of the growth period when YOY perch affect zooplankton availability (July–September) was much lower in years when YOY perch survival was high (adult perch density low) than in years when YOY perch survival was low (adult perch density high) (Persson et al. 2000, 2003). Therefore we restrict our analysis in this paper to the average (July–September) zooplankton levels.

Macroinvertebrates were sampled in August every year. In 1992, five macroinvertebrate samples were taken with an Ekman dredge (area 630 cm2) at one littoral station at a water depth of 0·5 m. In 1993–96, the macroinvertebrate sampling was extended to three littoral stations and from 1997 to five stations. For the years 1993–2000, six samples were taken at each station with a core sampler (area 63 cm2). In this study we concentrate on predator sensitive macroinvertebrates (PSM). This group consisted of organisms living on macrophytes, branches or on other substrates (Hirudinea, Ephemeroptera, Trichoptera, Odonata, Coleoptera, Megaloptera). The other group (mainly chironomids) consisted of organisms living in the sediment and that are less sensitive to fish predation (see Persson et al. 1996 and references therein). Although chironomids are included in the diet of perch we did not include them in the analysis of resource levels in this paper because it has been shown that they are not related to perch population density (Persson et al. 2000).

diet data analyses

The stomach contents were analysed under a dissecting microscope and were identified to order, family, or species and lengths of the 10 first prey of each group were measured to the nearest 0·1 mm. If there were less than 10 prey from a group then all prey in that group were measured. The lengths of all prey types were then converted to biomass (dry weight) using published length–mass relationships for zooplankton (Bottrell et al. 1976) and macroinvertebrates (Persson & Greenberg 1990). The biomass-based diet was separated into seven different diet categories. The diet categories were: (1) cladocerans, (2) copepods, (3) pelagic macroinvertebrates, (4) YOY perch, (5) predator-sensitive macroinvertebrates (PSM), (6) chironomid larvae and (7) terrestrial prey types. Pelagic macroinvertebrates consisted mainly of chironomid pupae and Chaoborus larvae, and terrestrial prey types included adult stages of Ephemeroptera, Odonata and Gerridae.

There are several methods to measure within-population variation in diet (Bolnick et al. 2002). To assess the within-population diet variation we compared the resource use distribution of an individual to that of its population by using a proportion similarity index (PS) (Schoener 1968; Feinsinger, Spears & Poole 1981; Bolnick et al. 2002). The diet overlap between an individual i and the population is:

image(eqn 1)

where pij is the frequency of diet category j in the individual i's diet, and qj is the frequency of diet category j in the population as a whole. For individuals that specialize on a single diet item j, PSi takes on the value qj. For individuals that consume resources in direct proportion to the population as a whole, PSi will equal 1. The overall prevalence of individual specialization (IS) in the population can be expressed by the average PSi value:

image(eqn 2)

Note that if all individuals have the same diets then IS will be 1, indicating no individual specialization, whereas values close to 0 indicate strong individual specialization.

At each sampling date (May, July, August and September), we calculated IS for two size classes (101–150 mm and 151–200 mm) of perch separately, comparing the individual diet within each size-class with the average diet of the size-class. We only calculated IS for the sampling dates and size-classes where more than five individuals of a size-class were caught.

The disadvantage of using IS as a measure of individual specialization is that when comparing two populations or the same population at two different times with different degrees of individual specialization, it is impossible to determine to what extent differences are due to changes in individual diet width or the population's diet width.

As an estimate of diet width of perch we used an index developed by Levins (1968). The index B is calculated as:

image(eqn 3)

where pj is the proportion of the diet that is represented by diet category j. The index has a minimum at 1·0 when only one prey type is found in the diet and a maximum at n, where n is the total number of prey categories, each representing an equal proportion of the diet. We used Levins’ index to calculate diet width in two ways for both size classes of perch: (1) the average individual diet width in the size class on each sampling date and (2) the total diet width of the size class on each sampling date. Diet width of individuals was calculated from the proportion of dry weight of each prey category in an individual's diet on each sampling date. The diet width of each size class was calculated from the average proportion of dry weight of each prey category in the size class’ diet on each sampling date.

A problem with indices 1–3 is that they have not been weighed for the different availabilities of different resources (Schoener 1974). Weighing for resource availability is essential when analysing mechanisms behind individual and population diet breadths, i.e. whether individual and population feeding simply reflect resource availability or involve active selection. Furthermore, even if data on resource abundance are present for different resource categories, as in our case, resource densities cannot be transformed directly into availabilities that depend on encounter rates with prey (Persson 1990; Mittelbach & Osenberg 1994). We quantified encounter rates for three major prey categories of perch (predator-sensitive macroinvertebrates, cladoceran zooplankton and YOY fish) for the two size classes studied (Persson et al. 2004). These encounter rates were calculated as functions of prey mass (see Persson et al. 2004 for functions and parameter values). These three prey categories also showed the largest changes in availability with changes in perch population density (Persson et al. 2000, 2003). We calculated an encounter based resource diversity index using eqn 3, where here pj stands for the mass encounter with prey j in relation to total mass encounter (for parameter estimations, see Persson et al. 2004).


Yearly values used in the analyses were calculated from averages of four samples (three for YOY perch in the diet and the condition factor). YOY perch were not hatched at the first sampling occasion (May), hence in this case the average biomass in the diet was based on the remaining three sampling occasions (July, August, September). For the condition factor, May sampling estimates were excluded because in some years perch had spawned at this sampling date, whereas in some years they had not. Excluding the May samplings for the condition measure meant that one year (1995) is missing in the analysis of condition factor as very low samples were obtained at the other dates this year. We tested the effect of the inclusion of vegetation in 1997 on all our models with ancova (vegetation as factor and density as covariate), but this effect was never found to be significant (P = 0·12–0·74). As a result, we excluded vegetation from our analysis although our conclusions would still be the same if vegetation had been included. All data were LN-transformed before analysis, except for proportions that were arcsine-square root transformed. Because our analyses are based on time series, tests for autocorrelations were carried out using autoregressive-integrated moving average (ARIMA) methods to test for the appropriateness of regression analyses. As no significant time lags were observed in the ARIMA (all ARIMA tested at lag 1; P = 0·21–0·82), we subsequently used ordinary regression analyses.


population numbers and condition of perch

During the study period perch population densities differed as much as 25-fold for ≥ 2-year-old perch and 120-fold for 1-year-old perch. Three distinct phases with respect to perch population numbers and structure were present during the study period. During the first phase (1992–93), high numbers of perch ≥ 2 years old were present whereas the density of 1-year-old perch was low (Fig. 1). The low density of 1-year-old perch was not a result of low reproductive output, but of high cannibalism on YOY perch (Persson et al. 2000, 2003). This period can be classified as a period of cannibal-driven dynamics with high cannibal control of victims (Claessen, De Roos & Persson 2000; Persson et al. 2000, 2003). During the second phase (1994–98), the density of ≥ 2-year-old perch had decreased to low levels, leading to high survival of YOY perch and consequently high densities of 1-year-old perch (Fig. 1). The number of perch ≥ 2 years old decreased because they had grown beyond the size that could be sustained on a long-term by their resource base (Persson et al. 2000). Increased per capita fecundity of the few remaining perch ≥ 2 years old resulted in the production of new YOY cohorts, and these YOY cohorts out-competed 1-year-old perch, in turn, leading to the reappearance of new strong cohorts of 1-year-old perch but continued low densities of perch of ≥ 2 years old for a number of years (Fig. 1). The third phase (1999–2000) resembles the first phase with high densities of ≥ 2-year-old perch, low survival of YOY perch and consequently low densities of 1-year-old perch (Fig. 1).

Figure 1.

Changes in population density for 1-year-old perch (means, filled circles) and perch ≥ 2 years old (means ± 95% CL, open circles) during the study period. Note log scale on y axis.

The length–mass relationships of perch 150 and 200 mm was correlated negatively with the density of perch ≥ 2 years old (correlation 150 mm; N = 8; −r = 0·79; P = 0·021; 200 mm; N = 8; −r = 0·79; P = 0·021). Condition of perch was low during the first phase (1992–93), with high numbers of perch ≥ 2 years old reflecting a high resource limitation (Fig. 2). The condition of both 150- and 200-mm perch increased when the density of ≥ 2-year-old perch was reduced to low levels, reflecting a decreased resource limitation for the remaining ≥ 2-year-old perch. The condition of both sizes of perch remained high during the whole second phase (1994–98). The increased condition of these perch was also reflected in an increased per capita growth, in turn, increasing per capita fecundity (Persson et al. 2000, 2003). In 1999, the condition of both sizes of perch started to decrease and continued to decrease during the third phase (1999–2000).

Figure 2.

Changes in condition (average weight) of a 150 mm perch (solid symbols) and a 200 mm perch (open symbols) in relation to the density of perch ≥ 2 years old.

resource densities

The average pelagic zooplankton biomass differed between years with high and low YOY survival, respectively, and was correlated negatively with YOY survival (correlation between average zooplankton biomass and the density of 1-year-old perch the subsequent year; N = 8; −r = 0·98; P < 0·001). Overall, zooplankton biomass was high in years with low YOY survival (1992–93, 1999–2000), but low during the years with high YOY survival (1994–98) (Fig. 3).

Figure 3.

Changes in average resource biomasses (pelagic zooplankton and macroinvertebrates) during the study period. Zooplankton biomasses are averages over the growing season.

The abundance of predator-sensitive macroinvertebrates differed between years of high and low ≥ 2-year-old perch densities. A negative correlation between adult perch density and density of macroinvertebrates has been shown previously for the years 1992–96 (Persson et al. 2000). The inclusion of data from the years 1997–2000 failed to find such relationship, although there was a negative trend between the density of ≥ 2-year-old perch and predator-sensitive macroinvertebrates (correlation; N = 9, −r = 0·522, P = 0·146) (Fig. 3). One reason for the non-significant result in our study is that the strong 1-year-old perch cohort in 1998 in contrast to previous 1-year-old cohorts survived the summer (Persson et al. 2003, 2004) and hence exerted a substantial predation pressure on macroinvertebrates at the same time, as they are not included in the analysis of the relationship between macroinvertebrates and ≥ 2-year-old perch density.

The encounter-based resource diversity was correlated negatively with adult perch density for both size classes of perch, although the relationship was weaker for larger perch (101–150 mm: N = 9, −r = 0·701, P = 0·035; and 151–200 mm: N = 9, −r = 0·606, P = 0·083). For both size-classes, the encounter based resource diversity was low during the first phase (1992–93) and the third phase (1999–2000) (Fig. 4), when the numbers of perch of ≥ 2 years was high. During low numbers of perch of ≥ 2 years (phase two, 1994–98), the encounter-based resource diversity was high. For 101–150 mm perch, the increase in resource diversity in years with low numbers of perch of ≥ 2 years was due to a shift in dominance in encounter with zooplankton to a dominance of both macroinvertebrates and YOY perch, whereas the increase for 151–200 mm perch was due to a shift from a dominance in encounter with macroinvertebrates to a dominance of both macroinvertebrates and YOY perch. These differences between size-classes can, in turn, be related to differences in size-dependencies in encounter rates on the two prey types macroinvertebrates and zooplankton (Byström & García-Berthou 1999; Wahlström et al. 2000).

Figure 4.

The relation between the density of perch ≥ 2 years old and resource diversity. Resource diversity is calculated from encounter probabilities on three resource categories (zooplankton, macroinvertebrates and YOY perch). The resource diversity was calculated for individuals of a size of 126 mm to represent the 101–150 mm size class and 176 mm to represent the 151–200 mm size class and is based on size-specific attack rates on the three resource categories (see text for more details).

diet use by the population

The diet of ≥ 2-year-old perch during years with low adult perch densities consisted mainly of littoral prey types, and in August and September also YOY perch. The proportion of pelagic zooplankton (cladocerans and copepods) in the diet of both size classes of perch was higher in years with a high density of perch ≥ 2 years old (1992–93, 1999–2000) than in years with a low density of perch ≥ 2 years old (1994–98) (correlation between proportion of pelagic zooplankton and perch density: 101–150 mm perch, N = 9, r = 0·845, P = 0·004; 151–200 mm perch, N = 9, r = 0·775, P = 0·014) (Fig. 5). The proportion of YOY perch in the diet of both size classes of perch decreased with perch ≥ 2-year-old density (Fig. 5), but was significant only for the 151–200 mm size class (correlation; 101–150 mm perch, N = 9, −r = 0·570, P = 0·109; 151–200 mm perch, N = 9, −r = 0·704, P = 0·034).

Figure 5.

Proportion of pelagic zooplankton (copepods and cladocerans) and YOY perch in the diet of perch related to the density of perch ≥ 2 years old for perch 101–150 mm (left panel) and perch 151–200 mm (right panel). Zooplankton is calculated as the average of all samplings within a year, whereas YOY perch in the diet is an average of the July, August and September samplings.

individual diet specialization

During the study period, individual specialization (IS) varied between 0·36 and 1 for the 101–150 mm size class and between 0·31 and 1 for the 151–200 mm size class. The index of individual diet specialization (values of IS) for both size-classes was negatively correlated with the density of perch ≥ 2 years old (101–150 mm; N = 9, −r = 0·770, P = 0·015, 151–200 mm; N = 9, −r = 0·883, P = 0·002) (Fig. 6). After the major die-off of ≥ 2-year-old perch in 1994, individual specialization decreased for both size-classes and stayed low throughout the period of low densities of ≥ 2-year-old perch. Following the increase in density of perch ≥ 2 years old from 1998 to 2000, individual specialization again increased for both size-classes reaching similar levels of individual specialization as in 1992–93 (Fig. 6).

Figure 6.

Top panels; changes in individual specialization (IS) within each size class related to the density of perch ≥ 2 years old for perch 101–150 mm (left top panel) and perch 151–200 mm (right top panel). Bottom panels; changes in diet breadth for the entire size classes (open symbols) and individual diet breath within each size class (filled symbols) related to the density of perch ≥ 2 years old for 100–150 mm large perch and 150–200 mm large perch.

The size-class diet breadth of both size-classes was positively related to the density of ≥ 2-year-old perch (correlation; 101–150 mm; N = 9, r = 0·763, P = 0·017, 151–200 mm; N = 9, r = 0·719, P = 0·029). In the years with a high density of ≥ 2-year-old perch, the size-class diet breadth of both size-classes was high (Fig. 6) related to the inclusion of pelagic prey types in the diet. In contrast, in years with low perch ≥ 2-year-old densities, the size-class diet breadth of both size-classes was low (Fig. 6) corresponding to the observation that most perch were feeding on macroinvertebrates (May and July) and YOY perch (August and September). These shifts in food resource use were related to density-dependent habitat use (see Persson et al. 2000 for habitat use results). The individual diet breadth (not weighed for resource diversity) was relatively constant for both size-classes throughout the study period (Fig. 6). However, considering that the encounter-based resource diversity was lower at high ≥ 2-year-old perch densities than at low ≥ 2-year-old perch densities actually implies that both individual niche breadth and population niche breadth was related positively to the population density of perch ≥ 2 years old.


In this study we found that the degree of individual specialization increased with increasing adult population size. This increase in individual specialization was due to an increase in both population diet breadth and individual diet breadth (when weighed for resource availability) with increasing density, indicating that both individuals and the population had more generalized diets at high perch densities. Based on these changes on degree in specialization, it is relevant to ask why individuals specialize. Some studies emphasize the importance of learning for foraging specialization (e.g. Heinrich 1976; Pietrewicz & Kamil 1979; Lewis 1986), whereas most studies of foraging specialization have been associated with morphological variation (Robinson & Wilson 1994; Smith & Skúlason 1996; Bolnick et al. 2003). In many cases where individuals of different morphologies have been found to specialize on different niches, evidence has also been found for efficiency trade-offs between alternate niches (e.g. Ehlinger 1990; Smith 1987; Schluter 1993; Svanbäck & Eklöv 2003; see also review by Bolnick et al. 2003). For example, in many lakes perch from the littoral zone are deeper-bodied than and forage on littoral prey types more than streamlined individuals, which are found in the pelagic zone and forage on pelagic prey types (Svanbäck & Eklöv 2002, 2003; Peter Eklöv & Richard Svanbäck unpublished data). This difference in morphology corresponds to functional expectations for fish species that occupy these different habitats (e.g. Webb 1984). In accordance with functional expectations, the difference in morphology and diet of perch between the two habitats have been shown to be a reflection of morphology dependent foraging efficiencies where deeper bodied individuals caught in the littoral zone had higher foraging efficiency in experiments with structure. The more streamlined individuals from the pelagic zone, on the other hand, had higher foraging efficiency in open water trials (Svanbäck & Eklöv 2003).

Ecologists have traditionally predicted that foraging efficiency trade-offs would favour generalist populations with individual specialists. Other conditions that could lead to this are high food predictability, high food availability and diversity, high intraspecific competition and low interspecific competition (e.g. Van Valen 1965; Roughgarden 1974, 1979). Many studies have shown individual specialists within generalist populations (see review by Bolnick et al. 2003), where feeding behaviour within a population can span those of several families (Werner & Sherry 1987). Most of these studies are, however, snapshots and represent only cases where individual specialization has been detected (Bolnick et al. 2003). More importantly, besides a study by Bryan & Larkin (1972) and Schindler, Hodgson & Kitchell (1997), neither variation in individual specialization over time nor the mechanisms behind temporal variation have been considered. In contrast to our study, Schindler et al. (1997) failed to show any correlation between largemouth bass (Micropterus salmoides, Lacépède) density and individual diet specialization over a 10-year period. This discrepancy in results may be due to that the variation in bass density was only threefold whereas we found a more than twofold difference in perch density.

The increase in individual and population diet breadth was because higher densities of adult perch led to a higher proportion of the population using the off-shore habitat, as the preferred prey types (YOY perch and macroinvertebrates) were scarce, and consequently a higher proportion of perch included zooplankton into their diet. In accordance with functional expectations, we have also found that it is the more streamlined individuals that move out to the pelagic zone in Lake Abborrtjärn 3 (Svanbäck et al. unpublished data).

ecology and evolution of individual specialization

It has been suggested that intraspecific competition is diversifying, in the sense that any individual able to efficiently use a new, exclusive resource will experience reduced intraspecific competition and have higher fitness (Roughgarden 1972). When competitive pressure increases, selection to switch to new resources becomes stronger so that previously suboptimal resources may confer a benefit (Wilson & Turelli 1986; Bolnick 2001). Furthermore, population niche breadth is thought to represent a balance between the diversifying effects of intraspecific competition and the constraints imposed by interspecific competition (Roughgarden 1972; Grant & Price 1981; Taper & Case 1985). Support for the different effects of inter- and intraspecific competition on population niche breadth and individual specialization stems largely from theory and from observational studies of character release and displacement (Grant 1972; Robinson & Wilson 1994; Robinson & Schluter 2000). However, these studies have not considered the effects of population feedbacks on niche breadths mediated via resource levels. In our study, the increase in population niche breadth and thus the increase in individual specialization with increasing adult density depend on two mechanisms. First, with a high density of adult perch both YOY perch and macroinvertebrate availabilities will be low due to heavy predation from adult perch. In this situation, YOY perch availability is low in terms of energy as YOY perch are cannibalized early on when they represent little energy to the cannibal. Secondly, the low survival of YOY perch at high adult density releases pelagic zooplankton from predation as adult perch are inferior foragers on zooplankton compared to YOY perch (Byström & Garc’a-Berthou 1999; Wahlström et al. 2000; Persson et al. 2003). The increase in zooplankton, in turn, allows adult perch to gain substantial energy from them (Persson et al. 2004). Overall, the change in individual specialization and population niche breadth over time is a result of variable cannibalism intensity and intercohort competition.

It has been suggested that once a subset of the population starts to use a new resource they are subject to selection pressures favouring new adaptations to use that resource (Bolnick 2001). In our studied perch population all adult perch were found within the littoral habitat at low adult densities (Persson et al. 2000). This means that deeper-bodied individuals will have the highest fitness in this phase (Robinson, Wilson & Shea 1996; Svanbäck & Eklöv 2003). In contrast, at high adult densities the same fitness scenario will be true for those staying in the littoral zone, whereas for the ones shifting to the pelagic zone the slender-bodied individuals will have the highest fitness (Svanbäck & Eklöv 2003). Thus, the fitness landscape for the adult population will fluctuate with population density and depend on both density and frequency of different phenotypes (Wilson & Turelli 1986). Based on a fluctuating fitness landscape, we therefore suggest that selection will favour phenotypic plasticity rather than genetically determined morphologies in this perch population. Conditions under which polymorphic populations are plastic and do not diverge genetically include, among other things, local extinction or colonization of competitors and seasonal or between year environmental variation driving relative prey abundance (Robinson & Parsons 2002). Our results indicate that internally driven population dynamics could have the same effect and favour plasticity rather than genetic diversification in polymorphic populations. Interestingly, it has been suggested that Arctic char (Salvelinus alpinus L.) show more stable population dynamics (Johnson 1994; Claessen et al. 2002) than perch. Correspondingly, Arctic char also show much more pronounced differences between different morphs compared to perch (Hindar & Jonsson 1982; Parker & Johnson 1991; Svanbäck & Eklöv 2002, 2003). Although not directly comparable, genetic studies also suggest that Arctic char populations may be more genetically variable than perch populations (Hindar, Ryman & Ståhl 1986; Magnusson & Ferguson 1987; Gerlach et al. 2001). However, the influence of population dynamics on evolutionary dynamics in populations is a field that needs further studies, both theoretically and empirically, before any general conclusions can be drawn.


We would like to thank the Åman Fishery Cooperative for access to the study lake. We also thank Jens Andersson, Eva Mari Diehl, Sebastian Diehl, Charlotta Halvarsson, Joakim Hjelm, Jens Karlsson, Emma Lindgren, Ann Lingerbrandt, David Lundvall, Johan Lövgren, Per Nilsson, Peter Nordin, Anders Persson, Kristina Samuelsson, Fia Staffans and Roger Wallin for field and laboratory assistance during different phases of the study. This manuscript was improved greatly by valuable comments on previous drafts of this manuscript by Daniel Bolnick, Peter Eklöv and two anonymous reviewers. This study was supported by grants from the Swedish Research Council and the Swedish Research Council for Environment, Agricultural Sciences and Spatial Planning to Lennart Persson.