1. Life-history omnivory or size-induced mixed competition–predation systems have under many conditions theoretically been shown to be fragile, whereas at the same time existing empirical data suggest such systems to be common in nature.
2. In a whole lake experiment covering 17 years, we analysed the effects of the introduction of the intraguild prey roach (Rutilus rutilus) on the population size and individual performance of the intraguild predator perch (Perca fluviatilis) and on resource levels in two low productivity systems.
3. A strong long-term effect of roach on the zooplankton resource but not on the macroinvertebrate resource was present. Competitive effects of roach on perch were observed in one of the lakes the first years after the introduction, but at the end of the study no competitive effect of roach on either size class of perch was observed in any of the two lakes. In contrast, a positive predatory effect reflected in improved growth rates of older perch was present.
4. The lack of a support for a competitive effect of roach on small perch raises the question of the importance of mixed competition–predation interactions in life-history omnivorous systems and the problem of comparing descriptive data on feeding relationships with theoretical predictions based on interaction modules.
The theoretical insights concerning the effects of omnivory have developed into a major concern for ecologists as a number of empirical overviews have suggested omnivory to be common in natural systems despite theoretical expectations (Polis 1991; Diehl 1995; Polis & Strong 1996; Arim & Marquet 2004). This contrast between observation and theoretical expectation has led to different attempts to find mechanisms that may increase the likelihood for coexistence between predator and intermediate consumer. First, many IGP modules are embedded in larger food webs that may increase coexistence (McCann, Hastings & Huxel 1997; Hart 2002; Daugherty, Harmon & Briggs 2007; Stouffer & Bascompte 2010). Second, adaptive behaviour may stabilize IGP interactions (Kondoh 2008; Abrams & Fung 2010). Third, feeding at different trophic levels is often a result of shifts in resource use over ontogeny (life-history omnivory), and IGP systems including life-history omnivory have been suggested to have stability properties different from those found in IGP models not considering life-history variation (Pimm & Rice 1987; Holt & Polis 1997). Considering the latter, theoretical studies have yielded different results depending on assumptions about how individual development takes place (Pimm & Rice 1987; Mylius et al. 2001; Van de Wolfshaar, De Roos & Persson 2006). Under the more realistic assumption that development rates of top predators and intermediate consumers depend on food conditions, it has been shown that life-history omnivory strongly demotes coexistence between the top predator and the intermediate consumer if adult predators can survive and reproduce while feeding on resource alone (Van de Wolfshaar, De Roos & Persson 2006). In contrast, IG predator and IG prey may coexist in the case that IG predator life history is characterized by a distinct niche shift such that IG prey is necessary for adult IG predator to reproduce (Hin et al. 2011).
Life-history omnivory in IGP systems gives rise to mixed competition–predation interactions and role reversals (Werner & Gilliam 1984; Persson 1988; Polis & Strong 1996) where small predators compete with the consumer whereas adult predators feed on the consumers. Mixed competition–predation interactions have been claimed to be of major importance in size-structured communities and even have been postulated to explain the lack of recovery of fish stocks (Olsson, Mittelbach & Osenberg 1995; Walters & Kitchell 2001). Here we analyse the extent to which mixed competition–predation interactions are of long-term dynamical importance based on a long-term (>15 years) experiment in a system where previous, short-term studies have suggested mixed competition–predation interactions to be important. The experiment involved the introduction of the intermediate consumer roach (Rutilus rutilus) into two lakes inhabited by the predator perch (Perca fluviatilis), which undergoes substantial niche shifts over its ontogeny. In a previous study, experimental evidence demonstrated that the predator perch cannot coexist with its prey roach but excludes the latter through predation (Persson, De Roos & Byström 2007). Two hypotheses can be forwarded for the observed exclusion of roach by perch: either the perch–roach system is an IGP system, in which IG prey exclusion occurs at high productivities (albeit that the experimental systems were unproductive lakes), or perch predominantly feeds on an exclusive resource in addition to preying on roach and competition between roach and perch is therefore weak or absent. Here we focus on two other unproductive lakes where perch and roach do coexist because of the presence of the top predator pike (Esox lucius) that selects perch over roach as prey (Schulze et al. 2006). Disregarding the strength of the interactions, the food web of roach, perch and pike consists of a mixture of competitive and predatory interactions (Fig. 1, left panel). Analysing the long-term community dynamics, we address the question whether all possible interactions between perch and roach in these systems are equally important for community dynamics and the community is hence most appropriately considered a mixed predation–competition system (Fig. 1, left panel; from here on referred to as ‘mixed interaction scenario’). Alternatively, the competitive and predatory relations between roach and perch may differ substantially in interaction strength, to the degree that the community is more appropriately represented as a pure-interaction system with community dynamics either determined mostly by competition alone (Fig. 1, middle panel; from here on referred to as ‘competition scenario’) or by predation alone (Fig. 1, right panel; from here on referred to as ‘predation scenario’) [see Vandermeer (2006) for transitions between different modules]. This question hence addresses whether this presumably omnivorous system may in the long run actually be structured in a more simple way (Fig. 1 left panel vs. middle and right panels). The experimental analyses include the effects of the introduced roach on shared resources, incidence of interspecific piscivory, size-specific growth rates and population size of perch. More generally, we thus ask whether the seeming contradiction between theory and data regarding IGP (i.e. Polis 1991; Diehl 1995; Polis & Strong 1996) may partly be a result of that interaction strengths in empirical systems are generally poorly known quantitatively.
Materials and methods
Species configuration studied and experimental systems
The study was focused on the relationship between two fish species: the predatory perch (Perca fluviatilis) and the consumer roach (Rutilus rutilus). Perch undergoes substantial ontogenetic niche shifts over its life cycle and starts to feed on zooplankton, to switch to benthic macroinvertebrates to finally end up as a piscivore (Byström, Persson & Wahlström 1998). Roach also undergoes ontogenetic niche shifts but does not become piscivorous. Previous experimental studies in ponds, mesocosms and whole lake experiments have provided ample evidence showing that the perch–roach system is an IGP system with potentially strong competitive (from roach on perch) and predatory (from perch on roach) interactions (Persson 1986; Persson & Greenberg 1990; Byström, Persson & Wahlström 1998; Persson et al. 1999; Persson, De Roos & Byström 2007). Short-term experimental studies show that roach are superior foragers on the shared resource zooplankton (Persson 1987; Byström & Gàrcia-Berthóu 1999). Furthermore, a recent long-term whole lake experiment and comparative lake data show, in accordance with predictions of food dependent IGP theory, that roach cannot coexist with perch in perch–roach systems only and that their common presence in lakes is always associated with the presence of pike (Esox lucius) or other piscivores that have a preference for perch over roach (Schulze et al. 2006; Persson, De Roos & Byström 2007). This conclusion is also supported by comparative data (Fig. 2). Besides zooplankton on which roach is a superior forager, perch and roach also feed on benthic invertebrates. On this resource, perch is a superior forager, and roach overall feeding efficiency on benthic invertebrates is poor (Persson 1988). In addition to interspecific interactions, cannibalism is a major regulatory mechanism in perch and pike (Claessen, De Roos & Persson 2000; Persson et al. 2004).
The two unproductive lakes studied were Lake Abborrtjärn 1 and 2 (LAT 1 and 2 in the following) that are small lakes (5·6–7·6 ha) with similar maximum depths (Table 1) situated in middle Sweden (64°29′N, 19°26′E) (Persson et al. 1996). Before the introduction of roach, both lakes (LAT 1 and 2) were inhabited by perch and pike (Persson et al. 1996). No other fish species are present in the lakes. Roach were introduced to one of the lakes (LAT 2) in the spring and late autumn of 1993. The spring stocking consisted of large roach (>100 mm: 71 individuals ha−1) and the autumn stocking of small roach (50–100 mm: 196 individuals/ha). The size structures and numbers of the introduced roach corresponded to natural densities found in lakes of similar productivity and species composition (including perch and pike) (Sumari 1971; M. Appelberg, unpublished data). In autumn 1996 and spring 1997, roach were also introduced into LAT 1 at numbers and size distributions similar to that stocked in LAT 2 in 1993 (>100 mm: 85 individuals ha−1, 50–100 mm: 207 individuals ha−1). In addition to LAT 1 and 2, the lake experimental area includes two other similar lakes that at the start of the study only contained perch (LAT 3 and 4). LAT 4 was stocked with roach at the same time as LAT 2, and LAT 3 was stocked with roach at the same time as LAT 1 (Byström, Persson & Wahlström 1998; Persson et al. 1999).
Table 1. Physical and chemical characteristics of Lakes Abborrtjärn 1 and 2 (LAT 1, 2) (data Persson et al. 1996)
Max depth (m)
Total phosphorus (μg/L)
Total nitrogen (mg/L)
Three different scenarios can be postulated for the dynamical importance of competition and predation in the interaction between perch and roach, two of which involve pure interactions only (predation or competition) and one of which involve mixed competition and predation (IGP) (Fig. 1). If competition is the dominant feature of the perch–roach interaction and the predation interaction is much weaker or absent, the roach introduction can be expected to negatively affect zooplankton and possibly macrobenthos densities and consequently also negatively affect YOY (young-of-the-year) perch growth and large perch (≥1-year old) growth and perch population size (Fig. 1, middle; Table 2). In contrast, if the perch–roach interaction is primarily a predator–prey relation and competition is weak or absent, perch can be expected to mainly obtain its energy from preying on roach and foraging on macroinvertebrates, while roach predominantly forages on zooplankton (Fig. 1, right). Then, the roach introduction can be expected to have a negative effect on zooplankton, but no effect on YOY perch growth (gape limitation in YOY perch means that predation by YOY perch on YOY roach is negligible), no effect on macroinvertebrates and a positive effect on large perch growth because of increased piscivory (Table 2). The effect on perch population size in the predation scenario is difficult to predict a priori, because the perch population could potentially be regulated by strong cannibalism, in which case positive effects of increased interspecific predation may not be observed at the population level. In a mixed interaction scenario, when competition and predation interactions are of comparable strength, roach introduction can be expected to have effects that are intermediate to the two pure-interaction scenarios. Most importantly, however, because of the life-history omnivory of perch, we expect the negative, competitive effect to dominate for YOY perch growth and the predatory effect to dominate for growth in older perch (Van de Wolfshaar, De Roos & Persson 2006). Because pike presence is necessary for the long-term coexistence of perch and roach (Persson, De Roos & Byström 2007), we cannot test for the dynamic importance of mixed interactions between these two species in a pure IGP setting. Nonetheless, despite the presence of pike we can still test for the presence of mixed interactions, that is, differential effects of roach introduction on perch dependent on perch body size (negative on small perch, positive on large perch) that have been suggested to be the rule in size-structured communities in general (Werner & Gilliam 1984; Persson 1988; Wilbur 1988; Olsson, Mittelbach & Osenberg 1995; Walters & Kitchell 2001).
Table 2. Expected responses of the perch populations to the roach additions for three scenarios in Fig. 1
Perch population size (YOY excluded)
−, 0, +
YOY perch individual growth
Perch ≥1 years individual growth
Macroinvertebrate (benthic) resource
Zooplankton (pelagic) resource
A further complicating factor is that the addition of roach may lead to an increase in pike densities that negatively affect perch densities through apparent competition (Holt 1984). The effect of the roach addition on pike densities is therefore an additional important aspect to include. The predictions of the three scenarios were contrasted at two different time-scales, one the first years after the roach introduction and one at the end of the study. The impact of roach immediately following their introduction has been considered in previous studies (Byström, Persson & Wahlström 1998; Persson et al. 1999, 2004), for which reason our treatment here will mainly be a summing up of these results.
Sampling methods for fish
The sampling in LAT 1 and 2 covers the period 1992–2007. Samplings were temporarily stopped in 2003 but were resumed in 2005. Sampling methods for perch in LAT 1 and 2 have been described in several previous articles see Persson et al. (1996, 1999, 2004) and are only briefly described here. Sampling methods for perch 1-year-old and older were based on capture–recapture methods. Methods for capturing perch (≥2 years) included traps and fyke nets (cf. Persson et al. 1996, 1999). One-year-old (OYO) perch were electrofished in spring from a boat when they were concentrated along the shore. Age and growth of perch were determined from opercular bones (Bagenal & Tesch 1978). From measurements of winter bands, the growth of individual fish in different years could be estimated by back calculations.
For pike, our estimates of population numbers are restricted to 2 years (1992 and 2005). Estimates in 1992 were based on catch effort methods (see Persson et al. 1996) and in 2005 on capture–recapture methods. For large roach, reliable estimates of population size based on mark–recapture were not successfully obtained in either LAT 1 or 2 (Persson et al. 1999). Therefore, standardized gill net catches in August were used as measures of the relative abundances of roach. Gill net catches were obtained with gill nets (benthic and pelagic) of survey link type (see Persson et al. 1999). The capture in the benthic and pelagic net was weighed for the volume of the lake section that the respective net represented to arrive at a habitat volume corrected capture per net (catch per unit effort, CPUE). As a measure of recruitment success, capture of roach in the size range 40–100 mm (essentially 1-year-old and 2-years-old roach) by standardized fyke nets and traps in spring each year was used.
Samples for diet analyses of perch in LAT 1 and 2 presented here were taken in late May, early July, early August and early September of 2006 and 2007 using stomach-flushing. In this study, the analysis was restricted to the incidence of interspecific fish prey (i.e. roach) in the diet of perch. Furthermore, as piscivory was found to be restricted to the summer period (July and early August sampling), only data from these two sampling periods are reported on.
Zooplankton in LAT 1 and 2 were sampled seven times during May–October from 1992 to 2008 except 2003 and 2004. Samples were taken at three pelagic stations in each lake and were collected with a 100-μm mesh net (diameter 25 cm) drawn vertically at an approximate speed of 0·5 m s−1. One tow was made at each pelagic station from the thermocline (estimated with a thermistor) to the surface. Zooplankton samples were preserved with Lugol’s solution. In the laboratory, animals were classified by species, counted and the lengths of 15 individuals (all, if fewer were collected) of each species from each sample were measured in an inverted microscope. Lengths were transformed to biomass using regressions relating length to dry weight (Bottrell et al. 1976).
Samplings of macroinvertebrates in LAT 1 and 2 were carried out in August each year during 1992–2002. 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 and 1994, the macroinvertebrate sampling included three littoral stations and from 1995 five stations. From 1993 and onwards, six samples were taken at each station with a core sampler. Macroinvertebrates were separated into two groups. One group consisted of organisms living on macrophytes, branches or on other substrates (Hirudinea, Ephemeroptera, Trichoptera, Odonata, Coleoptera, Megaloptera), and which are relatively sensitive to fish predation. The other group (mainly chironomids) consisted of organisms living in the sediment and which are less sensitive to fish predation (see Persson et al. 1996 and references therein).
The statistical analyses were carried out in R 2.6.1 (R Development Core Team 2007). We tested for differences between before and after the roach introduction using linear-mixed effect models with treatment as fixed factor and lake as random factor and random intercept (year as random slope was never significant). Also an AR1-autocorrelation term was fitted, but was never significant. If necessary, data were log-transformed to meet test assumptions of normal error distribution and constant variance. Terms were tested for significance with an alpha level of 0·05 using F-tests (likelihood ratio tests usually gave the same results). Because of the different sampling schemes used for the different test variables, degrees of freedom for the different tests differ. Pre- and post-perturbation years used in the different comparisons are therefore given in respective figure legends.
Dynamics of roach densities
After their stocking in LAT 1 and 2, roach have been captured repeatedly in both lakes and in large amounts especially during the three last years of the study (Fig. 3, top panel). Several successful recruitments of roach were observed over the study period reflected in peaks in abundance of roach in the size range 0–100 mm (Fig. 3, middle panel). In LAT 2, the highest numbers of roach 0–100 mm was observed in 1997 following the massive death of piscivorous perch in 1994 (Fig. 3, bottom panel). Thereafter, recruitment of roach has been lower associated with the recovery of perch in this lake but has still been regularly present (Fig. 3, middle panel). Recruitment patterns of roach in LAT 1 did not involve such a strong peak as observed in LAT 2 in 1997 but rather several small peaks that can be related to the absence of a massive die off of larger perch that was observed in the former lake (Fig. 3, bottom panel). Overall, both gill net abundance measures and recruitment patterns suggest that roach is well-established in both lakes and that no tendency for exclusion of roach is present.
Pike and perch abundance
Perch in LAT 2 were heavily affected by the roach introduction in 1993/1994 (Persson et al. 1999). Perch (≥2 years) abundance in LAT 2 decreased to very low numbers during 1994 (apparent in 1995 spring estimate) following the stocking of roach in late 1993–early 1994 (Fig. 3 bottom panel). The population size of perch ≥2 years remained very low during 1995–1997, but by the end of the study recovered to densities observed before the roach stocking (Fig. 3, bottom panel). Perch abundance in LAT 1 also underwent a decrease and a recovery over the study period, but this variation was not clearly related to the stocking of roach (Fig. 3, bottom panel). As for perch in LAT 2, perch densities in 2005–2007 were similar to those in 1992–1994 (both lakes 1992–1994 vs. 2005–2007; F1, 9=1·02, P =0·34).
Pike abundance in 1992 was 23 ± 3 in LAT 1 and 47 ± 102 in LAT 2 (means ± 95% CL) (Persson et al. 1996). In 2005, pike densities amounted to 23·3 ± 3 in LAT 1 and 20 ± 10 in LAT 2 (means ± 95% CL), hence no obvious changes in pike abundance following the roach addition took place in any of the two lakes (paired t-test, t =1·51, P >0·37). These results suggest that any indirect effect of roach via pike (i.e. apparent competition, Holt 1984) was absent. This conclusion is admittedly based on only 1 year before and 1 year after the introduction of roach, but at the same time the same pattern was present in both lakes.
Perch growth, diet and size distribution
The size of OYO perch in LAT 2 decreased in 1995 as a result of the roach introduction (Fig. 3 bottom panel) (Byström, Persson & Wahlström 1998). However, except for this year and only in LAT 2, the body mass of OYO perch at end of May each year did not show any relationship with the roach introduction in neither LAT 1 nor LAT 2 (F1, 12 =1·56, P =0·24) (Fig. 4). Following the roach introduction in LAT 1 in 1996/1997, perch ≥2-year old showed a strong increase in age-specific individual body weight that peaked in 2001, to thereafter rebound to a lower but still substantially higher age-specific body mass than before the roach introduction (Fig. 4 top panel). The same overall pattern was observed for the age-specific body mass of perch ≥2-year old in LAT 2 (Fig. 4 bottom panel) with the main difference that growth of perch ≥2 years in LAT 2 showed a decrease in growth during 1994 associated with a decrease in numbers. The following year, the age-specific mass increase in perch ≥2 years increased. The long-term, age-specific mass increase in 2-years- and 3-years-old perch showed a positive response to the addition of roach in both lakes (2-years-olds F1, 12 =69·4, P < 0.0001, 3-years-olds F1, 12=40·7, P <0·001). In contrast, 4-years-old perch showed no significant response in growth although their body mass were higher after the roach introduction (F1, 12=3·54, P =0·11) (Fig. 4). The size distributions of perch showed less response than growth rates, and the median size only increased marginally from 1992–1994 to 2005–2007 (F1, 9 = 4·72, P =0·06) and the 90% largest percentile did not change (F1, 9 = 0·18, P =0·68). The only marginally significant effect on median size and the lack of an effect on the 90% largest percentile is associated with that the perch population in LAT 2 in 2005–2007 was dominated by one growing cohort born in 2002.
The diet analyses had the purpose to investigate whether roach was present in the diet of perch after roach had been sustained in the lakes for a long time period (i.e. 2006, 2007). No roach was ever found in the stomachs of perch 70–150 mm. This does not necessarily preclude that these perch fed on roach as the main consumption of roach by these sizes of perch is expected to take place from the time of YOY roach hatching (second week of June) to the end of June when no samples for diet analyses were taken. In LAT 1, the proportion of perch stomachs with roach amounted to 18% (2 out of 11 stomachs) in 2006 and 22% (2/9) in 2007 for 151–200 mm perch. For perch >200 mm, the proportion was 27% (4/15) in 2006 and 18% (2/12) in 2007. In LAT 2, the proportion of perch stomachs with roach amounted to 14% (2/14) in 2006 and 10% (1/10) in 2007 for 151–200 mm perch. For perch >200 mm, the proportion was 13% (2/15) in 2006 and 9% (2/22) in 2007. The consistently higher incidence of roach in the stomachs of perch in LAT 1 compared to LAT 2 in 2006 and 2007 can be related to the higher recruitments of roach in LAT 1 in these 2 years (lake difference: F1, 7=17·4, P =0·06) (Fig. 2 middle panel).
Development of resources
The introduction of roach in LAT 2 was associated with a strong decrease in zooplankton in 2004 (Fig. 5 bottom panel and Byström, Persson & Wahlström 1998) but not in LAT 1. Zooplankton biomasses were lower in all years after the roach introduction in LAT 2 except in 1999. In contrast, cladoceran and copepod biomasses fluctuated substantially for several years in LAT 1 after the introduction of roach to settle at low levels from 2005 and onwards (note that data are lacking for 2003 and 2004) (Fig. 5). For both LAT 1 and LAT 2, the biomasses of cladoceran and copepod zooplankton were much lower at the end of the study than before the introduction of roach (1992–1993 vs. 2005–2007, cladocerans: F1, 7=22·6, P =0·002 copepods: F1, 7=61·5, P <0·001) (Fig. 5).
Comparing pre-roach introduction years (1992–1993) and the 3 last years when macroinvertebrates were sampled (2000–2002), no difference in benthic resources (chironomids and predator-sensitive macroinvertebrates) were observed (chironomids : F1, 7 = 3·34, P =0·11, predator-sensitive macroinvertebrates: F1, 7 = 2·14, P =0·11) (Fig. 6). A marginally significant negative relationship between perch ≥2-year-old density and chironomid biomass was present in LAT 1 (F1, 10 = 3·49, P =0·09) over the study years 1992–2002, whereas no significant relationship was found with predator-sensitive macroinvertebrates density in LAT 1 (F1, 10=0·29, P =0·60) and with chironomid biomass and predator-sensitive macroinvertebrates density in LAT 2 (chironomid biomass: F1, 10=2·47, P =0·15, PSM: F1, 10 = 2·26, P =0·16).
Summary: observations and expectations
In our comparison of observed patterns with expectations (Tables 2 and 3), we first consider the short-term response immediately following the roach introduction (Observed I). Our comparison here will, besides data from LAT 1 and 2, also include published results from LAT 4 that in 1993/1994 was a replicate treatment lake to LAT 2 (Byström, Persson & Wahlström 1998; Persson et al. 1999). First, perch population size showed a drastic decrease in LAT 2 and 4 that was related to starvation and decreased biomass of zooplankton. This pattern was, however, restricted to the perch populations in LAT 2 and 4, as no roach competition-related response in perch was ever observed in LAT 1. Second, growth of YOY perch decreased in the year following the roach introduction in LAT 2 and 4, but no decrease in growth of YOY perch was present in LAT 1. Third, growth rates of perch ≥1 year first decreased in LAT 2 and 4 to thereafter increase the next year (Persson et al. 1999). The increase in individual growth rates in LAT 2 and 4 the next year was related to an increase in the density of predator-sensitive macroinvertebrates (and not piscivory on roach) as a response to the drastic decrease in numbers of perch following the introduction of roach (Persson et al. 1999). In conclusion, the first-year response of perch and resources to the roach introduction was, if anything, mainly in correspondence with a competition only scenario (Tables 2 and 3). This conclusion is, however, only partly supported (patterns in LAT 2 and 4) as no responses in perch or resources were ever observed in LAT 1.
Table 3. Observed responses of the perch populations and resources to the roach additions and the correspondence to the predictions of the three trophic configurations in Table 2 (M, mixed interactions scenario; C, competition scenario; P, predation scenario). Comparisons are made at two time scales: the year following the roach introduction (LAT 2 1994, LAT 1 1997–1998, Observed I) and the last 3 years of the study (Observed II). The comparison made the year following the roach introduction also includes results from LAT 4 published before (Byström, Persson & Wahlström 1998; Persson et al. 1999)
The patterns regarding perch performance and resources at the end of the study were largely in correspondence with expectations from a mixed configuration or predation only scenario (Observed II, Table 3). Growth of small (YOY) perch is the critical criteria to differentiate between the mixed predation–competition scenario, and the predation only scenario as the former predicts a negative effect of the intermediate consumer on the growth of small individuals (Tables 2 and 3, Fig. 1) (Werner & Gilliam 1984; Persson 1988; Walters & Kitchell 2001). In other words, considering all response variables in both lakes at the end of the experiment, the observed patterns support the predation only scenario.
Whole lakes, time-scales and experiments
Both comparative studies and experimental studies have shown that the coexistence of perch and roach depends on the presence of top predators that may reduce the predatory effect of perch on roach as in LAT 1 and 2 (Sumari 1971; Persson, De Roos & Byström 2007). The long-term nature of the experiment in these two lakes gives us the unique possibility to scrutinize the interactions between roach and perch that are in the long run dynamically important compared to the potential ones depicted in Fig. 1.
The pros and cons of large-scale, whole-system (generally poorly replicated) vs. microcosm (generally highly replicated) experiments have led to considerable discussion in the ecological literature (cf. Carpenter 1996; Jessup et al. 2004; Lawton 2006; Benton et al. 2007). Schindler (1997) pointed out a number of limitations of small-scale experiments of relevance to the present study including inappropriate spatial scales to include whole communities, too short temporal scales to accurately assess the response of slow-responding organisms like fish and elimination of key littoral pelagic interactions. In the present experiment, we sacrificed proper control replicates for logistic reasons and basically used a replicated before/after design to be able to carry out a whole-system and long-term (15 years) experiment. Therefore, statistical differences found between pre- and post-stocking years show that the periods differed, but in a strict sense not that these differences were attributed to the treatment (roach addition).
Resource reponses and perch performance
We observed a long-term decrease in zooplankton levels following the roach introduction into LAT 2, which is in line with roach’s higher foraging capacity on this resource (Persson 1987; Persson & Greenberg 1990; Byström, Persson & Wahlström 1998). For LAT 2, zooplankton biomasses were except for 1 year (1999) consistently lower after the introduction of roach than before the introduction, whereas zooplankton biomasses in LAT 1 were relatively high up to 2001 (Fig. 5). A possible explanation for this high biomass of zooplankton in LAT 2 in 1999 is variation perch recruitment or environmental factors although neither temperature condition nor nutrient levels were different in this year compared to those in, for example, 1998 and 2000 (L. Persson, unpublished data). The between-year fluctuations in zooplankton biomasses observed in LAT 1 during 1998–2002, after which biomasses settled at low levels, may be related to at least 2 factors. First, roach abundance in LAT 1 was overall lower than that in LAT 2 up to 2005, and there was between-year variability in perch recruitment.
A possible reason for the different patterns in the short-term response of perch in the two lakes to the roach introductions in LAT 1 and LAT 2, respectively, is differences in the competitive effect of roach on perch between lakes as perch in LAT 1 were larger than in LAT 2 at the time of the roach stocking (Persson, De Roos & Byström 2007). The low biomass of zooplankton in both LAT 1 and 2 during the last years of the study had no negative effect on perch performance in either individual growth or in numbers. This lack of a negative response in perch may be observed because benthic prey forms the main energy source for non-piscivorous perch (Persson et al. 1999) and that this resource was unaffected by roach. Furthermore, although perch feed on zooplankton during early phases (Urho 1997; Byström, Persson & Wahlström 1998), they are during the larval phase more limited by digestive constraints than by food density (Persson et al. 2000b; Byström et al. 2003). The lack of a long-term negative effect of roach on OYO perch size provides clear evidence for a lack of competitive effect as in particular perch up to the age of 1 year are expected to have the largest potential diet overlap with roach.
We did not observe any long-term effect on benthic macroinvertebrates despite the recovery of the perch populations during the last years of the experiment, except for that chironomid biomasses tended to be negatively related to perch density in LAT 1. It should be noted that our sampling of benthic macroinvertebrates ended in 2002 when the perch population in LAT 1 was still at low densities why caution in interpretation is necessary. Still, the biomass of benthic macroinvertebrates in 2002 in LAT 2 was comparable to that in previous years despite an increased density of perch, suggesting that competition from roach on this resource was small corresponding with roach’s low foraging efficiency on this resource (Persson 1988).
We did not observe any change in population size of perch when comparing pre-stocking years and the last years of the experiment, whereas individual growth of perch 2–3 years old was higher after the roach introduction. We interpret the lack of a population response in perch as an indication that pike predation and perch cannibalism limit and regulate the population size of perch (Persson et al. 1996, 2003; Persson, Byström & Wahlström 2000a). The increase in individual growth of perch can be explained by the fact that roach (mainly YOY roach) represents a new resource to perch, and that the larger size of this resource compared to zooplankton and macroinvertebrates will allow for a larger maximum size of perch (Claessen et al. 2002).
Short-term vs. long-term responses
Previous within-generation studies on perch–roach interactions at pond and mesocosm scales have provided evidence for a potential competitive effect of roach on perch (Persson 1987; Persson & Greenberg 1990). Furthermore, whole lake experiments including the experimental lakes studied here have provided evidence for strong effects of roach introductions on both perch growth and mortality over time periods up to several years (Persson 1986; Byström, Persson & Wahlström 1998; Persson et al. 1999, 2004). Still, despite the fact that these studies covered several years, our results covering many years (11 years LAT 1, 15 years LAT 2) of data suggest that the competitive effect of roach on perch may be negligible in the long run (Table 3).
A relevant question to raise here is whether the patterns observed during the last years of the experiment represent the final asymptotic state of the system or not. Perch populations (Fig. 7) in single-species systems may show intrinsic cycles, but these are absent in the presence of pike (Persson et al. 1999, 2003). Van de Wolfshaar, De Roos & Persson (2006) showed theoretically that an alternative state in the perch–roach system with roach driving perch to extinction through competition is possible. The extent to which this alternative state can be realized is expected to depend on the potential for a strong competitive impact of roach on perch. The occurrence of a roach-dominated equilibrium state seems less likely in the experimental lakes we studied given the only short-term competitive effect of roach observed in LAT 2 and 4 and its absence in LAT 1. This conclusion is also supported by the fact that the overall resource production in the experimental lakes is dominated by benthic production (Karlsson et al. 2009). In contrast, in, for example, highly humic lakes with low benthic production and dominance of pelagic production, the room for such a competition dominated state to be present can be hypothesized to be more likely. The same may be true for highly productive systems dominated by pelagic production (Persson 1986). Still comparative data argue against a roach only state to be empirically present as lakes with roach and no perch are very rare whereas perch only lakes are more common (Sumari 1971).
Pure vs. mixed interactions in ecological systems
Overall, our experimental results from the unproductive LAT 1 and 2 suggest that in the long run the realized interaction web was substantially simpler than a priori postulated (Fig. 1a vs. b and c). In particular, the realized interaction web at the end of the study lacked evidence of interspecific mixed predation–competition interactions, as we did not find any evidence for a competitive effect of roach on perch. This result was unexpected given that previous experiments on this community configuration have suggested life history-related interspecific mixed competition–predation interactions to be important (Persson 1986, 1987; Persson & Greenberg 1990; Byström, Persson & Wahlström 1998; Persson et al. 1999). Admittedly, our experiments involved a stocking of roach into resident perch populations and do, hence, not address the question of a competitive effect of perch on roach (that would also lead to mixed interactions). Still, within-generation mesocosm experiments have shown that the competitive interactions between perch and roach, if present, are highly asymmetric, and only a competitive effect of roach on perch and not the reverse has been demonstrated (Persson 1987).
Considering the substantial attention that life history-related mixed interspecific interactions have received in the ecological literature (cf. Werner & Gilliam 1984; Wilbur 1988; Olsson, Mittelbach & Osenberg 1995; Byström, Persson & Wahlström 1998; Walters & Kitchell 2001), our results represent a major challenge to existing views. The long time-scale needed to reach our conclusions therefore raises questions about interpretations made in many of the field experimental studies in general because of the possible long transients present. Our long-term experimental data point to that mixed predation predation–competition interactions, although potentially there, are not realized in the long term. Whether this is the case for this species pair under all environmental conditions (i.e. along a gradient of benthic production/pelagic production ratio, see previous section) and the extent to which this generalize to other life-history omnivory systems remain to be investigated. Theoretically we know that the form of the niche shift in the IG predator is an important factor (Hin et al. 2011, in press). Still, we lack empirical studies on the relationship between the form of niche shift in top predators and whether (i) coexistence between IG predator and IG prey; and (ii) significant mixed predation–competition interactions are present.
The present study suggests that when perch and roach occur together because of presence of pike their interactions do still not involve any dynamically important interspecific mixed interactions. This conclusion was only possible to reach based on a long-term interpretation of both individual and population processes. In contrast, data on diet patterns that form the basis for overviews regarding the commonness of IGP in ecological systems in general (i.e. Arim & Marquet 2004) would certainly have provided evidence for the presence of interspecific mixed interactions. Unstructured IGP theory has provided a number of reasons for why IGP interactions may be viable in natural systems including that they are embedded in larger food webs, that weak omnivorous interactions may stabilize ecological systems and adaptive behaviour (McCann, Hastings & Huxel 1997; Hart 2002; Daugherty, Harmon & Briggs 2007; Kondoh 2008; Abrams & Fung 2010; Stouffer & Bascompte 2010). Our results relate to at least one of the first explanation as the coexistence of perch and roach depends on the presence of another top predator (usually pike). Furthermore, our results also suggest that the seeming contradiction between theory and field data to a substantial degree may be due to that theoretical interaction webs are compared with empirical descriptive data on food (qualitative) relationships rather than with results from experiments. In a broader sense, our study underlines the necessity of long-term and whole-system experimental studies to be able to unravel the essential processes driving the dynamics of ecological systems.
We thank E.G., M.H., E.L., J.L., K.M., K.N., A.P., L.S., S.S., M.S., R.S., E.W., R.W. and E.W. for field and laboratory assistance during different phases of this long-term study. We thank A.S. for help with the statistical analyses in R and G.M., A.S. and several reviewers for valuable comments on the article. We also thank the Åman Fishery Co-operative for access to the study lakes. The study was funded by the Swedish Research Council to L.P. A.M.D.R. was supported by the Netherlands Organization for Scientific Research.