The relaxation of predation and interspecific competition are hypothesized to allow evolution toward “optimal” body size in island environments, resulting in the gigantism of small organisms. We tested this hypothesis by studying a small teleost (nine-spined stickleback, Pungitius pungitius) from four marine and five lake (diverse fish community) and nine pond (impoverished fish community) populations. In line with theory, pond fish tended to be larger than their marine or lake conspecifics, sometimes reaching giant sizes. In two geographically independent cases when predatory fish had been introduced into ponds, fish were smaller than those in nearby ponds lacking predators. Pond fish were also smaller when found in sympatry with three-spined stickleback (Gasterosteus aculeatus) than those in ponds lacking competitors. Size-at-age analyses demonstrated that larger size in ponds was achieved by both increased growth rates and extended longevity of pond fish. Results from a common garden experiment indicate that the growth differences had a genetic basis: pond fish developed two to three times higher body mass than marine fish during 36 weeks of growth under similar conditions. Hence, reduced risk of predation and interspecific competition appear to be chief forces driving insular body size evolution toward gigantism.

Individual and population level variation in body size often correlates with the variation in many physiological and fitness traits (Peters 1983; Roff 1992; Stearns 1992), making body size a trait of fundamental ecological and evolutionary importance and interest. In general, fecundity selection, male–male competition, and female mate choice often favor larger body size (Wooton 1979; Clutton-Brock et al. 1982; Shine 1988, 1989; Andersson 1994). Still, in many cases, evolution toward larger body size is not taking place, and the question “What keeps organisms small?” is highly relevant (Stanley 1973; Blanckenhorn 2000). Because predation is one of the main factors behind viability selection against gigantism, it can be a critical factor keeping organisms “small” (Blanckenhorn 2000). Likewise, by constraining individuals’ possibilities for utilizing critical resources, interspecific competition can counteract selection favoring large body size (Wilson 1975; Lomolino 1985; Simberloff et al. 2000). Several other potential sources of interpopulation variation in body size have been proposed: random events (Wasserzug et al. 1979), different resource levels (Case 1978, 1979), prey size (Gittleman 1985; Forsman 1991; Boback 2003), and intraspecific interactions (Clegg and Owens 2002; Wu et al. 2006) can all significantly contribute to the body size variation among populations. Hence, in general, body size at the population level is a result of several co-occurring evolutionary and ecological forces that can be hard to disentangle.

Insular evolution has attracted considerable attention from ecologists as well as evolutionary biologists (MacArthur and Wilson 1967; Grant 1998), one reason being that island ecosystems are often less complex, with simpler selective pressures than mainland ones (Whittaker 1998; Schluter 2001). Studying evolutionary changes of a mainland organism following its establishment on islands is a powerful tool in the study of evolution; mainland–island systems can be viewed as natural experiments (e.g., Mayr 1967; Losos et al. 1998; Losos and Schluter 2000). Key features characterizing organisms in island habitats, as compared to mainland ones, are thought to be loss of dispersability (Whittaker 1998) and a relaxation of evolutionary forces stemming from predation pressure and interspecific competition (e.g., McNab 1994; Lomolino 2005).

Body size is among the traits that can change drastically as a response to insular environments, often being the only trait in which differences are ubiquitous (Case 1978). In the first comprehensive review of insular body size evolution of mammals, Foster (1964) pointed out that while some taxa tend to dwarf, others become giants on islands—a phenomenon coined the “island rule” by Van Valen (1973a,b). By revisiting the question, Lomolino (1985) supported Foster's (1964) finding that small-sized mammals tend to become giants, while larger species dwarf on islands. A popular and potential explanation for this pattern is that when a population is released from interspecific competition and predation, it evolves toward an intermediate “optimal” body size (Damuth 1993; Boback and Guyer 2003), which, however, might not represent a single value for large and diverse taxa like mammals (Meiri et al. 2005a). The island rule, originally formulated for mammals, has been widened to other taxa as well, earning support for vertebrates in general (e.g., Clegg and Owens 2002; Lomolino 2005). Parallel to the studies supporting the island rule, others arrived at opposite conclusions and rejected it (Lawlor 1982; Meiri et al. 2004, 2005b, 2006, 2008). Even inverse patterns, i.e., small species becoming smaller while large ones larger on islands, have been found among carnivorous lizards (Meiri 2007). Recent studies highlight the methodological challenges as well as the importance of proper null hypotheses and analytical tools for reliably supporting or rejecting the island rule (Lomolino et al. 2006; Meiri et al. 2008; Welch 2009).

However, the mechanism behind the insular body size evolution within species is also unclear. For instance, the possibility that the selective forces act only indirectly on body size by affecting life history traits is conceivable (Palkovacs 2003). Even though the ecogeographical rules were originally formulated in an interspecific context, intraspecific studies testing the given rule's predictions have also increased understanding of the mechanisms behind the large-scale patterns (see, e.g., Heaney 1978; Fairbairn and Preziosi 1994). In general, it seems that intraspecific studies, without bias from interspecific differences in physiological constraints and phylogenetic past, and with adequate knowledge about the target species’ biology and the studied environments, offer the possibility for in-depth analysis and interpretation of insular body size shifts (e.g., Palmer 2002; Keogh et al. 2005; Wu et al. 2006). Surprisingly, apart from McClain et al.'s (2006) study which compared coastal (treated as mainland) versus deep sea (treated as island) gastropods, all studies on insular body size evolution have focused on terrestrial organisms. However, the “attractiveness” of islands for evolutionary biologists, namely the decreased complexity and simplified selective forces, can be much more pronounced in small isolated water bodies than on terrestrial islands. First, reduced gene flow in the case of obligatory water organisms (e.g., fish) can be complete in isolated water bodies. This is very important given gene flow's power to counteract natural selection (e.g., Hendry et al. 2002; Moore et al. 2007). Second, a complete release from interspecific interactions (competition and predation) is possible in the simple fish communities characterizing some small isolated ponds.

The aim of this study was to test the hypothesis that evolution toward large size in small fish is found in island habitats (=isolated ponds) where the selective forces—stemming from predation and interspecific competition—acting against large body size are relaxed. We did this by comparing the variation in the mean body size among Fennoscandian nine-spined stickleback (Pungitius pungitius) populations. This species represents an excellent model for this purpose, because it is found in a variety of different habitats from seas, lakes, and rivers to small creeks and ditches, successfully persisting also in isolated ponds, often as the only fish species (Bănărescu and Paepke 2001; Östlund-Nilsson et al. 2007). By comparing geographically distinct marine, lake, and pond populations (Fig. 1), we tested the following predictions. First, P. pungitius should be larger in ponds where predator or competitor fish species are absent than in other habitats housing complex fish faunas. Second, the presumed body size differences in the wild should result from the variation in growth rates, not only in lifespan. Third, the presumed differences in growth rate among populations should have a genetic (evolutionary) basis. In addition, we investigated the effect of introduction of predatory fish on the mean body size of the sticklebacks in habitats that formerly lacked predators.

Figure 1.

Map of Fennoscandia showing the location of the sampling sites. Open circles denote small isolated ponds without predatory fish, plus signs denote ponds with introduced Salmo trutta, while crossed circles denote large lake, and filled circles marine populations. For the population abbreviations see Table 1. The underlined populations are the ones represented in the common garden experiment. The dark gray boxes show the regions used in the analyses of predator introductions.



Adult fish were collected from 16 Fennoscandian populations including four marine sites, five nonisolated large lakes (hereafter: lake), and seven isolated ponds (hereafter: pond; Fig. 1, Table 1) between 2003 and 2007. We define lakes as water bodies larger than 30 ha that are connected to creeks and rivers and thus, the stickleback populations in them are not isolated. On the other hand, ponds are smaller than 5 ha and completely isolated (both in- and outlets [when present] are small and steep), meaning no immigration possibilities at all. The sampled populations represent markedly different selective environments. Both marine and lake populations are members of a diverse fish fauna, including several predatory (e.g., pike, Esox lucius, perch, Perca fluviatilis, different salmonids, for example brown trout, Salmo trutta) and numerous competitor fish species (e.g., juveniles of almost all sympatric fish). In contrast, in the isolated ponds there are no predatory fish at all, and due to the relatively small surface area, they do not house permanent fishing birds either. Predation by aquatic insects and possible cannibalism (relevant in early life stages) might be present in all sites. Two facts demonstrate the relaxed predatory pressure in the ponds: first, P. pungitius have undergone strong reduction in defensive armor (reduction or loss of pelvic apparatus) in all studied ponds (G. Herczeg, M. Turtiainen, J. Merila, unpubl. data) and second, fish in the ponds live longer than those in other populations (see the Results section). In the three Russian ponds (Bolotnoje, Krugloje, Mashinnoje), P. pungitius is sympatric with large three-spined sticklebacks (Gasterosteus aculeatus) that heavily outnumber P. pungitius. In one of the Finnish ponds (Pyöreälampi), a few small whitefish (Coregonus lavaretus; still larger than sticklebacks) have been introduced in the last century. However, due to their dietary habits (feeding on invertebrates; see, e.g., Kahilainen et al. 2004), they can be only considered as competitors. Further, due to their low number (we did not catch a single individual among the thousands of sticklebacks), even this competitor effect might be minimal. We included two additional ponds (Kirkasvetinenlampi, Finland; Hansmyrtjärn, Sweden; Fig. 1) where a common fish predator, S. trutta, has been introduced in the last century (exact date unknown) in one analysis.

Table 1.  Sampling sites, their abbreviations, coordinates, sample sizes, surface area of water bodies, and diversity of fish community. Coordinates of the Russian sites are approximate ones based on Ziuganov and Zotin (1995). Gasterosteus aculeatus and Coregonus lavaretus are larger-bodied competitors, while Salmo trutta is a predator of Pungitius pungitius
Sampling siteAbbreviationCoordinatesSample size (aged)Surface area (ha)Fish community
  1. 1Heavily outnumbering P. pungitius.

  2. 2In extremely low numbers.

 Marine (coastal) populations
  Bölesviken, Baltic SeaBÖL63°39′N; 20°12′E 70N/AComplex
  Oulu, Baltic SeaOUL64°58′N; 25°29′E 30 (19)N/AComplex
  Helsinki, Baltic SeaHEL60°13′N; 25°11′E 59 (29)N/AComplex
  Levin Navolok, White SeaLEV66°18′N; 33°25′E 55N/AComplex
  KevojärviKEV69°45′N; 27°00′E 40 115Complex
  PulmankijärviPUL69°58′N; 27°58′E 421620Complex
  TuolpujärviYUO69°34′N; 28°02′E 40 (23) 185Complex
  PorontimaPOR66°12′N; 29°16′E 56 115Complex
  Vastre-SkavtrasketSKA64°26′N; 19°27′E 30  35Complex
  BolotnojeBOL66°18′N; 33°25′E 43<5 (isolated)G. aculeatus1
  KruglojeKRU66°18′N; 33°25′E 32 (26)<5 (isolated)G. aculeatus1
  MashinnojeMAS66°18′N; 33°25′E 63<5 (isolated)G. aculeatus1
  PyöreälampiPYÖ66°15′N; 29°26′E158 (21)<5 (isolated)C. lavaretus2
  RytilampiRYT66°23′N; 29°19′E115 (25)<5 (isolated)
  KirkasvetinenlampiKIR66°26′N; 29°08′E 73<5 (isolated)S. trutta
  AbbortjärnABB64°29′N; 19°26′E 80<5 (isolated)
  BynästjärnenBYN64°27′N; 19°26′E 91 (30)<5 (isolated) 
  HansmyrtjärnHAN64°33′N; 19°10′E 61<5 (isolated)S. trutta

The Baltic Sea (Helsinki, Oulu, Bölesviken) and the Russian samples (Levin Navolok, Bolotnoje, Krugloje, Mashinnoje) form two groups within which the populations are genetically indistinguishable, but these groups were found to be genetically isolated from each other and all the other studied populations (which were also highly isolated from each other in all possible combinations) based on highly variable microsatellite markers (Shikano et al. unpubl. data). The reason for the lack of genetic isolation within the Baltic and Russian groups is easy to understand: there is no barrier against the gene flow in the Baltic Sea despite the large geographic distance between the sample sites, while the Russian ponds became isolated only recently (within a century) from the White Sea (Ziuganov and Zotin 1995; information from the White Sea Biological Station).

In most cases, adult fish were collected in early summer (between 2003 and 2007), during the reproductive season (late May to early July) with the aid of seine nets and minnow traps. P. pungitius normally reaches maturity after its first wintering, and most of its growth is found before the first breeding season (Jones and Hynes 1950; Bănărescu and Paepke 2001), so distinguishing between adults and juveniles is very simple during the reproductive season. One lake, Tuolpujärvi, was sampled in early September. However, considering the general growth patterns (see above) and that data from Tuolpujärvi fitted with the patterns found (see the Results section), we did not exclude this population (exclusion did not alter our results qualitatively). It is noteworthy that we could not identify sex in some populations. However, analysis of a smaller dataset revealed that sexual size dimorphism is only relevant in the giant pond populations (G. Herczeg, A. Gonda, J. Merila, unpubl. data) and because they (1) represent one extreme of the size distribution and (2) had nearly equal sex representation in our sample, we did not consider sex differences in the analyses. The interpopulation size patterns are not qualitatively influenced by inclusion of sex as a factor in the analyses of the smaller set of populations where sex data were available (data not shown).

Collected fish were overanesthetized with MS 222 (tricaine methanesulfonate) at the site of capture and stored in 96% ethanol for about two months. After this, fish were moved to 4% formalin. Standard length (from the tip of the nose to the end of the tail base) was measured with a digital calliper to the nearest 0.01 mm at this time. Age was estimated in a subsample of populations (Helsinki, Oulu, Tuolpujärvi, Krugloje, Bynästjärnen, Pyöreälampi, and Rytilampi) from fin rays (and verified from otoliths for one population; e.g., Shirvell 1981). Fins were first cut as near to the base of the fin as possible and then air dried (dorsal, pectoral, or pelvic fin). After that fins were stained with a neutral red solution (with acetic acid). Annuli were evaluated under microscope with 30–100× magnification. Note that for the purpose of age determination, individuals were chosen to cover the full size range within populations. Hence, due to nonrandom sampling, the age data cannot be used for direct assessment of age structure. In the case of four populations (Oulu, Tuolpujärvi, Kevojärvi and Pulmankijärvi, see Table 1 for details), fish were stored in 96% ethanol for about three years before the samples were moved to formalin. Even though storage in alcohol can cause some shrinkage, length becomes stable during less than two months (e.g., Fox 1996; Kristoffersen and Salvanes 1998). Further, the length change was found to be minor (<3%) in fish species comparable in size to P. pungitius (Kristoffersen and Salvanes 1998). Hence, this together with the fact that all samples were stored in alcohol for at least two months suggests that our samples were comparable.


Four geographically isolated populations, representing two marine (Helsinki, Baltic Sea, and Levin Navolok, White Sea) and two pond (Bynästjärnen and Pyöreälampi, separated by >500 km) habitats were chosen for the common garden experiment (Fig. 1). The marine sites were shallow coastal bays close to creek inlets and thus, they represented low salinity marine habitats (Baltic Sea being a brackish water environment in general). Adult fish were collected before or during the early phase of the breeding season in 2007 and transported to the aquacultural facilities of the University of Helsinki. Fish were provided with frozen bloodworms (Chironomidae sp.) and kept at 17°C and a 24 h light period until enough fish from each population had turned to reproductive condition. Both the collected adults and their offspring were kept and reared in freshwater.

Five full-sib families were made in vitro from each of the four populations. The clutches were transported to 1.4 L tanks of two Allentown Zebrafish Rack Systems (Aquaneering Inc., San Diego, CA, USA). Racks were equipped with a multilevel filtering system including UV, physical, chemical, and biological filters, and had closed water circulation. Four days after hatching, when the fry started to swim freely, 10 fish from each family were randomly chosen and housed individually in the 1.4 L tanks, summing to 200 individually kept fish (four populations × five families × 10 individuals). Visual contact between the tanks was blocked. Fish were fed first with live brine shrimp (Artemia salina) nauplii, and then with frozen copepods (Cyclops sp.) and frozen bloodworms twice a day ad libitum. Temperature was set to 17°C throughout the experiment. We started with a 24 h light period (representative to high latitudes at summer) and changed it to a 12:12 h light–dark periodism gradually during one week after 12 weeks. Due to latitudinal differences between the source populations (Fig. 1), we did not aim to mimic the natural light regimes any closer and did not attempt to follow natural temperature regimes.

The fish were measured at the age of 36 weeks, when the growth curves of all populations approached their asymptotes (Herczeg et al. unpubl. data). Standard length was measured from digital photographs taken under standardized conditions with a ruler placed in each photograph for scaling using tpsDig 1.37 (Rohlf 2002) software. Body weight was taken to the nearest 0.01 g with a digital balance. Due to mortality, and because some randomly chosen fish were sacrificed earlier for other scientific purposes, we could use 86 fish, 21 from the Baltic Sea (family representations: 6, 5, 4, 3, 3), 23 from the White Sea (family representations: 7, 6, 5, 3, 2), 20 from Bynastjärnen (family representations: 7, 4, 3, 3, 3), and 22 from Pyöreälampi (family representations: 6, 5, 5, 5, 1). We note that although the use of the F1 laboratory generation should remove a large part of environmental effects on growth, some maternal effects or cross-generational influences may remain. However, most maternal effects on growth dissipate quickly and seldom explain any large interpopulational differences in growth (e.g., Green 2008).


We analyzed our data with the aid of General Linear Mixed Models (GLMMs) as implemented in SAS (Littell et al. 2006). In the first GLMM, we analyzed body size differences between habitat types. Standard length was the dependent variable (fresh weight was not available, while the differences in alcohol storage might bias weight measures considerably), habitat type (marine, lake, or pond) a fixed factor, and population nested within the habitat type a random factor. As most of our populations were isolated from each other, we did not enter geographic distances into our model. The analysis of the effect of recently introduced predators were based on a subset of the populations: a lake, a predator free pond, and a pond with introduced S. trutta—all being in close geographic proximity (Fig. 1). This setup was replicated in the Kuusamo region in Finland and the Åmsele region in Sweden. The ponds with introduced predators were Kirkasvetinenlampi in the Kuusamo and Hansmyrtjärn in the Åmsele region (Fig 1). Both regions had two ponds without predators (Bynästjärnen and Abbortjärn in the Åmsele region and Pyöreälampi and Rytilampi in the Kuusamo region), but we included only one pond per region for a symmetric setup. However, to provide a conservative comparison, we chose the pond population with the smaller mean body size in both cases (Abbortjärn and Pyöreälampi, see Figs. 1 and 2). Unfortunately, we are not aware of any other similar population system (i.e., one pond + one pond with introduced predator + one lake within a small geographic area), so our results based on only two replicates should be treated with caution. We used a GLMM again; standard length was set as the dependent variable, habitat type (lake, pond, pond with introduced S. trutta) as a fixed factor, and region (Kuusamo and Åmsele) as a random factor. Age–size relationships between habitats were compared with the third GLMM. The standard length was set as the dependent variable, habitat type as a fixed factor, population nested within habitat type as a random factor, and age as a covariate. The size and weight of the common garden experiment fish were analyzed with two GLMMs. The standard length or body weight was set as dependent variables, population as a fixed factor, and family nested within population as a random factor.

Figure 2.

Mean standard length (+95% CI) of wild caught nine-spined sticklebacks in different population and habitat types. For the population abbreviations see Table 1. The star symbols mark the recently isolated P. pungitius populations where they are heavily outnumbered by G. aculeatus.

All analyses were performed with the aid of the SAS 9.1 (SAS Institute, Cary, NC, USA).



The sticklebacks from different habitat types differed significantly in their standard length (GLMM: F2, 13.1= 13.51, P < 0.001; least-squares means ± S.E., all in mm: marine = 40.72 ± 2.84, lake = 41.02 ± 2.54, pond = 55.74 ± 2.14; Fig. 2). Post hoc (Scheffe) tests revealed that while marine and lake fish did not differ (P= 0.99), both were smaller than pond fish (marine–pond: P= 0.004; lake–pond: P= 0.003). The habitat differences were profound; pond fish were in general 35–40% longer than marine or lake fish. However, the population effect within habitat type was also significant (χ21= 569.8, P < 0.001), mainly due to the large variation among ponds (Fig. 2). While sticklebacks from all pond populations tended to be larger than their marine or lake conspecifics, some pond populations (Bynästjärnen, Pyöreälampi, and Rytilampi) contained adult fish that were strikingly larger than “ordinary” fish from the marine or lake sites.

Standard length of P. pungitius differed also between lakes, ponds, and ponds with introduced S. trutta (GLMM: F2, 454= 269.13, P < 0.001). Post hoc (Scheffe) tests showed that the habitat types differed in all possible combinations (P < 0.001 in all tests; Fig. 3). Ponds with introduced S. trutta housed smaller P. pungitius than the ponds lacking this predator (Fig. 3; least-squares means ± S.E., all in mm: lake = 40.56 ± 5.10, pond = 59.35 ± 5.07, pond with S. trutta= 51.31 ± 5.08). The effect of the region of origin was also significant (χ21= 194.2, P < 0.001): P. pungitius in ponds with S. trutta differed from lake conspecifics in the Kuusamo region, but not in the Åmsele region (Fig. 3).

Figure 3.

Mean standard length (+95% CI) of nine-spined sticklebacks in different populations, habitat types, and regions. For the population abbreviations, see Table 1.


A GLMM revealed a habitat-specific age effect on standard length (habitat: F2, 10.8= 1.28, P= 0.32; age: F1, 163= 46.62, P < 0.001; habitat*age: F2, 163= 10.43, P < 0.001; Fig. 4). The population effect within habitat type was significant (χ21= 48.5, P < 0.001). To test for habitat effects directly, we removed the interaction term. Here, habitat was only significant when we removed Krugloje, a pond where P. pungitius is found in sympatry with G. aculeatus (with Krugloje, habitat: F2, 4.23= 2.26, P= 0.21; age: F1, 167= 253.94, P < 0.001; without Krugloje, habitat: F2, 3.61= 34.00, P < 0.005; age: F1, 141= 226.32, P < 0.001; Fig. 4). The significant habitat effect (after the exclusion of Krugloje) suggests that body size is different at a given age between habitats when we consider only pond P. pungitius populations without competitor or predatory fish species being present, while the significant interaction term (it remained highly significant after removing Krugloje, data not shown) suggests that the effect of age on size is habitat specific (Fig. 4). Old fish from Pyöreälampi and Rytilampi ponds were two times longer than fish from lakes or marine sites, a phenomenon we recognize as gigantism. Interestingly, second year fish were not represented among the breeding adults of the ponds.

Figure 4.

Mean size (+95% CI) of nine-spined sticklebacks at given ages from different populations and habitat types. For the population abbreviations see Table 1. The numbers above the population abbreviations denote age in years. The star symbols mark a recently isolated P. pungitius population where nine-spined sticklebacks are heavily outnumbered by G. aculeatus.


The analyses of standard length and body weight variation from the common garden experiments revealed concurrent patterns: populations differed significantly both in standard length (F3, 14.9= 60.46, P < 0.001; least-squares means ± S.E., all in mm: Helsinki [Baltic Sea]= 46.52 ± 1.49, Levin Navolok [White Sea]= 60.80 ± 1.48, Bynästjärnen = 68.98 ± 1.51; Pyöreälampi = 72.74 ± 1.51; Fig. 5) and body weight (F3, 14.1= 81.50, P < 0.001; least-squares means ± S.E., all in g: Helsinki [Baltic Sea]= 0.91 ± 0.13, Levin Navolok [White Sea]= 1.69 ± 0.13, Bynästjärnen = 3.32 ± 0.13; Pyöreälampi = 3.09 ± 0.13; Fig. 5). Post hoc tests showed that Levin Navolok fish (White Sea) were longer and heavier than Helsinki fish (Baltic Sea; standard length: P < 0.001; body weight: P= 0.006). The pond populations (Bynästjärnen and Pyöreälampi) did not differ from each other (standard length: P= 0.41; body weight: P= 0.66), while fish from both marine populations were smaller than fish from both ponds (all P < 0.015). Size differences were profound: pond fish grew two to three times heavier than marine fish during 36 weeks (Fig. 5). The family effect was significant both in the case of standard length (χ21= 8.1, P < 0.005) and body weight (χ21= 4.1, P < 0.05).

Figure 5.

Mean (+95% CI) standard length and body weight of 36 weeks old individually reared common garden fish in different populations and habitats. For the population abbreviations, see Table 1.


According to theory, when the selective forces stemming from predation and interspecific competition are relaxed, small organisms should evolve to become larger because several forms of intraspecific competition and sexual selection all favor larger body size. By studying a widely distributed small-bodied fish species, we found support for this theory in an aquatic vertebrate. P. pungitius has a standard length around 4 cm (about 5–6 cm in total length) in habitats with diverse fish communities (e.g., Bănărescu and Paepke 2001). We found that it grew larger in isolated ponds, and turned into 8–9 cm (up to 10–12 cm in total length) giants in some populations. P. pungitius in ponds with introduced predatory S. trutta or sympatric competitor G. aculeatus were smaller than those in predatory and competitor fish-free ponds. Our size-at-age analyses demonstrated that the increased size was a result of both extended longevity and increased growth rates in the natural pond environments. Fish in the common garden study showed similar size trends to their conspecifics from the wild, pond fish reaching two to three times larger body mass than their marine conspecifics during 36 weeks of growth. This finding strongly suggests that the body size trends seen in the data from the wild have a genetic basis. The repeated, habitat-specific nature of the genetically based body size shifts suggests that observed divergence is adaptive and driven by natural selection (e.g., Clarke 1975; Endler 1986). Obviously, we cannot derive conclusions for the generality of the island rule in fish as we studied one species only. Our results, however, fit perfectly to the predictions drawn from the island rule: in the island-like ponds, where gene flow was zero, interspecific competition and predation negligible, P. pungitius was larger (sometimes reaching giant sizes) than that in the mainland-like large lakes or marine sites.

Apart from sexual selection, which generally favors larger body size within a population (e.g., Wooton 1979; Andersson 1994), several other agents of natural selection—or their absence—can influence the mean body size in a population. For instance, increasing population density and intraspecific competition as a consequence can lead to an increase in body size (Clegg and Owens 2002; Wu et al. 2006, but see, e.g., Boucher et al. 2004 for an opposite pattern in the case of resource limitation). This effect, coupled with that of sexual selection, can be expected to act particularly strongly in cases when predation acting against large or fast-growing phenotypes (Blanckenhorn 2000; Biro et al. 2004, 2006) or interspecific competition narrowing the population's niche (Wilson 1975; Lomolino 1985; Simberloff et al. 2000) are relaxed, a phenomenon typical on islands (McNab 1994; Lomolino 2005). These expectations match our observations: the largest sizes were observed in the Finnish and Swedish ponds where interspecific competition was negligible and predatory pressure low. We did not estimate population densities, but based on our catches, we believe that the densities of P. pungitius are much (by magnitudes) higher in ponds than those in the larger habitat types. Evidence for increased intraspecific competition in pond P. pungitius populations is also available: we found that the levels of aggression and drive to feed, and the social costs of group living are higher among pond than among marine fish as revealed by common garden experiments (Gonda et al. 2009a; Herczeg et al. 2009a,b). Further, brain parts relevant in chemical communication, habituation, or learning were found to be relatively smaller in pond than in marine sticklebacks (Gonda et al. 2009b). Another factor facilitating local adaptation in general and characteristic of island environments in particular is lack of gene flow (Whittaker 1998). This factor is also relevant in our case: all noncoastal ponds were found to be genetically highly differentiated and isolated from all other populations (Shikano et al. unpubl. data). Hence, considering that all the possible ancestor population types (marine or lake) consist of uniformly small sticklebacks (present study, see also Bănărescu and Paepke 2001; Östlund-Nilsson et al. 2007), repeated evolution of the large, fast growing phenotype seems to have happened.

It has been demonstrated in a series of studies focusing on the predation-related life-history evolution of guppies (Poecilia reticulata) that high-predation environments select for earlier maturation at smaller size and larger reproductive allotment (more frequent reproduction, larger number of smaller offspring) than low-predation environments (e.g., Reznick 1982; Reznick and Endler 1982). This pattern has evolved as a response to differences in age-specific mortality: high-predation environments are characterized by high mortality rates that are uniformly distributed across age classes while low-predation environments are characterized by lower mortality rates with relatively higher proportion of mortality prior maturity (Reznick et al. 1996). However, predation-related size divergence was not found in this system (Hendry et al. 2006). It is noteworthy that in the case of three-spined sticklebacks in which the dorsal and pelvic spines together with the supporting bony lateral plates provide effective means of antipredator defense (e.g., Hoogland et al. 1957; Reimchen 1983), large body size can serve fish to escape from gape-limited predators, and as a consequence, the presence of gape-limited predators can select for gigantism (e.g., Moodie 1972a,b; Moodie and Reimchen 1976; Reimchen 1988, 1991). However, the same would not work with P. pungitius because their spines are far less effective against predators (Hoogland et al. 1957). Further, in our case, the larger body size of pond P. pungitius is found parallel to the reduction or even complete loss of the pelvic apparatus in the absence of predatory fish (Herczeg et al. unpubl. data).

Among the ponds, the Russian coastal populations were the closest to marine and lake populations in body size, and the observed size differences appeared to be a result of extended longevity only. However, one of the factors expected to be responsible for insular body size evolution, namely the relaxation of interspecific competition, did not hold true due to the large numbers of sympatric larger bodied three-spined stickleback in these ponds. The recent isolation of these ponds from the White Sea and, as a consequence, the lack of genetic divergence from their marine conspecifics (Shikano et al. unpubl. data) is less likely to be the cause, because the time since isolation from the White Sea was enough for P. pungitius to lose/reduce its pelvic apparatus in these ponds (Ziuganov and Zotin 1995; Herczeg et al. unpubl. data). Hence, while evolution has clearly already occurred in these ponds, it does not apply to body size. This together with the fact that ponds with introduced predatory fish (S. trutta) exhibited lower or negligible size divergence from the marine and lake populations suggests that from the point-of-view of body size evolution, the biotic environment (i.e., predation and competition) is more important than the area of the habitat per se.

Reporting phenotypic body size trends from the wild does not tell us much about the functional (i.e., increased growth rate vs. increased longevity), or the evolutionary mechanism (i.e., local adaptation or phenotypic plasticity) behind the observed patterns. Analyzing size-at-age relationships showed that in the case of the Finnish and Swedish ponds, the maximal age of P. pungitius was two to three years higher and fish were also considerably larger at the same age than in marine or lake environments. In other words, pond fish grew faster and for longer periods than the marine and lake fish. Again, the Russian pond (Krugloje) was an exception: the size at a given age was similar to the marine and lake fish, while longevity was extended. This extended longevity might result simply from the low predation-caused mortality. However, it is noteworthy that—excluding Krugloje—the sizes at a given age were very consistent within the marine and lake fish and within the Finnish and Swedish ponds despite the often considerable geographic and genetic distances. Interestingly, the real giant fish were found in the sixth and seventh year classes in Pyöreälampi and Rytilampi showing a boost in growth after the fifth year. Intuitively, it might suggest some generation-specific environmental effects, but the fact that the existence of giant fish above 10 cm in total length in these populations has been observed during several years (Kuusela 2006; Merilä 2006; personal observations) contradicts this scenario. The cause and relevance of this “terminal gigantism” surely warrants further investigations. It is also noteworthy that we did not catch a single second year fish in the Finnish and Swedish ponds. The reason for this is unclear, but it might be that fish in the giant populations initiate breeding a year later—another pattern warranting further studies on the population dynamics of the giant populations. At any rate, this difference alone was clearly not enough to explain the habitat differences in body size.

Proving that the size differentiation observed in this study has a genetic basis is very important for any evolutionary inference, especially as it has been shown that divergence in certain ecological factors between island and mainland environments can result in profound intraspecific body size differentiation via phenotypic plasticity alone (e.g., Madsen and Shine 1993). In the common garden experiment, we found that pond fish grew two to three times larger in size than marine fish, irrespective of population origin. The four populations used in the common garden experiment are also known to be genetically isolated (Shikano et al. unpubl. data), and the within-habitat type (cf. pond vs. marine) replicates were separated by >500 km geographically. Such genetically based, repeated, and habitat-specific divergence strongly implies natural selection as the causal agent (e.g., Clarke 1975; Endler 1986; Schluter and Nagel 1995; McGuigan et al. 2005). The significant family effects within populations further suggest that there is also genetic variation within population in body size. Because first laboratory generation full-sib families were used, maternal and nonadditive effects cannot be isolated from additive genetic effects which need to be estimated with other types of breeding design (e.g., Lynch and Walsh 1998). However, considering the strength of differences (up to threefold, measured in 36-week-old fish), maternal and/or nonadditive effects are unlikely to account for the observed differentiation considering that maternal effects seldom explain any large interpopulation differences in growth and dissipate quickly during the early growth phase (e.g., Green 2008). Further, adult fish were kept at ad libitum food for weeks before the crosses were made, meaning that they did not face any energy limitation during that period.

In summary, we found that a small-sized teleost fish, P. pungitius, was generally larger in isolated ponds than in lakes or marine environments, sometimes reaching giant sizes. We demonstrated that this was a result of a combination of extended longevity and faster growth in the wild, and also that these patterns are likely to be genetically based as verified by a common garden experiment. We suggest that the relaxation of predation pressure and interspecific competition together with the increased intraspecific competition—and not habitat area per se—are the key factors behind insular body size evolution. Approaches such as those adopted in this study can provide important and complementary insights into the mechanisms—as well as their validity and generality—underlying body size–and island size correlations. According to Lomolino (2005), very large islands are “mainland like” in terms of predators and competitors. We propose that a single predator or larger and superior interspecific competitor species can transform even the smallest island into a “mainland like” habitat from the perspective of the selective pressures relevant for body size evolution.

Associate Editor: L. Kruuk


We thank Victor Berger, Göran Englund, Tuomas Leinonen, Daniel Lussetti, and Pirkko Siikamäki for their help in obtaining the samples. We are indebted to the Oulanka Research Station (University of Oulu) and the White Sea Biological Station (Russian Academy of Sciences) for their help with practicalities in the field. We thank Anders Forsman, Mark Lomolino, Shai Meiri, Loeske Kruuk, and three anonymous reviewers for their thorough comments leading to improvements of the manuscript, and John Loehr for correcting the English. We received financial support from the Academy of Finland, Ministry of Education and Centre for International Mobility (CIMO). The experiment was conducted under the license of the Helsinki University Animal Experimentation Committee.