• Eric B. Taylor,

    1. Department of Zoology, Beaty Biodiversity Research Centre and Museum, and Native Fishes Research Group, University of British Columbia, 6270 University Blvd., Vancouver, British Columbia V6T 1Z4, Canada
    2. E-mail: etaylor@zoology.ubc.ca
    Search for more papers by this author
  • Carling Gerlinsky,

    1. Department of Zoology, Beaty Biodiversity Research Centre and Museum, and Native Fishes Research Group, University of British Columbia, 6270 University Blvd., Vancouver, British Columbia V6T 1Z4, Canada
    2. Fisheries Centre, Aquatic Ecosystems Research Laboratory, 2202 Main Mall, University of British Columbia, Vancouver, British Columbia V6T 1Z4, Canada
    Search for more papers by this author
  • Nicole Farrell,

    1. Department of Zoology, Beaty Biodiversity Research Centre and Museum, and Native Fishes Research Group, University of British Columbia, 6270 University Blvd., Vancouver, British Columbia V6T 1Z4, Canada
    Search for more papers by this author
  • Jennifer L. Gow

    1. Department of Zoology, Beaty Biodiversity Research Centre and Museum, and Native Fishes Research Group, University of British Columbia, 6270 University Blvd., Vancouver, British Columbia V6T 1Z4, Canada
    Search for more papers by this author


Ecological speciation is the evolution of reproductive isolation as a direct or indirect consequence of divergent natural selection. Reduced performance of hybrids in nature is thought to be an important process by which natural selection can favor the evolution of assortative mating and drive speciation. Benthic and limnetic sympatric species of threespine stickleback (Gasterosteus aculeatus) are adapted to alternative trophic niches (bottom browsing vs. open water planktivory, respectively) and reduced feeding performance of hybrids is thought to have contributed to the evolution of reproductive isolation. We tested this “hybrid-disadvantage hypothesis” by inferring growth rates from otoliths sampled from wild, free-ranging benthic, limnetic, and hybrid sticklebacks in two lakes. There were significant differences in growth rate between lakes, life-history stages, and among years (maximum P = 0.02), as well as interactions between most factors, but not between hybrid and parental species sticklebacks in most comparisons. Our results provide little evidence of a growth disadvantage in hybrid sticklebacks when free-ranging in nature. Although trophic ecology per se may contribute less to ecological speciation than envisioned, it may act in concert with other aspects of stickleback biology, such as interactions with parasites, predators, competitors, and/or sexual selection, to present strong multifarious selection against hybrids.

Understanding the processes that drive speciation is a fundamental inquiry of evolutionary biology (Coyne and Orr 2004). Ecological speciation is a key deterministic process involved in speciation and is defined as the evolution of reproductive isolation owing to ecologically driven and divergent natural selection (Schluter 2000). Selection may directly favor reproductive isolation owing to hybrid disadvantage (i.e., via reinforcement) or indirectly as a byproduct of adaptation to contrasting environments (Schluter 2000). Evidence for the former includes instances of enhanced reproductive isolation when two species are sympatric compared to when they are allopatric (reviewed in Coyne and Orr 2004). Examples of the latter include reproductive isolation evolving as a byproduct of adaptation to different feeding substrates in Drosophila (Dodd 1989) or as a byproduct of divergence in body shape or size as an adaptation to distinct predator environments in Neotropical fish (e.g., Langerhans et al. 2007).

One system that has been studied heavily in terms of the role of ecology in speciation is that of the threespine stickleback (Gasterosteus aculeatus) “benthic” and “limnetic” species pairs (McPhail 1994; McKinnon and Rundle 2002). These small, freshwater fish are ecologically, morphologically, behaviorally, and genetically distinct in sympatry in several lakes in southwestern British Columbia, Canada (reviewed by McPhail 1994). They are, therefore, largely reproductively isolated from one another in sympatry and fulfill the criteria generally accepted to define biological species. The benthic form is a solitary, robust-bodied bottom browser of aquatic invertebrates in the littoral zone of lakes, whereas the limnetic is a schooling, slender-bodied planktivore of open waters. Consequently, the two species occupy distinct trophic niches (benthic browsing vs. offshore planktivory) and considerable evidence has accumulated indicating that the species are morphologically and behaviorally adapted to these alternative niches (e.g., Bentzen and McPhail 1984; Schluter and McPhail 1992; Schluter 1993, 1995; Matthews et al. 2010). In addition, laboratory-raised hybrids between the two species are morphologically and behaviorally intermediate relative to the parental species in traits related to foraging ecology (e.g., McPhail 1984, 1992), and are less proficient and grow more slowly when foraging in the niches of either parental species (e.g., Schluter 1995; Hatfield and Schluter 1999). Because increasing growth rate, particularly during early life-history stages of fishes, is typically associated with increased survivorship (by reducing susceptibility to gape-limited predators, increasing the potential prey base, or increasing overwinter survivorship), faster maturation, and in females, greater fecundity at a given age, growth rate is reasonably assumed to be an important component of fitness (e.g., Sogard 1997; Shima and Findley 2002; Berkeley et al. 2004; Morita and Nagasawa 2010). In summary, experimental evidence for fitness trade-offs experienced by one parental species in the trophic niche of the other and for the poor performance of hybrids in both parental niches is consistent with the idea that selection may favor mate discrimination owing to foraging-based hybrid growth disadvantage (e.g., Rundle and Schluter 1998; Schluter 2001) and, therefore, contribute to the evolution of reproductive isolation. In addition, genetic cohort analyses have shown a decline in abundance of hybrid genotypes with age in natural stickleback populations (Gow et al. 2007). This observation, coupled with evidence that survivorship of hybrid genotypes of various ages raised under laboratory conditions is no different than that of parental species (McPhail 1994; Hatfield and Schluter 1999) also implies that there is an ecological component to genetic divergence and reproductive isolation between benthic and limnetic sticklebacks (rather than a prominent role for intrinsic genomic incompatibilities).

Although these data provide compelling evidence consistent with the idea that trophically based natural selection against hybrids has contributed to the evolution of reproductive isolation, reproductive isolation may evolve as an incidental byproduct of trophic adaptation in sticklebacks (Vines and Schluter 2006) and perhaps by sexual selection against hybrids (Vamosi and Schluter 1999). Further, most of the evidence for a role for hybrid growth disadvantage in stickleback speciation is based on laboratory studies or micro/mesocosom field studies and direct tests of hybrid growth disadvantage under fully natural conditions are lacking. Mark-recapture studies of individual free-ranging sticklebacks in nature would be one way to collect such data, but they are logistically challenging. An alternative way to assess growth of wild-caught fish is to infer growth rates from otolith analysis.

Otoliths are calcareous structures found in the inner ear of teleost fishes (Campana and Thorrold 2001). The otoliths occur as three pairs on either side of the head of which the sagittal otoliths are the largest. Otoliths accrete layers of calcium carbonate throughout the life of a fish and the accretion rate and growth of the otolith varies with growth rate of the fish. In addition, accretion rate varies on a daily basis such that variation in daily growth rate of the fish is reflected in variation in daily growth rate of the otolith (Campana and Thorrold 2001). Daily variation in otolith growth appears as a series of alternating dark and light bands radiating out from the center of the otolith with one pair of light and dark bands representing 24 h. These patterns have been validated experimentally in many species, and once validated, it is possible to determine the age of a fish from its otoliths by counting the “daily rings.” In addition, because otolith growth is related to overall fish growth, it is possible to estimate the growth rate of a fish by measuring the size of an otolith over a given time period (i.e., days). Daily growth ring analysis has been widely used in studies of the biology and management of many fishes (e.g., reviewed by Campana 2005). In this study, we examined growth of wild, free-ranging benthic, limnetic, and hybrid sticklebacks inferred from the daily growth rings of otoliths to test the hypothesis of hybrid growth disadvantage—a key prediction of trophic-based ecological speciation in these fish.

Materials and Methods


To assess the utility of otolith growth rate as a proxy for overall fish growth, we made experimental crosses between marine fish from the Little Campbell River (located about 40 km southeast of Vancouver, British Columbia [BC], Canada, 49o04'44.37”N, 122o49'58.34”W) and fish from Cranby Lake (located on Texada Island in the southern Strait of Georgia region, southwestern BC, 49o41'44.15”N, 124o30'28.05”W) and raised them in the laboratory under different growth regimes. Methods of fish crossing and subsequent incubation and husbandry can be found at http://www.zoology.ubc.ca/~schluter/lab.methods.html. At 18 days of age posthatching, fish were randomly assigned to fast-growing and slow-growing treatments (N = 15 fish each). The slow-growing fish were kept in a smaller net pen inside the larger tank and fed half as often to induce a slower growth rate. The fish were then sacrificed at 69 days old and their otoliths were examined to compare known age with otolith ring counts and differences in growth rate with otolith growth rate. Growth rate was expressed as the change in length from hatching (4–5 mm in length) until 69 days of age.

Otoliths were then examined in eight groups of wild-caught stickleback sampled from Priest and Paxton lakes (Table 1). Priest Lake (49o44′42.39″N, 124o33′55.29″W) and Paxton Lake (49o42′37.38″N, 124o31′31.33″W) each contain a benthic–limnetic species pair and are part of independent watersheds located on Texada Island (see McPhail 1992; Taylor and McPhail 2000). These samples consisted of juvenile (age 0+) and subadult (between 1+ and prespawning fish) sticklebacks collected between 2003 and 2006. Juvenile sticklebacks were sampled using dip nets while canoeing along the littoral open water transition zone and ranged from 5–20 mm total length (TL). All subadult fish were sampled using minnow traps set within the littoral zone of each lake and ranged from 30–45 mm TL. All samples were obtained during May through October of each year as described by Gow et al. (2007). We were unable to include the adult fish examined by Gow et al. (2007) owing to small sample sizes of hybrid fish at this life-history stage (38 over four years and two lakes).

Table 1.  Summary of sample sizes for wild, free-ranging limnetic (NL), benthic (NB), and hybrid (NH) threespine sticklebacks (Gasterosteus aculeatus) sampled from Priest and Paxton lakes and examined for growth rate. Juveniles are in their first year of growth, subadults are in the second year of growth.
LakeYearLife-history stageSample sizes
Priest2004Juveniles2614 0
Priest2004Subadults12 919


All fish used in the otolith analysis were previously characterized genetically in the microsatellite analysis of Gow et al. (2007) that used a panel of nearly diagnostic microsatellite loci to identify benthics, limnetics, and their hybrids. Briefly, the genotype of each fish was summarized by calculating a “hybridity” coefficient, h, as,


where inline image is the average proportion of ancestry in the benthic population (calculated across five replicate analyses of genetic admixture of each fish's 10 microsatellite locus profile using STRUCTURE (Pritchard et al. 2000; Gow et al. 2007). The value of h ranges from 0 for parentals to 0.5 for F1 hybrids, and provides a measure of how intermediate an individual fish's multilocus genotype is on the microsatellite admixture scale. Fish with hybridity scores of 0 were previously identified as benthics or limnetics using the inline image genetic admixture scores (limnetics = 0, benthics = 1) and for each group (i.e., year × lake × life-history stage) used in the growth analysis we sampled as even a number as possible of parental limnetics and benthics (typically about 10–15 each). In addition to the inline image value for each fish, we created either two or three levels of hybridity for subsequent analyses: (1) all fish with h values of 0 (i.e., parental genotypes), and (2) all fish with h values of between 0.1 and 0.5 were pooled in a single category of “hybrids” (i.e., they represented somewhere between 10% and 50% admixture of benthic and limnetic gene pools), or (1) all fish with h values of 0 (parentals), (2) fish with h values between 0.1 and 0.29, and (3) fish with h values between 0.3 and 0.5.


Sagittal otoliths were sampled by removing the head of each fish, just behind the operculum, and then removing the lower jaw and gills with forceps. The remaining head portion was then placed dorsal-side-down on a glass slide, under a dissecting microscope, and held in place with one finger while prying back the “roof-of-mouth” tissue with forceps to reveal the otoliths encased in a thin layer of tissue on either side of the head. The sagittal otoliths were removed with forceps and placed immediately into a small drop of water on a glass slide. If tissue was present around the otolith it was then teased away with the otolith immersed in water to prevent it from flicking off the slide. Once the otoliths were cleaned they were immersed in 95% ethanol.

Methods of otolith preparation were modified from the general procedures described in Wright and Huntingford (1993). In preparation for mounting, small pieces of thermoplastic cement resin (∼2 mm chunks) were heated onto glass slides on a hot plate (set at 200°C). The glass slide with the softened resin dot was then transferred to the dissecting scope, and one otolith (sulcus groove side down) was transferred into the resin dot using forceps. The slide was then returned to the hot plate until the resin encased the otolith. The resin-encased otolith was then ground down using diamond-grit (3 and 30 μm) sanding paper under a 200×– 400× inverted microscope. Finally, a drop of immersion oil was placed on the sanded otolith to enhance the appearance of the rings.


A series of images of each otolith were taken using Auto-Montage Pro Image analysis software (Syncroscopy, Frederick, MD) and an inverted compound microscope fitted with a camera. A magnification level of 100×, 200×, or 400× was used depending on the size of the otolith. The radius of the otolith opposite the sulcus (Fig. 1) was measured after calibration using the image analysis software. Ring counts were made about two to four times for each otolith depending on overall clarity of the rings and the number of images needed to see all the rings clearly. The total ring count was started at the dark line designated as the “hatching activity check” that occurred at approximately 30 μm from the central primordium of the otolith, and ring count ended at the outer edge opposite the sulcus (Fig. 1; see also Wright and Huntingford 1993). The radius of the otolith was divided by the total ring count (representing days) to obtain the “otolith growth rate” in μm/day, which was used as a surrogate measure of overall fish growth rate (Campana et al. 1987; Wright and Huntingford 1993). If good quality counts were attainable on both otoliths, the two were then averaged, otherwise only the count associated with the clearest image was used. For a subsample of all fish examined, we also measured head length (tip of snout to posterior edge of gill cover) and standard length (tip of snout to beginning of caudal fin).

Figure 1.

Image of threespine stickleback (Gasterosteus aculeatus) otolith showing primordium center (white dot) and daily increments (x). The sulcus was used to orient the maximum otolith radius (r). Age was estimated by counting all daily growth rings along the radius after the hatching activity check (white solid arrow).


Differences in expected and actual mean daily ring counts in the 69-day-old laboratory-raised fish were tested using two sample t-tests (after verifying that variances were not significantly different) as were differences in mean otolith radius and head length using PAST (Hammer et al. 2001), a general spreadsheet-based statistical analysis package. Growth rates of otoliths in μm/day were expressed as the sulcus radius divided by the daily ring count (or average ring count if two counts were taken from the same fish). Otolith growth rates for individual fish were examined across years (2003, 2004, 2005, 2006), lakes (Priest or Paxton lake), life-history stages (juveniles, subadults), and hybridity value (grouped into two levels: h = 0, h = 0.1–0.5, and into three levels: h = 0, h = 0.1–0.29, and h = 0.3–0.5) in a four-factor ANOVA where all factors except hybridity were considered as fixed effects. We used ANOVA function (aov) in R to conduct the multifactor analysis of variance (R Development Core Team 2009). We also used regression analyses to evaluate linear or curvilinear relationships between inferred growth rate and the admixture coefficient, inline image, using PAST. In these latter analyses, we pooled inline image and growth rate estimates across years within each lake and within life-history stages for a total of four analyses. The “best” model was chosen based on a combination of statistical significance, assessments of model fit from the coefficient of determination and the chi-square statistic, subject to guarding against model over-fitting by changes in the Akaike Information Criterion (where models that deviated from the next simplest model by ≥2 AIC units were discarded, Hammer et al. 2001; Burnham and Anderson 2002; Arnold 2010). These analyses were conducted on the original growth rate data because log transformations did not change the results in any material way.



Newly hatched sticklebacks were between 4 and 5 mm in TL and reached a size of 6 mm TL when they began feeding exogenously. Fish were raised in the laboratory for a total of 69 days from hatching and demonstrated a clear relationship between known age and otolith daily ring count as well as between otolith size and overall body growth. The laboratory-raised fish in the “fast growth” group achieved a TL of about 18–22 mm after 69 days and the average number of otolith “daily” rings in these fish was 61.3 (SD = 3.97, N = 15). The “slow growth” group achieved a TL of 13–16 mm over the same time and had an average of 57.6 daily rings (SD = 15.4, N = 12). Daily otolith growth increment counts tended to underestimate the true age of fish both in the fast- and slow-growing groups. The difference between the expected (69) and actual daily ring count averaged 7.8 and 11.1 fewer rings in the fast and slow growth fish, respectively, both of which were significantly greater than 0 (one sample t-tests, both P < 0.001), but not from each other (P = 0.058). The otolith radius in the fast-growing group was significantly larger (mean = 213.1 μm, SD = 10.5 μm) than that in the slow-growing group (mean = 179 μm, SD = 47.7 μm, P < 0.001). The greater growth of the otoliths over 69 days in the fast growth group was also reflected in their longer heads compared to the slow-growing group (as a measure of overall body size and growth rate, mean = 5.36 mm, SD = 0.47 mm vs. 4.08 mm SD = 1.01 mm, respectively, P < 0.001). Overall, otolith growth rate was estimated as 3.09 μm/day (SD = 0.15) and 2.61 μm/day (SD = 0.14) in the fast- and slow-growing groups, respectively (t-test, P < 0.001).


Otoliths were collected and analyzed from a total of 364 wild threespine sticklebacks across the four years, two lakes, and two life-history stages. A total of 214 of these fish had hybridity values of 0, of which 118 were genetically classified as limnetics (inline image= 0) and 96 were benthics (inline image= 1.0, Table 1). The remaining 150 fish had an approximately uniform distribution of hybridity values of between 0.10 and 0.50 (Table 1, Fig. 2). We estimated the growth rates for sticklebacks across the various sample groups over a range of between 40 and 79 days (as estimated from daily growth increment counts). The average age over which growth was estimated for juvenile parentals (benthics and limnetics) and hybrids were not significantly different from one another (mean ± SD = 55.2 ± 14.1 vs. 54.0 ± 14.0, t-test, P > 0.5, Fig. S1). The average age over which growth was estimated for subadult fish was older than that of juveniles, but again these ages were not different among parental groups (benthics and limnetics) and hybrids (mean ± SD = 78.3 ± 10.6 vs. 77.0 ± 12.7, t-test, P > 0.5, Fig. S1). Within any one analysis group, the estimated mean ages of the parental and hybrid sticklebacks were always within four days of each other and in four of the treatments groups within one day of each other (i.e., daily otolith growth increment counts differed between one and four days). For fish with hybridity values equal to or greater than 0.1, hybridity was slightly higher in Priest Lake (0.369, SD = 0.100, N = 79) than in Paxton Lake (0.268, SD = 0.137, N = 78). Over all samples, there was a small negative correlation between hybridity values (i.e., h between 0.1 and 0.5) and otolith growth rate (rs=−0.17, P = 0.031, N = 150).

Figure 2.

Distribution of hybridity values (0 = parental limnetic or parental benthic, 0.5 = 50% admixture) of 364 wild-caught threespine sticklebacks (Gasterosteus aculeatus) examined for otolith growth.

Otolith growth rates of wild-caught fish were generally much higher than the laboratory-raised fish and ranged from 2.3 μm/day to 6.25 μm/day and were approximately normally distributed, but with a significantly positive skew (Wilks–Shapiro test, P < 0.001, Fig. 3). Over all samples, younger fish tended to grow faster than older fish; there was a modest, but significant, negative correlation between estimated total age and daily otolith growth rate (rs =−0.39, P < 0.001, Fig. S2). In addition, we observed a positive correlation between otolith-estimated growth rate and standard length at sampling (rs = 0.41, N = 208, P < 0.0001).

Figure 3.

Distribution of otolith inferred growth rates (μm/day) for 364 wild-caught threespine sticklebacks (Gasterosteus aculeatus). The “S” and “F” represent the mean otolith growth rates estimated for the slow-growing and fast-growing groups of lab-raised fish, respectively.

The four factor analysis of variance found significant effects of Year, Life-history stage, and Lake on inferred otolith growth rates (maximum P = 0.02), but there was no effect of Hybridity (two h levels: 0 and 0.1–0.5, P = 0.31) (Table S1 and Fig. 4). There were, however, significant interactions among the main factors; the effect of Year was dependent on Lake, there was an effect of Hybridity dependant on Life-history stage, as well as an effect of the combined effects of Year, Life-history stage, and Hybridity (Table S1, all P < 0.05). Of the first-order interactions, the main driver appeared to be the generally higher otolith growth rate of sticklebacks collected from Priest Lake (Fig. 4). Similarly, the relative growth of hybrid and parental genotypes was dependent on life-history stage; overall, hybrids grew faster than parentals as subadults, but at the same rate or slightly slower as juveniles (Table 2). Although growth rate of hybrids as subadults was consistently higher than for parentals in 2004, 2005, and 2006, the reverse was true for juvenile sticklebacks, except in 2006 when juvenile hybrids grew at the same rate or at a slightly greater rate than parental genotypes (Table 2). Across all eight groups, benthics tended to have higher otolith growth rates than limnetics, but growth rates of hybrids were lower than both parentals in only two instances (Table 2).

Figure 4.

Mean (±SD) otolith inferred growth rates (μm/day) for parental wild-caught limnetic or benthic threespine sticklebacks (Gasterosteus aculeatus, gray bars), sticklebacks of hybridity 0.1–0.29 (thick hatching), and sticklebacks of hybridity 0.30–0.5 (black bars) in Paxton and Priest lakes pooled across years and between life-history stages. Sample sizes (right to left) are 106, 39, 34, 111, 10, and 72, respectively.

Table 2.  Mean (±SD) growth rates (μm/day) for wild-caught free-ranging juvenile and subadult limnetic (GRL), benthic (GRB), pooled parental (GRP), and hybrid (GRH) threespine sticklebacks (Gasterosteus aculeautus) inferred from otoliths. NA = not applicable (no samples). Samples sizes are given in Table 1.
 Paxton20053.61 (0.31)3.60 (0.36)3.60 (0.32)3.43 (0.44)
 Paxton20063.42 (0.31)3.43 (0.23)3.43 (0.26)3.56 (0.38)
 Priest20043.98 (0.39)NANA3.74 (0.26)
 Priest20063.16 (0.39)3.33 (0.26)3.24 (0.33)3.38 (0.31)
 Paxton20033.84 (0.64)3.98 (0.24)3.92 (0.44)4.02 (0.38)
 Paxton20053.63 (0.56)3.59 (0.26)3.61 (0.43)3.74 (0.39)
 Priest20044.04 (0.35)4.39 (0.50)4.22 (0.45)4.48 (0.69)
 Priest20063.62 (0.44)3.54 (0.34)3.58 (0.38)3.86 (0.71)

The general results were very similar when Hybridity was considered across three levels (h = 0, h = 0.1–0.29, and h = 0.3–0.5) in terms of the main effects and interactions although levels of significance tended to be higher (Table S2). The effect of Hybridity, however, approached significance (P = 0.056) and there was an additional significant interaction between Lake and Hybridity (P = 0.018). In Paxton Lake, otolith growth rate tended to be highest in the h = 0.29–0.5 group, but in Priest Lake, variation was greater and fish in the h = 0.1–0.29 group had the highest otolith growth rates (Fig. 4).

When we measured admixture across a scale of inline image= 0–1, we detected a significant (P = 0.003, r2= 0.04) cubic relationship between inferred growth rate and genetic admixture (Fig. S3 and Table S3); and there was a slight tendency for benthic backcrosses (i.e., inline image= 0.7–0.8) to have the highest growth rates, whereas limnetic backcrosses (inline image= 0.1–0.3) tended to have the lowest. A simple linear model between genetic admixture and inferred growth rate was best in Paxton Lake, with benthics tending to be the fastest growers, limnetics the slowest and hybrids intermediate, although only the relationship in subadults was significant (P = 0.006, Fig. 5, Table S3). In Priest Lake, second-order polynomials tended to provide the best fit, with hybrids exhibiting the highest growth rates, although only the relationship in subadults approached significance (P = 0.06, Fig. 6, Table S3).

Figure 5.

Plots of genetic admixture coefficient (inline image, 0 = limnetic, 1.0 = benthic) versus inferred growth rate (um/day) for wild, free-ranging (A) juvenile and (B) subadult threespine sticklebacks (Gasterosteus aculeatus) sampled from Paxton Lake pooled across years. The mean growth rates are shown for limnetics (large open circle), hybrids (inline image between 0.1 and 0.9, large gray circle), and benthics (large closed circle).

Figure 6.

Plots of genetic admixture coefficient (inline image, 0 = limnetic, 1.0 = benthic) versus inferred growth rate (um/day) for wild, free-ranging (A) juvenile and (B) subadult threespine sticklebacks (Gasterosteus aculeatus) sampled from Priest Lake pooled across years. The mean growth rates are shown for limnetics (large open circle), hybrids (inline image between 0.1 and 0.9, large gray circle), and benthics (large closed circle).

Finally, head length and standard length measurements for a total of 204 Paxton Lake samples indicated that, in general, head lengths adjusted for differences in standard length (by analysis of covariance) tended to be largest for benthics, smallest for limnetics, and intermediate for hybrid sticklebacks (Table S4). These differences tended to be most apparent for subadults and absent or slight for juvenile samples, but in no case were hybrids significantly different both from benthics and limnetics (Table S4).



In our study, we raised juvenile stickleback from eggs fertilized in the laboratory under two growth conditions to: (1) validate that the growth rings that we counted were laid down at a rate of approximately once every 24 h, and (2) assess if otolith growth rate (size after a given number of days) was a valid proxy for overall body growth rate. In our laboratory-raised fish, we found that the age estimated by counting “daily” rings in the otoliths underestimated the known age by about eight days. Wright and Huntingford (1993) raised sticklebacks under laboratory and outdoor enclosure conditions and found that known and otolith estimated ages differed by an average of only about one day over a similar time period. The closer correspondence between daily ring count and known age in the latter study may have been due to the differences in temperature regimes (15°C in the laboratory and 2–20°C in the enclosures in Wright and Huntingford [1993] vs. 17°C in our study, respectively) and/or to the use of scanning electron microscopy (SEM) in some of the Wright and Huntingford (1993) samples. The use of SEM is known to provide greater resolution for closely spaced rings or in slowly growing individuals (e.g., Campana et al. 1987). Our laboratory-raised fish also demonstrated that fish held on a reduced ration had smaller body and head lengths, as well as smaller otolith radii than fish fed on a larger ration, indicating that otolith growth is a valid proxy for overall body growth rate. In addition, the slower growing fish showed a greater deviation (0.83:1 vs. 0.89:1) from the expected 1:1 relationship between otolith daily ring count and known age which is consistent with empirical observations of a poorer correspondence between daily otolith growth ring production and fish growth when fish grow slowly (e.g., Campana et al. 1987; Wright and Huntingford 1993). The poorer correspondence between ring count and known age in our slower growing laboratory fish might suggest that, in nature, their growth rates could be overestimated (because their age would be underestimated). The free-ranging wild fish that we examined, however, had estimated growth rates that were in excess of that both for fast- and slow-growing laboratory raised fish suggesting that otolith ring formation would closely approximate daily deposition in both groups of fish in nature. Altogether, our data support the use of daily growth increments to assess relative growth differences among individual sticklebacks (see also Wright and Huntingford 1993).


That hybrid sticklebacks tend to show reduced feeding performance relative to parental limnetics and benthics is a well-established observation. Initially, these observations were based on predictions stemming from the genetically controlled morphological differences between the species and functional morphological studies in fishes that have linked differential feeding performance, in terms of food particle sizes ingested, habitats exploited, and behavior of prey, to morphological differences (reviewed by Robinson and Wilson 1994; see also Langerhans et al. 2003; Carroll et al. 2004). Limnetic sticklebacks have more gill rakers, relatively larger, more protruding eyes, and a more slender snout compared to benthics, and F1 hybrids are morphologically intermediate to the parental species (McPhail 1984, 1992; Hatfield and Schluter 1999).

In sticklebacks, detailed studies of diet and foraging experiments demonstrated that limnetics and benthics tend to specialize on different food items in nature (limnetics on copepods, cladocerans, etc, and benthics on amphipods, isopods) and that each is better suited morphologically and behaviorally to exploit these distinct trophic niches (Bentzen and McPhail 1984; Schluter and McPhail 1992; Vamosi et al. 2000; Matthews et al. 2010). Consequently, there tends to be a strong association between morphology and diet in limnetic and benthic species pairs (although the strength of this association may vary among lake pairs—see Behm et al. 2010). In addition, benthic × limnetic F1 hybrids ingest prey of a maximum size that is intermediate to that of the parental species (Bentzen and McPhail 1984). These functional relationships between morphology and feeding performance have also been consistently associated with growth performance trade-offs in reciprocal habitat enclosure experiments, again with hybrids tending to grow at a lower rate than limnetics or benthics in their respective open water and littoral environments (e.g., Schluter 1993, 1995). These morphology-growth-fitness trade-offs have provided key evidence for ecological speciation in sticklebacks and have been suggested to explain, in part, how benthics and limnetics coexist as distinct gene pools in sympatry despite low, but persistent gene flow and the lack of intrinsic genomic incompatibilities between them (e.g., McPhail 1984, 1992; Hatfield and Schluter 1999; Rundle and Nosil 2005; Gow et al. 2007). Specifically, reduced foraging performance of hybrids may contribute to the evolution of reproductive isolation via selection for homotypic matings in adults (Schluter 2001; Rundle and Nosil 2005).

All of this previous work, however, has been conducted either in laboratory settings, experimental ponds, or in field enclosures (e.g., Bentzen and McPhail 1984; Schluter 1995; Vamosi et al. 2000), whereas more direct tests of foraging performance of wild, free-ranging hybrid sticklebacks relative to limnetics and benthics have never been attempted. Our study contributes to filling this gap in our understanding of the performance of hybrid sticklebacks in nature. Under the trophic “hybrid disadvantage” hypothesis, we expected that growth rates of hybrid sticklebacks would be demonstrably and consistently lower than one or both parental species. We did observe a low, but significant, negative correlation between hybridity and growth rate overall, but our samples were not balanced across age classes, lakes, or years. Consequently, these other factors (which clearly influenced growth rate) were confounded with hybridity in our study. When growth rate is examined in more detail across these factors, consistently reduced growth performance of hybrid sticklebacks was not observed.

Our results are consistent with those of Vamosi et al. (2000) who conducted large-scale (2400–5000 fish) introductions of juvenile (about one month old) Paxton Lake benthic, limnetic, and F1 hybrid sticklebacks into experimental ponds and field enclosures. These experiments ran for a three-month period and found that hybrids had equal (in lake enclosures) or only marginally lower growth rates (in experimental ponds) than benthics, whereas limnetics had the lowest growth rates (Vamosi et al. 2000). Our analysis of inferred growth rates of hybrid sticklebacks and those of Vamosi et al. (2000) measuring growth directly, therefore, both point to the lack of consistently poorer growth performance of hybrid sticklebacks under completely natural conditions (our study) or in experimental ponds.

Several factors may explain our observation of a lack of reduced growth performance of hybrid sticklebacks. First, we inferred growth rates over a relatively short time frame (∼60 days) and it is possible that when growth is assessed over the full life span of a stickleback (one–two years in our study populations), more consistent differences in growth rates would be observed. Still, in experimental field enclosures, consistent differences in growth rate have been observed across much shorter time frames (approximately two–three weeks, Schluter 1995; Hatfield and Schluter 1999). Second, our hybrids were a mix of genotypes, from F1 hybrids (hybridity of 0.5, ∼30% of all hybrids) to advanced generation hybrids or backcrosses (hybridity of 0.1–0.49). Genetic control of at least some of the trophically relevant morphological variation in sticklebacks tends to be polygenic (McPhail 1992; Peichel et al. 2001; Schluter et al. 2004; Albert et al. 2008), and morphological variance tends to be higher in hybrid crosses than within the parental species (e.g., Schluter et al. 2004). Consequently, the hybrid sticklebacks in our analyses probably represented an array of morphological types and trophic performance which presumably would increase the variability in growth relative to strictly F1 hybrids typically studied previously (e.g., Schluter 1995; Hatfield and Schluter 1999; Vamosi et al. 2000). Still, when we analyzed our inferred growth rates across three classes of hybridity, the fish with hybridity values most closely approximating F1 hybrids (i.e., h = 0.3–0.5) did not show depressed growth rates. Third, we inferred growth rates in juvenile and subadult sticklebacks (i.e., all fish were less than 40 mm TL, many were less than 22 mm) whereas most studies of morphological differences, feeding behavior, and growth have been conducted exclusively on larger fish (usually 35–60 mm TL). Given that trophically relevant morphology can change with ontogeny (e.g., MacNeill and Brandt 1990; Machado-Allison and Carmen Garcia 1986; Wainwright et al. 1991), it is possible that trophic character differences between benthics and limnetics (and therefore hybrids) are not fully expressed until a certain size, perhaps at least 40 mm (Day et al. 1994). Consequently, consistent differences in growth rate among benthics, limnetics, and their hybrids may not be observable until a specific minimum size has been reached. Indeed, diet differentiation between benthic and limnetic sticklebacks increases with body size (Matthews et al. 2010) and morphology in fishes may be plastic within individuals depending on the life-stage sampled, particularly for species that undergo a habitat shift (e.g., Nicieza 1995; Hjelm et al. 2003). All the genotypes we sampled in this study were collected from the same habitats; either via dip nets in the open water—littoral transition zone or by minnow traps in the littoral zone. The fact that we captured benthics, limnetics (in roughly equal proportion), and hybrids in the same habitats suggests that the full suite of morphological and/or behavioral components of the limnetic and benthic trophic phenotypes may not yet have developed in these fish. We were able to examine head length within some of our Paxton Lake samples which, indeed, suggested that differences between benthics and limnetics, and the intermediate nature of hybrids, are greater at the subadult stage than at the juvenile stage. Head length is, however, only one aspect of trophic morphology with somewhat inconsistent differences reported for benthics and limnetics (e.g., McPhail 1992). The ontogeny of morphological differences and their impact on feeding performance would be an interesting avenue of research in sticklebacks. Our adult samples, however, contained too few hybrids (38 total across the two lakes and four years) for analysis at this life-history stage, but genetic cohort analysis indicates that some selection against hybrids occurs prior to the adult life-history stage (Gow et al. 2007). Finally, the open water—littoral transition zone from which our juvenile samples originated could represent a trophically intermediate habitat at least in terms of spatial access both to limnetic and benthic food resources. Consequently, it is possible that intermediate hybrid phenotypes might perform relatively well in this habitat, at least in terms of feeding and growth, relative to strictly open water or littoral habitats, and performance of hybrids equal to or greater than that of parental species has been observed in several taxa under various environmental conditions (Arnold 1997; Arnold and Martin 2010). Behm et al. (2010) suggested that trophic ecological selection against hybrid sticklebacks appears to have disintegrated in Enos Lake as a result of changes to benthic and limnetic habitats following the appearance of an invasive crayfish (Taylor et al. 2006). Indeed, hybrid phenotypes have increased in frequency in this lake (Kraak et al. 2001; Taylor et al. 2006) suggesting that hybrid performance may depend critically on the distribution and/or relative abundances of trophic resources (Behm et al. 2010). The relative abundance of intermediate habitats and the ecological performance of hybrid phenotypes inhabiting them is a relatively understudied aspect of ecological speciation and adaptive radiation more generally (Hatfield and Schluter 1999; Schluter 2000; Seehausen 2004; Rundle and Nosil 2005).


Despite our inability to substantiate a prediction of trophically based ecological speciation in threespine sticklebacks (i.e., reduced hybrid growth performance), our results do not mean that ecological factors, including trophic ecology, are an unimportant component of speciation in this system. In addition to the caveats specific to feeding ecology discussed above, our study spanned only a four-year period and might not represent the long-term, average relative fitness of hybrid and parental genotypes. Arnold and Martin (2010) recently reviewed results from the relatively few studies of long-term studies of hybrid fitness that showed that the relative fitness of genotypes can fluctuate across generations. Perhaps our study period was associated with environmental or habitat conditions where selection against hybrids via trophic interactions may have been relaxed or variable.

In addition, other ecological interactions involving sticklebacks such as predation and parasitism may influence the relative fitness of hybrids. Although greater susceptibility of hybrids to predators has been demonstrated in some systems (Wahl and Stein 1989; Mallet et al. 1998), the only direct test in sticklebacks did not resolve higher susceptibility of hybrids to a fish predator in pond experiments (Vamosi and Schluter 2002). Similarly, in some situations, hybrid genotypes can be more susceptible to parasites and pathogens (Brun et al. 1992; Hjältén 1998), and parasites have been implicated in population divergence in sticklebacks (e.g., MacColl 2009; MacColl and Chapman 2010), but their role on hybrid fitness in benthic and limnetic sticklebacks is unexplored. One study of general susceptibility to an eye fluke and in general immunological response demonstrated that hybrids between two German populations of threespine stickleback were more susceptible than one of the parental populations (Kalbe and Kurtz 2006), but another study on major histocompatibility complex (MHC) variation and parasite load between river and lake phenotypes and their hybrids failed to show any evidence of differential effects on hybrids in either parental habitat (Rauch et al. 2006). Finally, sexual selection has been implicated in ecological speciation via divergent selection on secondary sexual traits or mate communication systems (Schluter 2000; Coyne and Orr 2004; Maan et al. 2006). Contrasting visual environments characterize at least some benthic and limnetic stickleback habitats (Boughman 2001), but again, evidence of reduced mating performance of hybrid sticklebacks has been reported in some instances (e.g., Vamosi and Schluter 1999), but not others (Hatfield and Schluter 1996; Raeymaekers et al. 2010).

In conclusion, we found little evidence of one of the key aspects of ecologically based divergent selection—reduced hybrid growth performance—as a component of ecological speciation (e.g., Schluter 2001, 2009; Hendry 2009). Our study adds to a growing body of literature on the stickleback system of divergent sympatric and parapatric pairs suggesting that extensions of foundational work in laboratory or seminatural systems to situations that more closely approximate conditions in nature are not always straightforward (e.g., see Raeymaekers et al. 2010). In addition, our work contributes to the idea that selection against hybrids involving single factors (e.g., feeding, mating, parasites, predators) may be slight or undetectable under most natural conditions. Rather, much stronger selection against hybrids may stem from a multitude of factors that act independently and/or in combination with one another (Schluter 2000; Nosil 2007; Raeymaekers et al. 2010). As with many other examples of multiple causality in nature, detecting and evaluating the relative magnitude of such factors will be a challenging, but important step in understanding the processes involved in ecological speciation (Quinn and Dunham 1983; Räsänen and Hendry 2008).

Associate Editor: C. Jiggins


We thank P. Tamkee and A. Smith for crossing and rearing the juvenile sticklebacks for the laboratory component of this study and R. Markel for assistance with otolith analyses. J-S. Moore, J. Mee, S. Dick, C. Crossman, D. Schluter, T. Ingram, and two reviewers provided helpful comments on the manuscript. Funding for this research was provided by a Natural Sciences and Engineering Research Council of Canada Discovery Grant awarded to EBT. N. Farrell was supported by an Environment Canada Science Horizons Grant. We thank L. Bernatchez for suggesting we examine growth rates via otoliths.