1999 mclaughlin et al

ORIGINAL ARTICLE

Robert L. McLaughlin á Moira M. FergusonDavid L.G. Noakes

Adaptive peaks and alternative foraging tactics in brook charr:evidence of short-term divergent selection for sitting-and-waitingand actively searching

Received: 27 July 1998 /Accepted after revision: 16 November 1998

Abstract Some recently emerged brook charr (Salve-linus fontinalis) inhabiting still-water pools along thesides of streams are sedentary and eat crustaceans fromthe lower portion of the water column. Others are moreactive and eat insects from the upper portion of thewater column. We provide evidence that this divergentforaging behavior re¯ects short-term divergent selectionbrought about by intraspeci®c competition in the pres-ence of alternative food sources. Rates of encountersand interactions between individuals were density de-pendent, and encounter and interaction events wereclosely timed with prey capture attempts. In addition,aggressive ®sh made more foraging attempts per minutethan nonaggressive ®sh. Aggressive ®sh were also eitherinactive or very active, while nonaggressive ®sh exhib-ited intermediate levels of activity. Growth rate poten-tial, an important component of ®tness during the earlylife stages of brook charr, was assessed using tissueconcentrations of RNA and found to be highest forsedentary ®sh and for active ®sh making frequent for-aging attempts, and lower for ®sh exhibiting intermedi-ate levels of activity. Our ®ndings support contentionsthat individual behavior plays an important role duringinitial steps in the evolution of resource polymorphisms.

Key words Charr á Divergent selection á Fish growthForaging tactics á Resource polymorphism

Introduction

In many vertebrate populations there is considerable,and sometimes apparently discrete, variation in feedingmorphology and behavior (Robinson and Wilson 1994

Wimberger 1994; Sku lason and Smith 1995; Smith andSku lason 1996). These resource or trophic polymor-phisms are often accompanied by corresponding di�er-ences in growth rate, age at maturity, and matingstrategies (Sku lason and Smith 1995; Smith and Sku la-son 1996). Resource polymorphisms are currently at-tracting much interest because they o�er an unparalleledopportunity to examine the roles that ecological andbehavioral processes (e.g., competition and resourcepartitioning) play in evolutionary processes (e.g., pop-ulation divergence, sympatric speciation, and adaptiveradiation). Resource polymorphisms occur in every classof vertebrate (Wimberger 1994), and recent reviewssuggest they are more common than originally appreci-ated (Robinson and Wilson 1994; Sku lason and Smith1995; Smith and Sku lason 1996). They are particularlycommon in ®shes inhabiting species-poor lakes withwell-de®ned benthic and limnetic niches (Robinson andWilson 1994).

Behavior is thought to be a crucial aspect of theevolution of resource polymorphisms (Wimberger 1994Sku lason and Smith 1995). In species-poor environ-ments with two or more available niches, morphologicaldivergence and has been proposed to occur in three steps(Wimberger 1994; Sku lason and Smith 1995). First, thepresence of stable, alternative food sources requiringdi�erent techniques for exploitation is expected to favorspecialization in the foraging behavior exhibited by be-haviorally ¯exible individuals. The specialization may beenhanced by biological interactions, and comparativeevidence suggests competition is the most likely form ofinteraction promoting the diversi®cation (Robinson andWilson 1994; Schluter 1996; but see BroÈ nmark andMiner 1992). Second, specialization in foraging behaviorfacilitates the evolution of specializations in feedingmorphology, as well as di�erences in patterns of growthand maturity. Third, the di�erences in feeding mor-phology alter the e�ciencies that morphs experiencewhile ®nding and capturing the alternative prey types,thereby reinforcing the behavioral di�erences. The mo-rphs may even evolve di�erences in habitat preference,

Behav Ecol Sociobiol (1999) 45: 386±395 Ó Springer-Verlag 1999

R.L. McLaughlin (&) á M.M. Ferguson á D.L.G. NoakesAxelrod Institute of Ichthyology and Department of ZoologyUniversity of Guelph, Guelph, Ontario, N1G 2W1, Canadae-mail: [email protected].: +1-519-8244120, Fax: +1-519-7633906

timing of breeding, and mate choice, possibly leading toreproductive isolation. Resource polymorphisms may bethe outcome of a genetic polymorphism or adaptivephenotypic plasticity (Robinson and Wilson 1994, 1996).In the latter case, behavior can in¯uence which devel-opmental trajectory a plastic genotype undertakes in asequence similar to that outlined in the steps above (e.g.,Meyer 1987; Metcalfe et al. 1989; Wimberger 1992;Metcalfe 1993).

If behavioral diversi®cation precedes morphologicaldiversi®cation, it should be possible to ®nd ecologicalsituations where divergent foraging tactics occur amongindividuals exhibiting no or only subtle morphologicaldi�erences. Investigations of these are important forunderstanding the mechanisms and conditions promot-ing the prerequisite specializations in behavior, prior totheir reinforcement by morphological di�erences (step 1vs step 3). Young brook charr (Salvelinus fontinalis)occupying still-water pools along the side of streamspromises to be a model study system in this regard (e.g.,Sku lason et al., in press). This ®sh exhibits conspicuous,short-term variation in foraging behavior which closelyparallels the benthic/limnetic distinction reported fre-quently for lacustrine ®shes (see Robinson and Wilson1994). Some charr tend to be sedentary (a sit-and-waittactic), feeding on mobile crustacean prey from thelower portion of the water column, while others tend tobe very active (an active search tactic), feeding on sed-entary insect prey from the upper portion of the watercolumn (McLaughlin et al. 1992, 1994). Individuals ex-hibiting intermediate levels of activity are also observed.The variation in foraging behavior is not correlated withdi�erences in body size or shape, but is it related to localenvironmental conditions within the pools (McLaughlinet al. 1994). In addition, similar variation in foragingbehavior has been reported for young-of-the-year brookcharr in much larger lakes (Biro and Ridgway 1995).This model system therefore presents an excellent op-portunity to test whether the divergent foraging behav-ior observed in the ®eld re¯ects the diversifying e�ects ofintraspeci®c competition in the presence of alternativefood resources. Further, the test can be carried out in aspecies where resource polymorphisms have been re-ported in other populations (e.g., Bourke et al. 1997),and a genus where resource polymorphisms occurcommonly (Robinson and Wilson 1994).

In this paper, we provide ®eld evidence of localcompetition for food among the charr. For a sample ofindividuals collected after our observations, we thendemonstrate that growth rate potential, an importantcomponent of ®tness during the early life stages of brookcharr, is linked to foraging behavior in a manner sug-gesting the presence of diversifying selection on foragingbehavior.

Growth rate potential was assessed using tissue con-centrations of ribonucleic acid (RNA), a biochemicalindex of short-term growth rate in ®shes (Ferron andLeggett 1994). This index was developed to reliablypredict rates of growth and mortality for wild ®sh

(Ferron and Leggett 1994). Because production of newtissue requires protein synthesis, and therefore ribo-somes, the tissue concentration of RNA, which is pre-dominantly ribosomal RNA, provides an index of thenumber of ribosomes per gram of tissue. Investigationsof ®shes in general, and salmonids in particular, havedemonstrated positive correlations between short-termgrowth rates and either tissue concentrations of RNA orthe ratio of RNA to DNA (Wilder and Stanley 1983;Miglavs and Jobling 1989; Ferguson and Danzmann1990; Bastrop et al. 1992; Mathers et al. 1993; Arndtet al. 1994). Unfortunately, the strength and form of therelationship varies among species and even life stageswithin species (Bulow 1987; Ferron and Leggett 1994).We therefore conducted a laboratory experiment toverify the relationship between speci®c growth rate andRNA concentration in young brook charr.

Methods

Behavior and growth rate potential in the ®eld

Data were collected from 5 to 25 April 1991 and 5 April to 7 May1992 at three sites along the Credit River: the North Branch atHighway 24 near Caledon, Ontario (43°52¢ N, 80°00¢ W); the WestBranch approximately 1 km northeast of Erin, Ontario (43°45¢ N,80°07¢ W); and Black Creek near Limehouse, Ontario (43°38¢ N,79°59¢ W). The Black Creek site was used only in 1991 because abeaver dam constructed in the summer of 1991 altered stream ¯owsubstantially thereafter. Brook charr at these locations are fromresident, headwater populations. Although these drainages werestocked in the 1960s and early 1970s (Ontario Ministry of NaturalResources, personal communication), studies of mitochondrialDNA revealed no evidence of hatchery stock remaining at the sites(R. McLaughlin and R. Danzmann, unpublished data). Our studysites encompassed 50- to 200-m segments of these streams con-taining areas of ¯owing water and still, backwater pools along theedges.

Observations of foraging activity were made daily between 0900and 1700 hours unless prevented by rain or snow. The ®sh were toosmall (20±30 mm) and delicate for us to mark and reobserve indi-viduals e�ectively. We therefore spread our sampling e�ort amongsites and among the pools at each site to minimize the probabilityof reobserving the same ®sh. After arriving at a pool, the observerwaited 10 min for the ®sh to return to the area and resume feeding.This wait minimized any bias toward bold or hungry ®sh becauserecently emerged brook charr return to their foraging locationwithin 2 min of a disturbance, on average (Grant and Noakes1987a). At the end of the 10 min, we usually counted all ®sh in thepool or within an area of approximately 1 m2, whichever wassmaller, and categorized the behavior of each ®sh as sedentary ormobile based on whether it was moving at the time it was counted.A focal ®sh was then selected and its fork length estimated visuallyto the nearest 0.1 cm.

We quanti®ed the behavior of the focal ®sh using the methoddescribed by McLaughlin et al. (1992, 1994). At 5-s intervals, theobserver recorded on microcassette tape (1) the distance, to thenearest body length, traversed by the focal individual, (2) whetherthe movement entailed an attempt to capture prey (a foraging at-tempt), a social interaction, or ¯ight to cover, and (3) whetherforaging attempts were directed at the substrate, the middle of thewater column, or the water surface (benthic, midwater, and surfaceforaging attempts, respectively). Intervals not involving a foragingattempt, social interaction, or ¯ight to cover were considered searchfor prey, a de®nition commonly used for search (Stephens and

387

Krebs 1986). We also noted incidences where a conspeci®c camewithin ®ve body lengths of the focal ®sh (an encounter), even whenthere was no obvious aggressive interaction. During agonistic in-teractions, the focal individual was categorized as aggressive if itchased a conspeci®c and nonaggressive if it was chased by a con-speci®c. The intended duration of observation varied from 5 minunder poor viewing conditions (e.g., low light, intermittent wind) to10 min under good viewing conditions (e.g., bright light, littlewind). The duration was varied in an attempt to balance the valueof obtaining longer periods of observation with the need to capturethe focal individuals in numbers large enough for the analysis ofgrowth rate potential.

Immediately following each observation, we attempted tocapture the focal individual using aquarium dip nets. Afterward,the site was marked. We captured 42 of 74 individuals in 1991 and42 of 69 individuals in 1992. Each captured ®sh was killed im-mediately with a blow to the head, its fork length measured to thenearest millimeter, and the carcass placed in a labelled plastic tubeand put on wet ice. Water temperature at the observation site wasthen recorded. At the end of the sampling day, we also measuredthe pool surface area and the current speed at each observationsite. For active ®sh we took the mean of several current speedmeasurements made over the area traversed by that individual.Upon return to the laboratory in the evening, the captured ®shwere frozen at )70 °C. Placing specimens on wet ice for less than24 h before freezing does not a�ect the measurement of tissueconcentrations of nucleic acids (Ferguson and Drahushchak1989).

Concentration of RNA and growth rate in the laboratory

In January 1992, we obtained approximately 1000 brook charrembryos (hatchery stock) from the Ontario Ministry of NaturalResources' Fish Culture Section. The embryos were reared inhatchery trays in recirculated, aerated well water at 8 °C, with10±20% of the water replaced daily.

The ®sh were moved into 54 ´ 50 ´ 31 cm stock tanks prior to®rst feeding. Mean water temperature was maintained at 8±9 °C tomatch the mean temperature encountered by recently emerged,wild brook charr at our ®eld sites near Guelph (McLaughlin et al.1992).

After the ®sh began feeding exogenously on commercial troutfood (BioDiet), we selected haphazardly 20 ®sh from the stocktanks. The ®sh were deprived of food for 1 day; then each ®sh wasanesthetized, measured for fork length (to the nearest millimeter)and wet weight (to the nearest 0.1 mg), placed singly into an ex-perimental tank (39 ´ 25 ´ 31 cm), and provided with food. Ex-perimental tanks were ®lled randomly to minimize any locatione�ects. The 20 ®sh were then ranked by wet weight and assignedto ®ve groups containing 4 ®sh of similar weight. Within each sizegroup, one ®sh was assigned at random to one of the followingtreatments: 0.5, 1.0, 2.0, or 4.0 times the daily ration (food al-lotment) recommended for a ®sh of that weight, following Bard-ach et al. (1972). Thereafter, the ®sh were fed BioDiet at theirassigned ration for 20 days. The ®sh were then deprived of foodfor 1 day, killed with a blow to the head, measured for fork lengthand wet weight, placed in a small labelled plastic tube, and put onwet ice. The samples were frozen at )70 °C within 2 h.Throughout the experiment, the ®sh were housed in water at8±9 °C, with ¯ow rates of 29±35 ml s)1 through the experimentaltanks. Photoperiod was altered to correspond with the naturalphotoperiod at this time of year (approximately 13 h light:11 hdark).

Measurement of tissue concentrations of RNA

Each ®sh was weighed (mg wet tissue) and gutted prior to itspreparation for measurements of nucleic acid concentrations. RNAand DNA concentrations were measured for the remaining, pri-marily muscle tissue according to the ¯uorometric technique of

Karsten and Wollenberger (1972, 1977), as described in Fergusonand Drahushchak (1989). The individual who measured the tissueconcentrations of RNA had no knowledge of the behavior each ®shhad exhibited in the ®eld.

Statistical analysis

Our behavioral observations were used to estimate for each indi-vidual the proportion of search time spent moving, the foragingattempts made per minute, and the proportion of foraging at-tempts directed at the water surface. We also estimated each focalindividual's rate of encounter and rate of interaction with con-speci®cs. Some of these variables were transformed before analysisto normalize their distributions as much as possible. The propor-tion of time spent moving and the proportion of foraging attemptsdirected at the water surface were arcsine-square-root transformed.Foraging attempts per minute was square-root transformed. Forklength was log10-transformed. In analyses considering all of the®sh we observed, we used our visual estimates of fork length forthose ®sh not captured after the observation period. Measuredvalues of fork length were used for ®sh that were captured. Ana-lyses for captured ®sh showed that our visual estimates of forklength (EFL) were good predictors of measured fork lengths(MFL). In each year, EFL and MFL were signi®cantly and posi-tively correlated (r = 0.74 and 0.84, respectively, both P < 0.001).In addition, for regressions of EFL versus MFL, the regressioncoe�cients (slopes) were not signi®cantly di�erent from 1(P > 0.10) and the intercepts were not signi®cantly di�erent from0 (P > 0.10).

All individuals considered in our analyses were observed for-aging in still-water habitats and not resting on the stream bottom.Our observation periods lasted from 4.1 to 21.3 min (mean =8.1 min). On 12 occasions, we quanti®ed the behavior of an indi-vidual, waited 10 min, and requanti®ed the behavior of that indi-vidual. Correlation coe�cients between estimates made in the ®rstand second observation periods were 0.99, 0.75, and 0.91 for theproportion of time spent moving, the proportion of foraging at-tempts directed at the water surface, and foraging attempts perminute, respectively.

Several Monte Carlo simulations were used to assess thesuitability of our sampling in relation to the activity and size ofthe ®sh. First, our selection of sedentary and mobile ®sh overeach ®eld season was not signi®cantly di�erent from that ex-pected for random sampling of the sedentary and mobile ®shavailable at the time of each selection, based on our visualcounts (P > 0.30 and P > 0.50 for 1991 and 1992, respective-ly). Second, for the 42 ®sh captured in 1991 and for the 42captured in 1992, the means and variances for the proportion oftime spent moving and for fork length did not di�er signi®-cantly from those expected for 1000 random samples of 42 ®shselected from the 74 and 69 ®sh observed in 1991 and 1992,respectively.

Data were aggregated across sampling sites because earlierwork has failed to reveal any signi®cant di�erences among them(McLaughlin et al. 1992, 1994; McLaughlin and Grant 1994). Itwas also necessary to aggregate data on social behavior for 1991and 1992 because of the low number of aggressive interactionsobserved.

For ®sh observed in the ®eld, bimodality in frequency distri-bution for the proportion of search time spent moving was assessedusing the saddle test in the MODECLUS procedure of SAS (1997).Fitness surfaces can be complex (e.g., have more than one peak)and di�cult to characterize, particularly with small sample sizes(see Schluter and Nychka 1994). For ®sh collected from the ®eld,we assessed how tissue concentration of RNA, our index of growthrate potential and short-term ®tness, changed in relation to theproportion of time spent moving and the foraging attempts madeper minute in four steps. First, we constructed contour plotscharacterizing the surface of this relationship. Second, we ap-proximated the surface statistically by ®tting the following poly-nomial equation:

388

Y � B0 � �B1 � x1� � �B2 � x2� � �B11 � x21� � �B22 � x22�� �B12 � x1x2� � �B111 � x31� � �B222 � x32� � �B122 � x1x22�� �B112 � x21x2� � �B1112 � x31x2� � �B2221 � x1x32�

where Y was the tissue concentration of RNA, x1 was the pro-portion of time spent moving, x2 was foraging attempts made perminute, and Bs represent the regression coe�cients. This equationwas selected because inspection of the contour plots suggested asimple second-order polynomial would not be adequate, particu-larly for our 1992 dataset, and preliminary analyses con®rmed this.Third, we determined whether the overall, ®tted equation wassigni®cant statistically and whether it described a saddle (minimax)system. A saddle system would indicate the existence of two distinctmaxima (peaks) in growth rate potential (Khuri and Cornell 1996).Fourth, we constructed contour plots using the predicted valuesfrom the ®tted equations and compared these visually with thesurfaces obtained in the ®rst step. The contour plots obtained fromthe ®tted equations are provided here. For the analysis, tissueconcentration of RNA was expressed on a relative scale, like rel-ative ®tness, by dividing each value by the sample mean.

For ®sh collected in our laboratory experiment, the relationshipbetween tissue concentration of RNA and speci®c growth rate wasanalyzed using analysis of covariance, with food ration as a nom-inal, ®xed e�ect and tissue concentration of RNA as a continuouse�ect.

Results

Behavior and growth rate potential in the ®eld

In 1991 and in 1992, frequency distributions of theproportion of search time spent moving were somewhatU-shaped indicating that, during our observation peri-ods, most individuals tended to spend little of theirsearch time moving or much of their search time mov-ing, and relatively fewer individuals tended to spendintermediate proportions of their search time moving(Fig. 1). For both distributions, the hypothesis of asingle mode was rejected in favor of the hypothesis oftwo modes (clusters) (saddle test: z = 2.3, P < 0.03 andz = 2.6, P < 0.03, for 1991 and 1992, respectively).

Individuals that spent a large proportion of timemoving directed more foraging attempts toward thewater surface than did individuals that spent a smallproportion of time moving (rs = 0.60, P < 0.0001,n = 74 and rs = 0.56, P < 0.0001, n = 69 for 1991and 1992, respectively; Fig. 2a). They also made moreforaging attempts per min (Fig. 2b). Overall, the varia-tion in foraging behavior was very similar between years,although in 1992, the rate at which foraging attempts perminute increased with the proportion of time spentmoving was signi®cantly higher than in 1991 (regressioncoe�cients: 0.016 vs 0.009, respectively; comparison ofslopes: F = 5.55, P < 0.02, df = 1,139). The individ-uals spending most of their time moving in 1992 mademore foraging attempts per minute than those spendingmost of the time moving in 1991, but there was littledi�erence between years for ®sh that spent little timemoving.

Fifty percent (71/143) of the individuals encounteredanother conspeci®c during an observation period. Of

Fig. 1 Frequency distributions of individual variation in the propor-tion of search time spent moving by recently emerged brook charr instill-water pools [n=74 individuals in 1991 (a) and n=69 in 1992 (b)]

Fig. 2 The proportion of foraging attempts directed at the watersurface (a) and the foraging attempts made per minute (FR) (b) inrelation to the proportion of search time spent moving (PM) byrecently emerged brook charr. Each point represents an individual ®shobserved in 1991 (closed circles) or 1992 (open circles). Proportions arepresented on an arcsine-square-root scale and foraging attempts perminute on a square-root scale. b Equations for regression lines areFR1=2 � 1:18� 0:009� arcsin �PM1=2� (r = 0.50, P < 0.0001,df = 72) and FR1=2 � 1:32� 0:016� arcsin (PM1=2� (r = 0.63,P < 0.0001, df = 67) for 1991 and 1992, respectively

389

these, 45% (32/71) were involved in at least one ag-gressive interaction and the probability of an individualbecoming involved in an aggressive interaction uponencountering another ®sh was 0.33 (SE = 0.03). Therate of encounters with conspeci®cs and the rate of in-teractions were higher, on average, when local densitywas higher (rs = 0.30, P < 0.025 and rs = 0.43,P < 0.005, df = 55, respectively).

Of the 32 ®sh involved in aggressive interactions, 17were categorized as aggressive and 13 as nonaggressive.Two could not be categorized. Frequency distributionsfor the proportion of time spent moving di�ered sig-ni®cantly between aggressive and nonaggressive ®sh(Kolmogorov-Smirnov two-sample test: P < 0.025;Fig. 3). Aggressive ®sh tended to exhibit either a low orhigh proportion of time spent moving (Fig. 3a), whilenonaggressive ®sh tended to exhibit intermediate pro-portions of time moving (Fig. 3b). These intermediatelevels of activity for nonaggressive ®sh were not a con-sequence of marked changes in activity before and afteran aggressive interaction. The proportion of time non-aggressive individuals spent moving after an aggressiveinteraction was positively correlated with the proportionof time they spent moving prior to the interaction(r = 0.63, P < 0.02). There was also no consistentchange in the proportion of time spent moving followingan aggressive interaction (Wilcoxon matched-pair sign-rank test: P > 0.60).

Social interactions were closely linked to prey captureevents with 36% of encounters (61/170) and aggressiveinteractions (23/64) occurring at the time a focal indi-vidual attempted to capture a prey item (Fig. 4). Afterstatistically adjusting for variation in the proportion oftime spent moving, moreover, aggressive individuals

exhibited a higher rate of foraging attempts per minutethan nonaggressive individuals (analysis of covariance:adjusted means = 4.4 and 2.4 foraging attempts perminute, respectively; F = 9.70, df = 1,27, P < 0.005;Fig. 5).

For ®sh collected in the ®eld, tissue concentrations ofRNA ranged from 401 to 2077 lg per gram tissue(mean = 1185) in 1991 and 521±2078 lg per gram tis-sue (mean = 1259) in 1992. In both years, the relativeconcentration of RNA (growth rate potential) observedamong individuals varied signi®cantly with the propor-

Fig. 3 The proportion of search time spent moving by recentlyemerged brook charr that were classi®ed as aggressive (a) ornonaggressive (b) in social contests

Fig. 4 The time interval between when a focal individual encountereda conspeci®c (a) or interacted aggressively with a conspeci®c (b) andthe focal individual's most recent foraging attempt

Fig. 5 Foraging attempts made per minute in relation to theproportion of search time spent moving by aggressive (closed circles)and nonaggressive (open circles) individuals. Proportion of search timespent moving is presented on an arcsine-square-root scale andforaging attempts per minute on a square-root scale

390

tion of time the individuals spent moving and thenumber of foraging attempts they made per minute(F = 2.30, P < 0.04, df = 11,30 and F = 2.53,P < 0.025, df = 11, 29, for 1991 and 1992, respective-ly). The ®tted response surfaces (see Methods) ac-counted for 46% of the variation in growth ratepotential in 1991 and 49% of the variation in 1992. Inboth years, growth rate potential was above average for®sh exhibiting a low proportion of time spent movingand few foraging attempts per minute and for ®sh ex-hibiting a high proportion of time spent moving andmany foraging attempts per minute (Fig. 6). Growthrate potential was average or below average for ®shspending intermediate proportions of time moving, andfor ®sh exhibiting a high proportion of time spentmoving, but low foraging attempts per minute.

Concentration of RNA and growth ratein the laboratory

In our laboratory experiment, the tissue concentrationof RNA and food ration explained 68% of the variancein speci®c growth rate. There was no signi®cant, statis-tical interaction between concentration of RNA andfood ration, suggesting they a�ected growth rate inde-pendently (F = 1.45, df = 3,11, P > 0.25). After sta-tistically adjusting for food ration, ®sh with higher tissueconcentrations of RNA grew faster than those withlower concentrations of RNA, as expected (partialr=0.60, F = 7.77, df = 1,14, P < 0.02; Fig. 7). Afterstatistically adjusting for the tissue concentration ofRNA, ®sh given a higher food ration grew signi®cantlyfaster than those given a lower food ration (F = 8.55,df = 3,14, P < 0.002).

Discussion

Our ®ndings are consistent with the hypothesis that thevariation in foraging behavior exhibited by recentlyemerged brook charr in the ®eld represents diversi®ca-tion brought about by intraspeci®c competition in thepresence of alternative food sources. The support comesfrom four signi®cant features of our brook charr system.First, the conspicuous, bimodal variation in the pro-portion of time spent moving during search for prey iscorrelated with diet and microhabitat (water column)use, with more sedentary individuals feeding on aquaticcrustaceans from the lower portion of the water columnand more active individuals feeding on insect prey fromthe upper portion of the water column (Grant and No-akes 1987b; McLaughlin et al. 1994; this study). Second,density-dependent rates of encounters and aggressiveinteractions between conspeci®cs, plus the temporalproximity between this social behavior and prey captureattempts, indicate that intraspeci®c competition (sensuMilinski and Parker 1991) is occurring on a local scale(this study). Intraspeci®c competition is a common fea-ture of this system in particular (this study; Grant 1990),

Fig. 6 Tissue concentrations of RNA, an index of growth ratepotential, in relation to proportion of search time spent moving andforaging attempts made per minute by recently emerged brook charrin 1991 (a) and 1992 (b). Like relative ®tness, concentrations of RNAare expressed as multiples of the sample mean. Positions of peaks inRNA concentration are indicated by a thicker contour line. Proportionof search time spent moving is presented on an arcsine-square-rootscale and foraging attempts per minute on a square-root scale

Fig. 7 Leverage plot showing the relationship between speci®cgrowth rate and tissue concentrations of RNA after statisticallycontrolling for daily food ration

391

and stream-dwelling populations of salmonid ®shes ingeneral (Elliott 1990b; Grant and Kramer 1990; Nakano1995; Fausch et al. 1997). Third, aggression is correlatedwith the variation in foraging behavior (this study).Aggressive ®sh tend to be either inactive or very active,while nonaggressive ®sh tend to exhibit intermediatelevels of activity. Aggressive ®sh also exhibit substan-tially higher rates of foraging attempts, for a given levelof activity, than do nonaggressive ®sh. Fourth, growthrate potential, an important component of ®tness duringthis early life stage, is also correlated with the variationin foraging behavior and supports the notion of short-term diversifying selection on foraging behavior (thisstudy).

The reason for the lower growth rate potential ofindividuals exhibiting intermediate levels of activitywhile searching for prey is not entirely clear. It appearsthat switching between foraging tactics entails a tradeo�,possibly due to learning (Bence 1986) or travel costs(Murdoch et al. 1975). Searching for benthic (crusta-cean) prey and surface (insect) prey is certainly mutuallyexclusive to some degree. For example, brook charrswitching from hovering to moving reduce the propor-tion of foraging attempts they direct at the substrate andincrease the proportion they direct at the water surface,and vice versa (Grant and Noakes 1987b; R.McLaughlin, unpublished data). Correlated shifts inmobility and water column use have also been reportedfor dolly varden charr (S. malma) inhabiting pools ofrunning water (Fausch et al. 1997), although the natureof the shifts is di�erent in running water than it is in thestill-water pools we have studied. Furthermore, brookcharr switching from hovering to moving also experiencea reduced probability of capture upon attack for benthicprey and an increased probability of capture upon at-tack for midwater prey, while the probability of captureupon attack for surface prey remains unchanged (R.McLaughlin, unpublished data). Lastly, analysis ofstomach contents has demonstrated the likelihood anindividual has eaten an benthic prey is negatively cor-related with the likelihood it has eaten a surface prey(McLaughlin et al. 1994).

The conclusion that diversifying selection on foragingbehavior is occurring within our study system dependsupon the adequacy of growth rate potential as an indexof ®tness, at least over the short term. We believe thatgrowth rate potential is the best available ®eld measureof ®tness for recently emerged brook charr. The evidenceavailable for stream-dwelling salmonids, includingbrook charr, indicates that increased growth rate im-proves ®tness through size-dependent e�ects on survivalduring the ®rst few weeks of high mortality followingemergence from the gravel (e.g., Elliott 1989, 1990a,1990b; Hutchings 1991). We do not know how long thegrowth di�erences we observed are maintained, but ev-idence from a variety of salmonid ®shes indicates thatbehavioral di�erences arising early in life can have im-portant longer-term consequences (e.g., Metcalfe et al.1989; Nielsen 1992). For brook charr, increased growth

rate over the longer term improves ®tness through size-selective overwinter survival (Hunt 1969; Shuter andPost 1990) and size-dependent e�ects on fecundity(Hutchings 1993, 1996). Finally, faster growth is favoredwhen mortality and competition are size dependent andwhen there are time constraints on reaching a minimumsize (Arendt 1997).

Recent reviews of the evolution of growth rate havereiterated that there can be tradeo�s associated withincreasing growth rate, although these tradeo�s are notall well understood (Arendt 1997; Conover and Schultz1997). Predation risk is the most obvious and important®tness component which is not considered by our mea-sure of growth rate potential, and animals will accept alower growth rate to avoid predators (Werner and Gil-liam 1984). Based on our experience with this system,predation risk, at least while the charr are foraging, islikely low and does not di�erentially a�ect one behav-ioral phenotype over another. This stream system isrelatively free of piscivorous ®shes and the individualslarge enough to be potential, facultative piscivores re-main in the main stream channel, away from the small,shallow pools occupied by recently emerged brookcharr. Larger semi-aquatic predators, such as mink,herons, and king®shers are seen occasionally, but at2±3 cm in length, recently emerged brook charr are wellbelow the minimum prey length of 5+ cm normallyeaten by these predators (Alexander 1991; R. McLau-ghlin personal observation). Invertebrate predators, in-cluding diving beetles, water scorpions, and larvae oflarge odonates, are also seen occasionally and can eatsmall ®sh, but in 10 years of ®eld research we have neverseen one attempt to capture a young brook charr. De-spite our observations, however, examining how indi-vidual brook charr adjust their foraging behavior inresponse to the presence of potential predators remains aquestion we need to address further.

Although we have provided evidence for two peaks ingrowth rate potential, the peaks sometimes did notcorrespond exactly with the peaks in the proportion oftime spent moving (Fig. 1) or in aggressive behavior(Fig. 4). The discrepancy may represent sampling error,as the data regarding each attribute come from di�erentsubsamples of the ®sh we observed. Alternatively, thismay re¯ect the impact that unmeasured components of®tness (see above) have on behavior, or re¯ect con-straints on the ability of these very young ®sh to assessthe level of activity where growth rate potential is high.The relationship between behavior and growth rate po-tential is also more complex than we expected, depend-ing on the rate of foraging attempts as well as theproportion of time spent moving. Super®cially, the de-pendency on foraging rate appears inconsistent with ourlaboratory experiment, where RNA concentration wasuna�ected by food ration. In our laboratory experiment,however, none of the ®sh were deprived of food. In the®eld, conversely, some ®sh are not receiving adequateamounts of food, as evidenced by empty digestive tracts(McLaughlin et al. 1994; McLaughlin and Grant 1994).

392

Concentrations of RNA will decline when charr aredeprived of food for periods of 5 days or more(McLaughlin et al. 1995).

Our research on recently emerged brook charr pro-vides a useful complement to the literature on resourcepolymorphisms in four ways. First, we have focuseduniquely on foraging tactics exhibited by behaviorally¯exible individuals from a population without distinctmorphological variants, whereas the existing literatureon resource polymorphisms has focused almost exclu-sively on ®xed di�erences in behavior exhibited bydistinctly di�erent morphotypes (see Smith and Sku la-son 1996). For example, the work of Schluter (1993,1995, 1996) demonstrating foraging and growth ratetradeo�s in three-spine sticklebacks (Gasterosteus acu-leatus) involved comparisons of species pairs. Morerecently, Robinson et al. (1996) demonstrated that in-dividual pumpkinseed sun®sh (Lepomis gibbosus) withextreme benthic and limnetic morphologies had highercondition factors and exhibited faster growth rates thandid individuals of intermediate morphology. Their ex-amination was particularly stringent because they fo-cused on intraspeci®c variation, and the morphologicaldi�erences among individuals were subtle in the studypopulation. Our brook charr system appears to repre-sent an even more subtle, but similar form of diversi®-cation in foraging behavior. Moreover, our ®ndingscome from a species where more discrete resourcepolymorphisms have been reported in other, lake pop-ulations (Robinson and Wilson 1994; Bourke et al.1997). Therefore, our ®ndings are relevant to specula-tion regarding the role of behavioral diversi®cationduring initial steps in the evolution of distinct feedingmorphs (Sku lason and Smith 1995; Smith and Sku lason1996).

Second, despite possessing the pitfalls of a descriptive®eld study, our study has provided evidence suggestingshort-term, divergent selection in a system where indi-viduals exhibit diverse foraging tactics. Testing the hy-pothesis that behavioral or morphological diversi®cationis brought about by divergent selection for e�cient re-source use is a signi®cant challenge and examples areuncommon (but see Hori 1993; Smith 1993; Schluter1993, 1994, 1995; Robinson et al. 1996). Although someuncertainties remain, our ®ndings are even more re-markable because the behavioral tactics we observed arenot reinforced by di�erences in body size or shape(McLaughlin et al. 1994) and individuals can switch be-tween tactics.

Third, because of our ability to observe the behaviordirectly in the ®eld, our ®ndings complement the largeliterature for lacustrine ®shes where the signi®cance ofbehavior has been largely inferred from conspicuousdi�erences in foraging morphology and diet, rather thandemonstrated through direct observation (see Robinsonand Wilson 1994; Robinson et al. 1996). The comple-mentarity is important for improving our understandingof the role of behavior in producing and maintainingresource polymorphisms, particularly because the

behavior of ®shes is often di�cult to observe undernatural conditions.

Finally, our system provides a unique example wheregeneralizations regarding the adjustments in prey-searching behavior and the tradeo�s reported in earlierlaboratory experiments examining ®shes switching be-tween benthic and limnetic niches (e.g., Murdoch et al.1975; Ehlinger 1989, 1990; Savino and Stein 1989) canbe observed and measured, albeit more crudely, in the®eld. Findings from small-scale, controlled laboratorystudies may not extrapolate well to larger-scale, morecomplex ®eld (lake) environments (e.g., Sku lason et al.1993), particularly when activity is the signi®cant be-havior of interest (Lindsey 1978). Results from thissystem, however, are pertinent to larger-scale, lake sys-tems, despite the simplicity and smaller scale of thesystem (see Robinson and Wilson 1994; Biro andRidgway 1995; Biro et al. 1996; Bourke et al. 1997;McLaughlin and Noakes, in press).

Why the conspicuous morphological variation ob-served relatively frequently in lacustrine ®shes is notobserved in our study system, in particular, and rarely instream ®shes in general (Wimberger 1994), is an inter-esting question. It could be that subtle variation doesexist, but has gone undetected. A more likely explana-tion, however, is that the temporal and spatial variabilityof benthic and limnetic resources is higher in streamsthan in lakes. Consequently, generations of stream ®shesare less likely to spend signi®cant portions of their livesin a speci®c niche (e.g., Fausch et al. 1997) in the waysome lake ®shes are thought to, hence the evolution ofalternative morphs would not be expected. With oursystem in particular, developmental changes in habitatuse as well as learning and swimming ability are prob-ably important. The young charr only inhabit the still-water pools during spring and early summer, and latermove into deeper, ¯owing water.

The temporal stability of the behavioral variationobserved among the charr in still-water pools remains arelated, interesting, but unanswered question because ofthe logistical di�culties of marking and reobservingthese small ®sh. Short-term repeated measurements ofbehavior (see Methods), the correlations between activ-ity and both diet and growth rate potential, and longer,qualitative observations in the ®eld, lead us to believethat some individuals exhibit longer-term repeatabilityin their behavior, while some probably do not. Further,a variety of studies have demonstrated longer-term, re-peatable, individual di�erences in social and foragingbehavior for wild salmonids (Bryan and Larkin 1972;Nielsen 1992; Nakano 1995; Bourke et al. 1997). Inparticular, one recent, ®eld enclosure study withmarked, recently emerged brook charr from a lakepopulation has revealed repeatable di�erences, overseveral weeks, in the proportion of time spent movingduring search for prey (P. Biro, personal communica-tion). In fact, we are unaware of a study on salmonidswhich has attempted and failed to ®nd individual dif-ferences in behavior, although it is also clear that

393

stream-dwelling salmonids will alter their behavior inresponse to changes in the abundances of benthic andsurface prey (Fausch et al. 1997). The fact that the be-havioral di�erences we observe for brook charr may notrepresent ®xed, long-term specializations may seem dis-advantageous given the current emphasis on resourcepolymorphisms exhibiting well-de®ned morphs. How-ever, from the perspective of understanding the earlystages of resource polymorphisms, when individuals arestill expected to possess a high degree of behavioral¯exibility, it is not. Part of identifying the ecologicalconditions favoring resource polymorphisms requiresthat we examine populations with more ¯exible behaviorand less pronounced variation in morphology.

For many years behavioral ecologists have promisedthat understanding the behavior of individuals couldprovide valuable insights into population-level processes(Krebs and Davies 1991), with some success (e.g. Met-calfe et al. 1989; Sutherland 1996; Fryxell and Lundberg1997). Our brook charr system provides an examplewhere ideas developed by behavioral ecologists could bevery useful for examining the role adaptive, individualbehavior plays in one important population-level phe-nomenon ± the evolution of resource polymorphisms.

Acknowledgements We thank P. Biro, J.W.A. Grant, and S.S.Snorrason, for commenting on earlier drafts of the manuscript, S.Arndt and T. Benfey for sharing unpublished manuscripts on nu-cleic acids and salmonid growth, and the Ontario Ministry ofNatural Resources for providing ®sh for our laboratory experimentand for granting permission to collect ®sh in the ®eld. This researchwas supported by a Natural Sciences and Engineering ResearchCouncil of Canada (NSERC) Postdoctoral Research Fellowship toR.L.M., NSERC Operating Grants to M.M.F. and D.L.G.N., anda Department of Fisheries and Oceans (DFO)-NSERC SubventionGrant to D.L.G.N. in liaison with R. Cunjak, DFO. Fish weremaintained and handled under conditions approved by the AnimalCare Committee, University of Guelph.

References

Alexander GR (1991) Trout as prey. In Stolz J, Schnell J (eds)Trout. Stackpole, Harrisburg, Pa, pp 112±117

Arendt JD (1997) Adaptive intrinsic growth rates: an integratingacross taxa. Q Rev Biol 72:149±177

Arndt SKA, Benfey TJ, Cunjak, RA (1994) A comparison of RNAconcentrations and ornithine decarboxylase activity in Atlanticsalmon (Salmo salar) muscle tissue, with respect to speci®cgrowth rates and diel variations. Fish Physiol Biochem 13:463±471

Bardach JE, Ryther JH, McLarney WO (1972) Aquaculture: thefarming and husbandry of freshwater and marine organisms.Wiley, New York

Bastrop R, JuÈ rss K, Wacke R (1992) Biochemical parameters as ameasure of food availability and growth in immature rainbowtrout (Oncorhynchus mykiss). Comp Biochem Physiol102A:151±161

Bence JR (1986) Feeding rate and attack specialization: the roles ofpredator experience and energetic tradeo�s. Environ Biol Fish16:113±121

Biro PA, Ridgway, MS (1995) Individual variation in foragingmovements in a lake population of young-of-the-year brookcharr (Salvelinus fontinalis). Behaviour 132:57±74

Biro PA, Ridgway MS, McLaughlin RL (1996) Does foragingattempt rate predict ingestion rate for young-of-the-year brooktrout (Salvelinus fontinalis) in the ®eld? Can J Fish Aquat Sci53:1814±1820

Bourke P, Magnan P, RodrõÂ guez MA (1997) Individual variationsin habitat use and morphology in brook charr. J Fish Biol51:783±794

BroÈ nmark C, Miner JG (1992) Predator-induced phenotypicchange in body morphology in Crucian carp. Science 258:1348±1350

Bryan JE, Larkin PA (1972) Food specialization by individualtrout. J Fish Res Bd Can 29:1615±1624

Bulow FJ (1987) RNA-DNA ratios as indicators of growth in ®sh:a review. In Summerfelt RC, Hall GE (eds) The age and growthof ®sh. Iowa State University Press, Ames, pp 45±64

Conover DO, Schultz ET (1997) Natural selection and the evolu-tion of growth rate in the early life history: what are the trade-o�s? In Chambers RC, Trippell EA (eds) Early life history andrecruitment in ®sh populations. Chapman & Hall, London, pp305±332

Ehlinger TJ (1989) Learning and individual variation in bluegillforaging: habitat-speci®c techniques. Anim Behav 38:643±658

Ehlinger TJ (1990) Habitat choice and phenotype-limited feedinge�ciency in bluegill: individual di�erences and trophic poly-morphism. Ecology 71:886±896

Elliott JM (1989) Mechanisms responsible for population regula-tion in young migratory trout, Salmo trutta. I. The critical timefor survival. J Anim Ecol 58:987±1001

Elliott JM (1990a) Mechanisms responsible for population regu-lation in young migratory trout, Salmo trutta. II. Fish growthand size variation. J Anim Ecol 59:171±185

Elliott JM (1990b) Mechanisms responsible for population regu-lation in young migratory trout, Salmo trutta. III. The role ofterritorial behaviour. J Anim Ecol 59:803±818

Fausch KD, Nakano S, Kitano S (1997) Experimentally inducedforaging mode shift by sympatric charrs in a Japanese mountainstream. Behav Ecol 8:414±420

Ferguson MM, Danzmann RG (1990) RNA/DNA ratios in whitemuscle as estimates of growth in rainbow trout held at di�erenttemperatures. Can J Zool 68:1494±1498

Ferguson MM, Drahushchak LR (1989) E�ects of tissue collectionand storage methods on nucleic acid determination in whitemuscle of ®shes. Trans Am Fish Soc 118:709±713

Ferron A, Leggett WC (1994) An appraisal of condition measuresfor marine ®shes. Adv Mar Biol 30:217±303

Fryxell JM, Lundberg P (1997) Individual behavior and commu-nity dynamics. Chapman & Hall, New York

Grant JWA (1990) Aggressiveness and the foraging behaviour ofyoung-of-the-year brook charr (Salvelinus fontinalis). Can JFish Aquat Sci 47:915±920

Grant JWA, Kramer DL (1990) Territory size as a predictor of theupper limit to population density of juvenile salmonids instreams. Can J Fish Aquat Sci 47:1724±1737

Grant JWA, Noakes DLG (1987a) Escape behaviour and use ofcover by young-of-the-year brook trout, Salvelinus fontinalis.Can J Fish Aquat Sci 44:1390±1396

Grant JWA, Noakes DLG (1987b) Movers and stayers: foragingtactics of young-of-the-year brook charr, Salvelinus fontinalis.J Anim Ecol 56:1001±1013

Hori M (1993) Frequency-dependent natural selection in thehandedness of scale-eating cichlid ®sh. Science 260:216±219

Hunt RL (1969) Overwinter survival of wild ®ngerling brook troutin Lawrence Creek, Wisconsin. J Fish Res Bd Can 26:1473±1483

Hutchings JA (1991) Fitness consequences of variation in egg sizeand food abundance in brook trout Salvelinus fontinalis. Evo-lution 45:1162±1168

Hutchings JA (1993) Adaptive life histories e�ected by age-speci®csurvival and growth rate. Ecology 74:673±684

Hutchings JA (1996) Adaptive phenotypic plasticity in brook trout,Salvelinus fontinalis, life histories. Ecoscience 3:25±32

394

Karsten U, Wollenberger A (1972) Determination of DNA andRNA in homogenized cells and tissues by surface ¯uorometry.Anal Biochem 46:135±148

Karsten U, Wollenberger A (1977) Improvements in the ethidiumbromide method for direct ¯uorometric estimation of DNA andRNA in cell and tissue homogenates. Anal Biochem 77:464±470

Khuri AI, Cornell JA (1996) Response surfaces: designs and ana-lyses. Dekker, New York

Krebs JR, Davies NB (1991) Behavioural ecology: an evolutionaryapproach. Blackwell, Oxford

Lindsey CC (1978) Form, function, and locomotor habits. In HoarWS, Randall JR (eds) Fish physiology. Vol 7: locomotion.Academic Press, New York, pp 1±100

Mathers EM, Houlihan DF, McCarthy ID, Burren LJ (1993) Ratesof growth and protein synthesis correlated with nucleic acidcontent in fry of rainbow trout, Oncorhynchus mykiss: e�ects ofage and temperature. J Fish Biol 43:245±263

McLaughlin RL, Grant JWA (1994) Morphological and behav-ioural di�erences among recently-emerged brook charr, Salve-linus fontinalis, foraging in slow- vs. fast-running water.Environ Biol Fish 39:289±300

McLaughlin RL, Noakes DLG (in press) Living with failure: theprey capture success of young brook charr in streams. EcolFreshwater Fish

McLaughlin RL, Grant JWA, Kramer DL (1992) Individualvariation and alternative patterns of foraging movements inrecently-emerged brook charr (Salvelinus fontinalis). Behaviour120:286±301

McLaughlin RL, Grant JWA, Kramer DL (1994) Foragingmovements in relation to morphology, water column-use, anddiet for recently emerged brook trout (Salvelinus fontinalis) instill-water pools. Can J Fish Aquat Sci 51:268±279

McLaughlin RL, Ferguson MM, Noakes DLG (1995) Tissueconcentrations of RNA and protein for juvenile brook trout(Salvelinus fontinalis): lagged responses to ¯uctuations in foodavailability. Fish Physiol Biochem 14:459±469

Metcalfe NB (1993) Behavioural causes and consequences of lifehistory variation in ®sh. Mar Behav Physiol 23:205±217

Metcalfe NB, Huntingford FA, Graham WD, Thorpe JE (1989)Early social status and the development of life-history strategiesin Atlantic salmon. Proc R Soc Lond B 236:7±19

Meyer A (1987) Phenotypic plasticity and heterochrony inCichlasoma managuense (Pisces, Cichlidae) and their implica-tions for speciation in cichlids ®shes. Evolution 41:1357±1369

Miglavs I, Jobling M (1989) E�ects of feeding regime on foodconsumption, growth rates and tissue nucleic acids in juvenilearctic charr, Salvelinus alpinus, with particular respect to com-pensatory growth. J Fish Biol 34:947±957

Milinski M, Parker GA (1991) Competition for resources. In KrebsJR, Davies NB (eds) Behavioural ecology: an evolutionaryapproach. Blackwell, Oxford, pp 137±168

Murdoch WW, Avery S, Smyth MEB (1975) Switching in preda-tory ®sh. Ecology 56:1094±1105

Nakano S (1995) Individual di�erences in resource use, growth andemigration under the in¯uence of a dominance hierarchy in¯uvial red-spotted masu salmon in a natural habitat. J AnimEcol 64:75±84

Nielsen JL (1992) Microhabitat-speci®c foraging behavior, diet,and growth of juvenile coho salmon. Trans Am Fish Soc121:617±634

Robinson BW, Wilson DS (1994) Character release and displace-ment in ®shes: a neglected literature. Am Nat 144:596±627

Robinson BW, Wilson DS (1996) Genetic variation and phenotypicplasticity in a trophically polymorphic population of pump-kinseed sun®sh (Lepomis gibbosus). Evol Ecol 10:631±652

Robinson BW, Wilson DS, Shea GO (1996) Trade-o�s of ecolog-ical specialization: an intraspeci®c comparison of pumpkinseedsun®sh phenotypes. Ecology 77:170±178

SAS (1997) SAS/STAT software: changes and enhancementsthrough release 6.12. SAS Institute, Cary, NC

Savino JF, Stein RA (1989) Behavioural interactions between ®shpredators and their prey: e�ects of plant density. Anim Behav37:311±321

Schluter D (1993) Adaptive radiation in sticklebacks: size, shapeand habitat use e�ciency. Ecology 74:699±709

Schluter D (1994) Experimental evidence that competition pro-motes divergence in adaptive radiation. Science 266:798±801

Schluter D (1995) Adaptive radiation in sticklebacks: trade-o�s infeeding performance and growth. Ecology 76:82±90

Schluter D (1996) Ecological causes of adaptive radiation. Am Nat148:S40±S64

Schluter D, Nychka D (1994) Exploring ®tness surfaces. Am Nat143:597±616

Shuter BJ, Post JR (1990) Climate, population variability, and thezoogeography of temperate ®shes. Trans Am Fish Soc 119:314±336

Sku lason S, Smith TB (1995) Resource polymorphisms in verte-brates. Trends Ecol Evol 10:366±370

Sku lason S, Snorrason SS, Ota D, Noakes DLG (1993) Geneticallybased di�erences in foraging behaviour among sympatricmorphs of arctic char (Pisces: Salmonidae). Anim Behav45:1179±1192

Sku lason S, Snorrason SS, Jo nsson B (in press) Sympatric morphs,populations and speciation in freshwater ®sh with emphasis onarctic charr. In Magurran A, May R (eds) Evolution of bio-logical diversity, Oxford University Press, Oxford

Smith TB (1993) Disruptive selection and the genetic basis of billsize polymorphism in the African ®nch Pyrenestes. Science363:618±620

Smith TB, Sku lason S (1996) Evolutionary signi®cance of resourcepolymorphisms in ®shes, amphibians and birds. Annu Rev EcolSyst 27:111±133

Stephens DW, Krebs JR (1986) Foraging theory. Princeton Uni-versity Press, Princeton, NJ

Sutherland WJ (1996) From individual behavior to populationecology. Oxford University Press, New York

Werner EE, Gilliam JF (1984) The ontogenetic niche and speciesinteractions in size-structured populations. Annu Rev Ecol Syst15:393±425

Wilder IB, Stanley JG (1983) RNA-DNA ratio as an index ofgrowth in salmonid ®shes in the laboratory and in streamscontaminated by carbaryl. J Fish Biol 22:165±172

Wimberger PH (1992) Plasticity in body shape: the e�ects of diet,development, family and age in two species of Geophagus(Pisces, Cichlidae). Biol J Linn Soc 45:197±218

Wimberger PH (1994) Trophic polymorphisms, plasticity, andspeciation in vertebrates. In Stouder DJ, Fresh KL, Feller RJ(eds) Theory and application in ®sh feeding ecology. BelleBaruch Library in Marine Science, no 18. University of SouthCarolina Press, Columbia, pp 19±43

Communicated by N.B. Metcalfe

395