Overriding the oddity effect in mixed-species aggregations: group choice by armored and nonarmored prey

Abstract

Because “odd” individuals often suffer disproportionately high rates of predation, solitary individuals should join groups whose members are most similar to themselves in appearance. We examined group-choice decisions by individuals in armored and nonarmored species and predicted that either (1) the oddity effect would result in preference for conspecific groups for solitary individuals of both species, or (2) individuals in the armored species would prefer to associate with groups containing individuals of the more vulnerable species. Armored brook sticklebacks (Culaea inconstans) and nonarmored fathead minnows (Pimephales promelas) have the same predators and often occur together in streams. In mixed-species shoals, yellow perch (Perca flavescens) attacked minnows earlier and more often than sticklebacks. We tested whether solitary minnows and sticklebacks preferred to associate with conspecific or heterospecific shoals under conditions of both low and high predation risk. When predation risk was high, minnows preferred to associate with conspecifics over heterospecifics, as predicted by the oddity effect. In contrast, sticklebacks preferentially associated with groups of minnows over groups of conspecifics when predation risk was high. When predation risk was low, solitary individuals of both species preferentially associated with conspecific over heterospecific shoals. Stickleback shoal choices under low-risk conditions may have been influenced by interspecific competition for food. In feeding experiments, minnows were more efficient foragers than sticklebacks, so it should benefit sticklebacks to avoid minnows unless predation risk is high. Therefore, for armored prey, the benefits of associating with more vulnerable prey appear to override the costs of both the oddity effect and food competition when predation risk is high.

In the presence of predators, individuals in groups often experience higher survival than solitary individuals. Defense benefits of groups can include dilution of predation risk, increased predator detection, increased efficiency of evasive maneuvers, the ability to hide behind group mates (selfish herd effects), transmission of information among group members, and predator confusion (reviewed in Godin, 1997; Pitcher and Parrish, 1993). Most defenses probably are more effective in single-species groups (Allan and Pitcher, 1986; Godin, 1997). However, the presence of heterospecifics can sometimes offer benefits such as increased efficiency of predator detection (FitzGibbon, 1990; Jacobsen and Ugelvik, 1994; Parker and Shulman, 1986; Thompson and Barnard, 1983).

The predator confusion effect appears to play an important role in determining which groups solitary individuals choose to join. According to this hypothesis, predatory attacks on a group are less likely or less effective when a predators' sensory system is confused by the movements of multiple potential prey (Ohguchi, 1981). Predator confusion is enhanced by the similarity of the appearance of group members (Godin, 1997). Group members that are odd in size (Theodorakis, 1989), color (Landeau and Terborgh, 1986; McRobert and Bradner, 1998; Mueller, 1975), or behavior (Parrish et al., 1989) suffer disproportionately higher rates of attack. If the oddity effect plays a substantial role in predation avoidance, solitary individuals should join groups whose members are most similar to themselves in appearance (Krause and Godin, 1994; Krause et al., 2000; Peuhkuri et al., 1997). Such preferences may be particularly pronounced when predation risk is high (Krause, 1994; Ranta et al., 1992).

Under what conditions should solitary individuals preferentially join groups of individuals that are different in appearance? That is, what conditions can override the oddity effect? The presence of morphological defenses (spines and armor) can provide individuals with some protection from predation. We hypothesize that armored individuals may make different group choices than those that are not armored because they could benefit from associating with more vulnerable prey.

Fish prey guilds in streams in south-central Saskatchewan, Canada, are relatively simple systems, often consisting solely of cyprinids (minnows, dace, shiners: Cyprinidae) and sticklebacks (Gasterosteidae). Cyprinids and sticklebacks are similar in body size but differ in their degree of body armor. Sticklebacks have pelvic armor and associated spines, whereas minnows lack body armor and spines. Fathead minnows (Pimephales promelas) and brook sticklebacks (Culaea inconstans), the most common species in many of these systems, occur together in the same microhabitats, eat similar prey (microinvertebrates), and are eaten by the same types of predators (Scott and Crossman, 1973).

We performed a series of experiments to examine the dynamics of single-species versus mixed-species shoals for fathead minnows and brook sticklebacks. We addressed the following questions: (1) in a mixed-species group, do minnows experience higher predation than sticklebacks? (2) Do minnows and sticklebacks differ in their foraging efficiencies? (Because shoal mates can influence foraging success, shoal choice decisions should also incorporate the effects of competition for food.) (3) Do solitary minnows and sticklebacks prefer conspecific or heterospecific shoals? (4) Does shoal choice differ according to level of predation risk?

METHODS

Collection and maintenance

For all but one of the experiments, fishes were collected and trials were conducted in August–October 2000. For the mixed-species foraging experiment, both collection and trials were done in April 2002. We captured all fishes in south-central Saskatchewan, Canada.

Perch (⁠

\(mean\ {\pm}\ SD\ standard\ length\ {=}\ 17.05\ {\pm}\ 0.840\ cm\)

⁠;

\(\mathit{n}\ {=}\ 19\)

⁠) were kept in 400-l holding tanks in the laboratory on a 14 h light:10 h dark light cycle at

\(17{^\circ}\ {\pm}\ 2{^\circ}C\)

and were fed approximately every 5 days with earthworms (Lumbricus sp.). All prey fishes were initially maintained in an approximately 5000-l outdoor pool but were moved to 74-l holding tanks in the laboratory a minimum of 5 days before testing. The light cycle for prey fishes was 16 h light:8 h dark, and water temperatures were

\(20{^\circ}\ {\pm}\ 1{^\circ}C\)

⁠. Fishes were fed a maintenance diet of commercial flake food and frozen brine shrimp (Artemia salina).

Predation experiment

The purpose of this experiment was to determine whether armored (sticklebacks) and nonarmored (minnows) prey differ in vulnerability to predation. We exposed mixed-species shoals of minnows and sticklebacks to predation by yellow perch (Perca flavescens) and compared the two species for (1) strike latency, (2) capture efficiency, and (3) survival.

For the last two feedings before testing, we fed the perch an equal number of fathead minnows and brook sticklebacks (approximately two prey/perch), so that the perch had recent experience with both prey species. There were 19 trials and perch and prey fishes were each tested only once.

Predation trials were conducted in 189-L plastic containers (⁠

\(100\ L\ {\times}\ 47\ W\ {\times}\ 40\ H\ cm\)

⁠) lined with a gravel substrate. A perch was placed at one end of each chamber inside a plastic bucket (⁠

\(diameter\ {=}\ 29\ cm\)

⁠) that was open on both ends (no floor or lid). The sides of the bucket had been drilled with several dozen holes to allow movement of water. Two brook sticklebacks and two fathead minnows were placed inside an open-ended plastic bucket (⁠

\(diameter\ {=}\ 24\ cm\)

⁠) at the opposite end of the container. Fishes were not measured before testing to minimize the stress associated with handling. We matched each group of four prey fishes for approximate size to minimize potential size-based oddity effects (Theodorakis, 1989) and because perch may generally prefer smaller fish prey (Paszkowski and Tonn, 1994). Surviving minnows and sticklebacks were measured at the end of the trial, and there was a maximum difference of 4 mm standard length between minnows and sticklebacks in the same group (⁠

\(mean\ difference\ {\pm}\ SD:\ 1.8\ {\pm}\ 1.09\ mm\)

⁠). For the nine trials in which at least one minnow and one stickleback survived, minnows were larger in three trials, sticklebacks were larger in three trials, and the largest minnow and stickleback were the same length in three trials. Overall mean standard length (± SD) was

\(43.4\ {\pm}\ 1.78\ mm\)

(⁠

\(\mathit{n}\ {=}\ 23\)

⁠) for surviving minnows and

\(44.9\ {\pm}\ 2.25\ mm\)

(⁠

\(\mathit{n}\ {=}\ 16\)

⁠) for surviving sticklebacks. Therefore, there is no evidence of any systematic differences in size between the two species of prey fishes in the experiment.

Trials began 5 h after the fishes were placed in the testing containers. Because individual prey fishes may have had variable experience with large piscivorous fishes, we “warned” the prey that a dangerous predator was nearby (Mathis and Smith, 1993) by adding 2.5 ml each of skin extracts (containing alarm pheromones) from fathead minnows and brook sticklebacks (see preparation methods below). After 3 min, we simultaneously removed the barriers from the perch and prey fishes. For 10 min, we recorded for each prey species: (1) time of each strike and (2) whether strikes resulted in capture. We also recorded number captured after 16 h. These time intervals were chosen arbitrarily but should be reasonable indications of short-term and long-term effects.

For latency to first strike, numbers of strikes, and number captured after 16 h, we used a Wilcoxon matched-pairs test (Siegel, 1956) to compare the two prey species. Matched pairs were strikes toward minnows and sticklebacks in the same trial. If no strikes occurred, a latency score of 600 s was recorded. Capture efficiencies for the two species were compared using a Fisher's Exact probability test (Siegel, 1956), with efficiency measured as proportion of trials where the first strike resulted in a capture. Only first strikes were included in the analysis to maintain statistical independence.

We prepared skin extracts (containing alarm cues) for fathead minnows and brook sticklebacks using the methods of Lawrence and Smith (1989). Donor minnows (⁠

\(mean\ {\pm}\ SD\)

⁠,

\(standard\ length\ {=}\ 43.1\ {\pm}\ 5.37\ mm\)

⁠;

\(\mathit{n}\ {=}\ 16\)

⁠) and sticklebacks (⁠

\(51.5\ {\pm}\ 6.09\ mm\)

⁠,

\(\mathit{n}\ {=}\ 13\)

⁠) were killed with a blow to the head, and patches of skin were removed from both sides of each fish. The skin samples were homogenized in distilled water using a Polytron homogenizer (minnows: 65 mm² of skin in 650 ml of water; sticklebacks: 42 mm² of skin in 420 ml of water). The solutions were filtered through glass wool to remove solid particles and were frozen at approximately −20°C. Distilled water was similarly frozen to use as a control (blank) stimulus.

Foraging experiments: mixed-species and single-species shoals

Group members can experience reduced foraging success through either exploitative or interference competition. In either case, individuals in the less successful species should experience reduced success in a mixed-species shoal. We compared the foraging efficiencies of minnows and sticklebacks in two experiments; the first examined mixed-species shoals, and the second examined single-species shoals. In a mixed-species group, differences in foraging success between species could be due either to exploitative or interference competition. If one species is more efficient at utilizing prey, even in the absence of the interspecific competitor, then the difference can be attributed to exploitative competition.

In the mixed-species experiment, we tested 18 shoals consisting of 3 minnows (⁠

\(mean\ {\pm}\ SD,\ standard\ length\ {=}\ 41.4\ {\pm}\ 3.66\ mm\)

⁠) and 3 sticklebacks (⁠

\(42.0\ {\pm}\ 3.09\ mm\)

⁠). Each shoal was housed in a 9.2-l aquarium and was fed daily with live Tubifex worms for at least 1 week before testing. Worms were always introduced into the bottom center of the tank. During each test, three worms were introduced to the tank, and we counted the number of worms eaten by each species. For the statistical analysis, we compared the number of worms eaten by minnows to an expected random distribution using a Wilcoxon test (Siegel, 1956). For the single-species experiment, testing tanks (9.2 l) contained either six fathead minnows (⁠

\(mean\ {\pm}\ SD\ standard\ length\ {=}\ 45.4\ {\pm}\ 4.68\ mm\)

⁠) or six brook sticklebacks (⁠

\(43.4\ {\pm}\ 4.14\ mm\)

⁠). Individual shoals (⁠

\(\mathit{n}\ {=}\ 8\)

for each species) were kept together for approximately 1 month before testing and were fed daily with brine shrimp (sticklebacks) or a mixture of flake food and brine shrimp (minnows). For 5 days before testing, all fishes were fed only with brine shrimp. The food was always introduced at the front left corner of the tank at approximately 1600–1800 h.

We used frozen brine shrimp as the prey in the single-species experiment. We thawed several grams of frozen brine shrimp in distilled water so that individual shrimp could be identified and filled pipettes with water containing 10 brine shrimp. The brine shrimp were pipetted into the front corner of the testing tank, and a stop watch was activated when the prey entered the water. We recorded latency to consume the 10 brine shrimp. If all 10 prey were not consumed in 15 s (two trials), we stopped the trial and recorded a latency score of 15 s. In all cases, at least five out of six fishes were located at the front one-third of the testing tank at the beginning of the trial. One tank of sticklebacks had a dead individual on the testing day, so this tank of five fish was not included in the trials. We compared the two species for the time used to consume 10 prey with a two-tailed Mann-Whitney U test (Siegel, 1956).

Shoal choice experiment

We tested whether solitary minnows and sticklebacks preferred conspecific versus heterospecific shoals. Because costs and benefits could vary depending on the level of predation risk, we tested individuals under both low-risk and high-risk conditions. In high-risk trials, we exposed focal fishes to alarm pheromones from either conspecifics or heterospecifics. In low-risk trials, focal fishes were exposed to a blank control. Our goals were to determine (1) whether minnows and sticklebacks differed in preferences of conspecific versus heterospecific shoals and (2) whether shoal choice is influenced by level of predation risk.

Each testing arena was a central 37-l focal tank with a 9.2-l stimulus tank positioned at each end of the focal tank's long axis. The focal tank had a gravel substrate and an airstone located in the center of the back wall. Plastic tubing was attached to the airline to serve as a stimulus injection tube. This position allowed the stimulus solution to be rapidly dispersed throughout the tank; dye trials indicated complete dispersion in less than 10 s. The stimulus tanks also were supplied with airstones, but none of the tanks contained filters because they would have impeded the view of both the focal fish and the observer. Water in both the stimulus and focal tanks was changed between trials. Focal tanks were separated from stimulus tanks by removable, white opaque dividers.

The stimulus tank on one randomly chosen end contained six fathead minnows, and the tank at the other end contained six brook sticklebacks. Stimulus fishes were the same as those used in the single-species foraging experiment. Three days before testing, either a fathead minnow (⁠

\(46.3\ {\pm}\ 9.08\ mm\)

⁠) or a brook stickleback (⁠

\(42.6\ {\pm}\ 9.38\ mm\)

⁠) was placed in the focal tank. Stimulus fishes and focal fish continued to be fed daily during this period. We tested focal fish only once, and stimulus fish were used in approximately 10 different trials with 4 days between trials.

Predation risk was manipulated by exposing the focal fishes to either a distilled water control stimulus (low risk) or to skin extracts (high risk) from either conspecifics or heterospecifics. The skin extracts were from the same batch as those used in the predation experiment. At the beginning of each trial, we removed 60 ml of tank water from the stimulus injection tube with a syringe and discarded it to remove any stagnant water that may have collected in the tube. An additional 60 ml of tank water was removed and saved. We injected 5 ml of a randomly chosen stimulus solution (distilled water control, fathead minnow skin extract, or brook stickleback skin extract) into the stimulus injection tube, followed by the 60 ml of tank water. The tank water injection ensured that all of the stimulus solution had been flushed from the tubing. The stimulus was injected at a rate of approximately 2 ml/s and reached the tank in approximately 20 s. We recorded a qualitative assessment of the focal fish's response to the stimulus as dashing (rapid, apparently disoriented swimming; Lawrence and Smith, 1989), freezing (cessation of all activity for a minimum of 20 s), decreased time active, increased time active, or no apparent change in time active. Dashing, freezing, and decreased activity are widely accepted to indicate fright reactions in fishes (e.g., Lawrence and Smith, 1989). We compared the proportion of fright responses for each treatment using chi-square tests.

Approximately 15 s after the stimulus injection was completed, we simultaneously removed the barriers between the focal tank and the two stimulus tanks. We began to record position data approximately 10 s after the barriers were removed. We divided the tank into three equal-sized compartments with a line marked on the glass. When a focal fish was in the middle compartment, it was considered to be in a neutral area; otherwise it was considered to be near one of the stimulus shoals. Trials lasted 10 min (600 s). We quantified the amount of time fish spent in each end compartment. For statistical purposes, we subtracted the amount of time spent in the neutral area and converted the data for the remaining two compartments to percentage of time spent near conspecifics. Thus, a minnow that spent 100 s in the compartment near sticklebacks and 300 s in the compartment near minnows received a score of 25%. Any focal fish that did not move for the entire test was deleted from the analyses. We compared scores for the three treatments using a Kruskal-Wallis one-way ANOVA followed by nonparametric multiple comparisons tests (Zar, 1999). In cases where statistical results were nonsignificant (⁠

\(\mathit{p}\ {>}\ 0.05\)

⁠), we did not conduct power analyses because values of n and were relatively high; power analyses are generally important only when these values are low (Johnsson, 1996). For each species, we compared treatments for (1) initial response to the chemical stimuli (before the barriers were removed), (2) time spent in the neutral compartments, and (3) time spent in the compartments near conspecific and heterospecific shoals.

RESULTS

Predation experiment

Perch struck at fathead minnows significantly sooner than they struck at sticklebacks (⁠

\(\mathit{Z}\ {=}\ 2.67\)

⁠,

\(\mathit{p}\ {=}.0076\)

⁠, two-tailed; Figure 1a). First strikes were directed toward minnows in 12 out of the 14 trials where perch struck at a prey fish during the first 10 min of the trial. Perch also directed significantly more strikes toward minnows than toward sticklebacks during the first 10 min (⁠

\(\mathit{Z}\ {=}\ 2.41\)

⁠,

\(\mathit{p}\ {=}.016\)

⁠, two-tailed; Figure 1b).

Although minnows were more likely to be attacked by perch, they also were more successful at evading capture. During the first 10 min, only two out of 11 first strikes at fathead minnows resulted in captures. In contrast, four out of five first strikes at sticklebacks resulted in captures (Fisher's Exact probability test,

\(\mathit{p}\ {=}.045\)

⁠, two-tailed).

Only five minnows and four sticklebacks were eaten during the first 10 min, so these data were not analyzed statistically. By the end of 16 h, at least one prey fish had been eaten in 18 of the 19 trials. The mean number of minnows (⁠

\({\pm}SE;\ 0.8\ {\pm}\ 0.20\)

⁠) and sticklebacks (⁠

\(1.2\ {\pm}\ 0.18\)

⁠) eaten were not significantly different after 16 h (⁠

\(\mathit{Z}\ {=}\ 1.29\)

⁠,

\(\mathit{p}\ {=}.20\)

⁠, two-tailed).

Foraging experiment

In the mixed-species experiment, minnows consumed more prey than expected by chance (⁠

\(\mathit{T}\ {=}\ 145\)

⁠,

\(\mathit{p}\ {=}.01\)

⁠, two-tailed; Figure 2a). In the single-species experiment, the 10 prey were consumed in 3.03–6.63 s by minnows and in 4.93–15.0 s by sticklebacks. Minnows consumed the brine shrimp significantly faster than sticklebacks (⁠

\(\mathit{U}\ {=}\ 6\)

⁠,

\(\mathit{p}\ {=}.01\)

⁠, two-tailed; Figure 2b).

Shoal choice experiment

Initial response to the chemical stimuli

Our data confirmed that the treatments were successful at simulating high-risk and low-risk conditions. For both solitary minnows and sticklebacks, the most common response to the blank stimulus was either increased activity or no change in activity. Fright responses (dashing, freezing, decreased activity) were often observed in response to both conspecific and heterospecific skin extracts but were rare in response to the blank (Table 1). The alarm treatments resulted in a higher proportion of fright responses (dashes, freezing, decreased activity) for both minnows (⁠

\({\chi}^{2}\ {=}\ 9.93\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {<}.01\)

⁠) and sticklebacks (⁠

\({\chi}^{2}\ {=}\ 30.70\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {<}.001\)

⁠).

Time in the neutral compartment

In general, solitary individuals appeared to be attracted to the stimulus fishes. Both solitary minnows and sticklebacks spent most of their time outside of the neutral compartments during trials. Minnows spent only 19% of their time in the neutral compartments, and there was no significant difference among treatments (⁠

\(\mathit{H}\ {=}\ 0.437\)

⁠,

\(\mathit{n}\ {=}\ 35\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {>}.80\)

⁠). Sticklebacks spent 22.5% of their time in the neutral compartment, and there also was no significant difference among treatments (⁠

\(\mathit{H}\ {=}\ 1.183\)

⁠,

\(\mathit{n}\ {=}\ 33\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {>}.30\)

⁠).

Shoal choice

In all three treatments, solitary minnows tended to spend more time near conspecific shoals than near heterospecific shoals, and there was no significant difference among the three treatments (⁠

\(\mathit{H}\ {=}\ 1.45\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {>}.30\)

⁠; Figure 3a). Minnows spent more than half of their time near conspecific shoals in 75% of blank trials, 64% of trials with the minnow alarm extract, and 67% of trials with the stickleback alarm extract.

In contrast, shoal choice by solitary sticklebacks was significantly influenced by treatment (⁠

\(\mathit{H}\ {=}\ 11.510\)

⁠,

\(df\ {=}\ 2\)

⁠,

\(\mathit{p}\ {<}.01\)

⁠, Figure 3b). Sticklebacks spent significantly less time near conspecifics in both the minnow alarm extract treatment (⁠

\(\mathit{Q}\ {=}\ 3.12\)

⁠,

\(\mathit{p}\ {<}.01\)

⁠) and the stickleback alarm extract treatment (⁠

\(\mathit{Q}\ {=}\ 2.43\)

⁠,

\(\mathit{p}\ {<}.05\)

⁠) than in the blank treatment. Shoal choice scores did not differ between the two alarm treatments (⁠

\(\mathit{Q}\ {=}\ 0.56\)

⁠,

\(\mathit{p}\ {>}.50\)

⁠). Brook sticklebacks spent more than half of their time near conspecific shoals in 82% of blank trials, but in only 17% of trials with the minnow alarm extract and 10% of trials with the stickleback alarm extract.

DISCUSSION

Shoal choice decisions were different for fathead minnows and brook sticklebacks, apparently reflecting differences in costs and benefits of shoaling for the two species. Our results suggest that shoal choice decisions incorporate information concerning the perceived levels of both predation risk and food competition. Assessment of predation risk apparently includes information about whether predators are active in the area (i.e., presence of alarm pheromones), potential influences of the oddity effect, and the vulnerability of group mates to predators (i.e., armored vs. nonarmored prey).

Under conditions of high threat, solitary minnows preferentially associated with conspecifics over heterospecifics as predicted by the oddity effect. Preferences for associating with similar individuals sometimes increases under conditions of high predation threat (Allan and Pitcher, 1986; Krause, 1994; Ranta et al., 1992; Wolf, 1985), but this was not the case in our study. Equal preferences for associating with similar-sized conspecifics under both low and high levels of predation threat have been reported for fathead minnows and for bluntnose (Pimephales notatus) and stoneroller (Campostoma anomalum) minnows (Theodorakis, 1989).

Behavior of sticklebacks under high-threat conditions was dramatically different from behavior of minnows. When risk was high, solitary sticklebacks preferentially associated with minnow shoals over conspecific shoals. We attribute this preference to differential vulnerability of the two species to predation. When minnows and sticklebacks were in the same shoal, perch attacked minnows earlier and more often than sticklebacks. Our experimental setup did not allow prey to hide or flee the area, so long-term (16 h) survival was the same for minnows and stickleback. In nature, however, sticklebacks in a mixed shoal should be able to perform successful avoidance behaviors while the predator focuses its attack on minnows. Fathead minnows also are more vulnerable than brook sticklebacks to predation by northern pike (Esox lucius; Robinson, 1989), and sticklebacks generally show less avoidance behavior than fathead minnows to large fish predators (Abrahams, 1995). The difference in vulnerability between the two species is probably due to the presence of body armor and spines on sticklebacks (McLean and Godin, 1989; Godin and Valdron Clark, 1997; Reist, 1980). Mixed-species aggregations occur in a variety of vertebrate taxa, and preferential association of protected species with more vulnerable species could play an important role in group dynamics. For example, unpalatable toad (Bufo bufo) tadpoles will associate with tadpoles of the palatable common frog (Rana temporaria), but common frog tadpoles avoid aggregations of toad tadpoles (Griffiths and Denton, 1992). Mixed-species groups also provide protection for Grant's gazelles (Gazella granti) because cheetahs preferentially attack the smaller Thomson's gazelles (G. thomsoni; FitzGibbon, 1990).

Under conditions of relatively low predation risk, solitary individuals of both species preferred to associate with conspecifics over heterospecifics. Several other studies have also shown that small fishes tend to associate more closely with conspecifics than with heterospecifics (Allan, 1986, Allan and Pitcher; 1986, Parrish, 1989). Although preferential association with conspecifics may offer a variety of benefits (Pitcher and Parrish, 1993), the oddity effect appears to be of primary importance for the minnows in our study and for at least some other species (Krause and Godin, 1994).

Why do sticklebacks preferentially associate with conspecifics only when predation risk is low? Sticklebacks could be attracted to conspecifics for reasons that are not related predation risk, such as possible benefits associated with social behavior. However, one explanation for the observed association may be that sticklebacks avoid minnows under low-risk conditions, due at least in part to competition for food. In mixed-species shoals, minnows consumed more prey (worms) than sticklebacks. When tested in the absence of heterospecific fishes, groups of minnows consumed prey almost twice as fast as groups of sticklebacks. This latter finding suggests that differential foraging success in this study is the result of exploitative competition rather than of interference competition. Differences among phenotypes in foraging abilities can result in assortative grouping (Peuhkuri, 1997; Ranta et al. 1993, 1994) and can play an important role in group stability (Griffiths, 1991; Griffiths and Denton, 1992; Lindström and Ranta, 1993).

If the oddity effect influences survival of individuals in groups, solitary individuals should associate preferentially with individuals that are similar in appearance. The behavior of the minnows in our study is consistent with this prediction under conditions of both low and high predation risk. In contrast, sticklebacks switched from association with conspecifics under low-risk conditions to association with minnows under high-risk conditions. Preference for conspecifics in low-risk conditions may have been due in part to avoidance of food competition with minnows. Sticklebacks prefer to associate with minnow shoals in high-risk conditions apparently because the benefits of associating with more vulnerable prey override any disadvantage due to either food competition or the oddity effect.

Figure 1

Attacks by yellow perch on mixed-species shoals of fathead minnows and brook sticklebacks in 10-min trials. Data are

\(means\ {+}\ 95{\%}\)

confidence intervals for (a) latency to first strike and (b) number of strikes. Sample size is indicated by the number in the box. The p value is from a Wilcoxon matched-pairs test

Figure 2

Foraging efficiency of fathead minnows and brook sticklebacks. (a) Percentage of worms eaten by minnows and sticklebacks in mixed-species shoals. Data were compared to a random expectation of 50% using a Wilcoxon matched-pairs test. (b) Time that it took monospecific groups of 6 minnows and 6 sticklebacks to consume 10 brine shrimp prey. The p value is from a Mann-Whitney U test. All data are presented as

\(means\ {+}\ 95{\%}\)

confidence intervals, and sample size is indicated by the number in the box

Figure 3

Shoal choice by solitary (a) fathead minnows and (b) brook stickleback under conditions of low (blank) and high (minnow and stickleback alarm chemicals) predation risk. Solitary individuals were given a simultaneous choice between a group of six minnows or six sticklebacks. Data are percent time (⁠

\(means\ {+}\ 95{\%}\)

confidence intervals) spent near the conspecific shoal. The dashed line indicates the null expectation of equal time spent near conspecific and heterospecific shoals. Sample size is indicated by the number in the box. The p values are from Kruskal-Wallis nonparametric ANOVA tests

Table 1

Initial behavioral responses of solitary fathead minnows and brook sticklebacks to chemical treatments.

	Blank	Conspecific extract	Heterospecific extract
Fathead minnows
Dashing (fright)	1	8	3
Freezing (fright)	0	0	1
Decreased activity (fright)	0	0	1
Increased activity (neutral)	4	3	6
No apparent change (neutral)	7	0	1
% Fright responses	8.3	72.7	41.7
Brook sticklebacks
Dashing (fright)	0	2	1
Freezing (fright)	0	6	4
Decreased activity (fright)	0	3	3
Increased activity (neutral)	1	0	1
No apparent change (neutral)	10	1	1
% Fright responses	0.0	91.7	80.0

Stimuli were distilled water (blank) and skin extracts (alarm cue) from conspecifics and heterospecifics. Dashing, freezing, and decreased activity are considered to be indicators of an antipredator (fright) response. Increased activity and no apparent change in activity are not considered to be antipredator responses (neutral).

We thank Reehan Mirza and Mike Pollock for help with the collection of fishes. Funding was provided by Southwest Missouri State University and the Natural Sciences and Engineering Research Council of Canada.