Growth rates of wild stream-dwelling Atlantic salmon correlate with activity and sex but not dominance

Authors

  • Keith M. Martin-Smith,

    1. †Fisheries Research Services Freshwater Laboratory, Faskally, Pitlochry, Perth PH16 5LB, UK; and ‡Fish Biology Group, Division of Environmental & Evolutionary Biology, Institute of Biomedical & Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK
    Search for more papers by this author
    • *

      Present address: Project Seahorse, Department of Biology, McGill University, 1205 Avenue Docteur Penfield, Montréal, Québec, Canada H3A 1B1.

  • John D. Armstrong

    1. †Fisheries Research Services Freshwater Laboratory, Faskally, Pitlochry, Perth PH16 5LB, UK; and ‡Fish Biology Group, Division of Environmental & Evolutionary Biology, Institute of Biomedical & Life Sciences, Graham Kerr Building, University of Glasgow, Glasgow G12 8QQ, UK
    Search for more papers by this author

J. D. Armstrong, Fisheries Research Services Freshwater Laboratory, Faskally, Pitlochry, Perth PH16 5LB, UK. E-mail: armstrongj@marlab.ac.uk

Summary

  • 1It is generally assumed that high social status confers benefits such as increased resource acquisition and growth rate, higher survival and/or increased reproductive output relative to subordinate individuals.
  • 2The hypothesis that dominant juvenile Atlantic salmon would have higher growth rates than subordinates in a flow-regulated natural stream was tested. Over 2 years, seven groups of eight size-matched wild fish were tagged with passive integrated transponders (PIT tags) and assessed for dominance using a serial removal method. Fish were then introduced together into an enclosed stretch of stream and allowed to forage freely for 14 days. Their use of space was recorded by an array of eight in-stream detectors. Using in-stream video recording, a correlation between dominance measured prior to and during the trials was confirmed.
  • 3Contrary to our hypothesis, there were no significant correlations between dominance status and specific growth rate.
  • 4Mean growth rate varied significantly between trials. In females, growth declined from May through to August. Maturing male fish exhibited a prepubertal growth spurt during June and July when their rates of growth significantly exceeded those of females.
  • 5There were significant positive relationships between rates of movement and growth of both sexes. Each salmon generally used only a restricted portion of the stream but there were considerable overlaps between the home ranges of individuals.
  • 6Sampling over space and time on two occasions revealed complex variations in rankings of patches with respect to the abundance of drifting invertebrates of aquatic and terrestrial origin.
  • 7Our data are consistent with the hypothesis that a high degree of unpredictability in food supply results in correlation between growth and movement but not the ability to dominate high-quality feeding patches exclusively. Variation in growth between and within sexes might result from trade-offs between the costs (e.g. increased predation risk) and benefits (increased food acquisition rate) of sampling movements. The marked prepubertal growth spurt in maturing males is consistent with a strong advantage of large size of male parr during spawning.

Introduction

The ability to defend resources is thought to be of fundamental importance to individual animals. In most vertebrates competition for resources results in intraspecific dominance hierarchies that commonly cause unequal outcomes for the competitors (Huntingford & Turner 1987). Across a range of taxa, individuals of high social status have been shown to have greater food intake or reproductive success than those of low rank (Fausch 1984; Packer et al. 1995; McElligott & Hayden 2000). However, increased metabolic or mortality costs may be incurred in achieving preferential access to resources (Hogstad 1987; Bryant & Newton 1994; Metcalfe, Taylor & Thorpe 1995). Hence, the net fitness of an animal in relation to dominance rank is best considered in terms of a trade-off function between the costs and benefits of dominance (Packer et al. 1995).

The precise relationships between measures of fitness such as growth rate and dominance status of most animals in natural habitats remain unclear. Roskaft et al. (1986) demonstrated that dominant great tits and pied flycatchers incurred higher metabolic costs but Bryant & Newton (1994) suggested these costs were trivial in relation to the benefits of defending better territories for dippers. In contrast, Huntingford & Garcia de Leaniz (1997) showed a negative correlation between growth rate and dominance of Atlantic salmon (Salmo salar L.), suggesting that under some (artificial) conditions costs could outweigh benefits of dominance. Furthermore, Byers (1997) found no relationship between social rank and foraging efficiency in female pronghorns, while, in the same species, Dennehy (2001) suggested a quadratic relationship with both high- and low-ranked pronghorn females performing better than middle-ranked animals.

Populations of salmonid fishes provide a good model system in which to investigate the costs and benefits of social dominance. Growth measurements provide a good proxy for fitness because growth is indeterminate and there are well-documented benefits of large body size in relation to fecundity, gaining access to mates and survival when smolting (Myers 1984; Quinn & Foote 1994; Thomaz, Beall & Burke 1997). Dominance hierarchies are ubiquitous among salmonids in natural (e.g. Hughes & Dill 1990; Nakano 1995) and artificial streams (e.g. Abbott & Dill 1989; Gotceitas & Godin 1991). In laboratory tanks, dominant fish grow relatively fast and consequently tend to smolt at a younger age (Metcalfe et al. 1990; Thorpe, Metcalfe & Huntingford 1992). Therefore, growth rate links dominance to life-history strategies and appears to be a correlate of the fitness of individuals. Positive correlations between growth and dominance rank can be very strong and may be measured within 6–7 days (Huntingford et al. 1998).

Although the positive correlation between dominance and growth of salmonid fishes in tanks and natural pools has generally been accepted as evidence of a direct relationship between dominance and fitness, there is also evidence that the influence of dominance may be profoundly modified by the distribution of resources. In laboratory studies, measures of aggression were higher in juvenile convict cichlids under predictable patterns of resource distribution (Grand & Grant 1994). In the same study, the number of food items obtained was correlated with dominance status only under predictable food supply whereas it was correlated with mobility (frequency of patch switching) under an unpredictable regime (Grand & Grant 1994).

It is possible that patterns of resource and fish distribution commonly used in laboratory tanks and streams to ascertain effects of dominance status may not reflect the situation in some types of habitat in the wild. Recent work showed that in a near-natural stream many individual Atlantic salmon were highly mobile, and rather than each fish defending a single local territory their home ranges overlapped extensively (Armstrong, Huntingford & Herbert 1999). These behaviour patterns probably resulted from fish tracking fine-scale spatio-temporal variations in the relative qualities of adjacent habitat patches (Gotceitas & Godin 1991; H. Grant & J. D. Armstrong, unpublished data). The distribution of food in shallow natural streams can be more complex than in laboratory tanks and the balance between costs and benefits of dominance may be different. The primary aim of this study was to test whether growth of Atlantic salmon parr was related to dominance and movement under a regime of natural food supply in a shallow riffle habitat.

Growth and dominance may vary with sex because there are profound differences in the life-history strategies adopted by male and female Atlantic salmon. In the UK, females have an invariant life history with little or no gonad development in fresh water and become reproductively active only after a period spent in salt water (Fleming 1998). Conversely males can develop mature gonads in fresh water without migrating and gain significant reproductive success during ‘sneak’ matings (Fleming 1998). Positive relationships between parr size and reproductive success have been reported (Thomaz et al. 1997; Koseki & Maekawa 2000). However, reproductive success is generally greater for larger, anadromous males that have spent time in salt water (Fleming 1998). Life-history theory predicts a trade-off between reproductive growth and measures of fitness (Stearns 1992; Hutchings & Jones 1998) and indeed lower over-winter and smolt survival has been reported for males that spawn as parr (Dalley, Andrews & Green 1983; Myers 1984). Arndt (2000) provides evidence that this may be a result of lower investment in short-term energy storage or growth, although Tucker & Rasmussen (1999) demonstrated that mature parr had assimilated significantly more food than nonmature individuals.

We tested whether relationships were apparent between growth rate, dominance and/or sex of wild juvenile Atlantic salmon under near-natural shallow stream conditions. By using continuous recordings of positions of fish in the stream using PIT tags, we could also test whether dominant and/or relatively fast-growing fish selected the same areas of stream in successive trials and if any relationship existed between growth and rates of movement of individuals.

Materials and methods

EXPERIMENTAL PROCEDURE

The study site for this experiment was a flow-regulated stream channel on the Girnock Burn, Aberdeenshire, Scotland (03°07′W, 57°02′N). A section of stream (19·5 m length, 46·6 m2 surface area) was fitted with mesh screens at the top and bottom to prevent fish from escaping and an array of eight PIT tag detector plates (40 × 100 cm2) were set into the stream bed approximately 1·8 m apart (mean = 1·81 m, SD = 0·13, n = 8). The areas between the PIT tag detectors (mean width = 2·12 m, SD = 0·39, n = 9; mean thalweg depth = 13 cm, SD = 2, n = 9) were landscaped with gravel, cobbles and boulders to resemble natural riffle areas in the adjacent Girnock Burn. Flow was regulated by an upstream sluice gate to give a mean surface current speed of 0·34 m s−1 in 1998 (SD = 0·18, n = 70) and 0·37 m s−1 in 1999 (SD = 0·22, n = 64). The mesh screens were cleaned when appropriate (every 3–4 days) and water temperature was recorded every hour with a submersible data logger.

Groups of eight size-matched wild fish (n = 3 in 1998, n = 4 in 1999) were caught by electrofishing in the Girnock Burn downstream of the experimental channel. Dates for each trial, mean water temperatures and initial size of fish are given in Table 1. The use of eight fish in the experimental area was based on the mean value of natural densities of 1+ and 2+ fish (range 0·04–0·27 m−2) in the Girnock Burn estimated by Buck & Hay (1984). Fish were anaesthetized with benzocaine solution and fork length (FL) to nearest mm and mass (Wt) to nearest 0·1 g were measured. They were then each implanted with a PIT tag in the peritoneal cavity and marked with a unique combination of alcian blue spots on the left-hand side of the body.

Table 1.  Mean daily water temperature, initial sizes of fish and sex ratio for each experimental trial. Note that the number of fish completing each trial does not always equal eight due to loss of fish by predation
Trial. no.DatesMean daily water temp. °C (range)Sex ratio (maturing M:immature M:F)Mean initial FL (mm ± SD)Mean initial Wt (g ± SD)
98/218/06–02/07/9813·9 (11·3–17·7)5 : 0 : 2 98 ± 511·2 ± 1·8
98/327/07–10/08/9814·3 (11·8–15·5)5 : 0 : 2106 ± 214·0 ± 1·4
98/424/08–07/09/9812·3 (9·7–14·2)4 : 1 : 2107 ± 414·8 ± 1·6
99/125/05–07/06/9911·3 (9·1–14·3)3 : 2 : 3103 ± 211·2 ± 1·0
99/223/06–07/07/9913·9 (11·1–15·8)3 : 0 : 3111 ± 315·2 ± 1·6
99/323/07–06/08/9915·2 (12·6–17·7)3 : 0 : 4117 ± 218·1 ± 1·4
99/423/08–05/09/9913·6 (10·9–16·0)3 : 0 : 3121 ± 220·3 ± 1·5

Fish were tested for dominance using a serial removal procedure adapted from Metcalfe et al. (1989). An in-stream flume (160 × 60 cm2) was constructed in the channel below the experimental section and manipulated so that a preferred location (high flow and a flat elevated rock) was created close to the observation window. Fish were introduced to the flume and allowed to settle for at least 2 days. All food items were natural drifting invertebrates entering from the experimental section with the incoming water. Fish were observed from a darkened hide for two 30-min periods, separated by at least 3 hours. During each 30-min period, fish were scored every 5 min for the spatial position they had adopted on a scale from 1 to 5 (lowest to highest current speeds). Furthermore, fish were continuously observed for feeding attempts, which were given a score of 1 (or 2 if the item was contested by another fish), and aggressive encounters won (score 5). At the end of the second observation period, the total score for each fish was calculated and the highest scoring fish defined as the dominant individual if there was a differential of 20 or more points from the next fish. If such a differential did not exist, a third observation period was undertaken. The dominant fish was given a dominance rank of 1 and removed to a holding tank. The flume was left undisturbed until the next day and the procedure was repeated with the remaining fish. This process was continued until all the fish had been ranked. Dominance relationships were usually clear for the first four or five fish, after which the flume had to be halved in size in order to force the remaining fish to interact with each other. All fish that had been ranked for dominance were held together in a tank with natural invertebrate drift as a food supply.

After dominance testing, fish were placed together into the middle of the experimental section of stream (between detectors 4 and 5) and then left undisturbed for 14 days. All movements of fish over detectors were logged continually with a computer system and the data were downloaded every 1–5 days. Fish were not fed; all food items were natural invertebrate production from the stream. At the end of the experimental period fish were caught by electrofishing, and FL and Wt were re-measured. At the end of each trial, fish were killed with an overdose of benzocaine and dissected for sex determination. Male fish were classified as immature or maturing based on macroscopic appearance of the testes.

VIDEO RECORDING AND ANALYSES

In 1999, four underwater video cameras (System-Q, Wolverhampton) were installed in the stream to provide behavioural observations of fish once they had been introduced into the stream. These cameras were placed in pairs on either side of the downstream edge of PIT tag detector Plate(s) 2 and 7 (numbering from the upstream end of the stream). Each camera had an angle of view of 84° and each pair of cameras could view the entire width of the PIT tag detector as well as the area immediately below the detector. The four cameras were connected to a multiplexer and video recordings were made sequentially from each camera on a time-lapse video recorder with a sampling rate of 10 frames per second. Date and time stamps were recorded by the multiplexer for each frame and regularly synchronized with the computer clock for the PIT tag detectors. Recordings were made for at least the final 7 days of each trial.

Video observations were confined to putative interactions between pairs of fish. Thus PIT tag records for the two detectors under observation were isolated and examined for records of two or more different fish within 60 s of each other. When such records were found, the video footage for the cameras concerned were viewed for a period of 5 min centred on the time of the PIT tag record. When pairs of fishes were observed interacting they were followed frame-by-frame during the course of their interaction. Individual fish were identified from left-side profiles when their alcian blue markings could be seen, especially when cross-correlated with PIT tag records of possible identities. Interactions were classified as neutral if only a single individual was observed or individuals did not approach within approximately three body lengths of each other. Interactions were classified as aggressive if any of the following behaviours were observed: parallel swimming, fin-erection, quivering, chasing, nipping or biting. In these cases an animal was considered to be a victor if it initiated the aggressive behaviour and returned to its former location following the interaction. This method could only provide relative ranks between pairs of fish, not absolute dominance ranks.

INVERTEBRATE SAMPLING

To ascertain the spatial and temporal variability in food supply, samples of invertebrate drift were collected at the end of trials 99/2 and 99/3. Drift nets (40 × 25 cm2 opening, 250 µm mesh) were placed across the stream at the lower edge of PIT tag detectors 2, 4, 6 and 8 to ensure that they had the same volume of water flowing through them. Invertebrate samples were collected over 6-h periods for 2 days starting at 16.00 on 8 July 1999 and 16.00 on 8 August 1999. Nets were removed from the stream from downstream and replaced from upstream to avoid disturbance to downstream areas. All samples were thoroughly washed from the net into 70% alcohol for preservation. Samples were completely enumerated and classified to Family for aquatic orders Diptera, Ephemoptera, Plecoptera, Megaloptera, Trichoptera and Coleoptera and to Order for a variety of low-abundance taxa. Wet masses were measured for total aquatic invertebrates and total terrestrial invertebrates.

DATA ANALYSES

Specific growth rate (SGR) as percentageWt d−1 was calculated as SGR = 100[ln(Wt2) – ln(Wt1)]/(T2 – T1) where Wt1 and Wt2 were masses at times T1 and T2, respectively (changes in length were within the range of measurement error and so were not used). The PIT tag data were parsed to eliminate multiple sequential readings from the same detector, since a fish that occupied an area near a detector plate would continually trigger the detector without necessarily crossing it. A new variable, crossing rate (Xrate), was created as a measure of fish movement and calculated as the number of sequential detector records per day. A pilot study on crossing rates over time, in addition to previous work (Armstrong, Braithwaite & Huntingford 1997), suggested that the fish took 3–5 days to establish home ranges. Thus only data from the last 7 days of each 14-day experimental period were considered for further analysis.

Variation in SGR was tested using three-way analysis of variance (anova) with fixed factors year, month and sex. To remove significant difference between months, SGR was standardized to Z-scores (Z[SGR]) for further analyses (Zar 1999). The relationship between Z[SGR] and dominance was tested using Kruskal–Wallis nonparametric anova on the whole data set and for each sex separately. The relationship between SGR and Xrate was tested using least-squares linear regression for each sex separately and comparison between sexes was made using analysis of covariance with Xrate as a covariate and sex as a fixed factor.

Static overlap between home ranges of dyads of fish was analysed by nonparametric correlation (Doncaster 1990). Significant positive correlations indicated that home ranges overlapped to a large extent while significant negative correlations showed that fish avoided each other’s home range. To ascertain whether particular areas of the stream were more profitable the weighted ‘centre’ of the home range was calculated and a nonparametric correlation performed against growth rate. χ2 statistics were calculated to assess the relationship between dominance before and during the trials with the null hypothesis that the victor in any interaction between pairs recorded by video was random with respect to the initial assessment of rank.

Invertebrate abundance and biomass were non-normally distributed (Shapiro–Wilk statistic), even following power transformations. Therefore, to look at changes in invertebrate abundance and biomass between nets, matrices of rank of each sample grouped by day or 6-h time period were compiled and Kendall’s rank correlation coefficient (W) was calculated. This was converted to the χ2 statistic and compared with the critical value of this statistic with a reduced significance level of 0·01 to control for multiple comparisons (Zar 1999). All other tests used a significance level of 0·05 and all statistical tests were conducted using SPSS 9·0.

Results

Eight fish were lost in total during the experiments, presumably due to predation, so that not all trials were completed with eight fish (Table 1). However some of these fish were lost close to the end of the trial so that space-use patterns had been established and could be used in further analysis although the SGR and sex of the fish were unknown. There were more males than females overall but this difference was not significant (χ2 = 2·08, d.f. = 1, P = 0·149). All males had maturing testes except two individuals in trial 99/1 and one in trial 98/4. These immature males were excluded from subsequent analyses.

There were significant effects of month, sex and month–sex interaction on SGR (Table 2, Fig. 1). Post-hoc Student–Newman–Keuls tests showed that there were two homogeneous groups of trials with significantly higher SGR earlier in the year (trials 98/2, 99/1 and 99/2) than later in the year (trials 98/3, 98/4, 99/3, 99/4). Overall, males grew significantly faster than females (mean ± SE: males 0·792 ± 0·114, females 0·367 ± 0·103% day−1t46 = –2·54, P = 0·013). There were no significant differences between SGR for males and females for trials conducted in May–June (99/1) or in August–September (98/4 and 99/4) (Fig. 1). However, there was an indication of higher growth rate for maturing males in trial 99/1 but power to detect differences was low because of small sample size. Growth rates ranged from –0·18 to 1·85% Wt day−1.

Table 2.  Results of three-way anova for effects of year, month and sex on specific growth rate of juvenile Atlantic salmon
Sourced.f.F-valueP-value
Year 1 0·06 0·817
Month 324·91<0·001
Sex 115·33<0·001
Year × Month 2 0·68 0·513
Month × Sex 3 5·14 0·005
Year × Sex 1 0·56 0·556
Year × Month × Sex 2 0·36 0·699
Error34  
Total48  
Figure 1.

Specific growth rate (mass) by month and sex for juvenile Atlantic salmon: ◆ maturing males, • immature males, ▪ females. Open symbols indicate range. Significant differences between sexes within months shown as *0·05 < P < 0·01, ***P < 0·001.

There was no correlation between dominance rank and SGR, either overall or within trials (Fig. 2). Kruskal–Wallis nonparametric anovas did not show any significant differences in SGR between dominance ranks either overall or for each sex (χ2 = 8·28, d.f. = 7, P = 0·309 for whole data set; χ2 = 6·77, d.f. = 7, P = 0·453 for males; χ2 = 10·18, d.f. = 7, P = 0·178 for females). These results were upheld even when the most different dominance ranks were compared (SGR for ranks 1 and 2 vs. ranks 7 and 8). Although there was no significant difference in mean dominance rank of males or females (Kruskal–Wallis: χ2 = 3·58, d.f. = 1, P = 0·059), there was a tendency for males to have higher dominance rank.

Figure 2.

Relationship between Z-score for specific growth rate (mass) and dominance rank for juvenile Atlantic salmon: ◆ males, ▪ females. Significant difference between sexes within ranks shown as *0·05 < P < 0·01.

Five fish showed ‘nonsettling’ behaviour whereby a home range was not established and the fish continued to move throughout the whole stream (see Armstrong et al. 1999). These were excluded from analyses of crossing rate data. There were significant linear relationships between crossing rate and SGR for females (y = 0·0160x + 0·0210, R2 = 0·405, F1,15 = 10·23, P = 0·006) and males (y = 0·0135x + 0·4231, R2 = 0·233, F1,20 = 6·07, P = 0·023) (Fig. 3). There was no significant difference in the slope of the line between sexes (ancovaF1,39 = 0·12, P = 0·754).

Figure 3.

Relationship between specific growth rate (mass) and movement rate for juvenile Atlantic salmon: ◆ males, ▪; females, ◆, □ nonsettling fish (see text for details). Solid line is least-squares regression for males, dashed line is least-squares regression for females.

Patterns of space use (location and amount of time spent in that location) varied considerably between trials (Figs 4 and 5). There were no areas that were particularly favoured by dominant fish across trials nor areas that gave better growth. There were no correlations between the location of home range and relative growth (Spearman’s ρ = –0·163, P = 0·284, n = 44 for all fishes; ρ = –0·296, n = 16, P = 0·252 for females; ρ = –0·047, n = 23, P = 0·826 for maturing males), i.e. fish that grew better did not use any particular part of the stream. Each fish generally occupied a limited area of the experimental channel, with 39 of 50 fish spending 90% of their time registering on only one to three detectors (Figs 4 and 5). Despite these limited home ranges, there were 41 significant dyadic interactions (with another 56 marginal at 0·05 < P < 0·10) out of a possible 157 (Table 3). Twenty-one of these were significant positive correlations (with a further 28 marginally significant), indicating that the pairs of fish had essentially the same rankings for their preferred areas of stream. There was considerable overlap between the home ranges and some individuals shared their home range with four other fish.

Figure 4.

Patterns of space use for juvenile Atlantic salmon in a flow-regulated natural stream during 1998. Width of polygon is proportional to the amount of time spent at that location (assuming space use was a linear monotonic function between each detector). Detector 1 was the upstream end of the stream. For each trial fish are presented in ascending order of dominance from top of graph to bottom (i.e. dominance rank 8 to dominance rank 1). Sex codes: mm – maturing male, im – immature male, f – female, u – undetermined.

Figure 5.

Patterns of space use for juvenile Atlantic salmon in a flow-regulated natural stream during 1999. See legend for Fig. 4 for explanation of space-use polygons.

Table 3.  Occurrence of significant (P < 0·05) home range overlap between dyad pairs of juvenile Atlantic salmon with additional occurrence at 0·05 < P < 0·10 in parentheses
TrialNo. possible dyadic interactionsNo. sig. +ve interactionsNo. sig. –ve interactions
98/2155 (5)2 (2)
98/3283 (5)4 (6)
98/4284 (5)5 (7)
99/1284 (4)2 (2)
99/2151 (1)3 (3)
99/3283 (4)3 (5)
99/4151 (4)1 (3)

Few aggressive interactions were observed between pairs of fish in the stream, 21 independent interactions in total with no more than 7 in any one trial (Table 4). Many of the putative interactions were either neutral or only one fish could be observed. Overall, there were significantly more victories by higher-ranked fish (Table 4).

Table 4.  Observed aggressive interactions between pairs of juvenile Atlantic salmon
TrialNo. aggressive interactionsNo. victories by higher-ranked individualχ2 statistic (1 d.f.)P-value
99/1 6 52·670·102
99/2 5 30·200·654
99/3 7 63·570·059
99/4 3 33·000·083
Total21178·050·005

Levels of invertebrates fluctuated in a complex manner over time of day, between days and between locations (Fig. 6). Overall, peaks of invertebrate drift were observed in samples spanning dusk. Rank-correlation revealed low consistency of ranks for any particular sampling location aggregated by either day (Table 5) or sampling period (Table 6).

Figure 6.

Patterns of invertebrate biomass from drift sampling. Different hatch patterns represent drift nets in order of position in stream (upstream to downstream from top of graph).

Table 5.  Kendall’s coefficient of concordance and associated Friedman’s χ2 for abundance and biomass of invertebrates collected in drift samples aggregated by day. Critical value for (χ2)0·01, 4, 4 = 9·6. Significant values shown in bold
DateAquatic invertebratesTerrestrial invertebrates
AbundanceBiomassAbundanceBiomass
Wχ2Wχ2Wχ2Wχ2
7–8/07/990·606 7·30·3754·50·90010·80·725 8·7
8–9/07/990·675 8·10·6507·80·543 6·50·519 6·2
7–8/08/990·85010·20·4755·70·775 9·30·531 6·4
8–9/08/990·375 4·50·4255·10·825 9·90·92511·1
Table 6.  Kendall’s coefficient of concordance and associated Friedman’s χ2 for abundance and biomass of invertebrates collected in drift samples aggregated by time of day. Critical value for (χ2)0·01, 4, 4 = 9·6. Significant values shown in bold
PeriodAquatic invertebratesTerrestrial invertebrates
AbundanceBiomassAbundanceBiomass
Wχ2Wχ2Wχ2Wχ2
04.00–10.000·95611·50·7759·30·675 8·10·5136·2
10.00–16.000·675 8·10·5756·90·92511·10·4755·7
16.00–22.000·625 7·50·3253·90·556 6·70·4755·7
22.00–04.000·575 6·90·7008·40·325 3·90·2002·4

Discussion

This study provides the first detailed analysis of interactions between growth, dominance, sex and local movements of Atlantic salmon in the type of habitat they typically occupy in nature. There was no correlation between growth and dominance. Instead, growth depended on the degree of movement between areas of stream. There were differences in growth between the sexes, which corresponded to a prepubertal growth spurt in males. The study confirmed previous observations of extensive space-sharing among stream-dwelling salmon and revealed a high level of temporal unpredictability in the relative spatial distribution of drifting invertebrates.

Growth decreased as runs progressed throughout the summer, probably reflecting the fixed number but increased biomass of fish and an expected reduction in invertebrate food and hence carrying capacity of the enclosure (Steingrímsson & Grant 1999). This scenario is probably a reasonable approximation of natural systems in which space becomes available in spring when smolts leave rivers, but may decrease along with food as habitats dry up as the summer progresses. Individual fish within groups lost weight in July and August, further indicating the high biomass relative to carrying capacity at these times. There were no relationships between growth and rank either in early summer when biomass was low relative to carrying capacity or in late summer when it was high. Further experiments could usefully examine variations in density relative to carrying capacities within seasons.

The absence of either positive or negative correlations between dominance and growth showed that the costs and benefits of social dominance were balanced for juvenile Atlantic salmon in a near-natural stream. Grant (1993) predicted that the net outcome of dominance behaviour (i.e. the attempted monopolization of resources) is highly dependent on the spatio-temporal distribution of resources. Our data on invertebrate abundance suggest that the profitability of patches varied substantially both spatially and temporally with respect to each other. Thus, dominant fish were not able to exploit spatially or temporally predictable food resources to the extent that they can do in stream pools, in which food is most abundant at the head of the pool, and artificial streams with point food sources (Fausch 1984; Huntingford et al. 1990; Nakano 1995). Salmon parr may be largely excluded from pool habitats by brown trout in natural streams (Kennedy & Strange 1986).

Unpredictability of food supply might reduce the differential in feeding rates between salmon because more time would be spent by dominant fish in suboptimal patches. However, this on its own would not eliminate a correlation between growth and dominance because, on balance, dominant fish would still be expected to have preferential access to good patches, when they find them, and hence to have a higher rate of resource gain (Ruxton, Armstrong & Humphries 1999). However, these gains would be less than in a predictable habitat and in our experiment appeared simply to have been countered by the metabolic costs of dominance (Roskaft et al. 1986; Metcalfe et al. 1995).

We suggest that the fitness (measured in terms of growth) of any given dominance strategy is determined by the variability of the environment. At one extreme, in a perfectly predictable habitat, dominant fish will have highest feeding rates and will grow most quickly (Fausch 1984; Hughes 1992; Nakano 1995). At the other extreme, in a completely unpredictable habitat, or with pulses of high food availability (Grand & Grant 1994; Huntingford & Garcia de Leaniz 1997), which prevent food from being defended effectively by dominant individuals, intake rate will be similar across ranks and subordinate fish will grow most quickly because of their low energy expenditure. What is particularly fascinating is that in our natural habitat the balance of costs and benefits appeared to be similar across ranks.

Growth rate correlated with rate of movement. This observation is consistent with the results of Grand & Grant (1994) who found, in the laboratory, that when food delivery was unpredictable, success was dependent on the rate of patch switching (equivalent to movement records in our data set). This correlation between movement and growth raises the question of why some fish adopted a relatively sessile strategy. It seems likely that mobile fish are relatively susceptible to predation because they lose their crypticity (Gotceitas & Godin 1991; Ryer & Olla 1996). Thus, increased growth associated with exploration may be traded against reduced survival probability (Stearns 1992). Predation risk has been demonstrated to change foraging tactics across many taxa but may interact in a complex manner with dominance and the type of predator. For example, dominant great tits fed for shorter periods than subordinates when exposed to two predators but the reverse was true when exposed to only a single predator (Krams 2000).

Nearly all of the male parr in our experiment were maturing in preparation for spawning later in the year when greater size would lead to increased spawning success (Thomaz et al. 1997; Koseki & Maekawa 2000). By contrast, none of the female parr showed signs of maturity. The significant difference in growth rates between males and females was probably a consequence of these different maturation constraints and may reflect trade-offs between reproductive investment, growth and mortality (Stearns 1992; Hutchings & Jones 1998). There was a prepubertal growth spurt in maturing male salmon parr in June and July supporting the results of Tucker & Rasmussen (1999) who found greater growth of mature male parr compared with immature male fish. However, there was no difference between growth of immature female and mature male parr in August. Arndt (2000) recorded low growth of males relative to immature fish in the final stages of gonad maturation (September), suggesting that there is a switch from high to low growth as maturation proceeds. This latter low growth phase may relate to increased maintenance costs associated with the increase in the mass of highly metabolically active tissues, such as heart ventricle, that occurs prior to spawning (Armstrong & West 1994).

The present study has important implications for the role of social dominance, which occurs across a wide range of invertebrate and vertebrate taxa (Huntingford & Turner 1987). The consequences of dominance may depend on interactions between social behaviour, resource distribution, predation and life-history strategies. Under predictable resource distribution with low predation risk, dominance may be strongly advantageous, while it may be disadvantageous when conditions are unpredictable. Manipulation studies in natural or near-natural habitats will be required to calculate the costs and benefits of dominance under various conditions.

Acknowledgements

We thank Christopher West, David Stewart, Neill Herbert, Helen Grant and Albert Davison for help with fieldwork, Janey Keay for processing of invertebrate samples and Darren Evans for video analyses. We also thank Neil Metcalfe, Felicity Huntingford, Dick Shelton and two anonymous referees for constructive comments and criticisms. Procedures on animals were carried out under Home Office Project Licence 60/2317. This work has been carried out with financial support from the Commission of the European Communities, Agriculture and Fisheries (FAIR) specific RTD programme, CT-97–3498, ‘Performance and Ecological Impacts of Introduced and Escaped Fish: Physiological and Behavioural Mechanisms’. It does not necessarily reflect its views and in no way anticipates the Commission’s future policy in this area.

Ancillary

Advertisement