J. Zool., Lond. (2000) 251, 377±384 # 2000 The Zoological Society of London Printed in the United Kingdom

Differential microdistributions and interspeci®c interactions in coexisting native and introduced Gammarus spp. (Crustacea: Amphipoda) Calum MacNeil* and John Prenter School of Biology and Biochemistry, The Queen's University of Belfast, Belfast BT9 7BL, Northern Ireland (Accepted 18 August 1999)

Abstract In Northern Ireland, the native Gammarus duebeni celticus and introduced Gammarus tigrinus occur in the same river and lake systems. This study examined the outcome of encounters between the two amphipod species in the same patch of lake/pooled area of river. A laboratory simulation of a lake/river habitat indicated that G. d. celticus and G. tigrinus differed in distribution within the same complex habitat. Cannibalism was low for both species in single and mixed species treatments. However, G. tigrinus suffered heavy intraguild predation (IGP) from G. d. celticus in mixed species treatments, with 22% of the laboratory population of G. tigrinus being eliminated within 4 days. There was negligible reciprocal predation of G. d. celticus by G. tigrinus. It is proposed that IGP may account for G. tigrinus occurring more frequently and in greater abundance in patches of lotic and lentic systems where predatory G. d. celticus is absent or scarce. Key words: Gammarus, microdistribution, intraguild predation, cannibalism

INTRODUCTION Use of habitat in amphipods as in all animals, is in¯uenced by a complex series of interacting abiotic and biotic factors (Dahl & Greenberg, 1996). Factors that in¯uence the distribution and abundance of amphipods such as Gammarus spp. can be generally categorized as: (1) those that determine a species' habitat use independent of other species, such as physiological tolerance (Savage, 1996); (2) biotic interactions such as competition and predation (Connell, 1975, Dick, Elwood & Montgomery, 1994; Dick, 1996; MacNeil, Elwood & Dick, 1999). These biotic interactions are themselves modi®ed by colonization and migrations (Jeffries & Mills, 1990). In fresh water in Northern Ireland, the native Gammarus duebeni celticus Stock & Pinkster and the introduced North American Gammarus tigrinus Sexton often occur together in the same locality (Dick, 1996; MacNeil, 1997; H. B. N. Hynes, pers. comm.). Niche overlap between these native and introduced species in rivers and lakes is unsurprising, because Gammarus spp. are not microhabitat specialists (MacNeil, Dick & Elwood, 1997). They exhibit phenotypic plasticity in growth and survival in relation to environmental parameters (Savage, 1996), extreme trophic versatility *All correspondence to current address: 24 The Wynd, Pelton, Chester-le-st, Co. Durham DH2 1EH, U.K.

(MacNeil, Dick et al., 1997), and many possess characteristics of fugitive species (Pinkster, Smit & Brandse-de Jong, 1977; Pinkster, Scheepmaker et al., 1992). Such traits allow the exploitation of a broad potential niche. However, these fundamental niches are modi®ed by biotic processes including competition and predation (MacNeil, Elwood et al., 1999). Although G. tigrinus is reported actively to avoid fast ¯owing lotic systems and wave stricken lake shorelines (Nijssen & Stock, 1966; Pinkster, Smit et al., 1977), this cannot account for its absence from the many sheltered, ponded bays in lakes such as Lough Neagh, Northern Ireland, where G. d. celticus is particularly abundant (Murphy & Carter, 1984; Dick, 1996; MacNeil, 1997). Intraguild predation (IGP), i.e. predation exerted by potential competitors belonging to the same ecological guild (Polis, Myers & Holt, 1989), is increasingly recognized as a powerful force underlying amphipod exclusions (Dick & Platvoet, 1996; Otto, 1998). Predation of G. tigrinus by G. d. celticus has been proposed as a mechanism to account for the spatially and temporally erratic distributions of G. tigrinus in several rivers and for the scarcity of G. tigrinus along the Lough Neagh shoreline (Dick, 1996; MacNeil, 1997). Indeed, despite G. tigrinus having a reproductive output 15 times greater than that of G. d. celticus (Hynes, 1954), it has been suggested that differential predation in favour of G. d. celticus (i.e. the superior ability of G. d. celticus both to resist G. tigrinus predation and to prey on

378

C. MacNeil and J. Prenter

moulted G. tigrinus) may over-ride this `advantage' when the two species come into contact, causing the introduced species to be eliminated (Dick, 1996). To understand how communities are structured, it is instructive to focus on differences in niche requirements of species living in the same habitat (Starr & Taggart, 1989). Therefore, in the present study we set up a laboratory simulation of a `typical' freshwater habitat, or rather micro-habitat, containing substrates and plants (see MacNeil, Elwood & Dick, in press). This represented a coarse template of the patches that could be found in areas of Lough Neagh and sluggish areas of the Lower Bann river, the latter being the system whereby G. tigrinus was originally introduced into N. Ireland (H. B. N. Hynes, pers. comm.). Distribution patterns obtained from ®eld surveys indicated that G. d. celticus and G. tigrinus, although co-occurring in certain habitats, were usually found in discrete patches of the same rivers and lakes (Murphy & Carter, 1984; Dick, 1996; MacNeil, 1997). In the present study, we examined the microdistribution of these two species with the same, albeit arti®cial habitat. Such a simulation could ascertain whether or not G. d. celticus and G. tigrinus are `ecological equivalents' at least in terms of microdistribution and habitat use when presented with the same habitat pro®les under the same physicochemical regime. A secondary aim of our study was to investigate and quantify the effects of cannibalism and IGP that may in¯uence the distribution and relative abundance of Gammarus spp. in northern Irish waters. Previous attempts have been made to set up `staged encounters' between G. d. celticus and G. tigrinus in the laboratory (Dick, 1996; Dick & Platvoet, 1996). However, the simulated habitats used, were necessarily highly simplistic with only rudimentary substrates, usually `housed' within a small area (plastic vessels of 5 cm diameter). The two species described, commonly inhabit the littoral zones of lakes, which are spatially patchy and often physically complex. Increased habitat complexity can have profound effects on amphipod species interactions (Dick, 1996; MacNeil, 1997). Increased complexity generates more microhabitat types, presenting an increased total niche space with more potential refuges (Crowder & Cooper, 1982) and this could allow the coexistence of even strong competitors and the persistence of predators and prey (MacNeil, Dick et al., 1997). Therefore, our study investigated whether previously reported incidences of cannibalism within and predation between, these two species would manifest themselves within a larger, more structurally complex habitat simulation than has been previously attempted. MATERIALS AND METHODS Collection of amphipods and set-up of habit simulations During late November 1998, G. d. celticus was collected by kick-sampling from Traad Point on the Lough

Neagh shoreline (U.K. grid ref. H957873). This involved systematically disturbing all major shoreline microhabitats by kicking the substrate and simultaneously sweeping a standard kick-net (1 mm mesh, frame aperture 230±250 mm) in front of and over the disturbed area. Gammarus tigrinus was collected from the midlough bottom by trawl netting (1 mm mesh-net) 1 km offshore from Traad point (only a few G. tigrinus being found along the shoreline). Collecting from singlespecies communities allowed us to mimic initial interspeci®c contact and invasion in the simulation (Dick, Montgomery & Elwood, 1993; Dick, 1996; MacNeil, Elwood et al., 1999). Specimens were maintained separately in large (36650615 cm deep), aerated tanks ®lled with lough water. They were also provided with substratum and habitat in the form of cobbles, pebbles, gravel and sand, together with ¯ora (i.e. Canadian pondweed, Elodea canadensis) and fauna from their sites of origin. Food in the form of leaves and pelleted cat®sh food (which both species have been previously observed to grab and devour; MacNeil, Elwood et al., 1999) was supplied as additional nutrition to these trophically versatile animals (MacNeil, Dick et al., 1997). In the absence of predators and cannibals, both species can be maintained for several months under these conditions (Dick, Montgomery et al., 1993; Dick, 1995). The light : dark cycle was kept constant at 9 : 15 h and the water temperature maintained at 6.0±7.0 8C, both regimes being appropriate for the time of year (see Dick, 1992, 1995, 1996; D. Jewson, pers. comm.). Microdistribution Circular tanks (10 replicates) measuring 22 cm in diameter and 25 cm deep and containing ®ltered Lough Neagh water, were covered on all sides with white plastic, leaving the tank surface clear. In pilot studies, individuals of each species survived for over 2 months under these tank regimes, indicating this water was physiologically suitable. Pilot studies also revealed that the presence of the observer had no effect on amphipod distribution (MacNeil, 1997; MacNeil, Elwood et al., 1999). Each tank was lit via overhead neon lights (light intensity 1.5 mE m-2 s-1, Glen Creston model LI-1858 quantum photometer) and the white background permitted clear observation of the amphipods. Each tank contained a washed Lough Neagh sand substrate, upon which were placed centrally 5 pieces of stone ®lter tube (each 1 cm long, 0.5 cm diameter hole). These tubes were placed upright in the sand, permitting access for amphipods into the stone substrate, while still maintaining the observer's view (Dick, 1996; Dick & Platvoet, 1996). The pondweed E. canadensis, was common in all of the sites where amphipods were collected. Therefore, 465-cm length strands of prewashed pondweed were weighted down and placed centrally in the sand next to the stones in each tank, to add a greater degree of realism to the simulation. Water conductivity was 263 mS cm-1 (Dist WP, Hanna Instru-

Microdistributions and interspeci®c interactions in Gammarus spp. ments) and air stones were used to maintain dissolved oxygen levels at 9.0 mg l-1 (Model 50B D.O. meter Y.S.I.), both these values being typical of levels found in Gammarus spp. collection sites (Dick & Platvoet, 1996; C. MacNeil, pers. obs.). Tanks were allowed to stand empty for 5 days before amphipods were introduced, to allow the water to become clear and suf®ciently mixed and aerated. We selected only healthy (free from the parasitic cystacanth Polymorphous sp.), non-gravid, adult G. d. celticus (c. 12.0±16.5 mm body length) and G. tigrinus (9.0±15.0 mm) for the simulation. These size ranges re¯ected the adult size of each species in the Lough (Dick, 1996). Twelve amphipods (6 of each species) were randomly introduced into each of the replicate tanks. Our observations took place 2 min, 10 min, 30 min, 1 h, 2 h, 4 h and 7 h after the initial introduction of amphipods (all within the light cycle). At each of these time intervals, the location of each separate species (6 individuals) were recorded for 6 habitat type categories; (1) sand; (2) weed; (3) stone surface; (4) under/in stone; (5) in the water column; (6) on tank walls. This range of monitoring times allowed shifts in microdistribution and habitat use by each species over both short and long timespans to be assessed. Because the number of amphipods declined differentially between subsequent observation times (see below), the relative proportions of surviving animals in each species were recalculated for each time interval, so that the microdistribution pattern could still be ascertained for the tank population for all 7 monitoring times, even in the event of mortalities. Concurrent with the mixed species microdistribution study, the distributions of animals in single species controls (12 individuals, 8 replicates for each species) were compared to their distributions in the mixed species treatment after 7 h. This allowed us to assess whether the presence of 1 species in¯uenced the distribution of the other species in the complex habitat. No `food' (discounting weed) was provided in the tanks during the microdistribution study, as location of food resources has a major in¯uence on Gammarus spp. habitat selection in freshwater ecosystems (Shannon, Blinn & Stevens, 1994). If pelleted food/leaf fragments had been added to tanks, the possibility that amphipods may have simply congregated around the food could not have been discounted. By keeping amphipods fed to excess for 3 weeks before this experiment began and limiting the observation period to 7 h, we assumed that any differences in habitat usage between the 2 species would be not be a result of hunger. All species data were arcsine transformed before statistical analysis, this method being appropriate for percentage and proportional data sets (Sokal & Rohlf, 1995). We examined the effects of time, habitat and species on amphipod microdistribution in a 3-factor repeated measures ANOVA (all factors as repeated measures). The effects of habitat and simulation type (mixed or single species) on individual amphipod species microdistribution after 7 h were examined in a 2-factor

379

repeated measures ANOVA (all factors as repeated measures). ANOVAs were performed using the SUPERANOVA statistical package (Abacus Concepts, 1989). Cannibalism and intraguild predation At each time interval during the 7-h observation period of the microdistribution study, we recorded the number and species of dead animals and any acts of cannibalism and/or predation. Amphipod mortalities, where there appeared no identi®able immediate aggressor or when deaths occurred unseen between monitoring periods, were categorized as unknown. Assuming these `unknown' deaths occurred by the same methods and in the same proportions (i.e. cannibalism vs predation) as observed amphipod deaths, we fairly sure of at least the relative strengths of predation and cannibalism simultaneously occurring within the same species. At the end of the 7 h period of the microdistribution study, food in excess (5 cat®sh food pellets and well processed leaf material) was added to each tank, ensuring that animals had a source of food other than each other. At the end of 4 days, we removed the weed/stones and identi®ed and counted survivors. This allowed an accurate assessment of mortalities owing to cannibalism/predation after a more extended time. Single species controls within the same complex habitat conditions were carried out concurrent with the mixed species treatment. These controls involved 12 individuals of each species (8 replicates), thereby presenting the same density as the mixed species simulations. Such controls allowed assessment of mortalities in monocultures, when each species was freed from competition and/or victimization by the other. The mean percentage mortality of each species at each time interval during the 7-h observation period was calculated as previously. Differences in the actual mean percentage mortality of each amphipod species (i.e. total percentage number of each species dead or missing) after 4 days in both mixed and single species treatments, were analysed by 2 separate 1-way ANOVAs (`species' and `treatment type') (once again percentage data were arcsine transformed before statistical analysis). RESULTS Microdistribution Although there was no signi®cant time effect or species effect on the relative abundance of amphipods, there was a highly signi®cant habitat effect (F5, 45 = 92.103, P < 0.0001), indicating that there was a signi®cantly higher percentage of the total amphipod assemblage in each tank congregating on speci®c habitat types. There was no signi®cant interaction between time and species on relative amphipod abundance. There was, however, a highly signi®cant time and habitat interac-

380

C. MacNeil and J. Prenter 2 min 10 min 40

30 min 1 hr 2 hr 4 hr

30

Mean %

7 hr

20

10

0 Sand

Weed

On stone Habitat

In stone

Water

Tank wall

Fig. 1. Mean percentage number (‹ s e) of all amphipods located on or in designated habitat types, over a 7 h period in mixed species complex habitat simulations (data arcsine transformed before statistical analysis).

50

G. d. celticus G. tigrinus

40 Mean %

tion effect (F30, 270 = 2.813, P < 0.0001; Fig. 1), indicating there were signi®cant shifts in the percentage of total amphipods from one habitat to another as the simulation progressed. There was a general decline in the percentage of total amphipods on the sand and on or in the stone tubes accompanied by a general increase on the weed, as the experiment progressed. The most profound differences in amphipod microdistribution were limited to the ®nal 7 h observation period. At this time, the percentages of amphipods located on sand, stone and in the water were lower and the percentage in the weed higher, than at all previous observation times. There was a highly signi®cant species and habitat interaction effect (F5, 45 = 76.342, P < 0.0001; Fig. 2). Least square means tests showed there was a signi®cantly higher percentage of the G. d. celticus located on the sand than on or in all of the other habitat types (higher than in the weed, t = 12.314; on the stone, t = 19.697; in the stone tube, t = 15.687; in the water, t = 16.769; or on the tank wall, t = 16.769; all at P < 0.0001). In contrast, there was a signi®cantly higher proportion of G. tigrinus in the weed than observed in/on any other habitat type except the tank walls (higher than on the sand, t = 9.270; on stone, t = 14.489; in stone, t = 13.470; water, t = 7.997; all at P < 0.0001). Indeed, the overall microdistribution pattern of G. tigrinus differed radically from that observed for G. d. celticus. There was a signi®cant difference in the percentage of G. d. celticus relative to G. tigrinus in ®ve of the six designated habitat types. There was a signi®cantly higher percentage of G. d. celticus located on the sand and in the stone than G. tigrinus (t = 15.051 and t = 3.564, respectively; both at P < 0.0001). In contrast, there was a signi®cantly lower percentage of

30 20 10 0 Sand

Weed

On stone In stone Habitat

Water

Wall

Fig. 2. Mean percentage number (‹ s e) of each amphipod species located on or in designated habitat types in mixed species complex habitat simulations (data arcsine transformed before statistical analysis).

G. d. celticus in the weed, in the water column, and on the tank walls than G. tigrinus (t = 6.534, P < 0.0001; t = 2.991, P < 0.01; t = 9.514, P < 0.0001, respectively). The surface of the stone tubes was the least used habitat by both species and was the only habitat category where the percentages of the two species did not differ signi®cantly. Therefore, the general pattern of amphipod microdistribution was one of G. d. celticus congregating on sand and in habitats at the bottom of the tanks, whereas G. tigrinus appeared to avoid the bottom, keeping to the pondweed, water column and tank walls. The lack of a signi®cant overall time, species and habitat interaction effect, indicated that there were no or only limited shifts in microdistribution patterns and speci®c species-habitat linkages between observation times.

Microdistributions and interspeci®c interactions in Gammarus spp. Mixed species treatment G. d. celticus

20

381

Single species treatment

G. d. celticus

G. tigrinus

G. tigrinus

% mean

15

10

5

0 0

1

2

3

4

5

6

7

Time (h)

Fig. 3. Mean percentage cumulative mortality (‹ s e) of different amphipod species in mixed and single complex habitat simulations (density of 12 amphipods per tank).

There were no signi®cant differences between the overall microdistribution of each of the two species between the mixed and single species treatment, indicating the presence of one species did not in¯uence the way the other species distributed itself. Cannibalism and intraguild predation (IGP) The mean (‹ se) percentage cumulative mortality (observed deaths) recorded at the end of the mixed species treatment after 7 h, was 11.67% (n = 10, se = 3.41) for G. tigrinus compared to only 3.33% (n = 10, se = 0.90) for G. d. celticus (Fig. 3). The rate of mortality of G. tigrinus was most rapid in the initial h of the simulation. The rate of mortality of G. d. celticus remained low throughout, with the ®rst observed deaths only occurring after the ®rst h of the simulation. In the single species controls (8 replicates of each type) mean cumulative percentage mortality of G. tigrinus was only 2.08% (n = 8, se = 0.91) at the end of 7 h and the mean percentage cumulative mortality of G. d. celticus was also low at only 6.25% (n = 8, se = 2.00) at the ®nal 7 h observation. In mixed species treatments, during the 7 h period, considering only observed amphipod deaths (i.e. where an amphipod was actually observed being killed and eaten), all such G. tigrinus deaths (5.00% of total tank population) were attributable to IGP by G. d. celticus and all G. d. celticus deaths (1.67% of tank population) were the result of cannibalism. After 4 days, mean mortalities of G. d. celticus in different treatment types were not signi®cantly different and ranged between 6.66% (n = 10, se = 2.32) and 8.33% (n = 8, s e = 2.78) in mixed and single species treatments, respectively. However, there was a highly signi®cant

difference among G. tigrinus mortalities in different treatment types (F1, 16 = 12.572, P < 0.01). Mean mortality of G. tigrinus was 21.66% (n = 10, s e = 4.32) recorded in the presence of G. d. celticus, compared to only 3.13% (n = 8, s e = 3.02) in the single species treatment. It is interesting to note that cannibalism and predation in all mixed and single species treatments continued after the 7 h observation period, despite the presence of excess food for the remainder of the 4 days. DISCUSSION Our study examines what may happen when two amphipod species, introduced and native, co-occur (even brie¯y) within the same patch of lake or river. Against a complex habitat template, numerous interactions were evident. Little could be inferred from total amphipod± microhabitat linkages, given that habitats were of different sizes and structural complexity and thus had inherently different capacities to contain amphipods. However, the observed differences and similarities in the overall microdistribution patterns of the two species proved enlightening. Gammarus d. celticus occurred mainly on or in the sand, in the weed or in the stone tubes. In contrast, G. tigrinus occurred mainly in the weed or crawling high up on the tank walls and swimming free in the water column, as opposed to the more `con®ned' habitats such as in/under the stone. For both Gammarus spp., species±habitat linkages observed in aquaria did not signi®cantly change over the observation period. Conservatism in habitat ®delity has been observed in other amphipods, for example Crangonyx (MacNeil, Elwood et al., 1999) and other crustaceans such as cray®sh (Stein & Magnuson, 1976).

382

C. MacNeil and J. Prenter

Rabeni & Sowa (1996) suggested that species : habitat linkages should be discussed cautiously, as biologists often assume a habitat element `used' disproportionately to its `availability' is either `preferred' or `avoided', when there is little evidence that habitat use equates with importance (MacNeil, Elwood et al., 1999). Despite this, the signi®cant disparity in microhabitat use between G. tigrinus and G. d. celticus would have allowed G. tigrinus to avoid encounters with G. d. celticus in many instances. The pondweed E. canadensis was the second most commonly used microhabitat type by both amphipod species. Van Dolah (1978) and Shannon et al., (1984) both reported a preference by Gammarus spp. for vegetation covered substrates as opposed to bare sand and gravel substrate types. Gregg & Rose (1982) suggested that the strong associations of Gammarus spp. with macrophytes was linked to strong current avoidance behaviour, as the physical heterogeneity of plant structures provided many living spaces with low current velocity (see Gee, 1982; Adams et al., 1987). However, in the present `static water' simulation, the relative abundance of the amphipods in the weed may be related to it being the only possible food source in the tanks (apart of from each other) for 7 h. While this confounding effect could have been overcome by using an arti®cial pondweed, this may have reduced the realism of this particular habitat. This highlights the dif®culty in attempting to isolate the effects of different factors on habitat use and in constructing a realistic habitat scenario without food in¯uencing animal distribution (Elton, 1927; MacNeil, 1997). Cannibalism occurred in both mixed and single species treatments. In G. d. celticus, cannibalism was present in both single species and mixed treatments, and in G. tigrinus, it took place in single-species treatments. Indeed, Dick (1995) noted that for several Gammarus spp., cannibalism is common even when space, refuges and food are all available in excess. Cannibalism can predispose individuals of one species towards the killing of other species, especially congenerics and members of the same ecological guild (Polis et al., 1989; Dick, 1995, 1996; Dick, Elwood & Montgomery, 1995; MacNeil, Dick et al., 1997). Some Gammarus species are also considered to be inherently more predatory than others (Gledhill, Sutcliffe & Williams, 1993; Dick, Elwood et al., 1995). The greater proportional losses of G. tigrinus in the mixed species treatments was clearly attributable to high rates of predation by G. d. celticus. There was no reciprocal predation observed by G. tigrinus on G. d. celticus (see also Dick, 1996). This may have been related to differences in the frequency of amphipod moult between the two species. In mixed simulations, while only one (1.67% of the total tank population) moulted G. d. celticus individual was noted, three (5.00%) G. tigrinus moults were noted during the 4 day study. These estimates were likely to be highly conservative given the high rates of predation found in tanks, as many newly moulted individuals and their moults were probably quickly devoured.

Although the native G. d. celticus and introduced central European G. pulex are physiologically adapted to fresh water (Sutcliffe, 1968; Dick, 1996), G. tigrinus occurs in oligohaline and brackish water in its native North America (Bous®eld, 1973). In Ireland and the Netherlands G. tigrinus is excluded from fresh water by G. pulex, but the reverse is evident in oligohaline waters (Pinkster, Scheepmaker et al., 1992; Dick, 1996). Dick & Platvoet (1996) proposed that this was the result of differences in moult frequency of the two species under different ionic regimes. Laboratory experiments revealed that, although mutual predation of moulting individuals occurred frequently between the two species, predation frequencies were differentially in favour of G. pulex in freshwater but not in oligohaline water to which G. tigrinus is physiologically adapted. Although it has not been investigated whether a similar number of moults is required during the life cycle of each species, an additional factor underlying differential mutual predation may be that the duration of the soft moulted state (i.e. before exoskeleton hardens) and therefore the period spent vulnerable to predation for each species, is itself strongly in¯uenced by ionic conditions (J. T. A. Dick, pers. comm.). In Lough Neagh, the introduced G. tigrinus may be at a similar physiological disadvantage as regards moult frequency and relative duration of the soft moulted state, in comparison to the native G. d. celticus. In other river and lake systems, the reverse may be true. For instance, Savage (1996), during a 27 year study of the population dynamics of macroinvertebrates in Watch Lane Flash (a small English lake), found that resident G. duebeni numbers signi®cantly decreased once introduced G. tigrinus were present in large numbers. He attributed this decline to direct predation of G. duebeni by the introduced species. The observed high mortalities and heavy losses owing to predation were suffered by G. tigrinus against a more complex habitat template than previously attempted in other amphipod IGP studies (i.e. Dick, 1996; Dick & Platvoet, 1996; Otto, 1998). Increased habitat structural complexity lowers predator ef®ciency, as the initial detection, pursuit and ®nal capture of prey is reduced (Crowder & Cooper, 1982). However, it is the degree of habitat complexity which may be important in determining whether the predator±prey relationship stabilizes, or if the prey is driven to extinction. There is a tendency for macroinvertebrate predators to have far stronger impacts on prey communities than many larger vertebrate predators such as ®sh (Walde & Davies, 1984; Colinvaux, 1986; Sih & Wooster 1994). Although the `refuges' that the tank habitat provided (i.e. in or under the stone tubes) may have been effective against large predators such as trout (MacNeil, 1997), such refuges were ineffective against predators that were a similar size to the prey. Where G. tigrinus could go, so could G. d. celticus (see Dick & Elwood, 1993). The designation of macroinvertebrates into functional feeding groups, based upon what they eat and how they eat it, is now a common practice in studies examining trophic interactions and community structures in fresh

Microdistributions and interspeci®c interactions in Gammarus spp. water (Cummins, 1974; Vannote et al., 1980). Freshwater amphipods are still considered principally as herbivorous `shredders' of coarse particulate organic matter such as leaf litter. Reported incidences of IGP and cannibalism are still commonly regarded as aberrations (MacNeil et al. 1997). Such assumptions need to be questioned for both amphipods and other macroinvertebrates (Dick, 1996; MacNeil, Dick et al., 1997; Otto, 1998). For instance, the larvae of limnephilid caddis ¯ies, which are also considered to be herbivorous `shredder-detritivores', can like amphipods, be both highly predatory and cannibalistic (Giller & Sangpradub, 1993). Indeed, Wissinger et al. (1996) showed that asymmetric IGP and cannibalism strongly in¯uenced the distribution and relative abundances of two caddis¯ies Limnephilus externus (Hagen) and Asynarchus nigriculus Banks. Both these species were previously assumed to subsist solely on leaf litter. Similarly, the present investigation shows that one amphipod shredder, G. tigrinus, may be locally excluded by ®erce interspeci®c competition and/or predation exerted by another putative shredder, G. d. celticus. This study provides additional evidence for the plasticity of Gammarus feeding mode and the importance of IGP in structuring amphipod communities and in¯uencing the lotic and lentic distributions of native and introduced species. Acknowledgements The assistance of Hugo McGrogan, Jaqui Hughes, Matt Quinn, Ewan Bigsby, Chris Harrod, Niall Clements, ZoÈe Ruiz, Dave Jewson, Dave Grif®ths, Clare Carter and everybody in the University of Ulster Freshwater Laboratory is greatly appreciated. Thanks to Jaimie Dick and Bob Elwood and everyone at the Queen's University of Belfast for pushing C. MacNeil in this direction. Thanks also to Professor H. B. N. Hynes for invaluable insights into the fauna of Lough Neagh. The additional help and advice of Mike Jeffries of the University of Northumbria is gratefully acknowledged. REFERENCES Abacus Concepts (1989). SuperANOVA. Berkley, CA: Abacus Concepts. Adams, J., Gee, J., Greenwood, P., Mckelvey, S. & Perry, R. (1987). Factors affecting the microdistribution of Gammarus pulex (Amphipoda); an experimental study. Freshwater Biol. 17: 307±316. Bous®eld, E. L. (1973). Shallow-water gammaridean Amphipoda of New England. Ithaca: Cornell University Press. Colinvaux, P. (1986). Ecology. Chichester: Wiley Connell, J. H. (1975). Some mechanisms producing structure in natural communities: a model and evidence from ®eld experiments. In Ecology and evolution of communities: 460±490. Cody, M. & Diamond, J. (Eds). Massachusetts: Harvard University Press. Crowder, L. B. & Cooper, W. E. (1982). Habitat structural complexity and the interaction between bluegills and their prey. Ecology 63: 1802±1813.

383

Cummins, K. W. (1974). Structure and function of stream ecosystems. Bioscience 24: 631±641. Dahl, J. & Greenberg, L. (1996). Effects of habitat structure on habitat use by Gammarus pulex in arti®cial streams. Freshwater Biol. 36: 487±495. Dick, J. T. A. (1992). The nature and implications of differential predation between Gammarus pulex and G. duebeni celticus. J. Zool. (Lond.) 227: 171±183. Dick, J. T. A. (1995). The cannibalistic behaviour of two Gammarus species (Crustacea : Amphipoda). J. Zool. (Lond.) 236: 697±706. Dick, J. T. A. (1996). Post-invasion amphipod communities of Lough Neagh, N. Ireland: in¯uences of habitat selection and mutual predation. J. Anim. Ecol. 65: 756±767. Dick, J. T. A. & Elwood, R. W. (1993). The mating system of Gammarus pulex : a negligible role for micro-habitat segregation. Anim Behav. 45: 188±190. Dick, J. T. A., Elwood, R. W. & Montgomery, W. I. (1994). Range expansion of the alien amphipod Gammarus pulex in the River Lagan, Co. Down. Irish Nat. J. 24: 403±404. Dick, J. T. A., Elwood, R. W. & Montgomery, W. I. (1995). The behavioural basis of a species replacement: differential aggression and predation between the introduced Gammarus pulex and G. duebeni celticus (Amphipoda). Behav. Ecol. Sociobiol. 37: 393±398. Dick, J. T. A., Montgomery, I. & Elwood, R. W. (1993). Replacement of the indigenous amphipod Gammarus duebeni celticus by the introduced G. pulex: differential cannibalism and mutual predation. J. Anim. Ecol. 62: 79±88. Dick, J. T. A. & Platvoet, D. (1996). Intraguild predation and species exclusions in amphipods; the interaction of behaviour, physiology and environment. Freshwater Biol. 36: 375±383. Elton, C. S. (1927). Animal ecology. New York: Macmillan. Gee, J. H. R. (1982). Resource utilization by Gammarus pulex (Amphipoda) in a Cotswold stream: a microdistribution study. J. Anim. Ecol. 51: 817±832. Giller, P. S. & Sangpradub, N. (1993). Predatory foraging behaviour and activity patterns of larvae of two species of limnephilid cased caddis. Oikos 67: 351±357. Gledhill, T., Sutcliffe, D. W. & Williams, W. D. (1993). British freshwater Crustacea Malacostraca: a key with ecological notes. Freshwater Biological Association. Kendal: Titus Wilson. Gregg, W. W. & Rose, F. L. (1982). The effects of aquatic macrophytes on the stream microenvironment. Aquat. Bot. 14: 309±324. Hynes, H. B. N. (1954). The ecology of Gammarus duebeni Lilljeborg and its occurrence in freshwater in western Britain. J. Anim. Ecol. 23: 38±84. Jeffries, M. & D. Mills. (1990). Freshwater ecology: principles and applications. London: Belhaven Press. MacNeil, C. (1997). The ecology of freshwater amphipods; a study of invasive and native species. Unpublished PhD thesis, The Queen's University of Belfast, Belfast. MacNeil, C., Dick, J. T. A. & Elwood, R. W. (1997). The trophic ecology of freshwater Gammarus (Crustacea: Amphipoda); problems and perspectives concerning the functional feeding group concept. Biol. Rev. 72: 349±364. MacNeil, C., Elwood, R. W. & Dick, J. T. A. (1999). Differential microdistributions and interspeci®c interactions in coexisting Gammarus and Crangonyx amphipods. Ecography 22: 415±423. Murphy, P. M. & Carter, C. E. (1984). A summer survey of the littoral macroinvertebrate fauna (excluding the Chironomidae) of Lough Neagh, N. Ireland. Proc. R. Ir. Acad. Sect. B Biol. Geol. Chem. Sci. 84: 103±108. Nijssen, H. & Stock, J. H. (1966). The amphipod Gammarus tigrinus Sexton, 1939, introduced in the Netherlands (Crustacea). Beaufortia 13: 197±206. Otto, C. (1998). Factors affecting the disjunct distribution of amphipods along a North Swedish river. Oikos 83: 21±28.

384

C. MacNeil and J. Prenter

Pinkster, S., Scheepmaker, M., Platvoet, D. & Broodbaker, N. (1992). Drastic changes in the amphipod fauna (Crustacea) of Dutch inland waters during the last 25 years. Bijdr. Dierkd. 61: 193±204. Pinkster, S., Smit, H. & Brandse-de Jong, N. (1977). The introduction of the alien amphipod Gammarus tigrinus Sexton, 1939, in the Netherlands and its competition with indigenous species. Crustaceana (Suppl. 1) 4: 91±105. Polis, G. A., Myers, C. A. & Holt, R. D. (1989). The ecology and evolution of intraguild predation : potential competitors that eat each other. Annu. Rev. Ecol. Syst. 20: 297±330. Rabeni, C. F. & Sowa, S. P. (1996). Integrating biological realism into habitat restoration and conservation strategies for small streams. Can. J. Fish. Aquat. Sci. 53 (Suppl. 1): 252±259. Savage, A. A. (1996). Density dependent and density independent relationships during a twenty-seven year study of the population dynamics of the benthic macroinvertebrate community of a chemically unstable lake. Hydrologia 335: 115±131. Shannon, J. P., Blinn, D. W. & Stevens, L. E. (1994). Trophic interactions and benthic animal community structure in the Colorado River, Arizona, U.S.A. Freshwater Biol. 31: 213±220.

Sih, A. & Wooster, D. E. (1994). Prey behaviour, prey dispersal, and predator impacts on stream prey. Ecology 75: 1199±1207. Sokal, R. R. & Rohlf, F. (1995). Biometry. New York: Freeman. Starr, C. & Taggart, R. (1989). Biology. The unity and diversity of life. Belmont, CA: Wadsworth. Stein, R. A. & Magnuson, J. J. (1976). Behavioral response of cray®sh to a ®sh predator. Ecology 57: 751±761. Sutcliffe, D. W. (1968). Sodium regulation and adaptation to fresh water in gammarid crustaceans. J. exp. Biol. 48: 359±380. Van Dolah, R. F. (1978). Factors regulating the distribution and population dynamics of the amphipod Gammarus palustris in an intertidal salt marsh community. Ecol. Monogr. 48: 191±217. Vannote, R. L., Minshall, G. W., Cummins, K. W., Sedell, J. R. & Cushing, C. E. (1980). The river continuum concept. Can. J. Fish. Aquat. Sci. 37: 130±137. Walde, S. J. & Davies, R. W. (1984). Invertebrate predation and lotic prey communities: evaluation of in situ enclosure/exclosure experiments. Ecology 65: 1206±1213. Wissinger, S. A., Sparks, G. B., Rouse, G. L., Brown, W. S. & Steltzer, H. (1996). Intraguild predation and cannibalism among larvae of detritivorous caddis¯ies in subalpine wetlands. Ecology 77: 2421±2430.