Escaping peril: perceived predation risk affects migratory propensity

Although migratory plasticity is increasingly documented, the ecological drivers of plasticity are not well understood. Predation risk can influence migratory dynamics, but whether seasonal migrants can adjust their migratory behaviour according to perceived risk is unknown. We used electronic tags to record the migration of individual roach (Rutilus rutilus), a partially migratory fish, in the wild following exposure to manipulation of direct (predator presence/absence) and indirect (high/low roach density) perceived predation risk in experimental mesocosms. Following exposure, we released fish in their lake summer habitat and monitored individual migration to connected streams over an entire season. Individuals exposed to increased perceived direct predation risk (i.e. a live predator) showed a higher migratory propensity but no change in migratory timing, while indirect risk (i.e. roach density) affected timing but not propensity showing that elevated risk carried over to alter migratory behaviour in the wild. Our key finding demonstrates predator-driven migratory plasticity, highlighting the powerful role of predation risk for migratory decision-making and dynamics.     

Heterogeneous landscapes and the role of refuge on the population dynamics of a specialist predator and its prey

How, and where, a prey species survives predation by a specialist predator during low phases of population fluctuations or a cycle, and how the increase phase of prey population is initiated, are much-debated questions in population and theoretical ecology. The persistence of the prey species could be due mainly to habitats that act as refuges from predation and/or due to anti-predatory behaviour of individuals. We present models for the former conjecture in two (and three) habitat systems with a specialist predator and its favoured prey. The model is based on dispersal of prey between habitats with high reproductive output but high risk of predation, and less productive habitats with relatively low risk of predation. We illustrate the predictions of our model using parameters from one of the most intriguing vertebrate predator–prey systems, the multi-annual population cycles of boreal voles and their predators. We suggest that cyclic population dynamics could result from a sequence of extinction and re–colonization events. Field voles (Microtus agrestis), a key vole species in the system, can be hunted to extinction in their preferred meadow habitat, but persist in sub-optimal wet habitats where their main predator, the least weasel (Mustela nivalis nivalis) has a low hunting efficiency. Re–colonization of favourable habitats would occur after the predator population crashes. At the local scale, the model suggests that the periodicity and amplitude of population cycles can be strongly influenced by the relative availability of risky and safe habitats for the prey. Furthermore, factors like intra-guild predation may lead to reduced predation pressure on field voles in sub-optimal habitats, which would act as a refuge for voles during the low phase of their population cycles. Elasticity analysis suggested that our model is quite robust to changes in most parameters but sensitive to changes in the population dynamics of field voles in the optimal grassland habitat, and to the maximum predation rate of weasels.     

Interspecific differences in antipredator strategies determine the strength of non‐consumptive predator effects on stream detritivores

Animal species differ considerably in their response to predation risks. Interspecific variability in prey behaviour and morphology can alter cascading effects of predators on ecosystem structure and functioning. We tested whether species‐specific morphological defenses may affect responses of leaf litter consuming invertebrate prey to sit‐and‐wait predators, the odonate Cordulegaster boltonii larvae, in aquatic food webs. Partly or completely blocking the predator mouthparts (mandibles and/or extensible labium), thus eliminating consumptive (i.e. lethal) predator effects, we created a gradient of predator‐prey interaction intensities (no predator < predator – no attack < predator – non‐lethal attacks < lethal predator). A field experiment was first used to assess both consumptive and non‐consumptive predator effects on leaf litter decomposition and prey abundances. Laboratory microcosms were then used to examine behavioural responses of armored and non‐armored prey to predation risk and their consequences on litter decomposition. Results show that armored and non‐armored prey responded to both acute (predator – non‐lethal attacks) and chronic (predator – no attack) predation risks. Acute predation risk had stronger effects on litter decomposition, prey feeding rate and prey habitat use than predator presence alone (chronic predation risk). Predator presence induced a reduction in feeding activity (i.e. resource consumption) of both prey types but a shift to predator‐free habitat patches in non‐armored detritivores only. Non‐consumptive predator effects on prey subsequently decreased litter decomposition rate. Species‐specific prey morphological defenses and behaviour should thus be considered when studying non‐consumptive predator effects on prey community structure and ecosystem functioning.     

Habitat-dependent foraging in a classic predator–prey system: a fable from snowshoe hares

Current research contrasting prey habitat use has documented, with virtual unanimity, habitat differences in predation risk. Relatively few studies have considered, either in theory or in practice, simultaneous patterns in prey density. Linear predator–prey models predict that prey habitat preferences should switch toward the safer habitat with increasing prey and predator densities. The density‐dependent preference can be revealed by regression of prey density in safe habitat versus that in the riskier one (the isodar). But at this scale, the predation risk can be revealed only with simultaneous estimates of the number of predators, or with their experimental removal. Theories of optimal foraging demonstrate that we can measure predation risk by giving‐up densities of resource in foraging patches. The foraging theory cannot yet predict the expected pattern as predator and prey populations covary. Both problems are solved by measuring isodars and giving‐up densities in the same predator–prey system. I applied the two approaches to the classic predator–prey dynamics of snowshoe hares in northwestern Ontario, Canada. Hares occupied regenerating cutovers and adjacent mature‐forest habitat equally, and in a manner consistent with density‐dependent habitat selection. Independent measures of predation risk based on experimental, as well as natural, giving‐up densities agreed generally with the equal preference between habitats revealed by the isodar. There was no apparent difference in predation risk between habitats despite obvious differences in physical structure. Complementary studies contrasting a pair of habitats with more extreme differences confirmed that hares do alter their giving‐up densities when one habitat is clearly superior to another. The results are thereby consistent with theories of adaptive behaviour. But the results also demonstrate, when evaluating differences in habitat, that it is crucial to let the organisms we study define their own habitat preference.     

Risk of predation as a promoting factor of species divergence in threespine sticklebacks (Gasterosteus aculeatus L.)

Icelandic freshwater systems are geologically young and contain only six species of freshwater fish. As these species colonized Icelandic fresh waters they were presented with a diversity of unique, uncontested habitats and food resources, promoting the evolution of new behaviour strategies crucial to the formation of new morphs and speciation. To determine the likelihood that predation threat could affect the antipredator behaviour and possibly the sympatric divergence of prey populations, we analysed antipredator behaviour of seven groups of Icelandic threespine sticklebacks ( Gasterosteus aculeatus ): two marine groups, one group from a lake without piscine predators, and two polymorphic lake populations, each with two groups occupying unique habitats. Shoaling cohesion, school formation and duration, and vigilance in predator inspection/avoidance behaviour varied greatly among groups. The differences appeared to be related to the risk of predation as well as to opportunities and constraints set by the different habitats. Antipredator behaviour was especially pronounced and differed extensively in two polymorphic forms from the lake Thingvallavatn, where predation risk is very high. By keeping the two morphs separate in their respective habitats, high predation risk may be a contributing factor in promoting the habitat-specific divergence of G. aculeatus seen in the lake. This suggests that in situations where refuge habitats are spatially separated, the risk of predation may contribute to the evolution of separate sympatric forms of small fish such as G. aculeatus .  © 2004 The Linnean Society of London, Biological Journal of the Linnean Society , 2004, 82 , 189–203.     

Turbidity amplifies the non‐lethal effects of predation and affects the foraging success of characid fish shoals

相似文献   

Scale dependent effects of predatory fish on stream benthos

In open predation experiments the effects of predators on prey densities can be influenced by predator consumption and by prey movements in to and out of experimental arenas. A published model predicts that the predator effects observed in such experiments are scale dependent over the scale range where there is a transition from movement control (of prey densities) to consumption control. The scale dependence follows from the assumption that per capita rate of emigration out of an experimental arena decreases with increasing arena size.
To test this model the effects of a small benthic fish ( Cottus gobio ) on densities of stream invertebrates was investigated in instream channels of different length (0.5, 2 and 8 m). The effect of fish predation was scale dependent for four prey taxa. For three of these taxa predator effects increased with experimental scale, which is in agreement with model predictions. However, this proved to be a case of "making the right prediction for the wrong reason" as the basic assumption of scale dependent emigration rate was not upheld. By analyzing the behaviour of the model, parameterized with emigration and consumption rates observed in the experimental channels, it was found that observed scale effects occurred because prey emigration in response to the predator treatment was modified by the experimental scale. Further analysis of the parameterized model suggested that the densities of most prey taxa were controlled by prey movements and not by consumption by the sculpins.     

Predictable changes in predation mortality as a consequence of changes in food availability and predation risk

Theory predicts that animals will have lower activity levels when either the risk of predation is high or the availability of resources in the environment is high. If encounter rates with predators are proportional to activity level, then we might expect predation mortality to be affected by resource availability and predator density independent of the number of effective predators. In a factorial experiment, we tested whether predation mortality of larval wood frogs, Rana sylvatica, caused by a single larval dragonfly, Anax junius, was affected by the presence of additional caged predators and elevated resource levels. Observations were consistent with predictions. The survival rate of the tadpoles increased when additional caged predators were present and when additional resources were provided. There was no significant interaction term between predator density and food concentration. Lower predation rates at higher predator density is a form of interference competition. Reduced activity of prey at higher predator density is a potential general mechanism for this widespread phenomenon. Higher predation rates at low food levels provides an indirect mechanism for density-dependent predation. When resources are depressed by elevated consumer densities, then the higher activity levels associated with low resource levels can lead to a positive association between consumer density and consumer mortality due to predation. These linkages between variation in behaviour and density-dependent processes argue that variation in behaviour may contribute to the dynamics of the populations. Because the capture rate of predators depends on the resources available to prey, the results also argue that models of food-web dynamics will have to incorporate adaptive variation in behaviour to make accurate predictions.     

Virtual plankton: A novel approach to the investigation of aquatic predator‐prey interactions

Small‐scale zooplankton swimming behaviors can affect aquatic predator‐prey interactions. Difficulties in controlling prey swimming behavior however, have restricted the ability to test hypotheses relating differences in small‐scale swimming behavior to frequency of predation by fish. We report here a Virtual Plankton (VP) system that circumvents this problem by allowing the observation of fish “preying"on computer‐generated prey images whose size, shape, color and swimming behavior can be precisely controlled. Two experiments were performed in which bluegill sunfish (Lepomis macrochirus) were given a choice of either two VP images, one of which moved twice as fast as the other, or six VP, one of which moved either faster (1.25 x, 1.5 x or 2 x ) or slower (0.5 x) than the other five. Current predator‐prey models based on encounter probabilities and prey visibility predict that moving faster increases predation risk and conversely, moving slower decreases predation risk. In agreement with existing predator‐prey models, in both experiments, fish chose faster moving VP significantly more often than their slower moving neighbors. Contrary to the predictions of existing models, in the second experiment with six VP, the rate at which fish chose a prey image moving half as fast as the five surrounding images did not differ significantly from the rate predicted by chance(l/6). These results suggest that current fish‐zooplankton predation models would benefit by the incorporation of small‐scale swimming behavior and assessments of its influence on overall prey visibility.     

A comparison of the feeding biology of Mink Mustela vison and otter Lutra lutra

The feeding habits and prey selectivity of Mink Mustela vison and otters Lutra lutra were compared in two localities in Devon: a eutrophic lake and a moorland river, in which both species occurred and had access to the same prey populations. The effects of prey availability on the predators' diets were assessed by comparing prey consumed, as revealed by scat analysis, with estimates of prey abundance and size range. Otters specialized in fish at all times of year but showed seasonal variation in species taken. Selection for slow-moving fish and seasonal changes in behaviour of some fish species were the probable causes of this variation. Otters diversified more into non-fish food in summer, when fish availability was reduced. The main alternative prey in the lacustrine habitat was waterfowl, but in the riverine habitat, rabbits. Mink were more generalized carnivores, taking a variety of fish, waterside and terrestrial prey in all seasons. These three prey categories were taken to an almost equal extent in the lake but terrestrial prey dominated in the riverine habitat. Fish were taken most frequently in winter and birds and mammals in summer. Neither predator showed selection in respect of prey size. In each area, about one third of the otter and Mink diets was common to both species. Fish was the principal group of the shared component, and dietary overlap in respect of them was greatest in autumn and winter. In view of the dietary preferences of each predator, the existence of alternative prey items and limited degree of dietary overlap, it is considered unlikely that the two species competed for food to any extent. Other factors must therefore be responsible for the spread of feral Mink and the decline in otter populations in many parts of Britain.     

A conceptual model of the dynamics of the Syamozero lake community

We have developed a conceptual mathematical model of the dynamics of a spatially heterogeneous population system, the prototype of which is the Syamozero lake fish community. Based on the analysis of solutions of the model, we show that interrelations between prey and predator populations in two neighboring habitats (pelagic and offshore zones) can lead to both undamped oscillations and stationary values of the population size. The population density was found to be close to the values oblained in the course of long-term observations of the biota of the Syamozero lake. Besides, we showed that the transition to the stationary states can be accompanied by long-term (dozens and hundreds of years) damped oscillations of the prey and predator population size. In natural waters, long-term transition periods can prevent the attainment of stationary regimes of fish community functioning.     

Metapopulation dynamics for spatially extended predator–prey interactions

Traditional metapopulation theory classifies a metapopulation as a spatially homogeneous population that persists on neighboring habitat patches. The fate of each population on a habitat patch is a function of a balance between births and deaths via establishment of new populations through migration to neighboring patches. In this study, we expand upon traditional metapopulation models by incorporating spatial heterogeneity into a previously studied two-patch nonlinear ordinary differential equation metapopulation model, in which the growth of a general prey species is logistic and growth of a general predator species displays a Holling type II functional response. The model described in this work assumes that migration by generalist predator and prey populations between habitat patches occurs via a migratory corridor. Thus, persistence of species is a function of local population dynamics and migration between spatially heterogeneous habitat patches. Numerical results generated by our model demonstrate that population densities exhibit periodic plane-wave phenomena, which appear to be functions of differences in migration rates between generalist predator and prey populations. We compare results generated from our model to results generated by similar, but less ecologically realistic work, and to observed population dynamics in natural metapopulations.     

Effects of spatial scale and foraging efficiency on the predictions made by spatially-explicit models of fish growth rate potential

Synopsis Spatially-explicit modeling of fish growth rate potential is a relatively new approach that uses physical and biological properties of aquatic habitats to map spatial patterns of fish growth rate potential. Recent applications of spatially-explicit models have used an arbitrary spatial scale and have assumed a fixed foraging efficiency. We evaluated the effects of spatial scale, predator foraging efficiency (combined probabilities of prey recognition, attack, capture, and ingestion), and predator spatial distribution on estimates of mean growth rate potential of chinook salmon,Oncorhynchus tshawytscha. We used actual data on prey densities and water temperatures taken from Lake Ontario during the summer, as well as, simulated data assuming binomial distribution of prey. Results show that a predator can compensate for low foraging efficiency by inhabiting the most profitable environments (regions of high growth rate potential). Differences exist in predictions of growth rate potential across spatial scales of observation and a single scale may not be adequate for interpreting model results across seasons. Continued refinements of this modeling approach must focus on the assumptions of stationary distributions of predator and prey populations and predator foraging tactics.     

Ultraviolet light influences habitat preferences in a fish under predation risk

Environmental light conditions are of general importance in predator–prey interactions. In aquatic systems, prey individuals experience different levels of predation risk depending on the properties of the visual environment, such as structural complexity or water transparency. To reduce the threat of predation, prey should move to habitats providing better protection against visual predators. We studied the role of UV wavelengths in habitat choice behaviour under predation risk in a fish, the three-spined stickleback (Gasterosteus aculeatus) that uses UV signals in different contexts of intraspecific communication. In a laboratory experiment sticklebacks were exposed to a predatory threat and given the choice between two escape habitats, one providing full-spectrum conditions including UV light (UV+) and one without UV wavelengths (UV−). Fish from two rearing treatments were tested, one group had been raised under natural lighting conditions (UV+), the other group under UV-deficient lighting conditions (UV−). Sticklebacks from the UV+ group preferred the UV− habitat as a refuge which suggests that predator avoidance behaviour is UV-related in this species with UV− conditions presumably being advantageous for prey fish. However, individuals from the UV− treatment group were equally attracted to both presented light habitats. It is possible that these fish could not discriminate between the two light habitats due to physiological limitations caused by their rearing conditions. Further control trials with neutral-density filters revealed that the UV− habitat preference of UV+ fish in the main experiment was rather not influenced by a difference in achromatic brightness between the UV+ and UV− habitat.     

Dying from the lesser of three evils: facilitation and non‐consumptive effects emerge in a model with multiple predators

Prey modify their behaviour to avoid predation, but dilemmas arise when predators vary in hunting style. Behaviours that successfully evade one predator sometimes facilitate exposure to another predator, forcing the prey to choose the lesser of two evils. In such cases, we need to quantify behavioural strategies in a mix of predators. We model optimal behaviour of Atlantic cod Gadus morhua larvae in a water column, and find the minimal vulnerability from three common predator groups with different hunting modes; 1) ambush predators that sit‐and‐wait for approaching fish larvae; 2) cruising invertebrates that eat larvae in their path; and 3) fish which are visually hunting predators. We use a state‐dependent model to find optimal behaviours (vertical position and swimming speed over a diel light cycle) under any given exposure to the three distinct modes of predation. We then vary abundance of each predator and quantify direct and indirect effects of predation. The nature and strength of direct and indirect effects varied with predator type and abundance. Larvae escaped about half the mortality from fish by swimming deeper to avoid light, but their activity level and cumulative predation from ambush predators increased. When ambush invertebrates dominated, it was optimal to be less active but in more lit habitats, and predation from fish increased. Against cruising predators, there was no remedy. In all cases, the shift in behaviour allowed growth to remain almost the same, while total predation were cut by one third. In early life stages with high and size‐dependent mortality rates, growth rate can be a poor measure of the importance of behavioural strategies.     

Interplay between temperature,fish partial migration and trophic dynamics

Whereas many studies have addressed the mechanisms driving partial migration, few have focused on the consequences of partial migration on trophic dynamics, and integrated studies combining the two approaches are virtually nonexistent. Here we show that temperature affects seasonal partial migration of cyprinid fish from lakes to predation refuges in streams during winter and that this migration in combination with temperature affects the characteristics and phenology of lower trophic levels in the lake ecosystem. Specifically, our six‐year study showed that the proportion of fish migrating was positively related to lake temperature during the pre‐migration growth period, i.e. during summer. Migration from the lake occurred later when autumn water temperatures were high, and timing of return migration to the lake occurred earlier at higher spring water temperatures. Moreover, the winter mean size of zooplankton in the lake increased with the proportion of fish being away from the lake, likely as a consequence of decreased predation pressure. Peak biomass of phytoplankton in spring occurred earlier at higher spring water temperatures and with less fish being away from the lake. Accordingly, peak zooplankton biomass occurred earlier at higher spring water temperature, but relatively later if less fish were away from the lake. Hence, the time between phyto‐ and zooplankton peaks depended only on the amount of fish being away from the lake, and not on temperature. The intensity of fish migration thereby had a major effect on plankton spring dynamics. These results significantly contribute to our understanding of the interplay between partial migration and trophic dynamics, and suggest that ongoing climate change may significantly affect such dynamics.     

Habitat overlap between predatory benthic fish and their invertebrate prey in streams: the relative influence of spatial and temporal factors on predation risk

1. The spatial heterogeneity of ecosystems as well as temporal activity patterns of organisms can have far‐reaching effects on predator–prey relationships. We hypothesised that spatiotemporal constraints in mesohabitat use by benthic fish predators would reduce habitat overlap with benthic invertebrates and lead to mesohabitat‐specific predation risks. 2. We analysed the spatiotemporal activity patterns of two small‐bodied benthivorous fishes, gudgeon (Gobio gobio) and stone loach (Barbatula barbatula), and of benthic invertebrates in a small temperate stream during three 24‐h field experiments. By applying a novel method of field video observation, we monitored the spatiotemporal foraging behaviour of the fish in their natural environment. A parallel analysis of invertebrate mesohabitat use by means of small area Hess sampling allowed a direct estimation of habitat overlap at a pool–riffle scale. 3. Gudgeon showed a dominant spatial activity pattern preferring pools at all times of day, whereas stone loach used both mesohabitats but with a distinct temporal (nocturnal) activity pattern. The patterns of residence were not identical with those of active foraging. Invertebrate community composition differed significantly between mesohabitats but not between times of day. More than half of the total dissimilarity between pools and riffles was accounted for by six invertebrate taxa. Five of these were subject to higher fish predation in pools than in riffles. The total prey consumption of the two fish species together in pools was about three times as high as in riffles. Trophic niche breadth of stone loach and thus its predation range was broader than that of gudgeon. 4. These results indicate that the potential predation risk for stream invertebrates depends on the combination of spatial and temporal patterns of both predator and prey. Given the distinct differences in predation risk found between pools and riffles, we conclude that spatial heterogeneity at the mesohabitat scale can influence mechanisms and consequences of selective predation. We also suggest that the analysis of spatiotemporal predator–prey relationships should not be based on the premise that the main residence habitat and active foraging habitat of a predator are identical.     

Forecasting the effects of global change scenarios on bioaccumulation patterns in great lakes species

Climate change will have substantial impacts on biodiversity, particularly for aquatic species. Warming temperatures and changing weather patterns will also remobilize and modify chemical partitioning. Holding millions of cubic yards of sediments contaminated with persistent legacy chemicals such as polychlorinated biphenyls (PCBs) and dioxins, the Laurentian Great Lakes are a laboratory for observing interactions between biological and chemical responses to climate change. They provide a wide range of habitat to a variety of species, from littoral forage fish to deep‐water predators. In this paper, we couple bioenergetic and bioaccumulation models to investigate the biological and chemical effects of climate change in the Great Lakes. We consider three species: round goby, a warm‐water invasive forage fish; mottled sculpin, a cool‐water native forage fish; and lake trout, a cold‐water native predator. Using our coupled models, we calculate the accumulation of a representative persistent chemical, PCB‐77, under four climate scenarios for Lake Erie and Lake Superior. Predator–prey (lake trout–round goby) interactions and food availability (high–low) are incorporated into our simulations. For cool‐ to cold‐water species (sculpin, lake trout) we find that warm temperatures limit growth. For warm‐water species (round goby) cold temperatures limit growth. The impact of climate warming on growth depends on the winter lows as well as the summer highs of the scenario, in combination with the species' critical upper and lower thermal limits. We find conditions for high growth and consumption rates generally lead to high bioaccumulation. However, this can be confounded by predator–prey dynamics, as mismatches in the temperature preferences of predator and prey can lead to mismatches in relative growth and uptake rates. As predator–prey dynamics are expected to undergo substantial shifts with changing climate, these relative thermal sensitivities will be key in determining the implications of climate change for bioaccumulation, particularly in top predator species.     

Coexistence in streams: do source–sink dynamics allow salamanders to persist with fish predators?

Theory suggests that source–sink dynamics can allow coexistence of intraguild predators and prey, but empirical evidence for this coexistence mechanism is limited. We used capture–mark–recapture, genetic methods, and stable isotopes to test whether source–sink dynamics promote coexistence between stream fishes, the intraguild predator, and stream salamanders (Dicamptodon aterrimus), the intraguild prey. Salamander populations from upstream reaches without fish were predicted to maintain or supplement sink populations in downstream reaches with fish. We found instead that downstream reaches with fish were not sinks even though fish consumed salamander larvae—apparent survival, recruitment, and population growth rate did not differ between upstream and downstream reaches. There was also no difference between upstream and downstream reaches in net emigration. We did find that D. aterrimus moved frequently along streams, but believe that this is a response to seasonal habitat changes rather than intraguild predation. Our study provides empirical evidence that local-scale mechanisms are more important than dispersal dynamics to coexistence of streams salamanders and fish. More broadly, it shows the value of empirical data on dispersal and gene flow for distinguishing between local and spatial mechanisms of coexistence.     

Differences in predator composition alter the direction of structure‐mediated predation risk in macrophyte communities

Structural complexity strongly influences the outcome of predator–prey interactions in benthic marine communities affecting both prey concealment and predator hunting efficacy. How habitat structure interacts with species‐specific differences in predatory style and antipredatory strategies may therefore be critical in determining higher trophic functions. We examined the role of structural complexity in mediating predator–prey interactions across several macrophyte habitats along a gradient of structural complexity in three different bioregions: western Mediterranean Sea (WMS), eastern Indian Ocean (EIO) and northern Gulf of Mexico (NGM). Using sea urchins as model prey, we measured survival rates of small (juveniles) and medium (young adults) size classes in different habitat zones: within the macrophyte habitat, along the edge and in bare sandy spaces. At each site we also measured structural variables and predator abundance. Generalised linear models identified biomass and predatory fish abundance as the main determinants of predation intensity but the efficiency of predation was also influenced by urchin size class. Interestingly though, the direction of structure‐mediated effects on predation risk was markedly different between habitats and bioregions. In WMS and NGM, where predation by roving fish was relatively high, structure served as a critical prey refuge, particularly for juvenile urchins. In contrast, in EIO, where roving fish predation was low, predation was generally higher inside structurally complex environments where sea stars were responsible for much of the predation. Larger prey were generally less affected by predation in all habitats, probably due to the absence of large predators. Overall, our results indicate that, while the structural complexity of habitats is critical in mediating predator–prey interactions, the direction of this mediation is strongly influenced by differences in predator composition. Whether the regional pool of predators is dominated by visual roving species or chemotactic benthic predators may determine if structure dampens or enhances the influence of top–down control in marine macrophyte communities.