Individuals from different-looking animal species may group together to confuse shared predators: simulations with artificial neural networks

Individuals of many quite distantly related animal species find each other attractive and stay together for long periods in groups. We present a mechanism for mixed-species grouping in which individuals from different-looking prey species come together because the appearance of the mixed-species group is visually confusing to shared predators. Using an artificial neural network model of retinotopic mapping in predators, we train networks on random projections of single- and mixed-species prey groups and then test the ability of networks to reconstruct individual prey items from mixed-species groups in a retinotopic map. Over the majority of parameter space, cryptic prey items benefit from association with conspicuous prey because this particular visual combination worsens predator targeting of cryptic individuals. However, this benefit is not mutual as conspicuous prey tends to be targeted most poorly when in same-species groups. Many real mixed-species groups show the asymmetry in willingness to initiate and maintain the relationship predicted by our study. The agreement of model predictions with published empirical work, the efficacy of our modelling approach in previous studies, and the taxonomic ubiquity of retinotopic maps indicate that we may have uncovered an important, generic selective agent in the evolution of mixed-species grouping.     

Time‐to‐kill: measuring attack rates in a heterogenous landscape with multiple prey types

Although spatial heterogeneity of prey and landscapes are known to contribute to variation around predator‐prey functional response models, few studies have quantified these effects. We illustrate a new approach using data from winter movement paths of GPS‐collared wolves in the Rocky Mountains of Canada and time‐to‐event models with competing risks for measuring the effect of prey and landscape characteristics on the time‐to‐kill, which is the reciprocal of attack rate (aN) in a Holling's functional response. We evaluated 13 a priori models representing hypothesized mechanisms influencing attack rates in a heterogeneous landscape with two prey types. Models ranged from variants on Holling's disc equation, including search rate and prey density, to a full model including prey density and patchiness, search rates, satiation, and landscape features, which were measured along the wolf's movement path. Movement rates of wolves while searching explained more of the variation in time‐to‐kill than prey densities. Wolves did not compensate for low prey density by increasing movement rates and there was little evidence that spatial aggregation of prey influenced attack rates in this multi‐prey system. The top model for predicting time‐to‐kill included only search rate and landscape features. Wolves killed prey more quickly in flat terrain, likely due to increased vulnerability from accumulated snow, whereas attack rates were lower when wolves hunted near human‐made features presumably due to human disturbance. Understanding the sources of variation in attack rates provides refinements to functional response models that can lead to more effective predator–prey management in human‐dominated landscapes.     

The oddity effect drives prey choice but not necessarily attack time

The tendency of predators to preferentially attack phenotypically odd prey in groups (the oddity effect) is a clear example of how predator cognition can impact behaviour and morphology in prey. Through targeting phenotypically odd prey, predators are thought to avoid the cognitive constraints that delay and limit the success of attacks on homogenous prey groups (the confusion effect). In addition to influencing which prey a predator will attack, the confusion and oddity effects would also predict that attacks on odd prey can occur more rapidly than attacking the majority prey type, as odd prey are more easily targeted, but this prediction has yet to be tested. Here, we used kerri tetra fish, Inpaichthys kerri, presented with mixed phenotypic groups of Daphnia dyed red or black to investigate whether odd prey in groups are preferentially attacked and whether these attacks were faster than those on the majority prey type. In agreement with previous work, odd prey were targeted and attacked more often than expected from their frequency in the prey groups, regardless of whether the odd prey was red in a group of black prey or vice versa. However, no difference was found in the time taken to attack odd vs. majority prey items, contrary to our predictions. Our results suggest that the time taken to make an attack is determined by a wider range of factors or is subject to greater variance than the choice of which prey is selectively targeted in a group.     

Prey capture and selection throughout the breeding season in a deep‐diving generalist seabird,the thick‐billed murre

Generalist seabirds forage on a variety of prey items providing the opportunity to monitor diverse aquatic fauna simultaneously. For example, the coupling of prey consumption rates and movement patterns of generalist seabirds might be used to create three‐dimensional prey distribution maps (‘preyscapes’) for multiple prey species in the same region. However, the complex interaction between generalist seabird foraging behaviour and the various prey types clouds the interpretation of such preyscapes, and the mechanisms underlying prey selection need to be understood before such an application can be realized. Central place foraging theory provides a theoretical model for understanding such selectivity by predicting that larger prey items should be 1) selected farther from the colony and 2) for chick‐feeding compared with self‐feeding, but these predictions remain untested on most seabird species. Furthermore, rarely do we know how foraging features such as handling time, capture methods or choice of foraging location varies among prey types. We used three types of animal‐borne biologgers (camera loggers, GPS and depth‐loggers) to examine how a generalist Arctic seabird, the thick‐billed murre Uria lomvia, selects and captures their prey throughout the breeding season. Murres captured small prey at all phases of a dive, including while descending and ascending, but captured large fish mostly while ascending, with considerably longer handling times. Birds captured larger prey and dove deeper during chick‐rearing. As central place foraging theory predicted, birds travelling further also brought bigger prey items for their chick. The location of a dive (distance from colony and distance to shore) best explained which prey type was the most likely to get caught in a dive, and we created a preyscape surrounding our study colony. We discuss how these findings might aid the use of generalist seabirds as bioindicators.     

The evolution of warning signals as reliable indicators of prey defense

It is widely argued that defended prey have tended to evolve conspicuous traits because predators more readily learn to avoid defended prey when they are conspicuous. However, a rival theory proposes that defended prey have evolved such characters because it allows them to be distinguished from undefended prey. Here we investigated how the attributes of defended (unprofitable) and undefended (profitable) computer-generated prey species tended to evolve when they were subject to selection by foraging humans. When cryptic forms of defended and undefended species were similar in appearance but their conspicuous forms were not, defended prey became conspicuous while undefended prey remained cryptic. Indeed, in all of our experiments, defended prey invariably evolved any trait that enabled them to be distinguished from undefended prey, even if such traits were cryptic. When conspicuous mutants of defended prey were extremely rare, they frequently overcame their initial disadvantage by chance. When Batesian mimicry of defended species was possible, defended prey evolved unique traits or characteristics that would make undefended prey vulnerable. Overall, our work supports the contention that warning signals are selected for their reliability as indicators of defense rather than to capitalize on any inherent educational biases of predators.     

Body masses,functional responses and predator–prey stability

The stability of ecological communities depends strongly on quantitative characteristics of population interactions (type‐II vs. type‐III functional responses) and the distribution of body masses across species. Until now, these two aspects have almost exclusively been treated separately leaving a substantial gap in our general understanding of food webs. We analysed a large data set of arthropod feeding rates and found that all functional‐response parameters depend on the body masses of predator and prey. Thus, we propose generalised functional responses which predict gradual shifts from type‐II predation of small predators on equally sized prey to type‐III functional‐responses of large predators on small prey. Models including these generalised functional responses predict population dynamics and persistence only depending on predator and prey body masses, and we show that these predictions are strongly supported by empirical data on forest soil food webs. These results help unravelling systematic relationships between quantitative population interactions and large‐scale community patterns.     

Intraspecific variation in responses to aposematic prey in a jumping spider (Phidippus regius)

Aposematic signals often allow chemically defended prey to avoid attack from generalist predators, including jumping spiders. However, not all individual predators in a population behave in the same way. Here, in laboratory trials, we document that most individual Phidippus regius jumping spiders attack and reject chemically defended milkweed bugs (Oncopeltus fasciatus), immediately releasing them unharmed. However, a small number of individuals within the population kill and completely consume these presumably toxic prey items. This phenomenon was infrequent with only 14% of our sample (17/122) consuming the milkweed bugs over the course of the study. Individuals that killed and consumed bugs often did so repeatedly; specifically, individuals that consumed a bug in their first test were more likely to kill a bug in their second test and also tended to consume them again. We explored what might drive some (but not all) individuals to consume these bugs and found that neither sex, sexual maturity, body size, laboratory housing type, nor being wild-caught or being laboratory-reared, predicted milkweed bug consumption. Consuming bugs had no negative effects on spider mass or body condition; contrary to expectations, individuals that consumed milkweed bugs actually gained more body mass and increased in body condition. We discuss potential behavioural and physiological variation between individuals that may drive these rare behaviours and the implications for the evolution of prey defences.     

Influence of predation risk on diet selection: a simple example in the grey squirrel

Current models of the optimal diet are special cases of a more general (and complex) model which incorporates the effects of predation risk on diet selection; this follows from an assumption implicit in current models that all prey items are eaten where they are encountered. Relaxing this assumption so that a forager might carry a prey item to protective cover for consumption leads to the conclusion that the value of a prey item is a function of its distance to cover as well as its energy content and handling time. Such considerations can significantly alter the outcome of diet selection relative to that expected from simple diet theory. We found that grey squirrels (Sciurus carolinensis) may reject more energetically profitable, but small food items in favour of locating larger, less energetically profitable items that can be carried to protective cover for consumption without greatly sacrificing foraging efficiency. The squirrel's tendency to reject a more profitable item is a function of its distance from cover and the size of the less profitable items. Such behaviour is inconsistent with predictions of current diet models, but is consistent with our qualitative predictions based on a previously determined predation-risk-foraging-efficiency trade-off in the grey squirrel.     

Warming up the system: higher predator feeding rates but lower energetic efficiencies

Predictions on the consequences of the rapidly increasing atmospheric CO₂ levels and associated climate warming for population dynamics, ecological community structure and ecosystem functioning depend on mechanistic energetic models of temperature effects on populations and their interactions. However, such mechanistic approaches combining warming effects on metabolic (energy loss of organisms) and feeding rates (energy gain by organisms) remain a key, yet elusive, goal. Aiming to fill this void, we studied the metabolic rates and functional responses of three differently sized, predatory ground beetles on one mobile and one more resident prey species across a temperature gradient (5, 10, 15, 20, 25 and 30 °C). Synthesizing metabolic and functional‐response theory, we develop novel mechanistic predictions how predator–prey interaction strengths (i.e., functional responses) should respond to warming. Corroborating prior theory, warming caused strong increases in metabolism and decreases in handling time. Consistent with our novel model, we found increases in predator attack rates on a mobile prey, whereas attack rates on a mostly resident prey remained constant across the temperature gradient. Together, these results provide critically important information that environmental warming generally increases the direct short‐term per capita interaction strengths between predators and their prey as described by functional‐response models. Nevertheless, the several fold stronger increase in metabolism with warming caused decreases in energetic efficiencies (ratio of per capita feeding rate to metabolic rate) for all predator–prey interactions. This implies that warming of natural ecosystems may dampen predator–prey oscillations thus stabilizing their dynamics. The severe long‐term implications; however, include predator starvation due to energetic inefficiency despite abundant resources.     

Prey capture and processing behaviors vary with prey size and shape in Australian and subantarctic fur seals

When hunting at sea, pinnipeds should adapt their foraging behaviors to suit the prey they are targeting. We performed captive feeding trials with two species of otariid seal, Australian fur seals (Arctocephalus pusillus doriferus) and subantarctic fur seals (Arctocephalus tropicalis). This allowed us to record detailed observations of how their foraging behaviors vary when presented with prey items that cover the full range of body shapes and sizes encountered in the wild. Small prey were captured using suction alone, while larger prey items were caught in the teeth using raptorial biting. Small fish and long skinny prey items could then be swallowed whole or processed by shaking, while all prey items with body depths greater than 7.5 cm were processed by shaking at the water's surface. This matched opportunistic observations of feeding in wild Australian fur seals. Use of “shake feeding” as the main prey processing tactic also matches predictions that this method would be one of the only tactics available to aquatic tetrapods that are unable to secure prey using their forelimbs.     

Experiments on predation and substratum choice by larvae of the muscid fly, Limnophora riparia

Larvae of the muscid fly Limnophora riparia live in lake outlets and prey on other invertebrates living there. In experiments, we demonstrated that larvae prefer moss as a substratum, though they will bury themselves in any suitable material to avoid light. The substratum is used to anchor Limnophora larvae as they attack their prey. When given a choice of prey they preferred chironomid and black fly larvae to oligochaetes and psyehodid larvae. Larvae of the black fly Simulium noelleri were used in laboratory experiments to test the interaction of predator and prey. Limnophora larvae attacked black fly larvae of all sizes, but preferred small larvae, the body contents of which were often removed completely. Increasing prey or predator density did not affect this latter preference, though an increase in predator density, or a decrease in prey density, did cause the predator to take prey of medium size as well as small prey. Limnophora larvae showed the same size preference when attacking dead (freshly-killed) prey and they preferred to attack larvae rather than pupae when both were available. They did not attack black fly eggs.     

The effects of hydrological disturbance on the densities of macroinvertebrate predators and their prey in a coastal stream

1. Environmental Stress Models (ESMs) predict that abiotic disturbance or harshness will differentially affect predators and prey. Consumer Stress Models (CSMs) predict that consumers will be relatively more inhibited by disturbance than prey, and therefore predator impacts will be reduced. Conversely, Prey Stress Models (PSMs) predict that prey will be more adversely affected and consequently predator impacts will increase in disturbed habitats. This study compared the relative tolerances of lotic invertebrate predators and their prey to hydrological disturbance in an Australian coastal stream to test the initial predictions of ESMs.
2. Macroinvertebrates were sampled with a suction sampler at monthly intervals and immediately following four high flow events at five sites on the Cumberland River, in south-west Victoria, Australia. Various statistical procedures were used to compare the relative resistance and resilience of predatory and prey taxa to each high flow event.
3. The relative resistances of seven predator and nine prey taxa to four floods over a 12-month period were highly variable between floods and between runs within the same flood. Prey taxa appeared to be more resilient than predators to the largest flood event, but there were no differences in the resilience of predators and prey following smaller floods. If disturbance tolerance is determined by resistance and resilience, then there was no consistent pattern of differential tolerance to floods among invertebrate predators and prey in this system.
4. The variability in the relative tolerances of taxa to different disturbance events makes general predictions about the effects of disturbance on the community-wide impact of predation extremely difficult.     

What you need is what you eat? Prey selection by the bat Myotis daubentonii

Optimal foraging theory predicts that predators are selective when faced with abundant prey, but become less picky when prey gets sparse. Insectivorous bats in temperate regions are faced with the challenge of building up fat reserves vital for hibernation during a period of decreasing arthropod abundances. According to optimal foraging theory, prehibernating bats should adopt a less selective feeding behaviour – yet empirical studies have revealed many apparently generalized species to be composed of specialist individuals. Targeting the diet of the bat Myotis daubentonii, we used a combination of molecular techniques to test for seasonal changes in prey selectivity and individual‐level variation in prey preferences. DNA metabarcoding was used to characterize both the prey contents of bat droppings and the insect community available as prey. To test for dietary differences among M. daubentonii individuals, we used ten microsatellite loci to assign droppings to individual bats. The comparison between consumed and available prey revealed a preference for certain prey items regardless of availability. Nonbiting midges (Chironomidae) remained the most highly consumed prey at all times, despite a significant increase in the availability of black flies (Simuliidae) towards the end of the season. The bats sampled showed no evidence of individual specialization in dietary preferences. Overall, our approach offers little support for optimal foraging theory. Thus, it shows how novel combinations of genetic markers can be used to test general theory, targeting patterns at both the level of prey communities and individual predators.     

Satiation and the functional response: a test of a new model

Abstract. 1. A model of the functional response to prey density is derived to include the reduction in time available for search, T_s , resulting from predator satiation.
2. For larger prey items predator satiation occurs at each prey capture and T_s is reduced by the attack time and digestive pause of a series of attack cycles. For small prey items predator foraging is continuous at low densities with T_s reduced solely by attack time. At higher densities predator satiation occurs after the capture of several small prey items and T_s is reduced by the attack time and digestive pause of a series of foraging cycles.
3. A comparison of the predicted asymptotic level of prey capture using experimentally estimated parameter values, with the maximum consumption of aphids by larval and adult coccinellids provides a test of the satiation model.
4. The limitation of prey capture by predator satiation is discussed with reference to handling time and the success of coccinellids in biological control.     

Do 3-D predators attack the margins of 2-D selfish herds?

To explain the evolution of grouping, Hamilton's selfish herdtheory assumes that predators attack the nearest prey and thatboth are acting on a 2-dimensional (2-D) plane. This proximityassumption in his theory is one explanation for marginal predation,the phenomenon whereby predators attack peripheral members ofa prey group. However, in some ecological circumstances, predatorsmove in 3-dimensional (3-D) space and prey in 2 dimensions.Because a predator coming from above or below the group mayhave relatively equal access to all members, marginal predationcannot be assumed. In this paper, we test whether marginal predationoccurs in such a 3-D/2-D geometry. We carried out 3 controlledlaboratory experiments in which fish attack prey grouped atthe water's surface. Predators were bass (Micropterous salmoides)or goldfish (Carassius auratus), and prey groups were eitherfree-swimming whirligig beetles (Dineutes discolor) or a constrainedgroup of tadpoles (Bufo bufo). In all 3 experiments, predatorswere significantly more likely to attack the periphery of preygroups. Our experiments also show that marginal predation isrobust to differences in overall density within a prey groupand that the fish are not reacting to observable state or behavioralcorrelates to position within a prey group. Furthermore, ourresults showed that predators will attack group margins evenwhen there is no variation, due to position, in nearest neighbordistance.     

State-dependent risk-taking by predators in systems with defended prey

Even defended prey items may contain nutrients that can sustain predators in times of energetic need. Conversely, a well-fed predator might be expected to avoid attacking prey items that have a chance of being defended, particularly if there is an abundance of familiar palatable prey to support it. To further understand the implications of optimal state-dependent foraging behaviour by predators in systems that contain defended prey, I developed a stochastic dynamic programming model. This state-dependent approach formally accounts for the trade-off between avoiding starvation and minimising harm from attacking defended prey. It predicts that the mean attack probability of predators on defended models and their undefended mimics should decline in a sigmoidal fashion with increasing availability of alternative undefended prey, and that the foraging decisions of predators should in general be relatively insensitive to the probability that a potentially defended prey item is indeed defended. Some implications of these predictions are that conspicuous warning signals are more likely to evolve in systems that contain an abundance of alternative undefended prey, and that imperfect mimicry will provide almost complete protection to the mimic when predators are readily supported by alternative food sources. Somewhat surprisingly, increasing the density of nutritious undefended mimics while keeping the densities of all other prey types constant tended to decrease the attack rates of predators on encounter with mimics and their defended models. This increase in dietary conservatism arose because in these cases there would be more prey available to sustain the predator if it ever found itself critically low in energy.     

Variation in predation risk and vole feeding behaviour: a field test of the risk allocation hypothesis

Many prey animals experience temporal variation in the risk of predation and therefore face the problem of allocating their time between antipredator efforts and other activities like feeding and breeding. We investigated time allocation of prey animals that balanced predation risk and feeding opportunities. The predation risk allocation hypothesis predicts that animals should forage more in low- than in high-risk situations and that this difference should increase with an increasing attack ratio (i.e. difference between low- and high-risk situations) and proportion of time spent at high risk. To test these predictions we conducted a field test using bank voles (Clethrionomys glareolus) as a prey and the least weasel (Mustela nivalis nivalis) as a predator. The temporal pattern and intensity of predation risk were manipulated in large outdoor enclosures and the foraging effort and patch use of voles were measured by recording giving-up densities. We did not observe any variation in feeding effort due to changes in the level of risk or the proportion of time spent under high-risk conditions. The only significant effect was found when the attack ratio was altered: the foraging effort of voles was higher in the treatment with a low attack ratio than in the treatment with a high attack ratio. Thus the results did not support the predation risk allocation hypothesis and we question the applicability of the hypothesis to our study system. We argue that the deviation between the observed pattern of feeding behaviour of bank voles and that predicted by the predation risk allocation hypothesis was mostly due to the inability of voles to accurately assess the changes in the level of risk. However, we also emphasise the difficulties of testing hypotheses under outdoor conditions and with mammals capable of flexible behavioural patterns.     

Shoals Receive more Attacks from the Wolf-Fish (Hoplias malabaricus Bloch, 1794)

We investigated whether the piscivorous wolf‐fish, Hoplias malabaricus (Bloch, 1794), showed preferences for attacking a ‘straggler’ or a shoal using guppies, Poecilia reticulata (Peters, 1860), as prey. Predators should show a preference to attack single prey items rather than grouped prey to overcome the confusion effect, which has been shown to negatively affect their capture success. However, they may attack groups more frequently because they are more likely to detect them because of their greater conspicuousness. In our study we looked at predator choice for grouped or single prey and whether this was affected by larger shoals being more conspicuous. We offered Hoplias binary choices of a single guppy (straggler) and shoals of 2, 6 and 10 fish. Hoplias preferred to attack the shoal over the straggler with regards to both frequency of attacks (initial and subsequent) and attack duration. The implications for predator–prey interactions of such a preference are discussed.     

Effects of temporal variation in the risk of predation by least weasel (Mustela nivalis) on feeding behavior of field vole (Microtus agrestis)

Predation risk tends to vary in time. Thus prey animals face a problem of allocating feeding and antipredator effort across different risk situations. A recent model of Lima and Bednekoff (1999) predicts that a prey should allocate more feeding effort to low risk situations and more antipredator effort to high risk situations with increasing relative degree of risk in high risk situations (attack ratio). Furthermore when the proportion of time the prey spends in the high risk situation (p) increases, the prey have to eventually feed also in the high risk situations. However the increase in feeding effort in low risk situations should clearly exceed that in high risk situations as p increases. To test these predictions we measured feeding effort of field voles (Microtus agrestis) exposed to varying presence of least weasel (Mustela nivalis) and its feces in laboratory conditions. We generated quantitative predictions by estimating attack ratios from results of a pilot experiment. The model explained 15% of the observed variation in feeding effort of voles. Further analyses indicated that feeding effort was lower in high risk situations than in low risk situations at high attack ratio, but not at a lower one. Voles exposed to a presence of a weasel for extended periods showed signs of nutritional stress. Still we did not find any increase in feeding effort with increasing p. This was obviously due to the relatively low maximal p we used as we included only conditions likely to occur in nature.     

Genetically-based variation between two spider populations in foraging behavior

Summary Optimal foraging theory is based on the assumption that at least some aspects of foraging behavior are genetically determined (Pyke et al. 1977; Kamil and Sargent 1980; Pyke 1984). Nonetheless, very few studies have examined the role of genetics in foraging behavior. Here, we report on geographical differences in the foraging behavior of a spider (Agelenopsis aperta) and investigate whether these differences are genetically determined. Field studies were conducted on two different populations of A. aperta: one residing in a desert riparian habitat, and the other in a desert grassland habitat. Data from the spiders' natural encounters with prey demonstrated that grassland spiders exhibited a higher frequency of attack than riparian spiders towards 13 of 15 prey types, including crickets and ants. Grassland spiders also had shorter latencies to attack 12 of 15 prey types, including crickets and ants, than riparian spiders. Subsequently, we reared grassland and riparian spiders under controlled conditions in the laboratory and observed their interactions with prey to determine whether the populational differences we found in the field could be genetic. Again, grassland spiders showed a shorter latency to attack prey (crickets, ants) than riparian spiders. These latencies were not significantly affected by the hunger state or age of the spiders. Finally, we reared a second generation (F2) of grassland and riparian spiders in the laboratory and observed their interactions with prey to determine whether the populational differences in the previous generation were due to genetic effects or maternal effects. As before, grassland spiders exhibited a shorter latency to attack prey (crickets) than riparian spiders. We conclude that the foraging differences we observed between these two populations of A. aperta are genetically determined. These differences probably have resulted from either natural selection acting directly on attack frequency and the latency to attack prey, or natural selection acting on traits which are genetically correlated with these aspects of foraging behavior.