Article Navigation
Article Contents
-
Abstract
-
INTRODUCTION
-
METHODS
-
Results
-
DISCUSSION
-
SUMMARY
-
FUNDING
-
References
- < Previous
- Next >
Journal Article
, Kate R. Searle Commonwealth Scientific and Industrial Research Organization (CSIRO) Sustainable Ecosystems, Davies Laboratory, PMB PO Aitkenvale, QLD 4814, Australia Search for other works by this author on: Oxford Academic Chris J. Stokes Commonwealth Scientific and Industrial Research Organization (CSIRO) Sustainable Ecosystems, Davies Laboratory, PMB PO Aitkenvale, QLD 4814, Australia Search for other works by this author on: Oxford Academic Iain J. Gordon Commonwealth Scientific and Industrial Research Organization (CSIRO) Sustainable Ecosystems, Davies Laboratory, PMB PO Aitkenvale, QLD 4814, Australia Search for other works by this author on: Oxford Academic
Behavioral Ecology, Volume 19, Issue 3, May-June 2008, Pages 475–482, https://doi.org/10.1093/beheco/arn004
Published:
19 February 2008
Article history
Received:
23 May 2007
Revision received:
30 November 2007
Accepted:
06 December 2007
Published:
19 February 2008
- Split View
- Views
- Article contents
- Figures & tables
- Video
- Audio
- Supplementary Data
-
Cite
Cite
Kate R. Searle, Chris J. Stokes, Iain J. Gordon, When foraging and fear meet: using foraging hierarchies to inform assessments of landscapes of fear, Behavioral Ecology, Volume 19, Issue 3, May-June 2008, Pages 475–482, https://doi.org/10.1093/beheco/arn004
Close
Search
Close
Search
Advanced Search
Search Menu
Abstract
Anthropogenic environmental change is escalating in magnitude, rate, and extent, inducing cascading effects across trophic levels. Assessing the nature of these alterations to trophic interactions requires an understanding of how species' demography and behavior are altered by simultaneous, complex pressures. For predator–prey relationships, “landscapes of fear” have been used to measure the trade-off prey animals make between maximizing energy gain and minimizing risk of predation. However, hierarchical foraging theory predicts that the degree to which aggregations of resources are used will depend upon the context in which they occur, not merely on the predation risk associated with those patches. We develop a conceptual framework that synthesizes theories of foraging hierarchies and landscapes of fear to show how predation risk and resource variation may interact to influence foraging behavior. We show, experimentally, that northern brown bandicoots (Isoodon macrourus), do respond to the likely predation risk when making their foraging decisions; however, the food resources in the habitat surrounding the food patch also play a significant role in the degree to which food patches are used. This result has important implications for the accuracy of assessments of landscapes of fear and habitat use using observations of animal foraging behavior.
INTRODUCTION
Ecologists have only recently begun to understand the importance of nonlethal effects of predation risk on prey behavior (Brown et al. 1999; Ripple and Beschta 2004; Creel et al. 2007). However, although the direct effects of predation can be obvious (e.g., changes to survival rates and population dynamics), the indirect costs of predation caused by changes to prey behavior can be more subtle. Typical approaches for quantifying the influence of fear on animal behavior involve implementing experiments that use observations of a behavior (e.g., quitting harvest rate) to titrate the effects of simultaneous pressures against one another (e.g., Kotler and Blaustein 1995). Much of behavioral ecology has been derived from experimental studies in artificial or semicontrolled natural environments that have allowed individual aspects of animal behavior to be isolated for study. Such work has greatly advanced our understanding of animal behavior, but has given rise to several independent branches of theory. The challenge that remains is to synthesize and apply these theories in real world environments, where animal decisions routinely involve complex trade-offs among multiple competing influences.
Ecological interactions across many ecosystems are strongly influenced by predator–prey relationships, and animals must balance their needs to meet biological and physical requirements with risk of predation (McNamara and Houston 1987). Doing so means that animals must consider spatial variation in predation risk. Understanding the influence of fear on animal behavior has posed a challenge for ecologists because it is unclear how to measure the predation risk an individual animal perceives in a particular environment. The prevailing approach has used foraging theory models that define the cost of predation relative to foraging activity as a mathematical equation for fear. Marginal rates of substitution of food for safety are used to measure the nature of predation risk for foraging animals (Brown 1988). Combining these models with empirical data from measurements of “giving-up densities” (GUDs, the amount of food remaining in a depletable food patch after a forager has ceased harvesting the patch, Brown 1988), in artificial food patches embedded in a landscape, allows ecologists to quantify the trade-off in energy for safety made by foraging animals when managing risk of predation. These GUDs can then be physically mapped as contour lines, and compared with physical features of the underlying habitat, to create a “landscape of fear” for the focal species (Laundré et al. 2001; Brown and Kotler 2004).
Foraging models predict that foragers should leave a patch when the fitness gains from foraging no longer exceed the fitness costs of potential predation (Brown 1988, 1999; Brown and Kotler 2004); in a more risky environment, the GUD of a forager should be higher than for a low-risk environment because of the relatively greater risk of predation associated with the foraging activity (Brown 1988). These data can then form an explicit measure of the topography of an animal's landscape of fear, providing valuable ecological insight into how animals perceive and exploit complex landscapes, and informing management (Morris and Davidson 2000; Schmitz 2005). However, missing from this approach has been consideration of the modifying influence of spatial variation in the prey's food resources. The theory underlying landscapes of fear includes the marginal value of energy and its relationship with fitness when determining the risk of predation for a forager, which will be influenced by food availability (Brown 1988; Brown and Kotler 2004). However, the practical application of this theory has not, thus far, attempted to explicitly account for the influence of spatial variation in food resources on measurements of GUDs as predicted by hierarchical foraging theory. This consideration has particular importance today, as animals face simultaneous changes in both predation risk, and habitat loss, transformation, and fragmentation.
Studies on the influence of spatial variation in resources on foraging behavior of animals have led to the development of hierarchical foraging theory (O'Neill 1986; Senft et al. 1987; Bailey et al. 1996); however, the development of theory supporting landscapes of fear has progressed in isolation from hierarchical foraging theory. This isolation has potentially large consequences for the way that experimenters routinely assess the landscape of fear for foraging animals. If, as predicted by hierarchical foraging theory, foraging animals respond to spatial variation in both predation risk and variation in food resources, accurate quantification of the food for safety trade-off will be reliant upon the marriage of these 2 branches of theory.
There are strong a priori reasons for using insights from hierarchical foraging theory to inform assessments of a foraging animal's landscape of fear. Hierarchy theory has been widely used to quantify scales of resource variation that are of functional significance to foraging animals (O'Neill et al. 1986; Senft et al. 1987; Kotliar and Wiens 1990; Bailey et al. 1996; Hobbs 1999). Foraging hierarchies identify key scales of spatial variation in resources that are functionally defined using differences in the rate of a behavior or process (Sih 1980; Senft et al. 1987; Bailey et al. 1996; Pinaud and Wiemerskirch 2007)—upper levels form constraints on the operation of processes at lower levels, and properties of lower levels explain the mechanism of an upper level process (O'Neill et al. 1986). Therefore, behavior at any one level of the foraging hierarchy is influenced by both upper and lower levels (Bailey et al. 1996; Pietrzykowski et al. 2003; Searle et al. 2006). The dynamics of upper hierarchical levels are typically much larger in space than the more fine-scale processes of lower levels (O'Neill et al. 1986; Senft et al. 1987). When observing a phenomenon at lower hierarchical levels, this means that upper level properties tend to appear constant and relatively unaffected by the level of interest (O'Neill 1986). This property of hierarchical systems has important implications for behavioral interactions between predators and prey. We contend that in the case of a prey species and its predator, the spatial overlap of the foraging hierarchies of both species bounds the region over which behavioral interactions between predator and prey are relatively predictable. This spatial overlap in the foraging hierarchies of predator and prey can be illustrated by comparing the extent of the spatial scale of each nested level in the respective foraging hierarchies. For example, a foraging herbivore may respond to its food resources at the scale of a feeding station (less than 1 m), a patch (10s of meters), and a plant community (100s of meters). Similarly, the predator may respond to variation in herbivore distribution at the scale of foraging sites utilized by individuals or small groups of herbivores (10s of meters), vegetation communities that convey a particular vulnerability for prey animals (100s of meters), and home ranges of prey populations (1000s of meters). In this example, the spatial overlap of the predator and prey foraging hierarchies occurs on the scale of 10–100s of meters, and it is this scale that delineates the region over which behavioural interactions are relatively predictable. As such, this region of overlap forms the landscape of fear for the prey species (Figure 1). At these overlapping scales, the outcome of prey–resource and predator–prey interactions drives the emergent behavior of both species. For instance, consider a predator–prey game in which the prey's best behavioral option depends on that of the predator, and vice versa (Sih 1998; Brown et al. 2001; Lima 2002). In this game, prey animals respond to variation in food resources and predation risk to maximize fitness gains from foraging and future survival, and predators respond by focusing their own foraging on areas of greatest prey density and vulnerability, acting to manage the fear level of their prey (Brown et al. 1999). Prey may then respond to the behavioral strategies of the predator by selecting safe places, or times, for foraging activity (e.g., Kotler et al. 1991; Wirsing et al. 2007). These reciprocal interactions between the foraging behavior of prey and predator mean that spatial and temporal variation in the prey's food resources at these scales is implicitly incorporated into the behavioral reactions of the predator (e.g., Kotler et al. 2002; Rosenheim 2004). Variation in forage resources of the prey transcends trophic levels, such that environmental variation and resource availability are explicitly incorporated into the foraging game (Kotler et al. 2002).
Figure 1
Open in new tabDownload slide
Illustrative foraging hierarchies for a predator and its prey, defining scales of resource variation for a landscape of fear. Spatial overlap of levels in foraging hierarchies for predator and prey forms the extent of the prey's perception of the landscape of fear. This delimits the zone over which predator–prey interactions are predictable in time and space (gray). At lower hierarchical levels, the prey perceives risk of predation to be constant, because the predator does not respond to variation in prey behavior at these scales (green). Resource heterogeneity at these lower hierarchical levels in the prey's foraging hierarchy (white) will influence experimental measurements of the landscape of fear, identifying the scales at which linkages between resource variation and the landscape of fear become important.
For predators feeding on mobile prey, there is often no benefit derived from investing effort in predicting fine-scale behaviors of the prey species, because prey are simply too variable over time and space to provide useful information (Zhang and Sanderson 1993). This has important implications for linking predator–prey interactions with hierarchical variation in resources—when a portion of the prey's foraging hierarchy exists at finer spatial scales than the predator's foraging hierarchy, resource heterogeneity at those finer scales may not be explicitly incorporated into behavioral interactions between predator and prey. Resource variation at finer scales of the prey's foraging hierarchy is not perceived by the predator, meaning that the prey's perception of predation risk at these finer scales is constant because the behavior of the predator is relatively invariant at these scales. So although the prey's behavioral response to this fine-scale resource variation is constrained by its perception of predation risk at coarser scales, the precise nature of its fine-scale response will be unaffected by predator–prey interactions at coarser scales (Figure 1). We propose that it is at these scales that linkages between resource variation and the landscape of fear will be most important. At its simplest, this logic leads to the hypothesis that the assessment of a prey species' landscape of fear using observations of foraging behavior will be modified by spatial variation in foraging resources within the landscape. We test this hypothesis by experimentally determining if the way foraging animals display their landscape of fear can be altered by varying the resource density in the landscape. We use foraging theory to derive 3 predictions for how GUDs will be altered by changing the resource density in a forager's landscape:
using the marginal value theorem (Charnov 1976),
using Brown's (1988) “H = C + P + MOC” model (where H is the harvest rate, C the metabolic cost, P the predation cost, and MOC the missed opportunity cost) in relation to addition of supplemental food, and
applying hierarchy theory in conjunction with the marginal value theory (MVT).
METHODS
We designed an experiment to test the prediction that variation in resources at lower levels in the foraging hierarchy of a prey species can influence the assessment of its perception of its landscape of fear. We measured habitat use of wild Isoodon macrourus, Gould 1842 (northern brown bandicoots), in a woodland–grassland landscape in northern Australia as they foraged at artificial feeding stations. Isoodon macrourus inhabit grassland, woodland, and open forest in eastern and northern Australia, preferring areas of ground cover such as tall grass and dense shrubbery (Gordon 1983). This species typically forages for invertebrates such as insects, spiders, and earthworms, although it will also consume plant products (berries and grass seeds) (Gordon 1983). Bandicoot species are preyed upon by several native and introduced mammals (Canis familiaris, Felis catus, Dasyurus spp.) and raptors (Aquila audax), and seek shelter and nest sites in areas of dense, low-lying vegetation (Gordon and Hulbert 1989; Southgate et al. 1996). Predation risk could, therefore, be manipulated by placing feeding stations at different distances from the protective cover of woodland habitat. A steady increase in fear with distance from cover is reliant upon individual bandicoots having a reasonable chance of escaping a predator attack once detected. Because bandicoots rely on rapid flight to escape predatory attacks, it is reasonable to assume that the effectiveness of this response is dependent on the animal's proximity to protective cover. We, therefore, manipulated predation risk by placing an array of feeding stations within a mown grassland at 3 distances (0, 20, and 35 m) from the protective cover of an adjacent woodland, with an understorey of tall grass. Within this array of feeding stations, we measured GUDs over 2 observation periods. In the first observation period (hom*ogeneous array), feeding stations were placed in the natural landscape without modification of the context around each feeding station. GUDs in this hom*ogeneous array were measured to quantify a “baseline” landscape of fear in the vicinity of the woodland margin. We hypothesize that this baseline assessment of the landscape of fear can be altered by changing the resource density within the landscape. Therefore, in the second observation period (modified array), we altered the landscape context by providing additional food (low or high levels) around some feeding stations and leaving some unchanged. The same total amount of food was available in the modified array for every night of observation. We specifically chose to add extra food in the modified array, rather than simply varying the density of food already present in the hom*ogeneous array, because this additional food represents naturally occurring food that has previously been ignored in experiments used to measure landscapes of fear. By varying the amount of extra food added around each of the 18 feeding stations in the modified array, we created variation in the resource density in the context of each feeding station used to measure GUDs; variation that is hypothesized to be important under considerations of hierarchical foraging theory. The change in GUD at each station (from the hom*ogeneous array to the modified array) was related to the changed context of surrounding food distribution across a range of scales (circles of radius 1, 21, 34, and 50 m) to examine how manipulated resource density affected changes in GUD. We used these 4 different scales of measurement for the resource density surrounding focal feeding stations because foraging animals are known to respond to variation in resources at different spatial scales.
Predictions from current theory
Foraging theory provides a number of hypotheses regarding how altering resource density, in the context of patches, should change patch exploitation. The most basic prediction, using MVT (Charnov 1976), expects foragers to quit patches when their instantaneous intake rate falls to the average rate for the habitat (Charnov 1976). The additional of supplemental food in the modified array serves to increase the average rate of energy intake in the habitat in comparison to that available in the hom*ogeneous array. As a consequence, MVT predicts that GUDs in all feeding stations should increase, to equate to the new, higher, quitting harvest rate.
Brown et al. (1988, 2004) have developed models to predict the exploitation of depletable food patches by foragers foraging under predation risk. These models are an extension of the marginal value theorem (Charnov 1976), but include important additions that allow for their application to much less restrictive, and more realistic sets of ecological conditions. The essential outcome of these models demonstrates that a forager seeking to maximize probability of survival and fitness during some fixed time period should exploit a patch until its harvest rate (H) falls to equal the sum of its metabolic (C), predation (P), and missed opportunity costs (MOC) of foraging—the H = C + P + MOC model (Brown 1988). When supplemental food is added to the environment, the foragers' marginal valuation of food (dF/de; where F is fitness and e is energy) decreases, given that fitness increases asymptotically with energy gain (Brown 1992; Brown et al. 1992; Kotler 1997). Presenting supplemental food should increase the energy reserves of the forager, affecting the marginal value of energy. Given this outcome, the marginal rate of substitution of energy for risk of predation is altered, such that foragers with higher energy reserves have more to lose from predation and less to gain from foraging (Brown 1992; Brown et al. 1992; Kotler 1997). Under these conditions, the model predicts that the addition of supplemental food should cause GUDs to increase in both safe and risky patches (Brown 1992); but that the differences between GUDs in safe and risky environments will be exaggerated, such that GUDs increase considerably more in risky patches than in safe patches (e.g., Brown et al. 1992; Kotler 1997). As a consequence, we expect that in the modified array, GUDs will increase considerably more in risky patches than in safe patches.
As a third alternative, we contend that current applications of foraging theory to predict changes in GUDs under varying resource conditions are flawed because they do not allow for a risk-independent effect of resource variation on GUDs at fine spatial scales, as predicted by hierarchy theory. We expect that there will be a component of foraging behavior that leads to a change in GUDs in response to fine-scale spatial variation in resources and that this component will operate independently from variation in predation risk. A basic formulation of this prediction implements the MVT and implies that in comparison to the hom*ogeneous array, the introduction of fine-scale resource variation in the modified array will lead to higher GUDs in feeding stations surrounded by high-resource density and lower GUDs in stations surrounded by low-resource density.
The presence of I. macrourus in this area was determined by nonlethal trapping at the woodland/grassland interface prior to the onset of the experiment. It was determined that at least 3 individuals were using the area regularly. We used video recording at selected feeding stations to confirm that I. macrourus were the only foragers responsible for the observed consumption of food.
hom*ogeneous array
In the hom*ogeneous array, feeding stations were placed at 0, 20, and 35 m from protective cover, with 6 replicates of each distance from cover treatment spaced 20 m apart (Figure 2). This created an array of 18 feeding stations (3 distances from cover × 6 replicates). Each feeding station consisted of 45 mealworm grubs (larval stage of Tenebrio molitor, mean mass 0.14 g) mixed into a container (30 × 20 × 10 cm, 0.06 m2) of sand. Fifteen live grubs were mixed into the sand within each third of the feeding station to create an even distribution with depth and to ensure decelerating gain. A wire mesh cage prevented birds from gaining access to the grubs, but a small opening at ground level permitted the bandicoots to enter. Feeding stations were placed out at dusk each night and collected at dawn. Animals were familiarized with the experimental layout by placing feeding stations nightly for 2 weeks prior to the first measurements. GUDs were then measured at each station over 9 consecutive nights of foraging by measuring the number of remaining grubs at dawn (grubs per 0.6m2, averaged for each station over the 9 nights).
Figure 2
Open in new tabDownload slide
Experimental layout of feeding stations and altered resource density in the modified array, within the mown grass/eucalypt woodland landscape. Black dots are the locations of feeding stations. The 3 modified resource densities are as follows: dark gray for high, light gray for low, and white for unchanged. hom*ogeneous array is identical, minus the addition of extra resources.
Modified array
The modified array used the same layout as the hom*ogeneous array, but the context of feeding stations was altered by systematically varying the resource context around each of the 18 feeding stations (Figure 2). Resource density was altered by placing 4 additional food trays around each feeding station (within a 1-m radius). Three levels of modified resource density were used; unchanged (no additional trays), low (2 grubs/tray), and high density (10 grubs/tray) in a replicated Latin square design (3 resource densities × 3 distances from cover × 2 replicates). This design allowed us to examine the influence of spatial variation in resource density on the behavior of I. macrourus within feeding stations. Grubs were buried in a shallow layer of sand approximately 2 cm deep in each tray (30 × 20 × 5 cm). The modified array was presented after observations had been made in the hom*ogenous array to establish the baseline measurement of the landscape of fear. Animals were then allowed 9 nights to become familiar with the modified array, after which GUDs were measured for the next 9 nights in the modified array.
Number of visits to feeding stations
The number and timing of visits per feeding station in each array were measured each night using pressure plates placed at the entrance to each feeding station. These pressure plates were not completely reliable (e.g., on a few nights bandicoots scooped sand from the feeding stations onto the plates so complete times for the night could not be recorded). We therefore restrict the use of this data to approximating the mean number of visits recorded at feeding stations, given that a feeding station was visited during the night, within each distance from cover (1, 20, and 35m). These data were used solely to examine any changes in the mean number of visits per feeding station, given a feeding station was visited, in relation to distance from cover, and array (hom*ogeneous vs. modified). This was to establish that differences in GUDs observed were due to individual animals altering their foraging behavior within feeding stations with respect to either distance from cover or array, rather than due to different numbers of individuals visiting feeding stations at different distances from cover or in the different arrays offered. The mean number of visits per feeding station, given that it was visited during the night, was averaged over the 9 nights of observation in the hom*ogeneous and modified arrays. A general linear model (PROC GLM, SAS Institute Inc., 2001) was fit to the data set for each array, using distance as a class variable, to determine if distance had a significant effect on the mean number of visits. Data from the hom*ogeneous and modified arrays were then combined, and a general linear model was fit using array (hom*ogeneous or modified) and distance (1, 20, 35 m) as class variables. This analysis was used to look for an effect of array (hom*ogeneous or modified) on the mean number of visits and for an interaction between array and distance from cover.
Analysis
Assessing the initial landscape of fear
To examine how I. macrourus traded-off in food for safety in this landscape, we analyzed the effect of distance from cover on GUDs in the hom*ogeneous array using an analysis of varinace with mean GUD per feeding station (n = 6 per distance from cover), averaged over the 9 nights of observation, as the dependent variable and distance from cover (0, 20, and 35 m) as an independent class variable (PROC GLM, SAS Institute Inc., 2001).
Examining the effect of introduced resource variation on assessment of the landscape of fear
The nature of the influence of altered resource context on foraging behavior was quantified using the change in GUDs that occurred in the modified array relative to the hom*ogeneous array. Baseline GUD measurements from the hom*ogeneous array accounted for pre-existing spatial variation in factors such as predation risk, distance from cover, and microhabitat variation. Subtracting this baseline GUD from the GUD measurement for each feeding station in the modified array provided a measure that resulted solely from the altered context of food distribution, the , where is the mean GUD at the ith feeding station averaged over the 9 nights of observation in the modified array and is the same quantity in the hom*ogeneous array.
To look for a modification of the way I. macrourus displayed their landscape of fear, once resource density in the landscape had been altered, we used a linear contrast to examine the interaction between distance from cover and GUDdiffi. If the MVT prediction is correct, when resource density is manipulated in the modified array, the intercept of the relationship between distance from cover and GUD should change, but the slope should remain the same. Alternatively, if the H = C + P + MOC model is correct, the intercept and the slope of the relationship between distance from cover and GUD should change in the modified array, with the slope becoming steeper due to foragers' placing a higher premium on foraging from risky patches. Finally, if GUDs are modified by fine-scale variation in resources irrespective of variation in risk as predicted by hierarchy theory, GUDs in the modified array should change predictably in response to the modified spatial variation in resources. Because foraging animals perceive and respond to variation in resources across multiples spatial scales, we examined how the behavior of animals responded to variation in resource density when measured to incorporate information over a range of distances from the focal feeding station. We used a general linear model to examine the effect of resource density averaged over different spatial scales (using circles with a radius of 1, 21, 34, or 50 m) on the change in GUD once the habitat had been modified. Resource density was calculated as the number of additional grubs within the specified distance of the focal station (1, 21, 34, or 50 m) divided by the area of the circle given by the specified radius. These spatial scales were chosen because they split resource density into categories such that points with geometrically equivalent configurations were grouped into the same distance category. Using a radius of 1 m captures variation in resource density immediately surrounding each feeding station; 21 m captures resource variation around the nearest horizontal and vertical neighbor of each feeding station; 34 m captures resource variation around the nearest horizontal, vertical, and diagonal neighbor of each feeding station; and finally, 50 m captures resource variation around the 2 nearest feeding stations along vertical, horizontal, and diagonal axis (Figures 2 and 4). The change in GUD once the habitat had been altered was calculated as
(1.1)
where GUDmodified,i,j is the GUD at the feeding station at the ith location on the jth night of observation in the modified array and is the mean GUD at the feeding station at the ith location in the hom*ogeneous array, averaged over the 9 nights of observation in the hom*ogeneous array. We used a repeated-measures general linear model with night as a class variable and resource density as a continuous variable (PROC MIXED, SAS Institute Inc., 2001). Given considerations of hierarchy theory and the MVT (Charnov 1976), we expected that in comparison to the hom*ogeneous array, GUDs in the modified array would increase in areas of high-resource density and decrease in areas of low-resource density.
Figure 3
Open in new tabDownload slide
Trade-off in food for safety made by Isoodon macrourus foraging at a woodland/grassland interface. Mean GUDs of I. macrourus foraging in feeding stations placed in a woodland/grassland landscape in the hom*ogeneous array (black circles, solid line), and in the modified array (white circles, dashed line). Error bars ± 1 SEM.
Figure 4
Open in new tabDownload slide
Influence of resource density on the change in GUDs between the modified and hom*ogeneous arrays. Slope estimates (±1 SEM) from general linear model of change in GUD () against resource density averaged over geometrically equivalent configurations of neighboring feeding stations. Point patterns show geometric configurations of focal and neighboring feeding stations for each radial distance used to measure resource density.
Results
When a feeding station was visited during the night, the mean number of visits per feeding station at 1, 20, and 35 m from cover were 1.96 (standard error of the mean [SEM] = 0.50), 1.05 (SEM = 0.12), and 1.25 (SEM = 0.50) in the hom*ogeneous array and 2.17 (SEM = 0.66), 1.23 (SEM = 0.24), and 1.00 (SEM = 0.00, n = 3) in the modified array. Distance from cover had no effect on the mean number of visits, given a station was visited, in the hom*ogeneous (F1 = 0.42, P = 0.67) or modified arrays (F1 = 0.23, P = 0.80). Furthermore, the mean number of visits, given a station was visited, did not differ between the hom*ogeneous and modified arrays (F1 = 0.17, P = 0.68), and there was no interaction between array and distance from cover (F2 = 0.07, P = 0.94).
Assessment of the initial landscape of fear
As predicted by the landscape of fear theory, in the hom*ogeneous array animals traded off food for safety by decreasing time spent foraging (leaving feeding stations at a higher GUD) at feeding stations as risk of predation increased (Figure 3). Distance from cover accounted for 80% of the variation in mean GUDs in the hom*ogeneous array, demonstrating a significant effect of distance from cover on GUDs (F2,15 = 30.65, P < 0.0001, R2 = 0.80). GUDs were approximately 11 times lower in feeding stations at the woodland edge compared with feeding stations furthest out into the open grassland (35 m from cover; F1 = 60.07, P = 0.0001) and approximately 7.5 times lower than in feeding stations partway out in the open grassland (20 m from cover; F1 = 23.38, P = 0.0002).
The effect of introduced resource variation on assessment of the landscape of fear
As predicted, the modified feeding array altered the way that I. macrourus displayed their landscape of fear. A linear contrast between distance from cover and the difference in mean GUD between the modified and hom*ogeneous arrays, GUDdiffi, showed that there was a significant interaction between distance and feeding array (t16 = 1.85, P = 0.041). In the modified array, GUDs in feeding stations at the woodland edge were about 4.5 times lower than those farthest out in the grassland (35 m from cover; Figure 3) and approximately 2.3 times lower compared with feeding stations partway out in the open grassland (20 m from cover; Figure 3). The slope of the relationship between distance from cover and GUD became less steep once resource variation in the context had been altered in the modified array (Figure 3).
We examined the effect of resource density on the change in GUD between the modified and hom*ogeneous arrays, , using a general linear model. Resource density had a significant influence on the change in GUD between the modified and hom*ogeneous feeding arrays over radial distances of 21 and 34 m (Table 1). Model selection revealed the model with the greatest support in the data averaged resource density over the nearest horizontal and vertical neighboring stations (radial distance of 21 m: Wr = 0.64, Table 1), although the model that averaged resource density over the nearest horizontal, vertical, and diagonal neighbors also received reasonable support (radial distance of 34 m: Wr = 0.34, Table 1). In contrast, the model that used resource density in the immediate context of feeding stations (radial distance of 1 m), ignoring the resource density of neighboring stations, received almost no support in the data (Wr = 0.014, Table 1). There was a negative relationship between increasing resource density in the context of feeding stations and the change in GUD between the modified and hom*ogeneous arrays (Figure 4). Comparing GUDs before (hom*ogeneous array) and after (modified array) manipulating the resource context of feeding stations shows that variation in resource density caused animals to lower GUDs in feeding stations surrounded by higher resource density and to increase GUDs in stations surrounded by lower resource density (Figure 4).
Table 1
Influence of resource density on the change in GUD between the modified and hom*ogeneous arrays
Radius of resource density (m) | F | P | AIC | Δr | Wr |
1 | 0.08 | 0.78 | 1162.1 | 7.7 | 0.0014 |
21 | 8.7 | 0.0037 | 1154.4 | 0 | 0.64 |
34 | 8.46 | 0.0042 | 1155.7 | 1.3 | 0.34 |
50 | 1.25 | 0.27 | 1163.4 | 9 | 0.0071 |
Radius of resource density (m) | F | P | AIC | Δr | Wr |
1 | 0.08 | 0.78 | 1162.1 | 7.7 | 0.0014 |
21 | 8.7 | 0.0037 | 1154.4 | 0 | 0.64 |
34 | 8.46 | 0.0042 | 1155.7 | 1.3 | 0.34 |
50 | 1.25 | 0.27 | 1163.4 | 9 | 0.0071 |
Restricted maximum likelihood was used to assess a suite of general linear models incorporating night as a repeated measure and resource density as a continuous covariate. Each general linear model calculated resource density using different geometric arrangements of neighboring feeding stations (determined as the average density of additional trays within a radius of 1, 21, 34, or 50 m of the focal feeding station). Model selection was used to quantify the support in the data for each model (Akaike's information criterion, AIC), given the data and the competing models (Burnham and Anderson 2002). Akaike weights (Wr) were used to assess the models, based on the likelihood of a candidate model given the data, and the difference between the value of the AIC between the best approximating model and the model in hand (Δr).
Open in new tab
Table 1
Influence of resource density on the change in GUD between the modified and hom*ogeneous arrays
Radius of resource density (m) | F | P | AIC | Δr | Wr |
1 | 0.08 | 0.78 | 1162.1 | 7.7 | 0.0014 |
21 | 8.7 | 0.0037 | 1154.4 | 0 | 0.64 |
34 | 8.46 | 0.0042 | 1155.7 | 1.3 | 0.34 |
50 | 1.25 | 0.27 | 1163.4 | 9 | 0.0071 |
Radius of resource density (m) | F | P | AIC | Δr | Wr |
1 | 0.08 | 0.78 | 1162.1 | 7.7 | 0.0014 |
21 | 8.7 | 0.0037 | 1154.4 | 0 | 0.64 |
34 | 8.46 | 0.0042 | 1155.7 | 1.3 | 0.34 |
50 | 1.25 | 0.27 | 1163.4 | 9 | 0.0071 |
Restricted maximum likelihood was used to assess a suite of general linear models incorporating night as a repeated measure and resource density as a continuous covariate. Each general linear model calculated resource density using different geometric arrangements of neighboring feeding stations (determined as the average density of additional trays within a radius of 1, 21, 34, or 50 m of the focal feeding station). Model selection was used to quantify the support in the data for each model (Akaike's information criterion, AIC), given the data and the competing models (Burnham and Anderson 2002). Akaike weights (Wr) were used to assess the models, based on the likelihood of a candidate model given the data, and the difference between the value of the AIC between the best approximating model and the model in hand (Δr).
Open in new tab
DISCUSSION
GUDs allow us to measure the trade-off in food for safety that animals routinely make. This approach has vastly augmented our understanding of many species (Brown et al. 1988; Kotler 1997; Mohr et al. 2003; Druce et al. 2006), and the ecological communities they inhabit (Brown et al. 1997; Morris and Davidson 2000; Pusenius and Schmidt 2002; Yunger et al. 2002). However, this trade-off does not operate in isolation; the foraging behavior of animals is an emergent property of all relevant sources of variation that exist within an environment. As such, it is embedded within the resource heterogeneity of the landscape and is inseparable from the resource variation that pervades the context of all decisions.
In this experiment, we have shown that animals simultaneously moderate their foraging behavior in response to both perceived predation risk and resource variation in the landscape context of their foraging hierarchy. Measurements in the hom*ogeneous array showed a clear decrease in foraging investment with distance from cover and, therefore, likely perceived predation risk. We contend that this effect was due to foragers trading off food for safety because the average number of visits per feeding station, given that a station was visited during the night, did not differ between the 3 distances from cover in either array. This indicates that the lower GUDs closest to cover were due to foragers increasing patch residence time, not from more individuals encountering and using those feeding stations closest to cover.
Most importantly, after altering landscape context, the animals spent more time foraging in feeding stations within areas of the landscape that had greater resource density (measured at a scale of approximately 20–35 m), independent of predation risk. The fact that there was no effect of array (hom*ogeneous or modified) on the mean number of visits per feeding station, given that a station was visited, supports our contention that the altered GUDs in the modified array were not a consequence of different numbers of individuals foraging within the array, but rather the same individuals choosing to alter residence time within feeding stations in response to changed resource context.
The first 2 predictions, which used MVT and the H = C + P + MOC models to predict changes in GUD in response to supplemental food, failed to explain the observed changes in GUDs once resource variation in the context had been altered. The MVT predicted that the slope of the relationship between distance from cover and GUD should not change between the hom*ogeneous and modified arrays, and yet it clearly did (Figure 3). Moreover, the slope of this relationship in the modified array did not become steeper as predicted by the H = C + P + MOC model but instead became less steep (Figure 3). The failure of the MVT to predict the observed changes in behavior is likely due to its restrictive assumptions, including nondepleting resources and constant risk of predation over all activities. We believe the failure of the H = C + P + MOC model lies in its omission of the influence of fine-scale spatial variation in resources in the local neighborhood of feeding stations.
Instead, our results demonstrate the importance of incorporating fine-scale spatial variation in resources when assessing landscapes of fear using GUDs. Isoodon macrourus foraging in this landscape interacted strongly with resource variation when it was expressed so as to include information about resource density at neighboring feeding stations. This finding validates the prediction from hierarchy theory that a component of the foragers' patch exploitation decision responds to fine-scale spatial variation in resources, independent from their response to spatial variation in risk. However, within this fine spatial scale, the feeding station exploitation behavior of I. macrourus did not respond as predicted by the MVT. We found that once resource density had been altered, animals invested more foraging effort in areas of the greatest resource density and less in areas of the lowest resource density. This is the opposite of predictions derived from the MVT. Thus, although the precise mechanisms driving I. macrourus foraging decisions remain unclear, our results clearly demonstrate that spatial context of food resources influences behavior of foragers and therefore needs to be better incorporated into foraging theory.
These results do provide some evidence that I. macrourus use a Bayesian approach when forming patch exploitation decisions. Bayesian models of patch utilization combine information contained in a prior distribution with information gained during patch sampling to generate a posterior distribution, upon which foraging decisions are based (Valone and Brown 1989; Valone 2006). The prior distribution may be a previously experienced distribution of the qualities of food patches over some temporal or spatial scale (McNamara et al. 2006). Our observations of the influence of the spatial context of food resources on I. macrourus feeding behavior may be partially explained in terms of Bayesian foraging. Isoodon macrourus appeared to utilize information about previously visited, neighboring feeding stations when making exploitation decisions in the current station. The immediate resource context of feeding stations (i.e., within 1 m) had no detectable influence on behavior within stations, rather it was the resource context provided by neighboring feeding stations that was responsible for the change in GUDs observed. These results suggest a degree of complexity to the influence of resource density on the GUD of foragers, which appears to vary nonlinearly with scale.
Importantly, our results demonstrate that fine-scale resource heterogeneity can influence measurement of a prey species' landscape of fear. Using the GUDs in the hom*ogeneous array to infer I. macrourus's food for safety trade-off produces a different landscape of fear than that created using the GUDs in the modified array. If we assume that measurements of GUDs in the hom*ogeneous array were an accurate assessment of the animals' landscape of fear, then the measurements of GUDs in the modified array represent a modification of this landscape of fear that is caused by introducing resource variation in the context of feeding stations. This experimentally introduced resource variation was meant to simulate resource patchiness present in natural landscapes that is typically ignored when conducting GUD experiments in the field. Therefore, if we were to solely measure the landscape of fear using observations from the modified array, we would underestimate the effect of predation risk on foraging investment on the order of approximately 250% at the furthest distance from cover (see below) and by around 330% at the middle distance from cover.
To illustrate this finding, we calculated the effect of additional resource variation on measurements of fear at 35 m as follows (refer to Figure 3 for GUDs at each distance from cover):
This underestimation is a result of the failure to recognize the influence of resource variation in the context of feeding stations to which foraging animals respond when deciding how much to invest in foraging within patches. Although these calculations are necessarily rough, they do demonstrate that experiments that use GUDs to measure a forager's landscape of fear without taking the effect of existing natural resource density into account are at risk of significant inaccuracies. This approach assumes that all of the differences in GUDs are attributable to variation in risk of predation, when in fact a considerable portion of those differences could be due to variation in food resources. Our experimental design did impose a temporal component to variation in resource variation (from the hom*ogeneous array to the modifed array) that may have introduced additional changes in drivers of foraging behavior, such as a potential change in the level of predation risk over time. However, the clear and demonstrable effect of variation in resource density on the change in observed GUDs is persuasive evidence for resource variation in the context of feeding stations being the primary driving force underlying the shift in behavior. Given these findings, it is evident that only by combining measurements of the simultaneous effects of resource variation and risk of predation can accurate assessments of the landscape of fear be produced.
For many species the food introduced by the experimenter in behavioral titration experiments is likely to be only a fraction of the “functional” patch through which an animal assesses its foraging environment. Measurements of the landscape of fear can, therefore, be greatly enhanced by matching the functional scale of resource manipulation to the foraging hierarchy of the focal species. For instance, the scale at which animals perceive variation in resources can be measured by testing the ability of foragers to equalize GUDs in neighboring experimental food patches with differing initial levels of food abundance (e.g., Valone and Brown 1989).
SUMMARY
This experiment clearly demonstrates that variation in resources can substantially modify the way that foraging animals display their landscape of fear. It is, therefore, imperative that researchers account for the modifying influence of natural variation in resources when designing experiments to measure how animals trade-off food for safety. The experiment and framework devised here represent an initial exploration into the simultaneous effects of variation in resources and predation risk on the behavior that foraging animals display. The results raise important questions for further study. We hope that these findings will lead to more detailed analyses of the complexities underlying behavioral expressions of landscapes of fear and habitat use. If we are to effectively confront the challenges of understanding and managing our complex changing natural environments, we need to build our predictive ability for assessing the likely outcomes of interactions between multiple ecological processes and disturbances. This will require the continued synthesis of disparate branches of ecological theory and practice.
FUNDING
Commonwealth Scientific and Industrial Research Organization (CSIRO) Sustainable Ecosystems Division, Australia.
We thank L. Labbé and B. Abbott for fieldwork assistance. Insightful reviews and discussions were provided by B. Kotler, N. Thompson Hobbs, G. Colinshaw, C. McArthur, A. Illius, J. Fryxell, R. Lawes, W. Cresswell, and one anonymous reviewer.
References
Bailey DW Gross JE Laca EA Rittenhouse LR Coughenour MB Swift DM Sims PL
Mechanisms that result in large herbivore grazing distribution patterns
,
J Range Manage
,
1996
,vol.
49
(pg.
386
-
400
)
Brown JS
Patch use as an indicator of habitat preference, predation risk, and competition
,
Behav Ecol Sociobiol
,
1988
,vol.
22
(pg.
37
-
47
)
Brown JS
Patch use under predation risk: I. Models and predictions
,
Ann Zool Fenn
,
1992
,vol.
29
(pg.
301
-
309
)
OpenURL Placeholder Text
Brown JS
Vigilance, patch use and habitat selection: foraging under predation risk
,
Evol Ecol Res
,
1999
,vol.
1
(pg.
49
-
71
)
OpenURL Placeholder Text
Brown JS Kotler BP
Hazardous duty pay and the foraging cost of predation
,
Ecol Lett
,
2004
,vol.
7
(pg.
999
-
1014
)
Brown JS Kotler BP Bouskila A
The ecology of fear and foraging games between owls and rodents
,
Ann Zool Fenn
,
2001
,vol.
38
(pg.
71
-
87
)
OpenURL Placeholder Text
Brown JS Kotler BP Mitchell WA
Competition between birds and mammals: a comparison of giving- up densities between crested larks and gerbils
,
Evol Ecol
,
1997
,vol.
11
(pg.
757
-
771
)
Brown JS Kotler BP Smith RJ Wirtz WO II
The effect of owl predation on the foraging behaviour of heteromyid rodents
,
Oecologia
,
1988
,vol.
76
(pg.
408
-
415
)
Brown JS Laundré JW Gurung M
The ecology of fear: optimal foraging, game theory, and trophic interactions
,
J Mamm
,
1999
,vol.
80
(pg.
385
-
399
)
Brown JS Morgan RA Dow DD
Patch use under predation risk: II. A test with fox squirrels, Sciurus niger
,
Ann Zool Fenn
,
1992
,vol.
29
(pg.
311
-
318
)
OpenURL Placeholder Text
Burnham KP Anderson DR
Model selection and multimodel inference: a practical information-theoretic approach
,
2002
New York
Springer
Charnov EL
Optimal foraging, the marginal value theorem
,
Theor Popul Biol
,
1976
,vol.
9
(pg.
129
-
136
)
Creel S Christianson D Liley S Winnie JA Jr
Predation risk affects reproductive physiology and demography of elk
,
Science
,
2007
,vol.
315
pg.
960
Druce DJ Brown JS Castley JG Kerley GIH Kotler BP Slotow R Knight MH
Scale-dependent foraging costs: habitat use by rock hyraxes (Procavia capensis) determined using giving-up densities
,
Oikos
,
2006
,vol.
115
(pg.
513
-
525
)
Gordon G
Australian museum: complete book of Australian mammals
,
1983
Sydney (Australia)
Angus and Robertson
Gordon G Hulbert AJ Walton DW Richardson BJ
Peramelidae
,
Fauna of Australia, Mammalia
,
1989
Canberra (Australia)
Australian Government Publishing Service
(pg.
603
-
624
)
OpenURL Placeholder Text
Hobbs NT Jung HG Fahey GC
Responses of large herbivores to spatial heterogeneity in ecosystems
,
Nutritional Ecology of Herbivores: Proceedings of the Vth International Symposium on the Nutrition of Herbivores
,
1999
Savory (IL)
American Society of Animal Science
(pg.
97
-
129
)
OpenURL Placeholder Text
Kotler BP
Patch use by gerbils in a risky environment: manipulating food and safety to test four models
,
Oikos
,
1997
,vol.
78
(pg.
274
-
282
)
Kotler BP Blaustein L
Titrating food and safety in a heterogeneous environment: when are the risky and safe patches of equal value?
,
Oikos
,
1995
,vol.
74
(pg.
251
-
258
)
Kotler BP Brown JS Dall SRX Gresser S Ganey D Bouskila A
Foraging games between gerbils and their predators: temporal dynamics of resource depletion and apprehension in gerbils
,
Evol Ecol Res
,
2002
,vol.
4
(pg.
495
-
518
)
OpenURL Placeholder Text
Kotler BP Brown JS Hasson O
Owl predation on gerbils: the role of body size, illumination, and habitat structure on rates of predation
,
Ecology
,
1991
,vol.
71
(pg.
2249
-
2260
)
Kotliar NB Wiens JA
Multiple scales of patchiness and patch structure—a hierarchical framework for the study of heterogeneity
,
Oikos
,
1990
,vol.
59
(pg.
253
-
260
)
Laundré JW Hernandez L Altendorf KB
Wolves, elk and bison: reestablishing the “landscape of fear” in Yellowstone National Park, USA
,
Can J Zool
,
2001
,vol.
79
(pg.
1401
-
1409
)
Lima SL
Putting predators back into behavioral predator-prey interactions
,
Trends Ecol Evol
,
2002
,vol.
17
(pg.
70
-
75
)
McNamara JM Green RF Olsson O
Bayes' theorem and its applications in animal behaviour
,
Oikos
,
2006
,vol.
112
(pg.
243
-
251
)
McNamara JM Houston AI
Starvation and predation as factors limiting population size
,
Ecology
,
1987
,vol.
68
(pg.
1515
-
1519
)
Mohr K Vibe-Petersen S Jeppesen LL Bildsoe M Leirs H
Foraging of multimammate mice, Mastomys natalensis, under different predation pressure: cover, patch-dependent decisions and density-dependent GUDs
,
Oikos
,
2003
,vol.
100
(pg.
459
-
468
)
Morris DW Davidson DL
Optimally foraging mice match patch use with habitat differences in fitness
,
Ecology
,
2000
,vol.
81
(pg.
2061
-
2066
)
O'Neill RV Risser PG
Hierarchy theory and global change
,
Spatial and temporal variation of biospheric and geospheric processes: research needed to determine interactions with global environmental change
,
1986
Paris
International Council of Scientific Unions Press
OpenURL Placeholder Text
O'Neill RV Deangelis DL Waide JB Allen TFH
A hierarchical concept of ecosystems
,
1986
Princeton (NJ)
Princeton University Press
Pietrzykowski E McArthur C Fitzgerald H Goodwin AN
Influence of patch characteristics on browsing of tree seedlings by mammalian herbivores
,
J Appl Ecol
,
2003
,vol.
40
(pg.
458
-
469
)
Pinaud D Wiemerskirch H
At-sea distribution and scale-dependent foraging behaviour of petrels and albatrosses: a comparative study
,
J Anim Ecol
,
2007
,vol.
76
(pg.
9
-
19
)
Pusenius J Schmidt KA
The effects of habitat manipulation on population distribution and foraging behaviour in meadow voles
,
Oikos
,
2002
,vol.
98
(pg.
251
-
262
)
Ripple WJ Beschta RL
Wolves and the ecology of fear: can predation risk structure ecosystems?
,
Bioscience
,
2004
,vol.
54
(pg.
755
-
766
)
Rosenheim JA
Top predators constrain the habitat selection games played by intermediate predators and their prey
,
Isr J Zool
,
2004
,vol.
50
(pg.
129
-
138
)
SAS Institute Inc
,
SAS release 8.02
,
2001
Cary (NC)
SAS Institute Inc
OpenURL Placeholder Text
Schmitz OJ
Scaling from plot experiments to landscapes: studying grasshoppers to inform forest ecosystem management
,
Oecologia
,
2005
,vol.
145
(pg.
225
-
234
)
Searle KR Vandervelde T Hobbs NT Shipley LA Wunder BA
Spatial context influences patch residence time in foraging hierarchies
,
Oecologia
,
2006
,vol.
148
(pg.
710
-
719
)
Senft RL Coughenour MB Bailey DW Rittenhouse LR Sala OE Swift DM
Large herbivore foraging and ecological hierarchies
,
Bioscience
,
1987
,vol.
37
(pg.
789
-
799
)
Sih A
Optimal foraging: partial consumption of prey
,
Am Nat
,
1980
,vol.
116
(pg.
281
-
290
)
Sih A Dugatkin LA Reeve HK
Game theory and predator-prey response races
,
Game theory and animal behaviour
,
1998
New York
Oxford University Press
(pg.
221
-
238
)
OpenURL Placeholder Text
Southgate R Palmer C Adams M Masters P Triggs B Woinarski J
Population and habitat characteristics of the golden bandicoot (Isoodon auratus) on Marchinbar Island, Northern Territory
,
Wildl Res
,
1996
,vol.
23
(pg.
647
-
664
)
Valone TJ
Are animals capable of Bayesian updating? An empirical review
,
Oikos
,
2006
,vol.
112
(pg.
252
-
259
)
Valone TJ Brown JS
Measuring patch assessment abilities of desert granivores
,
Ecology
,
1989
,vol.
70
(pg.
1800
-
1810
)
Wirsing AJ Heithaus MR Dill LM
Living on the edge: dugongs prefer to forage in microhabitats that allow escape from rather than avoidance of predators
,
Anim Behav
,
2007
,vol.
74
(pg.
93
-
101
)
Yunger JA Meserve PL Gutiérrez JR
Small-mammal foraging behaviour: mechanisms for coexistence and implication for population dynamics
,
Ecol Monogr
,
2002
,vol.
72
(pg.
561
-
577
)
Zhang ZQ Sanderson JP
Spatial scale of aggregation in 3 acarine predator species with different degrees of polyphagy
,
Oecologia
,
1993
,vol.
96
(pg.
24
-
31
)
© The Author 2008. Published by Oxford University Press on behalf of the International Society for Behavioral Ecology. All rights reserved. For permissions, please e-mail: journals.permissions@oxfordjournals.org
Issue Section:
Articles
Download all slides
Advertisem*nt
Citations
Views
2,157
Altmetric
More metrics information
Metrics
Total Views 2,157
1,573 Pageviews
584 PDF Downloads
Since 12/1/2016
Month: | Total Views: |
---|---|
December 2016 | 6 |
January 2017 | 6 |
February 2017 | 11 |
March 2017 | 12 |
April 2017 | 6 |
May 2017 | 14 |
June 2017 | 2 |
July 2017 | 2 |
August 2017 | 3 |
September 2017 | 11 |
October 2017 | 11 |
November 2017 | 6 |
December 2017 | 9 |
January 2018 | 16 |
February 2018 | 21 |
March 2018 | 21 |
April 2018 | 17 |
May 2018 | 15 |
June 2018 | 33 |
July 2018 | 18 |
August 2018 | 17 |
September 2018 | 13 |
October 2018 | 29 |
November 2018 | 30 |
December 2018 | 23 |
January 2019 | 19 |
February 2019 | 29 |
March 2019 | 35 |
April 2019 | 45 |
May 2019 | 23 |
June 2019 | 25 |
July 2019 | 10 |
August 2019 | 32 |
September 2019 | 28 |
October 2019 | 30 |
November 2019 | 28 |
December 2019 | 24 |
January 2020 | 21 |
February 2020 | 22 |
March 2020 | 21 |
April 2020 | 17 |
May 2020 | 26 |
June 2020 | 30 |
July 2020 | 17 |
August 2020 | 17 |
September 2020 | 15 |
October 2020 | 18 |
November 2020 | 15 |
December 2020 | 29 |
January 2021 | 19 |
February 2021 | 38 |
March 2021 | 30 |
April 2021 | 24 |
May 2021 | 30 |
June 2021 | 20 |
July 2021 | 13 |
August 2021 | 16 |
September 2021 | 9 |
October 2021 | 46 |
November 2021 | 23 |
December 2021 | 18 |
January 2022 | 17 |
February 2022 | 14 |
March 2022 | 21 |
April 2022 | 42 |
May 2022 | 15 |
June 2022 | 32 |
July 2022 | 31 |
August 2022 | 30 |
September 2022 | 33 |
October 2022 | 29 |
November 2022 | 31 |
December 2022 | 16 |
January 2023 | 32 |
February 2023 | 16 |
March 2023 | 38 |
April 2023 | 23 |
May 2023 | 24 |
June 2023 | 12 |
July 2023 | 9 |
August 2023 | 29 |
September 2023 | 17 |
October 2023 | 59 |
November 2023 | 65 |
December 2023 | 68 |
January 2024 | 44 |
February 2024 | 34 |
March 2024 | 33 |
April 2024 | 129 |
Altmetrics
Email alerts
Article activity alert
Advance article alerts
New issue alert
In progress issue alert
Receive exclusive offers and updates from Oxford Academic
Citing articles via
Google Scholar
-
Latest
-
Most Read
-
Most Cited
More from Oxford Academic
Animal Behaviour and Behavioural Ecology
Biological Sciences
Science and Mathematics
Zoology and Animal Sciences
Books
Journals
Advertisem*nt