, Volume 9, Issue 1, pp 70–74

The Ecology of Fear: Host Foraging Behavior Varies with the Spatio-temporal Abundance of a Dominant Ectoparasite

Short Communication

DOI: 10.1007/s10393-012-0744-z

Cite this article as:
Fritzsche, A. & Allan, B.F. EcoHealth (2012) 9: 70. doi:10.1007/s10393-012-0744-z


Prey engage in myriad behaviors to avoid predation, and these indirect effects of predators on their prey are often measured by the amount of food abandoned by a forager (the “giving-up density”, or GUD) in a given habitat. Recent evidence suggests that hosts may engage in comparable behaviors to avoid exposure to parasites. We investigated changes in local foraging and regional space use by mammal hosts for the lone star tick (Amblyomma americanum), using GUDs as an indicator of the perceived risk of parasitism. At eight study sites at the Tyson Research Center (Eureka, MO), we placed two feeding trays, one on the ground and one at 1.5 m height in a tree, in order to assess how the emergence of ground-dwelling ticks affected foraging by several mammal species both locally (between the two GUD stations) and regionally (among the eight sites, mean distance 1064 m apart). Though GUDs did not differ between the ground and tree GUD stations, we did find that greater amounts of food were “given-up” at sites with higher abundances of ticks. This increase in food abandonment suggests that hosts respond to the risk of parasitism and alter their space use accordingly, potentially affecting a cascade of other ecological interactions across large spatial scales.


host–parasite interactionsticksgiving-up density (GUD)ecology of feartick-borne diseases


The mere presence of predators has been shown to elicit fear-based behavioral changes in foraging, space use, and allocation of time towards vigilance in prey species, a phenomenon referred to as “the ecology of fear” (Pekarsky et al. 2008; Kotler et al. 2010; Laundré 2010). These indirect, nonconsumptive effects of predators can be as ecologically important as consumption because the changes in prey behavior can scale up to affect community and ecosystem structure (Ripple and Beschta 2004; Mouritsen and Poulin 2005). In light of the recent conceptualization of parasites playing ecological roles similar in magnitude to predators (Raffel et al. 2008), it seems a natural extension to incorporate the interactions between hosts and parasites into the framework of the ecology of fear (Rohr et al. 2009). Although parasites often do not cause immediate mortality as do predators, they do impose physiological costs as well as long-term reductions in fitness, and there can be strong selection pressure on hosts to reduce their exposure to parasites (James et al. 2008; Witter et al. 2011).

Ecologists studying the ecology of fear in a predator–prey context have cleverly used tools derived from optimal foraging theory (MacArthur and Pianka 1966) to assess how prey balance the costs and benefits of activities such as foraging or social interactions, which will have different optima based on varying levels of predation risk and the allocation of effort towards vigilance (Brown et al. 1999; Jacob and Brown 2000). A forager is predicted to abandon a resource patch when the costs due to predation exposure outweigh the benefits of continued foraging (Brown et al. 1999). This measure of food abandonment, or “giving-up density” (hereafter GUD), quantifies the tradeoff of foraging benefits against the costs of predation risk (MacArthur and Pianka 1966). Although this framework has become well-established in predator–prey ecology, it has only recently been used to assess the ecology of fear in host-parasite interactions (Allan et al. 2010a).

In Missouri, lone star ticks (Amblyomma americanum) are an abundant ectoparasite of wildlife and feed on a variety of vertebrate hosts (Allan et al. 2010b). The larval life-stage, upon its emergence in mid- to late-summer, heavily parasitizes small mammals such as gray and fox squirrels (Sciurus carolinensis and S. niger), and the common raccoon (Procyon lotor) (Kollars et al. 2000; Allan et al. 2010b, Goessling et al. in press). We used this host-parasite system to assess the effects of escalating parasite pressure on host behavioral responses, measured via GUDs. This system is particularly amenable to the study of parasite-mediated behavioral effects within the framework of optimal foraging theory because: (1) the density of larval A. americanum increases during the period of emergence, thereby creating a temporal gradient in risk of parasitism (Kollars et al. 2000), (2) fitness consequences of ectoparasite exposure may be considerable, either due to exposure to vectored pathogens or the direct effects of substantial ectoparasite burdens such as anemia (Allan et al. 2010a), (3) important mammalian hosts are easily attracted to novel food resources after minimal adjustment time, and (4) these mammals are heavily parasitized by larval A. americanum, increasing the probability that this level of exposure results in a detectable change of behavior in response to the threat of parasitism. Since larval ticks are confined spatially to the ground, we established pairs of feeders (hereafter GUD stations) on the ground and in a nearby tree in order to detect changes in foraging location by hosts on a relatively small spatial scale. We further established GUD station pairs across a spatial gradient in tick density. We hypothesized that (1) locally, foraging on the ground would decrease over the study period as larval ticks emerged, and (2) regionally (on the scale of kilometers), greater amounts of food would be abandoned at sites with high tick densities.

Materials and Methods

Our study was conducted at the Tyson Research Center (Eureka, MO), a ~ 809 ha biological field station comprised primarily of oak-hickory upland deciduous forests. We established eight replicate locations (mean ± S.E.: 1,065 ± 503 m apart), all in highly similar closed-canopy oak-hickory forest, to serve as sites for pairs of GUD stations. GUD stations were 40 × 40 × 6 cm plywood trays affixed either to a tree at ~1.5 m of height, or placed on the ground 5–8 m from the tree. GUD stations were baited with 150 g of sunflower seed mixed into 1.5 kg of sand, an inedible substrate causing diminishing returns in a depletable food patch, an essential component of a GUD study (Jacob and Brown 2000). Trays were baited daily between 0900 and 1200 h for a period of 3 days; every ~24 h, seed was filtered from the sand, weighed, and trays were replenished with another 150 g of seed. GUD station trays remained un-baited during the 4 days between data collection periods. We calculated a GUD estimate for each station at each site as the mean weight of seed remaining after each of the 3 days, yielding 4 weeks of GUD estimates per station between 25 June and 26 July, 2010.

Within a day after each 3-day series of GUD data collection, we measured tick density at each site using the drag-sampling technique (Schulze et al. 1997). We dragged a 1 × 1 m canvas cloth along a 60 m transect established 40 m from each pair of GUD stations and removed ticks every 20 m for counting and identification in the laboratory. To prevent resampling bias in subsequent surveys, we sampled parallel transects 1–2 m away from the original transect.

We analyzed the effects of resource location (ground or tree GUD station), log-transformed larval tick density, and time (week number) on host GUDs using an Analysis of Covariance (ANCOVA). ANCOVA results were qualitatively similar before and after removal of three log-transformed tick density samples considered statistical outliers; here we present the results without outlier removal. By the fourth week of our study all eight sites had non-zero tick densities, so we also analyzed the association between tick densities and GUDs by ANCOVA and linear regression for Week 4 alone. To account for potential spatial autocorrelation among sites, we calculated pair-wise distance matrices representing the linear distance between sites and conducted a partial Mantel test. ANCOVA analyses were conducted using SAS version 9.2 (SAS Institute Inc., Cary, NC), and the partial Mantel test was conducted using the “ecodist” package in R (R Core Development Team 2007).

Our GUD stations were not designed either to target or to exclude certain species. Therefore, to assess qualitatively which host species were feeding at GUD stations, we set out two infrared beam-triggered digital camera traps at two randomly selected sites at the end of our study. Camera traps were deployed for 72 h/site, during which seed was replenished every 24 h.


Tick numbers varied greatly among the eight study sites and across the 4 weeks of the study. The mean (± S.E.) tick density per 60 m2 across all sites increased from 0, to 3.1 ± 2.7, to 24.3 ± 13.9, to 109.3 ± 79.8 from Weeks 1–4, respectively. By Week 4, the gradient of tick density among sites ranged from 1 to 667 ticks per 60 m2 (mean 113.8 ± 80.1).

The amount of seed abandoned by hosts was significantly affected by week (F3,56 = 8.47, P < 0.0001; Table 1), such that less seed, on average, was abandoned as the study period progressed. Resource location was not a significant predictor of GUDs (F1,56 = 0.22, P = 0.64; Table 1), indicating hosts removed seeds equally from the ground and tree GUD stations. Although there were no differences between the two GUD stations within each site, GUDs were strongly affected by variation in tick densities among sites (F1,56 = 10.71, P = 0.002; Table 1). In Week 4 in particular, there was a strong positive relationship between tick abundance and host GUDs (F1,13 = 9.77, P = 0.008, r2 = 0.43 for both resource locations combined; Table 2; Fig. 1) meaning that hosts abandoned significantly more seed at sites with high tick densities. This effect remained after controlling for spatial proximity among the sites (Mantel’s r = 0.57, P = 0.043), suggesting the positive correlation between tick abundance and host GUDs was not a spurious result due to spatial autocorrelation.
Table 1

Results of full-model ANCOVA analysis of factors determining giving-up densities (GUDs) of tick hosts.



F ratio


P value

Resource location










Tick density





Week * tick density










Effects sums-of-squares are Type III. P values were accepted as significant (*) at P < 0.05.

Table 2

Results of ANCOVA analysis for Week 4 of the study.



F ratio


P value

Resource location





Tick density










Effects sums-of-squares are Type III. P values were accepted as significant (*) at P < 0.05.
Figure 1

Effects of tick density and resource location (tree or ground GUD station) on food abandonment by hosts for the tick Amblyomma americanum for Week 4 of the study. Data for ground (filled circle) and tree (open circle) GUD stations are shown separately, but the linear regression trendline shows the effect of larval tick abundance on GUD at both resource locations combined. Resource location did not significantly affect GUD (P = 0.89) but GUD was significantly related to tick density (P = 0.008, r2 = 0.43).

The camera traps we deployed at the end of the study at two of the study sites yielded a total of 875 images containing an identifiable animal foraging from the GUD stations. A total of four species were identified in the images: common raccoons (65% of images), gray squirrels (31%), Peromyscus spp. mice (2%) and eastern cottontail rabbits, Sylvilagus floridanus (2%). Both of the dominant foragers (raccoons and squirrels) were observed to forage from both the ground and tree GUD stations. It is important to emphasize that when we discuss the response variable (GUD) below, we are referring to a response to tick parasitism for several host species combined.


Our results did not support the hypothesis that hosts for A. americanum would abandon more food at ground GUD stations as the seasonal emergence of larval A. americanum increased risk of parasite exposure. However, we did find that GUDs increased with larval tick density across a larger spatial gradient as larvae became more prevalent, yet still patchy across the landscape, suggesting that hosts may recognize the threat of parasitism in their environment and adjust their large-scale space use accordingly. We cannot discount that other factors (e.g., predation risk, food availability) that could potentially covary with tick abundance may have influenced food abandonment at our study sites, potentially leading to a spurious correlation. However, the increasing abandonment of food at sites of high tick abundance as tick populations became more ubiquitous is strongly supportive of our hypothesis.

The literature on behavioral mechanisms of parasite avoidance has historically focused on very local behaviors, such as grooming, “nest fumigation”, grouping, or avoidance of feces-contaminated foods (Hart 1990). Lozano (1991) recognized a link between parasites and optimal foraging, though focus was placed on how dietary choices impact parasite exposure. Recently, harassment by parasitic flies was found to substantially alter the behavioral time-budgets of Canadian reindeer, such that time spent feeding was strongly decreased in the presence of high parasite intensities (Witter et al. 2011). Local-scale behavioral avoidance of ticks was also noted by Hart (1990), who observed that cattle avoided paddocks seeded with tick larvae. At spatial scales comparable to this study, Allan et al. (2010a) found that white-tailed deer (Odocoileus virginianus), also a host for larval A. americanum, abandoned more food at locations with high densities of ticks. And at much larger spatial scales, Altizer et al. (2011) recently proposed the “migratory escape hypothesis”, which suggests that host animals can escape their parasites by way of large-scale movements or migrations. Our findings complement these results by demonstrating that a wide breadth of hosts may respond behaviorally to the risk of parasitism, and our empirical test of multiple spatial scales indicates that different outcomes may be observed at different scales of inquiry.

Our choice of system also affords the opportunity to consider the implications of our results for vector-borne disease dynamics. Members of the genus Sciurus are zoonotic reservoir hosts for several bacteria vectored by A. americanum in the genera Borrelia and Ehrlichia, which cause diseases such as Ehrlichiosis in humans (Allan et al. 2010b, Goessling et al. in press), though the impacts of these pathogens on wildlife health are less well known. Importantly, altered space use by host species that are competent pathogen reservoirs may affect pathogen prevalence and associated risk to human health. Including these interactions in spatio-temporal models of pathogen transmission will facilitate an improved understanding of tick-borne disease dynamics. We advocate for the continued use of manipulative experiments to further quantify the impact of parasites on host behavioral ecology, especially when such interactions involve vectors and reservoir hosts for pathogens that imperil human health.


We thank B. Decker, C. Hilliard, A. Narla, J. Pivor and TERF interns for assistance with research, Tyson Research Center staff for logistical support, and K. Smith and J. Chase for intellectual feedback and guidance. Comments from the S. Altizer and V. Ezenwa lab groups, C. Nunn, and two anonymous reviewers greatly improved the manuscript. Funding was provided by Environmental Protection Agency grant 834495 to J.C. and B.F.A.

Copyright information

© International Association for Ecology and Health 2012

Authors and Affiliations

  1. 1.Tyson Research CenterWashington University in St. LouisEurekaUSA
  2. 2.Odum School of EcologyUniversity of GeorgiaAthensUSA
  3. 3.Department of EntomologyUniversity of Illinois Urbana-ChampaignUrbanaUSA