Most organisms are at risk of being consumed by a predator or infected by a parasite at some point in their life. Theoretical constructs such as the landscape of fear (perception of risk) and non-consumptive effects (NCEs, costly responses sans predation or infection) have been proposed to describe and quantify anti-predator and anti-parasite responses. How prey/host species identify and respond to these risks determines their survival, reproductive success and, ultimately, fitness. Most studies to date have focused on either predator-prey or parasite-host interactions, yet habitats and ecosystems contain both parasitic and/or predatory species that represent a complex and heterogenous mosaic of risk factors. Here, we experimentally investigated the behavioural responses of a cactophilic fruit fly, Drosophila nigrospiracula, exposed to a range of species that include parasites (ectoparasitic mite), predators (jumping spiders), as well harmless heterospecifics (non-parasitic mites, ants and weevils). We demonstrate that D. nigrospiracula can differentiate between threat and non-threat species, increase erratic movements and decrease velocity in the presence of parasites, but decrease erratic movements and time spent grooming in the presence of predators. Of particular importance, flies could distinguish between parasitic female mites and non-parasitic male mites of the same species, and respond accordingly. We also show that the direction of these non-consumptive effects differ when exposed to parasitic mites (i.e., risk of infection) versus spiders (i.e., risk of predation). Given the opposing effects of predation versus infection risk on fly behaviour, we discuss potential trade-offs between parasite and predator avoidance behaviours. Our findings illustrate the complexity of risk assessment in a landscape of fear and the fine-tuned non-consumptive effects that arise in response. Moreover, this study is the first to examine these behavioural NCEs in a terrestrial system.

Video analysis: behavioural assays were documented using digital video recorded in a DanioVision^TM observation chamber (Noldus Information Technology, Leesburg, VA, USA) and response variables were measured with EthoVision^TM computer tracking software (Noldus Information Technology, Leesburg, VA, USA). In total, we generated 286 videos of fly behaviour in response to male mites (n=30), female mites (n=30), ant (n=30), weevils (n=30), spider (n=22), control (n=144); flies were only tested once. Once the arenas were placed in the DanioVision^TM chamber, 5 minutes of unused video was recorded before the five-minute experimental recording period. The experimental five-minute video was analysed using Ethovision^TM software that recorded all response variables 12.5 times per second. Response variables reported were velocity (mm/s), distance travelled (mm), ambulatory movement (yes/no), distance between fly and cage (mm), and meander (the number of degrees through which the fly changed direction for each mm traveled - °/mm) where an increase in the meander value corresponds to an increase in erratic movement or path complexity (Richardson et al. 2018).

We manually calibrated the raw data such that all samples displaying velocity below 1.5mm/s, and all corresponding response variables (distance, ambulatory movement, meander), were counted as zero (distance and meander) or no movement because below this threshold EthoVision^TM often registered false positives. Average velocity and meander values for each fly were recalculated excluding all zero scores, essentially recreating the variables as velocity or meander while mobile. These adjustments were made because most flies spent the majority of the five-minute observational period immobile, and the large number of zeros in velocity or meander caused unwanted smoothing in the data and masked actual differences among treatment groups. Some flies remained immobile for the entire observation period and so were removed from analysis; these flies were all active and groomed periodically, but none moved sufficiently to record velocity or meander measurements. A generalised linear model with a binomial error distribution and a logit link function was used to test for any effect of stimuli species on the proportion of immobile flies, but none was found (see Appendix1: Table S2). Time spent grooming was recorded via direct observation by a single observer who noted time spent grooming with a stopwatch; any flies that spent >15s (5% of video) obscured from observer were excluded from analysis. Unfortunately, our stimuli species were sporadically available during the 20 weeks we ran video analysis: male and female mites were maintained in-lab, although availability was highly variable; weevils were provided by an outside lab on an ad hoc basis; while ants and spiders were collected from the field. Consequently, stimuli species were not evenly distributed throughout the trials (see Appendix S1: Table S3). To ensure that small variations in RH and temperature during acclimation and recording did not have a confounding effect on fly behaviour, we ran a series of models on all response variables using only the control video data, as no other treatment was present in all trials (see Statistical Analysis section and Appendix S1: Table S2).