Inducible defence strategies evolve in response to temporally and spatially variable predation risk. Selection should favour the expression of these strategies when risk crosses a threshold, but expression may also depend on how effective a strategy is across contexts. For neotropical harvestmen, aggregation behaviour is a group-level defence strategy, but can be supplemented by additional behaviour to enhance antipredator defence. A secondary behavioural strategy termed “bobbing”, in which individuals rapidly move their bodies up and down, has been proposed to confuse or deter predation as the group moves en masse. While this behaviour may be performed in isolation, we propose that its net pay-off may depend on the presence of conspecifics and group size. Therefore, due to the benefits of plasticity in a behaviour with such an apparent context-dependent pay-off, we hypothesize that the aggregation size of neotropical harvestmen affects the propensity of individuals to engage in the antipredator bobbing defence. We predicted that individuals would be more likely to perform bobbing in aggregations as opposed to in isolation. We collected harvestmen (Prionostemma sp.) at La Selva Biological Research Station, Costa Rica, and exposed them to a simulated predator cue when housed within experimental chambers. Individuals were more likely to engage in bobbing when in a group than alone. Moreover, we performed field observations that found bobbing to be a more frequent response as group size increased. Our results show social context-dependence of an inducible antipredator defence behaviour that aligns with proposed costs and benefits of its expression.

Lab-based experiments were conducted to determine whether the social context of harvestmen (i.e., isolation or aggregation) affects the propensity of individuals to perform antipredator bobbing following a predator simulation. As control experiments for potential confounding factors from the isolation versus aggregation experiment, subsequent trials were conducted to determine whether the behaviour was affected by sex and density within an enclosure. In addition to the lab-based experiments, field observations were carried out to assess the propensity of harvestmen to exhibit bobbing as affected by group size.

Analyses were performed in R v 4.3.2 (Team, 2020). Generalized linear and mixed effects models with the lme4 (Bates et al., 2015) and glmmTMB (Brooks et al., 2017) packages were used to analyse the bobbing behaviour during experimental trials. The propensity of individuals to exhibit bobbing in experiment 1 and experiment 2 was fitted using a generalized linear mixed effects model with a binomial distribution {bobbing ~ treatment + (1 | block:ID)}, where the response is explained by the treatment (i.e., aggregation or sex), with container ID nested within block included as a random effect. As individuals were observed alone in the density experiment, a generalized linear model with a binomial distribution was used for the analysis. Estimated marginal means, odds ratios, and associated 95% confidence intervals were extracted from all models using the package emmeans (Lenth et al., 2020).

To analyse the field observations of the occurrence of bobbing behaviour, a generalized additive model with a binomial distribution was fit using the package gam (Hastie & Tibshirani, 1986). An initial model was fit including location, date, and the number of individual harvestmen observed at a tree. The number of harvestmen included a smooth function for non-linear fitting. Non-significant terms were removed from the final model. For the data used in the model, any aggregation with the number of individuals recorded as greater than 25, due to the inability to accurately determine the exact number, was given a value of 25.