Testing the stability of behavioural coping style across stress contexts in the Trinidadian guppy

1. Within populations, individuals can vary in stress response, a multivariate phenomenon comprising neuroendocrine, physiological and behavioural traits. 2. Verbal models of individual stress “coping style” have proposed that the behavioural component of this variation can be described as a single axis, with each individual's coping style being consistent across time and stress contexts. 3. Focusing on this behavioural component of stress response and combining repeated measures of multiple traits with a novel multivariate modelling framework, we test for the existence of coping style variation and assess its stability across contexts in the Trinidadian guppy (Poecilia reticulata). 4. Specifically, we test the following hypotheses: (1) there exists repeatable among‐individual behavioural (co)variation (“personality”) within a mild stress context consistent with a risk‐averse–risk‐prone continuum of behavioural coping style, (2) there is population‐level plasticity in behaviour as a function of stressor severity, (3) there is among‐individual variation in plasticity (i.e. IxE), and (4) the presence of IxE reduces cross‐context stability of behavioural coping style. 5. We found significant repeatable among‐individual behavioural (co)variation in the mild stress context (open field trial), represented as an I matrix. However, I was not readily described by a simple risk‐averse–risk‐prone continuum as posited by the original coping style model. We also found strong evidence for population‐level changes in mean behaviour with increasing stressor severity (simulated avian and piscine predation risks). 6. Single‐trait analyses did show the presence of individual‐by‐environment interactions (IxE), as among‐individual cross‐context correlations were significantly less than +1. However, multitrait analysis revealed the consequences of this plasticity variation were minimal. Specifically, we found little evidence for changes in the structure of I between mild and moderate stress contexts overall, and only minor changes between the two moderate contexts (avian vs. piscine predator). 7. We show that a multivariate approach to assessing changes in among‐individual (co)variance across contexts can prevent the over‐interpretation of statistically significant, but small, individual‐by‐environment effects. While behavioural flexibility enables populations (and individuals) to respond rapidly to changes in the environment, multivariate personality structure can be conserved strongly across such contexts. A plain language summary is available for this article.