A novel method to measure the impact of water quality on judgement bias in wild juvenile fish

Methods to examine judgement bias in free-living animals in situ are required to address ecological, conservation and animal welfare questions. Wild animals make behavioural decisions based on complex information, and judgement bias is an experience-induced adjustment in the cognitive appraisal of ambiguous information. Following on from recent research showing judgement bias in fish, we developed a novel approach to measure population-level judgement bias using the natural tendency of juvenile Murray cod ( Maccullochella peelii) ) to approach a light- source, but move away from larger, potentially predator fish. Population-level judgement bias was determined from the number of Murray cod caught in three slightly different light traps containing; 1) a light-only (positive stimulus), 2) a predator-model (negative stimulus) and 3) an ambiguous-model (ambiguous stimulus). Ten water quality parameters were also recorded at each site. All combinations of water quality parameters were included in models to examine how well they explained (i) the presence of Murray cod and (ii) in sites where Murray cod were present the population-level judgement bias. The best models were selected using Akaike's Information Criterion. We caught 113 Murray cod at 19 out of 33 sites and modelling highlighted the importance of dissolved oxygen (P=0.02-0.05 =0.02-0.05 in top logistic models) to explain presence/absence of fish, confirming the threat of low dissolved oxygen for this species. More Murray cod were caught in light-only (positive stimulus) traps than in predator-model (negative stimulus) traps (P=0.04). =0.04). Population-level judgement bias was overall negative, indicating a general tendency to avoid the ambiguous-model light trap. The top linear model (AICc=57.71, =57.71, R2=0.63, 2 =0.63, P=0.025) =0.025) indicated that in combination, there was greater avoidance of the ambiguous stimulus (i.e. a more pessimistic response) as salinity (P=0.043) =0.043) and filterable reactive phosphorous increased (P=0.055) =0.055) and pH decreased (P=0.013). =0.013). The above water quality parameters were not near known lethal levels, indicating a need to better understand the sub-lethal effects of water parameters on fish behaviour and physiology. Our findings indicate that methods to measure population-level judgement bias can support research on the function of judgement bias and its possible relation to affect in fish. More generally, the method provides a potentially useful tool to bring together conservation biology and animal welfare disciplines.