Complex strategies: an integrative analysis of contests in Siamese fighting fish

Animals use contests to attain resources and employ strategic decisions to minimise contest costs. These decisions are defined by behavioural response to resource value and competitive ability, but remain poorly understood. This is because the two factors are typically studied separately. Also, their study relies on overgeneralised assumptions that (i) strategies are fixed, (ii) modulated by the motivation or drive to fight and (iii) used to manage costs proportional to the timing of the loser's retreat. To address these problems, we adopt an integrative sequential analysis that incorporates competitive ability and resource value factors, to characterise territorial contest decisions in male Siamese fighting fish (Betta splendens).

Individuals exhibited a chronological organisation of behaviour, engaging opponents first with frontal display, then switching to lateral display before deciding to attack, and reserved retreats for later stages. Using asymmetries in retreats as a proxy for outcome, the likelihood of winning was found to be mostly dependent on display. However, resource and contest conditions affected initiation latency, display, attack and retreat, suggesting that strategic decisions influence all behaviour. Overall, sequential behaviour varied consistently with individual aggressiveness and resource-value factors, and increasingly with information on competitive ability collected during the contest. This enabled shifts in tactics, such as disadvantaged individuals responding first with aggression and later with submission. Motivation to continue fighting, after interruption by startle, was also adjusted to information gathered during the contest and progressively with energetic state. Two clusters of correlated behaviours were identified, cost-mitigation (display and retreat) and escalation (initiation and attack), but changes in motivation were associated only with cost mitigation.

Our findings contrast dominant assumptions that strategic decisions are fixed, controlled by motivational state and sufficiently described by outcome-dependent measures. We instead demonstrate that strategic decisions are complex, comprising functional changes in assessment, information use and motivational effects, which are not always inter-dependent.