Ants Learn a Maze Fast but Forget It Within Days

Worker ants of the species Dinoponera quadriceps can learn to navigate a visual maze to reach food, but they fail to show that learning when tested hours or days later, according to a new open access study published in the journal Animal Cognition. The findings suggest that in these solitary foraging ants, learned route information may be rapidly lost or suppressed rather than retained as lasting memory.

The research team, led by Igor Eloi and colleagues, tested associative learning, memory retention and individual learning profiles using a bifurcated maze with visual cues. Eighty-two workers drawn from ten colonies each underwent four days of individual training to locate a food reward placed in one arm of the maze. After training, each ant received an unrewarded retention test at one of three intervals: 24, 48 or 72 hours.

Thirty-eight of the 82 workers were classified as learners, defined by a consistent reduction in the time they took to find the reward across training sessions. Performance improved significantly by the final training session, and the learners were significantly faster than non-learners in their average latency to reach the reward. This indicates that the learner group was not simply composed of individuals with poorer overall foraging ability.

An unexpected pattern emerged when the researchers examined daily foraging activity. They found a significant negative correlation between an individual’s foraging activity and its learning performance, suggesting an activity-learning trade-off in which the most active foragers were slower to master the maze task.

The critical result concerned memory. Performance across all retention tests, from 24 hours to 72 hours after training, was static and statistically equivalent to the ants’ initial, naive performance in the very first training trial. In other words, although the workers had acquired a visually guided route during training, the behavioral expression of that learned information was not maintained in delayed unrewarded tests.

The authors propose several possible explanations for this retrieval deficit. It may reflect rapid forgetting, extinction-like responding in the absence of reward, retrieval failure, or interference from experiences after training. They also note that the pattern at intermediate intervals is consistent with non-monotonic retention, a shape of memory curve in which performance dips before potentially recovering.

From an ecological perspective, the researchers suggest that such a retrieval deficit could constitute an adaptation that prioritizes up-to-the-minute information. D. quadriceps is a ground-dwelling ant of dynamic forest habitats, where conditions on the forest floor can change quickly, making recently acquired information more valuable than older memories.

The study also highlights substantial individual variation in learning performance within a single species. The authors conclude that memory expression in solitary foraging ants may be strongly shaped both by ecological context and by individual foraging profiles, rather than following a uniform pattern across all workers.

Because the article is being shared early as a peer-reviewed, accepted version of record pending final edits, some details may change in the final published version. Even so, the central observation stands out: these ants can learn, but the learned route does not persist in delayed tests, a combination the authors summarize as associative learning without lasting memory.

Subject of Research: Associative learning and memory retention in the solitary foraging ant Dinoponera quadriceps

Article Title: Associative learning without lasting memory: a recency bias in Dinoponera quadriceps

Article References: Eloi, I., de Lima, M. L. C., Luchiari, A. C., & Araújo, A. (2026). Associative learning without lasting memory: a recency bias in Dinoponera quadriceps. Animal Cognition. https://doi.org/10.1007/s10071-026-02120-x

Image Credits: AI Generated

DOI: 10.1007/s10071-026-02120-x

Keywords: Dinoponera quadriceps, ants, associative learning, memory retention, maze navigation, foraging behavior, individual variation, Animal Cognition, insect cognition, recency bias, Associative, learning