BEHAVIORAL IDENTIFICATION OF AN INSIGHTFUL MOMENT IN TOLMAN’S MAZE

Sudden and discontinuous changes in problem-solving behavior, often interpreted as evidence of “insight,” have long been described in humans and were demonstrated in animals in the classic maze experiments of Tolman and Honzik. Such transitions are thought to reflect a reorganization of how task information is represented and used. However, the mechanistic nature of insight remains elusive, in part because behavioral measures often lack the resolution needed to identify when such reorganizations occur.
Here, we aim to identify the precise moment of insight. We define insight as the discovery of an optimal problem-solving strategy that is not explicit in the task, emerging from the sudden integration of previously underweighted variables. To study this process, we developed a behavioral re-implementation of Tolman’s maze and examined insight-like problem solving in rats, combined with AI-based behavioral analysis to quantify trial-by-trial spatial and sequential features across task experience.
This approach revealed behavioral signatures that distinguish insight-like strategy discovery. High-performing rats flexibly departed from previously learned routes when task demands changed, whereas low-performing rats largely perseverated on prior movement patterns. Notably, increased time spent at choice points emerged near the formation of a stable strategy. Together, these findings provide an operational framework for temporally localizing insight-like cognitive transitions, enabling future work to link discrete behavioral shifts to neural dynamics underlying cognitive flexibility and problem solving.