We investigated whether mice can discriminate static dot arrays differing in relative numerosity and whether stimulus magnitude influences side choice. In Experiment 1, mice were trained in a touchscreen-equipped operant-conditioning chamber to select the larger of two dot arrays (8 vs. 4) and then completed reinforced test blocks that introduced untrained numerosity pairs and stimulus manipulations of continuous physical variables. Mice showed evidence of transfer to novel comparisons, and accuracy varied systematically with the numerical ratio, consistent with Weber-like scaling. Accuracy remained above chance with open-circle displays, but performance declined when cumulative surface area was equated and when discrete item structure was disrupted by phase scrambling. These findings indicate that mice can learn and generalize relative numerosity discriminations, although their performance was also influenced by correlated continuous visual properties. Thus, the results do not establish stimulus control by numerosity independently of these properties. In Experiment 2, we examined whether stimulus magnitude influenced side choice under noncontingent probe conditions, in which identical dot arrays were presented simultaneously on the left and right sides of the screen. Under these conditions, the probability of choosing the left side increased with numerosity at the group level. Together, these results demonstrate that mice can learn and generalize discriminations between static dot arrays and provide limited evidence of a magnitude-related bias in side choice.
Our study examined whether reward magnitude affects long-term spatial memory retention in rats performing a radial-arm maze task. Previous studies, using a forced-choice learning phase followed by a free-choice test, have reported retention of up to 72 hours, but they typically used a large reward (10 pellets per arm; 45 mg each) rather than the standard reward (one pellet per arm), leaving open a reward-based explanation. In Experiment 1, we conducted a between-subjects comparison of two reward conditions: a standard-reward group (one pellet per arm) and a large-reward group (10 pellets per arm). The delay interval was progressively extended from 1 min to 1 hr, 24 hr, 48 hr, and 72 hr. Both groups performed significantly above chance at all delays, including 72 hr, and performance did not differ significantly between reward conditions. In Experiment 2, we tested whether relative changes in reward magnitude modulate retention. Within each reward group, reward magnitude was increased or decreased for only one arm during the learning phase, followed by a 24-hr retention test. Accuracy for the arm with the changed reward did not differ significantly from that of the unchanged arms. Together, these findings show that 72-hr retention can be obtained, even with the standard small reward (one pellet), and that radial-arm maze performance is robust to both absolute reward magnitude and relative changes in reward.
The present study examined the effects of varying the events occurring during the delay interval (DI) and the intertrial interval (ITI) on asymmetries in temporal discrimination in pigeons. Four groups were exposed to combinations of equal or different events during DIs and ITIs and consistent or inconsistent arrangements between short-sample and long-sample trials. Results showed that differentiating short from long samples shortened acquisition time and eliminated asymmetries in performance. The specific form of performance depended on the interaction between interval similarity and task arrangement: Equal intervals with consistent arrangements produced a choose-short effect, whereas different intervals with consistent arrangements produced a choose-long effect. Inconsistent arrangements led to symmetrical performance, either stable or decreasing as the DI increased. These findings indicate that asymmetries in temporal discrimination under delayed matching-to-sample procedures depend not on memory mechanisms such as code loss or subjective shortening, but on how the organization of task events within the delay and intertrial intervals enhances or reduces stimulus control. The study thus supports interpreting temporal discrimination in this class of procedures as a problem of stimulus control shaped by the task’s event structure rather than by memory decay.
Boyle and Brown’s (2025) functional account and the Temporal Context Model–Successor Representation (TCM-SR; Zhou et al., Psychological Review, 132(1), 18–49, 2025) approach episodic memory from different angles, yet they converge on a similar claim: episodic memory supports adaptive behaviour by flexibly using past experience. In this commentary, we highlight this convergence and build on Boyle and Brown’s framework by focusing on how this function can be mechanistically specified and empirically tested. Boyle and Brown outline why episodic memory may have evolved to support learning from sparse, non-repeating experiences, whereas TCM-SR formalises how contextual reinstatement can supply predictive information about what typically follows, allowing past experience to guide evaluation. We propose that the probed recall task offers a direct way to test these shared claims, that episodic retrieval makes related experiences accessible, and this process can support learning beyond simple associative mechanisms. The contiguity effects seen in this task, in which retrieval is cued by a studied item, reflect contextual reinstatement rather than simple associations, capturing both the integrative processes emphasised by Boyle and Brown and the retrieval dynamics central to TCM-SR. By linking function, mechanism, and behavioural signature within a single paradigm, we outline a path toward a more unified understanding of episodic memory that explains how it accomplishes one of its main functional goals.
In a new study, researchers offer evidence that chimpanzees rationally revise their beliefs in light of new evidence, a hallmark of epistemic rationality.
