Recency bias; a strategy to compensate for the lack of evidence

Perceptual decisions often rely on integrating multiple discrete pieces of sensory evidence over time. Previous studies have shown that these integration strategies are dynamic and can give rise to temporal biases such as the recency effect, where more recent information receives greater weight in the decision process. While some have attributed the recency effect to memory leak, recent findings suggest it may instead reflect a strategic modulation of attention. In this study, we investigated whether attention increases uniformly over time or selectively in response to the amount of accumulated evidence. Using a direction discrimination task with single-, double-, and triple-pulse random dot motion stimuli, we tested competing models of evidence integration and attentional allocation. Consistent with prior work, we observed marginally better performance in weak-strong double-pulse trials compared to strong-weak ones. Notably, in triple-pulse trials, improved performance was evident only when considering the first and third pulses, but not the first two or last two, arguing against a uniform time-based increase in attention. Crucially, when the first two pulses were treated as a unit, a weak-strong pattern relative to the third pulse produced significantly higher accuracy. These findings support a model in which attention increases not over time, but over evidence: reduced early evidence leads to increased attentional allocation to later samples. Our results provide further insight into the adaptive nature of perceptual decision-making and the role of online attentional strategies in optimizing performance in environments with discrete, time-separated sensory events.