Abstract
In free recall procedures, the order and timing of participants' responses can inform our understanding of human memory. Analyzing a corpus of more than half a million freely recalled words from 127 young adult participants, we develop a statistical model of interresponse times (IRTs) as a function of the temporal and semantic relations among the recalled items and their positions in the output sequence. Residual IRTs exhibited strong sequential dependencies, being positively correlated at lags of one and two transitions. We used this IRT model to evaluate the hypothesis that chunking helps participants learn unstructured materials. Specifically, we hypothesized that chunks would appear as a slow IRT, indicative of a boundary, followed by a sequence of fast IRTs. Model-based analyses that included sequential dependencies in IRTs offered evidence for spontaneous chunking in free recall of lists that lacked any grouping or hierarchical structure. (PsycInfo Database Record (c) 2026 APA, all rights reserved).