Local predictive context accelerates evidence accumulation during semantic priming: a drift-diffusion approach

Facilitation in word processing, identified as faster responses to semantically related words, has traditionally been viewed as evidence of the core cognitive mechanisms that support language processing. However, it remains uncertain whether this facilitation arises from bottom-up spreading activation, top-down prediction, or a combination of both. To address this, we developed a novel semantic priming paradigm that systematically manipulates local contextual buildup and applied drift-diffusion modeling (DDM) to differentiate the decision dynamics associated with each mechanism. In our experiment, 20 participants performed a lexical decision task on targets preceded by either one (short condition) or three (long condition) semantically related primes. We found that, as the local semantic context became more enriched (short vs. long condition), participants adjusted their decision strategies, engaging different processing mechanisms. In the short condition, participants relied on spreading activation, as indicated by increased drift rate (faster evidence accumulation), suggesting that a single related prime pre-activated the target's mental representation. In contrast, in the long condition, reduced boundary separation (participants required less evidence before responding) implied greater confidence driven by enhanced predictability from multiple primes, consistent with the predictive processing account. Together, our findings offer new insights into theoretical models of language processing; demonstrate the practical relevance of DDM in language processing and offer potential implications for translational research.