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

Imagine reading a shopping list: words such as apple, banana, and pear immediately create a strong sense of coherence, forming a predictable semantic context so that when the word orange appears, it is readily anticipated. In contrast, when the list begins with only apple followed by orange, the prediction is weaker despite semantic closeness. To capture this effect experimentally, participants (n = 20) performed a lexical decision task on targets preceded by either one prime (short condition) or three semantically related primes (long condition). This design allowed us to examine how local semantic predictability accumulated across multiple primes. Our results showed that repeated exposure to semantically related primes substantially accelerated decision-making — an effect we attribute to faster evidence accumulation for semantically similar prime-target pairs, as quantified by a drift-diffusion model. Our results thus extend our understanding of language prediction to minimal contexts, demonstrating that top-down facilitation can emerge rapidly from just a few preceding items. This study advances our understanding of how prediction shapes language processing and demonstrates the relevance of hierarchical decision-making models in behavioral research.