Evidence accumulation models of Navon interference

Abstract

The Navon task is a classic paradigm for studying selective attention, in which participants identify target letters presented at either a local or global level within compound figures. Traditionally, differences in reaction time (RT) between local and global trials—so-called precedence effects—are used to quantify individual biases toward local or global processing. However, such RT-based measures conflate attentional bias with general differences in decision caution and speed–accuracy trade-offs. Here, we applied evidence accumulation models (EAMs) to dissociate these mechanisms and evaluate whether Navon interference reflects conflict resolution processes similar to those in tasks such as the Flanker and Simon paradigms. Participants performed a letter-Navon task with congruent and incongruent trials under local and global focus instructions. We fitted four EAMs—the Diffusion Decision Model (DDM), Dual-Stage Two-Phase model (DSTP), Shrinking-Spotlight model (SSP), and Diffusion Model of Conflict (DMC). All models captured the overall response time distributions and conditional accuracy functions, but model comparison favored the DSTP model for most individuals, indicating that Navon processing involves an initial stage of dimension selection followed by evidence accumulation for response selection. Importantly, individual differences in distractor processing within the DSTP predicted behavioral global–local precedence, suggesting that variation in interference resolution underlies observed performance differences. The findings support interpreting the Navon task as a conflict-resolution paradigm and demonstrate that evidence accumulation modeling provides a principled framework for disentangling attentional and strategic contributions to selective attention performance.