Abstract
The flanker effect shows that responses to a central target, flanked by distracting items, are slowed when the flanker items are linked to the opposite response to that of the target. There is general agreement that this interference is strongest at stimulus onset and is suppressed over time. However, debate remains regarding whether this improvement occurs discretely or continuously. White, Ratcliff & Starns’s (2011) Shrinking Spotlight Model (SSP) realizes a gradual narrowing of attention within a single stage (i.e., zoom lens model). Conversely, Hübner, Steinhauser & Lehle’s (2010) Dual-Stage Two-Phase (DSTP) model proposes that this increase in selectivity is discrete, marked by the selection of a candidate stimulus on which to focus attention. Previous attempts to compare these models have used static, noise-free stimuli. To compare these models, we examined their performance not only in terms of reaction time distributions and error responses, but also how well they capture the influence of perceptual noise on the flanker effect. Therefore, we utilized data from our recent flanker study where we manipulated perceptual noise with random-dot kinematograms (RDKs). Subsequently, we estimated the marginal likelihood using Thermodynamic Integration via Differential Evolution (TIDE, Evans & Annis, 2019). While the results were mixed, there was an overall preference for SSP, suggesting a gradual improvement in selectivity offered a better account of the data. However, both models struggled to produce good fits for the highest noise condition compared to lower noise conditions. At a population level, SSP underestimated the effect in reaction time, while DSTP overestimated the effect in error. In the poster, we discuss how the models account for these effects and the implication of these results.