Perceptual noise disrupts flanker suppression: Evidence from a novel type of noise in the colour domain and Bayesian modelling

Poster Presentation 43.329: Monday, May 22, 2023, 8:30 am – 12:30 pm, Banyan Breezeway
Session: Attention: Temporal, templates, memory

Dietmar Heinke1 (), Jordan Deakin1; 1University of Birmingham

Despite extensive research into the flanker congruency effect, most studies use stimuli free from perceptual noise. Here we explored the flanker effect in a new paradigm inspired by random-dot kinematograms (RDKs) that introduces perceptual noise in the colour domain. Like RDKs, participants were presented with circular clouds of dots. However, unlike RDKs, the dots were static and perceptual noise was implemented by assigning noise dots a random hue while signal dots were either red or cyan. The level of noise was manipulated by varying the percentage of signal dots across four levels (20%, 30%, 60% and 90%) and fitted two drift diffusion models (DDM) of the flanker task, the Dual-Stage Two-Phase Model (DSTP, Hübner, Steinhauser & Lehle, 2010) and the Shrinking Spotlight Model (SSP, White, Ratcliff & Starns, 2011). A model comparison based on marginal likelihood (Bayes Factor) found that DSTSP was the superior model for the majority of participants despite its higher complexity due to its two-stage architecture. DSTP’s first stage consists of an early selection process vulnerable to flanker interference while the later, second stage is a selection process responsible for flanker suppression. DSTP indicates that colour noise influences behaviour in two ways. First, early interference from flankers increases with decreasing noise in line with their increase of perceptual strength. Second, the threshold to engage the late selection stage is lowered with decreasing noise countering the increased early This effect may be due to the increase of perceptual contrast between target and flanker in line with other research showing that increasing discriminability between target and flankers can lead to a reduction of the flanker effect. We will discuss these results further and compare them with results from a flanker task using typical RDKs.