Modeling negatively accelerating search slopes in a relational search task

Logan (1994) showed that visual search for a target defined by the spatial relations among its parts (e.g., a plus above a dash among dashes above plusses) gives rise to a steep linear search slope. In seven previously reported experiments, we (Heaton et al., 2021; Heaton, 2023) conceptually replicated Logan with slightly different stimuli (colored Xs and Os in above-below relations) and with a larger range of set sizes than those used by Logan. These experiments revealed steep but negatively accelerating functions of response time vs set size. Here we report simulations with the most recent version of the CASPER model of visual search (Heaton, 2023) that account for the negative acceleration in these search functions in terms of the role of an emergent feature in the negative space between the X and the O in our stimuli. The influence of this feature is modulated by factors that affect perceptual grouping strength. In the absence of this feature, CASPER produces steep linear functions of response time over set size. These results suggest that the visual system is capable of searching for targets defined only by spatial relations but will exploit any simple visual features that are confounded with the correct response, and in so doing highlight the difficulty of conducting search experiments that we can be confident depend exclusively on relational processing.