Abstract
A recent study suggested that stimuli leading to configural superiority effects (CSE) have higher information processing costs than their constituent parts presented alone (Bratch et al., 2016). Embedding the stimuli in noise revealed lower contrast thresholds for part-stimuli compared to configural stimuli. We reasoned that the use of external noise disrupted the processing of emergent features at low contrasts, and that the benefit for part-stimuli would not hold across the whole contrast range. We reanalyzed Bratch et al.'s data and observed that the slopes of the psychometric functions showed a part-based benefit at lower contrasts and a configural benefit at higher contrasts. Such a slope difference was absent for their control experiment, in which the configural stimulus was not intended to elicit any CSE. To corroborate the results of this reanalysis, we conducted a replication and extension experiment of Bratch et al. We measured contrast thresholds in an odd quadrant task for three types of displays: "good" Gestalts (inducing CSE), "bad" Gestalts (not inducing CSE) and their constituent parts. This design allowed us to also directly compare good and bad Gestalts. In a second experiment, we assessed the processing of Gestalts at threshold when external noise is absent. The results revealed: (1) a successful replication of Bratch et al.'s results; (2) lower thresholds for good Gestalts compared to bad Gestalts in both experiments; (3) lower thresholds for good Gestalts compared to their parts in the no-noise experiment; (4) steeper slopes of the psychometric functions for good Gestalt stimuli, suggesting nonlinearity of emergent features; (5) threshold differences for good Gestalts and their parts were strongly dependent on quantifications of contrast: RMS contrast revealed benefits for parts whereas Weber contrast revealed benefits for Gestalts. In sum, these results challenge the generality of the claim that configural processing involves less efficient information extraction.
Meeting abstract presented at VSS 2018