Eyes on emergence: Fast detection yet slow recognition of emerging images

Abstract

Visual object recognition occurs at the intersection of visual perception and visual cognition. It typically occurs very fast and it has therefore been difficult to disentangle its constituent processes. Recognition time can be extended when using images with emergent properties, suggesting they may help examining how visual recognition unfolds over time. Until now, their use has been constrained by limited availability. We used a set of stimuli with emergent properties-akin to the famous Gestalt image of a Dalmatian-in combination with eye tracking to examine the processes underlying object recognition. To test whether cognitive processes influenced eye movement behavior during recognition, an unprimed and three primed groups were included. Recognition times were relatively long (median ∼ 5s for the unprimed group), confirming the object's emergent properties. Surprisingly, within the first 500 ms, the majority of fixations were already aimed at the object. Computational models of saliency could not explain these initial fixations. This suggests that observers relied on image statistics not captured by saliency models. For the primed groups, recognition times were reduced. However, threshold-free cluster enhancement-based analysis of the time courses indicated that viewing behavior did not differ between the groups, neither during the initial viewing nor around the moment of recognition. This implies that eye movements are mainly driven by perceptual processes and not affected by cognition. It further suggests that priming mainly boosts the observer's confidence in the decision reached. We conclude that emerging images can be a useful tool to dissociate the perceptual and cognitive contributions to visual object recognition.

Perceptual reorganisation from prior knowledge emerges late in childhood

“…In addition, the use of visual stimuli and blurred transitions reduces the demand for working memory, cross-modal integration and configurement of mental representations. Alternative cueing methods, such as non-sequential paradigms and semantic cues, are sufficient to trigger perceptual reorganisation in adults (González-García et al, 2018; Ludmer et al, 2011; Nordhjem et al, 2015; Samaha et al, 2018). While in the present study we prioritised reducing load for cross-modal and working-memory processes known to concurrently develop throughout childhood (Buss et al, 2018; Jüttner et al, 2006; Swanson, 2017), future work should test how these processes affect cue integration across development.…”

The Use of Prior Knowledge for Perceptual Inference Is Preserved in ASD

“…Little is known about the actual format of the prior induced with this kind of Mooney templates. A recent study in TD participants showed that word primes (e.g., “gorilla” if the Mooney contained a gorilla) or image primes of different exemplars of the same category as the object in the Mooney image (Nordhjem, Kurman Petrozzelli, Gravel, Renken, & Cornelissen, 2015) helped recognition as well, suggesting that an activation of a conceptual prior is sometimes sufficient to improve Mooney perception. Indeed, neural changes related to prior information for Mooney images can be found both in higher-level visual and in lower-level retinotopic areas, and those activity patterns are different from the neural correlates of conventional (feature-based) priming (Gorlin et al, 2012; Hsieh et al, 2010).…”

Perceptual reorganization from prior knowledge emerges late in childhood

“…Alternative cueing methods, such as non-sequential paradigms and semantic cues, are sufficient to trigger perceptual reorganization in adults. 25 , 40 , 41 , 42 While in the present study we prioritized reducing load for cross-modal and working-memory processes known to concurrently develop throughout childhood, 43 , 44 , 45 future work should test how these processes affect cue integration across development.…”