Yale Perception & Cognition Lab

VSS '08 Abstracts
Jump to:  
Flombaum, J. J., & Scholl, B. J. (2008). How does attention operate during multiple object tracking?: Evidence from the 'slot-machine' task for parallel access to target features. Talk given at the annual meeting of the Vision Sciences Society, 5/10/08, Naples, FL.  
In multiple object tracking (MOT), observers track a subset of haphazardly moving and featurally-identical objects. The fact that MOT is possible in the first place is often taken as implicit evidence that the moving targets are all attended in parallel, but there has never been a direct experimental test of this critical hypothesis. We tested this possibility using novel combinations of MOT and probe detection. In the MultiProbe task, observers detected small simultaneous 80ms probes that appeared in the centers of targets once on each trial. In particular, observers determined whether there were as many probes as targets, or one fewer. Performance was well above chance when tracking multiple targets among an equal number of distractors -- an ability that would be impossible without simultaneous access to each of the targets. In the Slot Machine task, observers tracked three targets among three distractors, but these objects were not identical. Instead, each object's color changed every 250ms, and each object possessed a distinct color throughout a trial (ensuring that tracking was still necessary for target identification). At one key moment in two thirds of trials, however, either two or three of the targets' colors momentarily matched -- as in the congruence of wheels on a slot-machine (jackpot!). Observers readily determined whether the match involved two, three, or none of the targets -- an ability that would not be possible without sustained parallel access to each of the targets' features. (This result is also striking due to the fact that observers generally have very poor access to objects' surface features during MOT.) This novel slot-machine method can also be used to directly explore the extent to which these parallel resources are divided between noncontiguous regions of the display. Overall, this work begins to reveal the underlying attentional dynamics that make MOT possible.
Gao, T., Newman, G. E., & Scholl, B. J. (2008). The psychophysics of chasing. Poster presented at the annual meeting of the Vision Sciences Society, 5/10/08, Naples, FL.  
The currency of visual experience consists not only of features such as color and shape, but also higher-level properties such as animacy. Psychologists have long been captivated by the fact that even simple moving geometric shapes may be perceived in animate, goal-directed terms. However, the study of such phenomena has been limited by two major challenges: (1) Previous research has had difficulty measuring animacy with quantitative precision, given the haphazard construction of typical stimuli. (2) Task demands have made it difficult to distinguish the perception of animacy from higher-level inferences, especially when using simple rating scales. Here we introduce two new converging methods that address both concerns. In the Search for Chasing task, subjects viewed many identical moving discs, and had to detect whether a chase was present: on half the trials, one disc (the 'wolf') pursued another disc (the 'sheep'). Across trials, we manipulated 'chasing subtlety' - the degree to which the wolf could deviate from a perfectly 'heat-seeking' trajectory. Detection accuracy revealed both a robust perception of chasing (with small subtlety values), and an ability to infer chasing without direct perception (with larger subtlety values). The Don't Get Caught! task was similar, but now subjects controlled the sheep's trajectory via the computer mouse, with the goal of avoiding contact with the wolf after identifying it. Performance was a U-shaped function of chasing subtlety. Subjects readily avoided being caught with both large deviations (when the chasing was highly inexact in the first place) and small deviations (when the wolf was easily identified and thus avoided). With intermediate deviations, however, performance was poor: the wolf essentially 'stalked' the sheep in a manner that was difficult to detect. These results collectively demonstrate how the perception of animacy can be measured with precision and can be distinguished from higher-level inferences.
New, J. J., Schultz, R. T., Wolf, J., Niehaus, J. L., Klin, A., German, T., & Scholl, B. J. (2008). The scope of social attention deficits in autism: Prioritized orienting to people and animals in static natural scenes. Talk given at the annual meeting of the Vision Sciences Society, 5/12/08, Naples, FL.  
A central feature of autism spectrum disorder (ASD) is an impairment in 'social attention' -- prioritized processing of socially-relevant information. For example, people with ASD do not show the same visual interest in (and biased processing of) the eyes and face. Beyond such specific cues, however, socially relevant stimuli are preferentially attended in a broader categorical sense: in particular, observers orient preferentially to people and animals (compared to inanimate objects) in complex natural scenes (New, Cosmides, & Tooby, 2007, PNAS). To determine the scope of social attention deficits in autism, we explored whether this bias was evident in people with ASD (both children and adults, with IQs ranging from 56 to 140). In each trial, observers both with and without ASD viewed alternating versions of a natural scene, and had to 'spot the difference' between them. This difference involved either an animate object (a person or animal) or an inanimate object (a plant or artifact) that either reversed its orientation or repeatedly disappeared and reappeared. Participants were not made aware of these categories, and change detection performance (in terms of both speed and accuracy) was measured as an index of automatic attentional prioritization. Control participants without ASD showed prioritized attention to people and animals, replicating previous work. This could not be explained by lower-level visual factors, since the effect disappeared when using blurred or inverted images. Our primary discovery was that individuals with ASD also showed the same prioritized social attention for animate categories. This prioritized social attention increased slightly with age, and was generally unrelated to their clinically-evaluated social abilities. These results suggest that social attention -- and its impairment in autism -- is not a unitary phenomenon: specific impairments in processing faces and eyes may occur despite the intact categorical prioritization of visual social information.
Turk-Browne, N. B., Johnson, M. K., Chun, M. M., & Scholl, B. J. (2008). Neural evidence of statistical learning: Incidental detection and anticipation of regularities. Talk given at the annual meeting of the Vision Sciences Society, 5/12/08, Naples, FL.  
Our environment contains many regularities distributed in space and time that can be detected by way of statistical learning. This unsupervised learning occurs without intent or awareness, but little is known about its component processes, how it manifests over time, or how it relates to other types of learning. Here we use fMRI as a measure of statistical learning to explore these questions. Participants viewed short blocks of novel shapes appearing one at a time, while performing a motion-detection cover task. The underlying sequence of shapes constituted our primary manipulation. Structured blocks contained deterministic sub-sequences of shapes. Random blocks lacked this structure but were otherwise identical. Sensitivity to statistical structure was assessed by comparing fMRI responses to these two block types. This approach resulted in several discoveries about the nature of statistical learning. (1) Robust neural responses to statistical structure were observed during learning, despite weak subsequent explicit familiarity judgments --indicating the utility of fMRI as a measure of statistical learning. (2) This neural evidence of learning emerged after surprisingly little exposure -- as made possible by our use of an online measure of learning. (3) The brain regions that were sensitive to statistical structure overlapped with those underlying other well-studied forms of learning and memory -- helping to characterize the nature of the component processes that support statistical learning. (4) Responses to statistical structure were also observed in visual cortical regions -- suggesting that these regions are sensitive to temporally contiguous relations in addition to static visual features. (5) Several regions involved in reflective processing exhibited enhanced responses to the beginnings of deterministic subsequences -- suggesting that anticipation per se need not be conscious, and may be a natural perceptual process. Collectively, these results emphasize both the power of statistical learning and its integration with other cognitive processes.