Visual Cognition Lab

Projects

Our research focuses on mental representations of visual information. In particular, we are interested in what people represent about objects and scenes, and the role of attention in the development of those representations.

Scene Perception

Given the richness of one's perceptual experience, it is tempting to conclude that mental representations of the world are equally rich and complete. The truth, however, appears to be more complicated. Demonstrations of change blindness have revealed a striking insensitivity to changes in scene content, leading to questions concerning the accuracy of scene representations. While some have argued that scene representations are quite sparse, others have argued that representations are relatively detailed, and that scene changes are missed because comparisons are rarely made across scene views.

Both of these accounts share a common assumption that visual attention plays a critical role in the development of mental scene representations. For that reason, we have focused on identifying those factors that guide attention during early scene viewing.

In one set of experiments, for example, we used a spatial probe to assess participants' attention while viewing scenes containing semantically consistent (e.g., a globe in a classroom) or inconsistent (e.g., a lawnmower in a classroom) objects. Our results suggested that participants are biased to attend to inconsistent objects just 150 ms after the scene is presented.

We have found converging evidence for this claim using a procedure in which we use scenes to prime responses on a subsequent lexical decision task. In those experiments, we also found evidence that suggests that inconsistent objects draw attention not because participants know that they are incongruous, but because they are especially difficult to identify without contextual support.

We have also found evidence for individual differences in biases to attend to objects in scenes. For example, we have found that participants who are high in trait anger are biased to attend to non-hostile cues in scenes depicting ambiguously hostile events. We are also currently examining smokers' biases to attend to smoking-related objects embedded in complex scenes.

Relevant Publications:

Object Representation

We inhabit a dynamic world, in which objects are constantly moving and changing, and in which our own movements change our views of objects in the environment. Successful interaction with objects in such a dynamic environment requires mental object representations that are sufficiently flexible to accommodate object changes. An example of such a representation is the “object file” representation described by Anne Treisman and her colleagues.

Object files are temporary, episodic representations of objects in the environment. As mid-level representations, they mediate interactions between low-level perceptual representations and higher-level semantic representations. In addition to maintaining perceptual stability across object changes, it has been suggested that they play an important role in the development of complex scene representations and in the integration of information across saccadic eye movements.

We have conducted a number of experiments examining the nature of object file representations. Much of our early work in this area suggested that object file representations are very abstract, so that even very large perceptual changes that leave the object's identity intact will not lead to perceptual instability. More recently, however, we have argued that object file representations include perceptual object features, but that those features to not play a role in processes that support object stability.

Relevant Publications: