Academic journal article Perception and Psychophysics

Attentional Spread in the Statistical Processing of Visual Displays

Academic journal article Perception and Psychophysics

Attentional Spread in the Statistical Processing of Visual Displays

Article excerpt

We tested the hypothesis that distributing attention over an array of similar items makes its statistical properties automatically available. We found that extracting the mean size of sets of circles was easier to combine with tasks requiring distributed or global attention than with tasks requiring focused attention. One explanation may be that extracting the statistical descriptors requires parallel access to all the information in the array. Consistent with this claim, we found an advantage for simultaneous over successive presentation when the total time available was matched. However, the advantage was small; parallel access facilitates statistical processing without being essential. Evidence that statistical processing is automatic when attention is distributed over a display came from the finding that there was no decrement in accuracy relative to single-task performance when mean judgments were made concurrently with another task that required distributed or global attention.

As we navigate through the environment, we sometimes focus attention on a single object, such as a tree, and sometimes spread it over a wider area to see the forest. The forest, in turn, can yield two different kinds of information. It may have global properties of its own, such as a shape, a color, an orientation, and a density, and it also comprises the aggregate properties of its individual elements, which can be statistically summarized. The individual trees have a mean size with a particular variance and range, a mean separation, a mean leafiness, and so on. Most environments are hierarchically structured, as in this example. Perceptual mechanisms have evolved to form representations that connect objects hierarchically, giving properties of the wholes, as well as properties of their component parts. We perceive the global features that emerge from the arrangement of elements, the individual properties of particular local elements, and the statistical properties of the ensemble. How do we extract these three kinds of information-global, local, and statistical?

A possible hypothesis, some aspects of which we explore in this article, is that there are different modes of processing and that each is favored by a particular distribution of attention. At one extreme, when attention is focused on a single object, we bind its features, determine its structure, and perhaps identify it. At the other extreme, we may attend globally to the scene as a whole. With attention set to the global scale, we can access the gist, or semantic interpretation, of the scene (for example, a beach in Summer, a modern kitchen) and its global layout. When attention is distributed over a set of items but the scale is adjusted to that of individual elements, we have access to their statistical properties. Between these extremes, we may attend to pairs or triplets of objects to determine the relations between them-for example, the book is under the table, Mary is stroking the cat. The apparently complete and veridical representation of the surrounding scene that we normally experience may be an illusion generated from occasional detailed samples, together with statistical summaries of remaining areas and an overall interpretation of the meaning or gist. This sampling and summarizing at different levels of representation may help account for the striking change blindness recently explored by Rensink (2002), Simons (Simons & Levin, 1997), and others.

Global Versus Local Properties

In research pioneered by Navon (1977), global and local processing of shape have been compared, using large shapes made of smaller shapes. These paradigms varied both the size of the attended area and the scale of the attended object. An early issue concerned the order of processing: Is the overall structure identified earlier in time than the component parts? Navon reported evidence supporting global precedence. Global precedence, in a different sense, is also reflected in the asymmetry of interference between global and local forms: A global form typically creates greater interference in the processing of a local form than the converse (Hoffman, 1980; Miller, 1981; Pomerantz, 1983). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.