A number of spatial metaphors have been used to describe the manner in which attention is deployed in the visual environment. For example, Posner (1980); suggested that visual attention operates like a spotlight with objects falling within its focus being preferentially processed. Other spatial metaphors such as a zoom lens and gradient have also been suggested as descriptors which better convey the flexibility with which attention can be deployed and reoriented in the visual field. Indeed, these space-based models of visual attention have received a good deal of empirical support from both psychophysical and neuropsychological studies.

However, more recently it has become increasingly obvious that attention can also be used to select objects and perceptual groups rather than simply regions of visual space. One of the initial studies which supported this object-based conceptualization of visual selection was performed by Duncan (1984). In Duncan’s studies subjects were presented with two overlapping objects, a box and a line, and were asked to identify two properties of the objects. The target properties, the orientation and texture of the line and the size and side of a gap in the box, could either be located on a single object (e.g. the orientation and texture of the line) or distributed between the two objects (e.g. the orientation of the line and size of the box). Despite the fact that the target properties were equidistant regardless of whether they were located on one or two objects, subjects were more accurate in their identification of the properties when they were located on a single object. This effect has subsequently been referred to as the same-object benefit. Such a pattern of results is inconsistent with space-based models since such models suggest that attention is focused independently of the objects or structure in the environment.

Additional research supportive of object-based attentional selection has been performed in focused attention, divided attention and cueing paradigms. Furthermore, several neuropsychological studies have reported cases of neglect not only for the side of the visual field contralateral to a patient’s lesion but also for a particular side of an object independent of the area of the visual field in which the object appeared.

In an important series of studies Egly, Driver & Rafal (1994a), Egly, Rafal, Driver & Starrveveld (1994b) obtained evidence which suggests that both space-based and object-based attention can operate concurrently within the time course of a single experimental trial. Egly et al. presented subjects with two horizontally or vertically oriented rectangles, one on each side of a fixation cross. Subjects were required to respond if one of the ends of the two rectangles flashed. One of the rectangle ends was cued prior to the occurrence of the luminance increment target. The cue was predictive of the location of the subsequent target.

Egly et al. found that RTs were fastest when the luminance target occurred at the cued location, slower if the target appeared at the uncued end of the cued rectangle and slowest if the target appeared in the uncued rectangle. The difference in RT between the cued and uncued ends of the cued rectangle was interpreted in terms of the time required to re-orient a spatial beam of attention while the difference in RT between the uncued end of the cued rectangle and the uncued rectangle was interpreted in terms of the cost incurred when attention had to be shifted from one to the other rectangle. Thus, these data suggest that spatial and object-based modes of attention can operate cooperatively in the service of selective processing of information in the visual environment.

The original purpose of our current studies was to further investigate the interaction between space-based and object-based modes of attentional selection. To that end, we briefly presented subjects with stimuli like those illustrated in Fig. 1 (i.e. line drawings of two horizontally or vertically oriented wrenches). The subjects’ task was to determine whether the two open-ends either matched (i.e. two rectangular open-ends or two rounded open-ends) or mismatched (i.e. one rectangular open-end and one rounded open-end). Prior to the presentation of the wrench pair one of the wrench ends was cued by a bar marker.

Fig. 1. A graphical illustration of the stimuli presented in Experiment 1. The top panel presents an illustration of a different-object trial (requiring a mismatch response). The bottom panel presents an illustration of a same-object trial (requiring a match response).

To our great surprise, results indicated a different-object performance benefit. That is, subjects were faster and more accurate when one target property (i.e. open wrench end) appeared on one wrench and the other target property appeared on the other wrench then when the two target properties appeared on a single wrench. Given that the wrench ends were both equidistant from fixation as well as equidistant from each other, whether they appeared on one or two wrenches, this pattern of results is consistent with neither space nor object-based models of attention. Thus, given the consistent same-object benefit that has been obtained in numerous previous studies when multiple target properties are to be identified our finding of a different-object benefit was an anomaly which required further investigation.

The present experiment was conducted in an effort to replicate the different-object benefit that we obtained in our pilot study. In the pilot study, described above, subjects were presented with a spatially informative cue at the location at which one of the two target properties would subsequently appear. The two wrenches were then presented and subjects decided whether the two target properties, which appeared either on one wrench or distributed between two wrenches, were the same or different.

In the present study we decided to simplify the paradigm by presenting the two wrenches in the absence of the spatial cue. If the same pattern of results is obtained as in the pilot study, that is a different-object benefit, then we can be assured that whatever selection strategies the subjects are employing are not the result of any potential capture of attention engendered by the presentation of the spatially informative onset cue.