Abstract
Previous research suggests that visual attention can be allocated both to locations (space-based) and to objects (object-based). Space-based attention effects are consistent across experiments, and tend to be large, especially within simple cuing paradigms. Object-based attention effects, however, are smaller and less consistent. Here, we address the possibility that variability in object-based attention effects across studies can be explained by variability across individual observers. We tested 60 observers in each of two configurations of a common object-based attention paradigm (Moore, Yantis, & Vaughan, 1998). In one configuration, two vertically oriented rectangles were presented to the right and left of fixation; in the other, the rectangles were horizontally oriented above and below fixation. Theories of object-based attention predict responses to targets in uncued locations should be faster on trials when the target and cue are presented on the same object than on trials when they are presented on different objects. Using ANOVA and t-tests, we found a space-based cuing effect in both configurations. However, we found the anticipated pattern of object-based results only in the horizontal configuration, and the opposite pattern in the vertical configuration. We then used bootstrapping methods to estimate effect sizes for individual subjects. The analyses found that 99% of observers exhibited significant space-based attention effects, but only 26% of observers exhibited significant object-based attention effects. Furthermore, another 26% of observers exhibited significant effects in the direction opposite to predicted object-based attention effects. This individual variability accounts for the contradictory findings of the two experiments, reveals a potential source of the small and inconsistent findings across the object-based attention literature, and suggests that object-based attention may not be as robust as previously assumed. These results highlight the importance of using techniques like bootstrapping to analyze experimental data at the individual observer level.