Locus of dimension weighting: Preattentive or postselective?

Abstract

In visual search for singleton feature targets, detection RTs are faster when the target-defining dimension is constant across trials rather than variable; in the latter case, RTs are faster when the target dimension on a trial is the same, rather than different, relative to preceding trial (with little effect of a feature change within a repeated dimension); and RTs are expedited when the target dimension is validly indicated by a symbolic precue on a given trial (e.g., Found & Müller, 1996; Müller, Heller, & Ziegler, 1995; Müller, Reimann, & Krummenacher, 2003). Müller and his colleagues have explained these effects in terms of a “dimension-weighting” account, according to which these effects arise at a preattentive, perceptual stage of saliency coding. In contrast, Cohen (e.g., Cohen & Magen, 1999) and Theeuwes (e.g., Theeuwes et al., 2006 this issue) have recently argued that these effects are postselective, response-related in nature. The present paper examines these challenges and puts forward counterarguments in support of the view that dimensional weighting operates at a preattentive stage of processing.

Notes

¹Found and Müller (1996) obtained evidence of some feature-specific intertrial facilitation only for the colour-defined targets (there was a RT advantage for, say, a red target preceded by a red target, relative to a preceding blue target), but not orientation-defined targets, consistent with work on colour and orientation grouping (Nothdurft, 1993; Wolfe, Chun, & Friedman-Hill, 1995).

²Consistent with this, Meeter and Theeuwes (2006) have recently reported that precueing the dimension or, respectively, feature of an upcoming target obligatorily sets the system to (all) odd-one-out stimuli defined within the precued dimension, rather than the precise feature defining the target (at the exclusion of features defining distractors). Meeter and Theeuwes presented an odd-one-out “distractor” (in a display matrix of homogeneous nontarget elements) that could be defined within either the same precued dimension as the singleton target or within a noncued dimension (same-dimension distractor, e.g., “colour”-cue, colour-defined distractor; different-dimension distractor, e.g., “colour”-cue, orientation-defined distractor). The featural identity of the distractor was fixed (e.g., the colour-defined distractor was always red), providing observers with an incentive to ignore/suppress the distractor feature—so as to be able to respond as fast as possible to the presence of a target, avoiding the standard distractor interference effect demonstrated by Theeuwes (1992). Besides a general RT cueing benefit relative to a neutral-cue condition, Meeter and Theeuwes found that a same-dimension distractor produced greater interference than a different-dimension distractor, whether the cue specified the dimension of the upcoming target or the precise target feature. For target-present trials, the distractor effects were assessed against a redundant-target condition (in which the singleton target was defined in both the colour and the orientation dimension). Yet, the results showed a pattern of distractor interference effects only relative to the cued dimension of the target. Meeter and Theeuwes took this pattern to suggest that observers can set can themselves, in top-down manner, only to the precued dimension generally, but not to specific features within this dimension.

³In this context, it is worth noting that Müller and O'Grady (2000) have already shown that performance in a nonsearch task, even with only a single stimulus object in the display, is dependent on dimensional processing limitations. In the study of Müller and O'Grady, stimulus displays were presented briefly (for less than 50 ms) and then masked, and the emphasis was on response accuracy, rather than RT. Yet, essentially the same cross-dimension cost was obtained, even when there was only one object in the display. In the latter case, the task required observers to judge one-and-the-same object in terms of dual colour (saturation and hue) or dual form (line size and texture)—that is, within-dimension—judgements or in terms of one colour (saturation or hue) and one form (line size or texture)—that is, cross-dimension—judgements (e.g., hue and texture). Judgement accuracy was reduced for cross-dimension, relative to within-dimension, judgements (essentially the same cross-dimension cost was obtained when judgements were to be directed to two separate objects, though in this case there was an additional object-based effect). Judgements/responses were made by observers mouse-clicking one of several visually displayed response alternatives (click panels, presented 1033 ms after display termination): The panel that matched the perceived (to-be-reported) stimulus properties (i.e., observers had to select the appropriate panel, move the mouse cursor there, and click it—without any time pressure). Response-selection processes (i.e., processes that select a response appropriate to the information derived from the brief displays) were unlikely to be a performance-limiting factor in this situation. Rather, the finding of cross-dimensional cost effects in this condition argues that these effects arise at a perceptual “locus” of stimulus processing.