Dividing attention across feature dimensions in statistical processing of perceptual groups
DOI: 10.3758/pp.70.6.946
URL: https://doi.org/10.3758/pp.70.6.946
archive: archived pipeline: cataloged verified
Summary
Three experiments testing whether dividing attention between feature dimensions impairs statistical (mean) judgments of perceptual groups. Participants viewed displays of moving circles or tilted bars and indicated which side had the larger mean size, faster mean speed, or more tilted mean orientation. A pre-cue (before display) versus post-cue (after display) manipulated whether attention could be focused on one dimension. Across all three experiments post-cue thresholds were higher than pre-cue thresholds, indicating a cost of dividing attention across dimensions. By contrast, prior work (Chong & Treisman, 2005) found no cost for averaging two spatially intermixed sets within a single dimension. The authors interpret results as consistent with dimensional weighting: statistical processing benefits from attention being directed to a single feature dimension, even though averaging multiple groups within one dimension can proceed in parallel.
Key finding
Statistical averaging across perceptual groups carries a cost when participants must divide attention across feature dimensions (size vs speed; size vs orientation), but not when averaging multiple groups within a single dimension; cuing the relevant dimension before stimulus presentation lowers thresholds relative to post-cuing.
Methodology
Three within-subjects psychophysical experiments using brief (520 ms) displays of moving circles or tilted bars on a CRT (75 Hz). Pre-cue vs post-cue blocks indicated which dimension to judge (size, speed, or orientation). 75% accuracy thresholds estimated via probit analysis; repeated-measures ANOVA on thresholds and raw accuracy.
Sample size: Exp 1: N=24 (3 excluded, analyzed N=21); Exp 2: N=18 analyzed; Exp 3: N=13 analyzed.
Quality score: 5 / 5