Statistical processing: Not so implausible after all
DOI: 10.3758/pp.70.7.1327
URL: https://doi.org/10.3758/pp.70.7.1327
archive: archived pipeline: cataloged verified
Summary
Theoretical/empirical reply to Myczek and Simons (2008), defending the hypothesis that the visual system uses a distributed-attention statistical processing mode to extract set-level properties (e.g., mean size) of groups of similar items, rather than focused attention to small samples of one to four items. Reviews prior dissociations between performance on individual-item and averaging tasks, then presents three new experiments with circle arrays (200-500 ms exposures) testing predictions of the sampling account against the distributed-attention account.
Key finding
Three experiments contradict the focused-attention sampling account: (1) mixing display types that would require different sampling strategies did not impair mean-size accuracy relative to blocked presentation; (2) participants estimated the mean size significantly worse when shown small samples (one or two circles per side) than when shown the full eight-circle display; (3) accuracy was equivalent on trials where the largest item was on the same versus the opposite side as the larger mean, indicating participants did not rely on a largest-item comparison strategy. Results support a separate distributed-attention statistical processing mode.
Methodology
Three psychophysical experiments with circle arrays presented for 200 or 500 ms; participants judged which side of the display had the larger mean size. Exp 1 compared blocked vs. mixed presentation of three display types (uniform-uniform, uniform-heterogeneous, frequency) that would each demand a different focused-attention sampling strategy. Exp 2 manipulated the number of items shown (1, 2, or all 8 per side) and sample location (foveal vs. random). Exp 3 manipulated whether the largest single circle was on the same or opposite side as the larger mean. Comparisons of mean accuracy were made against Monte Carlo simulations of optimal sampling strategies.
Sample size: Not reported in extracted text
Quality score: 5 / 5