Priming by the mean representation of a set

Abstract

Recent evidence suggests that sets of similar objects can be represented in terms of their statistical parameters, such as mean size. Observers are more likely to indicate that a probe item was part of a previously presented set of items when the probe has the same size as the mean size of the set than when it has the same size as one of the set members (e.g., Ariely, 2001). Here we provide further evidence for set representation by statistical properties, by showing priming by the mean size of a set of circles. Observers were presented with a set of circles followed by a degraded outline of a single target circle and were asked to judge the contrast of the target. Target contrast was reported to be significantly higher when the target circle had the same size as the mean size of the preceding set of circles than when it had the same size as one of the members of the preceding set. These findings show that conceptual priming by a summary description can be stronger than exact repetition priming.

Keywords:

• Statistical processing
• Size
• Mean
• Priming
• Perceptual representation

Notes

¹ We use the term conceptual priming to differentiate it from exact repetition priming and to suggest that the resultant priming is due to the existence of an abstract concept of the set rather than exact repetition of one of its members. This is somewhat different from the standard usage of conceptual priming, which is normally related to priming after semantic processing.

² We also ran the principal analysis including only trials on which participants made correct mean estimations. This analysis showed the same pattern of results, and this time both pairwise comparisons were statistically significant: Following a set with small mean size, the proportion of “high-contrast” responses was greater for a small target circle (mean = .53, SE = .065) than for a large target circle (mean = .30, SE = .047), t(12) = 2.19, SEM = 0.105, p < .05. Conversely, following a set with large mean size, the proportion of “high-contrast” responses was greater for a large target circle (mean = .62, SE = .049) than for a small target circle (mean = .37, SE = .055), t(12) = 2.66, SEM = 0.097, p < .05.