Abstract
People can make accurate ensemble perception (EP) judgments about features of groups of objects (e.g., the mean size of apples). Discussions about the generality of EP for different features have been mostly based on correlations between pairs of EP tasks. Correlations between pairs of EP tasks can usually be explained by common stimuli (e.g., faces in both tasks) or common judgments (e.g., two mean judgments or two variance judgments). We seek evidence for a general EP ability that underlies estimation of different summary statistics and different object categories, distinct from object recognition ability. Two-hundred participants completed (1) a diversity judgment with blob ensembles, (2) a mean judgment with ensembles of transformer toys, and (3) the novel object memory task which uses multiple items simultaneously but requires only object recognition (OR) and no EP. We hypothesized that EP contributes to ensemble judgments (Tasks 1 and 2) and that OR contributes to tasks with objects (Tasks 2 and 3). We conducted confirmatory factor analyses using the R library lavaan. We used the MLR estimator and created 4 parcels of trials per task. A correlated three-factor model (equivalent to a model with one higher-order factor) fit the data well (robust χ2(51) = 79.02, robust RMSEA = .055, CFI = .97). A likelihood ratio test revealed that the correlations between factors assumed to share influence from EP (Tasks 1 and 2) or from OR (Tasks 2 and 3) was higher (r = .56) than those between the factors that we expected not to share these influences (Tasks 1 and 3, r = .41, χ2 (1) = 4.65, p = .03). The results suggest that performance on EP judgments that differ in task, stimuli, and summary statistics are supported by mechanisms distinct from those responsible for general object recognition.