Abstract
Humans are highly efficient in identifying individuals or ensembles of animals and man-made objects. For single items, although animacy information could be extracted based on low- or mid-level visual shape features (e.g., animals tend to be curvy and man-made objects are often elongated, Long et al., 2017), category selectivity for individual animals and man-made objects remains after such visual differences are minimized across the categories (e.g., He et al., 2020). For ensemble perception, it remains unclear whether the rapid extraction of animacy across multiple complex, real-world animals or man-made objects may depend on naturalistic variations in low- or mid-level visual features, or category information, between the two categories. To minimize low- or mid-level visual differences among the categories, we used grayscale images of real-world animals and man-made objects, either of round or elongated shapes, which shared comparable gist statistics across categories. Across two experiments, participants judged the relatively numerosity on briefly presented (500ms) displays of six animal and man-made object images of the same shape (round/elongated), with either a numerosity ratio of 4:2 or 5:1. The items were presented either at six fixed locations (Experiment 1, N=43) or six random (among 12) locations (Experiment 2, N=44). We found evidence of ensemble processing for both categories, with significantly higher accuracy in numerosity judgment, compared with the respective baselines for each of the numerosity ratios, calculated based on the expected performance if participants randomly subsampled only one item from a display. Moreover, ensemble perception for animals and man-made objects appears to be facilitated by shape information, with significantly better and faster ensemble performance for round than elongated animals, and for elongated than round man-made objects. These results suggest that ensemble processing of animacy can depend on rapid extraction of category information, and is facilitated by the expected shapes of the respective categories.