Abstract
Face recognition may differ from object recognition in being more affected by spacing between parts (e.g., distance between eyes) than by the features themselves. Here we examined adults' ability to discriminate and recognise houses in a design similar to that used previously with faces. Stimuli were photographs of houses in three conditions: houses that differed only in spacing between the windows and door; houses that differed only in features (the particular windows and door); and houses that were completely different. Subjects (N=24) completed a match-to-sample task followed by a 2AFC recognition memory task for each condition. The mean spatial frequency amplitude was matched across all sets. Overall, subjects were less accurate at the spacing task than the feature task (81% vs. 90%, t(23) = 4.45, p lt; .001 for discrimination; 47% vs. 89%, t(23) = 4.95, p lt; .001, for memory). Importantly, even for sets for which accuracy on the spacing and feature tasks were matched at the discrimination stage (81% vs. 84%, t(11) = 1.15, p [[gt]] .05), memory for previously seen spacing was at chance (50%, t(11) lt; 1), while memory for particular features remained high (85%, t(11) = 5.3, p [[lt]] .001). This contrasts with a previous study with faces in which adults remembered spacing information learned as part of another task at well above chance levels (Gilchrist & McKone, 2003). The current results for houses are consistent with different real-world processing demands for faces and houses. Facial features change as individuals talk, turn their head, or show facial expressions, and spacing information can be deduced relative to the same basic layout in every face. Houses, conversely, have constant features that can be distinctive and far fewer restrictions on basic structure (the door goes at the bottom but windows can go in any number of locations).
This work was funded by grants from CIHR.