Abstract
We applied the Bubbles technique to reveal directly the spatio-temporal features of uppercase Arial letter identification. We asked four normal readers to each identify 26,000 letters that were randomly sampled in space and time; afterwards, we performed multiple linear regressions on the participant's response accuracy and the space–time samples. We contend that each cluster of connected significant regression coefficients is a letter feature. To bridge the gap between the letter identification literature and this experiment, we also determined the relative importance of the features proposed in the letter identification literature. Results show clear modulations of the relative importance of the letter features of some letters across time, demonstrating that letter features are not always extracted simultaneously at constant speeds. Furthermore, of all the feature classes proposed in the literature, line terminations and horizontals appear to be the two most important for letter identification.
We thank those who took part in this study. This research was supported by a grant from the Canadian Institute of Health Research (CIHR) to Martin Arguin, Frédéric Gosselin, and Daniel Bub; by grants from the Natural Sciences and Engineering Research Council of Canada (NSERC) to Martin Arguin and to Frédéric Gosselin; by a scholarship from the James S. McDonnell Foundation (Perceptual Expertise Network) and by a postdoctoral scholarship from the Fonds Québécois de Recherche en Nature et Technologies (FQRNT) to Daniel Fiset; and by a FQRNT graduate scholarship to Caroline Blais and to Karine Tadros.