Purchase this article with an account.
Jordan Suchow, Denis Pelli; Letter learning: Feature detection and integration. Journal of Vision 2008;8(6):1133. doi: https://doi.org/10.1167/8.6.1133.
Download citation file:
© ARVO (1962-2015); The Authors (2016-present)
Letters are identified in two stages: feature detection and feature integration. At which stage does the learning of unfamiliar letters occur? To make this distinction, we designed three letter identification tasks, matched in feature detection statistics. One challenged both detection and integration, a second challenged only detection, and a third challenged only integration. While previous learning studies have been unable to distinguish learning detection from learning integration, we used gabor letters, which are uniquely suited to distinguishing the stages. Each “gabor letter” is a 2×2 array of four gabors (grating patches), each of which may be horizontally or vertically oriented. Our alphabet contains all 16 possible combinations. By displaying letters at threshold contrast, we challenged both detection and integration. By presenting single features at threshold contrast, we challenged only detection. By presenting only some of the letters' features at a high contrast, we challenged only integration. The results show that learning unfamiliar letters reflects improvement at both stages. Plotting threshold contrast as a function of number of trials, we find that learning feature detection is very slow (log-log slope of 0.03) and learning feature integration is very fast (log-log slope of 2).
This PDF is available to Subscribers Only