Learning to identify letters: Generalization in high-level perceptual learning

Learning the perceptual task of letter identification is crucial to reading. The beneficial effects of training have been well documented for a variety of perceptual tasks. This improvement is often highly task and location specific. However, unlike the specificity of perceptual learning, many studies of conceptual learning have found a great deal of generalization between related tasks and stimuli. Three new experiments explore the specificity of letter learning. In Experiment I, observers learned to identify letters in one subset of a foreign alphabet (Chinese) before learning a second subset of that same alphabet. Observers are found to receive no benefit from having partially learned the alphabet, proving that letter learning is letter specific. In Experiment II, observers trained on the components of Chinese characters (individual brushstrokes and combinations of brushstrokes) before learning the characters themselves. The results show that observers, when learning to identify a new object, need not relearn combinations of features with which they are already familiar; in fact, knowledge of an object’s parts instills a more effective learning strategy in the observer. Experiment III explores the specificity of letter learning with regard to location in central and peripheral vision. Observers’ efficiency for foreign letter identification (Armenian) is found to be highly dependent on eccentricity (distance from central fixation, measured in degrees of visual angle). In sum, the results reveal two mechanisms that identify letters: a process that isolates an object’s individual parts present only in central vision, and a ubiquitous process based on holistic letter shape. These two letter identification processes are likely the substrate for the letter decoding and whole-word processes that have been postulated for reading. 3 of 20 Suchow, Jordan W.