Abstract
Background: Visual crowding the failure to identify an object in clutter, imposes significant constraints on reading and has been linked to reading difficulties and developmental dyslexia. Previous studies in alphabetic scripts have demonstrated that letter recognition within a trigram string is more accurate when the string forms a word compared to a pseudoword (the well-known "lexicality" effect). This effect occurs both in the fovea and the parafovea. However, words and pseudowords differ not only in their lexical properties, such as print frequency, but also sublexically in the transitional probabilities of their letters (n-grams). These probabilities which capture the likelihood of the occurrence of a specific letter given its neighboring letters, play a crucial role in reading. The precise mechanism through which transitional probabilities facilitate reading, however, remains unclear. Objective: We investigated the effects of transitional probability (bigram/trigram frequency), lexicality (words vs. pseudowords) and visual hemifield on crowded letter recognition among skilled readers in Hebrew, a right-to-left script. Method: In two experiments (N = 27), we measured font-width threshold in three conditions: uncrowded (an isolated letter), crowded word, and crowded pseudoword. In Experiment 2, observers also performed several blocks of crowded word and pseudoword tasks at threshold level. We used logistic regression analysis to determine the contribution of transitional probability to performance. Results: We revealed two language-universal effects: a lexicality effect and a right hemifield (left hemisphere) advantage, as well as a strong language-specific effect a left bigram advantage stemming from the right-to-left reading direction of Hebrew. This finding suggests that transitional probabilities are a significant factor in parafoveal letter recognition. Conclusions: These results shed light on the visual system's processing of crowded stimuli in general and in printed words in particular. They reveal that script-specific contextual information, such as letter combination probabilities, influences letter recognition in crowded displays.