A study investigated the roles of phonological encoding and visual processes in word recognition in American learners of Arabic as a foreign language. Subjects were 36 individuals with proficiency ranging from beginning to native. Two experiments in word recognition were conducted, one at word and one at sentence level. At each level, the word recognition task was either phonological or visual. Both experiments were conducted with a personal computer. Results show the learners of Arabic used both visual and phonological strategies to assess Arabic words at both word and sentence levels, continuing to make incorrect judgments through all proficiency levels as manifested in both visual and phonological errors. As reading proficiency increased, error rates of both phonological and visual tasks decreased. Phonological errors were virtually absent in native speakers, and visual errors were infrequent. Recommendations for instruction include: (1) emphasis on critical sounds and letters that may cause confusion in identification; (2) encouragement of a learning strategy using phonology plus spell-checking, especially at lower reading levels; (3) attention to vocabulary and structures that contain critical sounds or have high visual similarity in context; and (4) maximum manipulation of the linguistic codes, especially at lower reading levels. (Contains 27 references.) (MSE)