Abstract. A model is described attributing narrow perceptual spans of poor readers to abnormally slow phonological coding speed. In a test of the model with elementary-age school boys, (1) poor readers were slower than average readers on a digit naming task; (2) perceptual span for random digits was impaired for poor readers; (3) a linear relation was found between perceptual span and naming speed; and (4) within the limits of reliability, perceptual span and naming speed accounted for the same portion of reading ability variance. Discrepant results are also presented, and possible modifications of the model are discussed.It has long been known that perceptual span, defined as the number of items reported from brief tachistoscopic exposures, is greater for good readers than for poor readers. It is commonly thought, however, that perceptual spans of good readers are greater than those of poor readers only for meaningful material, such as words and sentences. Reviewing early studies, Anderson and Dearborn (1952) concluded that perceptual spans of good and poor readers do not differ for nonsense material such as strings of unrelated letters. They interpreted this to mean that, compared to good readers, poor readers do not actually perceive fewer letters, but that they extract less meaning from the letters they do perceive. This interpretation is similar to the view of contemporary theorists who emphasize the psycholinguistic nature of reading. For example, Smith asserts that skilled readers extract no more graphemic information from an eye fixation than beginning readers, and explains that their wider perceptual span is due to knowledge of redundancies of the language (Smith,