Not all letters are alike; the role of phonology in reading

The major goal of a series of four experiments was to evaluate whether printed words contain units that are similar to the onset and rime constituents found in spoken words. Recent research in speech perception has provided the impetus for this investigation by showing a level of linguistic analysis intermediate between phonemes and syllables--an analysis into onsets and rimes. Linguistic theories have provided converging support. One might expect onsets and rimes to be found in written English because of the close correspondence between print and speech. Therefore, these studies investigate whether onsets and rimes exist is word recognition, the first stage of reading. Previous research in reading which investigated a similar question used a speech-based task and confounded written and spoken English. Previous research then has failed to provide a clear understanding of the role of letter units in word recognition. To conduct this evaluation in a less-flawed manner, this series of experiments predominantly utilized a non-speech based task. Also, these experiments utilized carefully constructed nonwords (words that look real; consider VUST) to minimize interpretive ambiguities due to phonological inconsistency in some English words. The nonwords were phonologically consistent (e.g., nonwords unlike TAVE; consider have and save) and were not homophonic with real words (e.g., nonwords unlike BRANE which sounds like a real word).; Experiments 1, 2, and 3 utilized a letter detection task. A phoneme replacement task was used in Experiment 4. In the letter detection tasks, subjects viewed a nonword, then a target letter, and decided if the target letter was present in the preceding nonword. Target letters that alone represent a linguistic constituent should be faster and more accurately detected than target letters which are part of a larger letter unit--a linguistic constituent. Although the results of Experiments 1 and 2 failed to find support for the role of onsets in printed word recognition, the findings of Experiments 1-3 found evidence for an internal organization among letters within a graphemic rime. An organization that is determined, in part, by the letter's phonological attributes. The findings suggest that graphemic rimes have two parts--a vowel unit and a coda unit. The results show that sonorants following a vowel appear to occupy a "seat" in the nucleus. In contrast, postvocalic obstruents are classified as part of the graphemic coda. The null findings of the speech-based experiment, Experiment 4, hint that the cohesion of sonorants to preceding vowels is unaffected by vowel length in the analysis of spoken English; a finding that may generalize to a property of letter units in word recognition.