Variant Grapheme-Phoneme Correspondences in Unfamiliar Polysyllabic Words

Abstract

Ten college students and ten PhDs read aloud 30 unfamiliar English words, two to five syllables in length, of Greek, Latin, and Germanic origin. The average number of different subject pronunciations per word was five (range one to ten). Each response was compared to the rule-predicted (Venezky, 1970), dictionary-prescribed, and most frequent pronunciation for that word. The subjects agreed more with each other than with the dictionary, and with the latter more than with the rules. However, the rules predicted half of the prescribed pronunciations, which was better than the average number of individual subject's responses agreeing with the dictionary. The most frequent response to each word demonstrated considerably more agreement with both the dictionary and the rules than did the average number of responses for the subjects individually. The etymological source of the test words had no effect. While the PhDs as a group did significantly better than the students on most measures, the differences were small. In view of previous research, the frequent vowel laxing in open third and fourth syllables from the end was unexpected. Several models of reading are examined in the light of these results.