Predicting Reading Error Rates With Frequency, Function, and Sound-to-Spelling Rules

• In Event: Roundtable Session 29
  In Roundtable Session: Examining Literacy Issues and Practices in Elementary School

Abstract

While phoneme-to-grapheme correspondences have been studied with regard to curricula comparison (Ehri et al. 2001) and expected errors (Labov et al. 1998), there are few if any analyses on if and how these correspondences predict errors. We analyze 408,253 tokens of correctly and incorrectly read words from 446 African American adults. Phoneme-to-grapheme categories significantly predict errors only for readers whose error rates are above 30%. Frequency affects readers with error rates above 10%, while readers with error rates below 10% are unaffected by complexity and frequency. We demonstrate that complexity and frequency predict reading accuracy, and will use these results to develop an online interface that teachers and parents can use to identify, understand, and target early readers’ errors.

Authors

• Brittany McLaughlin, Swarthmore College

• John R Rickford, Stanford University