Raskind, M. H., & Higgins, E. (1995). Effects of speech synthesis on the proofreading efficiency of postsecondary students with learning disabilities . Learning Disability Quarterly , 18 (2), 141–157. https://doi.org/10.2307/1511201

Participants were given the choice of writing by hand or using a word processing program (without spell-checking). Participants returned for a second session to proofread and locate errors in their essays under three conditions. The first was using a speech synthesis/screen review system that enabled participants to select text on the computer screen and hear the words spoken as they were simultaneously highlighted. It was possible to review the text by word, line, sentence, or paragraph. Students could modify the rate of speech, volume, pitch, and the colors of the background and highlighted text for maximum contrast and readability. The second was having the text read aloud by a human reader. The third was having no assistance, and proofreading the hard copy independently. Also, no time constraints were placed on the participants in any of these conditions.

The study included 33 students with learning disabilities (19 male and 14 female) at California State University, Northridge (CSUN). Students were 19 to 37 years old with a mean age of 24.9 years. Participants were predominantly Caucasian and middle class with 25 identified as Caucasian, 4 as Hispanic, 3 as African-American, and 1 as Asian-American.

Participants wrote an essay of three to five typewritten pages on a topic of their choice or from a list of six topics. Nine categories were used to score the essays: capitalization, punctuation, spelling, usage, grammar- mechanical, grammar-global, typographical, content/organization, and style. The total number of errors found by each subject was divided by the number of errors found by the raters. This resulted in the percentage of total errors found by each subject for each condition.

Results indicated that under the speech synthesis condition, participants found significantly more of the total errors (35.5%) than in either the read-aloud (32%) or the no assistance (25%) conditions. The difference between the read-aloud and the no assistance conditions also was significant. The use of a speech synthesis system also outperformed the other two proofreading conditions in seven out of nine categories of written language errors, four of them at a statistically significant level.