Computer-assisted accent modification is a promising new technology that may increase the efficiency of training by providing users with customized drill opportunities.

Accent modification has traditionally been carried out in one-on-one or small group teaching situations. As such, it is time-consuming and of doubtful efficacy. It also depends on the trainer having some level of competence in phonetics or speech-sound production.

A previous study (Ferrier, Walsh and DiGiacomo, 1993) indicated that a group of international teaching assistants improved their speech-sound production using the Speech Works accent modification software in one-on-one sessions with a speech-language pathologist (student trainer) more than a control group who were simply encouraged to practice independently.
.

This follow-up study asked the following questions:

• Would ESL students provided with training in CAAM show greater improvement than a control group who simply attended ESL classes?

• Would automatic monitoring of the time spent in independent practice increase gains?

• Could CAAM be provided by a person without training in speech-language pathology?
.

The program used, Speech Works, was designed for high school students and above and was developed to train large numbers of nonnative speakers with limited human resources available.

The program contains model words and sentences (both male and female voices) recorded in natural-sounding, digitized speech. Using the sound and recording capabilities of computers, users can listen to, practice and record their own versions and then compare the two.

Self-recording and comparison with a digitized model form an essential part of using the program. Speech Works also allows user progress to be followed and hence provides a built-in instrument for documenting efficacy.
.

A. The Subjects

The subjects included thirteen ESL students enrolled in the beginning level (ESL 099) course at Dean College. Twelve of the students were Japanese and one was Chinese. All students had been learning English in their home countries for six to seven years. Most had been speaking English on a regular basis for less than one year.

B. The Trainer

The trainer was a professor with a background in voice, theater and public speaking, but with no training in phonetics, speech-language pathology or ESL.

C. Procedures
The subjects were assigned to one of three groups, two treatment groups and one control group.

• Control Group (3 subjects, all Japanese, mean age: 20) No intervention, but attended ESL classes.

• Unmonitored (5 subjects, 4 Japanese and 2 Chinese, mean age: 19) One-on-one weekly sessions and expected to practice a minimum of one additional hour per week. The mean number of one-on-one sessions for the unmonitored group was 10.40.

• Monitored (4 subjects, all Japanese, mean age:21) One-on-one weekly sessions and expected to practice a minimum of one additional hour per week. Independent practice time was recorded by Speech Works and reported. The mean number of one-on-one sessions for the monitored group was 10.00. The mean number of minutes spent in independent practice by the monitored group was 68.42 per session as measured by the automatic program log in Speech Works.

D. Measures of Progress
All subjects were measured pre and post with two computer tests, the Speech Works Diagnostic Test and the Speech Works Screening Test. Content and scoring of the two tests is summarized below.

E. One-on-One Session Content

Both treatment groups, monitored and unmonitored, followed the sequence of lessons for each student provided by the Speech Works Diagnostic Test. The trainer provided speech-sound positioning cues based on his intuitions as a speaker, as well as auditory and visual cues for correct intonation and stress placement. Students were provided with individualized practice, targeting sounds in their plan to practice independently (using the Speech Works Personal Practice program). The trainer required students to submit recorded sounds on disk from their independent practice for review and feedback. Students received a grade for each lesson that contributed to their overall (ESL 099) course grade. In the monitored group only, the students were informed of the total time they had logged in independent practice.

Both groups, monitored and unmonitored, improved significantly more from the pre-tests to post-tests than did the control group.

In the Diagnostic Test (Table 2), the unmonitored group changed on average 6.4 points and the monitored group by 11 points compared to the 1.33 points by the control group. There was a similar pattern in the progress of the three groups as measured by the Speech Works Screening Test although these changes were not statistically significant.

On average, the unmonitored group changed by 6.25 points, the monitored group by 10.75 points as compared with a change of only 3 points by the control group.
.

We conclude that CAAM was effective. The effectiveness of the training was enhanced when monitored by an automatic log. And, this training could be provided effectively by a trainer without a specific background in ESL, speech-language pathology or phonetics.
.