RESNA 26th International Annual Confence

Technology & Disability: Research, Design, Practice & Policy

June 19 to June 23, 2003
Atlanta, Georgia


John Todman and Halina Rzepecka
Department of Psychology
University of Dundee
Dundee, DD1 4HN, Scotland


For people who use whole-utterance AAC devices, the pauses preceding their utterances set a limit on their conversational rate and have a negative affect on their perceived communicative competence. Average pause times preceding utterances in recorded social conversations by people using a text-based AAC device (TALKBoards) were experimentally manipulated by replacing natural pauses with pauses of specified lengths (2 s, 6 s, 10 s, Natural = 16 s). Communicative competence ratings of these modified conversations showed a highly significant linear trend, with ratings increasing from the 16 s condition through the 2 s condition. This has important implications for AAC system design.


Word-by-word generation of utterances at the time they are needed in a conversation is extremely slow for most people who use an AAC system with synthetic speech. Even with word prediction facilities, typical conversational rates are too slow to sustain effective social conversation. This tends to result in negative evaluations of intelligence, social competence etc.

Systems designed for the storage of whole utterances in preparation for output during a later conversation, may, in principle, result in much faster conversational rates. This potential is achievable when pragmatic features are "built into" the device to make it relatively easy to anticipate utterances that are likely to be useful in later conversations and to find an appropriate utterance when it is needed. Conversational rate has been greatly increased in this way for social conversations (1, 2) and transactional conversations (3). This rate enhancement has been achieved without compromising the quality of the conversational content (4).

A question that arises is the extent to which more positive attributions of communicative competence are made when conversational rate is increased. For text-based systems that use word-by-word entry during a conversation, conversational rate depends primarily on the pauses between the entry of individual letters and words. For whole-utterance systems, however, it is primarily the average length of pause between utterances that determines conversational rate. So, for whole-utterance systems, the question becomes, "does average pause time preceding user utterances affect perceptions of communicative competence?"

There have been correlational studies linking pause times with social competence, enjoyableness etc. (1, 5, 6), but an experimental manipulation of pause time in recordings of whole-utterance aided conversations failed to find clear evidence of a positive effect of pause time on ratings of communicative competence (7). However, that study involved transactional conversations in which the goal was to purchase a particular book at a bookstore.


The present study investigated the experimental effect of variations in average pause time on perceptions of communicative competence in social, as opposed to transactional conversations. It was hypothesized that, for social conversations, where the goals are enjoyment of the interaction, creating a favorable impression, etc., shorter average pause times would result in more positive attributions of communicative competence.


Three participants who were unable to speak due to cerebral palsy each used TALKBoards (a whole-utterance AAC system) to have social, "getting to know you", conversations with three speaking partners. The conversations were audio-recorded and five- minute extracts from the middle of each conversation were used. Pauses preceding the AAC users' turns at speech were replaced with pauses whose average length was 2 s, 6 s, and 10 s. Twenty-eight volunteer raters listened to each of the nine conversation extracts. For each extract, a rater heard it in one of four conditions with a specified average pause time (2 s, 6 s, 10 s, Natural = 16 s). The raters expressed their degree of agreement with 14 statements on a seven-point scale (1 = total disagreement, 7 = total agreement). The 14 statements comprised a communicative competence scale.


Inter-rater reliability was high (Kendall coefficient of concordance = .91, p < .001) and the internal consistency of the scale was satisfactory (coefficient alpha = .89).

Figure 1. Mean communicative competence ratings of social conversations with four average lengths of pause preceding the AAC user's turns at speech
Figure 1
The mean communicative competence ratings for the four pause time conditions are:
Condition		Rating
2 s			6.03
6 s			4.92
10 s			3.20
16 s (natural)		1.98

Mean communicative competence ratings for the four pause conditions are shown in Figure 1. A one-way repeated measures analysis of variance of the ratings was carried out, using the Greenhouse-Geisser correction for lack of sphericity. This found a statistically significant effect of pause time (F (2, 63) = 996, p < .001) and the linear trend that is apparent in Figure 1 was significant (F (1,27) =2145, p < .001).


The hypothesis that more positive attributions of communicative competence would be made as the average pause time preceding AAC user utterances in social conversations was experimentally reduced was supported. There was a clear linear trend over the full range of average pause times used in the study.

The theoretical implication of this result is that any consideration of the effectiveness of different trade-offs by AAC users between the competing demands of speed and precision of utterances (7) will need to take account of the goals motivating the conversation in which a given trade-off occurs. Goals that are typical for transactional conversations, such as transmitting information accurately in order to accomplish a task, may be expected to make a trade-off favoring precision more effective. On the other hand, goals typical of casual social conversation, such as enjoying the interaction and creating a favorable impression, may be more likely to result in trade-offs favoring speed being more effective. Future research will need to consider the dynamic balance between the demands of speed and precision in conversations in which the importance of transactional and social goals varies over the course of the conversation.

The practical implication of the results of the present study concerns the design features required in AAC devices that are intended to support conversations of different kinds, in different contexts. It seems that, where it is intended that a device should support social communicative goals, effective whole utterance features should be well represented. 


  1. Todman, J. (2000). Rate and quality of conversations using a text-storage AAC system: A training study. Augmentative and Alternative Communication, 16, 164-179.
  2. Todman, J., Rankin. D., & File, P. (1999). The use of stored text in computer-aided conversation: A single-case experiment. Journal of Language and Social Psychology, 18, 287-309.
  3. Moulton, B. (1999). Frametalker: An utterance-based augmentative device. (Final Report, NIH SBIR Phase I Grant 1R43-CA80715-01), Lockport, NY: Enkidu Research.
  4. Todman, J., Elder, L., & Alm, N. (1995). An evaluation of the content of computer-aided conversation. Augmentative and Alternative Communication, 11, 229-234.
  5. Todman, J., & Lewins, E. (1996). Conversational rate of a non-vocal person with motor neurone disease using the 'TALK' system. International Journal of Rehabilitation Research, 19, 285-287.
  6. Todman, J., Lewins, E., File, P., Alm, N., & Elder, L. (1995). Use of a communication aid (TALK) by a non-speaking person with cerebral palsy. Communication Matters, 9, 18-21.
  7. Bedrosian, J.L., Hoag, L.A., & McCoy, K.F. (2002). Communication rate and the use of utterance-based technology. Proceedings of the 10th Biennial Conference of the International Society for Augmentative and Alternative Communication (pp. 129-130). Odense, Denmark, ISAAC.


The authors wish to thank the three participants who used the AAC device to provide the conversations. Thanks also go to their vocal partners and the volunteer raters.


Prof. John Todman
Department of Psychology
University of Dundee
Dundee, DD1 4HN, Scotland, UK

Phone: 011 44 1382 34461
Fax: 011 44 1382 229993

RESNA Conference Logo