The development of a voice-recognition Census short form represents the Census Bureau's first attempt to simulate the verbal transaction between a live interviewer and a respondent. The goal of our research was to learn how respondents felt about this simulation. Relative to the interviewer- administered questionnaire, respondents in this study reacted quite favorably towards the voice-recognition questionnaire. They ranked the voice- recognition questionnaire just slightly below a comparable interviewer- administered questionnaire in terms of most characteristics. They anticipated a touch-tone system and were clearly pleased with the voice-recognition system from that perspective. Despite the system's achievements, respondents identified two areas in need of improvement. They were: 1) the error rate and 2) the pace of the interview. Respondents also made some excellent suggestions for decreasing the error rate and increasing the pace of the interview. Although we found nothing in our research to suggest that, in general, respondents would refuse to be enumerated simply because we offered them the option of answering the census by computer, the limitations of this research must be kept in mind. Experimental research with a representative probability sample is necessary before we can confidently conclude what the effect of this new technology would be on response rates.
