The mother tongue of the majority of black workers in the industrial workforces of the Witwatersrand is either Zulu or S Sotho/Tswana, and these are also the main languages spoken in the area. However a substantial minority cite one of several other languages as their mother tongue. In consequence, in occupational surveys using a respiratory questionnaire, the accepted practice has been to use multilingual interviewers who administer the questionnaire in whichever of the two main languages the interviewee requests, but work from an English language version. To test the reproducibility of the information so gathered, a standardized questionnaire was administered to 56 subjects on two occasions approximately three weeks apart by each of two interviewers, one of whose mother tongue was Zulu and the other S Sotho. Reproducibility was little affected by whether the interviewer used her mother tongue or not, pointing to the multilingual competence of both interviewers. Reproducibility was, however, less in subjects not interviewed in their mother tongue compared to those who were. In general, reproducibility was also less for symptom than for history questions, though the phlegm question performed comparably to that reported for more homogeneous language/ethnic groups. Thus, even if internally valid, respiratory symptom information obtained by this method may have limited generalizability. However respiratory history information appears more reliable and may prove of value in the investigation of the natural history of chronic lung disease in this and similar communities.