The purposes of audiotaping of social survey interviews for evaluation can be formulated easily: locate prevalent response errors, locate prevalent recording errors, assess misunderstanding between interviewer and interviewee, measure the "cost" in terms of time of each question or group of questions on the questionnaire.
It is more difficult to translate these purposes into testable hypotheses, and still more difficult to devise means of measuring the taped data in an objective manner. Furthermore, the difficulties of execution of such an exercise are considerable and require, for success, a set of favourable circumstances that occur only fortuitously.
Working out objective means of analysis is fundamental.
Otherwise, the analysis deteriorates into the pursuit of features that may be anthropologically interesting but are largely impressionistic and statistically nonquantifiable.
An interestingly written paper, spiced with the more bizarre citations from the tapes, may be merely depicting events beyond two sigmas that are unrepresentative and in their total impression utterly misleading. Such analysis will then conclude with the trivial advice that "enumerators should be better trained". The empirical data used in this note come from Gambia and Morocco where interviews were taped respectively during the 1972 pilot census taken for the 1973 national census and the national census of 1971. The author had personal experiences in Morocco (Quandt, 1972) and had access to documents from both countries made available by the statistical authorities concerned.
Apart from the impressionistic listening, which leads to misleading results, four means of analysis are available: timing-counting-measuring,