-19-
developed, the more interviewers tend to maximize their own utility at the expense of
research interests (sloppy work, outright cheating, reduced output).
(2) These processes (alienation) in a survey organization lead to an increase in 'invisible
errors: interviewers learn to avoid visible traces of 'inadequate behavior.
(3) Supervisors will not be aware of these processes because they concentrate on visible
sources of error. They will reinforce interviewers 'strategic' conformity to the rules while not
being able to reward 'authentic' research goal-oriented behavior (Roth 1966 119781, Argyris
1968, 1980).
While I will not be able to test these hypotheses explicitly, I will address some related
questions. The approach described above suggests a tension between appearances and meaningfulness
of data. " Are these two aspects of data quality really different in terms of response effects influencing
them, if any? Do invisible errors increase over time due to alienation processes while visible errors
decline due to unintended learning processes? Survey researchers should be concerned with these
questions both for evaluating and monitoring survey data quality.
2My study of the survey methodology literature did not provide me with an adequate
terminology tapping these two aspects of data quality. By analogy, they might be viewed as syntax
and semantics. A sentence with a perfect syntax might be utterly meaningless when the semantics
are wrong and vice versa. Although both are necessary for assessing the quality of a text, they are
independent concepts.