Since the 1970s the United States and other nations have conducted reg
ular statistical monitoring of the prevalence and patterns of drug use
in their populations. Given the importance of such surveys for policy
making, their quality is a critical issue, and the biases that may aff
ect their measurements become a major concern. An increasing volume of
empirical evidence shows that the mode of administration of a survey
can strongly influence the validity of respondents' reports. Compared
with interviewer-administered questionnaires, self-administered forms
appear to elicit more complete reporting of drug use, but the challeng
es they pose to the literacy skills of respondents may result in measu
rement biases. In addition, processes of social change may confound tr
ue shifts in drug use with changes in the willingness of respondents t
o report such use. The authors propose several strategies to improve m
onitoring of trends in drug use. Those approaches include 1) more freq
uent use of a survey technology-audio computer-assisted self-interview
ing-that ensures full privacy for all survey respondents but does not
require literacy; 2) increased use of time-series of indicators of dru
g use consequences built from blinded surveys of medical records; and
3) population-based surveys that collect biological specimens (e.g., h
air samples). Data from the latter two sources are not subject to the
same constellation of biases that afflict self-reports of drug use. Ti
me-series of those data can be integrated with self-reports to provide
a better understanding of changes over lime in the prevalence and pat
terns of drug use.