The analysis examined data obtained from 335,940 adult (18 years old and older) arrestees. Most of these sites collected data every year through 2001. Three procedures were used to examine the degree of agreement in reported and detected use of marijuana over time: time-variation, regression, and graphical trend. Analyses were performed separately for each of the 23 ADAM sites. The disclosure and overreporting of marijuana use among arrestees varied significantly across sites and over time. Overall, 60 percent of the variation in reported use reflected changes in the marijuana use detected through urinalysis. Trends in reported and detected use of marijuana suggested different dynamics for the increase in marijuana use during the 1990s. The increase derived from urinalysis testing started later than that revealed through self-reports, increased more, and lasted longer. Several factors apparently contributed to this divergence between the two measures: the percentage of marijuana users who self-reported their use changed over time; the accuracy of ADAM's urinalysis test increased over time; and the percentage of infrequent users changed over time. The latter factor is significant, because urinalysis tests are less likely to detect infrequent users. Five recommendations address inaccuracies in biological drug tests: use a more accurate test, such as gas chromatography/mass spectrometry; monitor drug-potency changes and infrequency of use; confirm trends with multiple indicators; place greater confidence in trends that involve large changes over time; and treat indicators as proxies. 1 table, 4 figures, 28 references, and appended discussion of calculations of the relationship between reported and actual use