Timeline Follow-Back (TLFB) is a widely used, calendar-based measure of self-reported use of (among other things) illicit substances. We examined agreement between TLFB and biological measures for illicit substances. PubMed, PsycINFO, Cochrane CENTRAL, and EMBASE were searched in December 2010. 16,633 papers screened to identify those that measured illicit substance use by both TLFB and biological measures. We extracted data on agreement between TLFB and biological measures, sample size, study type, inclusion criteria of participants, and length of recall of TLFB. Twenty-nine papers were included, almost exclusively in substance-use-disorder populations. Some studies reported several overall agreement rates, e.g. over time. Lowest and highest weighted average agreement rates were: for cannabis, 87.3% (95% confidence interval 86.9% to 87.7%) and 90.9% (90.5% to 91.4%); for cocaine, 79.3% (79.1% to 79.6%) and 84.1% (83.9% to 84.2%); for opiates 94.0% (93.5% to 94.5%) for both weighted averages; and for studies not distinguishing between substances, 88.5% (88.4 to 88.7%) and 91.0% (90.7% to 91.2%). Higher agreement was found in populations without psychiatric comorbidity, and lower agreement in randomized controlled trials. Publication bias or selective outcome reporting bias was not detected. TLFB validly detects use of illicit substances in populations with substance use disorders. Using TLFB may limit the need for biological samples, making information on illicit substance use easier and less costly to obtain and analyze. Copyright © 2011 Elsevier Ltd. All rights reserved.