Addressing adherence to medication is essential and notoriously difficult. The purpose of this study was to determine physicians' ability to predict patients' adherence to antihypertensive therapy. Primary care physicians were asked to predict the adherence to medication of their hypertensive patients (n=42) by using a visual analogue scale (VAS) at the beginning of the study period. The patients were asked to report their adherence to medication using a VAS. The adherence was then monitored by using a Medical Event Monitoring System (MEMS) for 42+/-14 d. The means+/-SD (range) of MEMS measures for timing adherence, correct dosing, and adherence to medication were 82+/-27% (0 to 100%), 87+/-24% (4 to 100%), and 94+/-18% (4 to 108%), respectively. The physicians' prediction of their patients' adherence was 92+/-15%. The Spearman rank correlations between the physician's prediction and the MEMS measures of timing adherence, correct dosing, and adherence to medication was 0.42 (p=0.006), 0.47 (p=0.002), and -0.02 (p=0.888), respectively. The patients reported their own adherence to medication at 98+/-2% (range 83 to 100%). The Spearman correlations between the reported and actual behaviours were 0.27 (p=0.08) for timing adherence, 0.25 (p=0.12) for correct dosing, and 0.11 (p=0.51) for adherence to medication. The physicians' ability to predict patients' adherence to antihypertensive medication is limited and not accurate for identifying non-adherent patients in clinical practice. Even patients themselves are unable to give accurate reports of their own adherence to medication.