Studies of mental health services have emphasized that people in need are not receiving treatment. However, these studies, based on self-reported use, may not be consistent with administrative records. This study compared self-reports of mental health service use with administrative records in a large representative sample. Respondent reports within the Ontario portion of the 1994/95 Household Component of the National Population Health Survey (NPHS) were individually linked to the provincial mental-health physician reimbursement claims. A total of 5,187 Ontarians, aged 12 years or more, reported on their use of mental healthcare within the NPHS and 4,621 (89%) consented and were successfully linked to administrative records. Comparisons between the two sources identified that the agreement for any use and volume of use was moderate to low and varied according to select respondent characteristics. These differences affected estimates of the associations with use and volume of use. People who reported high levels of distress reported more visits than those who did not and this effect was stronger in the self-reported data. These results suggest that recall bias may be present. Regardless of the definition of care, access for those in need remains a concern despite universal medical insurance coverage.