Whilst steady progress has been made in the development and dissemination of valid and reliable instruments for evaluation of patient outcomes, progress in establishing standard methodologies for description and classification of mental health services has been limited. Valid and reliable methods of describing and classifying services are needed in mental health services research for purposes including; i) comparison of levels of provision between catchment areas and countries; ii) reaching an understanding of the relationship between socio-demographic indicators and service use; iii) investigation of the degree to which one service may be substituted for another; iv) evaluating programme implementation; and v) understanding why different outcomes are observed from apparently similar treatment programmes. Where programmes of research encompass different countries, the need for methods of comparing the local service contexts is particularly acute. In this paper, the various contributions which have been made to the somewhat fragmented body of research in this area are reviewed, and the problems which have hampered the development of satisfactory instruments are discussed. A set of criteria for satisfactory service measurement methodologies is proposed: these include development of standard nomenclature and operational definitions of service types; clear demonstration of validity and reliability; comprehensiveness; and applicability and similar interpretation in a range of countries.