The last decade of empirical research on the added value of HRM, also known as the 'HRM and performance' debate, demonstrates evidence that HRM does matter (Huselid, 1995; Guest et al, 2003; Wright et al, 2003). Unfortunately, the relationships are often statistically weak and the results ambiguous. This article reviews and attempts to extend the theoretical and methodological issues in this debate. Its aim is to build an agenda for future research in this area. A brief overview of achievements to date is followed by the theoretical and methodological issues related to what constitutes HRM, what is meant by the concept of performance and the nature of the link between these two. In the final section, it is argued that research designs should start from a multi-dimensional concept of performance, including the perceptions of employees; and build on the premise of HR systems as an enabling device for a whole range of strategic options. This implies a reversal of the strategy-HRM linkage.