One response to the possible outsourcing of the human resource (HR) function is to turn to 'science' and seek to demonstrate an empirical association between HR practices and increased organizational performance. This paper critically examines the shortcomings of the 'scientific' approach by first, reviewing the three distinctive versions of research on the Human Resources Management (HRM)-Performance link in an attempt to demonstrate their commitment to a common 'scientific' meta-theory. Second, we use critical realism to demonstrate: (1) that theoretical underdevelopment and lack of explanatory power are encouraged by the use of an inappropriate 'scientific' meta-theory; (2) the possibility of meta-theorizing the causal connection between HRM and performance without seeking statistical associations; and (3) how all this is in-keeping with Institutional theory. Finally, all of this is achieved by inserting evidence from interviews with HR professionals to demonstrate not only that they are sceptical of the 'scientific' approach, but also that they hold views of the world not dissimilar to the critical realist approach we advocate. Volume 13(5): 677-699