[Excerpt] A few trends have emerged in the field of Strategic Human Resource Management (SHRM) over the past few years. First, and most obviously, has been the extensive effort to demonstrate a link between HRM practices and firm performance (Becker & Gerhart, 1996). Researchers such as Huselid (1995), MacDuffie (1995), Delery and Doty (1996), and Guthrie (2000) have published empirical studies showing a statistically significant linkage between HRM practices and some measures of organizational performance. A second trend has been to try to understand the mechanisms through which this relationship takes place. Authors such as Becker & Gerhart, (1996), Dyer and Reeves (1995), Guest, (1997) and Wright and Gardner (2003), have all called for research that uncovers some of the mediating relationships that must exist between the HRM practices and organizational performance. A final trend has been the recent interest in taking a multi-level approach to understanding SHRM. Wright and Boswell (2001) reviewed the SHRM literature and categorized this research as being differentiated along one dimension representing whether the focus was on single or multiple practices, and along a second dimension dealing with the unit of analysis, specifically the individual versus the group or organization. Ostroff and Bowen (2000) and more recently Bowen and Ostroff (2004) have developed the most extensive multi-level model of SHRM to date. Their theoretical approach argues that HR practices serve as communications mechanism signaling employees to engage in certain behaviors; relying on communications theory they contend that different aspects of HRM systems impede or facilitate this communication process. The purpose of this paper is related to these last two trends: we conceptually examine some of the mediating processes that might occur in the HRM – performance relationship, and try to make explicit their multilevel nature. In order to accomplish this, we will first explore the concept of variance, which is crucial to the analysis of any phenomena across multiple levels. We will show how virtually all existing SHRM research focuses on variance at one level of analysis while assuming constancy at other levels. We will next discuss the process through which HRM practices must act, and identify some of the relevant variables that have heretofore been virtually ignored in the empirical SHRM literature, specifically focusing on variance at different (unit vs. individual) levels of analysis. Finally, we will present some implications for theorizing and research in this area.