Human Resource Management–Performance Research: Is Everyone Really on the Same Page on Employee Involvement?

Differences in the treatment of involvement in the human resource management (HRM)&#8211;performance research stream have been underplayed, as commentaries concentrate on showing that HRM produces a performance premium, and more recently on exploring the mechanisms explaining this. This paper first identifies the two initial concerns of the research stream &#8211; the value of employee involvement and the holistic treatment of HRM &#8211; and the way these are joined to present a unified view of the area. It then reviews the studies, confirming that involvement has been underplayed or neglected completely, and is only prioritized in a minority. A divide is identified between HRM as an orientation towards fostering employee involvement &#8211; seen as a managerial philosophy &#8211; and as a technology &#8211; a set of practices constituting high&#8208;performance work systems. The paper then argues that acknowledgement of this divide matters, and concludes by drawing out some implications for how we should progress the research stream.

A strong relationship between human resource management (HRM) and the performance of an organization is now generally accepted by HRM academics and practitioners, as is the view that this relationship is supported by a solid evidence base, namely a stream of research studies which emerged in the 1990s that used quantitative methods to evaluate the relationship between the modernization of HRM and organization performance. This modernization entailed two elements: a focus on employee involvement and the holistic coordinated use of human resource practices, including involvement practices. Reviews of the stream, and introductions to later studies, present it as a largely homogeneous set of studies (e.g. Combs et al. 2006;Guest 2011;Wall and Wood 2005;Wright and Gardner 2003). However, employee involvement has been underplayed or I thank Nick Catley for his editorial assistance. neglected in many studies and the focus has been on the combined use of a set of practices that often excludes involvement in favour of skill-acquisition and motivation-enhancing practices (Wood and Wall 2007). The objectives of this paper are to: (1) review the treatment of involvement in the HRMperformance studies; (2) expose how the homogeneous portrayal of the HRM-performance research stream is inaccurate; (3) identify how the neglect of involvement reflects a schism between centring on employee involvement and on high-performance work systems; (4) suggest how a lack of appreciation of this has stymied the development of the field and led to an overconcentration on the mediation issue at the expense of more fundamental concerns; and (5) draw out the implications for the field if we are to correct the neglect of involvement.

Introduction
Reviews of the HRM-performance studies tell the same story: HRM is good for the economic performance of firms and public sector organizations. The term 'HRM' can be used to refer either to a generic range of sophisticated personnel or human resource practices, or to a more specific approach built around employee involvement and development. Both concerns -the holistic approach to personnel management and employee involvement -were major impetuses behind both the HRM-performance research and its subsequent presentation as a unified stream. Within these, HRM is used in a way that twins these two concerns to signify a more strategic and holistic approach to personnel management, which entails a greater emphasis on employee involvement and development and diverges from a rigid Taylorist model based on a narrow division of labour and low employee involvement.
The portrayal of the field as self-contained and homogeneous assumes that the studies adequately measure such a concept and that the consistent finding of positive relations between the measures and performance can be readily generalized. Allied to this has been the use of high-commitment management, highinvolvement management, and other such terms as synonymous labels for the modern HRM (e.g. Jiang et al. 2012;Latorre et al. 2016). It is as if the proliferation of terms reflects a desire on the part of authors to differentiate their work, but they are all ultimately assumed to amount to the same, post-Taylorist HRM. The perceived security of the evidence for the superiority of this model is reflected in the widespread use of the term 'high-performance work system' (a term originated by the U.S. Department of Labor 1993) as one of the synonyms for post-Taylorist HRM. Paul andAnantharaman (2003: 1246) go so far as to say that its connection to performance has been 'proved'. Such confidence in the research has led to calls to move on to explain this relationship between HRM and performance, rather than simply proving its existence, as initial studies were criticized for treating this link as a 'black box' (Edgar and Geare 2007;Evans and Davis 2005). These calls have been heeded, as the field has become dominated by studies assessing mediators of the relationship (e.g. Beltrán-Martín et al. 2008;Jiang et al. 2012;Takeuchi et al. 2007). However, while this is important, a more fundamental issue of the variety of concepts of HRM across the studies remains.
Acknowledgement of variations between studies in reviews has largely been confined to differences in the practices included in HRM measures, with occasional consideration given to whether studies test contingent relationships between HRM and performance in addition to the universal one (Wood 1999a). The variation in practices has not been presented as a serious problem, as most studies are assumed to cover a sufficiently comprehensive set of these practices for the field to be presented in a unified way. It has become commonplace to accept that highperforming HRM is made up of three types of practices -skill acquisition, motivation, and opportunities for participation -a notion that has been codified in the Abilities-Motivation-Opportunities for Participation (AMO) framework (Appelbaum et al. 2000), and reviewers of the area appear to assume that most studies cover all three types. Nonetheless, in one review that focuses on the variation in practices, Wood and Wall (2007) show that this is inaccurate; in particular, employee involvement, the core of opportunities for participation, has been underplayed in a significant number of studies.
The focus of this paper is on the inconsistency in the treatment of employee involvement across the quantitative studies of the HRM-performance relationship and the implications of this. Employee involvement is widely seen as having two dimensions, defined by their focus. First is role or job involvement, which concentrates on employees' core jobs, and is concerned with ensuring they have an element of autonomy and responsibility in their jobs (Wall et al. 2004;Wood et al. 2012). Second is organizational involvement, which entails employees participating in decision-making beyond the narrow confines of their jobs, so they are involved in work-organization decisions, other immediate aspects of their environment, and the 'business as a whole' (Benson and Lawler 2003: 156);or, in Wall et al.'s (2004: 19) terms, employees having 'a say in decisions about the management and strategy of their organization'. This is often referred to as employee participation. Throughout the paper we will adopt the contrasting role-and organizational involvement terminology.
Organizational involvement is itself differentiated by whether employees are directly involved or involved indirectly through trade unions or other representatives. The focus in the HRM-performance stream has been on direct involvement, potentially as a competitor to formal processes, but it need not be 410 S. Wood at the expense of these (Walton 1985;Wood 2013). Nonetheless, a key element of the argument for the modernization of HRM is that giving employees a voice in matters beyond their wages and other terms and conditions of employment is crucial for enhancing organizational performance. This argument has concentrated on voice and participation through briefings, idea-capturing schemes, team working and job design (Markey and Townsend 2013), and focused on improvement-oriented voice, which is at the centre of Morrison's (2011: 375) definition of voice as 'discretionary communication of ideas, suggestions, concerns, opinions about work-related issues with the intent to improve organizational and unit functioning'. Such a definition does not imply that the improvement must be on managerial terms and hence an underlying assumption that employees and employers have the same interests, as is too readily assumed by those who dismiss HRM as a unitary management ideology (Wood 2013: 101).
I focus this review on employee involvement for two main reasons. First, the concept enjoyed wide currency in academic and practitioner circles when the HRM-performance research began. It was part of an argument for a more participative management -in McGregor's (1960) terms 'Theory Y' management -that could herald a personnel management that had a significant bottom-line effect (Pfeffer 1998). Organizational involvement was being given prominence in socio-technical theory (Trist and Bamforth 1951), theories of job autonomy's effects on employee well-being (Hackman and Oldman 1980;Karasek 1979), the Japanese management and Total Quality vogues (Wickens 1995;Womack et al. 1990;Wood 1989), and Peters and Waterman's (1982) In Search of Excellence. Alongside these developments, it was argued that leading-edge practice in firms like IBM, the role of employee involvement in Japanese management methods, and examples of effective user involvement in new technologies all illustrated the merits of involvement. Discussion also connected with the changing nature of employment relations, which centred on direct and indirect methods of participation as alternatives to collective bargaining (Kochan et al. 1994).
Reflecting these concerns for employee involvement, the second reason for concentrating on involvement is that it is the cornerstone of two key concepts in the advent of the HRM-performance research stream: Lawler's (1986) high-involvement management and Walton's (1985) high-commitment management, which in turn have been equated with the notion of high-performance work systems in the formation of the unified account of the stream. The centrality of involvement to the inception of the HRMperformance research stream provides us with one benchmark for assessing variation within it. Moreover, Wood and Wall's (2007) analysis of the inclusion of employee involvement in the original HRMperformance studies suggests that its neglect may reflect more fundamental differences between the studies that go beyond their measures of practices. Nonetheless, this analysis needs updating to include the mediation studies, as it is possible that the neglect of employee involvement is less prevalent -a task that I have conducted and report as part of this review.
My analysis entails a review of three types of literature: reviews of the field, the original HRMperformance (black-box) studies, and the HRMperformance studies that include mediators. I relied for the data on the original studies, primarily on other reviews (Boon et al. 2019;Combs et al. 2006;Wall and Wood 2005;Wood and Wall 2007). I conducted an original search of the mediation studies using sources including PsycInfo, Business Source Elite, Sage Journals Online, and reference lists of the articles reviewed. The inclusion criteria followed those of Wood and Wall (2007), namely refereed journal articles concerned with the relationship between multiple HRM practices and economic measures of organizational performance, with the addition that tests for the mediation of the practiceperformance relationship were included. The review is directed at a specific question: 'To what extent is there variation across the HRM-performance research stream in the treatment of employee involvement?' The analysis of the articles thus focuses on the substance of the literature and its treatment of involvement.
I first introduce the role of high-involvement management and holism in the birth of the research stream, and then discuss how it has typically been treated in the narratives about the stream's nature and development. Next, I present a narrative review of how employee involvement has been tackled in the empirical research. Having identified differences in the treatment of involvement, and how these reflect a distinction between high-involvement management (a managerial orientation centred on designing practices and expectations based on the principle of employee involvement) and high-performance work systems (a set of best practices), I discuss why acknowledging this distinction and variation within the field matters. Finally, I present some implications of this acknowledgement for how we might take the field forward.
Employee-involvement and holism concerns at the outset of the HRM-performance research stream I will now discuss the concerns from which the HRM-performance research stream developed: the need for employee involvement and the holistic use of sophisticated HRM practices in an increasingly competitive world.

The employee-involvement concern
The first concern for employee involvement was a reaction to the past dominance of Taylorist methods, with their tight division of labour, narrow job descriptions, and over-reliance on pay as a motivational tool and concomitant low workforce involvement and commitment. This concern emerged from Lawler's (1986) high-involvement management and Walton's (1985) high-commitment management, both of which were influenced by then-emerging best practice. The foundation of their concepts was the empowerment of employees at the job level, but they also embraced organizational empowerment and involvement. This addition of organizational involvement was the distinctive feature of high-involvement management relative to earlier concerns with job redesign. Above all else, high-involvement management required a deep philosophy, or set of principles to which management is committed, that guides the design of practices, reactions to key events, and everyday leadership behaviours. Following it entails that all practices, and not simply job design, are constructed with this involvement philosophy in mind (Walton 1985: 80). Walton's (1977) and Hackman and Lawler's (1971) earlier work had focused on job design and was consistent with the psychological literature, which was finding that job autonomy and variety enhanced employees' well-being and in turn their performance (Fried and Ferris 1987;Karasek 1979;Loher et al. 1985). As the experiments in redesigning jobs developed, however, the need to go beyond role involvement was identified (Bowen and Lawler 1992). Changes were required to ensure the job design worked and to extend involvement beyond its focus on the job. Successful job design required people to be trained differently, as well as changes in supervisory, appraisal, and selection systems. Employees also needed greater awareness of the wider context of their job in order to use their autonomy effectively. Moreover, giving individuals more autonomy in their jobs would not guarantee organization-wide innovation; it became clear that new quality and logistic methods that were allied to a quest for continuous improvement methods -variously known as lean production, Total Quality Management, and Toyotaismrequired inputs from all employees regardless of their level (Wood 1989). A wider organizational involvement through teamwork and idea-capturing methods was needed, even (or perhaps especially) where achieving high levels of role involvement was constrained, as in assembly-line operations.
Organizational-involvement practices such as idea-capturing schemes, problem-solving groups, and information sharing, along with job design, are then the core of high-involvement management. Lawler (1986) viewed these as forming the core power dimension of high-involvement management, the other dimensions being information, skills, and rewards. The crucial point is that the skills and motivation elements of the AMO model are supporting the central involvement element of high-involvement management. For example, training will be concerned with developing skills required for involvement, such as problem-solving and team-working skills.

The holism concern
The second concern that precipitated the stream, the need for holistic methods, was a reaction to management's use of human resource (HR) practicesoften not the most advanced ones -in a piecemeal or ad hoc way. On the research front it also represents a reaction to the predominance of individuallevel studies that examined the independent effects of job design, selection, appraisal, and training methods on individual performance. This led to a perceived need for research at the organizational level (more normally termed the firm level).
It was argued that organizations using the full complement of best practices in each area of HRM would outperform others (Fey et al. 2000;Guest and Hoque 1994;Guthrie 2001;Ichniowski et al. 1997). Such practices were seen as complementing each other, each filling a gap in personnel management left by the others, implying that the larger the number of practices used, the greater the performance S. Wood effects. The HRM system is a set of practices that has horizontal fit, so no practices jar or conflict with one another. Involvement practices need not be prioritized. This horizontal fit may also mean that the performance effects of each individual practice are enhanced by the presence of others, so interactions or synergistic effects exist between the practices.
In addition, concern was shown at the outset of the stream for vertical fit -that the HRM system should be aligned with the business strategy or competitive context of the organization. This implies a contingency approach to HRM, and thus any HRM model represents the high-performing option only in particular situations. Following this, high-involvement management would only yield high performance in certain contexts, typically taken to be ones where product differentiation and innovation are required (Schuler and Jackson 1987).

Unifying the two concerns
The employee-involvement and holism concerns have been combined in portrayals of the area, with reviewers creating the sense of a unified area through narratives joining them (Combs et al. 2006;Huselid 1995;MacDuffie 1995;Raineri 2017). The issue of whether HRM is best universally is typically resolved through assuming the practices included are optimal in their domain and form the coherent best-practice combination. Allied to this, Becker and Huselid (2006) and Lepak et al. (2006) have argued that the high-performance HRM model is universally appropriate, and external fit may be managed through tailoring the detail of the practices to the organization's strategy and competitive situation. Thus, for Lepak et al. (2006), high-performance work systems take different forms, one of which is high-involvement management. The device of assuming practices are tailored to circumstances is also used to reconcile the universal best-practice model with the demand for unique core competencies, which is associated with resource-based theory, since the fine tuning of practices ensures that these practices 'are deeply embedded in the organization' (Becker and Gerhart 1996: 782).
The treatment of high-involvement and highcommitment management as either synonymous with high-performance management or work systems, or as dominant forms of such a system, is a key element of the unified picture of the research stream: their superiority is enshrined in the high-performance labelling. Another contributing factor has been treat-ing the variety in the practices across the studies as not constituting a large problem, and one which can be resolved through greater unanimity on the identity of high-performance practices. We now demonstrate how the reviews of the HRM-performance field, while acknowledging the variety of practices and their measurement within studies, have not treated this variety as fundamental.

Reviewers' downplaying of the significance of variation in practices
The existence of a wide variety of practices across the initial studies in the field quickly became apparent. Dyer and Reeves (1995) identified this in the first review of the area -although limited to four studies, there were 28 different practices spread across them. The review focused on the holism concern and concluded that bundling of practices was superior to their isolated use. Significantly, Dyer and Reeves (1995) noted that all the studies included employee involvement, and that this could be integral to superior bundles -along with careful selection, extensive training, and contingent compensation. The next review, by Becker and Gerhart (1996), showed that the number of practices in the five studies reviewed varied from 4 to 11, and only 7 of a total of 26 practices were present in more than one study. They attributed this to a lack of agreement about what techniques constituted high-performance work practices, concluding that the variety of practices illustrated that 'researchers have much to learn about what constitutes a high-performance HR strategy' (Becker and Gerhart 1996: 784). In contrast to Dyer and Reeves (1995), involvement is downplayed, reflecting a focus on holistic HRM as a source of unique capability. The significant point for Becker and Gerhart was that regardless of what practices were included in the measure of HRM, the studies consistently found HRM had positive effects on performance.
Subsequent reviewers also gloss over differences in practices and convey a sense that this is not a fundamental problem. For instance, Wall and Wood (2005), having drawn attention to the range of practices, played down its significance, writing: 'Nonetheless, there is much commonality as studies typically cover a substantial range of the following: sophisticated selection, appraisal, training, teamwork, communication, job design, empowerment, participation, performance-related pay/promotion, harmonization, and employment security' (p. 435). Consistent with this, Boselie et al. (2005) argued that underpinning the commonality was an increasing acceptance of the AMO framework, and predicted that a consensus on high-performing practices would emerge as research tackled the mechanisms inside the black box. A year later, Lepak et al. (2006), having attributed the range of practices to a lack of strong theory underlying the choice of practices in the studies, presented the AMO framework to fill this gap, in effect attempting to hasten the uniformity of the research stream that Boselie et al. (2005) perceived was already underway. In so doing, Lepak et al. (2006) were implying that not all studies covered all the AMO elements; they were thus less sanguine than Wall and Wood (2005) about the multiplicity of practices.
The lack of a strong theory explaining the HRMperformance association, coupled with the lack of exploration of mediators, has been the focus of most subsequent reviews and, understandably, the focus of reviews in the mediation studies has been on their black-box nature. Paauwe (2009: 131-132), for example, urged that the pleas from some reviewers for better theory meant that the next step is to identify the intervening variables that explain the HRMperformance relationship. Introductions to the mediation studies continue to underplay the heterogeneity of the black-box studies, with no attention paid to the variety in the significance given to involvement.
The review by Wood and Wall (2007) is exceptional in highlighting the neglect of involvement. Examining the 27 studies that, by the mid-2000s, had regressed economic performance against measures of the multiple use of practices, it demonstrated that involvement's neglect, relative to the attention to the other dimensions of AMO, had become more prominent. They associate this 'marginalization' (p. 1366) of employee involvement with the increasing salience of resource-based theory, with its stress on unique competencies. They also suggested that the neglect of role involvement reflects the way the HRM-performance research stream was built on observations of leading-edge companies, 'where work redesign did not necessarily amount to work enrichment; and the opportunities for organizational voice … were limited' (p. 1367).
Despite Wood and Wall's (2007) intervention, the two most recent reviews of the area do not mention the underplaying of involvement in the research stream, and concentrate on the holistic concern. However, they differ significantly -one, a meta-analysis by Saridakis et al. (2017), continues to portray the field as unified, while the other, by Boon et al. (2019), hones in on the variation in studies. Saridakis et al.'s (2017) study concentrates on eight longitudinal studies; unwittingly, they expose the diversity of practices, as the studies use very different HRM concepts and practices, some of which bear little relationship to the AMO model (Kim and Ployhart 2014;Snell and Youndt 1995;Welbourne and Andrews 1996). Tellingly, involvement is only included in two studies (Guest et al. 2003;Sheehan 2014) and longitudinal studies where it is included are not (inadvertently, one assumes) in their study (Birdi et al. 2008;Cappelli and Neumark 2001). Boon et al. (2019) also concentrate on the holistic concern underlying the research stream, but their focus is on the different ways of approaching this concern, of which the variety of practices is only one part. Boon et al. (2019) nonetheless analysed the inclusion of practices over time, and show that agreement on them has declined over time. Notably for our present concerns, the percentage of studies including involvement practices declined significantly in the period 1991 to 2017. They also show that the use of the label 'high involvement' has remained stable and low (averaging 8%) compared with a rise of the use of 'high performance'. Significantly, though, Boon et al. (2019) do not draw our attention to these and other developments. Instead, for them the key issue is that most studies begin with an emphasis on the synergistic effects of practices yet do not adequately test it, as they examine linear relationships between composite measures based on aggregating practice use. This disjunction means all we can conclude from existing studies is that 'investments in some broad set of HR practices yields returns', but 'whether and how practices jointly [my emphasis] affect outcomes remains unclear' (p. 19). The urgent priority for Boon et al. (2019) is comparisons of different methodologies for capturing the holistic or systems concept of HRM, a priority which, I gauge, is in danger of underestimating the substantive issues about the nature of HRM.
I have shown that the neglect of involvement continues to be underplayed. Nonetheless, the case still needs to be made that this variety in the treatment of involvement is large, and is a genuine problem. I now assess the extent of the variety before outlining why acknowledging such differences is important.

Variation in the treatment of involvement
The limited coverage of employee involvement The variety of practices used means that no single practice is consistently included in the studies. Posthuma et al.'s (2013) analysis of a broad range of HRM literature, which goes beyond studies of the HRM-performance relationship, reviewed 181 articles published between 1992 and 2011 and identified 61 practices. In Combs et al.'s (2006) meta-analysis of 92 studies that examined the relationship between at least one HRM practice and organizational performance, the most popular of the 13 practices covered in five or more studies was included in only 31 of them. Significantly, this practice was incentive compensation. The next most popular was training (29). Participation was included in 18 studies, teams in 8 studies, and information sharing in 7. Job-design concepts were not even categories in the review. Wood and Wall's (2007) analysis, which concentrated on black-box studies that had measures of multiple uses of practices, revealed that none included a job-level measure of involvement or used the terms 'work enrichment' or 'job enrichment', and only three studies included a similar concept: Vandenberg et al. (1999) included empowerment, Fey et al. (2000) decentralized decision-making, and Guest and Hoque (1994: 12) 'jobs designed to make full use of skills'. Organizational-involvement practices were items in 80% of the studies, and included self-directed teams, idea-capturing schemes, and information-sharing practices. Even the widely cited pioneer studies (Arthur 1994;Huselid 1995;Ichniowski et al. 1997;MacDuffie 1995) excluded role involvement and neglected, to varying degrees, organizational-involvement practices.
The similar review I conducted on the 30 mediation studies I identified revealed a neglect of involvement practices in around half the studies. Nonetheless, a greater proportion of studies included a measure of work enrichment than was the case amongst the black-box studies. Job autonomy or empowerment was measured in 13, while one measured 'interesting and varied jobs' (Latorre et al. 2016) and another 'high quality work' (Bonias et al. 2010). In contrast, the proportion with a measure of organizational involvement was lower than for the black-box studies -63%. Self-managed teams or similar concepts like decentralized decision-making were included in nine studies, suggestion-making in six, and participation in four. Around half of the studies attempted to cover all AMO dimensions (e.g. Bonias et al. 2010;Hauff et al. 2018;Piening et al. 2013;Raineri 2017). The majority of the others focused on one or both of skills-and motivation-enhancing practices (e.g. Liao et al. 2009;Mansour et al. 2014;Martínez-Del-Río et al. 2012;Paul and Anantharaman 2003;Takeuchi et al. 2007). Within these are studies that include involvement in their opening discussion without including a measure of it (e.g. Farouk et al. 2016).
Especially striking is the increase, compared to the black-box studies, in the use of the label highperformance work system (or practices). Five of the 27 black-box studies used it (Wood and Wall 2007), compared with 17 of the 30 mediation studies. Moreover, the emphasis in the majority of studies is on the holistic focus in defining their HRM concept. For example, Raineri (2017: 3150) justifies his preference for high-performance work systems on the grounds that the practices 'work in an interrelated manner when affecting employees' performance' through impacting not only 'employees' motivational states but also … their knowledge and skills'. Martínez-Del-Río et al. (2012: 829) are unique in mentioning both holism and involvement in their definition of HRM (labelled high-involvement work practices), defined as 'mutually reinforcing and synergistic HR practices that are intended to acquire, train, retain, involve and motivate employees and to improve communication'.
High-involvement management is also neglected in the theoretical discussions, which are dominated by resource-based theory and social exchange theory. Employee involvement or empowerment is, however, a mediator in five studies (e.g. Bonias et al. 2010;Liao et al. 2009), and related concepts involving proactive behaviours such as innovative work behaviour (Fu et al. 2015) and organizational citizenship behaviour (Messersmith et al. 2011) are mediators in three studies. The range of mediators is large, adding to the variation in the field. Only three mediators figure in more than three studies: organizational commitment is in 10 studies, while job satisfaction and human capital both appear in 8.
Three studies, however, labelled their HRM concept in terms of involvement: two using the term high-involvement management (Martínez-Del-Río et al. 2012;Wood et al. 2012) and one employee empowerment (Fernandez and Moldogaziev 2013). I will now outline these and the black-box studies that focus on involvement within their core HRM concept.

Studies focused on involvement
Within the black-box studies, Cappelli and Neumark (2001: 742), while recognizing the lack of agreement on what constitutes high-performance practices, take the focal question to be 'whether employee involvement is associated with improved performance'. Consequently, their study concentrated on involvement practices, including job rotation, quality circles, and teams. Vandenberg et al. (1999) defined their object of study in Lawler's (1986) original terms and found that the base -empowerment -and the other three dimensions -information-sharing, skills and knowledge, and rewards -were associated with performance. Consistent with the emphasis on orientations underlying the practices, these were measured by latent variables. Similarly, Guerrero and Barraud-Didler (2004) define high-involvement management in Lawler's terms and measure it by latent variables and find, using French data, that all dimensions except the motivational one (financial compensation based on performance) are related to organizational performance.
Of the mediation studies, Fernandez and Moldogaziev (2013: 492) also focus on Lawler's four dimensions of high-involvement management, which they term empowerment following Bowen and Lawler's (1992) earlier work. With US data, they measured each dimension using a latent variable from which they formed a unidimensional measure of empowerment, which was found to be related to performance and mediated by employees' job satisfaction and innovativeness.
Wood, along with colleagues in various studies in Britain, used similar latent variable modelling techniques to first assess, with some success, whether an orientation underlies the use of management practices, since its existence, they argue, cannot be taken for granted (De Menezes and Wood 2006;Wood 1999b;Wood and Albanese 1995;De Menezes 1998, 2008). The two elements of high-involvement management -role-involvement and organizational-involvement management -were found in several of their studies to be discrete; roleinvolvement management was measured by the degree of employee discretion and variety included in the design of jobs, while measures of organizationalinvolvement management included team working, idea-capturing schemes, disclosure of financial information and training specifically for involvement (De Menezes and Wood 2006). Where measures of motivational practices and general training were analysed, they were unrelated to involvement, and the motivational practices were not found to be strongly correlated with each other.
The two dimensions of involvement have been found, in a study using Britain's Workplace Employment Relations Survey (WERS) of 2004, to be related to both the level and rate of change of productivity (Wood and De Menezes 2008). However, the effect of role-involvement management was found to be stronger. In a mediation study they found similar results, but only the relationship with roleinvolvement management was mediated by employee well-being and job satisfaction, the two mediators included in the study (Wood et al. 2012). In a study using data from the 2011 WERS, Wood and Ogbonnaya (2018) found similar results, and speculate that the unexplained positive effect of organizationalinvolvement management on performance reflects the way it improves work organization, coordination, and collective action.
Finally, Gittell et al. (2010), using US data, focus on organizational involvement (high-performance work systems in their terms) achieved through interteam and -departmental mechanisms that facilitate cross-functionality (in selection, conflict resolution, and performance management). Gittell et al. (2010) found a relationship between such practices and performance, and that it was mediated by relational coordination, 'a mutually reinforcing web of communication and relationships carried out for the purpose of task integration' (p. 491), which is echoed in Wood and Ogbonnaya's (2018) speculation about how organizational involvement generates high performance. Gittell et al. (2010) see their view as distinct from the dominant focus on abilities and motivation, but it is unclear whether they see it as a competitor or a complement to this.
The studies concentrating on involvement thus make up a small proportion of the HRMperformance research stream. In AMO terms, these studies are distinct in giving priority to O. In Gittell et al.'s (2010) case, A and M are ignored; in Wood et al.'s (2012) case, attention is on skillsand motivation-enhancing practices that support employee involvement; while those studies that follow Lawler (1986) include amongst these practices items such as performance-related pay that can support other approaches, which is why they are found in Wood et al.'s (2012) work to be discrete from their measure of high-involvement management. With the possible exception of Gittell et al. (2010), the studies also highlight that high-involvement management S. Wood reflects an underlying management orientation that values and fosters this involvement above all else, which can be captured using latent variable modelling. These studies have been more faithful to Lawler's (1986) and Walton's (1985) original concern for involvement and have not, I contend, been given sufficient recognition for this. Their difference from the high-performance studies has been similarly neglected. Involvement is thus a major way of identifying the key divide within the HRM-performance research stream.

Basic divides
This review has revealed two sources of variation within the HRM-performance research stream. First, there are differences between studies that focus on involvement as an underlying approach to HRM that transcends previous command-and-control approaches, and those that focus on HRM as a set of designated practices that together should produce high performance. Despite the melding in the reviews of the two concerns that ignited the research streaminvolvement and holism -in practice the studies are fragmented. Second, within the set-of-practice studies there are significant differences in the practices considered and most notably varying degrees of neglect of involvement.
The majority of studies refer to involvement in some way in their introductions, and this has continued to be the case in the mediation studies. The treatment of high-performance work systems and high-involvement management as synonymous was encouraged by the early studies, and the subsequent emphasis on AMO in introductions to studies may have given the impression that coverage of practices was comprehensive. The divide between the studies focusing on involvement and the others has nonetheless become more apparent as the proportion of studies concentrating on skills or motivational practices has increased (e.g. Beltrán-Martín et al. 2008;Bloom and Van Reenan 2007;Paul and Anantharaman 2003). When involvement practices are included, the focus is on their role as motivational devices, with insufficient attention paid to their potential role in innovation, work redesign, relational coordination, and restructuring relationships -high-involvement management is aimed at improving all of these. The extreme case of neglecting involvement is the measures of HRM for the Anglo-American Management and Organizational Practices Survey (MOPS); centred on motivational aspects, its core practices are monitoring, targets, and incentives (Buffington et al. 2016). Illustrative of the inattention to involvement's role in HRM, one of the leaders of the MOPS project has written: 'we believe these three functions [monitoring, targets, and incentives] are the core of what business schools and consultancies claim is the essence of good management' (Van Reenen 2017: 10).
It has even become commonplace to define the opportunities in the AMO model in ways that reduce the significance of involvement. For example, Lepak et al. (2006: 233) call it 'opportunities to contribute', and Fu et al. (2015: 212) refer to 'opportunities [for employees] to express their talents', as if the issue is simply that skills and motivation count for nothing if there is no job discretion, rather than involvement being key to organizational success. Indeed, in some cases the O in AMO is simply referred to as opportunity.
The marginalization of involvement in such cases as MOPS reflects a literal interpretation of the highperformance work system as an emphasis on the direct management of performance through incentives. This can be put alongside the main influence behind the neglect of involvement that Wood and Wall (2007) suggested -that the importance given to the resource-based theory of the firm has led to a focus on skills. However, the varied treatment of involvement across the whole field is also indicative of the focus on holism in the development of the concept of high-performance work systems. Indeed, Chadwick (2010: 85) identifies 'the organizational system level approach to HRM [as] … a defining characteristic of this [HRM-performance] research'. It forsakes the emphasis on a philosophy underlying the adoption and execution of high-involvement management. This could simply have meant disregarding the bedrock status of involvement, making it just one element of the AMO triad, or not acknowledging the role-/organizational involvement distinction. However, as we have seen, involvement has been relegated below the other elements. This process has been unwitting in some cases, as introductions to studies are often dominated by vague specifications of the key concepts, while equating high-involvement or -commitment management with high-performance management means the distinctiveness of involvement is lost.
Allied to the creation of the unified view of the field, or perhaps more of a consequence of it, has been a supposition that the various modes of composing measures of HRM -aggregation of practice use, latent variable modelling, cluster analysis -are substitutes for each other, and results based on measures derived by various methods are comparable. These are, however, distinct. The main difference between them is in the objects they can measure. We have already seen that latent variable modelling can be used to develop measures of an underlying orientation, in this case a high-involvement philosophy, whereas the aggregation of practice use is measuring the extent of an HRM system. In measurement theory terms, the former produces reflective scales, as the practices are the effects of the orientation, while using the latter, akin to formative scales, means the practices included define the system (Diamantopoulos and Siguaw 2006).
The issue of synergy is distinct from these differences as it relates to the effects of individual practices on performance. Latent variable modelling is not concerned with this, and none of those using it think they are testing for synergy. Nor can cluster analysis, a case-based approach that (with the exception of latent class models) defines cases on the basis of their similarity of use of practices, capture synergies. They can only be tested using aggregate measures if we test for an exponential relationship between the total use of practices and performance, which has only been done in two studies (Godard 2001(Godard , 2010; but, since synergy is about the combined effects of practices, testing the effects on performance of the interactions between pairs of practices (or more) is most appropriate.
An appreciation of these research design protocols highlights that a perspective based on viewing highinvolvement management as an orientation is distinct from one based on the wholesome use of the gamut of modern personnel techniques. The former is based on an integrated use of involvement practices and their supporting practices, the latter on the total use of a broader range of less specific practices, whose relationship to performance is not given by its nature and which may be linear or non-linear in theory or empirically. But does following all these distinctions really matter?

Why does the different treatment of involvement across the studies matter?
It might be argued that variation in the treatment of involvement, or more generally in the practices across the studies, does not matter as the results in studies of HRM's link to performance are uniformly strong: any differences between HRM definitions are less significant than this commonality. Guest et al. (2003: 293) certainly play down its significance, writing: 'it is worth noting that the positive results appear to be reported irrespective of the definition of HRM, and therefore the precise contribution of practices, utilized by researchers'. It may even be argued that the achievement of common results across studies with different measures is a strength. The results might suggest that the practices are substitutes for each other, and that their nature, and any interactions between them, are less important than the quantity used. The results may also be viewed as consistent with Lepak et al.'s (2006) theory that variety in the details of practices reflects the tailoring of a general set of best practices so they are aligned to strategic objectives (assuming that organizations are competent at this alignment), so high-involvement management is a variant of high-performance work systems.
Following this logic would suggest that the difference between the latent variable and aggregation methodologies is unimportant, as the total score will be correlated with the latent scores. Even if there are synergistic effects (unless these are concentrated in a small number of practices), the implication is the same -that the practices are best in class. Consequently, the differences are secondary and the policy implication of the studies is clear: use these practices.
This all seems a little unsatisfactory, and appears to endorse the claims that the area is under-theorized (Fleetwood and Hesketh 2006;Gerhart 2012) and encourages a sidestepping of potential differences in the ontology of the underlying concepts. Gerhart (2012: 158), commenting on prescriptions based on the total use of broadly defined practices, asks: 'Would we be content with saying, if doing medical research, that more surgery and more hospital stays should be implemented to achieve better health, without getting more detailed recommendations?' To continue the use of Gerhart's medical metaphor, we might also ask what policymakers seeking evidence bases might think of a set of studies where some measures included either hospitals or surgery, but not both. Alternatively, what would they make of a field where the variety in therapies across the studies was so large that some only included conventional drugs, while others included behavioural therapies or homoeopathy, or even neglected drugs completely? Moreover, if involvement is crucial, is its neglect akin to ignoring the lifestyle component of modern health advice?

S. Wood
I contend that high-involvement management and high-performance work systems can and should be distinguished from each other. The differences between the two perspectives are ontologically significant and go beyond differences in the measures used to assess practices. There is a basic difference between a focus on HRM as an approach or philosophy towards personnel management centred on involvement, and HRM as the use of a comprehensive set of HR practices that complement each other. Highinvolvement management is an integrated management approach centred on extending employee involvement and voice, whereas the high-performance work systems perspective is about the quest to identify a set of practices that can guarantee high performance through aligning employees' behaviours with the strategic goals of the organization. The former is more centred on employees' needs as a potential end in itself and is, because their fulfilment is assumed to be beneficial for their and the organizational performance, a stakeholder perspective. In contrast, the latter is more directly driven by employers' concerns for best-in-class performance, and is therefore a shareholder perspective in the private sector context. In AMO terms, the O -the opportunity for participation -is the fulcrum of high-involvement management, while it is only one potential element of the highperformance work system, and if primacy is given to skills-and motivation-enhancing practices, it may be relegated to a moderator of effects of A and M on performance.
The high-involvement management concept as an integrated approach to personnel means that the selection of practices reflects an underlying managerial philosophy centred on employee involvement. Practices are tied together and made consistent by following high-involvement principles; but the enactment of this approach explains the performance effect, not the practices. The selection of practices is only one manifestation of the approach, as it will influence reactions to key events and day-to-day interpersonal relationships. For example, managements in a high-involvement regime may handle downturns in demand through short-time working or providing extra training rather than laying workers off. The precise nature of the practices will also vary both between organizations and over time. For example, idea-capturing schemes can be group-type quality circles, suggestion schemes, or surveys. Moreover, their properties, and not just the approach's effect (as is the case in a synergistic model), will be transformed when designed to fit with or sup-port practices. For example, attitude surveys will entail feedback and participation in their design, so they are transformed into the survey feedback method.
In contrast, if the high-performance work systems perspective is to be coherent, the HRM system must be viewed as a set of practices that are individually the best in their domain, each being positively related to organizational performance. This systems perspective is a technocratic methodology, based on treating the management practices as technologies which can be applied selectively to different groups in the organization. The system is then a set of complementary practices, implying that the more practices are used, the higher the performance effects. Any synergistic effects of practices relate to their relationship to performance, making the effect of the system greater than the sum of its parts, without changing the nature of the practices.
Both approaches imply some level of shared intentionality amongst management. In the highinvolvement perspective, an intention to involve employees is its defining characteristic; in the highperformance perspective, the intentions to create a system may vary, as Lepak et al. (2006) suggest. Accordingly, in the case of high-involvement management, the extent to which intentionality underlies the use of practices will vary with its level. A high level of shared consciousness of employees' needs is only likely to exist at a high level of high-involvement management. At lower levels, managements may not be conscious of the advantages of involvement or the stress caused by low involvement. Similarly, the various possible intentions underlying high-performance work systems may only exist with high use of the practices, whether these be to adopt a strategic approach to HRM so practices are integrated with the business strategy, to ensure a committed workforce, or to enhance the human capital of the organization.

Implications for future research
In drawing out the implications I assume the objective is to continue the tradition of using practices to measure orientations and systems, while being more faithful to the original concern of the stream in relation to employee involvement. In the case of the high-involvement management perspective, this entails both a re-establishment of the bedrock status of role involvement in high-involvement management and acknowledging the complementary role of organizational involvement in distinguishing it from the earlier job design movement. In the case of high-performance management, it means including involvement practices within the set of potential best practices so they have equal status relative to skills-and motivation-enhancing practices, or at least a role as a moderator of their effects. This is not to say that forsaking involvement is not legitimate, but may be better viewed as a human capital perspective. Nor does concentrating on practices mean that research on the intentions underlying the adoption (or low use) of high-performance work systems would not be fruitful or feasible. Equally, research directly on the principles and intentions behind high-involvement management could usefully be pursued.
The first implication of differentiating between the two perspectives is for the composition of composite variables, since the object being measured in the high-involvement perspective is an orientation, but is a system in the high-performance work system. In contrast, Boon et al.'s (2019) and other recent methodological papers (Delery and Gupta 2016; Hauff 2019) assume the object in HRM-performance studies is the human resource system. Boon et al. (2019) thus treat latent variable and aggregation methods as comparable. A crucial difference, though, is that the practices included in reflective scales of high-involvement management developed from latent models do not alter the nature of the underlying variable (assuming they reflect it), whereas those included in additive indexes or formative measures of high-performance work systems define the nature of what they are measuring and thus the selection of items is absolutely crucial. The meaning one can attach to the scales differs, and even if they were to include the same practices, the latent scores are unlikely to coincide perfectly with the total use of practices, as Wood and Albanese (1995: 232) show, and they may not predict performance equally.
The high-involvement management perspective requires role-and organizational involvement practices that reflect the orientation and measures available in the UK's WERS (Wood 2009;Wood and Ogbonnaya 2018;Wood et al. 2012) offer a good starting point: idea-capturing methods, teamwork and briefing, functional flexibility, appraisal, information sharing, and selection and training for involvement. Given the neglect of role-involvement and voice mechanisms, such as idea-capturing schemes, it is vital these are included, not least as both are crucial for proactivity. Gittell et al.'s (2010) focus on crossfunctional practices could also be taken on board.
Taken literally, the high-performance work systems perspective means measuring a set of highperformance practices, which requires knowing in advance that each has a performance effect which is best in class. This chicken-and-egg problem may explain the so-called lack of consensus about the practices that might go into a high-performance work systems model. We can specify the practices to explore based on the frequency of their appearance across studies, as Boon et al. (2019) and Posthuma et al. (2013) have, but the risk is that these practices may be biased towards skills and motivational ones, or may not be based on studies that show they are predictive of performance -neither author confines themselves to HRM-performance studies. Nonetheless, Boon et al.'s (2019) recommendation of centring on six complementary practices that cover all three dimensions of AMO in equal proportions -training and development, participation/autonomy, incentive compensation, performance evaluation, selection, and job design -may be a good start. However, these are only topics, and serious consideration needs to be given to the items in the measures. Similar involvement practices to those used in the high-involvement literature may be used, and adopting the same factual style for the others is necessary to avoid the increasing tendency, which Boon et al. (2019) expose, to include evaluative questions or items other than practices (such as transformational leadership, trust, vertical integration, policies, and programmes). By focusing on the practices used, it also avoids the potential problem of intended or espoused practice versus realized or enacted practices (Boon et al. 2019;Simons 2002).
The second implication is that synergistic relationships should be distinguished from the concepts of the complementarity of practices and integrated approach to management. Practices that complement each other are distinct, each playing a unique role, whereas positive synergistic relationships are about their combined effect -when complementary practices enhance each other's impact. Synergy is about the performance effects of practices, whereas complementariness is about the nature of the HRM system. Under the high-involvement management approach, any synergistic relationships between practices reflect their being designed to be integrated. The correlations between practices are indicative of this approach, but do not indicate synergy.

S. Wood
Following the high-performance work systems perspective, we first have to specify the nature of the system -is it simply a set of complementary practices, or of ones that have synergistic relationships with each other? If the former, then the predominant approach so far of regressing performance on the total use of a high-performance work system is the correct test. If synergistic relations are expected, then the interaction effects of pairs of practices (or larger combinations) on performance need to be tested. As 'high-performance work system' is a label for a set of practices, we can legitimately ask whether it might be simpler to treat each item in an index of practice use as separate predictors, as one would need to do if testing for synergies directly through practices rather than subsystems. The advantage of accepting Boon et al.'s (2019) six practices is that this approach is manageable.
The third implication is that the mediators of any relationship between HRM and performance will differ between the objects whose effect is being mediated. Under the high-involvement management perspective, the object is the involvement orientation, not the practices themselves, and potential mediators could include collective psychological or organizational phenomena. The starting point must be the expectation that employee involvement transcends high levels of motivation or engagement, with their emphasis on effort levels, and entails psychological empowerment, proactivity, and other cognitive processes (Evans and Davis 2005;Griffin et al. 2007). A small number of the mediation studies have included Spreitzer's (1995) psychological empowerment, but these are within the high-performance work systems perspective (Bonias et al. 2010;Liao et al. 2009;Messersmith et al. 2011). Proactivity has been neglected across the board (Evans and Davis 2015). However, the high-involvement approach implies the achievement of a proactive workforce, which can contribute to innovation, particularly in processes. The relative importance of individual-and organizational-level proactivity may differ between role-and organizational-involvement management.
Two other types of mediators that have figured in the high-involvement literature may have a role -satisfaction and well-being measures, and relational coordination and flexible working processes. The issue then becomes the extent to which these are the cause or effect of proactivity, or are generated directly by high-involvement management. Consideration of reciprocal relationships between this triad of possible mediators -proactivity, satisfaction/well-being, and relational coordination -may be fruitful. For example, the autonomy provided by role-involvement management may contribute to improved coordination of activities, or the improved coordination derived from organizational-involvement management may increase employees' social satisfaction. So far, well-being has been associated with pleasure and the satisfaction of need, but the meaningful dimension of happiness associated with a sense of purpose should be explored (Baumeister et al. 2013). We might predict that organizational-involvement management has a greater effect in this area than on the satisfaction of need, which is more affected by roleinvolvement management.
For the high-performance management perspective, the object whose effect is being mediated is not so clear. Is the object individual practices, categories of practices, or the total set? The effect of individual practices may be mediated by what they are assumed to entail -skill acquisition corresponds to training, and motivation to reward systems; and the potential mediators of high-involvement management discussed above apply to the involvement subsystem, though each type of practice may affect all the potential mediators. The intensification of work (assuming it refers to increases in given demands) that critics of HRM practices often point to is likely to be most affected by motivational practices. Changes in the nature of demands, however, may be affected by either skills-acquisition or involvement practices.
However, in line with the holistic argument, should we presume that an explanation is being sought for the effect of the combined use over and above the effect of each practice? The added value of the overall ensemble of practices, we might predict, for example, is achieved through increasing organizational commitment, which is consistent with the increasing significance given to social exchange theory in the mediation studies. The signalling theory of HRM's effect, whereby HRM practices have a symbolic effect through conveying the employer's concern for employees, which figures in mediation studies (e.g. Liao et al. 2009;Messersmith et al. 2011;Razouk 2011), applies most readily to explaining the added value of the wholesale use of practices rather than individual ones. This implies that the strong focus on affective commitment in mediation studies has some merit. Alongside this, the cognitive dimension of commitment -the alignment of employees' orientation with the organization's strategy, or what Buller and McEvoy (2012) call their 'line of sight'could be considered, and may have more effect than affective commitment. Such considerations reinforce the case for inclusion of involvement practices, as a set of such practices seems more likely to enhance employees' feelings that they are valued, or their awareness of the organization's strategy, than is an HRM based on human capital or incentives.
If the object whose effect is being mediated is interactions between practices, as in the synergistic model, mediators are likely to be quite specific. For example, the interaction between appraisals and training might be explained by more targeted training and organization-specific skills, and by focused appraisals if training affects how they are conducted. If synergistic relationships are found, it is important to examine in detail which elements of a mediation chain between one practice and the outcome is moderated by another practice. For example, if appraisal has a positive effect on labour productivity through organizational citizenship, its relationship with organizational citizenship may be stronger in organizations with a high degree of training, or it may be that the effects of organizational citizenship on performance will be stronger in such organizations.
Regardless of the perspective, modelling multiple mediators requires their simultaneous inclusion in models to ensure each controls for the other. Potential causal links between them, and thus explanations in terms of sequential, mediated processes, should be developed. 1 Examination of potential negative effects and countervailing influences -such as increased workloads -must also be considered.
Additionally, the arguments of the paper have implications for issues beyond organizational performance that have been explored in the wake of this 1 In some mediation studies, employees' perceptions of -or attributions to -the HRM system or orientation are posited as mediators. These are seen as the second in a chain of factors that lead from HRM to performance (Elorza et al. 2011;Purcell et al. 2003). However, this seems to negate the reason they need to be considered. It implies that they will authentically reflect the practices or intentions behind them, and it is precisely because this is not the case that such perceptions may be important. They could be considered as potential moderators of the relationship between HRM and any mediators, regardless of which perspective is taken. For example, even if management are genuinely oriented towards employee involvement, employees may be sceptical that their ideas will be implemented or rewarded fairly, view management's motives as being instrumental (Koys 1989), or gauge that the costs of involvement are too high in terms of increased anxiety or workload. Such responses will reduce the orientation's capacity to increase potential mediators such as proactivity or job satisfaction. stream of research, notably the links between HRM and trade unionism or Works Councils (Wood 2013) and employee outcomes (Boxall and Macky 2014;Mohr and Zoghi 2008;Wood and De Menezes 2011). The need to differentiate between high-involvement management and high-performance work systems applies equally to these issues. For example, the relationship with well-being may differ between them, and recent arguments from Guest (2017) for emphasizing the increased well-being that follows from HRM may apply most readily to high-involvement management.
The differentiation between perspectives also has implications for qualitative case studies of the HRMperformance relationship. Current studies, for example questioning the implementation or sustainability of HRM, often do not include core involvement practices (especially idea-capturing ones), which means they only have a high-performance work system perspective (e.g. Cook et al. 2016;Tregaskis et al. 2012). Qualitative studies on high-involvement management are needed to complement the quantitative research stream, particularly to address how and why it has developed over time.

Implications for policy
The policy implications of the two perspectives are different then. High-involvement management means management should adopt a particular orientation and recruit and develop managers accordingly, ensuring that they can involve and lead employees in a way that is consistent with high-involvement principles, as well as design and operate practices that reflect this orientation. The general principle of employee involvement should apply to all staffit is an inclusive approach. While not all employees should be involved, for example, in a formal problem-solving group at any one time, the principle that they may be involved extends to all. The high-performance work systems perspective is a literal evidence-based approach, suggesting a formulaic methodology for the management of employees, and not the adoption of certain values as in highinvolvement management. Once a set of the best practices has been designated or agreed, this must be adopted, although coverage might be concentrated on 'strategic jobs' (Becker and Huselid 2006: 919). The current proliferation of practices in the literature means that there is no definitive set, and thus until such a set is decided, it may be argued that there is a 422 S. Wood limit to the perspective's value, which is not the case for high-involvement management.
The differences between the two perspectives extend to their implications for management education. To ensure the high-involvement management perspective is implemented effectively, a management education oriented towards deep learning and developing dispositions and attributes is required. It is only with such developments, coupled with accumulated experience, that managers can engage in the sort of 'careful self-analysis' which Vandenberg et al. (1999: 329) recommend, if they are to enact employee involvement. In contrast, under the highperformance system, the reality is that management education is a process of knowledge exchange that alerts managers to best practices. The danger is that it only generates surface learning. However, attention could be given to deep understanding of the individual practices and how to design them for optimal performance.

Conclusions
Studies within the HRM-performance research stream differ in a number of ways. My review has shown that underlying the variations in treatment is a differentiation between studies according to whether they view high-involvement management as an approach to HRM that can maximize performance through giving primacy to involvement, or see bestpractice HRM as a set of high-performance work practices. Within the latter group, involvement has not been given the priority that the concept of highinvolvement management gives it. Such differences have been passed over in introductions to research papers and reviews of the area; I have shown the unified view of the field to be unwarranted, and that the extent of unity has not increased as the field has moved into testing mediators of the HRM-performance relationship. Nor have the implications of the diversity in practices across the studies been given much attention, as the focus has most recently been on the different methodologies for combining practices or conceiving of synergies. This lack of attention to the detailed substance of the objects in the studies, coupled with a lack of precision or theoretical underpinning in much of the writing, means our conceptions need to be sharpened. 2 Arresting the neglect of involvement entails restoring the bedrock status of role involvement to highinvolvement management and, in the case of highperformance work systems, emphasizing opportunities for involvement -the O in AMO -as a potential element of them. The two perspectives have different implications for designing future research and policy, and both can legitimately be taken forward and evaluated against each other. Addressing the mechanisms underlying any relationships between HRM and performance can still be given priority, providing the distinction between high-involvement management and high-performance work systems is observed. Such knowledge is relevant to managers, especially those who find that they are not getting the expected results from one or other of the approaches, as it allows them to diagnose whether the implementation is the problem, or the approach is not having the desired effect on an assumed mediator.