Terry & Jimmieson discuss reasons for these inconsistencies. One suggestion is that the interaction only occurs for some outcomes, i.e. they argue that they may be specific to non-affective outcomes such as absenteeism (Dwyer and Ganster, 1991; Parkes, 1991) and some physiological measures (Fox et al., 1993). Another possibility is that it is highly dependent on the measure of control. This argument is put strongly by Wall et al. (1996), who suggested that Karasek’s predicted interaction effect is seldom demonstrated because the decision latitude measure used lacks specificity. They compared a focused measure combining timing and method control with the conventional decision latitude measure in a study that aimed to predict psychological strain. They found a significant interaction effect with demand for the specific measure but not for the conventional measure. However, Terry & Jimmieson argue that many studies using focused measures of control are not consistent with Wall et al. (e.g. Carayon, 1993; Kushnir & Melamed, 1991).

The use of multidimensional measures assessing different aspects of control shows some promise in clarifying the nature of relationships. For example, Sargent & Terry (1998) suggest that specific types of control that are relevant to the task will buffer the effects of demands, whereas more peripheral forms of control will not.

It has also been suggested that the relationship between work stressors, including job control, is curvilinear, i.e. not only are low levels of job control harmful but very high levels may also be harmful (Warr, 1987). This poses some potential problems for the demand– control model because models which fail to take into account nonlinear relationships may showspurious interactive effects due to nonlinear effects of either of the variables comprising the interaction term (Lubinski&Humphreys, 1990). Fewstudies have examined curvilinear relationships, leaving open the possibility that where interactions were significant these may have been spurious. Warr (1991), in a study of over 1600 employed people, found significant nonlinear components and taking these into account in the final regressions found no evidence for an interactive effect in predicting job related anxiety, depression and job satisfaction.

Furthermore, a study by Fletcher & Jones (1993) examined the effects of demand and control in a heterogeneous sample of over 3000 men and women, and found, for men, that there was some evidence that the two variables interacted to predict job satisfaction and life satisfaction. However, this effect disappeared once the nonlinear relationships were taken into account, suggesting that it was indeed a spurious effect.

A further reason for the inconsistency in the findings in relation to the interactive effects, which is increasingly attracting attention, is that it may be highly dependent on individual differences. These are discussed below.

Personality Variables

Karasek’s model of job control was designed to provide a general model ofwork redesign and hence did not incorporate individual difference variables. However, a range of personality variables have been found to moderate the impact of demands and control on strain outcomes and are now thought to be one possible explanation for the ambiguous findings in relation to the model (Parkes, 1991). Over the past few years the focus on finding such variables has increased. Typically, personality variables examined are those most closely related to perceptions of personal control such as perceptions of  Type A personality, locus of control and self-efficacy. There are two possible ways that personality can act as a moderator. It can either be a buffer or resistance factor, decreasing the impact of stressors or strains, or it can act as a vulnerability factor, increasing the impact. For example, Type A behaviour pattern, characterised by high achievement orientation, time urgency and hostility (Friedman & Rosenman, 1974), is seen as a vulnerability factor. Type A personalities have been shown to be more prone to psychological strain than Type Bs under high-demand working conditions with little control (Kushnir & Melamed, 1991).

Locus of control refers to whether an individual generally believes that the causes of events in their lives are substantially within their control (internal locus) or controlled by external factors (Rotter, 1966). This characteristic has been viewed as both a buffer and a vulnerability factor. For example, Parkes (1991), in a study of civil servants, found that external locus of control increased the vulnerability of driving examiners and student teachers to the psychological impacts of job strain. Thus, the interactive demand-discretion model predicted anxiety for subjects with external locus of control but not for those with internal locus. Daniels & Guppy (1994), however, suggest that having an internal locus of control buffers the effect of stressors on psychological well-being.

Belief that the individual has internal control may not always be coupled with confidence that the individual is able to take that control and succeed in taking a particular desired course of action. This kind of belief has been termed “self-efficacy” (Bandura, 1977) and has also been shown to moderate the impact of job strain. For example, Schaubroeck et al. (2000), in a study of American bank tellers, found that perceived job control moderated the impacts of job demands on psychological strain only for those who were high in selfefficacy. Where they had low self-efficacy, having high control exacerbated the effect of demands. Schaubroeck&Merritt (1997) further found that self-efficacy moderates the effect of job strain on blood pressure such that for individuals with high self-efficacy, job control mitigated the effects of job demands.

Related to the concept of self-efficacy is the dispositional trait of proactive personality. Characteristics of proactive people include being unconstrained by situational forces, effecting environmental change, showing initiative and taking action (Bateman & Crant, 1993). Parker & Sprigg (1999) found that this characteristic, like self-efficacy, moderated the effect of job strain. The effect was such that job demands caused psychological strain except in circumstances where employees had opportunities to reduce demand—that is, they had job control, and they had proactive personality, enabling them to take advantage of this opportunity. For passive employees (those lacking a proactive personality), job control did not mitigate the effects of demands, and for such employees interventions to increase control may have little impact.

Proactive people are likely to use coping strategies that involve taking control, such as analysing the problem and taking action. Such strategies have been characterised as active coping (de Rijk et al., 1998). The role of this type of coping in moderating job strain has been examined by de Rijk et al., who found that for nurses high on active coping, job control was a stress buffer, whereas for those low on active coping, it seemed to increase the effects of job demands on emotional exhaustion.

Overall, these studies suggest that only for individuals high in personality traits such as internal locus of control, self-efficacy and with proactive styles of personality and coping are the impacts of demands reduced by job control. For those low in such characteristics then enhancing control will have little benefit and may even make matters worse. Most studies have considered the moderating effects of personality variables on psychological well-being or minor physical symptoms. There is less evidence in relation to coronary heart disease. However, Bosma et al. (1998) examined the impact of several personality variables in the relationship between job control and heart disease in the Whitehall II study of British civil servants. They suggest that psychological characteristics such as hostility, need for control, negative affectivity, angry coping and unassertive coping had little impact on the association. They acknowledge that they have not considered the effects of all possible psychological factors. For example, they excluded self-efficacy (a factor which Schaubroeck & Merritt (1997) suggest moderates the effects of stressors on blood pressure). Nevertheless, they argue that their study suggests that “increasing job control could, in principle, lower risk of heart disease for all employees” . Thus, it seems likely that different mechanisms are implicated depending on the outcome measure considered. One possibility discussed by Bosma et al. (1998) is that work stressors impact on bodily responses without conscious awareness of being under stress, hence personality styles are less important when physical well-being is the outcome. However, in the case of psychological well-being, the individual’s personality may substantially influence appraisals and, ultimately, well-being. In terms of implications, interventions which focus on the individual’s appraisals and coping (for example, by increasing self-efficacy or active coping) may be effective in reducing psychological distress among workers lacking in job control. However, based on current research evidence, this may be less likely to have an impact on physical outcomes such as heart disease.

Individual differences inwork motivation and needs are seldom considered as moderators of the effect of job strain although de Rijk et al. (1998) found that need for control did not have any moderating effect. However, other kinds of needs and motivation may be important moderators. For example, Homer et al. (1990) used the model to predict women’s risk of delivering preterm, low-birth-weight babies. They found that women working in jobs high in demand and low in control were more likely to deliver such babies than women in other jobs, but only where they did not want to remain working.

Unlike Karasek’s model, the moderating effect of needs is built into the job characteristics model in the form of the variable growth need strength (GNS). This variable assesses the need for personal accomplishment and for learning and development. While there have been inconsistent findings (Roberts & Glick, 1981), meta-analyses support the importance of this variable (Loher et al., 1985; Spector, 1985). For example, Loher et al. (1985) found that the relationship between job characteristics and job satisfaction was stronger for those employees higher in GNS. More recently, Champoux (1991) looked in more detail at relationships and found different patterns of relationship dependent on level of GNS. Those with high levels responded positively to enriched jobs, as predicted; however, distinctly negative responses were found in individuals low in GNS.

More recent work by Landeweerd & Boumans (1994) looked at need for autonomy as a moderator (a variable which they consider to be close to GNS) in a study which used the JDS to predict both satisfaction and health in nurses. They found that perceived autonomy was related to job satisfaction and health complaints but not to absenteeism. When the moderating effect of preference for autonomywas added they found that more job autonomy leads to increased absenteeism in nurses with relatively little preference for autonomy. No significant relationship was found for those who have greater preference for autonomy. Overall, therefore, notwithstanding the indications that high levels of control are generally associated with positive outcomes, individual differences do seem to be important, and for some people increasing autonomy may actually be related to negative outcomes.

Gender

Many early studies examining job control focused exclusively on men’s work and health, and although studies have increasingly looked at gender differences, it remains unclear whether effects are the same for women. Schnall et al. (1994), in their review focusing on CVD, found that in eight out of eleven studies in which comparisons between men and women could be made effect sizes were similar for both men and women. While conflicting evidence still exists, a number of more recent studies have suggested that effects of demands and control may be somewhat lower for women than for men. For example, Weidner et al. (1997) found that having a high-demand/low-control jobwas unrelated to standard coronary risk factors in both sexes but that it was related to increased medical symptoms and health damaging behaviour in a sample of men but not women. Unlike Weidner et al., Wamala et al. (2000) found that even after controlling for standard risk factors, women in the lowest occupational class had a heart disease risk more than twice that of executive and professional women. However, while job control did contribute to this risk, they concluded that neither lack of job control nor the combination of low demand and high control made a substantial contribution. They argue that the impact of such factors is much less for women than for men, perhaps because, for women, job control may not “adequately capture all the negative aspects of their jobs”. Furthermore, women may face multiple stressors, including a range of non-work stressors. This point is reinforced by Brisson et al. (1999), who found that a combination of family responsibilities and high-strain jobs had a greater effect on the blood pressure of white-collar women than exposure to either factor alone. Similar patterns emerges for psychological well-being. Studies reviewed by Van der Doef & Maes (1999) also find less support for the effects of low control and high demands for psychological health forwomen. However, not all studies support this view. For example, Mausner-Dorsch & Eaton (2000) found that job strain was more strongly related to major depression in women.

Overall, when looking at redesigning work for women, employers may need to consider a wider range of influences on psychological well-being.

HAS JOB CONTROL RESEARCH HELPED GUIDE INTERVENTIONS AND JOB REDESIGNS?

A startling feature of Karasek’s model is that despite the fact that it was explicitly designed to aid job redesign, as far as can be ascertained from the published literature, it has had minimal impact in this area. Instead, it has had huge impact in terms of enabling dimensions of work to be incorporated in medical and epidemiological studies investigating the effects of work on health. As a result we now have a large literature providing quite convincing evidence about the negative effects of job control for health, but little use of the model to guide job design.

One reason why this may be so is that the construct of control included in psychological theories tends not to be specific enough to guide interventions. For example, Reynolds (1997) argues that the concept of job control is “somewhat diffuse” and that within a single organisation different individuals will have different levels of control over different aspects of work, as well as different preferences. Reynolds further suggests that we do not have sufficiently well-developed methods for bringing about changes in organisations. These factors help explain her findings in a study comparing an organisational change intervention (increasing control) and a counselling scheme. Reynolds found that neither had any influence on people’s perceptions of work stressors. However, the counselling intervention, but not organisational change, led to improvements in psychological well-being.

A further reason that the model may not be used to guide interventions in practice may be due to its emphasis on such limited aspects of work. It may be clear to employers that problems in a particular occupational group are related to a much wider range of variables (Jones et al., 1998).

Intervention studies that do aim to enrich jobs by increasing control have had mixed results. They also highlight difficulties with both implementing these kind of changes and assessing their effectiveness in the context of rapidly changing organisations (Landsbergis & Vivona-Vaughan, 1995; Maes et al., 1998). For example, Landsbergis & Vivona- Vaughan in the USA (1995) used an intervention based on a range of approaches, including Karasek’s model, to reduce stress in a public service agency. Two matched pairs of departments were assigned to either a treatment or a waiting list control group. The intervention involved employees participating in problem solving committees. They identified stressors and developed action plans to reduce them and to encourage and assist management to implement changes. Priority stressors included such factors as uneven or repetitive workload and poor communication. At the end of one year there were mixed results in one department and negative or negligible impact in the other. It seems likely that the failure of the intervention can be attributed to the fact that it was only department wide while workers were also affected by a range of organisational changes. In this case these included a major agency reorganisation without any employee or union involvement. This produced frustration and disappointment. Both departments lost directors who had supported the changes and some planned changes were not completed. This study clearly demonstrates the wide range of impediments to this kind of organisational change.

It is possible, however, that although employers have not explicitly used the demand– control model, they have nevertheless taken on board the concept of control as a necessary part of job design. The frequent rhetoric of empowerment and the incorporation of control concepts in the management literature suggests that this might be the case. For example, Landsbergis et al. (1999) suggest that new working methods such as “lean production” incorporate total quality management approaches and multi-skilling, techniques which are described in texts as involving increased employee control. However, Karasek (2001) considers that such job enriching approaches are often also accompanied by tougher managerial control. While research is limited on the impact of these new methods of working, Landsbergis et al. (1999) reviewed 38 studies which look at the impact of a range of new work systems and conclude that there is little to suggest that they empower employees. They argue that job control typically remains low while pace of work increases—creating highstrain jobs in Karasek’s terms. Overall, a recent report by Merlli´e & Paoli (2001), which draws on surveys spanning Europe over the period from 1990 to 2001, states that while work intensity has grown, work control has not increased significantly. More specifically, they suggest that increases in employee control from 1990 to 1995 were not sustained in the following five years. In fact, some groups, such as plant and machine operators and sales and service workers, report a decline in the control they have over work.

Evidence for interventions based on the JCM is similarly unimpressive. For example, Kelly (1992) reviewed longitudinal studies of 31 field experiments in job redesign, all of which changed one or more of the dimensions of the JCM (i.e. not necessarily autonomy). They found only limited support for the model, with job redesign leading to significant improvements in job satisfaction in only 17 out of 30 instances. However, where employees perceived an improvement in job content they were more likely to experience increased job satisfaction (though not necessarily motivation or performance). Others have also found inconsistent results in predicting other psychological well-being outcomes. For example, Wall et al. (1986) studied the introduction of semi-autonomouswork groups in a longitudinal study and found that while there were increases in extrinsic satisfaction, there were no effects on mental health. Briner & Reynolds (1999) reviewed the evidence for job redesign interventions generally and concluded that there is limited evidence for their effectiveness, and effects are not uniformly positive. They attribute much of this failure to theoretical limitations.

It is both a strength and a limitation of models such as the demand-discretion model and the job characteristics model that they try to provide simple parsimonious models to fit all situations. By necessity they ignore job-specific factors (Jones et al., 1998) and they ignore contextual factors such as social class (Muntaner & O’Campo, 1993) or organisational factors (Cordery & Wall, 1985). Increasingly, there are calls for much more explicit theories to explain particular phenonoma (Briner & Reynolds, 1999) or to apply to particular occupational groups (Sparks & Cooper, 1999). It is becoming clear that simple parsimonious theories which try to explain a broad range of outcomes can only take us so far in indicating the likely impact of stressors. In the applied context, we need to know what particular specific stressors lead to specific outcomes in particular situations. Theoretical developments towards more specific job control constructs and measures are a helpful move in this direction.

The amount of control or autonomy an employee has over his or her ownwork is perhaps one of the most crucial aspect of working life and one which has been extensively researched. It is a key feature of major theoretical approaches to stress (e.g. Karasek, 1979; Payne, 1979; Warr, 1987), and, together with the related concepts of discretion and autonomy, is a central feature of job design theories (Hackman & Oldham, 1980;Wall et al., 1990a). Furthermore, the importance of job control and related concepts is recognised in the management literature, where it is generally seen as important for releasing employee potential and increasing performance. For example, Peters & Waterman’s (1982) analysis of innovative companies intimately links autonomy with entrepreneurship. The related concept of “empowerment” is a central motivation for many organisational changes and total quality approaches. While control has been linked to a wide range of outcomes, including improving performance and motivation, this post focuses on the implications of the construct for health and well-being.

In the psychological literature, “control” can be viewed as both a characteristic of the environment and a characteristic of the individual. For example, machine-paced assembly work is intrinsically lacking in control; however, the employees themselves may nevertheless be viewed as being high in control, in terms of having high mastery or self-efficacy (Bandura, 1977). The individual may further be seen as having greater or lesser need for control (Burger & Cooper, 1979). Typically, models of stress and job design have treated “control” (and the similar concepts of “decision latitude”, “discretion” and “autonomy”) as characteristics attached to a particular work role or job task, and this approach is the focus of this post. However, the fact that individuals vary in terms of their preferences and perceptions of personal control has implications for these models. Hence individual differences are increasingly being incorporated into approaches to stress and job design. This post aims to introduce two key theoretical approaches to the topic of control in the workplace, the demand–control (or job strain) model (Karasek, 1979), and the job characteristics model (Hackman&Oldham, 1976). Both of these approaches were designed to provide practical guidance to practitioners and managers on the design of work. More recent theoretical developments focusing on more specific types of job control will also be discussed. The post provides a brief overview of the evidence relating to the damaging effects of low control for health and well-being. This includes consideration of the nature of the relationship between job control and health and the effects of individual differences on these relationships. The post discusses the implications of this research for job design interventions and the evidence concerning the effectiveness of such interventions. It finally considers job control in the context of new technology and new work designs.

The literature on job control and health is now vast and a number of comprehensive reviews of various aspects of this literature exist (Ganster & Fusilier, 1989; Kristenson, 1996; Schnall et al., 1994; Terry & Jimmieson, 1999; Van der Doef & Maes, 1998, 1999). This post therefore does not aim to duplicate such reviews, but rather to summarise recent key issues and trends in the research literature and their implications for managers and practitioners.

JOB STRESS AND WORK DESIGN THEORIES

The Demand–Control (“Job Strain”) Model

While job control is an important feature of a number of approaches to occupational stress, the most well known approach, which has been responsible for most of the impetus of research in this area, is Karasek’s demand–control model. However, the key feature of the model is that both psychological and physical strain can be predicted from a combination of demands and control. The model has evolved over the years and, perhaps as a result, there is now a certain amount of uncertainty in its specification (Kasl, 1996). In particular, there is a lack of clarity concerning the exact nature of the hypothesised relationship between the two key variables in the model and strain outcomes. One hypothesis examined in many studies focuses on the negative impacts of jobs that are high in demand and low in control, suggesting that these two variables have an additive effect. A further hypothesis, mooted by Karasek (1979), is that the two variables have an interactive effect such that the combination of high demand and low control produces greater strain than the simple additive effect. This is often viewed as a central feature of the model (e.g. Terry & Jimmieson, 1999). However, the implications of the interactive model are clearly not the same as the additive model. If the effect is interactive (and demand is harmful primarily in conditions of low control) strain could, in practice, be reduced by increasing control without reducing workload (Karasek, 1979; Parkes, 1991). This strategy would be less effective in the case of additive effects.

Further lack of clarity exists concerning the actual nature of the interaction. Kasl (1996) points out that some researchers seem to regard decision latitude as buffering the effect of job demand, such that risk due to high demands will only be present at low levels of control. However, others regard the interaction as synergistic, i.e. both low demand and high control are associated with risk but the combination of the two increases the risk beyond the additive effect. These two approaches are usually not clearly distinguished. Kasl suggests that in many studies the data are presented in such a way that it is not even possible to interpret whether the effect is truly additive or interactive. It is often also the case that the variables are combined into a single “job strain” variable so that it is not possible to separate the effects of demand and control.

A further area of ambiguity which has been widely discussed in the literature is the operationalisation of the key variables (e.g. Ganster & Fusilier, 1989; Jones et al., 1998; Smith et al., 1997). In its original version, Karasek’s model measured decision latitude, a construct which was rather wider than the popular meaning of control, by combining a measure of decision authority with a measure of skill discretion (Karasek, 1979). While decision authority assesses the extent to which people have freedom over how they do their work and have a say over what happens, skill discretion is concerned with the level of skill required and the non-repetitive nature of work. While the two concepts have been shown to be related, they are clearly distinct. Over the years a range of different measures of job control has been used; however, in recent years, a key development in the model has been the increased sophistication in the operationalisation of this variable. This will be discussed further below.

The model has been further developed to include social support (Johnson & Hall, 1988). This is sometimes referred to as the “iso-strain model”. Here again two alternative hypotheses have been identified, the additive hypothesis and the interactive hypothesis, suggesting that support acts as a buffer (Van der Doef & Maes, 1999). Karasek & Theorell (1990) have further proposed the addition of job insecurity and physical demands to the model. However, these latter two variables are seldom included.

The job strain or iso-strain models have been tested using a variety of methods and levels of analysis. At one extreme, long-term epidemiological studies have followed individuals over years to predict coronary heart disease and other disease outcomes (Alterman et al., 1994). At the other extreme, researchers have used laboratory tasks varying in degree of demand and control to examine immediate changes in pulse or other physiological responses (Perrewe & Ganster, 1989). However, the most common approach is the cross-sectional study using self-report measures of demands and control to investigate the predictors of psychological well-being (also based on self-reports) or other symptoms (Dollard et al., 2000; Fletcher & Jones, 1993).

The model has been subject to considerable criticism (Ganster & Fusilier, 1989; Jones et al., 1998; Kristenson, 1996; Muntaner & O’Campo, 1993). For example, concerns have been expressed about the nature and subjectivity of measurement, the statistical tests used to test the interaction, the tendency for its core dimensions to be confounded with socioeconomic status and the lack of account it takes of both wider sociocultural issues and individual differences. It has also been criticised for its simplicity and lack of scope. Nevertheless, perhaps because of its clarity and simplicity, it has stimulated a vast amount of research into the effects of job control on health, in addition to research aiming to improve the model itself. While Karasek’s model is probably the most influential approach to job control and health, a substantial strand of work investigating psychological well-being and job satisfaction is based on the similar construct of autonomy contained in the motivation and job design literature, frequently in the context of the job characteristics model.

The Job Characteristics Model

In motivation and job design research, the notion of autonomy is implicit in the earlywork of Maslow (1970) and Herzberg et al., (1959). However, its key place in job design research is ensured by the job characteristics model (JCM) (Hackman&Oldham, 1976) and its accompanying measurement instrument, the job diagnostic survey (JDS) (Hackman & Oldham, 1980; Idasdak&Drasgow, 1987). This model suggests that there are five core job characteristics: skill variety, task identity, task significance, autonomy and feedback. The construct of “autonomy” is similar to Karasek’s notion of decision authority and includes items concerned with freedom to decide how to do the job and opportunity to use your own discretion. It is hypothesised that these dimensions (via the mediation of the critical psychological states of experienced meaningfulness, experienced responsibility and knowledge of results) predict work motivation, job satisfaction and work effectiveness. In calculating the overall motivating potential of a job, autonomy and feedback are particularly important and are each given equal weighting to the other three elements added together. It is hypothesised that the relationship between job characteristics and the above outcomes is moderated by the individual difference variable, “growth need strength”. The model did not initially suggest relationships between the core job characteristics and health, but later researchers added mental health as an outcome variable (Wall et al., 1978, 1986).

The model has also been adapted to look at job enrichment at both an individual and a group level. For example, in semi-autonomous work groups or self-directed teams, the group may have increases in autonomy and other task characteristics, but the individuals’ jobs may not be improved (Langred, 2000; Wall et al., 1986).

As with Karasek’s model, there has been considerable criticism of the research investigating the JCM, particularly in relation to the strong bias towards the use of cross-sectional studies relying on self-reports alone (Roberts & Glick, 1981). The model is based on the premise that the direction of causation is such that job characteristics cause high or low satisfaction and also that self-reports of job characteristics are an accurate reflection of the objective characteristics of a job. As a result, it is assumed that improving the task to afford greater control will lead to people becoming more motivated as they perceive themselves to have more control. However, it is now widely recognised that there are some problems with this assumption and that self-reports may not always closely correspond to more objective ratings (Sanchez & Levine, 2000; Spector & Jex, 1991). Furthermore, perceptions of work characteristics may be quite easily manipulated (Adler et al., 1985) and the direction of causation may be reciprocal (Wong et al., 1998). The relationship may also be influenced by individual difference variables such as positive and negative affectivity (Munz et al., 1996). Nevertheless, despite these limitations, the model has been extremely influential and has been successful in predicting a number of outcomes, but most frequently it has been used to investigate the predictors of job satisfaction (Champoux, 1991; Loher et al., 1985). This research is considered further below.

Theoretical Developments and the Concept of Job Control

More recent theoretical developments in stress and job design have helped to clarify what is meant by “job control” and have improved the measurement of the construct by differentiating between various specific types of control. This idea is not new. Karasek himself, in the original paper introducing the model in 1979, suggested that future work should distinguish between different aspects of decision latitude. A number of researchers have since developed typologies. This is potentially useful both to improve the predictive power of the theories but also to provide clearer guidelines for practitioners.

Breaugh (1985), for example, suggested three specific facets of autonomy. Work method autonomy refers to the degree of discretion concerning the procedures and methods used in conducting the work. Work scheduling autonomy is the extent of control over scheduling, sequencing or timing of activities and work criteria autonomy describes the degree to which workers can determine the criteria for evaluating their performance.

More recently, Jackson et al. (1993) have developed measures of control over how to do the work (method control) and when the work is done (timing control), in addition to more specific demand concepts. These were originally developed in the context of advanced manufacturing technology but norms have subsequently been developed for a wider range of shop-floor and related jobs, including administrative and managerial posts(Wall et al., 1995). While these factors refer essentially to direct control over the task, Mullarkey et al. (1995) have also suggested a measure of individual role breadth, which encompasses influence over and involvement in decisions affecting broader aspects of work that impact upon the employee. Other types of control may also be important in particular contexts. So, for example,Wall et al. (1990b) consider boundary control (that is, control over secondary aspects of the task such as maintenance and servicing of machines) to be crucial in the context of advanced manufacturing technology. Similarly, Soderfeldt et al. (1996) suggest taking into account a wide range of aspects of control in a study of human service organisations. These include administrative control, ideological control and control within and over a situation.

These specific approaches to job control are nowincreasingly being utilised in the context of Karasek’s model and have potential for extending the model (e.g. Sargent & Terry, 1998). There is, as yet, however, little research investigating either the differential effects of these variables on health or the effectiveness of interventions targeting specific types of control.