Over the past two decades, research into worker well-being has been greatly influenced by the demands-control-support models of Robert Karasek (1979; Karasek & Theorell, 1990). These models propose that worker strain and active learning are determined by particular combinations of job demands, job control and social support. Specifically, incumbents of jobs that are high in demands, low in control, and low in support are expected to show high levels of strain, whilst incumbents of jobs that are high in all three job factors are expected to display high levels of activity, learning and participation, both on and off the job. The models also propose that prolonged exposure to combinations of these job conditions influence workers' accumulated anxiety and sense of mastery. If empirically substantiated, Karasek's models have profound implications for the design of "healthy work" environments. This thesis represents an attempt to clarify, critically evaluate, extend and test Karasek's models. Self-report data, as well as information obtained from a collateral source, are used to assess the independent linear, quadratic, additive and interactive effects of Karasek's job factors. Multivariate models of the direct and indirect relationships between the job factors and a range of possible antecedents and consequences are proposed and submitted to empirical test. Two major, and several minor, studies, all using samples of school teachers, are reported. The first major study used a cross-sectional design, and self-report measures of demands, control and job stressors to predict several indices of worker strain (e.g., stress, job dissatisfaction, somatic complaints). Analyses of data from 421 teachers revealed independent and additive effects of demands and control on strain, but few quadratic or interactive effects. Demands and control also predicted job stressfulness, with additional evidence showing that the effects of demands on this outcome were buffered by perceived job control. The second major study tested Karasek's models using a two-wave full panel design, and an expanded set of predictor, moderating and outcome variables. Data were collected from 987 teachers, as well as from a significant other person nominated by the majority of these teachers. Demands, control and social support were shown to predict stressors and strain, both cross-sectionally and longitudinally. Results generally confirmed Karasek's additive strain hypotheses. Consistent evidence of interactive effects of the job factors on strain was not found in the full sample of teachers, but was obtained when analyses were limited to a sub-sample of new-start workers. Relationships between the job factors and measures of worker activity and participation were also found. However, rather than all three job factors acting conjunctively, demands and support from supervisors predicted some outcomes (e.g., the number of hours teachers worked), control and support from co-workers predicted other outcomes (e.g., self- and other-reported levels of vigour-activity), whilst a further set of indices (e.g., participation in organized activities outside of work) were not strongly related to any of the job factors. Exposure to highly demanding jobs was associated with increases over time in levels of neuroticism, whilst exposure to high control job conditions was associated with increases in levels of mastery. The demands-mastery relationship was buffered by perceptions of job control. Limited support was obtained for a set of additional predictions regarding the role of leadership style in shaping job factors, and hence worker strain. In general, the results from this research confirm past findings regarding the effects of job demands, control and social support on strain. The research makes several important contributions to the literature. From a research methods viewpoint, new, congruent and specific measures of the job factors, activity-participation outcomes and leadership dimensions were developed. Also novel was the use of structural equation techniques to test competing longitudinal models that involved continuously-measured interaction terms. Substantively, the research represents one of most comprehensive investigations yet conducted into Karasek's models. The test of the dynamic person-environment hypotheses is believed to be unprecedented in the literature, as is the attempt to demonstrate links from leader behaviour, through the job factors, to strain. Implications for Karasek's models include the need to reject the additive hypothesis in relation to activity-participation, and to extend the models to incorporate organizational antecedents of demands, control and support. More practically, the research reinforces the importance of providing "control-enhancing" opportunities for workers exposed to highly demanding jobs. Recommendations for future research include the need to test an expanded model of healthy work using multi-wave longitudinal designs, samples of new-start workers, and multiple (including objective) measures of key variables.