Burnout and engagement at work as a function of demands and control

Picture a simple map of working life. On one axis, how much the job asks of you. On the other, how much say you have in how to do it. That's Karasek's demand and control model, and it makes two bold promises. First, the worst strain shows up where demands are high and control is low. Second, the richest learning and motivation happen where both are high. The quieter corners—low demand with either high or low control—should be calmer, maybe even a little sleepy. It's a clean, intuitive picture. The argument for decades has been about whether those two axes actually interact the way the model suggests, or whether they mostly run on separate tracks: demands driving strain, and resources like control fueling motivation. Demerouti, Bakker, de Jonge, Janssen, and Schaufeli decided to press that question with a broader lens. Instead of focusing only on strain, they put strain and motivation on the table at the same time, and asked the model to separate four kinds of jobs: low demands and low control, low demands and high control, high demands and low control, and high demands and high control. It's a more complete test of the promise, and it leans on a distinction that matters in real life: feeling burned out is not the same thing as feeling engaged, and, crucially, you can feel some of both. Let's unpack those pieces for a second. Burnout, in the Maslach tradition, has three faces: exhaustion, that bone-deep weariness; cynicism, the distancing and indifference that creeps in; and professional efficacy, a sense that you can still do things well. When Christina Maslach and Michael Leiter adapted their measure for general work, the Maslach Burnout Inventory General Survey, they kept those three dimensions. Engagement, in Wilmar Schaufeli and Arnold Bakker's framing, also has three parts: vigor, the energy and persistence you bring; dedication, the sense of meaning and enthusiasm; and absorption, that flow state where time disappears. Decades of psychometrics show these are not just mirror images. They're distinct, measurable states that often move separately. That matters because if strain and motivation really trace different pathways, a good test has to measure both. So here's how they did it. They surveyed three hundred eighty-one employees at an insurance company—about two thirds men, average age around forty, with a dozen years on the job—after on-site briefings and confidential participation. Demands and control came from the Dutch version of the Job Content Questionnaire. Burnout was captured with the Maslach Burnout Inventory General Survey; engagement with the Utrecht Work Engagement Scale; psychosomatic health complaints with a thirteen-item checklist; and commitment to the organization with a short affective commitment scale. These aren't improvised instruments. They've been validated across occupations, and in this sample they behaved reliably. The team then split demands and control into high versus low using a median cut—after putting the scores on the same standardized scale—so each person landed in one of the four quadrants. The analytic move they made is worth translating because it fits the question. They used discriminant analysis. Think of it as asking: if all I know about you is your levels of exhaustion, health complaints, vigor, dedication, absorption, professional efficacy, and commitment, can I tell which of the four kinds of jobs you're in? The math finds combinations of those outcomes—weighted blends—that best separate the groups. If the demand and control story is right, one blend should align with control and motivation, another with demands and strain. And if the classic interaction is strong, the four groups should fall into the distinct corners Karasek drew. Now the results. The outcomes did, in fact, separate the four job types better than chance. The overall test was solid—Wilks' lambda of 0.73 with a chi-square of about one hundred nine, and a p-value well below 0.001. Two discriminant functions carried the signal. The first had a canonical correlation of 0.45, which in plain terms means it captured a moderate chunk of the variation across groups. The second was smaller, a correlation of 0.29, but still meaningful. Put them together, and the model guessed the right quadrant forty-two percent of the time, compared to twenty-five percent if you were just rolling dice. What did those two functions represent? The first one mapped onto control and motivation. High scores on dedication, vigor, absorption, professional efficacy, and organizational commitment stacked on this axis. When that blend was high, you were more likely in a high-control job. The second function mapped onto demands and health impairment. Exhaustion and psychosomatic complaints defined that axis. When that blend was high, you were more likely in a high-demand job. Cynicism behaved awkwardly—loading on both axes but not really helping the model sort people—so much so that dropping cynicism didn't change the story. If you place the four job types in that two-dimensional space, you get a picture that's both familiar and surprising. Low demand and low control jobs—the quiet corners—sat low on motivation and low on impairment. Not much strain. Not much spark. Low demand and high control jobs lit up on motivation without triggering health complaints. That's the "good, easy job" many people imagine: autonomy without overload. High demand and low control jobs showed exactly what you might fear: elevated exhaustion and health complaints, with dampened motivation. And high demand and high control—active jobs—were the interesting ones. They brought both. People reported high dedication and vigor, and they also reported strain. Motivation and wear and tear living together. That last quadrant is the hinge. If high control always buffered high demands, you'd expect strain to drop when control was present. It didn't. Demerouti and colleagues found that demands were tied to impairment largely regardless of control, and control was tied to motivation largely regardless of demands. That's the job demands and resources view in the data: two partly independent routes—one risk, one motivation—running side by side. A couple of fine-grained notes complete the picture. The statistical separation was sharpest for control when demands were low; it blurred more for high demands plus low control, the worst case, where many people share symptoms. And the model didn't make wild errors—almost no one in an easy job with low demands and low control was misclassified as being in a job with high demands and high control. But the key takeaway is how the outcomes clustered: the first function said "this feels like control," the second said "this feels like demand." Let's pause on the measures for a moment because they ground that interpretation. Exhaustion and health complaints are classic health-impairment signals. They track the body and mind saying "enough." Vigor, dedication, and absorption aren't just moods; they predict persistence and performance. Professional efficacy—feeling capable—belonged with those motivational signals. And organizational commitment is the attitude that links personal energy to the firm's goals. By building both families of outcomes into one analysis, Demerouti's team could ask which axis of the job map each family naturally gravitates toward. The answer was asymmetric: health outcomes with demands, motivational outcomes with control. Does that mean Karasek was wrong? Not exactly. The model's big intuition—that jobs live in a space defined by how much they demand and how much they let you shape the work—holds up. But the strict interaction, where high control neutralizes the harm of high demand, only showed partial support. The active learning side thrived where control was high, whether or not demands were high. The strain side rose with demands, even when control was also high. Active jobs are energizing and taxing at once. Quiet, high-control jobs can be motivating without burning you out. That's more nuanced than a single diagonal. There are caveats. This was a cross-sectional snapshot and all self-report. That limits causal claims and invites concerns about common method variance. The authors addressed some of that with factor checks, and importantly, they point to longitudinal work showing that changes in perceived demands and control tend to precede changes in burnout and engagement rather than the other way around. Still, we should treat the pathways as strongly suggested, not definitively proven. So what do we do with this two-track story? It offers a clean, practical split. If you want to protect health, aim at the demand side: workload, time pressure, emotional load. Those are the dials that turn exhaustion and health complaints up or down. If you want to build motivation and learning, aim at control: decision latitude, skill use, and participation in how the job is done. Those are the dials that turn vigor and dedication up. And don't expect one lever to do both jobs. More control is great, but it won't magically erase the physiological and cognitive costs of too much demand. Likewise, trimming demands may make people feel better, but without autonomy, don't expect an explosion of engagement. There's also a subtle cultural message here. We sometimes treat engagement as the antidote to burnout, as if getting people more excited will fix exhaustion. Demerouti's data say engagement is its own channel. You can be dedicated and tired. You can be calm and bored. Managing one does not guarantee movement in the other. That aligns with the lived experience of a lot of high-skill, high-control work: it can be deeply meaningful and still leave you wrung out. If you're a researcher, the methods piece is a nudge, too. Testing models of work with only negative outcomes hides half the picture. This study showed how bringing positive and negative outcomes into the same analytic frame can clarify which job features align with which human responses. The numbers—forty-two percent correct classification, two clear discriminant functions with canonical correlations of 0.45 and 0.29—aren't just technicalities. They mark a model that captures real structure in messy, human data, but not the neat interaction the field once hoped for. Where does that leave Karasek's map? Still useful, now annotated. Demands and control matter, and they matter differently. Plot your job, or your team's jobs, and ask two questions: what's likely to sap health here, and what's likely to feed energy and growth? Those answers will often come from different corners of the room. The art is designing work that keeps the learning channel open while turning the strain channel down. Active jobs can be the best jobs—if we respect both sides of the curve.