Aims/hypothesis The natural history and physiological determinants of glucose intolerance in subjects living in Europe have not been investigated. The aim of this study was to increase our understanding of this area. Methods We analysed the data from a population-based cohort of 1,048 non-diabetic, normotensive men and women (aged 30–60 years) in whom insulin sensitivity was measured by the glucose clamp technique (M/I index; average glucose infusion rate/steady-state insulin concentration) and beta cell function was estimated by mathematical modelling of the oral glucose tolerance test at baseline and 3 years later. Results Seventy-seven per cent of the participants had normal glucose tolerance (NGT) and 5% were glucose intolerant both at baseline and follow up; glucose tolerance worsened in 13% (progressors) and improved in 6% (regressors). The metabolic phenotype of the latter three groups was similar (higher prevalence of familial diabetes, older age, higher waist-to-hip ratio, higher fasting and 2 h plasma glucose, higher fasting and 2 h plasma insulin, lower insulin sensitivity and reduced beta cell glucose sensitivity with increased absolute insulin secretion). Adjusting for these factors in a logistic model, progression was predicted by insulin resistance (bottom M/I quartile, OR 2.52 [95% CI 1.51–4.21]) and beta cell glucose insensitivity (bottom quartile, OR 2.39 [95% CI 1.6– 3.93]) independently of waist-to-hip ratio (OR 1.44 [95% CI 1.13–1.84] for one SD). At follow up, insulin sensitivity and beta cell glucose sensitivity were unchanged in the stable NGT and stable non-NGT groups, worsened in progressors and improved in regressors. Conclusions/interpretation Glucose tolerance deteriorates over time in young, healthy Europids. Progressors, regressors and glucose-intolerant participants share a common baseline phenotype. Insulin sensitivity and beta cell glucose sensitivity predict and track changes in glucose tolerance independently of sex, age and obesity.