Aims/hypothesis Most studies of diet in glucose intolerance and type 2 diabetes have focused on intakes of fat, carbohydrate, fibre, fruits and vegetables. Instead, we aimed to compare diets that were available during human evolution with more recently introduced ones. Methods Twenty-nine patients with ischaemic heart disease plus either glucose intolerance or type 2 diabetes were randomised to receive (1) a Palaeolithic ('Old Stone Age') diet (n=14), based on lean meat, fish, fruits, vegetables, root vegetables, eggs and nuts; or (2) a Consensus (Mediterranean-like) diet (n=15), based on whole grains, low-fat dairy products, vegetables, fruits, fish, oils and margarines. Primary outcome variables were changes in weight, waist circumference and plasma glucose AUC (AUC Glucose 0-120 ) and plasma insulin AUC (AUC Insulin 0-120 ) in OGTTs. Results Over 12 weeks, there was a 26% decrease of AUC Glucose 0-120 (p=0.0001) in the Palaeolithic group and a 7% decrease (p=0.08) in the Consensus group. The larger (p= 0.001) improvement in the Palaeolithic group was independent (p=0.0008) of change in waist circumference (−5.6 cm in the Palaeolithic group, −2.9 cm in the Consensus group; p=0.03). In the study population as a whole, there was no relationship between change in AUC Glucose 0-120 and changes in weight (r=−0.06, p=0.9) or waist circumference (r=0.01, p=1.0). There was a tendency for a larger decrease of AUC Insulin 0-120 in the Palaeolithic group, but because of the strong association between change in AUC Insulin 0-120 and change in waist circumference (r=0.64, p=0.0003), this did not remain after multivariate analysis. Conclusions/interpretation A Palaeolithic diet may improve glucose tolerance independently of decreased waist circumference.