Dietary changes can be difficult to effect both at an individual and at a population level, and even when changes do occur they are often far slower and less pronounced than might be expected. Three possible reasons for this situation will be considered: the complexity of food choice and competing influences, attitudinal ambivalence and optimistic bias. Food choice is influenced by a large number of factors, not only health considerations, and therefore it is not surprising that interventions based primarily on health concerns have been ineffective. Another concern is that people do not always have clear-cut attitudes, but rather can be ambivalent about foods and about healthy eating, and this factor might impact on the translation of beliefs and attitudes into behaviour. A third possible reason is optimistic bias, where individuals believe themselves to be at less risk from various hazards than is the average person. This effect has been demonstrated for nutritional risks, and this factor might lead people to take less note of health education messages. The stages-of-change model from health psychology has been proposed as a method for improving the effectiveness of behaviour change interventions. However, there are a number of problems in transferring such a model from smoking, where it was originally developed, to dietary behaviours, including the lack of clear-cut specific behaviours and behaviour change targets in the dietary field.