A question that arises in the application of environmental choice modelling (CM) studies is whether to present the choice sets in a generic or labelled form. The former involves labelling the policy options to be presented to respondents in a generic way, for example, as 'option A', 'option B', etc. The labelled approach assigns alternative-specific descriptors to each option. These may relate to the names of proposed policies, different locations or any other policy-relevant details. Both approaches have their advantages. A potential advantage of using alternative-specific labels is that respondents may be better able to base their choices on the true policy context. This can increase predictive validity whilst at the same time reducing the cognitive burden of the CM exercise. A potential advantage of the generic labelling approach is that respondents may be less inclined to base their choices wholly or largely on the labels, and as a consequence, may provide better information regarding trade-offs among attributes. The two approaches to choice set design are compared in the context of a CM study of the values of remnant vegetation in the Desert Uplands of Central Queensland. Results indicate a difference in the cognitive processes generated by choice models using the different approaches. This difference is reflected in both the alternative-specific constants and the taste parameters, and cannot be accounted for by differences in error variance across the two treatments. The implications for environmental valuation are discussed. (C) 2000 Elsevier Science B.V.