One of the problems of brand image research lies in identifying which are the particular beliefs about and associations with the brands in question which ought to be measured. This paper discusses a number of different techniques which can be used for eliciting what the various attitude-dimensions for a product field are, and then for selecting which of these dimensions should be covered in a quantified survey. The paper reports on an experimental study of five different elicitation techniques. On the basis of this study, it is suggested that an approach which invites consumers to compare brands with each other is likely to be the most fruitful. Alternative comparative techniques to repertory grid interviews are described, which appear to avoid a number of the 'drawbacks' which the repertory grid has tended to manifest in a number of market research applications.