Many types of environmental impacts are multidimensional in character. What this means is that an environmental resource that is affected by a proposed project or policy often will give rise to changes in component attributes each of which command distinct valuations. One tool that can elicit respondents’ distinct valuations of these multiple dimensions, and account for trade-offs between them, are (discrete) choice experiments (DCEs). DCEs share strengths and weaknesses with contingent valuation but also have some distinctive characteristics that may differentially affect its performance and accuracy. A number of developments, on the face of it, appear to work against one another. The selection of the experimental design, i.e. the combination of attributes and levels to be presented to respondents in the choice sets, is a key stage and the tendency has been to opt for increasingly complex designs, to improve response efficiency. Yet this creates inevitable cognitive difficulty for respondents, associated with making multiple complex choices between bundles, with many attributes and levels. There is a limit to how much information respondents can meaningfully handle while making a decision, possibly leading to error and imprecision, depending on whether fatigue or learning dominate respondent reactions. The links to behavioural research are again highly relevant such as on heuristics, and filtering rules guiding choices that are “good enough” rather than utility-maximising. The growing sophistication of statistical modelling of responses is another notable characteristic of this work and has enabled far better account for considerations such as preference heterogeneity. While the domain of specialists, this modelling is increasingly accessible more broadly via a growth in training opportunities and free statistical software.