Which preferences have voters for the descriptive representation of different societal groups in parliament? While a large corpus of literature argues that descriptive representation is vital for substantive representation, empirical research largely focuses on analyzing voter preferences for individual candidate characteristics. I argue that such studies are ill-equipped for assessing voters' preferences for descriptive representation. Voters might favor being represented by, for example, a well-educated, mid-aged and politically experienced woman, but this does not necessarily indicate that voters want all legislators to look like that. I design a novel conjoint experiment in which respondents have to decide between two differently composed parliaments and use this design for testing theories about representation preferences. These findings provide a novel perspective on voter preferences for descriptive representation and provide empirical evidence regarding the question of which societal groups voters want to be accurately represented in parliament.