This paper uses an agent-based model to study party system dimensionality. Empirically, patterns of party competition vary widely across political systems, but it seems unclear how exactly they come about. One specific strand of research has grappled with explaining the tendency of some party systems towards a single dominant dimension of competition - commonly referred to as the left-right dimension. I build on this research in assuming that voters, depending on the party system they are situated in, come to regard dimensions of political conflict as differing in salience, and crucially, as non-separable. This leads to their utility function not being directly proportional to Euclidean distance anymore, but to stretch out in some directions and compress in others, thus discriminating party positions more strongly among a subset of political space. I explore the possibility of idiosyncratic unidimensionality, i.e. that salience and non-separability parameters differ across voter (implying they have different understandings of what left-right means), to further elucidate how the interplay of parties and voters leads to the emergence of party system structures.
