Many areas of research in ecology and evolutionary biology depend on the quantification of dietary niche width. For herbivorous insects, diet breadth has most often been measured as simply the number and type of host plant taxa attacked. We propose an index of host range (which we refer to as “ordinated diet breadth”) based on observed associations between plants and insects, and the calculation of multivariate distances among plants in ordination space. Similarities and distances are calculated based on host association and, in this context, potentially encompass multiple properties of plants, including phytochemistry, phenology, and other plant traits. This approach can distinguish between herbivores that utilize suites of hosts that are commonly used together and herbivores that attack unusual host combinations, and thus have a relatively broad diet breadth. For illustration, we use a data set of nymphalid butterfly host records, and compare taxonomic and ordinated host range. For a large number of butterfly taxa, we find that host use is clustered in multivariate space with respect to associations observed across all of the butterfly taxa. Applications are discussed, including a hypothesis test of nonrandom host association, and prediction of shifts and expansions of diet breadth.