The phonological inventories of the world’s languages vary remarkably in their size and constituency, when modeled as sets of phonemes or systems of distinctive features. An alternative approach to the analysis of inventories can be made, based on the premise that the phonological primitives represented in the inventory, the lexicon and the speech signal are one and the same – coordinated actions of the vocal tract. Described in articulatory terms, the differences between inventories of different languages may not be as significant as feature and segment-based characterizations suggest. Comparative estimates of the entropy of different inventory structures suggest that an articulatory model may provide a more parsimonious account of the salient contrasts than a feature-based approach. Under an articulatory account, both consonantal and vocalic inventories can be explained using the same theoretical apparatus, and ‘complex’ segments can be explained in temporal terms.