The status of the Upper Devonian Acanthostega gunnari as a tetrapod is considered in the light of the apomorphic characters which it shares with other tetrapods, and of the plesiomorphic characters which it retains in common with osteolepiform fishes. It shares about two-thirds of the characters recently listed as tetrapod autapomorphies. Of the plesiomorphic features which it retains, some may be truly primitive, whereas others may be associated with secondary modification for an aquatic habitus. After consideration of these characters, we conclude that parsimony demands that they be interpreted as primitive. We then address the question of how tetrapods may be defined, given the likely discovery of more transitional forms in the future. Possession of digits remains the most compelling character in practice at the present, but we point out its potential inadequacies for the future.