Features: what we learn from indigenous languages

I revisit the conjecture that languages take their features from a ‘fixed list’, and consider in particular what indigenous languages bring to this issue. Identifying features (such as case) can involve serious problems of analysis; identifying the values of features (such as dative) is equally challenging. I review the contribution of the Set-theoretical School, and analyse some significant problem instances. Key examples include those which show the rise of and loss of features and their values (the rise of person in Archi, the loss of the dual number value in most of Slavonic). This leads on to the ‘correspondence problem’, the way in which features and their values correspond across and within languages. It is tempting to assume that languages either have or do not have a particular feature. But we find examples of features being ‘partitioned’ (as in Upper Sorbian possessive constructions) and languages which have ‘concurrent’ systems, that is, that have orthogonal systems involving one and the same feature (Mian). The positive conclusion is that indigenous languages are still leading us to reconsider and refine our analyses of features, through the challenging new data we find there.