The aim of the previous chapters has been to present a unified and consistent analysis of a number of problematic nominal constructions, trying to reconcile the sometimes contradictory semantic, syntactic and pragmatic evidence available. Although the resulting underlying representations give the impression of a neat classification of clearly distinguishable construction types, the discussion throughout this part of the study will have made clear that such a conclusion is not warranted. In reality, we have found that only a subset of the examples found in the corpus exhibit all (or most) of the criteria for a particular category. All the other examples turn out to be ‘in-between’ cases defying straightforward categorization. It was therefore suggested that the classifications given be regarded as based on prototypical cases, and the underlying structures proposed as representing the best members of the category in question.

A crucial distinction in the preceding chapters has obviously been that between left-headed and right-headed constructions. Although this distinction is still considered to be useful, it has at the same time proved to be quite problematic, in particular if one holds on to the idea that it ought to be possible to determine the head of a construction unequivocally. Such an approach, we have found, is problematic on various counts. First of all, there is the question of which criteria we are to use to determine headedness, and of how reliable these criteria are.