Headedness, again*

Headedness is an intriguing feature of language design. On the one hand, headedness manifests itself very clearly; preposed relative clauses are visibly different from postposed ones, and postpositions are easily distinguished from prepositions. More generally, structural heads (the constituents which determine the category of their phrase) either precede or follow their dependents. On the other hand, there is room for disagreement and variation in the assessment of headedness. For instance, the ordering of determiners and nouns can be viewed differently depending on what counts as the head, causing disagreements over the headedness of nominal constituents. Furthermore, even if all linguists agree on what counts as the head and what counts as a dependent, there is no requisite consistency in the way dependents and heads are ordered across different phrases within the same language. An otherwise dependably head-final or head-initial language may exhibit exceptions; the results are messy, and researchers get discouraged. There is, as of yet, no good explanation for headedness. The phenomenon is visible; it is rather robust; it seems easy to learn (Lupyan and Christiansen 2002; Van Everbroeck 2006), but what is it? This question has puzzled many linguists, and we still do not know its answer. When dealing with something that is unfamiliar it is often tempting to just toss it out as unnecessary or superficial. Researchers now and again have suggested that headedness is no more than a trivial pattern-recognition device without much deep meaning and with no value in linguistic theory. Yet it is hard to dismiss a device that is so pervasive. This squib presents a new argument as to why theoreticians should still give headedness a chance. I am not prepared to explain headedness, but I will bring in a new dimension in which its effects are apparent: the proportions of lexical categories. If my results are on the right track, they add further evidence in support of the hypothesis that headedness is still relevant for theory construction and needs to be accounted for.