Alex Fink, On 11/06/2014 11:36: > This relates to something I've complained about here before: I forget > the example, but it was something like casually saying "who" had been > _dropped_ in the English "the man I saw". Well, I won't rule out that > analyses could show it had, but that IMO certainly shouldn't be your > initial assumption; it's much simpler, much Occamier, to just say > English has two forms of the relative construction, one with a > relative pronoun and the other with zero in the same place. Why is > this principle so tempting that if two forms mean the same, one > should be an _alteration_ of the other? Even if the two constructions > stand in relation, even if one's the default and the places where the > other appears can be expressed by reference to the first -- why can't > this just be analogy and not masquerade? On the contrary, if the only difference between bare and nonbare relatives was the presence of the relativizer, optional phonological nullity of the syntactically-present relativizer would be the simplest explanation. There is abundant evidence that the mapping from sentence syntax to sentence phonology is messy (i.e. the whole world of inflectional morphology), so positing a bit of extra messiness there is trivial. On the other hand, elaborating the *syntactic* rules to generate distinct syntactic structures for bare and nonbare relatives both necessitates substantially more complexity, in multiplying the syntactic possibilities and fails to account for the (putative) identical syntactic behaviour of the two. As it happens, bare relatives are sufficiently distinct from nonbare that it's at least not readily apparent that they're syntactically identical. For instance, bare relatives don't allow pied-piping and -- very very mysteriously to me -- aren't recursible. But it may be that the most satisfactory account of them treats them as syntactically identical. For example, I have argued (<https://drive.google.com/file/d/0B223OeOg5j3yekpCdFhPNm54dzg/edit?usp=sharing>) that wh-relatives and that-relatives are syntactically identical, differing only in phonology. But at any rate, my general point is that regular syntax but moderate shenanigans at the syntax--phonology interface is in the grand scheme of things simpler than regularity at the syntax--phonology interface but greater complexity in the syntax. (Other examples: I tend to think that subordinate clauses with and without _that_ are syntactically identical; but I am unpersuaded by the idea that proper nouns and nonpartitive mass nouns b egin with a null determiner.) --And.