Where do relative specifiers come from?

This paper argues that the emergence of dependent wh-relatives in Middle English was an instance of diffusion of a novel syntactic property through the set of wh-forms, construed as a series of reanalyses of individual lexical items. This seemingly parochial claim takes on a broader significance in the context of two well-established findings about the typology of relativization. First, similar constructions (dependent relatives containing a filled [Spec,CP], or relative specifier) have an unusual typological distribution: they are largely clustered in Indo-European languages, but probably not present in Proto-Indo-European. This leads us to ask where relative specifiers come from, or more broadly, how daughter languages can share an otherwise rare property which they did not inherit from a common ancestor. Second, the research tradition stemming from Keenan and Comrie (1977) associates relative specifiers with positions low on the Accessibility Hierarchy. Romaine (1980, 1982, 1984) describes the genesis of English dependent wh-relatives in just those terms: they were initially confined to low-accessibility functions, and subsequently spread up the hierarchy. This pattern is very common among languages with relative specifiers (Hendery 2012), and may be a diachronic universal. These findings invite analyses based on two tenets: (1) a near-complementarity between the initial distribution of dependent wh-relatives and that-relatives, the primary relativization strategy in Keenan and Comrie’s terms, which suggests that whrelatives emerged to replace the earlier demonstrative series of relativizers in functions which could not be relativized by that (Romaine 1982:450); (2) work by Keenan and Hawkins (1987), Hawkins (1995), and Kirby (1996) grounding the Accessibility Hierarchy in processing factors, which suggests that patterns described in the terms of the Accessibility Hierarchy should have explanations in those terms. These functionalist analyses ask “in Language L at time t, which forms were available for which types of relativizaton?”. This approach leads to a surprising position, where formally distinct elements (such as demonstratives and interrogatives) are considered part of the same system, provided that they do the same job. Accordingly, we find quotes like this: