Matching relatives in Middle English
Rob Truswell
There is an ongoing debate about the division of labour between the
three analyses of externally headed relative clauses in (1).
1a) RAISING ANALYSIS: [DP D NP_i [CP [DP Op t_i]_j ... t_j ...]]
1b) OPERATOR ANALYSIS: [DP D NP_i [CP [DP Op_i]_j ... t_j ...]]
1c) MATCHING ANALYSIS: [DP D NP_i [CP [DP Op (NP_i)]_j ... t_j ...]]
Implementational details may vary, but the core differences between the
analyses concern the position in which NP is base-generated: internal to
the relative clause in (1a), external to the relative clause in (1b), or
in both positions in (1c). It is clear that NP is pronounced externally
to CP, and yet typically functions as an argument within CP. The
question is how it does that. (1a) assumes a movement relation between
NP and Op, (1b) assumes a predication relation, and (1c) assumes
deletion under identity or near-identity. Sharpening our understanding
of the nature of the relationship that holds across the CP boundary is
one way to sharpen our understanding of which unbounded dependencies
movement must, can, or cannot, be held responsible for.
It has been uncontroversial since Carlson (1977) that at least some
externally headed relatives are structurally ambiguous, between a
raising structure and something else. The outstanding questions are
which analyses are appropriate for which classes of examples, and
whether (1b) or (1c) (or both) is the appropriate choice for "something
else". An influential line of research (Vergnaud 1974, Safir 1999,
Hulsey & Sauerland 2006) uses reconstruction phenomena to gain insight
into these questions, typically proposing that the raising analysis is
appropriate wherever a CP-external element is dependent on a CP-internal
element (2a), and the matching analysis is appropriate elsewhere (2b).
2ai) I saw the picture of himself that John liked.
2aii) Mary praised the headway that John made.
2bi) *I saw the picture of himself yesterday that John liked.
2bii) *Mary praised the headway last year that John made. (Hulsey &
Sauerland 2006: 114-5)
In this talk, I examine a new source of evidence bearing on the division
of labour between (1a-c). The evidence comes from examples like (3),
part of the grammar of English from c.1340-1800, in which the head noun
is present in both external and internal positions.
3) thilkegoodewerkes that been mortefied by oftesynnyng, whichegoode
werkes he didewhil he was in charitee
"the GOOD WORKS that are mortified by often sinning, which GOOD WORKS he
did while he was in charity" (Chaucer, Parson's Tale, c.1390)
These overtly matching relatives are always nonrestrictive, and in fact
there are good reasons, based on Sells (1985), Heim (1990), Elbourne
(2001), to expect such matching structures to be interpreted as
nonrestrictive. This raises the following questions:
A) Are covert matching relatives, as in (1b), always nonrestrictive like
the overt matching relative in (3)? If not, why not?
B) If matching relatives ARE necessarily nonrestrictive, what are the
implications for the role of movement in accounting for reconstruction
phenomena?
Although I do not expect to have clear answers to these questions
(largely because of the difficulties associated with investigating
reconstruction effects in historical data), I show that the Middle
English data do at least rule out certain current approaches to the
structure of relative clauses.