In certain contexts first/second person pronouns can be interpreted like bound variables. Both sentences in (1) admit a bound reading in which nobody else did their homework, in addition to an unremarkable strict reading. The bound reading interests semanticists because it seems to suggest, contrary to Kaplan (1989), that such pronouns have their reference mediated by some operator.
(1) a. Only I am proud of my students.
b. I’m the only one [RC who is proud of my students. ]
Free reading: Nobody else is proud of my students. (Context: Divisive Students)
Bound reading: Nobody else is proud of their students. (Context: Mean Colleagues)
In this talk, I report joint work with Isabelle Charnavel and Dominique Sportiche on indexical binding in relative clauses (RCs) like (1b). Focusing on French and English we identify two restrictions which reveal the phenomenon’s syntactic underbelly. We see in both languages a subject/non-subject asymmetry: indexical binding is only possible in subject relatives (2). The object relative in (2b), for example, does not permit a bound reading (even having controlled for Weak Crossover!).
(2) a. The Subject Restriction
For an indexical to be bound in certain RCs, the syntactic subject must be relativised.
b. I’m the only one [RC to whom David announced t [ that myFree/*Bound students are happy ]]
In French we observe a further restriction in (3). On the bound reading of a pronoun, local verbal agreement needs to match in person value with that pronoun. Thus in (3b), the bound reading forces the appearance of 1SG copula (suis) rather than 3SG one (est). This is clearly reminsicent of interactions between agreement and indexical binding noted by Kratzer (2009) for German.
(3) a. The Agreement Restriction
If a pronoun receives a bound reading, agreement local to pronoun must match in person.
b. Je suis le seul [RC qui suis/#est fier de mesBound enfants ] 1SG:ok / 3SG:*
I’m the only one who is proud of my children.
To explain (2) and (3) we adopt an analysis of bound indexicals as bound pronouns with local person features (Charnavel & Sportiche 2024). This meshes with several commitments of interest to syntacticians, summarised in (4). In short, wh-elements need to serve as intermediate binders for indexicals, but they can only do this if they are local to a T with features matching the indexical.
(4) a. Wh-elements are syntactically underspecified for person. (WSSCABI)
b. T starts its syntactic life already specified for person/number values.
c. Wh-elements acquire at LF, under certain semantic conditions, person features from T.
These commitments require a rethinking of the syntax of agreement. Contrary to Probe-Goal theories (e.g. Chomsky 2001), nominals do not supply values to underspecified clausal ϕ-features. Instead agreement reflects feature-matching between a syntactically local nominal and T, enforced at LF. This view is independently motivated by cases of semantic agreement (e.g. Corbett 2023).