Tom Meadows, University of Geneva
Spooky Agreement, Agree and Phi-syntax (advanced)
[This course explores theoretical and empirical issues in the analysis of agreement phenomena, and the nature of phi-features. It probes an incomplete line of reasoning and feedback/pushback is highly welcome from attendees!]
It widely assumed that agreement involves an asymmetric syntactic relationship between a probe and a goal (Chomsky 2001, Deal 2023). A probe is representationally deficient, lacking values for a given feature. A goal supplies values via Agree to a probe, which is realised as agreement morphology. The underlying “pretheoretic” intuition is that nominals supply verbs with some kinds of information. We will review the motivation for this now fairly standard view in modern generative grammar.
But a spectre is haunting such Agree-based modeling! There are phenomena from familiar languages in which verbal agreement realises feature values without an obvious goal, or where the goal doesn’t appear to supply all of information realised as agreement morphology (Ackema & Neeleman 2018). I refer to these as cases of Spooky Agreement, so named because they are disturbing for standard Agree-based approaches. We will review several defensive analyses, with the aim to convince you that Spooky Agreement is best handled by remodeling Agree (not abandoning it completely!).
If it looks like a given probe couldn’t not have acquired its featural information from anywhere, the first step to an analysis is to abandon the assumption that Agree is a mechanism for supplying information. Suppose that functional heads like T or v enter the derivation prespecified for phi-feature values. What Agree does is to search for the closest phi-features which sufficiently match those of the trigger/probe. The relevant pretheoretic intuition here is that the verb requires its arguments to be compatible for certain properties.
Sufficient matching is crucial, since anything stronger (e.g. exact matching) will miss the point about spooky agreement. These are cases where probe and goal don’t exactly match, yet that seems to be unproblematic. We could think of the conditions on matching as referring to components of a feature-geometry (e.g. Harley & Ritter 2002). Phi-features on T will be satisfied to match at syntactic level with spooky goals, which themselves bear the absolute bare minimum phi-information. In other words, spooky goals are syntactically underspecified for phi-information but the matching condition of Agree is sufficiently weak that spooky goals still count.
In order to avoid that massive overgeneration in agreement possibilities, we must assume that LF is actually largely responsible for the correspondence that is characteristic of agreement phenomena (e.g. Charnavel, Meadows & Sportiche 2025). What Agree does in merely pair phi-features in the syntax, according to fairly weak conditions. This determines what the controller of agreement is. This pairing is a necessary ingredient for a stricter Compatibility Filter at LF, which is still weaker than exact matching/feature unification.
Remaining issue concerns the syntactic representation of phi-properties, particularly given the use of syntactic underspecification and feature geometries. How can we implement syntactic underspecification of spooky goals? What information do they have that makes them count as goals, and how is it represented? How can we use feature-geometric notions responsibly, without letting hierarchies float free of the syntax and without committing to syntactic representations we cannot justify the richness of?
- Selected References
Ackema, Peter and Ad Neeleman. 2018. Features of Person: From the inventory of persons to their morphological realization. MIT Press.
Chomsky, Noam. 2001. Derivation by phase. In Ken Hale: A Life in Language, ed. Michael Kenstowicz, 1–52. Cambridge, Mass: MIT Press.
Charnavel, Isabelle, Tom Meadows & Dominique Sportiche. 2025. Meaningful Agreement Features: Evidence from Indexical Binding. Proceedings of NELS54.
Deal, Amy Rose. 2023. Current models of Agree. Lingbuzz: https://ling.auf.net/lingbuzz/006504
Harley, Heidi and Elisabeth Ritter. 2002. Structuring the bundle: A universal morphosyntactic feature geometry. In Pronouns: grammar and representation, Eds. Horst J. Simon & Heike Wiese. Linguistik Aktuell.