φ-Agree Versus φ-Feature Movement: Evidence from Floating Quantifiers

Chomsky (2000) proposes a key shift in the analysis of φ-agreement. In Chomsky 1995, agreement is the displacement of a pronounlike terminal containing interpretable φ-features to the agreeing head, or D-agreement. In Chomsky 2000, agreement becomes the valuation of uninterpretable φ-features on an independent terminal, Agree. Here, a new argument for the Agree model is developed from floating quantifiers. The argument permits directly contrasting agreement with minimal pronouns: clitics and pro. Agree is the right analysis for φ-agreement alone, and D-agreement for clitics, pro, and clitic doubling.

1 The Shift to Agree

In Chomsky 1995:272-276, φ-agreement moves, from an agreeing argument to a target such as T⁰, a minimal X⁰ containing the argument's interpretable φ-features and other formal features, (1a). This X⁰ should have properties similar to those of a pronoun at LF, such as the ability to control and bind (p. 272). Here this is called D-agreement. It reflects a common diachronic origin and synchronic shape of agreement morphology (Fuß, to appear). In Chomsky 2000: 119, 146n71, movement of interpretable material is eschewed, and agreement reflects the valuation of uninterpretable features on T0 via Agree, (1b). Being uninterpretable, these features are deleted prior to LF and thus have no effect there.

(1)

1. a.

2. b.

The empirical motivation for the shift from D-agreement to Agree is Den Dikken's (1995) and Lasnik's (1999) demonstration that φ-agreement is invisible for syntax and interpretation. In (2b), if agreement were a pronounlike φ-set on the verb, it would be expected to license the anaphor, (1a), much like the A-moved (pro)noun in (2a). It does not. This follows if agreement is uninterpretable and deleted after Agree, so that it cannot bind at LF: (1b).

(2)

1. a. Some linguistsi/Theyi seem to them(selvesi) to have been given good job offers.

2. b. There seem to them(*selves_i)/*each otheri to have been some linguists_i given good job offers.

The same inertness of agreement for binding has been found in Icelandic (Jónsson 1996:206), Italian (Cardinaletti 1997a:526n7, Chomsky 2000:147n71), and Tsez (Polinsky and Potsdam 2001:620, 2006:178), which differ with respect to pro-drop, expletives, and the definiteness effect. φ-agreement lacks the binding capacity of pronouns. Agree predicts it: agreement is syntactic φ-feature transmission deleted by LF.¹

2 Floating Quantifiers

This section adds a new diagnostic for the inertness of φ-agreement in the Agree model: its failure to license floating quantifiers (FQs). It extends evidence beyond expletive constructions to other subject "inversions," permitting a direct contrast with pronouns of the poorest kind, which do license FQs: clitics and pro.

FQs are licensed by (pro)nouns in c-commanding A-positions (see section 3 for the mechanism). Agreement by itself fails to license them. An instantiation of this generalization is the expletive construction in (3). The underlined FQ all is licensed by the c-commanding subject in (3a), but not in (3b), where only agreement with the subject c-commands the FQ. The paradigm is part of Baltin's (1978) argument that FQs are anaphoric. It also follows from it that agreement, unlike (pro)nouns, does not license FQs.²

(3)

1. a. <All of> the portraits of Picasso had <all> hung.3PL over the fireplace.

2. b. There had <*all> hung.PL over the fireplace <all of> the portraits by Picasso.

In English and French, the FQ diagnostic can be applied to other "inversion" constructions, with an agreeing subject below the FQ. Baltin (1978:28) finds that FQs are not licensed in French stylistic inversion, (4), and later work concurs (Déprez 1990:56, Hulk and Pollock 2001:8, Kayne and Pollock 2001:157n77, Lahousse 2006:437, 445). Culicover and Levine (2001:301) find the same for English locative inversion, (5).

(4)

1. a. Je voudrais  savoir ce  que <tous> les hommes
   I would.like know what that <all> the men
   ont    <tous> mangé.
   have.3PL <all>  eaten
   'I would like to know what (all) the men have (all) eaten.'

2. b. Je voudrais savoir ce qu' ont <*tous> mangé <tous> les...