Subject-Verb Agreement

Wolfram & Christian (1976:83) observe that in AppE, on average 59.1% of non-pronominal plural subjects fail to show regular agreement with the finite verb (thus, (1a) and (1b) alternate in AppE), while the percentage of pronominal non-agreeing subjects (cf. (1b)) is extremely small (0.1% with non-be verbs; 0.7% with present-tense be). Similar results have been found for other varieties of English, including Belfast English (BelE; Milroy 1981, Henry 1995), Scots English (see e.g. Adger & Smith’s 2005 study of the Buckie corpus, and also Hazen 1996, 2000), and Ocracoke English (Hazen 1996, 2000).

In seminal work on the morpho-syntax of BelE and its relationship with SE, Henry (1995) analyzes this contrast between pronominal and non-pronominal subjects by exploiting the two subject positions provided by the expanded, Pollockian structure of the clause (see Pollock 1989, Chomsky 1995:Ch. 2,3), arguing that while BelE non-pronominal subjects can stay in SpecTP, pronominal subjects must raise to SpecAgrSP. By adopting the assumption that SpecAgrSP is the locus of subject agreement, Henry can then account for the fact that only pronominal subjects are forced to agree with the finite verb.

Henry’s (1995) analysis was developed on the basis of Belfast English; but in AppE as well non-pronominal plural subjects may fail to agree with the finite verb. From Henry’s perspective, this suggests that these subjects may stay in SpecTP and raise no higher. That subjects can indeed stay below SpecAgrSP in AppE is suggested by the fact that, in negative concord constructions, a negative subject may surface to the right of the finite auxiliary (hosting -n’t), producing a subject-auxiliary inversion pattern (see Sells et al. 1996). Note that, unlike Standard English subject-auxiliary inversion, this word order characterizes a certain class of declaratives and manifests itself even in relative clauses, as shown in (2c) (data from M&H and Wolfram & Christian (1976:113)):

This word order plausibly results not from movement of wouldn’t to a position outside the inflectional domain (which is generally impossible in relatives; cf. ?*a room that under no circumstances would he stay in), but instead from the subject failing to move up to SpecAgrSP, as in (3):

Further investigation of the ‘two subject positions’ hypothesis and its applicability to AppE remains to be done. The fact that it is impossible to place really to the left of a non-agreeing subject without comma intonation (Really*(,) the children likes pizza; Henry, p.c.) is unexpected from the ‘two subject positions’ point of view. Also, in light of Henry’s claim that finite non-lexical verbs are allowed (in BelE as well as in SE) to raise to T and no further, one is led to ask whether finite non-lexical verbs are ever in AgrS, and how one can tell: a ‘subject – Auxfin – adverb’ word order is derivable, on Henry’s assumptions, by having Auxfin raise to AgrS across a TP–adjoined adverb, but it is equally derivable by adjoining the adverb to VP with the finite auxiliary sitting in T. Theoretically, the ‘two subject positions’ hypothesis needs to be reassessed in light of two important recent proposals, (i) the abolition of AgrP, and (ii) Cinque’s (1999) detailed argument to the effect that adverbial modifiers are typically inserted in fixed, highly hierarchically stratified positions in the syntactic tree. Cinque’s proposal is difficult to reconcile with Henry’s assumption that adverbs like really and probably can be adjoined either to VP (as in BelE The children really likes pizza) or to TP (as in BelE The children really are late), without there being any apparent difference in interpretation. A re-evaluation of Henry’s account of BelE with an eye toward addressing these issues will be an integral part of our research.