The Structural Deficiency of Verbal Pro-Forms

In this squib, I discuss two kinds of verbal pro-forms in British English, do and do so. Examples of these forms are given in (1).

(1) British English

1. a. Terry will eat pasta and Ines will do, too.
   [with second sentence interpreted as 'Ines will eat pasta']

2. b. Terry will eat pasta and Ines will do so, too.
   [with second sentence interpreted as 'Ines will eat pasta']

In the spirit of Cardinaletti and Starke's (1999) typology of strong, weak, and clitic pronouns, I will argue that do is a structurally deficient relative of do so. In particular, while both of these pro-forms are headed by v, do-but not do so-lacks a VP complement. As I will show, this approach accounts for certain prosodic and semantic differences between these forms. The analysis, if correct, suggests that aspects of Cardinaletti and Starke's and Déchaine and Wiltschko's (2002) decompositional approaches to pronouns extend to the lower functional sequence of the clause.

1 British do as a Verbal Pro-Form

The construction in (1a) is sometimes referred to in the literature as "British do" in view of the fact that it is most robust in United Kingdom varieties of English. Speakers of American English, in particular, typically lack the option in (1a) on any reading.

(2) American English

A:

Will she eat?

B:

*She should do.
[* on any reading]

In addition, British English, like other varieties of English, allows true elisions as in (3).

(3) British English
Terry will eat pasta and Ines will, too. [with second sentence interpreted as 'Ines will eat pasta']

A traditional distinction in the literature is between anaphora that have internal structure (typically, bare deletions as in (3)) and those that do not have internal structure (which often take the form of some phonetically overt pro-form). In a derivational framework, this difference is often expressed by positing a derivational difference between these two kinds of anaphora such that true elisions have a full-fledged structure that undergoes deletion, while pro-forms are base-generated as anaphors.

In two recent papers, Baltin (2004, 2005) argues that the do of (1a) is a pro-form. In particular, he points out that if we view sentences like (1a) as true elisions, it is mysterious why elided constituents under do cannot contain internal structure. Wh-traces, for example, are impossible in the VP "covered up" by do, as shown in (4), and inverse scope is likewise unavailable out of the understood VP, as shown in (5). Similar evidence to this effect comes from the fact that do is unavailable with antecedent-contained deletion, passivization, and topicalization, as illustrated in (6)-(8).¹ As the following examples show, these properties are also shared by do so, traditionally treated as a verbal pro-form (Ross 1970, Johnson 2001, Stroik 2001, Horvath and Siloni 2003).²

(4) Wh-traces
*Although I don't know which book Fred will read, I do
know which book Tom will do/do so.
(Baltin 2004)

(5) Inverse scope
Some man will read every book and some woman will do/
do so, too.
(Baltin 2004)
[only interpretable with some taking scope over every in both clauses]

(6) Antecedent-contained deletion
*Bart can eat anything that Homer can do/do so.

(7) Passivization
*The steak was eaten by Bill, and the fish was done/done
so too.
(Baltin 2006)

(8) Topicalization
*Hazelnuts I like, peanuts I don't do/do so.
(Chalcraft 2006)

By contrast, true elisions have none of these properties, as shown in (9)-(13).

(9) Wh-extraction
Although I don't know which book Fred will read, I do
know which book Tom will _____.

(10) Inverse scope
Some man will read every book and some woman will _____,
too.
✓ ∀>∃
✓ ∃>∀

(11) Antecedent-contained deletion
Bart can eat anything that Homer can _____.

(12) Passivization
The steak was eaten by Bill, and the fish was _____, too.
(Baltin 2006)

(13) Topicalization
Hazelnuts I like, peanuts I don't _____.
(Chalcraft 2006)

2 do versus do so

In view of these data, let us follow Baltin (2005...