An Argument for Multiple Spell-Out

In this squib we present a new argument based on Bulgarian (Bg) clitic data for the kind of "multiple Spell-Out" theory recently argued for in works such as Epstein et al. 1998, Uriagereka 1999, and Chomsky 2000; for an early proposal along these lines, see Bresnan 1971. In particular, we will show that if information is sent from the syntax to the phonology at more than one point, and if CP but not IP is a "phase," then certain otherwise mysterious Bg clitic-ordering facts follow straightforwardly.

The squib is organized as follows: in section 1 we describe the basic Bg data, in section 2 we discuss several approaches to these data, in section 3 we introduce new data that reveal the derivational nature of clitic-verb linearization, and in section 4 we demonstrate the empirical advantages of adopting multiple Spell-Out to accommodate the clitic-ordering paradigm.

1 Verb-Adjacent Clitics in Bulgarian and Macedonian

The South Slavic languages, including Bg and Macedonian (Mac), have complex systems of pronominal and verbal auxiliary clitics.¹ In Bg and Mac these elements typically precede the verb, as in (1).²

(1)

1. a.  Včera Petko mi go dade.

    yesterday Petko me(DAT) it(ACC) gave

    'Yesterday Petko gave it to me.'

2. b.  Včera mi go dade Petko.

3. c. *Petko dade mi go včera.

4. d.  . . . če/deka Petko mi go dade

       that(Bg/Mac) Petko me(DAT) it(ACC) gave

    včera.

    yesterday

    '. . . that Petko gave it to me yesterday.'

Furthermore, they cannot be separated from the verb, as shown by (2).

(2) *Petko mi go včera dade.

These simple data are sufficient to illustrate the basic properties of Bg and Mac pronominal clitics: they are verb-adjacent (and not, e.g.,second position) and they precede the verb.

While these generalizations are correct for Mac finite clauses,³ there is one consistent exception that arises in Bg. If the verbal complex is initial, then the clitics must follow rather than precede the verb.

(3)

1. a. (Bg: */Mac: OK) Mi go dade Petko včera.

2. b. (Bg: OK/Mac: *) Dade mi go Petko včera.

This is the well-known Tobler-Mussafia (TM) effect. It results from a phonological difference between the relevant Bg and Mac clitics: in Bg these elements are strictly enclitic, whereas in Mac they are not. When no nonverbal lexical material that could support the clitics occurs in front of the clitics, the verb must precede the clitics in order to provide phonological support for them. This "repair" strategy of pronouncing the verb to the left of the clitics occurs in Bg, but not Mac.

We emphasize that the need to invoke the inverted word order in Bg (3b) is sensitive to the absence of material that could support the preverbal enclitics. Franks (1998, 1999b, 2000) points out that even i 'and' is sufficient to serve this purpose, as shown in (4a).

(4)

1.  I mi go dade Petko včera.

    and me(DAT) it(ACC) gave Petko yesterday

    'And Petko gave it to me yesterday.'

2. b. *I dade mi go Petko včera.

This curious fact will play a pivotal role in the discussion in sections 3 and 4.

2 Approaches to the Tobler-Mussafia Effect

While it is fairly obvious that phonological factors are responsible for the initial "V + clitic" order in Bg, there is less general consensus about how to implement them. In this section we mention three conceptually distinct ways of accommodating the TM effect.

One standard account in the literature involves the PF reordering mechanism of prosodic inversion (PI), popularized by Halpern (1995).⁴ Under this view the output of the syntax leaves the clitics left-adjoined to the verb, and in the phonology the clitic cluster makes a minimal move rightward, to the right edge of the first prosodic word to its right (in this instance, the verb), in order to be properly supported prosodically.

(5) mi go [dade]_ω Petko včera → [dade]_ωmi go Petko včera

PI takes place in Bg, but not in Mac, since only in the...