Past participle agreement: split auxiliary selection and the null-subject parameter
Past participle (pp) agreement in Romance languages is usually thought to be governed by the
mechanisms proposed by Kayne (1989) and Belletti (2001). According to Belletti, pp agreement
always takes place with internal arguments. Here, we present counterevidence to this generalization,
and propose a novel analysis of pp agreement, as well as person-driven auxiliary selection, using
previously unknown data from Abruzzese, a southern Italian dialect spoken in Abruzzo.
In Abruzzese, pp agreement may take place either with the subject (1) or with the object (2) of
transitive verbs, with the subject of unergative (3) and unaccusative verbs (4). The pp always agrees
with a plural DP, be it the subject or the object; see (5).
Abruzzese has a "split-ergative" system of auxiliary selection (Manzini & Savoia (2005)).
Many dialects in Central and Southern Italy show this type of auxiliary selection. In these varieties, the
argument structure of the verb is not the conditioning factor, but rather the person-number
specification of the subject. There is much variation, but Abruzzese represents probably the most
common pattern, that where "be" appears with a 1st- or 2nd-person subject, and "have" with a 3rd-
person subject is ((Cocchi (1995, Chapter 4), Kayne (2000, Chapter 7), Ledgeway (2000, Chapter 6),
Loporcaro (1998), Tuttle (1986)). If the auxiliary is merged in v, this indicates that v is sensitive to the
-features of the subject; it is in this sense that such systems are ergative (see Müller (2004)).
We propose that all agreement facts in Abruzzese are resolved within the vP. v has -features
that probe the subject (Müller (2004), Bejar & Rezac (forthcoming)), and V has -features, inherited
from v (Chomsky 2005), which probe the object. In Abruzzese there is no gender agreement, only the
number feature of a DP values the number features of v or V. The pp is merged in V. In (2), where the
pp agrees with the plural object, the object Agrees with V (= with pp). As a result, the object's
Accusative Case is licensed and V-pp has its number feature valued as plural. v Agrees with the
subject, which is merged in Spec, VP. As a result, the subject is valued as Nominative and values v's
-features as 3rd person, and the auxiliary have is selected. The pp in Abruzzese does not raise higher
than V, unlike the Italian pp, which raises at least to v. This is shown in (6), where the adverb tande
(=It. molto, `much') precedes the pp in Abruzzese and follows it in Italian.
In the case of (3), where the pp agrees with the subject, the singular object Agrees with V. As a
result, the object receives Accusative. If we assume that only plural number activates Agree -- in other
words, that [singular] is the default, underspecified value for [number], then V's number feature
remains unvalued. The number feature on the pp remains consequently unvalued. v Agrees with the
subject, and is valued as Nominative. The plural subject values the number feature on v as plural. The
feature content is transmitted from v to V by feature inheritance in the vP phase (analogous to the C-T
transmission in the CP phase proposed by Chomsky 2005), and therefore V, and consequently the pp,
become plural. Hence pp agreement is always with a plural argument, either the subject or the object.
If Nominative case is assigned by v in Abruzzese, it follows that the role of T in Abruzzese is
reduced. T's -features do not probe the subject; in fact they seem to probe only the inflected verb
(given that V raises to T). T's EPP-feature therefore can only trigger verb-movement, not subject
raising to SpecTP. We conclude that varieties with the kind of auxiliary-selection and participle-
agreement system described here will show the phenomenology of null-subject languages of the Greek
type (see Alexiadou & Anagnostopoulou (1998)) in that the Spec,TP position is unavailable. Support
for this claim comes from the impossibility of inserting a subject immediately after a low
complementizer in Abruzzese. In (7a), no subject may appear after ocche (`that'), a low jussive
complementizer. Conversely, this subject position is available in Standard Italian, as shown in (7b).
The correlation with the null-subject parameter explains a significant cross-linguistic
generalization. Although both argument-structure-driven auxiliary selection and the generalized use of
"have" as perfect auxiliary are found in Germanic (German vs. English), the Central-Southern Italian
person-driven system is not found. Our analysis entails that a person-driven system can only be found
in a null-subject language with a defective T of the type described above. Since no Germanic language
is a null-subject language, we understand the typological generalization as a non-trivial case of
parameter interaction.
(1) Giuwanne e Mmarije a pittite nu mure
have-3rd sg/pl painted-pp pl
John and Mary a-sg masc wall-sg masc
`John and Mary have painted a wall'
(2) Giuwanne a pittite ddu mure
has-3rd sg/pl
John painted-pp pl two walls
`John has painted two walls'
(3) Nu seme tilifunite
we-1st pl are-1st pl called-pp pl
`We called'
(4) Vu sete arrivite
you-2nd pl are-2nd pl arrived-pp pl
`You arrived'
(5) a. Giuwanne a pittate nu mure/pittite ddu mure
has-3rd sg/pl painted-pp sg a
John wall / painted-pp pl two walls
`John has painted a wall/painted two walls'
b. Giuwanne e Mmarije a pittite nu mure/pittite ddu mure
have-3rd sg/pl painted-pp pl a wall /painted-pp pl two walls
John and Mary
`John and Mary have painted a wall/painted two walls'
(6)a. Ni li so tante capite b. Non ho capito molto
not it am-1st sg much understood-pp sg not have-1st sg understood much
`I didn't understand much'
(7)a.*Ocche Giuwanne zi li magne
b. Che Gianni se lo mangi
that John himself it ate
`Let John eat it'