Syntactic doubling and the encoding of voice
Central and Southern varieties of Italian and Italian dialects are famously subject to so-called
Raddoppiamento fonosintattico (RF), or syntactic doubling, which lengthens the first consonant of a
word when it immediately follows another word that either features final stress or consists of a strong
monosyllable cf. (1). Syntax is often said to play a conditioning role in this phenomenon as there are
various cases where doubling fails to apply despite the fact that the appropriate phonological
conditioning requirements are met (cf. Nespor & Vogel 1986, Loporcaro 1997, and see 2-3). However,
though RF may sometimes carry gender and number information, it is believed not to be able to carry
other kinds of syntactic information. In this paper, we focus on a peculiar doubling phenomenon in
Eastern Abruzzese (EA), a southern Italian dialect, where RF functions specifically to express voice
information. Moreover, we argue that RF in EA offers a transparent window on the nature of the
syntactic factors that trigger doubling. RF in fact differentiates between active (2, no RF) and passive
(3, RF) constructions, which would not otherwise be distinguishable by means of auxiliary selection
(which is person- and not verb class-driven in EA) or any alternative means.
Given the right phonological environment, passive in EA therefore triggers RF, while active does not.
One possible analysis of the alternation shown in (2-3) would be to adopt a perspective in terms of
which the difference between them follows straightforwardly from a lexically encoded difference
between active and passive participles in EA. That this analysis cannot, however, be correct is clearly
shown by (5), which registers two important points: (i) participles which receive a passive
interpretation do not obligatorily exhibit RF and (ii) RF obligatorily applies to whichever element
surfaces in the position adjacent to the passive auxiliary (Aux).
(4-5) thus show that RF in EA is conditioned by adjacency in a specific syntactic context, namely
passive structures. Taking this into account, we therefore propose an alternative phase-based syntax-
PF mapping analysis of RF in EA; more specifically, we propose that RF results whenever the stressed
Aux is sent to Spellout at the same time as the element which exhibits doubling, with the result that
these two elements are assigned to the same phonological phrase (: cf. Selkirk 1995, Truckenbrodt
1995); correspondingly, RF fails to appear whenever these conditions do not apply.
A crucial piece of evidence in favour of our analysis is that the active auxiliary occupies a position
higher than the passive one. Assuming that adverbs occupy a fixed and hierarchically ordered position
in the clause (Cinque 1999), the distribution of ggià relative to so' in (5-6) suggests that the active
auxiliary is located higher than the passive one. This proposal is further substantiated by the behaviour
of EA compound auxiliaries, where the lower auxiliary clearly encodes lexical-aspectual information
and is thus merged in v, whereas the higher auxiliary, which encodes person, number, and tense
information, is merged higher. Significantly, compound auxiliaries may not appear in passive
constructions cf. (8). We take (8) to indicate that the passive auxiliary and the aspectual auxiliary in
compound auxiliary constructions compete for the same position, namely v.
With the positional distinction between active and passive auxiliaries in place, the EA RF patterns
readily follow, on a phase-based analysis. We assume the Phase Impenetrability Condition (PIC) of
Chomsky (2000) and additionally follow Chomsky (ibid.) in identifying active vP, but not passive vP
as a phase. Active structures will thus involve spellout of VP as soon as vP is completed. As active
Aux is only merged after completion of the vP phase, Aux and the active participle can never be sent
to Spellout together and consequently they also cannot be part of the same . This derivational fact
clearly rules out the possibility of RF in active contexts. In passives, by contrast, Aux is merged in
passive (i.e. non-phase head) v; consequently, Aux and adjacent material are sent to Spellout together
after CP is completed, and the phonological environment for RF to apply is thereby created.
If our analysis is correct, the prediction will be that RF will always obtain wherever a phonologically
appropriate Aux is located in the domain of a non-phasal vP, as Aux and the adjacent element will
always be spelled out together. This prediction is in fact borne out: RF always applies in predicative
constructions, which we analyse as vP-internal small clauses, following Moro (2000), Costa & Pereira
(2002) cf. (9).
The EA doubling data presented here provide important insights into the role that syntax plays in the
conditioning of RF phenomena. They also supply an important piece of evidence for the claim that
passive and active auxiliaries can occupy different positions in the clause. Finally, these data show that
the first version of Chomsky's PIC is to be preferred over the most recent one (Chomsky 2001 et seq.),
which would not allow us to derive our syntactic facts correctly.
Data
(1) a. Tre ccase b. Città ccara
three houses-RF city expensive-RF
`three houses' `expensive city'
(2) So' viste
am seen-past participle
`I have seen' (active)
(3) So' vviste
am seen-past participle
`I am seen' (passive)
(4) a. So' sembre viste b. So' ssembre viste
am always seen am always-RF seen
`I have always seen' (active) `I am always seen' (passive)
(5) Ggià so' (*ggià) vviste da tutti quinde
already am already seen by everyone
`I am already seen by everyone'
(6) Ggià li so' (ggià) viste cullù
already him am already seen him
`I have already seen him'
(7) Li so' `ve viste
it am-1st sg had-imperfective seen-pp
`I had seen that'
(8) *Ji so' `ve state viste
I am had been seen
`I had been seen'
(9) Esse jè ssembateche
she-he is nice-RF
`(S)he is nice'