Summary

The Unaccusative Hypothesis (Perlmutter 1978, Burzio 1986) stipulates that, across languages, intransitive verbs are of two types, called unaccusative and unergative, which have distinct syntactic and semantic properties. A lack of correspondence has often been observed, however, between the semantic components postulated for a verb and the syntactic behavior that might be predicted on the basis of those components. Since most of the diagnostics of unaccusativity/unergativity (such as auxiliary selection in French, impersonal passives in Dutch, resultative constructions in English, etc.) tend to identify semantically coherent subsets of verbs, the challenge for a theory of unaccusativity is to single out the syntactically relevant components of meaning in different languages. Focusing on the systematic linking of a multi-categorial lexical-semantic level to a necessarily binary and discrete syntactic level, Sorace (Sorace 1992, 1993, 1995, forthcoming) has shown that (potentially universal) gradient continua- "hierarchies" - distinguish core unaccusative and unergative verbs from progressively more peripheral verbs. These hierarchies, which are based on aspectual parameters, identify the notion of "telic dynamic change" at the core of unaccusativity and that of "atelic non-dynamic activity" at the core of unergativity. The closer to the core a verb type is, the stronger the link between its single argument and the position of internal or external argument at argument structure, and the more determinate its syntactic status as either unaccusative or unergative. This theory substantiates the intuition that, within the class of unaccusatuve verbs, some are "more unaccusative" than others, and within the class of unergative verbs, some are "more unergative" than others (Legendre, Miyata and Smolensky 1995).

The field of Romance dialectology represents an extraordinarily rich source of experimental data for linguistic theories: one of the advantages of working with variants of the same language type is the possibility to investigate the limits of variation for given syntactic phenomena while at the same time keeping other syntactic variables constant (cf Benincà 1994). This paper applies the theoretical framework described above in order to explore, both at the synchronic and at the diachronic level, the syntactic manifestations of split intransitivity in the grammar of Paduan, a dialect spoken in the Northeastern region of Veneto. In the existing descriptions of auxiliary usage in Paduan (Benincà and Vanelli 1984; Benincà 1994; Loporcaro and Vigolo 1995; Loporcaro 1998), attested variation in auxiliary selection is well-documented but tends to be left unexplained, or is simply attributed to lexical idiosyncrasy. This paper will show that the pattern of variation is systematic and that it can be accounted for in terms of Sorace’s unaccusative and unergative hierarchies.

The data collection took place in June 1997. Thanks to the help of local contacts, in-depth interviews were conducted with a total of 24 native Paduan speakers. Interviews were conducted both in Padova itself and in the small adjacent towns of Conselve and Ponte Vigo d'Arzere. Lexical and, to a lesser extent, syntactic variation was observed between geographically adjacent dialectal varieties; such variation, however, did not affect auxiliary usage in any significant way.

Speakers interviewed were representative of different socio-economic backgrounds (roughly corresponding to working class, middle class, and upper-middle class) and of different age groups (age range: 18 to 76). While all speakers had good receptive command of standard Italian, their productive command varied as a function of age and socio-economic background, with older, lower-status speakers tending towards exclusive use of the dialect in production and younger, middle class speakers tending towards exclusive use of Italian.

The format of interviews was the following. Speakers were given sentences in Italian and were asked to translate them into their dialect. Questions followed a pre-determined order in a questionnaire (visible only to the interviewers) that had been specially designed for this purpose: sentences including reflexive verbs were presented first, followed by unergative verbs and unaccusative verbs, and finally by impersonal constructions. For most verbs, informants were asked to supply verbal forms in all persons. The order of question presentation was flexible to allow a more in-depth investigation of interesting points arising during individual interviews. For certain verbs, representative of particular syntactic classes, informants were also presented with ungrammatical sentences including the 'wrong' auxiliary, and asked whether they regarded these sentences as totally impossible, or possible only in some contexts or when pronounced by other speakers.

Informants were (willingly) tape-recorded throughout the interview, which lasted about one hour. Extensive written notes were taken by the interviewers. Most interviews took place at the informants' home; some took place at Conselve Hospital.

The fieldwork tested a variety of syntactic constructions and verb types and their sensitivity to auxiliary selection, ne-cliticization and past participle agreement. The general pattern revealed by the data suggests that these three properties cluster together for verb categories corresponding the core of the unaccusative/unergative classes, but are more loosely related, or not related at all, with non-core verb categories. This paper will primarily deal with auxiliary selection since in Paduan it appears to be the most consistent manifestation of split intransitivity: unlike standard Italian, neither ne-cliticization not past participle agreement are, by themselves, reliable diagnostics of the unaccusative/unergative distinction. The remainder of this outline will focus only on the patterns of variation in auxiliary choice.

As predicted, two classes of verbs present no variation at all in auxiliary selection: (a) verbs denoting inherently telic change of location/state (e.g. vignere ‘come’, rivare ‘arrive’) and (b) verbs denoting non-motional volitional activity (e.g. lavorare ‘work’). These verb types have been identified in previous work as core unaccusative and core unergative, respectively: they have been found to be consistent in selecting auxiliaries essere ‘be’ (henceforth E) and avere ‘have’ (henceforth A) in various languages that have a choice of auxiliaries. Paduan is no exception in this regard: all informants, regardless of age, select E with inherently telic verbs of change of location and A with verbs of non-motional volitional activity.

On the other hand, alternations in auxiliary selection in Paduan appear to occur with the following classes of verbs:

a. anticausative verbs in reflexive form (e.g. romperse ‘break’)

b. anticausative verbs in non-reflexive form (e.g. aumentar ‘increase’)

c. monadic verbs denoting non inherently telic change of state/location (e.g marcire ‘rot’)

d. stative verbs denoting continuation of a pre-existing state (EX) or existence of a state

(e.g. durar ‘last’, bastar ‘suffice’)

e. verbs denoting manner of motion (e.g. correre, ‘run’))

f. verbs of (smell, light, substance and sound) emission (e.g. spussare ‘stink’, splender ‘shine’).

The patterns of variation appear to be sensitive to age differences among the informants: in general, the younger the speaker, the stronger the influence of standard Italian, and thus the more extensive the use of E. The data have been analysed according to four age bands: young (age range: 18-22), early middle-age (age range: 40-50), late middle age (age range: 50-60), older (age range: 70-85).

The variation in the Paduan data confirms that, in general, inherent lexical aspect determines auxiliary choice with core verb categories, whereas compositional aspect (i.e. the event structure of the whole predicate) affects auxiliary selection with peripheral verb categories. The data support the prediction that in Paduan, as in other languages, auxiliary selection with core verb types is a lexical phenomenon and is relatively insensitive to compositional factors. The degree of sensitivity to these factors increases for non-core verb types as they get more distant from the core.

4.1 Unaccusative verbs

One of the interesting facts uncovered by our analysis is the behaviour of reflexive anticausative verbs which, unlike in standard Italian, display a sensitivity to parameters such as external/internal causation and the eventive/stative nature of the predicate.

(1) I pomi i ze kaska e se ga makka

the apples they-masc are fallen and refl-have bruised

(2) Go provà metare la ciave ma la porta no se ga verto / *no se ze verta

have-1s tried put the key but the door not refl-has opened / not refl-is opened

(3) Go butà par tera el vaso ma no se ga roto / *no se ze roto

have-1s thrown on the floor the vase but not refl-has broken / not refl-is broken

(4) El vaso ze andà in tera e se ga spakà / se ze spakà

the vase is gone to floor and refl-has broken / refl-is broken

The choice of E reflects the lack of external causation, so that a sentence such as

(5) La porta se ze verta de colpo

the door refl-is opened suddenly

is acceptable only if it is understood that the door opened by itself, or by virtue or an inanimate (and unexpressed) causer (e.g. the wind). In contrast, the choice of A in (2) is related to an interpretation in which the event is seen as being brought about by an external (animate) causer.

On the other hand, in (5) the selection of A seems to signal the eventive nature of the situation described by the predicate, which is explicitly signalled by the adverbial di colpo ‘suddenly’. When there is no indication of the eventive/stative or internally/externally caused nature of the situation, as in (4) both auxiliaries are possible.

The conceptualization of the situation as eventive or stative also appears to determine A/E alternations with non-reflexive anticausative verbs. The past participle of these verbs may in fact be ambiguous between a verbal and an adjectival interpretation. This may in turn be related to the lack of the reflexive morpheme, which in many languages functions as a marker of telicity.

(6) a. A ferita a ga gwario in freta

the wound it-fem has healed quickly

b. A ferita a ze gwaria

the wound it-fem is healed

With verbs denoting non-inherently telic change of state/location, middle-aged and older informants show A/E alternations which appear to be sensitive to the degree of subject control and to the degree of telicity of the action. Thus, (7a) implies that Maria slid, but did not fall, whereas (7b) implies that she actually fell. In (7c), E is the only possible auxiliary.

(7) a. Maria la ga sbrissà su na scorsa de banana

Maria she has slid on a skin of banana

Maria slid on a banana skin

b. Maria la ze sbrissa su na scorsa de banana

Maria she is slid on a skin of banana

Maria slid on a banana skin

c. Maria la ze sbrisà su na scorza de banana e se ga spakà na gamba

Maria she is slid on a skin of banana and refl-has broken a leg

The choice of auxiliary with stative verbs (both those denoting continuation of a pre-existing state and those denoting existence of a state) is highly variable for the majority of informants. Both A and E are accepted with most verbs by the same speakers, although older informants prefer A and younger informants prefer E. Exceptions are restare ‘stay’, which is preferred with E, and parere ‘seem’, which is preferred with A.

(8) Il detersivo me ga dura / ze dura poko

the washing powder to-me has lasted / is lasted little

The washing powder lasted for a short time

(9) La gwera (la) ga dura /ze dura dieze ani

the war it-fem has lasted / is lasted ten years

The war lasted for ten years

(10) La farina me ga basta / ze basta par un dolce solo

the flour to-me has sufficed / is sufficed for a cake only

The flour was sufficient only for one cake

(11) A notisja me ga parsa strana

the news to-me has seemed strange

(12) Maria ze sta / ga sta a kaza koi fijoli

Maria is stayed / has stayed at home with the children

Maria stayed at home with her children

4.2 Unergative verbs

Variation in auxiliary selection is displayed by verbs denoting manner of motion. In most cases, auxiliary alternations are constrained by compositional aspect (like in standard Italian), and in particular by the presence of a directional phrase indicating telicity. Thus,

E and A are the only possible choices in (13a) and (13b) respectively; (13c) implies that Maria kept going in and out of the park.

(13) a. Maria la ze corsa casa

Maria she is run home

Maria ran home

b. Maria la ga corso par tre ore in tel parco

Maria she has run for three hours in the park

c. Maria la ze corsa par tre ore in tel parco

Maria she is run for three hours in the park

With ne-cliticization the telic/ atelic distinction no longer holds, and both A and E may occur even if the directionality is overtly expressed, as in (14).

(14) Ghe ne ga corso /ghe ne ze corsi tanti a casa

of-them have run many / of-them are run many home

Many of them ran home

In other cases, the choice of auxiliary is determined by whether the emphasis is on the process or on the endpoint of the action. In (15a), the interpretation is that the ball rolled for a while before it fell into the hole; in (15b), the focus is on the ball in the hole: a natural continuation might be "now take it out". If directionality is not specified, as in (15c), both auxiliaries are possible.

(15) a. La bala la ga rodola in tel buzo

the ball it-fem has rolled in the hole

The ball rolled into the hole

b. La bala la ze rodolada in tel buzo

the ball it-fem is rolled in the hole

c. Ghe ne ga rodola tre / ghe ne ze rodolae tre

there of-them have rolled three / there of-them are rolled three

With verbs denoting emission, most informants accept both auxiliaries, with only marginal (and sometimes inconsistent) preferences for one or the other.

(16) Ea giaka ga spusa / ze spusa de fumo

the jacket has reeked / is reeked of smoke

My jacket reeked of smoke

(17) L’akwa a ze skorsa /la ga skorsa par tante ore

the water it-fem is run / it-fem has run for many hours

The water ran for many hours

(18) el tron el ze rimbomba / ga rimbomba jeri per tre volte

the thunder it-masc is rumbled / has rumbled yesterday for three times

The thunder rumbled three times yesterday

(19) L’ortensia a ga fjoria / ze fjoria in tre giorni

the hortensia it-fem has blossomed / is blossomed in three days

(20) L’eco ga rizona /ze rizona

the echo has resounded / is resounded

The echo resounded

(21) la luna a ga spleduo / ze splendua

the moon it-fem has sparkled / is sparkled

The moon shone

(22) L’akwa la ga sgosada / ze sgosada per tuto el giorno

the water it-fem has dripped / is dripped for all the day

The water dripped all day

However, verbs of sound emission tend to be preferred with A. This preference is likely to reflect the greater agentivity of these verbs since, as Levin and Rappaport Hovav (1995: 170) suggest, they "require an input of energy for the maintenance of the eventuality".

Conclusions

The Paduan data provide support for the hypothesis that generated this research: namely, that auxiliary selection is not only determined syntactically by the unaccusativity or unergativity of the verb, but also semantically conditioned by aspectual factors in a hierarchical way. Auxiliary A has a wider distribution in Paduan than in standard Italian, but variation affects only peripheral verb classes. In both languages, core verb classes exhibit stronger unaccusativity / unergativity and thus are categorical in auxiliary selection. Paduan non-core verb types, to a greater extent than their Italian equivalents, display a sensitivity to factors other than the inherent lexical features of the verb, such as the eventive nature of the predicate and the degree of affectedness of the subject.