The syntax of subject clitics in the dialect of Genoa
Research MA Thesis in Linguistics
August 2018
Lorenzo van Velzen
supervisors: dr. M Scorretti & prof. dr. E.O. Aboh
Universiteit van Amsterdam
Abstract
This MA thesis adds to the corpus of research on subject clitics (SCLs) in Northern Italian dialects (NIDs). It is the first study to focus on the subject clitics of the dialect of Genoa: it provides a comprehensive overview of the distribution of Genoese SCLs and proposes a structural syntactic account for these SCLs. The results indicate that in many respects, Genoese patterns with what has been found for other NIDs with respect to their SCLs: it possesses an obligatory 2nd person singular SCL, ti, and an optional 3rd person singular SCL,
o/a. With respect to the syntax of these SCLs, I argue that they occupy different structural
positions. Regarding the optionality of the 3rd person singular SCL, I argue that it depends on specific semantic and pragmatic features: specificity, animacy, referentiality and impliciteness. The presence of any of these factors contributes to the likelihood of appearance of the 3rd person SCL.
Acknowledgements
I would like to thank the supervisors of this thesis for their guidance, which has been a mix of relaxed, pleasant and enthusiastic. This created an atmosphere in which I could start, continue and finish working on this thesis after having lost academic direction. To Mauro, thank you for your enthusiasm and curiosity throughout both my BA and MA, and for always caring about the subject matter above anything else. To Enoch, thank you for your eternal honesty with respect to this thesis and previous works and courses, and for your interest in my general development as a student. Further thanks are to prof. dr. Jeannette Schaeffer, who throughout my MA has always been prepared to lend a listening ear and give sound advice, either solicited or unsolicited. Finally, I wish to thank my parents, who never cease to support me and who assured me that it was OK if I struggled for a while. You are always soft and gentle with me, but thank you mama for taking off the white gloves and telling me to get this done.
Introduction
For the last few decades, the study of Northern Italian dialects has proven to be a rich field of enquiry with great relevance for the theory of syntax. Northern Italian dialects (NIDs) show extensive syntactic variation between themselves and with other languages and dialects in the
Romance language family; the study then of Italian syntax (Rizzi 1982), French syntax (Kayne 1975, 1983) and of the syntax of the NIDs is highly relevant to them and has born and continues to bear fruit to new analytical perspectives. Especially the investigation of NIDs that have not yet been studied with as much scrutiny as others may help refine syntactic theory.
This MA Thesis dives into the syntax of subject clitics, a much debated subject in Romance syntax. It does so by focusing on the dialect of Genoa, the capital of the region of Liguria in northwestern Italy. For this dialect, a comprehensive study of its subject clitics (SCLs) has been lacking in academic literature. By investigating the SCLs of this dialect, this study does two things: first, it fills an empirical gap; second, from the data it extracts an analysis with as broad a theoretical relevance as possible. The question that this study asks is the following: how can we explain the distribution of subject clitics in Genoese, and what
does this tell us about the typology of SCLs in Northern Italian dialects and other Romance languages?
It is important to note that this study mostly looks at the subject matter from a generative perspective. The reasons for this are the following: most work in SCLs has been done within the generative framework; within this framework it is possible to investigate details of the distributive properties of clitics; using this framework, generalizations can be made over the typology of clitics within Romance languages; the proposed typology of Romance clitics bears on general questions of generative concepts of clause structure and licensing of agreement.
I will first give an overview of previous work on subject clitics in Northern Italian dialects, the theoretical proposals to which they have borne fruit and the criticisms that these are subject to. I will then explain my hypothesis and predictions, and the method that I employ to test these. An overview and analysis of my results follows, after which I add a few concluding remarks on unresolved issues and this thesis in general.
1. The syntax of Northern Italian subject clitics
Studying the syntax of Northern Italian subject clitics, and arguably Romance subject expression in general, is meaningful by virtue of the extensive variation that we find in this domain. It has become commonplace to treat Northern Italian subject clitics in the context of a comparison between subject expression in French and Standard Italian. On the one hand, this is because French utilizes clitic pronouns for subject expression as well, and these have been studied extensively (Kayne 1975, 1982; Sportiche 1999; etc.); on the other hand, NIDs share many syntactic features with Standard Italian, of which most NID speakers are native or at least fluent speakers. Further, the variation between the NIDs is crucial in trying to determine the syntactic status of their SCLs. In this section, I give an overview of the literature that has been dedicated to this topic over the years, including theoretical proposals that have been made and the criticisms that they are subject to.
1.1 What are subject clitics?
Subject clitics can roughly be described as follows: they are pronominal elements that express features of the subject and cannot occur independently of the verb. For instance, a subject clitic cannot be used as fragmented answer or with contrastive focus:
Genoese1
(1) Q: Chi l2’ è vegnùo?
who l’AUX come-PTCP
‘Who came?’ A: *O / Lé.
SCL.3SG.M / PRN.3SG.M
‘He.’
(2) Ghe son anæto mi, no *o / lé
LOC AUX go-PTCP PRN.1SG NEG SCL.3SG.M / PRN.3SG.M
‘I went there, not him.’
In the Genoese examples above, use of the subject clitic o is ungrammatical; rather, use of the full pronoun lé is required. Subject clitics only appear with finite verbs, and may express different (numbers of) features of the subject. Their position is always dependent on the verb and other clitics (if present). Depending on the definition used, they may appear in combination with a full lexical subject DP or not, they may be omitted in various syntactic contexts, they may appear in different position with respect to other constituents such as negation, and so on. A more detailed explanation of this follows under 1.2.
Early seminal works concerning SCLs as they appear in Northern Italian dialects include Brandi & Cordin (1981) and Renzi & Vanelli (1983): they are arguably the first to focus on SCLs in NIDs, providing a base that is followed or elaborated upon by many authors afterwards. On the basis of Trentino and Florentine, Brandi & Cordin (1981) identify Northern Italian SCLs as similar, but not identical, to French SCLs. In their comparison of Northern Italian and Standard Italian subject expression, they find that SCLs serve some of the purposes that in Standard Italian are fulfilled by verbal inflection. Renzi & Vanelli (1983) investigate the variation that is found between NIDs with respect to their SCLs, first establishing a typology of subject expression in about 30 NIDs and binding solid generalizations to their findings; these include the fact that there exist implicational relations between the different types of SCLs that any dialect may exhibit, and also which types of SCLs are more likely to be obligatory than others.
1.2 Pro-drop
Arguably the most important syntactic feature with which to start an analysis of Northern Italian SCLs is the pro-drop parameter. The pro-drop parameter, or null subject parameter, determines whether a language must overtly express the subject of a tensed predicate (i.e.
1 All Genoese data in this work is from original research.
2 The nature of l’ not at all clear. I will dedicate more time to this element later in this work; here, I will merely gloss it as ‘l’’.
non-null subject languages) or not (i.e. null subject languages). Many languages, such as English, Dutch, German, and French do not allow the omission of the subject in this context; other languages, such as Italian and Spanish, do.
English
(3a) *(He) doesn’t eat meat
Dutch
(3b) *(Hij) eet geen vlees
he eat-PRS.3SG no meat
Italian
(3c) (?Lui) non mangia carne3
he not eat-PRS.3SG meat
In generative theory, it is assumed that languages that do not overtly express a subject are not lacking a subject; rather, the subject position is filled by a morphologically empty element called pro (Jaeggli & Safir 1989). One way to put the idea of pro-drop is as follows: languages that allow for the omission of the subject have ‘strong’ verbal morphology from which certain properties of the subject, such as person, number and gender features (φ-features), can be recovered. This would mean that pro is licensed by the rich inflection on the verb. On the contrary, languages that do not allow this have ‘weak’ verbal morphology, as a consequence of which pro cannot be licensed since its syntactic features cannot be recovered. Such languages must therefore overtly express the subject. In Rizzi’s (1982) terms, pro in subject position [Spec, VP] is allowed in languages in which INFL(ection) carries a feature [+pronoun]; if a language’s INFL does not have this feature, it cannot, thus requiring a full subject DP in [Spec, VP].
Null-subject language Non-null subject language
In the above schematic representations, pro is allowed when (I) carries the same feature [+pronoun] as [Spec, VP]; in modern terms (Adger 2003), this means that the [+pronoun] feature on (I) can ‘check’ the [+pronoun] feature in [Spec, VP]. This is not the case for the non-null subject language, where the [+pronoun] feature of [Spec, VP] cannot be checked by (I): agreement being weak in this language, (I) lacks this feature. As such, further linguistic
3 Though grammatical, expression of the pronominal subject in non-marked cases (as opposed to cases of contrastive stress, focalization etc.) is in fact often pragmatically infelicitous in Italian.
material is required that can satisfy the feature specification of [Spec, VP], i.e. a subject DP. The above example is one way in which pro-drop can be accounted for, but the precise nature of this parameter remains subject to debate:
“While different theories assign a different role to this relation [between agreement and identification of null subjects], all of them agree that it is the special status of the inflectional system that allows null subjects.” (Jaeggli & Safir 1989:21).
The pro-drop parameter is crucial for the characterization of Northern Italian subject clitics, because whether the SCLs appear in non-null subject languages (NNSLs) or null-subject languages (NSLs) has direct consequences for their syntactic status. Early studies on Northern Italian SCLs have to a large extent compared Northern Italian SCLs with subject clitic pronouns in French pronouns, which is usually considered a NNSL. However, most scholars to date agree that there is evidence that NIDs are in fact NSLs, similarly to Standard Italian; arguments for/against such an analysis are explained in the following.
1.2.1 Relation to subject DPs
Many Northern Italian dialects allow the expression of a lexical subject DP in combination with a subject clitic. In (4a), the subject clitic appears on its own, i.e. without a lexical subject DP. In (4b), the subject DP el Gianni appears alongside the subject clitic. As the asterisks in both examples show, the subject clitic el may not be omitted in either example, whereas the subject DP is free to be present or absent.
Trentino (Rizzi 1986b4)
(4a) *(El) magna.
SCL.3SG.M eat-PRS.3SG]
‘He eats.’
(4b) El Gianni *(el) magna.5,6
DET.SG.M Gianni SCL.3SG.M eat-PRS.3SG ‘Gianni eats.’
What these examples from Trentino show is that 3rd person singular SCL el is obligatory in
combination with the finite verb and that it may appear alongside a lexical subject DP (4b). The subject DP in (4b) bears no focal stress and has no topicalized interpretation, so a left dislocated analysis is unlikely; further support for this follows later in this section. Further, (4a) and (4b) show that the verb in Trentino may not appear without the subject clitic, and that a lexical subject DP must be doubled by the clitic.
In some NIDs, however, the clitic is not (always) obligatory: Genoese is such a dialect. In any case, NIDs contrast with Standard Italian, which has no such element:
Genoese
(5) (A) beive vin.
4 All examples in section 1.2.1 are from Rizzi (1986b) unless noted otherwise. 5 In many NIDs, it is common that a personal name is preceded by a determiner.
(SCL.SG.F) drink-PRS.3SG wine
‘She drinks wine.’ Italian
(6) Beve vino.
drink-PRS.3SG wine
‘She drinks wine’
Example (5) shows that in some NIDs, the verb may appear on its own, as SCLs may be optional; example (6) shows that Standard Italian allows the verb to appear on its own, i.e. without an overt subject DP and/or clitic. In fact, it is impossible to insert an element like the Northern Italian SCLs: either the verb appears with a covert subject, or with a full subject DP.
Within Romance, Standard Italian behaves differently from Standard French, an apparent non-pro-drop language. In Standard French, subject clitics or weak pronouns je, tu, il, elle,
on, nous, vous, ils, elles are in complementary distribution with DPs in the subject position7:
(7a) Jean mange.
Jean eat-PRS.3SG ‘Jean eats.’ (7b) Il mange. he eat-PRS.3SG ‘He eats.’ (7c) *Jean il mange. Jean he eat-PRS.3SG
‘Jean (he) eats.’
(7c) is marked ungrammatical here for the unmarked, non-topicalized reading. If this is the case, then this may mean that lexical subject DP Jean and clitic pronoun il compete for the same position. However, we cannot uncontroversially explain the ungrammaticality of (7c) based on the intended reading. Indeed, many modern speakers of French find example (7c) perfectly acceptable even without any intonational break between Jean and il (Renzi 1992; Zribi-Hertz 1994). This either means that left dislocated constructions in French need no particular intonational contour, or that the syntax of DP + SCL is in fact changing. I cannot analyze these questions in detail here, so the status of these constructions remains unconcluded8.
7 Sportiche (1999) does not treat this complementarity as true competition for the same position (section 1.3.1).
8 It might be the case that Poletto’s (1993) diachronic analysis for Northern Italian SCLs is applicable to French as well: Poletto (1993a:12) argues that SCLs originate as free pronouns, then becoming clitics in the phonological, but not in the syntactic sense (such as in standard, ‘correct’ French); then, they are reanalyzed as syntactic heads that move from the VP; finally, they acquire functions similar verbal inflection, licensing and/or identifying pro and/or playing a role in Case-assignment. Modern, spoken French would fall into either one of the latter categories, with SCLs coming to behave as syntactic clitics.
While linguists debate as to whether the French structures above are left dislocations, there is evidence that a left dislocated analysis cannot extend to NIDs. As Rizzi (1986b) explains, pronouns cannot appear in combination with co-referential QPs that are left dislocated (the reader is referred to Rizzi (1986b) for a detailed explanation). They must, however, appear with co-referential, left-dislocated lexical DPs. For Standard Italian, this gives the following results:
(8a) Gianni, *(lo) conosco
Gianni, *(OCL.3SG.M) know-PRS.1SG
‘John, I know (him).’
(8b) Nessuno, (*lo) conosco in questa città
no-one, (*OCL.3SG.M) know-PRS.1SG in this city
‘No one, I know (him) in this city.’
Now, if subject DPs would occupy a left dislocated position in NIDs, one would expect the same contrast: lexical DPs would allow the SCL, whereas QPs would not. However, this is not what is found; QPs can freely appear with subject clitics in many NIDs, which by consequence cannot be called (full) pronouns, at least per Rizzi’s (1986b) analysis.
Let us first establish that NIDs, like Standard Italian, do in fact employ left dislocation productively. NIDs are SVO languages like Standard Italian; the examples below deviate from this order, but are perfectly grammatical:
Veneto (Goria 2004:76)
(9) Nane, el gelato, el lo ga za magnà
John, DET.SG.M ice-cream, SCL.3SG.M OCL.SG.M AUX already eat.PTCP
‘John, the ice cream, he’s already eaten it.’ Genoese
(10) A mèia a veuggio
DET.SG.F apple OCL.SG.F want-PRS.1SG
‘The apple, I want it.’
(11) A menestra no a mangia nisciùn
DET.SG.F soup NEGOCL.SG.F eat-PRS.3SG no-one
The soup, no-one is eating it.
These examples show that NIDs exhibit left dislocation and that the subject is doubled by an object clitic in such structures. Now, as Rizzi (1986b) points out, left dislocated QPs (or indefinites in general) typically cannot be doubled by a clitic/pronoun. The fact then that we find indefinite subjects in NIDs that are doubled by a subject clitic suggests that these subjects are in fact not left dislocated.
(12) Gnun l’ha dit gnent
no-one SCL.3SG.M AUX say.PTCP nothing ‘No one’s said anything.’
Fiorentino
(13) Nessuno l’ha detto nulla
no-one SCL.3SG.M AUX say.PTCP nothing ‘No one’s said anything.’
In other NIDs however, we do find that indefinites prevent the appearance of SCLs (Poletto 1993b:210):
Paduan/Venetian/Triestino
(14) Nisun (*el) vien
no-one SCL.3SG.M come-PRS.3SG
‘No one’s coming.’
Thus, subject clitics in several NIDs may appear with preverbal subject DPs and QPs, which in those cases are not in a left dislocated position and as such most logically occupy the subject position. Consequently, their SCLs cannot occupy the same position. The fact then that the subject position need not be phonetically expressed in NIDs (cf. (1a-d)) means that NIDs are in fact null subject languages: they allow an empty element in subject position, which many agree to be pro. Many have argued that the subject clitic in NIDs, in combination with the verbal inflection, serves to license and/or identify this pro. The difference with Standard Italian would be that in Standard Italian, the inflection visible on the verb licenses and identifies pro, whereas the verbal inflection is not ‘strong’ enough to do this autonomously in NIDs, thus requiring another element. Countering consensus, Sportiche (1999) argues for a similar analysis of French: SCLs license a pro, but differently from NIDs, the specifier of the projection that the SCL heads cannot be filled (due to the ‘doubly filled NomP constraint’ explained under section 1.3.1). In other words, if in French there is an SCL, then the subject position must be filled by pro, whereas NIDs have the option of filling it with either pro or a lexical subject.
Rizzi (1986b) argues that SCLs are, in fact, part of the verbal inflection. Structurally this means that the SCL heads INFL, or the inflectional phrase (IP):
(15)
Of course, this proposal was made before Pollock’s (1989) Split INFL hypothesis, which argues for a more fine-grained structure of the IP. I will not go into much detail about this here, but minimally it means that the IP is split into a tense phrase (TP) and an agreement phrase (AgrP), respectively carrying tense features, and person, number and gender features, with SCLs heading the AgrP. It is currently agreed upon by many that Northern Italian SCLs are syntactic heads in the agreement domain, e.g. by Brandi & Cordin (1989), Belletti (1993), Poletto (1993a,b, 1996, 2000), Zribi-Hertz (1994), Goria (2003, 2004). This view is compatible with the idea that agreement morphemes may develop from clitic pronouns, which in turn may originate from lexical pronouns. It is also compatible with what is known of subject and object markers in other languages, e.g. Bantu languages which mark the subject and the object on the verb.
1.2.2 Negation
More evidence that Northern Italian SCLs occupy different positions than French SCLs comes from their position with respect to negation: Northern Italian SCLs may appear either before or after a preverbal negative marker, whereas the postnegative position is barred for their French counterparts:
Romagnolo
(16) (Maria) la ‘n ve brisa
(Maria) SCL.3SG.F NEG come-PRS.3SGNEG
‘(Maria) isn’t coming.’ Standard French
(17) Elle ne vient pas
SCL.SG.FNEG come-PRS.3SG NEG
‘She isn’t coming.’ Fiorentino
(18a) (La Maria) la un parla
(DET.SG.F Maria) SCL.3SG.F NEG speak-PRS.3SG
(18b) (Te) t’un parli
(You) SCL.2SG NEG speak-PRS.2SG
‘(You) don’t speak.’
(18c) (Te) un tu parli
(You) NEGSCL.2SG speak-PRS.2SG
‘(You) don’t speak.’
Rizzi (1986b) argues that this shows that in NIDs, SCLs occur in the same clitic cluster (under INFL) as does negation, whereas this is not the case for French.
The precise ordering of the subject and negation clitics varies among the dialects and the SCLs. Poletto (1993a) sees this as part of the evidence for an elaborate, fine-grained distinction between different types of SCLs that occupy different positions. This is illustrated by the following Genoese examples:
Genoese
(19a) No ti veddi
NEGSCL.2SG see-PRS.2SG
‘You don’t see.’
(19b) O no vedde
SCL.3SG.M NEG see-PRS.3SG
‘He doesn’t see.’
In these examples, negation precedes the 2nd person SCL, but follows the 3rd person SCL. One
could argue, as Poletto (1993a) does, that this entails a different structural position for the 2nd
person SCL and the 3rd person SCL. This is compatible with independent work by Zanuttini
(1997) on the different structural positions of negation within a clause.
1.2.3 Coordination
Further arguments for a pro-drop analysis of NIDs come from coordination. The repetition of SCLs in both conjuncts of coordinated sentences is often obligatory in NIDs, whereas this is not the case for subject DPs in NNSLs:
English
(20) John reads books and rides his bike.
Dutch
(21) Jij drinkt koffie en eet taart.
you drink-PRS.3SG coffee and eat-PRS.3SG cake ‘You drink coffee and eat cake.’
This is to be contrasted with NIDs: Trentino (Rizzi 1986b)
(22) La canta e *(la) bala.
SCL.3SG.F sing-PRS.3SG and *(SCL.3SG.F) dance-PRS.3SG
‘She sings and dances.’ Genoese
(23) (Lé) o mangia pan e *(o) beive vin.
(PRN.3SG.M)SCL.3SG.M eat-PRS.3SG bread and SCL.3SG.M drink-SCL.3SG.M wine
‘He eats bread and drinks wine.’
Examples (22) and (23) show that Trentino and Genoese SCLs may not be omitted in the second conjunct of coordination; however, the full pronoun lé in (23), presumably in subject position, need not be repeated. One might take this as further support that Northern Italian SCLs are structurally situated lower than the subject position. However, Poletto (1993a) shows that many elements that she analyzes as subject clitics may or may not be repeated in coordination. This is again a motivation for her to propose a multi-layered functional field to account for the different types of SCLs found in NIDs (Poletto 2000):
Veneto (Poletto 1993:18,23)
(24a) La magna pomi e *(la) beve cafè.
SCL.3SG.F eat-PRS.3SG apples and *(SCL.3SG.F) drink-PRS.3SG coffee
‘She eats apples and drinks coffee.’
(24b) A9 magno pomi e (?a) bevo cafè.
SCL eat-PRS.1SG apples and (?SCL) drink-PRS.1SG coffee
‘I eat apples and drink coffee.’
In (24a), the exclusion of the second la entails ungrammaticality, whereas the inclusion of the second a in (24b) is slightly awkward, much like the overt pronunciation of a pronoun in a non-marked sentence in Italian:
(25) ?Lui verrà domani.
?he come-FUT.3SG tomorrow
‘He’ll come tomorrow.’
Some NIDs however seem sensitive to which types of verbs/phrases are coordinated: the examples above either consist of two separate verbs and their complements (22, 24) or two separate verbs with a shared object (23); these are what Poletto (1993) calls ‘type 1’ and ‘type 2’ coordination, for which NIDs typically require repetition of the SCL. There is however a
9 Given that it is unlikely that we should ascribe any person, number or gender features of the subject to this clitic (see below), I gloss it merely as ‘SCL’ with no further specifications.
third type of coordination: it consists of the repetition of the same verb, only with different aspectual specifications:
(26) T’ lo lese e rilese continuament (Piedmontese, Goria 2004)
SCL.2SG it read and reread continuously
‘You read and reread it all the time.’
In this example a non-iterative verb (‘read’) is coordinated with its iterative counterpart (‘reread’, expressed by suffix ri-/re-). This coordinated sentence is grammatical in Piedmontese; however, its counterpart in which the SCL is repeated in both conjuncts is less marked. Still, this type of omission is more frequent with coordinated sentences of type 3 than of types 1 and 2.
Traditionally, Northern Italian coordination data has been contrasted with French. Akin to the discussion about (non-)complementarity of SCLs with subject DPs above, Northern Italian coordination has been argued to differ from French on the basis of the pro-drop status of NIDs: if we accept that NIDs are pro-pro-drop, and that their SCLs are situated somewhere lower than the subject position (e.g., in AgrP which is a sister to the VP), then coordinating predicates could mean the requirement of the multiple expression of the SCL. If we then accept that French SCLs are in fact subject DPs and occupy the specifier of TP, then coordinated predicates would not require the repeated expression of the SCL. Rizzi (1986b) utilizes this assumption to contrast NIDs with French (example (22) repeated here as (28)): French
(27) Elle chante et dance.
she sing-PRS.3SG and dance-PRS.3SG
‘She sings and dances.’ Trentino
(28) La canta e *(la) bala.
SCL.3SG.F sing-PRS.3SG and *(SCL.3SG.F) dance-PRS.3SG
‘She sing and dances.’
However, modern spoken French treats its SCLs quite differently than formal, ‘bonne usage’ French. Modern Standard French (MSF) allows the expression of SCLs alongside lexical subject DPs (which is formally not allowed; recall section 1.2.1) in structures that we arguably should not call instances of left dislocation (Renzi 1992, Zribi-Hertz 1994). Assuming that this entails a lower structural position for its SCLs, we would predict that they behave like their Northern Italian counterparts in coordination. This is exactly what we find, as the second SCL may be optionally repeated for some speakers:
MSF (Zribi-Hertz 1994:462)
(29) Il mangera beaucoup de viande et (il) boira du bon vin.
He eat-FUT.3SG much DET.INDF meat and (he) drink-FUT.3SG DET.INDF good wine
‘He’ll eat lots of meat and drink good wine.’
As with the discussion about the left dislocated status of subject DPs in French, the status of these coordinated structures is subject to debate and a fine-grained analysis of French SCLs is beyond the scope of this thesis. However, I believe it is relevant to speak about these French data: they seem to constitute empirical support for Poletto’s (1993a) proposed diachronic development of SCLs, explained further under section 5.
Nonetheless, there is another way to look at the single/multiple expression of SCLs in coordination. It is largely agreed upon that Northern Italian SCLs are heads; French SCLs on the other hand are weak pronouns which are assumed to be maximal projections (Kayne 1975, Cardinaletti & Starke 1996). As Rizzi (1986b) explains, this bears directly on the possibility of their omission in coordination: major categories (= maximal projections) such as verbs, adjectives and nouns allow for omission in coordinated small clauses, whereas minor categories such as prepositions and others do not:
Italian (Rizzi 1986b:406-407)
(30a) Ho visto [Mario leggere un libro]
e [Piero [_] un articolo]
‘I saw Mario read a book and Piero an article’
(30b) Ritenevo [Mario arrabbiato con noi]
e [Piero [_] con voi]
‘I believed Mario angry at us and Piero at you’
(30c) Ritenevo [Mario sostenitore dei repubblicani]
e [Piero] [_] dei democratici]
‘I believed Mario (a) supporter of the republicans and Piero of the democrats’
(30d) *Ritenevo [Mario in buona forma]
e [Piero [_] forma eccellente]
‘I believed Mario in good shape and Piero excellent shape’
(30e) *Ho visto [un passero su un albero] e
[un gatto [_] un tetto]
‘I saw a sparrow on a tree and a cat a roof’
Rizzi (1986b:407) argues that “a zero pro-form in the second conjunct, in order to survive alone, must be a major category”.
Cardinaletti & Starke (1996) further discuss the categorical status of pronouns and clitics, making a tripartite distinction between strong pronouns, weak pronouns and clitics. I
refer the reader to their article for further detail, but they argue in line with Rizzi (1986b) that coordination asymmetries between subject and object clitics in French are the result of differences in categorical status:
Cardinaletti & Starke (1996:30-31)
(31a) Il travaille à son article et – pense à ses problèmes ‘He works at his paper and – thinks of his problems.’ (31b) *Jean le lavera soigneusement et – remettra en place
‘Jean it will-wash carefully and – will-put back into place.’
The object clitic in (31b) is not a constituent of itself. Rather, it forms a complex head with the verb; by virtue of this, it cannot be omitted independently of the verb. The subject pronoun in (31a) however can be omitted because it is an autonomous constituent and a maximal projection. Cardinaletti & Starke (1996) give the following schematic representations:
(32) (33)
Such an analysis extends to Northern Italian SCLs, as they explicitly discuss: as these do not occupy a specifier position as full DPs, but rather are clitics, i.e. agreement ‘words’ attached to the verb, they cannot be omitted in the second conjunct of coordination.
1.3 Proposed locations of SCLs
The discussion above served to establish that NIDs should be considered NSLs. This would make (34) a possible candidate for the structure of Northern Italian SCLs, as opposed to (35) for, for example, Standard French SCLs.
In a language with structure (34), content from AgrP would license and identify pro, leaving [Spec TP] phonetically null but syntactically present. In a language with structure (35), content from AgrP would not suffice to saturate [Spec TP], which by consequence cannot be null and has to be filled. Recall that these structures resemble those proposed by Cardinaletti & Starke (1996) (section 1.2.3) and that they imply that Northern Italian SCLs are heads rather than maximal projections. However, structures like (34) cannot account for all NIDs: such a structure prohibits the omission of SCLs in coordinated structures, which in fact does happen, especially in ‘type 3’ coordination (i.e. where the same verb is repeated with different aspectual specifications (Poletto (1993a, 2000); see example (26) above). In this section I will explain the proposals that try to account for the distributional variation of different SCLs across NIDs.
1.3.1 Sportiche’s ‘uniform’ analysis
Sportiche (1999) argues for a quite different perspective on Northern Italian, and French, subject clitics than others. In fact, he proposes that French and Northern Italian SCLs are formally alike. By consequence, French and NIDs are formally alike when it comes to the null subject parameter: French, contra common analysis, is argued to be a NSL.
An important argument for this type of analysis comes from the idea that the complementary distribution of subject DPs and SCLs in French is actually a language-specific syntactic constraint rather than competition for the same structural position. Sportiche points out that the complementarity is between a phrasal category and single word, rather than between equivalent elements. He likens the complementarity to that between a complementizer that heads CP and a wh-phrase in the specifier of CP; these do not compete for the same position, but they are nonetheless barred from occurring together in some languages. This is called the ‘double filled CP filter’, which has a long history (going back to Chomsky & Lasnik (1977)); Sportiche proposes that a similar filter may also be active for the lexical DP subject and the SCL in French, but not in Trentino or certain dialects of French. This means that the languages are structurally similar, but that they differ in whether they have such a filter.
Doubly filled CP filter:
(36a) I saw that you beat him
(36b) I saw who (*that) you beat
In the example above we see that for (standard) English, the complementizer must be deleted when the dependent phrase is introduced by a wh-phrase. The same holds for Dutch; standard Dutch does not allow the joint expression of a wh-phrase and a complementizer, but substandard varieties exist that do allow this:
Standard Dutch:
I see that you him have beaten
(37b) Ik zie wie (*dat) je hebt geslagen
I see who (that) you have beaten Substandard Dutch:
(38a) Ik zie dat je hem hebt geslagen
I see that you him have beaten
(38b) Ik zie wie dat je hebt geslagen
I see who that you have beaten
Other evidence against French subject DPs and SCLs occupying the same position comes from Q-float structures: quantifiers that refer to the subject may, for some speakers, be moved from the embedded clause to the main clause without affecting the position of the (embedded) SCL:
(39) Il a tous fallu qu’ils partent
it is all necessary that they leave ‘It’s necessary that they all leave.’
Sportiche (1999) argues that this means that the SCL is not part of the subject DP10;
otherwise, it would be moved along with it. The idea that SCLs and subject DPs in French do not occupy the same structural position is not new in Sportiche (1999): Kayne (1975) and Jaeggli (1982) already proposed this; Sportiche (1992) analyzes object clitics following the same logic, i.e. the clitic and the full DP object do not occupy the same position. Sportiche (1999) straightforwardly extends this idea to subject clitics.
Structurally, this means that each clitic heads its own projection, a ClP, and that if there are more clitics, there are more projections. For subject clitics, this ClP is NomP (for Nominative P). In NIDs, the specifier of this NomP may contain the subject DP. Sportiche argues that this is ruled out in French because of a “doubly filled ClP prohibition” (Sportiche 1993:9), which takes its name from the doubly filled CP filter explained above. The doubly filled ClP filter functions in the same way as its CP counterpart in that it does not allow to be filled by two elements.
This has some important implications. The most advantageous one is that this type of analysis argues for the same underlying structure in French, Italian and NIDs: they are all null subject languages, and they all have a complete set of SCLs, or NomP projections. The differences lie in how the languages license and identify their null subjects; Italian does so through strong verbal agreement, thus not requiring a phonetically filled NomP; NIDs are like Italian, and any SCLs present are in fact redundant (cf. Rizzi 1986) or “superfluous” (Sportiche 1993:15) – any gaps in an SCL paradigm bear no major consequence. French
10 Sportiche (1999:10) proposes a structure [tous pro] for this DP, “which is a possible DP as shown by Tous sont venus”.
verbal agreement is not strong enough to license and identify pro, so it requires the SCL (or the presence of a full subject DP).
Sportiche (1999) does not explain the systematic distributional variation between NIDs and between their SCLs, and any patterns in optionality are left undiscussed. This might be a logical consequence of his characterization of these SCLs as “superfluous” material (Sportiche 1999:15). However, this evades the question of why there are systematic differences in which SCLs have to appear, which do not, and which may or may not appear.
1.3.2 Poletto’s Agreement Field
Poletto’s is probably the most extensive work on the syntactic status of Northern Italian SCLs. In her comprehensive works on the subject clitic variation of NIDs, she argues that only a complex functional field can account for the variation of SCLs across the dialects: the variation is not random and can be caught looking at only a handful of morphosyntactic features encoded by the SCLs (1993a, 1999, 2000).
Essentially, she argues for the existence of four distinct groups of subject clitics: invariable clitics, deictic clitics, number clitics and person clitics. Invariable clitics and deictic clitics are argued to be situated higher than number clitics and person clitics: invariable clitics and deictics clitics are necessarily situated above Neg, while number and person clitics are merged below Neg. Of the lower clitics, only the number clitics may move to a position above Neg: “for every semantic morphological SCL type, there is a basic syntactic position where it is realized, and when we find a SCL higher than its basic position, it must have moved there.” (Poletto 2000:35)
Invariable SCLs, as the name suggests, do not vary in form: they have the same form, typically a, regardless of the person/number/gender specification of the subject. They seem to encode a topic-comment feature, i.e. signaling the whole sentence as thetic, and they necessarily cluster with a complementizer (if present). They occur before negation, and they can always be omitted in the second conjunct of coordinated structures. They may occur in zero-place predicates, and they are incompatible with left dislocated, focalized and wh-elements. This leads Poletto (2000:24) to the following hypothesis:
“…invariable SCLs are able to move through the head of the CP projections where wh-items occur […] to the head of FocusP, and then to the head of the left dislocation position, thus preventing wh-, focalized and left-dislocated items from occurring in the sentence. […] This amounts to saying that the invariable SCL is a sort of expletive for the LDP, FocusP and interrogative CP, as it realizes the default feature of each of these projections and not the marked value that is realized when a wh-, focalized, or left-dislocated item enters its SpecC position.”
Polesano (Poletto 1999:587,589,591)
(40a) A no magno
SCL NEG eat-1SG.PRS
‘I don’t eat.’
(40b) A no piove
SCL NEG rain-3SG.PRS
‘It’s not raining.’
(40c) A magno patate e bevo caffè
SCL eat-1SG.PRS potatoes and drink-1SG.PRS coffee ‘I eat potatoes and drink coffee.’
(40d) Quando ch’a vegno
when that SCL come-1SG.PRS
‘When I come’
(40e) *Co a vegno
when SCL come-1SG.PRS
‘When I come’
(40f) Co vegno
when come-1SG.PRS
‘When I come’
Deictic clitics take their name from expressing a [+/-deictic] feature of the subject: they vary in form between 1st and 2nd singular and plural as opposed to 3rd singular and plural. Like
invariable clitics, they necessarily cluster with a complementizer (41d). However, they encode no theme/rheme distinction, so they can also appear with ‘old information’. They are compatible with some wh-elements, but incompatible with others:
S. Michele al Tagliamento, Friulian (Poletto 2000:25,72-73)
(41a) Dulà a vanu?
where SCL.3PL go-3PL.PRS
(41b) Do (*a) vanu?
where SCL.3PL go-3PL.PRS
‘Where are they going?
(41c) Quant *(i) mangi - tu?
when i11 eat-2SG.PRS - SCL.2SG
‘When are you eating?’
(41d) Quant ch’al rivi?
when COMP SCL.3.SG come-3SG-SUBJ
‘When could he come?’
Deictic SCLs are allowed with left dislocated items: Friulian (Poletto 2000:26)
(42) A ciasa o soi già laat
at home SCL.1SG AUX already go-PTCP
‘I have already been at home.’
Deictic SCLs can be omitted in coordinated structures. The extent to which this is possible depends on the dialect, but omissions are found with every type of coordination considered by Poletto: those where two verbs and their complements are coordinated (‘type 1 coordination’); those where two different verbs and a shared object are coordinated (‘type 2 coordination’); and those where the same verb is repeated with different aspectual specifications (‘type 3 coordination’):
Turinese (Poletto 1999:591)
(43) I cantu cun ti e balu chun chiel
SCL.1SG sing-1SG.PRS with you and dance-1SG.PRS with him
‘I sing with you and dance with him.’ S. Michele al Tagliamento, Friulian (Poletto 1999:591)
(44) I cianti cun te e *(i) bali cun lui
SCL.1SG sing-1SG.PRS with you and SCL.1SG dance-1SG.PRS with him
‘I sing with you and dance with him.’
Finally, deictic clitics cannot appear when an invariable clitic is present. However, they do for example show different behavior in coordination and with respect to left dislocation, so it is not obvious that they occupy the same position. Poletto (2000) argues that the reason for their incompatibility is that deictic SCLs interfere with the earlier mentioned movement of invariable SCLs from a wh-position. Poletto (2000) argues that deictic SCLs are situated
higher than simple wh-items (but lower than wh-phrases12), and as such are generated higher
in the structure than invariable SCLs.
Person clitics take their name from expressing strictly a person feature. However, this cannot be just any person feature: only the 2nd and 3rd person singular are encoded by this type
of clitic. This leads Poletto (1993a, 2000) to the conclusion that these SCLs carry a [+/-hearer] feature. They are usually not omitted in any type of coordination, and do not move past negation.
Trentino (Poletto 1993a:5)
(45) La magna pan e *(la) beve vin
SCL.3SG.F eat-3SG.PRS and SCL.3SG.F drink-3SG.PRS wine
‘She eats bread and drink wine.’
Number clitics are responsible for distinguishing the different [-hearer] persons, adding a [+/-plural] and a [+/-gender] feature. They sometimes move across negation from their base-generated position below it, and they can sometimes be omitted in type 3 coordinations: Venetian (Poletto 1999:588)
(46) No la vien
NEG SCL.3SG.F come-3SG.PRS
‘She isn’t coming.’ Fiorentino (Poletto 1999:588)
(47) La un viene
SCL.3SG.F NEG come-3SG.PRS
‘She isn’t coming.’ Oneglia, Ligurian (Poletto 1999:588)
(48) U/a nu catta
SCL.3SG.M/SCL.3SG.F NEG buy-3SG.PRS
‘She isn’t buying.’ Venetian (Poletto 1993a:6)
(49) La canta e ricanta sempre la stesa musica
SCL.3SG.F sing-3SG.PRS and resing-3SG.PRS always the same music
‘She sings the same song time and again.’
Arguably, this means that number clitics are merged higher in the structure than person clitics, which cannot be omitted in any type of coordination.
It is useful here to relate to Renzi & Vanelli (1983), who find that there is an implicational relation between the persons that any NID may have SCLs for:
12 I refer the reader to Poletto (2000) for the detailed analysis she proposes for different types of wh-items, their structural positions and their bearing on different types of SCLs.
a) if a dialect has only one subject clitic, it is for 2nd singular
b) if a dialect has two subject clitics, they are for 2nd and 3rd singular
c) if a dialect has three subject clitics, they are for 2nd and 3rd singular and 3rd plural Table 1
1 2 3 4 5 6
a - + - - -
-b - + + - -
-c - + + - - +
This fits rather neatly with Poletto’s proposed morphological features of SCLs:
Table 2 1 2 3 4 5 6 invariabl + + + + + + deictic + + + + + + number - - + - - + person - + + - -
-Renzi & Vanelli’s (1983) statement c) corresponds to a dialect having person and number clitics, and statement b) allows only for person clitics. However, Poletto (1993a, 2000) does not treat 2nd and 3rd person singular as different types of SCLs, so how can we account for
statement a)? Quite simply, she argues that we can hypothesize that there are dialects that only encode the marked value of [+/-hearer], i.e. only [+hearer] and not [-hearer]. The syntactic structure that Poletto (2000) proposes is the following:
(50)
She supports the hypothesis that structures are built bottom-up, i.e. that “it is possible to build up higher projections only if the lower ones have been projected. When the lower projections
are not there, their higher ones are obviously not accessible” (Poletto 2000:40). Further, she states that optionality varies between the different types of clitics: invariable and deictic clitics are optional, while number and person clitics are not. The underlying reason would be that the latter are needed for pro licensing and case assignment, whereas this is not the case for the former (Poletto 1999).
Poletto’s theory makes certain predictions for Genoese, which are problematic. As will be explained later on in further detail, Genoese possesses SCLs for the 2nd and 3rd person
singular, the former obligatory and postnegative and the latter optional and prenegative. Poletto’s model predicts that the Genoese 3rd person SCL is a deictic clitic, as it is optional,
and that the Genoese 2nd person SCL would be a person clitic, as it expresses [+hearer]. We
now see that this would mean that the functional projection for the deictic SCLs is projected without first having a projection for the number clitic, which is awkward in her account.
Further, the Genoese 3rd person SCL does not necessarily cluster with a complementizer,
which she argues is the case for deictic clitics.
1.3.3 Goria’s Optimal Agreement
Goria (2003, 2004) is fundamentally unsatisfied by the main structural proposals about SCLs that rely on the presence of Agr, and especially Poletto’s Agreement Field. Goria proposes a complementary working of minimalist generative syntax and Optimality Theory to account for the distributional variation of SCLs across the dialects. Following Chomsky’s (1995) Agr-less T-model, Goria (2004) wishes to reduce structural complexity. She argues that the plurality of functional layers of Poletto’s Agreement Field is not only incompatible with an MP approach, but that it is also self-contradictory and that it is empirically problematic. In this section I will cite her theoretical objections to Poletto’s Agreement Field; for her intricate, data-based argumentation against it, I will refer the reader to Goria (2004). Finally, I will explain Goria’s own structural proposal.
In her work on Piedmontese SCLs13, Goria (2004) points out that Poletto’s theory is
incongruent with the current theoretical developments and assumptions, which can account for SCLs as well. Within a modern MP approach, the functional category of Agr is dispensed with; rather, agreement is a long distance operation that does not require movement to satisfy any features (cf. Adger 2003). As such, SCL movement is unwarranted, which solves the problem that in Goria’s (2004) view, Poletto (2000) motivation for what exactly makes these SCLs move to the right position is lacking:
“The Agreement Field is based on the Correspondence Hypothesis, which says that ‘for each semantic morphological SCL type, there is a syntactic position where it is realised […]’ (Poletto 2000b:35). In other words, agreement features head their own independent functional projections and in the NIDs these projections are overtly realised by SCLs. As a first observation, under Minimalist assumptions there is no such correspondence between agreement features and syntactic positions as the former are either intrinsic to the lexical items or optionally added before Numeration.” (Goria 2004:72)
Further, she argues that the morphosyntactic features proposed by Poletto as relevant for the SCLs actually show overlap, and as such cannot be taken to be predictors of distinct structural
positions. In other words, Poletto’s proposal is inconsistent with itself. For example, it is argued that 2nd person singular clitics can only be ‘person’ clitics, i.e. encoding [+hearer].
However, this feature overgenerates: it is the 2nd person singular SCL that has a special status
in the SCLs, not the 2nd person plural. This means the 2nd person singular has to encode a
number feature as well: [-plural]. In Poletto’s analysis however, this would make it a ‘number’ clitic, moving to the same position as other number clitics without distinction. To go even further, Goria (2004) argues that the [+/hearer] feature is also necessarily [+deictic], which would have to activate the [deictic] functional layer in Poletto’s analysis. For more detail on this argumentation, the reader is referred to Goria (2004), chapter 3; here it suffices to illustrate her logic.
Her own structural proposal follows Chomsky’s (1995) T-model. The model contains the smallest amount of substantive and functional categories that can be motivated by phonetic and semantic properties. Functional categories C, D, and T respectively express mood/force, referentiality and tense/event structures; another functional category, Light-v, is motivated by θ-role assignment. Functional category Agr is eliminated, because agreement is motivated by an operation, rather than a position, Agree.
Goria (2004:68) proposes the following syntax for SCLs: “ i) SCLs are heads [D] in TP;
ii) SCLs do not check the EPP. They encode T’s EPP feature by expressing T’s φ features overtly.”
However, the most relevant part of her proposal is the way in which she accounts for the optionality that we find. She points out that it is hard to account for optionality within Minimalism, since variation is explained by parametric variation, which actually serves to distinguish one language from another. This would mean that an MP analysis of the optionality of SCLs in some NIDs would posit different underlying structures for any sentence with an SCL and its counterpart which lacks the SCL. Assuming that the optionality we are concerned with is indeed free variation, this is an undesirable result.
Goria solves this problem by recurring to Optimality Theory (Prince & Smolensky 1993), in which UG is argued to contain a set of universal constraints; grammars then are rankings of constraints, and differences in grammars stem from different rankings of constraints. In Optimality Theory (OT), different ‘candidates’ for output compete with each other, with the candidate that incurs the least violations of the constraint rankings being the ‘optimal’ candidate. We might say that an OT account of UG is less strict than a generative account, allowing for imperfect candidates to still be grammatical, rather than requiring a clear structural yes/no for any input.
From an OT perspective then, Goria (2004) argues that NID speakers may have multiple SCL paradigms at their disposal that they rank and re-rank according to the situation at hand. The systems themselves are based on a ranking of morphosyntactic features that different SCLs encode. Crucially, she argues that the morphological realization of agreement takes place after narrow syntax, such that sentences possessing SCLs are syntactically no different than the same sentences which lack them. The absence or presence of the SCL thus constitutes the overt or covert realization of features.
As we have seen, Northern Italian SCLs minimally distinguish the 2nd person singular;
if it distinguishes two, they are the 2nd person and 3rd person singular; if there are three, they
are the 2nd and 3rd person singular and 3rd person plural; if there are more, then Renzi &
Vanelli (1983) and Poletto (2000) disagree on which these are, but in any case it means that the 2nd person singular, 3rd person singular and 3rd person plural are likely to be obligatory
(Renzi & Vanelli 1983). Goria (2004) captures this in the following way: there are feature bundles, of which some dialects possess more than others, which are ranked differently with respect to each other in the different dialects.
The features taken by Goria to play a role are: [+/-(addressee,singular)], [+/-speaker], [+/-participant], [+/-singular] and [+/-masculine]. The features and their different rankings give rise to the different paradigms we find across the NIDs. The feature [+/-(add,sg)] is the most prominent among the features, as the 2nd singular is always the first person to take its
own SCL in any variety. Only after this feature a variety may add any of the above features. Goria (2004) mostly concerns two Piedmontese dialects: Turinese and Astigiano. Below I will give the SCL paradigms of both and the feature rankings, or systems, which Goria attaches to them. Note that both these dialects have multiple systems from which speakers can choose, thus allowing for optionality:
Table 3 (Astigiano) SCLs features 1sg a [-(add,sg)] 2sg at [+(add,sg)] 3sg a [-(add,sg)] 1pl a [-(add,sg)] 2pl a [-(add,sg)] 3pl a [-(add,sg)]
Astigiano uses what Goria calls the ‘Basic System’, which has the following feature ranking: [+(add,sg)] >> [-(add,sg)] >> NO [φ]. In other words, it is a system with SCLs that express the distinction between 2nd person singular vs. all other persons. NO [φ] is ranked lowest,
because in these dialects all persons have SCLs.
Table 4 (Turinese) SCLs features 1sg i [-(add,sg)] [+part] 2sg it [+(add,sg)] [+part] 3sg a [-(add,sg)] [-part] 1pl i [-(add,sg)] [+part] 2pl i [-(add,sg)] [+part] 3pl a [-(add,sg)] [-part]
The Turinese paradigm reflects the ‘Deictic System’, of which the feature ranking is [+ (add,sg)] >> [+part] >> [-(add,sg)] >> [-part] >> NO [φ].
Speakers of these dialects may omit their SCLs however, resulting in the following paradigms:
Table 5 (Astigiano) All persons: Basic System Person Optionality: Deictic System 2 Full Optionality: Ø SCL System 1sg a Ø Ø 2sg at at; ‘t Ø 3sg masc
a; al; el; ‘l; ir; o
a; al; el; ‘l; ir; o Ø
3sg fem a a Ø 1pl a Ø Ø 2pl a Ø Ø 3pl a a Ø Table 6 (Turinese) All persons: Deictic System 1 Person Optionality: Deictic System 2 Full Optionality: Ø SCL System 1sg i Ø Ø 2sg it; ‘t it; ‘t Ø 3sg masc a a Ø 1pl i Ø Ø 2pl i Ø Ø 3pl a a Ø
As can be noted, terminology has been added: Deictic System 1 and Deictic System 2, Person Optionality, Full Optionality. Deictic System 1 is the Deictic System described just above: [+ (add,sg)] >> [+part] >> [-(add,sg)] >> [-part] >> NO [φ]. Deictic System 2 differs from it in omitting the 1st singular and plural and 2nd plural. The corresponding feature ranking is [+
(add,sg)] >> [-part] >> NO [φ]. Person Optionality can briefly be described as a system that entails optionality for some, but not other persons; Full Optionality means any SCL is optional.
To conclude here, Goria (2004) argues that NID speakers have multiple SCL systems at their disposal and may choose which one to adopt. Contact with Standard Italian is arguably a driving force behind the adoption of the Ø SCL system, or in any case a system with fewer SCLs than the dialects ‘proper’. It is justified to ask ourselves here if it is preferable to let optionality be the result of speakers with multiple spell-out mechanisms at their disposal, which in turn may have their origin in different grammars, rather than be the result of different structures entirely, as would be predicted from a strict MP account. The former view relies heavily on the assumption, intuitive as it may be, that the variation really is free and that there is no evidence for different structures between possessing and SCL-lacking sentences. It may be the case that this assumption is not justified; one argument in favor of this may be that there actually are clear patterns in the distribution of these optional SCLs, as we will see from Genoese.
2. Hypothesis & predictions
NIDs all possess SCLs. However, they all have different SCL paradigms and their SCLs behave differently from dialect to dialect and from clitic to clitic. The question to be asked is whether we can capture their syntactic nature within a unifying analysis, or rather if they are to be explained as different syntactic elements, as many previous authors have claimed:
“The notion of subject clitics is spurious from the syntactic point of view, in that it does not identify a class of elements with uniform syntactic properties, as pointed out in Kayne (1975), Rizzi (1986), and Poletto (1991, 1993a), among others.” (Zanuttini 1997:28)
I believe it is reasonable to explain the distributional variation of SCLs in NIDs on the basis of how the SCLs interact with the pro-drop parameter and on which morphosyntactic purposes they serve in any dialect. Adopting the view that NIDs are pro-drop languages, I hypothesize that any distributional variation of the Northern Italian SCLs is microvariation: in essence, Northern Italian SCLs share the same structural properties, and variation in the distribution of their SCLs is to be explained by the differences in the precise character of the pro-drop parameter in the different NIDs, e.g. how strong agreement is in any given NID.
Besides the matter concerning whether a NID has certain clitics or not, omission or inclusion of SCLs that are ‘optional’ in a dialect may happen along systematic lines, rather than constituting real, free variation. If this is the case, then it is necessary to tease apart the features and contexts of influence on the position and presence of these SCLs. Here I expect that semantic and pragmatic factors may be of influence on the way in which pro is licensed and agreement is instantiated.
The analysis found in this thesis for Genoese SCLs may serve as a first step to understanding other NIDs. My analytic tool is that features or contexts that influence the expression of SCLs in Genoese can be examined in any other variety to see which influence they have on SCL expression there. Where this fails, the same logic used in the analysis of Genoese is to be employed there: if a SCL is obligatory, which features does it encode and which purpose does it serve? If it is (or appears) optional, is it possible to find features or contexts that influence its absence/presence? If it is absent, what characterizes the contexts in which it is?
3. Method
This thesis adds Genoese data to the discussion on Northern Italian subject clitics. Precise data on Genoese was lacking in the relevant literature: data of other Ligurian varieties did not reflect Genoese (i.e. Ligurian from Genoa proper), and data and/or analyses that did claim to reflect Genoese were incomplete, imprecise, or at the very least outdated. As such, I decided to collect data from native speakers of both Genoese and Standard Italian.
Seven native speakers of Genoese were included in the study: 4 male, 3 female, ages 51-85. Speakers are from urban Genoa and are all (near-)native speakers of Standard Italian. Genoese is employed at home and with friends, family and acquaintances; it is also employed when non-Genoese speakers are present, as long as the addressee is a speaker. Genoese
generally is not employed for formal use; however, it might be used if for example a shop keeper/restaurant owner addresses clients in Genoese.
Data was gathered in the following way: participants were asked to translate 152 Standard Italian sentences into Genoese. The sentences were made to reflect different morphosyntactic contexts so as to get a broad impression of SCL distribution in Genoese; the factors taken into account can be found in the appendix. After examination of the translations, phenomena that resulted deserving further scrutiny were investigated by further contact with the participants, through further translations and acceptability judgments.
The orthography used for Genoese in this work follows the grafîa ofiçiâ (‘official spelling’) proposed by the Académia Ligùstica do Brénno. Founded in 1970, one of the aims of this institute is the codification of the Genoese script. The reason for employing this orthography here is ease of use as a result of my earlier work on Genoese and a sense of respect for the dialect. The phonemes and graphemes used in the grafîa ofiçiâ are shown in the appendix.
4. Distribution of Genoese subject clitics
4.1 Paradigm
Genoese has subject clitics for two persons: the 2nd person singular and the 3rd person singular.
Below is a table of tonic and clitic nominative pronouns for Genoese:
Table 8
tonic clitic
1st singular mi
-2nd singular ti ti
3rd singular lé (masc. & fem.) o (masc.), a (fem.); l’ (masc., fem.)
1st plural noiätri
-2nd plural voiätri
-3rd plural liätri
-The second person singular SCL ti is obligatory, and the 3rd person singular SCLs o
(masculine) and a (feminine) are optional; l’ is a somewhat peculiar item in the 3rd person
singular paradigm which is discussed in more detail further on. The 2nd person SCL appears
after negation, whereas the 3rd person SCL appears before negation.
(51a) No ti a veu
NEG SCL.2SGOCL.F want-2SG.PRS
‘You don’t want it/her.
(51b) O no a veu
SCL.3SG.M NEG OCL.F want-3SG.PRS
Genoese exhibits no verb-clitic inversion in interrogatives, as several other NIDs do. Subject clitics do not appear in zero-place predicates (cf. Renzi & Vanelli (1983), who report the possibility of an SCL with weather verbs in Ligurian dialects):
(52)(*O) cieuve
SCL.3SG.M rain-3SG.PRS
‘It’s raining.’
Subject clitics are compatible with DP subjects (including QPs and pronouns), but only (in general) when these are preverbal:
(53)O diretô (o)l’è arivòu
the manager SCL.3SG.M AUX arrive-PTCP
‘The manager has arrived.’
(54)Quarchedun o m’ha visto
someone SCL.3SG.M me AUX see-PTCP
‘Someone saw me.’
(55)(*O) no m’ha visto nisciùn
(*SCL.3SG.M) NEG me AUX see-PTCP nobody ‘No one saw me.’
In coordination, both the 2nd person and 3rd person clitic do not allow for omission in the
second conjunct:
(56)Ti lezi e *(ti) rilezi sempre o stesso libbro
SCL.2SG read-PRS.2SG and SCL.2SG reread-PRS2SG always DET same book
‘You keep reading the same book over again.’
(57)O preuva e *(o) ripreuva ma o no ghe riesce
SCL.3SG.M try-PRS.3SG and SCL.3SG.M retry-PRS.3SG but SCL.3SG.M NEG at-it
succeed-PRS.3SG
‘He keeps and keeps trying but he doesn’t succeed.
Genoese SCLs may appear in clitic clusters, with negation always preceding the 2nd person
singular, and always following the 3rd person singular. For monotransitivies, the order is
SCL-OCL-verb:
(58)(No) ti me veddi
NEG SCL.2SG OCL.1SG see-2SG.PRS
(59)O (no) me vedde
SCL.3SG.M NEG OCL.1SG see-3SG.PRS
‘He (doesn’t) see me.’
The situation is more complicated for ditransitives. NIDs generally do not exhibit clitic clusters of three or more; Genoese does allow for clusters of three or more, but the clitics present in the cluster typically do not align with the arguments of the verb. For example: there may be three clitics present with a ditransitive verb, but these do not express the subject, the direct object and the indirect object. Rather, one of the clitics takes on a benefactive meaning:
(60)O me l’avisa
SCL.3SG.M for-me DCL.3SG advise-3SG.PRS
‘He advises him for me.’
In this example, o expresses the subject and l’ expresses the object. Me however does not express the direct object, but rather the person whose wish it is that someone advise someone else. Thus, the direct object clitic and the indirect object clitic interfere with each other in ditransitives14. The subject clitic, however, is unaffected by this.
Further, Genoese possesses a locative clitic ghe:
(61)O me gh’ha visto
SCL.3SG.M OCL.1SG there AUX see-PTCP
‘He saw me there.’
The locative clitic also does not interfere with the subject clitic.
4.1.1 The case of l’
For the third person singular, there’s variation between only expression of the SCL o or a, expression of the SCL+l (ol or al) and expression of only l’. As Rohlfs (1968) already finds, the constraints are phonological: if the SCL precedes a consonant, only the vowel appears; if the SCL precedes a vowel, it appears as ol/al :
(62a) O te deve dâ quelo libro
SCL.3SG.M DCL.2SG must-3SG.PRS give that book
‘He has to give you that book.’
(62b) O l’ariva doman
SCL.3SG.M l’ come-3SG.PRS tomorrow
‘He’s coming tomorrow.’
14 This is not an uncommon phenomenon and is known as the Person-Case Constraint (‘PCC’, Bonet (2008)). The PCC entails the following: if there are two clitics in a ditransitive structure (other than the subject clitic and/or locative clitics) and one of them is for the third person, then this can only be the direct object and not the indirect object.
Support for the claim that the constraints on the appearance of l’ are phonological is the following: if another clitic appears between the SCL and the vowel-initial verb, such as OCL
m’ (me), l’ is absent:
(63a) Credo ch'o l'aggia dovùo riçeivi-me pe forsa.
believe-1SG.PRS COMP SCL.3SG.M l’ AUX must-PTCP receive-OCL.2SG for force
‘I think he was forced to receive me.’
(63b) Credo ch'o m'agge riçevùo pe forsa
believe-1SG.PRS COMP SCL.3SG.MOCL.2SGAUX receive-PTCP for force
‘I think he was forced to receive me.’
If we do indeed analyze these various instantiations of the SCL as a singular clitic that shows phonological variation, it is not unlike Standard Italian definite articles: underlyingly they are
il15 (masculine) and la (feminine), but they drop their vowels when they follow prepositions or
when they precede vowels:
(64a) il teatro the theater (64b) de-l teatro of-the theater (64c) l’amico the friend (64d) (del)la squadra
(of) the team
(64e) l’anima
the soul
It is however not entirely obvious that the vowel and the consonant are parts of one single clitic that shows phonological variation. Most importantly, negation appears between the o/a and l’; if a predicate containing a vowel-initial verb is negated, the ordering is o/a – negation – l’ – verb. This is strong evidence against a ‘single clitic’-analysis, as infixation of this type is untypical of modern Romance languages:
(65a) Credo ch'o no l'agge dovùo riçeivi-me pe forsa
believe-1SG.PRS COMP SCL.3SG.M NEG l’ AUX must-PTCP receive-OCL.2SG for
force
‘I think he wasn’t forced to receive me.’
(65b) Credo ch'o no m'agge dovùo riçeive pe forsa
believe-1SG.PRS COMP SCL.3SG.M NEG OCL.2SG AUX must-PTCP receive for
force
‘I think he wasn’t forced to receive me.’
Second, l’ may appear on its own, i.e. without o/a preceding it, at the start of a sentence before a vowel-initial verb:
(66) L'è vegnù o cädo
l’ AUX come-PTCPDET heat
‘The heat has come.’
(67) O l'ha corìo rapidiscimo
SCL.3SG.M l’ AUX run-PTCP very.fast
‘He’s run very fast.’
These examples also exhibit a difference in animacy, which seems to influence o/a but not l’, which occurs with any type of subject: o/a regularly appears with animate subjects, but not with inanimate subjects.
(68) O figgieu ch'o no l'è vegnùo véi, o vegne doman
DET boy COMP SCL.3SG.M NEG l’ AUX come-PTCP yesterday SCL.3SG.M come-3SG.PRS tomorrow
‘The boy that didn’t come yesterday will come tomorrow.’
(69) No caze a miägia
NEG fall-3SG.PRS DET wall
‘The wall isn’t coming down.’
Poletto (1993a) and Zanuttini (1997) report the existence in other NIDs of elements similar to Genoese l’: an example is l’ in Piedmontese, “a type of subject clitic that is attested in the majority of northern Italian dialects and appears only in the presence of auxiliary verbs” (Zanuttini 1997:28), which may co-occur with other subject clitics:
(70) La barca a l’a andà a fond.
DET boat SCL l’ AUX go-PTCP to bottom
‘The boat has sunk.’
Poletto’s (1993a) analysis is that these elements only appear with auxiliaries, that they are in complementary distribution with clitics on Agr, and that they are only compatible with vocalic clitics. She proposes an analysis which follows Roberts (1991) in which auxiliaries head a different projection than lexical verbs, in which these ‘auxiliary clitics’ serve to make a complex head.
Such an account is difficult to maintain for Genoese, however: l’ does appear with SCLs that should be analyzed as occurring under Agr (i.e. clitics that encode subject features),
