Partitive objects in negative contexts in Northern Italian Dialects

1 Introduction

The goal of this article is twofold: to provide an overview of the variation patterns in the syntax of partitive objects under negation (Negated Partitive Objects: NPOs) in the Northern Italian dialectal domain, and to analyze them in a diachronic perspective comparing them with two different syntactic items: true partitive structures (TPSs) and the so-called “partitive articles” (PAs). The phenomenon we intend to investigate is exemplified in (1) ^[1]:

The sentences in (1) were produced as translations of a sentence with a bare indefinite plural object. Notice that, while the Lombard variety of Frontale di Sondrio displays the simple preposition de ‘of’, the Ligurian variety of La Spezia has the conflated form of the preposition plus what looks like the definite plural article di ‘of the’ (in Ligurian the simple ‘of’ is de), So, the NIDs can be split into two types: the French-like type, where NPOs display the bare preposition ^[2] and those that have the preposition plus the definite article. Analogous forms have been analyzed for French by Ihsane (2013) as containing an agreement marker, and not a true definite article. ^[3] We will discuss this type of variation, since it is relevant when comparing NPOs with both TPSs and PAs. The article has the following structure: in 1.1, we distinguish the three types of syntactic items we take into consideration. In Section 2, we present the microvariation data of NPOs in Northern Italian dialects. In Section 3, we discuss the historical development of NPOs in the context of the grammaticalization process from TPSs to PAs. In Section 4 we propose a formal analysis explaining the relation of NPOs with both TPSs and PAs. Section 5 contains an account of the variation observed in the Italo-Romance domain, which is the main point of our investigation, although we will occasionally refer to standard French and Italian for comparison. Section 6 summarizes and concludes the analysis.

2 Partitive Objects under Negation in NIDs

In this section we describe, from a comparative and microvariationist point of view, the phenomenon of NPOs across the Northern Italo-Romance domain. The data come from the database of the ASIt project (Atlante Sintattico d’Italia ‘Syntactic Atlas of Italy’). The dialect sentences of the database were collected as translations of standard Italian sentences, grouped in general and specific questionnaires (see Benincà and Poletto 2007 for a description of the field methods used for the database). The positive side about having such a database is that the stimulus sentences are always the same for all the dialects; the problem is that not all variables we might want to investigate are present in the data pool. Our main reason for using the ASIt corpus is that it provides a large set of comparable positive and negative sentences with indefinite objects for a considerable amount of dialects. ^[6] In any case, another large collection of comparable data from the Italo-Romance domain, the corpus of Manzini and Savoia (2005: III, 285) confirms that varieties with NPOs do not display a similar construction in positive contexts. We have selected eight sentences from all the questionnaires available in the ASIt: they all contain a sentential negation and an indefinite count plural or mass singular noun in object position. Two of these sentences contain a plural, two contain a true mass noun as the object (carne ‘meat’ and pesce ‘fish’), but there also four sentences where the object is the singular collective frutta ‘fruit’, which behaves as a singular mass nouns regarding this phenomenon. In (8) we provide the whole set of stimulus sentences we have extracted from the database for all Northern Italian dialects available ^[7]:

The phenomenon has a “leopard-spot” distribution in the whole Gallo-Italic domain (with a strong presence in Ligurian) ^[8] and is completely absent in Venetan, in Friulian and in Tuscan, which display regular PAs whether negation is present or not. ^[9] Figure 1, taken from Garzonio and Poletto (2018), shows the distribution of NPOs in the dialectal translations of Non compri mai mele ‘You never buy apples’.

Figure 1:

NPOs in NIDs.

In the following subsections, we present the different types we have singled out based on the ASIt data.

3 The development of NPOs and PAs

Before turning to our analysis, in this section we discuss the diachronic development of NPOs and PAs in Italo-Romance, comparing it with the development of PAs in French, which has been recently examined by Carlier (2007), and Carlier and Lamiroy (2014). We cannot summarize the whole literature on French PAs and NPOs here and only underline that there are three main lines of thought. The first is the one going back to Foulet (1965: 65) who assumed that in Old French PAs emerged from cases of a null quantifier in the structure. This idea has been exploited by Rowlett (1998: 66) for NPOs. He considers them a type of pseudopartitives, for which the null quantifier is actually not null, but corresponds to the negative marker pas (or to point or mie in older stages), which has moved from within the object position to the NegP in the sentence structure.

According to this view, NPOs encode a type of ‘part-whole’ relation, where the ‘part’ is a negated minimizer and the ‘whole’ is an indefinite mass or plural entity. In this sense they are distinct from TPSs, which are characterized by a closed set for the ‘whole’, which is clearly d-linked either through a demonstrative or some pragmatic device.

The second line of thought, developed by Kupferman (1994, 1998) relates the emergence and presence of PAs with a specific type of transitive verbs, the so-called fragmentation verbs, like boire ‘to drink’, which encode the partition of the object. Following Kupferman’s idea, PAs can be considered as case markers selected by these verbs. ^[15] It is indeed true that in Old French the vast majority of objects introduced by de are found with this type of verbs, although this is by no means mandatory. On the other hand, when one tries to provide a precise definition of “fragmentation” and tests the properties that define this type of value of a verb, the task is by no means an easy one.

Finally, Carlier (2007) rejects both the original analysis with a null quantifier proposed by Foulet as well as the one of PAs as case markers proposed by Kupfermann on the basis of a number of arguments against the two ideas. According to her proposal, the Old French de has an ambiguous status between full preposition and determiner, in the sense that it does not relate to the verb or a noun, as prepositions do, but only to the internal nominal expression. Objects introduced by de are used in Old French when the predicate presupposes an indefinite partition of a contextually specific set or amount of something (as already pointed out by Foulet 1965: 69), i. e., they were TPSs. For this reason, they usually appear with fragmentation verbs. ^[16]

The fact that in examples like (24b) de selects the oblique form of the relative pronoun can be considered evidence that it still behaves as a preposition. At this stage there are cases of nouns introduced by de in existential predicates, like (25). It is debatable that in (25) de sains moines is a subject, as suggested by Carlier (2007: 14)(see also Luraghi 2012: 12), since the existential predicate is a form of the verb ‘have’. However, this example could be an instance of the bridge context for the next stage of the process, where the partition meaning disappears and only the indefinite quantity is expressed.

In the Middle French stage the ambiguity is resolved, with de entering into the paradigm of determiners, combining with the definite article encoding phi-features:

“From the point of view of its meaning, this insertion into the paradigm of articles results in loss of meaning (the notion of a contextually specified partition object fades away) but also in enrichment (the notion of indefiniteness emerges). The loss of the prepositional status and the concomitant desemantization lead to an extension of the conditions of use. The partitive is still dominantly used in object position, but not only with verbs of the group boire, manger, etc. Moreover, the partitive also appears in other syntactic positions, including those introduced by a preposition.” (Carlier 2007: 33).

In Italian the development of PAs is similar. In Old Italo-Romance varieties it is possible to find examples where ‘of’ plus the definite article is used to introduce direct objects. Luraghi (2012: 14) points out that in most cases these constructions have to be interpreted as TPSs, since the referenced entity is contextually definite, as in (26a). However, from the end of the thirteenth century, we also find cases like (26b) where the noun introduced by ‘of’ plus the definite article is the subject of an existential predicate with an indefinite interpretation (in this case under negation):

In Tuscan varieties, the attestation of PAs derived from ‘of’ plus a form that looks like the definite articles becomes more robust from the end of the fourteenth century (at the end of a significant change in the language during the previous century; cf. Poletto 2014).

Furthermore, before the completion of the grammaticalization process of PAs at the end of the fourteenth century, Old Italo-Romance varieties display also cases like the following ones:

In these examples the object is a complex item formed by a minimizer, which was already used as negation during this stage (punto ‘point’, miga ‘crumble’), and a PP headed by ‘of’. We interpret similar examples as the contexts from which NPOs emerged. What is relevant here is the categorial ambiguity of the minimizer between a quantity noun and a postverbal negative adverb. This suggests that the reanalysis of the partitive preposition as a determiner (i. e., the ambiguity of this P) is linked to the ambiguity of the quantificational item expressing the ‘part’ of the partitive relation. We will take this idea into consideration for our proposal in Section 4.

We have mentioned that in French PAs and NPOs are distinguished because the former present conflation with what looks like a definite article, while the latter correspond to the simple P de. However, as we have seen in Section 2, in NPO contexts in the Gallo-Italic domain we can find the conflated form typical of French PAs. A similar variation regarding the form of PAs is observable across the domain of Romance languages that have PAs. While in French and in Italian the PA is always the conflated form, there are also cases of the converse, where the PA is the simple ‘of’, as in central Occitan varieties (and Gascon, where furthermore PAs are optional). Compare the examples in (29a) and (29b) with the one with the conflated form in (29c):

Something similar happens in Gallo-Italic. For instance, the Sprach- und Sachatlas Italiens und der Südschweiz (Jaberg and Jud 1928–1940) maps 636–637 (andare a cercare delle mammole ‘go and collect some violas’) display translations with a simple P in the Piedmontese area. Compare (30a) and (30b), which is a close Gallo-Romance variety, with the other varieties with a conflated form. ^[17]

This is a very interesting parallelism regarding the distribution of PAs and NPOs, suggesting that the two constructions developed in a parallel way from TPSs and definitely have something in common. For this reason, in our analysis we will consider NPOs as an intermediate stage in the grammaticalization process of PAs, which has been maintained independently by some of the Romance varieties that have PAs.

4 A broad view of Partitives

As we have seen above, there are at least three lines of thought concerning the development of TPSs into PAs and NPOs. In what follows we would like to present all the intuitions from the three analyses put forth above and combine them to develop a more specific structural analysis. The three points we intend to take over into our analysis are a) Foulet’s intuition that some TPSs contain a null quantifier; b) Carlier’s intuition that the preposition in these constructions is not a true preposition because it does not have two relations (one to the selecting verb and one to the DP it takes as its complement) but only one, i. e., the one internal to the nominal expression; c) Kupfermann’s intuition that the “fragmentation” semantics must somehow be encoded in the syntax, i. e., TPSs and NPOs must share some structural properties that PAs lack. We implement this into the internal structure of these three types of nominal expressions with the idea that NPOs constitute an intermediate structure between TPSs and PAs, sharing some properties with both of them.

We will start by taking up Carlier’s observation that TPSs derive from a locative construction that has the meaning of ‘taking away something’, i. e., subtracting a portion from the whole. This property is analogous to inalienable possessive constructions, where we define a part and a whole. Since there is general consensus in the literature that possession expressions can be analyzed as small clauses (cf. Hulk and Tellier 2000; den Dikken 2006), we will adopt an idea first proposed by Sleeman and Kester (2002) and argue that the relation ‘part/whole’ in TPSs is expressed just like in inalienable possession through a small clause, of which the nominal expression representing the part is the specifier, the nominal expression representing the ‘whole’ is in the complement position and the preposition realizes the head (see also Alexiadou 2003 for the idea that the difference between alienable and inalienable possession lies in the respective positions of the possessor and the possessum). The semantic value of this head can correspond to the feature Belong as proposed by Hulk and Tellier (2000):

This has the advantage of formally encoding the semantics of TPSs through a specific structural configuration, i. e., the small clause. At the same time, it captures Carlier’s intuition that the preposition is not selected by anything, either a verb or a noun in itself, since it is the realization of the head of a small clause. Carlier proposes that de is a marker of partition, and this is clearly the case in our structure, since the P is the realization of the head of a partitive small clause. This idea is also in accordance with recent treatments of PAs, as the one proposed by Cardinaletti and Giusti (2016) for Italian del/dei, which see them as indefinite articles. Since PAs represent the last step of the grammaticalization process of TPSs, here the small clause does not exist any longer, since the silent ‘part’ is not interpreted (and hence not projected) and the head of the small clause becomes the highest portion of a regular DP, as we will see below.

In addition to the small clause idea, we have to assume that real partitives contain a null Amount (for mass nouns) or Number (for plurals) functional light noun n. This null n is licensed through the presence of an adjectival Q in its specifier (like assez ‘much’ in French) which provides the null n with one of the two values, either mass or countable plural. ^[18] Several authors have followed Foulet’s intuition (see Gross 1967; Milner 1978) in assuming the presence of the null quantifier. We would like to be more specific and say that the null Amount/Number sits embedded inside the nominal expression located in the Spec of the small clause, i. e., in the structural portion representing the part and can be modified by a lexical specifier, as in (32):

In TPSs the nominal expression representing the ‘whole’ can be definite or indefinite, as Foulet already pointed out for Old French (although, as Carlier notices, the cases of indefinite ‘whole’ are less frequent).

If we now turn to the final stage of the historical reanalysis, i. e., PAs, we propose that these are regular DPs, so there is no small clause any longer. This corresponds to the semantics of PAs, which do not have any partitive interpretation, but simply a standard indefinite one. We adopt here Cardinaletti and Giusti (2016) proposal that PAs are actually normal DPs. As for the PA itself, it is clear that this is a bimorphemic element where the preposition is a type of Case marker. As for the additional marker element, there are two possibilities: we can exploit the fact that it is independently necessary to postulate the existence of expletive definite articles of the type already identified by Longobardi (1994) in his work on articles with proper nouns. As such, expletive articles are not referential, as already pointed out by Longobardi. The alternative is to adopt the cartographic view proposed in Ihsane (2013) and assume that the definite article appearing with PAs is actually an agreement marker and not a true article. In any case, PAs are made of two morphemes, one located in the IP-equivalent of the internal structure of D, which expresses phi-features, and the higher portion, historically derived from the preposition, which is a case marker. In our structure, however, even the higher morpheme of the article must be located rather low in the left periphery of the DP, in a position which does not encode ‘familiarity’, since the interpretation is the one of an indefinite. The structure of a PA is thus the following:

According to this view PAs do not have the syntax of real partitive expressions, which require a relation between two DPs, the part and the whole, and they just express indefiniteness (see Storto 2003 on this).

Suppose now that an NPO corresponds to an intermediate stage between the original structure of TPSs and the one of PAs, since it has a partitive meaning like TPSs, but the ‘whole’ is forcedly indefinite like in PAs and not d-linked as in TPSs. Suppose that what NPOs and PAs have in common is that the structure has already been reanalyzed as a regular nominal expression and is not a small clause any longer. On the other hand, what NPOs have in common with TPSs, is the presence of a null quantified Amount/Number, licensed by its measure specifier, which in the case of other pseudopartitives is the lexical quantifier beaucoup ‘a lot of’, combien ‘how many’, and in NPOs is the negative element pas. The presence of the preposition corresponding to ´of´ is justified by the AMOUNT/NUMBER, which, being of nominal character assigns genitive to its complement. In other words, NPOs are in-between TPSs and PAs, since they share some properties with both. Here the core idea is that NPOs are still a special type of partitive expressions where sentential negation is the modifier of the silent AMOUNT/NUMBER, but they do not have the small clause structure typical of TPSs. As already proposed by Rowlett (1998)(see also Manzini and Savoia 2011), we surmise that the negative marker pas originates as part of the nominal expression, and in this sense it can be compared to those languages that express negation by means of negating the object. Pas is then moved out of the QP and attracted to the Spec of NegP via the usual probe-goal mechanism. ^[19]

Since in cases of NPOs like (34) the quantifier has scope over its c-command domain, hence it requires the following nominal expression not to be definite, because a quantifier cannot have scope over a definite expression. ^[20] On the contrary, TPSs are complex structures that contain two DPs, one of which is under the scope of the quantifier, while the other is not c-commanded by it, and is definite, and generally d-linked. On the basis of this structural hypothesis, let us now turn to the different Northern Italian dialects that we have seen so far.

5 An account of variation

The analysis above is based on French, where the structural distinction between the two sentences in (35) is the one between a normal DP (as in 35a) whose internal left periphery contains both the marker d- as well as the lower portion of the definite article (much as in Ihsane’s (2013) proposal). The negative case is a quantified nominal expression, where there is still a null Amount/Number as there is in TPSs, although there is no small clause here as in TPSs. Furthermore, in NPOs, the visible modifier of the null Amount/Number, i. e., pas, has been extracted out of the specifier of the part containing the null Amount/Number to reach SpecNegP.

Dialects that behave like French with respect to NPOs have the same structure as French, i. e., they have a negative marker originating in the object position, which is then raised to the NegP position and can license a null Amount/Number through its copy in the SpecQ position. ^[21] This is true independently of the final position reached by the negative marker: as proposed by Poletto (2016) all sentential negative markers are originated in the same position and then raise to different heights in the clausal spine due to their feature endowment, so in principle also higher negative markers can be originated within the object and then raised.

Since Modern Italian and all the varieties of the North East of Italy do not have NPOs of the French type but have PAs for both negative and positive clauses, it is straightforward to assume that the structure of positive and negative sentences is similar. In other words, in standard Italian, and all the dialects that behave alike either there is no null or lexical postverbal negative marker or the null Amount cannot be licensed through its specifier or both. The fact that in the following examples also (37b) is ungrammatical leads us to think that the second condition, i. e., that the null Amount/Number can be licensed through its specifier, is not met in standard Italian and all varieties that behave in the same way. Therefore, these varieties just use PAs in all contexts where it is possible:

All varieties that do not have NPOs of the French type either use PAs or an alternative strategy, (for instance bare nouns).

Old Italo-Romance varieties, just like modern French and modern Gallo-Italic varieties which have NPOs, had the possibility to generate the postverbal negation (like miga) inside the object and it could in turn license the null Amount/Number through a spec-head Agree relation. Examples like (28b), repeated here as (38), are analyzed as in (39) according to the derivation we propose.

Notice that even beyond this stage miga could still be used as a true quantifier:

We are now left with two types of variation described in Section 2: varieties where NPOs are optional and varieties where the NPO must or can present the conflated additional marker, like in PA cases. In the case of the first ones, we have observed that the presence of NPO structures is more frequent with plural countable indefinites. This suggests that there is a difference between the silent Amount and the silent Number. Our hunch is that this concerns different licensing conditions for the two silent items. It appears that Amount is “easier” to license than Number, since in these varieties only Number requires the full NPO structure licensing it through negation. This is probably in line with the semantics of the two quantifiers, as it is generally acknowledged that mass nouns are semantically (and if we are right also syntactically) less complex than plurals. We will not proceed any further along this line of thought, since it would require further data.

Finally, varieties with mandatory or optional conflated additional markers inside the NPO construction can be analyzed assuming that the grammaticalization of the NPO form is based on partitive structures where the conflated marker already lost the definite interpretation and only expressed phi-features of the noun. As we have observed in examples like those in (29) and (30) a similar variation regarding PAs exists in Gallo-Romance and in Piedmontese.

6 Conclusions

In this article, we have tried to provide an overview of the complex microvariation of a specific construction, NPOs, inside the bigger realm of partitive constructions. We have noticed that among the Northern Italian dialects there are some that behave like French, in having a bare P ‘of’ introducing the object in negative sentences, while others use normal PAs or bare nouns in these contexts or even both, like standard Italian. We have proposed a structural formalization of the derivational cline of TPSs to NPOs and to PAs translating into structural terms Carlier’s (2007) idea that TPSs are derived from a motion structure of removal and have gone even further in equating TPSs to inalienable possession, which is standardly assumed to be a small clause, where the ‘part’ is in the specifier and the ‘whole’ is in the complement position, while the P lexicalizes the head of the small clause. In addition to this, we have embedded Foulet’s intuition that these structures must contain a null Amount/Number quantifier, which can be licensed by the overt quantificational modifier in its specifier. As for PAs, we have essentially adopted Ihsane (2013) and Cardinaletti and Giusti (2016) idea that here there is no partitive structure and that the P plus the article is an indefinite with a KP on top realized by the prepositional element, which is, as often assumed in the literature, no more than a type of Case marker. From these two poles, we have proceeded to analyze the intermediate structure, i. e., NPOs, which have in common with PAs the fact that they are regular DPs and not small clauses, and with TPSs the presence of a null Amount/Number, licensed by its overt modifier. All variation we found can be accounted for by using these three structures and can be argued to derive from the possibility to a) have an overt modifier of Amount/Number, b) that this modifier can indeed license the null Amount/Number through a spec-head procedure, and c) that in some languages only Number is licensed in this way.