On the property-denoting clitic ne and the determiner de/di: a comparative analysis of Catalan and Italian

2.1 The definite partitive complement is not resumed by NE

A true partitive constituent cannot appear without an overt indefinite quantifier.^[4] This is clear in Catalan (10), in which the absence of the quantifier alguns ‘some’ is ungrammatical.

However, this claim seems to be challenged by Italian (11), which displays two cases in which the quantifier is apparently optional: the so-called bare partitive (Le Bruyn 2007a, 2007b) in (11a) and the so-called partitive article in (11b).^[5]

If a hidden quantifier was to be assumed in (11) when the quantifier is not overtly realized, we would expect bare partitives (11a) and nominals introduced by the partitive article dei (11b) to be resumed by NE when they are dislocated, contrary to fact. The contrast in (12) and (13) shows that both Catalan and Italian require the quantifier to be overt when the partitive complement is merged at the left periphery (12), while no overt partitive complement can be merged in the absence of the quantifier responsible for the part-whole relationship (13):^[6]

Furthermore, in both languages, clitic resumption of the nominal complement of an indefinite quantifier is mandatory. The ungrammaticality of (14), parallel to (12) without NE, shows that the dislocated partitive is not sufficient to license the empty category in the complement of the quantifier:

On the contrary, the true partitive can be missing if the indefinite complement of the quantifier is pronominalized with NE as in (12) above or is realized as a full nominal as in (15):

Furthermore, NE cannot be missing when the nominal complement of the quantifier is null, as in (16):

Therefore, the co-occurrence of NE and the partitive constituent in (12) is no compelling reason to suppose that NE resumes the true partitive complement introduced by DE, pace Chierchia (1997) and Zamparelli (2008) for Italian and Martí i Girbau (2010) for Catalan. Rather, NE resumes the indefinite nominal complement of the weak quantifier. This is confirmed by the co-occurrence of NE and the partitive in constructions, in which a moved constituent cannot be resumed by a clitic, as is the case of wh-phrases, relative phrases, and focused phrases. Compare (17), in which NE is obligatory in Italian (and improves the grammaticality in Catalan), with (18) in which the accusative clitic is ungrammatical in both languages and the empty category left by the fronted constituent is perfectly legitimate without the clitic:

As originally proposed by Belletti (1979) and discussed in Cardinaletti and Giusti (2006, 2017, the cooccurrence of NE and the overt dislocated partitive in (17), as opposed to the ungrammaticality of the accusative clitic in (18), shows that NE does not resume the true partitive. In these clauses, the displaced constituents move from inside the sentence and bind a null constituent in this position, while NE is associated with the null nominal complement of the in situ weak quantifier.

A final confirmation of the hypothesis that NE does not resume the dislocated partitive constituent is given by the fact that NE cooccurs with the partitive located in situ in sentences like (19a). This is because NE resumes the nominal complement of the quantifier, which is realized by the bare nominal libri/llibres in (19b). The ungrammaticality of NE in (19c) further shows that NE cannot resume the dislocated partitive (dei libri della biblioteca/dels llibres de la biblioteca) in the presence of the overt indefinite complement (romanzi/novel·les) of the quantifier (molti/moltes).

Up to now, we have argued that:

1. the partitive structure requires the insertion of a weak quantifier, NE does not;

2. the definite partitive complement is optional, NE is either mandatory or ungrammatical; and

3. NE resumes the nominal complement of the weak quantifier rather than the partitive complement of a partitive structure.

Having shown that NE does not resume a definite partitive complement is the first step to argue that NE is a non-quantified (i.e., non-partitive) clitic pronoun; more precisely, it is a property-type anaphora. This is the goal of the remaining part of Section 2.

2.2 NE is a non-quantified indefinite clitic pronoun

In (1)–(4), we observed that NE resumes a bare indefinite. In (20), we see that this pronoun cannot resume a quantified expression, and in (21) we show that it cannot resume a nominal introduced by uns in Catalan or the partitive article in Italian, as already argued by Cardinaletti and Giusti (2016) and Espinal and Cyrino (2022a):

Note that this pattern unifies the categorial status of quantified phrases and indefinite nominals introduced by an overt determiner with definite and universal quantifiers in both languages. These are full arguments that refer either to sets of individuals (22a) or sets of sets (22b); given this, they are expected to introduce discourse relationships with entity-type anaphors, which is exactly what has been illustrated in (20b) and (21b), while bare nominals (see (1)–(2)) and DE-phrases (see (3)–(4)) denote properties and as such they are expected to introduce discourse relationships with property-type anaphors (Espinal and Cyrino 2022a):

Catalan provides one more argument for the non-quantificational property-type denotation of NE. First consider that this language allows for what looks like bare singulars (i.e., bare nominals unspecified for Number; Espinal 2010; Espinal and McNally 2011) in the object position of a restricted class of predicates (so-called have-predicates). This nominal in object position is assumed to denote a property of kinds (of type ⟨e^k,t⟩), while Number − analyzed semantically as a realization or instantiation operator (Carlson 1977) – is a morphosyntactic category that applies to properties of kinds (the meaning of the noun) to yield properties of objects (of type ⟨e^o,t⟩; Borik and Espinal 2012, 2015). Pronominalization by NE is, therefore, a privileged diagnostic for indefiniteness. Unfortunately, it does not apply to Italian because bare singular nominals are not productive in this language.^[7]

Thus, the impossibility for NE to resume a nominal introduced by an overt determiner or quantifier in both languages, together with the possibility for Catalan NE to resume a bare singular argument, strongly argues for the indefinite nature of NE, for its lack of partitive status and, more precisely, for its status as a property-type anaphora.

From the discussion so far, we conclude that Catalan and Italian NE is the pronominal counterpart of an indefinite nominal expression that may have various morphophonological forms: either a bare nominal or a DE-phrase (see (1)–(4)). NE neither resumes a true partitive complement (see (19c)) nor a dislocated internal argument introduced by a partitive article in Italian (see (21a)).

In the next section, we discuss a surprising contrast between Catalan and Italian concerning the (im)possibility of resuming the predicate of a predicational copular sentence by means of NE.

4.1 DE as the determiner counterpart of the clitic NE

Following recent work by Cardinaletti and Giusti (2015, 2016 and Espinal and Cyrino (2022a, 2022b), we assume that DE is a determiner that conveys indefiniteness with plural or mass nouns, no matter whether at Spell-Out it is instantiated as de/di or it is silent, thereby yielding a bare noun.^[11]

Espinal and Cyrino (2022a, 2022b) observe that DE-phrases alternate with bare plurals and, like them, have an indefinite meaning and denote properties. Catalan and Italian are languages that allow bare mass nouns and bare plurals in direct-object position of transitive verbs (29a) (see also (1)) or in postverbal subject position of unaccusative verbs (29b). None of these positions allow overt DE, but still a null indefinite DE has been hypothesized by these authors to account for the indefinite reading of the nominal. Overt DE appears to be mandatory in the indefinite complement of semi-lexical nouns (30) and optional with some quantifiers in Catalan (31).^[12]

On a different trend of research, Cardinaletti and Giusti (2015, 2016 propose that the partitive article is formed by the indefinite determiner DE in SpecDP, which is uninflected and needs to be merged with gender and number features in D (namely, the morphology of the definite article, henceforth ART). Looking at Italian dialects, Cardinaletti and Giusti (2018) further argue that DE and ART alternate with silent counterparts across varieties: DE-phrases that appear in basic positions, as in (32a), alternate with bare nominals, as in (32b), nominals introduced by an apparent definite article, as in (32c), or by the partitive article, as in (32d). The examples in (32), taken from the AIS map 1343, are all in Piedmont, suggesting that the overt or non-overt realization of DE and/or ART are subject to fine-grained variation:

The four possible combinations are claimed to create competing forms, that specialize for different “flavours” of indefiniteness when they are coexistent in the same language, but also give rise to variation across dialects and regional varieties of Italian (Cardinaletti and Giusti 2020; Giusti 2021a). Figure 1 presents the rendering by Lebani and Giusti (2022) of the distribution of the four possible determiners in the AIS maps 637, 1037, 1343, corresponding to ‘[to go look for] violets’ (left); ‘[if there was] water’ (central), [to go to the cellar] to take wine’ (left). In blue we find silent DE and silent article (ZERO), in red the combination of overt DE and silent ART, in orange the combination of overt DE and overt ART and in green the combination of silent DE and overt ART:

Figure 1:

Rendering of the AIS maps numbers 637, 1037, and 1343 (Lebani and Giusti 2022: 4, Figure 1, revised).

From the discussion above, we conclude that DE is an indefinite determiner that may but need not occur with overt ART across Romance varieties; it alternates with a silent counterpart, which also may but need not occur with overt ART. In this perspective, Catalan qualifies as a language with silent DE and silent ART in basic object position (29), but overt DE may be required for the expression of indefiniteness with semilexical nouns and indefinite complements of weak quantifiers (30)–(31).

4.2 DE with dislocated indefinite arguments

Recall that exactly like dislocated bare plurals, dislocated DE-nominals denote indefinite expressions that are antecedents of NE. Thus, in (3), repeated here as (33) for convenience, both the dislocated nominal expression and the clitic pronoun are interpreted as indefinites, associated with a non-unique, non-specific, non-familiar reading. Note that DE is mandatorily present in Catalan but optional in Italian left-dislocation (33a), while it must be overt in right-dislocation (33b):

We deal with the contrast in (33) first showing that the difference does not reside in the optionality of DE in left-dislocated indefinite nominals but in the necessity for Catalan to have DE even in indefinite hanging topics. Following a suggestion by Cardinaletti and Giusti (1992, 2006, 2017 for Italian, we assume that di libri in (33) is a left-dislocated topic, while bare libri is possible only in a hanging topic construction.

As noted in much literature stemming from Cinque (1977, 1983, 1990, a hanging topic is less dependent on the host structure than a left-dislocated topic and, for this reason, it may appear as an independent nominal expression without displaying argumental marking, such as the dative preposition a or the overt determiner di (34a). In Catalan both the dative preposition and the indefinite determiner are mandatory (34b):^[13]

Note that the sequences in (34) are ambiguous between a hanging topic and a left-dislocated structure. There is however a battery of diagnostics to disambiguate between the two. First, a property distinguishing hanging topics from left-dislocated topics is the fact that the resumptive element in the former can be a clitic but can also be a strong pronoun as in (35a), or a demonstrative pronoun as in (35b). Once again, Italian prefers no marker on the hanging topic (neither the preposition nor an overt DE), while Catalan rules out the hanging topic structures without overt dative/indefinite markers:

A second diagnostic is that unlike left dislocated topics, hanging topics do not obey the restrictions that apply to extractions, for example the wh-island restriction displayed in (36). Italian prefers the bare noun in (36) whereas Catalan is reluctant to allow this construction.^[14]

A third diagnostic is that unlike left-dislocated topics, hanging topics are root phenomena, they cannot be iterated and can only appear as the highest element in the left periphery of the main clause. Thus, only a left dislocated topic can be the second in the main clause, as in (37), or appear after the complementizer of an embedded clause, as in (38). In (37)–(38), we observe that in Italian, dislocated topics following a hanging topic must display overt DE (37a)–(38a), parallel to the overt dative preposition a (37b)–(38b). Indefinite hanging topics in (37b)–(38b) are completely acceptable without DE, parallel to the missing preposition on the dative hanging topic in (37a)–(38a).

In Catalan (39)–(40), we observe only a slight possibility for the dative preposition a to be missing in the hanging topic position, sentence initial in (39a)–(40a), while DE must always be overt in any position (39b)–(40b):

Summing up, property-denoting hanging topics in Italian may be bare, while in Catalan they must have overt DE. Moreover, DE is mandatory in both languages with dislocated indefinites, as further illustrated by right-dislocation in (33b) above and (41) below:^[15]

Note that the comparative perspective has allowed us to discover an independent common property of Catalan and Italian: both languages do allow indefinites as hanging topics but differ in that Catalan requires DE to be overtly spelled out.

As a last remark, also note that from a semantic perspective, all the examples considered so far show DE preceding dislocated indefinite arguments whose denotation is a property of individual entities. This is expected since, as previously mentioned regarding (28b), both Catalan and Italian allow NE to be a property of individuals anaphora. Likewise, DE is a property-denoting determiner. This notwithstanding, Catalan DE may also introduce a dislocated argument that denotes a property of kinds (42a) (Espinal 2010; Espinal and McNally 2009), a possibility excluded in Italian (42b) because bare singular nouns are excluded in this language (see examples in (23)):

4.3 Stranded DE with adjectives

In both Catalan and Italian, DE cannot precede the adjective in situ (43a) but must do so when the adjective is stranded by NE-cliticization (43b).

Thus, stranded adjectives are similar to bare plurals in that they show overt DE only when they are adjacent to an indefinite nominal expression (44a)–(44b).

In the presence of an overt quantifier that does not select overt DE nominal complements in situ (45a), DE emerges in combination with stranded adjectives (45b), no matter whether a dislocated nominal appears to the left or to the right of the sentence. This shows that stranded DE correlates with but is distinct from the determiner DE that precedes dislocated nominals. The stranded adjective mandatorily displays overt DE in Catalan, while DE may be overt or silent in Italian (45b).

As observed by Cinque (1991) and Cardinaletti and Giusti (1992) for Italian and McNally and Boleda (2004) for Catalan, not just any adjective can be stranded. Apart from intersective adjectives (e.g., interessants/interessanti), classifying adjectives (e.g., pulmonars/polmonare) may be stranded as well, whereas those adjectives that constitute the prototypical examples of prenominal predicate modifiers (e.g., presumptes/presunti) cannot (see McNally and Boleda 2004: examples (17) and (13)).

According to Cardinaletti and Giusti (1992) for Italian and Martí i Girbau’s (1995) for Catalan, the clitic ne/en and the marker di/de that precedes adjectives in structures with elliptical nominals are the manifestation of the same phenomenon: partitive case marking. We reformulate this analysis by postulating that the DE that introduces stranded adjectives is an overt concord marker of indefiniteness. Given the contrast in (43), we postulate that DE with stranded adjectives concords with the indefinite DE of the null nominal it modifies. As the examples in (44) illustrate, this nominal expression may very well be expressed by means of a dislocated DE phrase. This means that one single indefinite argument (the indefinite nominal expression) may be discontinuous at Spell-Out provided that the scattered portions display overt concord for relevant features. In other words, the co-presence of the dislocated DE, the clitic NE and the stranded DE must be considered discontinuous Spell-Outs of indefiniteness and therefore of the same property-type denoting constituent. If this is correct, it must be concluded that adjectives project a position that realizes concord for the indefiniteness of the nominal (as claimed by Giusti 2015: Ch. 4).

Finally, note that if the object nominal is definite, the stranded adjective cannot be introduced by a DE marker, thus supporting our hypothesis that DE with stranded adjectives is the expression of a concord relationship with a property-denoting internal argument constituent. To illustrate this claim, consider the examples in (47) which show that, when the object nominal is definite and is in situ, DE cannot precede the adjective (47a), as was already observed with respect to (43a) for indefinite objects; when the object nominal is a dislocated definite, as expected, DE is also discarded (47b).

4.4 DE with dislocated predicates

Recall from Section 3 that predicates of copular sentences in Catalan, but never in Italian, are resumed by NE. In this section, we concentrate on the occurrence of DE introducing dislocated predicates, which constitutes one of the major points of divergence between the two languages. Observe that in (48) the dislocated predicate may be preceded by de in Catalan irrespective of whether it is resumed by en or ho and also irrespective of the nominal versus adjectival nature of the predicate. However, this is not possible in Italian under any condition.

Note that a lack of dislocated predicate adjectives and nominals with DE and a lack of resumption of these elements by NE in Italian cannot be related to competition with another strategy in the language, since in both Catalan and Italian a neuter non-inflecting clitic is available (Cat. ho/It. lo). We propose that the impossibility of di in dislocated predicates in Italian is directly related to the absence of a predicate ne anaphora in this language, as discussed in Section 3. Therefore, we claim that resumption by NE and the possibility of the determiner DE must be considered two sides of the same phenomenon since NE is a property-type anaphora and DE is a property-type determiner. We understand that the difference between the two languages is a lexical one: namely, whereas Catalan allows that the lexical Spell-Out of both NE and DE denote at least three different types of properties, Italian only admits one.