Number Morphology and Bare Nouns in Some Romance Dialects of Italy

: This paper explores aspects of microvariation concerning the morphological realization of the feature Number within nominal structures in a selected subset of Romance dialects of Italy. First, the different strategies adopted in the dialects of the dataset for the realization of number alternations on various nominal categories (nouns/adjectives, articles, demonstratives, and possessives) are presented. Then, the relation between the latter and the distribution of “bare” argument nominals (i.e., of nominal structures which, in argument position, occur without any lexicalized determiner) is explored. It will be observed that the distribution of bare arguments in the dialects of the dataset is consistent with the hypotheses made in the literature, which suggest that there is a correlation between the realization of number alternations on nouns and the possibility for “null” (i.e., unpronounced) determiners to be licensed.


Introduction
The relation between the morphological representation of the feature Number on nouns and the possibility for nouns to be realized as bare (i.e., not introduced by any overt determiner) in argument position was explored by Delfitto and Schroten (1991). 1 They propose that, in English, the licensing of argument bare nouns depends on the realization of number alternations through overt affixes "attached to a 'free' morpheme" (Delfitto and Schroten 1991, p. 157): a silent "plural quantifier" is licensed when the affix raises to D at LF, "providing the correct quantificational representation" (Delfitto and Schroten 1991, p. 162). By contrast, in French, where morphological number exponence on nouns is generally absent, 2 bare arguments are ungrammatical: argument nominal structures, including indefinite mass and plural nouns, require a visible "quantification operator" in D, as shown in (1).
A similar proposal is presented in Crisma and Longobardi (2020, pp. 51-54); in their analysis, the possibility of identifying the feature Number via a null D is connected to the availability of overt morphological Number exponence on nouns, which distinguishes, for instance, languages like English, Italian, or Spanish-where number alternations are lexicalized on (almost) all nouns and bare arguments are possible-from languages like French, where nouns are generally unmarked for Number and bare arguments are impossible.
Concerning the realization of number alternations on suffixes, English, Italian, and Spanish exhibit three different strategies. In English, as already mentioned, plural suffixes In the Romance dialects of Italy, number marking on nouns is realized through various strategies (for a detailed survey, see Manzini and Savoia 2005, vol. III). For example, there are dialects, such as Ladin or certain Friulian varieties, where "syncretic" suffixes of the Italian type alternate with the combination [WM + -s] (see, among others, Pescarini 2020). Since none of the dialects considered in this paper exhibit -s suffixes, this phenomenon will not be further explored here. By contrast, we focus on the effect of the loss/weakening of final vowels (Tagliavini [1949] 1972) on the realization of "syncretic" suffixes. We observe two types of dialects. In the dialects where final vowels were not lost/weakened, number alternations are realized on suffixes, which, like in Italian (Manzini and Savoia 2005, vol. III, pp. 547-48), collapse class/gender and number information. By contrast, in the dialects where final vowels were dropped or became indistinct, the overt realization of number alternations on suffixes was blurred as well: in some such dialects, these phenomena affected almost all noun classes; in others, the weakening process affected only some final vowels: therefore, number alternations were retained on some suffixes and became lost in others. In turn, in some dialects, alternative strategies were developed to overtly mark singular vs. plural distinctions, through the re-analysis of stressed vowel alternations originally induced by metaphony (Tagliavini [1949(Tagliavini [ ] 1972Fanciullo 1994). 5 With respect to these phenomena, the languages of our dataset can be classified into two major types: (1) languages where final vowels were preserved and, consequently, number alternations are realized on suffixes on all/most nouns, and (2) languages where, as a consequence of the weakening/loss of final vowels, number alternations on suffixes were (entirely or partially) lost. The latter, in turn, split into two further types: (2a) languages where, due to the loss of final vowels, suffixes were lost as well on entire classes of nouns, and (2b) languages where, due to the weakening of final vowels, suffixes were retained but have lost (either partially or entirely) number distinctions. These outcomes are discussed in the first part of the paper and are summarized in (26). The first part of the paper also contains a survey of the strategies available in the languages of the dataset for marking number alternations on nouns, adjectives, articles, demonstratives, and possessives.
In the second part of the paper, we focus on the relation between the morphological realization of number on nouns and the availability of bare argument nouns. 6 Using original data collected from native speakers, we sketch a survey of the distribution of bare arguments in our sample of dialects. Two generalizations emerge from our data: (a) languages where there is regular/systematic number exponence on nouns have bare nouns; (b) the absence of suffixes on nouns seems to correlate with the absence of argument bare nouns.
The structure of the paper runs as follows. Section 2 introduces the languages of the dataset and presents some properties relevant for our discussion. Section 3 presents three parameters that govern the realization of the feature Number on nominal structures and describes the morphological strategies for number marking in the dialects of the dataset, In what follows, we provide a brief description of some phenomena traditionally observed in these four groups of dialects that are relevant for our discussion.
In the dialects where final vowels became -@ (i.e., most dialects of group (4b)), nouns (and adjectives) originally ending in -E, which realized plural number through the suffix -I, were affected by metaphony in the plural (triggered, in fact, by -I), while no change happened in the singular (because final -E does not trigger metaphony). Consequently, when final unstressed -E and -I changed into -@, singular vs. plural alternations were preserved through the alternation, on the stressed root vowel, between non-metaphonetic (singular) and metaphonetic (plural) outputs (e.g., mes@, 'month.SG' / mis@, 'month.PL'). By contrast, nouns (and adjectives) ending in -U, whose plural was -I, were affected by metaphony both in the singular and in the plural (because both final -U and final -I trigger metaphony); thus, after the weakening of final -U and -I, no alternation was preserved (e.g., nwov@, 'new.M.SG'/ nwov@, 'new.M.PL'). Finally, nouns and adjectives ending in -A, whose plural was -AE>-E, and were thus etymologically unaffected by metaphonetic changes, did not retain any singular/plural alternation after the weakening of final vowels (for example, in Campania, the item rOt@, 'wheel', encodes both singular and plural number). 37 Nouns/adjectives of all classes whose stressed root vowel was -a-, -i-, or -uare not expected to display any metaphonetic output. Yet, there are exceptions. For example, Fanciullo (1994) reports cases of lexical roots etymologically unaffected by metaphony whose stressed vowels display alternations signaling singular vs. plural interpretation; he concludes that, in these dialects, metaphony, which was originally a phonetic/phonological process, was turned into a morphological procedure (that he calls morphometaphony) to maintain/restore the overt realization of singular/plural alternations originally lexicalized on suffixes (Fanciullo 1994, pp. 574-77).
Various instances of these phenomena are visible in Teramano and other dialects of Abruzzo (Fanciullo 1994;Savoia and Maiden 1997;D'Alessandro and Van Oostendorp 2014, a.o.), where the extension of stressed root vowel alternations to non-etymological contexts also affected -a-, as shown in (8) (from Mantenuto 2015b, p. 11).
In various dialects of northern Italy (Rohlfs 1966, § § 141-47;Loporcaro 2009, p. 80;Foresti 1988), metaphony is triggered only by final -I, as in the following examples (from Foresti 1988, p. 579): In the dialect of Savignano sul Rubicone (Pelliciardi 1977, pp. 45-48), like in other dialects of Romagna (Foresti 1988), the realization of number distinctions through alternations of the stressed vowel was extended to almost all classes of nouns (and adjectives) which, because of the loss of final vowels, have lost their suffixes (see Tables A4 and A5). Thus, in this dialect, three different types of nouns are identified on the basis of number marking strategies: indeclinable nouns (10a), nouns where singular vs. plural alternations are realized on suffixes (-a vs. -i, or -a vs. no suffix, as shown in (10b.i) vs. (10b.ii)), and nouns that have no suffix and realize number oppositions through stressed vowel alternations (10c).
To sum up, the dialect of Savignano sul Rubicone shares the absence of suffixes on most noun (and adjective) classes with the other dialects of group (4a), while it shares the overt realization of Number through alternations of the stressed root vowel with the dialects of group (4b), where, in contrast, all nouns/adjectives have suffixes (although often undistinguished for Number). In Section 3.3.1, we briefly explore these differences and their morphosyntactic consequences.

Number in Nominal Structures
In this section, we first describe three syntactic parameters that are responsible for the representation of Number within nominal structures (Section 3.1) and observe their manifestations in our sample of dialects (Section 3.2). Then (Section 3.3), we describe the morphological strategies observed in our sample of dialects to mark number alternations on nouns and adjectives (Section 3.3.1), articles (Section 3.3.2), demonstratives (Section 3.3.3), and (pronominal) possessives (Section 3.3.4). Concerning the latter two, we also highlight some aspects of variation in their distribution across the languages of the sample.

Number in DPs
We start from a brief survey of three parameters that have been proposed in the literature to account for cross-linguistic variation in the representation of the feature Number in D and within DPs. 38 Number is one of the features that can be realized in D. 39 Languages differ according to whether Number is "grammaticalized" or not, where "grammaticalized" means "obligatorily valued through some overt exponence in syntactically defined contexts" (Crisma and Longobardi 2020, p. 21). 40 In the comparative parametric analysis of the nominal domain proposed since Guardiano and Longobardi (2005) and Longobardi and Guardiano (2009), until its most recent instantiations , this cross-linguistic distinction is encoded by parameter Grammaticalized Number, whose empirical manifestations are summarized in (11). The parameter is active (i.e., Number is grammaticalized) in languages that display (at least one of) the patterns in (11) (e.g., English and Italian). By contrast, the parameter is not active in languages that do not display any such manifestations (e.g., Mandarin or Japanese).
In turn, languages where parameter Grammaticalized Number is active are of two types : languages (e.g., Basque) where singular vs. plural alternations are realized overtly only in D, where the latter must thus be systematically lexicalized (i.e., bare arguments are ruled out, Longobardi 2021), and languages where number distinctions are overtly realized (also) on nouns and other nominal categories (e.g., Italian, English, and, more generally, Indo-European languages). This difference is assumed to follow from a further parameter, Number spread to N (Longobardi and Guardiano 2009;Longobardi et al. 2013;, that is active in Indo-European and non-active in Basque. Finally, among languages where Number spread to N is active (i.e., where at least some nouns exhibit overt singular vs. plural alternations), there are languages where overt number exponence is generalized to all (or most) classes of nouns (like English, Italian, or Spanish) and languages (like French) where number alternations are visible only on a lexically restricted set of nouns (see note 2). The latter do not allow argument bare nouns, thus superficially behaving like Basque, while the former can allow empty Ds in argument nominals. This difference is encoded by parameter Number on N, which is active in English, Italian, and Spanish (see examples (2) and (3)), and non-active in French (see example (1)).
The implicational relations between these three parameters are summarized in (12). 41

Parameter Manifestations in the Dialect Dataset
In what follows, we briefly explore the evidence available in the dialects of our sample concerning some of the patterns listed in (11), which define the empirical manifestations of parameter Grammaticalized Number. 42 In (13)- (19), we provide examples of manifestations (11b) (subject-verb agreement) 43 and (11c) (agreement in number between a noun or a definite article/demonstrative/quantifier and adjectives). Concerning (11a), an overview of how number alternations are realized on major nominal categories is provided in Section 3.3.
Example (13) is from Casalmaggiore (group (4a)). In this example, agreement in number between subject and verb is visible only on the participle: as in several other dialects of northern Italy, the auxiliary be does not agree in number with the subject (Manzini and Savoia 2005, chps. 2 and 5). Concerning DP-internal agreement, in the feminine, number agreement is visible on all DP-items (noun, adjective, definite article).
Examples (14) and (15) are from Teramo and Santa Maria Capua Vetere (group (4b)), respectively. 44 Both examples show subject-verb agreement. DP-internal agreement is realized on all DP-items: nouns and adjectives realize number alternations on the stressed vowel.
Examples (16)-(18) are from dialects of group (4c). Here, again, number agreement is visible between subject and verb and on all the items that belong to the same DP.
Example (19) is from Aidone (group (4d)). Here, too, the verb agrees in number with the subject ((19a) vs. (19b)) and there is DP-internal agreement in number: the latter is only visible between items interpreted as feminine singulars. 45 To sum up, all the dialects of our sample display subject-verb agreement in number (11b), and number agreement within DPs (11c). This constitutes positive evidence for Grammaticalized Number.
Concerning parameter Number spread to N, all the dialects of our sample display at least some nouns where number alternations are realized overtly (see Section 3.3.1 and Table A4); this means that, in these languages, parameter Number spread to N is active. In contrast, the empirical manifestations of parameter Number on N are more variable. We present the relevant data in Section 3.3.1.

Nouns (and Adjectives)
As shown in Sections 1 and 2, across the dialects of our dataset, two strategies (often in complementary distribution) for making number on nouns 46 are observed: suffix alternations and root vowel alternations 47 . The former strategy, as remarked in Manzini and Savoia (2005, vol. III, p. 583), is "clearly computational, namely syntactic" ("chiaramente computazionale, cioè sintattico"). The latter has the properties of an "introflexive" mechanism ("un paradigma di flessione interna"), which yet is not generalized to all inflectional classes and, especially in some dialects, is not productive (Fanciullo, p.c.).
In the dialects of group (4c) and in Felitto, number alternations are visible on almost all classes of nouns (with isolated lexical exceptions), 48 and are realized through suffixes, which, like in Italian, collapse class/gender and number information, as shown in (20). 49 In most such dialects, gender alternations are visible in the singular and blurred in the plural, as shown in (21). 50 The other dialects of our dataset display a "reduction of the internal articulation of the inflectional system" ('una riduzione dell'articolazione interna del sistema flessivo', Manzini and Savoia 2005, vol. III, p. 574) as compared to Italian. For example, in the dialects of group (4a), number alternations are visible on suffixes only on one class of nouns, i.e., feminine nouns ending in -a, with plural -i. 51 The other classes do not exhibit any suffix (example (22) is from Casalmaggiore). 52 Finally, in Savignano sul Rubicone, like in other dialects of Romagna, there are nouns that have no suffixes but overtly mark singular and plural interpretation through stressed vowel alternations (see the examples in (10)).
In group (4b) (except for Felitto) and Aidone (4d), most final vowels have been weakened, and are now realized as -@. Thus, most number alternations on suffixes were lost. As already mentioned in Section 2 (see examples (5), (6), and (7)), in Verbicaro (23) and Aidone (24), some alternations on suffixes have been maintained, because of the retention of final -A. In these languages, there are two different suffixes: -a, which corresponds to (feminine) singular interpretation (originally -A), and -@, which collapses masculine singular (originally -U) and plural (originally -I). 53 (23)  By contrast, in most dialects of group (4b), where final -A, -U, and -I turned into -@, there is only one suffix, which is not specified for number (i.e., -@). Yet, in some such dialects, suffixes -a, -u, and -i re-emerge in certain contexts/classes of items, which might suggest that suffixes still encode number information, albeit often non-overtly. 55 For example, Ledgeway (2007, p. 106) observes that, in the dialect of Napoli (like in several other dialects of Campania), (some) prenominal adjectives (as well as other prenominal modifiers) retain the suffixes -a (feminine, singular), -u (masculine singular), and -i (plural), while postnominal adjectives generalize -@. Another case is exemplified by Francavilla in Sinni (examples in (25)). Here, on nouns, the feminine singular suffix is realized as -a when the noun is followed by an adjective (see (25a.i)); otherwise, it is realized as -@ (see (25a.ii)). Similarly, on adjectives, the feminine singular suffix is realized as -a when the adjective precedes the noun and as -@ when the adjective is post-nominal, as shown by the contrast between (25b.i) and (25a.i). 56 Similar patterns have been described for Bari (Andriani 2017, p. 92) and Teramo (Savini 1881, p. 58 To sum up, concerning morphological exponence of Number on nouns, four major types of languages are identified in our dataset: 58 In Section 4, we explore the relations between these number marking strategies and the distribution of bare nouns in the languages of the dataset, to check whether the predictions made by Delfitto and Schroten (1991) and  are met in this domain of languages.
Before closing this Section, we provide a short survey of three categories (articles, demonstratives, and possessives) that display variation across the dataset in terms of how they realize number alternation and their DP-internal distribution.

Definite Articles
All the dialects of our sample have a definite article, normally inflected for gender and number, as shown in Table 1 (full paradigms are given in Table A1). 59 Few dialects display four-member paradigms; by contrast, most of the dialects have three-member paradigms, with only one plural item that syncretizes masculine and feminine.

Demonstratives
In all the dialects of the database, demonstratives are incompatible with articles (i.e., they are D-checking, Guardiano and Stavrou 2020). Most dialects display two different items for adnominal and pronominal demonstratives: adnominal demonstratives are usually "reduced" as compared to pronominal ones (monosyllabic vs. disyllabic). 61 Almost all the dialects of the sample realize number distinctions on both forms (full paradigms are given in Tables A2a and A2b). Like definite articles, only a few dialects display four-member paradigms, e.g., Casalmaggiore (27) and Palma Campania (28); in the latter, feminine plural demonstratives trigger Rafforzamento Fonosintattico (see note 60). In most of the other dialects, masculine and feminine gender are collapsed into one and the same item in the plural, while they take two separate forms in the singular.
In Francavilla in Sinni, pronominal demonstratives realize number distinctions through alternations of the root vowel, as shown in (29a), while suffix alternations are visible only on adnominal demonstratives (29b-d).
A further aspect of variation in the languages of our dataset concerns the realization of deictic interpretation (Guardiano and Stavrou 2020). Some dialects of group (4a) (Casalmaggiore, Correggio, Novellara, and Reggio Emilia) are like French: adnominal demonstratives are usually realized as two separated lexical items (as shown in (27) and (30), from Vezzosi 2019, p. 26), wherein one (ku/ki, kla/kli) occurs in the D-area, does not encode any deictic information (i.e., it is "deictically neutral") and is inflected for gender and number, and the other (ke/le, a deictic "reinforcer", Bernstein 1997) occurs DP-finally (after adjectives and prepositional phrases), realizes deictic reference and is uninflected. A difference between Casalmaggiore and the other dialects is that, in Casalmaggiore, the reinforcer seems to be required in all contexts (and with all interpretations, Vezzosi 2019), while in the other dialects (like in French) it is not obligatory when the demonstrative does not have deictic meaning.
In Parma and Savignano sul Rubicone, and in the other groups, deictic demonstratives do not require reinforcers; like in Italian, they take different forms according to whether they encode proximal or distal distinctions, as shown in (31).
A peculiar case is instantiated by the dialect of Teramo. As noted in , Teramano features "demonstrative doubling": two demonstratives can occur in the same DP, one DP-initially, in the reduced form of adnominal items, and the other DP-finally, in the non-reduced form of pronominal items. Doubling is not obligatory but is preferred when the DP has deictic interpretation.

Possessives
By the label "possessive", we refer here to pronominal forms interpreted as arguments of the head noun (i.e., expressing one of the following relations: Possessor, Subject, Object; Crisma et al. Forthcoming). In several Romance dialects of Italy, 62 like in Italian (and unlike in English), adnominal possessives must co-occur with a determiner (e.g., an article or a demonstrative: il/un/questo mio libro, lit. 'the/a/this my.M.SG book.M.SG' vs. *mio libro) and do not assign any definite reading to the noun phrase they modify. 63 In various dialects of our dataset, possessives agree in gender and number with the head noun, like in Italian, 64 while in others they don't. 65 For example, in group (4a), adnominal possessives are prenominal (with exceptions, see Vezzosi 2019, p. 50) and uninflected. 66 Pronominal possessives display number alternations only on some forms: 1st and 2nd person plural, except for Savignano (nOstra.F.SG vs. nOstre, vOstra.F.SG vs. vOstre), exhibit the same suffix alternations as adjectives (like in most other dialects, Manzini and Savoia 2005, p. 573), i.e., -a/-i in the feminine (nOstra/nOstri, vOstra/vOstri) and no alternations in the masculine (nOster, vOster); 3rd person plural forms are uninflected (lor); 1st, 2nd, and 3rd person singular are uninflected for gender when modifying a plural head noun (me, to, so), while, in some dialects, gender alternations (through the suffixes -o/-a, masculine and feminine, respectively) are visible on possessives modifying a singular head noun (e.g., mio/mia, tuo/tua, suo/sua in Reggio Emilia and Novellara).
More variation is observed in group (4b), mostly resulting from the combination of the weakening of final vowels and metaphony. Barletta has uninflected items for all persons but 1st singular, which has two forms opposing masculine (mej@) vs. feminine (ma), with no number oppositions. In Bari (Andriani 2017, p. 106), Amalfi, and Palma Campania, possessives are inflected for gender but not for number, except for 1st person singular in Amalfi and Palma Campania, where mij@ encodes singular (masculine and feminine), mjej@ encodes masculine and plural, and mEj@ encodes feminine and plural-note that no such alternation is realized through suffixes. In Verbicaro, all items are invariable (Silvestri 2016, p. 135), except for 1st and 2nd person plural, which, as in most other dialects, display gender/number alternations identical to adjectives. In Teramo, except for 1st person singular (mi), possessives are inflected for number (to vs. tu, so vs. su, nOstr@ vs. nustr@, vOstr@ vs. vustr@, so vs. su) but not for gender. In Francavilla in Sinni, 3rd person plural is uninflected (lor@) and 1st and 2nd person plural, like adjectives, have two forms, one encoding feminine singular (nOst@/vOst@), and the other (nwost@/vwost@) encoding plural and masculine singular. The other forms (1st, 2nd, and 3rd person singular) are inflected for number (mij@ vs. mej@; tuj@ vs. toj@, suj@ vs. soj@) but not for gender; in all forms, number distinctions are realized through root vowel alternations only. In Taranto, possessives exhibit the same gender/number alternations as adjectives (see Table A4). In Santa Maria Capua Vetere, the suffix -a encodes feminine gender and singular number on all persons but 3rd plural. Plural number and masculine gender are realized as -@. In the 1st person singular, root vowel alternations encode singular vs. plural and masculine vs. feminine oppositions (mij@.M.SG, mija.F.SG, mjej@.M.PL, mEj@.F.PL). Finally, in Felitto, number alternations are realized through suffixes (-u/-a for the singular, masculine and feminine, respectively, -i/-e for the plural); in 1st and 2nd person plural, gender alternations are also visible on the stressed vowel (-wo-vs. -O-, masculine and feminine, respectively).
In group (4c), the dialects of Sicily and Reggio Calabria have two types of possessives (Manzini and Savoia 2005, vol. III, pp. 552-74;Guardiano et al. 2018). Pre-nominal possessives are uninflected and realized attached to D (called "Wackernagel possessives" in Guardiano et al. 2018); by contrast, inflected adnominal possessives are post-nominal and identical to pronominal "strong" items. The latter are inflected for gender and number in Ragusa, Trapani, and Reggio Calabria; in Ribera, San Filippo del Mela, and Mussomeli, possessives are uninflected (except for 1st and 2nd person plural, which display the same suffix alternations as adjectives). In Salento (Cellino San Marco, Mesagne, Botrugno) and the rest of Calabria (Cutro, Catanzaro and Nicastro), possessives are post-nominal only; in Cellino San Marco, they are inflected for number but not for gender, except for 1st and 2nd person plural, which display the same suffix alternations as adjectives and stressed vowel alternations (nweSSu/vweSSu.M.SG, nOSSa/vOSSa.F.SG, nweSSi/vweSSi.PL), and for 3rd person plural, which is invariable (lOru); in Nicastro, Mesagne, and Cutro, possessives are uninflected (except for 1st and 2nd person plural and, in Nicastro, also 1st person singular), and in Botrugno and Catanzaro (except for 3rd person plural), they are inflected for number and, in the singular, also for gender.
In Aidone (4d), possessives are uninflected, except for 1st and 2nd person plural, which, as in the other dialects, display the same alternations as adjectives.

The Distribution of Bare Nouns
Based on the premises sketched in Section 3.1, and on the data discussed in Section 3.2, we can now attempt some preliminary predictions concerning the availability of bare nouns in argument position in the dialects under investigation.
In what follows, we illustrate the distribution of bare nouns in argument position in the dialects of the dataset, to check whether the predictions suggested in (33) are borne out. To do so, we test (a) whether bare arguments are grammatical in the dialects investigated and, if they are, (b) whether their distribution displays any differences with respect to Italian.

Data Collection
The data used to test the predictions in (33) were collected from native speakers and further integrated, when possible, with those available in the literature. When not otherwise specified, the examples presented in this Section come from elicited data. For each dialect, we interviewed one speaker. If needed, speakers were consulted multiple times to refine and double-check the material they provided.
We arranged a list of syntactic environments where testing the degree of acceptability of various types of nominals when used with no visible determiner (i.e., bare). Our starting point is the distribution of bare nouns in Italian , which is summarized in (34) and exemplified in (35)-(39).
Based on this material, we created a list of 41 different patterns/sequences (see Table A6) that contain modified and unmodified plural nouns, singular count nouns, and mass nouns, in the following positions in non-negative sentences: The list was then used as a "lexically-flexible" guide to collect the relevant data from the speakers and test their judgements. 67 During the interviews, for each of the 41 patterns, speakers were asked to generate a sentence in their language with the same properties of that pattern. The language used in the interviews was Italian. It has been shown (Pinzin and Poletto 2022) that the use of Italian (i.e., the other native language for most of the speakers) as the input language in this type of task is likely to induce non-trivial priming effects (Cornips and Poletto 2005;Van Craenenbroek et al. 2019). To control for this, when possible, we provided sentences in the dialect of the speaker, asking about their grammaticality. 68 Each pattern was tested twice, using different lexical items. When the answers provided by a speaker were unclear or inconsistent, we tested the relevant pattern multiple times, controlling for potentially infelicitous lexical choices and/or pragmatic contexts.
In what follows, we present the data that are relevant to test the predictions in (33).

Data Description
In the dialects of group (4c) and in Felitto (type (26a)), bare argument nouns are grammatical, and their distribution is constrained by the same restrictions as in Italian (see (34)): (a) bare singulars are ungrammatical in all argument positions (see examples (41a-c) from Ragusa); (b) plural (and mass) nouns are ungrammatical as pre-verbal subjects (41d) and very marginally accepted when modified by an adjective, a relative clause, or a PP (41e); and (c) unmodified bare plural (and mass) nouns can only occur as objects (41f-g), post-verbal subjects (41h-i), and pivots of existential sentences (41j-k). Prediction (33a) is therefore borne out.
Let's now explore (33b). In the dialects of type (26d), bare nouns are ungrammatical in all contexts, as shown in (42). By contrast, in type (26b), bare nouns are possible and occur with almost the same distribution as (34). As shown in (43), a difference with respect to (34) is that plural (and mass) nouns are only marginally accepted in pre-verbal position, even when modified.
As mentioned, there is one difference between type (26d) and type (26b), i.e., the absence, in type (26d), of suffixes generalized to all nouns. In this latter group, like in French, there are nouns that do not exhibit any specification for Number and have no suffixes; according to Delfitto and Schroten's (1991) analysis of French, this type of nouns do not allow the licensing of empty Ds. Apparently, the persistence of number alternations on a restricted class of suffixes (e.g., -a/-al/-el) and, in some dialects (e.g., Savignano sul Rubicone), the realization of number alternations through other strategies (i.e., alternations of the stressed vowel) are not sufficient to license bare arguments. By contrast, one can assume that, in type (26b), because of the alternation between the suffix -a and the suffix -@, the latter is assigned number/gender interpretation (e.g., non-singular, non-feminine) even though it is not overtly specified for number/gender. If this line of reasoning is on the right track, a possible conclusion is that, among languages that partially mark number on nouns, what sets a difference between those which allow bare nouns and those which do not is not the amount of nouns that overtly display number exponence, but rather the morphological structure of nouns themselves, and the mechanisms that, through this structure, make the retrieving of number information possible even when the latter is "silent" (i.e., not overtly specified).
A further similarity between the dialects of type (26d) and French is that, in all the dialects of this group, like in French, the item that overtly realizes D with non-definite plural/mass nouns is a "partitive-like" article (DE+definite article; for a recent detailed survey of the distribution of this item in northern Italy, including Emilian varieties, see Pinzin and Poletto 2021). In all the other dialects of our dataset, this item is never used to lexicalize D with non-definite plural/mass nouns in argument position. 69 Finally, concerning prediction (33c), the data collected from languages of type (26c) show the highest variability. Speakers of Francavilla in Sinni, Taranto, and Amalfi accept argument bare nouns under the same conditions as types (26a) and (26b), as shown in (44) In Teramo, according to Mantenuto (2015a, bare nouns are ungrammatical, as shown in (47). The ungrammaticality of bare nouns in this dialect seems to be a recent phenomenon: older varieties of Abruzzese accepted bare (plural) nouns at least in some argument function, for example in object position, as shown in (48)-from Finamore (1882, pp. 112, 148, respectively). 71 Similarly, our speakers of Santa Maria Capua Vetere, Palma Campania, Bari, and Barletta do not accept bare nouns in argument position: argument nominals require an overt D-like item (e.g., an article, a demonstrative, or a quantifier); a difference with respect to French and type (26d) is that these dialects never use the "partitive article" to introduce indefinite arguments (see Pinzin and Poletto 2021). However, in all these dialects, there are signals that bare nouns are not entirely ruled out. For example, the speaker of Palma Campania accepts bare plurals as pivots of existential constructions when the coda is a relative clause, as shown in (49). Moreover, in the literature about other dialects of Campania (e.g., the dialect of Napoli, Ledgeway 2009, p. 191), instances of argument bare nouns are attested, as shown in the examples in (50), which show that, in Napoli, plural bare nouns in object position have been grammatical at different diachronic stages, and mass nouns as pivots of existential clauses were possible at least until the 20th century. Instances of argument bare nouns are also reported in the literature about the dialect of Bari, as can be seen in (51).
(51) Bari (Lacalendola 1972, p. 22, in Andriani 2017 The conclusion seems to be that, in the dialects of type (26c) where bare nouns are not currently accepted by the speakers, they were possible, and presumably productive, at older diachronic stages. Prediction (33c) is therefore partially met. What remains to be explained is why-although there is no visible difference among languages of type (26c) in the realization of number on nouns-in some dialects bare nouns have become lost while in others they remained productive. Table 2 compares the two groups of phenomena considered so far in the dialects of the dataset: the morphological representation of number on nouns and the availability of bare arguments. In the Table, the label S indicates that all nouns have suffixes that realize number alternations systematically. The label S/@ indicates that all nouns have suffixes, but number alternations are overtly realized only in a subset of nouns. The label S/0 indicates that suffixes (which also show number alternations) are visible only on a subset of nouns, while the other nouns have no suffixes and do not realize number alternations. The label S/0/M indicates that some nouns have suffixes specified for number, while other nouns have no suffixes, but some of them realize overt number alternations on the root vowel. The label @/M indicates that all nouns have suffixes unmarked for number (-@) and number distinctions are realized through stressed vowel alternations. The label YES indicates that bare nouns are grammatical, (roughly) in the same syntactic configurations as in Italian (summarized in (34)). 72 The label NO indicates that speakers do not accept bare nouns under any condition. Finally, the label NO* signals that bare nouns are generally ungrammatical, but there are exceptions.

Summary
The data observed in Section 4 support the hypothesis that there is a relation between the possibility for nominal structures to occur bare in argument position and the realization of the feature Number on nouns. In our dataset, languages where number alternations are systematically realized on nouns can have bare arguments (type (26a)). This sets a first split between languages where number marking is generalized to all (or most) nouns and those where it is not (see the schema in (52)).
Concerning languages where number alternations are only found on some (classes of) nouns, it appears that generalized number marking is not a necessary condition for bare nouns to be licensed; as a matter of fact, bare nouns are also possible in languages where not all noun classes exhibit overt number alternations (type (26b)). Yet, having "partial" number alternations is not sufficient by itself for bare nouns to be licensed (type (26d)). As seen above, a difference between types (26d) and (26b) is that, in the former, there are nouns that do not have suffixes at all. By contrast, in type (26b), all nouns have suffixes, although not all suffixes display number alternations. Thus, the property of having suffixes on all nouns vs. not having suffixes (or having suffixes only on a subset of noun classes) seems to set a further split between languages that can have bare nouns and languages that never license them, respectively (i.e., the second split in (52)).
Finally, among languages where all nouns have suffixes but only some noun classes display overt number alternations (i.e., where number marking is not generalized to all nouns, types (26b) and (26c)), there is more variability. Concerning languages where bare nouns are allowed, we make the hypothesis that, in these languages, there exists some mechanism that allows the speakers to extract number information even from nouns where such information is not visible on the surface. In this respect, the nature of the suffix -@ and the way it interacts with the alternations of the stressed root vowels (in type (26c)) require better investigation. In what follows, we provide some preliminary hints.
We make the hypothesis that final -@ is not merely the "relic" of a phonetic change but has rather retained the morphological properties of a suffix that contains number information. In the dialects where final -a was not lost (type (26b)), the plural interpretation of -@ is presumably induced by pairs like a fava (lit. 'the.F.SG fava bean.F.SG') vs. i fav@ (lit. 'the.PL fava bean'), where -@ is associated with plural interpretation by opposition to singular -a. By contrast, in the dialects where overt number alternations are realized through the stressed root vowel only, one might assume that -@ harmonically agrees  in number with it, thereby replicating number information "silently".
The realization of plural number through multiple morphemes on the same item has been recently explored by Koopman (2020) who, capitalizing on previous proposals by Alexiadou (2011) and Schwarzschild (2015), analyzes "two plural morphemes" in Dutchone realizing the "inner plural" (often through vowel alternations), and the other realizing the "outer plural", in the form of a suffix. This analysis can be tentatively extended to our dataset under the assumption that suffixes realize the outer plural and root vowel alternations realize the inner plural. As a rule, only one such morpheme overtly realizes number alternations, while the other silently agrees with it. Yet, in the dataset, there are also instances (again, like in Dutch) of nouns (and/or other DP-items, e.g., possessives or adjectives) where both the root vowel and the suffix overtly encode number information (see for instance the examples from Felitto in Table A4).
What also remains to be explored, through the investigation of diachronic data, is whether the different degrees of acceptability of bare argument nouns in the dialects of type (26c) correlate with potential differences in the productivity of the number marking strategies available in these dialects, how the latter are related with class/gender marking, and how the phonetic processes of weakening/loss of final vowels are diachronically related with the persistence of morphological information on suffixes. Funding: This work is part of the project MIUR PRIN 2017K3NHHY Models of language variation and change: new evidence from language contact: the data discussed in this paper (when not taken from the literature) have been collected as a part of this project and are the property of the project database (http://www.parametricomparison.unimore.it/site/home/projects/prin-2017/romanceand-greek-dialects-the-database.html; last accessed: 18 August 2022; the content of this website is regularly updated as work progresses).

Informed Consent Statement:
Informed consent was obtained from all subjects involved in the study.

Data Availability Statement:
Most of the data we collected are listed in the examples and in the tables found in the text and in the Appendices A-C. Further data and material can be found on the project website: www.parametricomparison.it, in particular at this link: http://www.parametricomparison. unimore.it/site/home/projects/prin-2017/documents-and-materials.html (last accessed: 18 August 2022; the content of this section is regularly updated as work progresses).

5
See also Maiden (1991), Fanciullo (1995), Loporcaro (2011) and literature therein. 6 See, on this same topic, recent work by Pinzin and Poletto (2022). 7 The data were collected from native speakers and, when possible, double-checked against the existing literature. A description of the areas under investigation, with the relevant literature, can be found at http://www.parametricomparison.unimore.it/site/ home/projects/prin-2017/documents-and-materials.html (accessed on 18 August 2022; the content of this section is regularly updated as work progresses). For a discussion of their classification and major features, we refer to Pellegrini (1977), Maiden and Parry (1997), Cortelazzo et al. (2002), Manzini and Savoia (2005), Loporcaro (2009), Ledgeway and Maiden (2016), among many others.
8 Rohlfs ( , 1968Rohlfs ( , 1969, Pellegrini (1977). 9 Vezzosi (2019). 10 In the transcriptions of the examples, we mark only the allophones which are relevant for the purposes of our description, are peculiar of individual dialects, or oppose different dialects. To signal such phonetic peculiarities, we adopted conventional IPA symbols (https://www.internationalphoneticassociation.org/content/ipa-chart, accessed on 3 May 2022), with one exception: the symbol <å> signals the low-mid central vowel (allophone of /a/ in open stressed syllable) found in Barletta and Taranto  See also Fanciullo (1986Fanciullo ( , 1988, Rizzi and Savoia (1993), Silvestri (2011, 2015). 35 See also Idone and Silvestri (2018, Section 2), to which we also refer for a description (and examples) of the conditions on metaphony in Verbicarese. 36 Metaphony has different manifestations across the Romance dialects of Italy. We refer to the literature for more detailed typologies and examples, e.g., among many others, Rohlfs ( , 1968Rohlfs ( , 1969, Calabrese (Calabrese 1984(Calabrese -1985(Calabrese , 1998(Calabrese , 2008, Maiden (1991), Fanciullo (1994), De Blasi and Fanciullo (2002), Russo (2007), Barbato (2008), Loporcaro (2016), and literature therein. We also refer to Savoia and Maiden (1997) for a detailed survey of the internal variability concerning these phenomena in the Romance dialects of Italy. For the purposes of the present paper, we want to stress the role of metaphony, originally a phonetic/phonological phenomenon, in preserving morphological number alternations on nominal structures; this, in turn, has consequences on the realization of bare nouns in argument position, i.e., a syntactic process. For this reason, in what follows, we mostly refer to those dialects (especially group (4b) and Savignano sul Rubicone, (4a)), where metaphony impacts the morphological realization of Number.
In some dialects, root vowel alternations superficially matching singular vs. plural interpretation also result from propagation (Rizzi and Savoia 1993). Manzini and Savoia (2016, p. 221) describe propagation as "the result of the spreading of [U] properties from an unstressed nucleus to the stressed nucleus (or [a] vowel) immediately to the right". Phenomena of this type are visible for instance on the stressed vowel of nouns preceded by the masculine singular form of the definite article (e.g, u lwibbr@ 'the book' vs. i libbr@ 'the books' in Verbicaro; see also Idone and Silvestri 2018). 37 Yet, in some dialects (e.g., Francavilla in Sinni, Taranto: see Tables A4 and A5) adjectives ending in -u/-a developed a different paradigm: in the masculine (-u/-i), as expected, the combination of metaphony and weakening of final -u/-i generated one item undistinguished for singular and plural (e.g., nwov@ < NOVU(M) and NOVI); in the feminine, the expected form nOv@ (< NOVA(M) and NOVAE) is only used in the singular, while the plural analogically generalizes nwov@.

38
Detailed descriptions of these parameters and their internal dependencies, which are summarized in (12), can be found in Longobardi et al. (2013, Appendix) and . The updated list of their manifestations can be found in Crisma et al. (2020, Supplementary Material). 39 For a recent description of the featural composition of the head D, see . 40 For a typological analysis of Number systems across languages, see Corbett (2000). 41 For the representation of parameter dependencies and implications, see Longobardi and Guardiano (2009), Guardiano and Longobardi (2017), Roberts (2019), and literature therein. 42 We also refer to Manzini and Savoia (2005, vol See also Manzini and Savoia (2005, chps. 2, 4, 5). 44 See also Bari (Andriani, p.c.): stu pum@dor@ jE da S@ttà ('this tomato must be thrown away') vs. sti pum@dur@ so da S@ttà ('these tomatoes must be thrown away'). 45 See also Manzini and Savoia (2005, vol. I, pp. 52, 200, 229, 302, 415, vol. II, pp. 528, 802). 46 In the languages of the sample, nouns and adjectives display very similar patterns concerning number marking. In some dialects, metaphony affects the representation of gender on adjectives (on the strict relation between Gender and Number in these dialects see also note 49 below). Here, we focus on nouns only. Examples of number marking on adjectives in the dialects of the sample are reported in Table A5. See also Manzini and Savoia (2005, vol. III, pp. 574-660). 47 However, items exhibiting both strategies are found across Italy (cf. Foresti 1988): see, for instance, fjore vs. fjiuri 'flower/s' in Padova (Trumper 1972, pp. 13-18), lEpre vs. lepri 'hare/s' in Macerata (Biondi 2012cited in Fanciullo 2015, fOrte vs. fuerti 'strong.SG/PL' in Central Salento (Fanciullo 1994, p. 574). 48 See also Manzini and Savoia (2005, vol. III, pp. 583-90). 49 But see Manzini and Savoia (2005, vol. III, pp. 642-60) and Pescarini (2020). The literature on Romance nominal systems has shown that the realization of Number on nouns is strongly related to that of Gender: "the assignment of grammatical Number depends on the assignment of a formal class to a linguistic category" (Picallo 2008, p. 47). We refer to Picallo (2008), and to work by (e.g., Manzini and Savoia 2005 and Poletto (2021, 2022) for a discussion and a summary of the literature. To account for the relation between Number, Gender and inflectional Class, and for their morphosyntactic realization, the hypothesis of a "layered view of plural" (Manzini 2020, p. 6 These -a plurals are well-known to the literature: we refer to Rohlfs (1968, § 368) and Sornicola (2010) for an overview. 51 Manzini and Savoia (2005, vol. III, pp. 590-99) suggest that -a is to be analyzed as a noun class morpheme, while -i is a "quantificational denotation morpheme" ("morfema a denotazione quantificazionale", 596), denoting both plural number and feminine gender. In other items, such as demonstratives and quantifiers, -i would only express quantificational information (i.e., plural number, 596-597). See also Pescarini (2020). It is not unreasonable that the plural suffix -i instantiates an innovation probably introduced after the loss of final -I, -U, and -E. The origin of this suffix is unclear. Rohlfs (1968, § 363) suggests it to be an analogical creation based on Latin feminine nouns ending in -ȊAE (such as in BESTȊAE > bestij), where final -i was reanalyzed as a plural feminine morpheme. Reasonably, the creation of plural -i happened after the loss of final vowels.

52
There are exceptions: for example, like in French (see note 2), masculine nouns ending in -al/-el in the singular take the suffix -ai/-ei in the plural (kaval/kavai, 'horse.SG, horse.PL'; kavel/kavei 'hair.SG, hair.PL'). 53 See, for a discussion of these systems, Manzini and Savoia (2005, vol. III, pp. 637-42). As it can be seen in the examples (23) and (24), gender alternations are maintained on (some) adjectives. 54 The form bbwell@ results from propagation (see note 36). 55 See, e.g., Manzini and Savoia (2005, vol. III, pp. 637-39), Giuliani (2001, pp. 145-46), Ledgeway (2007, pp. 106-7). 56 For a discussion of similar phenomena in other Romance dialects of Southern Italy, see Manzini and Savoia (2016, Section 3). 57 In Francavilla in Sinni, most adjectives are only post-nominal; by contrast, the adjective bbell@ (along with few additional others) can be realized either pre-or post-nominally. 58 See Tables A4 and A5  See also Manzini and Savoia (2005, vol. III, pp. 574-75). 60 In several dialects of Campania, the plural form of the definite determiner triggers Rafforzamento Fonosintattico (Fanciullo 1997;Loporcaro 1997) in the feminine: a fiLL@~e ffiLL@ 'the daughter~the daughters' vs. o fiLL@~e fiLL@ 'the son~the sons'. On the relation between RF and morphosyntactic structures, see also D'Alessandro and Scheer (2013). 61 See also Manzini and Savoia (2005, vol. III, pp. 582-83). 62 Manzini and Savoia (2005, vol. III, pp. 552-74), Guardiano (2014), Guardiano et al. (2016Guardiano et al. ( , 2018, Silvestri (2020), and references therein). 63 Kinship expressions are exceptional: when a possessive modifies a kinship noun in the singular, and the latter refers to a unique individual, it does not co-occur with any determiner and has a "definite" reading only: mio padre (lit. 'my father') vs. *il mio padre (lit. 'the my father'). In some dialects of our sample (e.g., Salentino, Santa Maria Capua Vetere), when occurring with a kinship noun in the singular, with the interpretation described above, possessives are realized as enclitic (D'Alessandro and Migliori 2017 and literature therein; Manzini and Savoia 2005, vol. III, pp. 660-749 In several dialects, adnominal possessives display "weaker" morphophonological structure as compared to pronominal ones (Cardinaletti 1998;Cardinaletti andStarke 1994, 1999;Manzini and Savoia 2005, vol. III, pp. 570-74, a.o.). In Table A3, for each dialect, we list the pronominal forms, whose paradigms are more variable than those of articles and demonstratives with respect to the realization of number alternations. 66 For further examples, see Manzini and Savoia (2005, vol. III, pp. 554-55). 67 For some dialects (e.g., those used in Guardiano et al. 2016), data concerning the distribution of bare nouns had been collected during previous fieldwork. These data were integrated with novel ones, with the exception of two dialects: Santa Maria Capua Vetere (because the speaker was no more available) and Teramo. For the latter, we found extensive material in the literature, especially Mantenuto (2015a, and the data found in the TerraLing group SSWL (http://test.terraling.com/ groups/7, accessed on 3 August 2022; Koopman andGuardiano 2014-2018): properties O 01 1_Indef mass_can be bare to O 09 5_PN+A_Order PN A and S01_Existential constructions to S 04 3_Indef Pl Ns (Subj) must have an article. 68 The sentences provided by the speakers for each dialect can be found here: http://www.parametricomparison.unimore.it/site/ home/projects/prin-2017/documents-and-materials.html; accessed on 18 August 2022 (the content of this section is regularly updated as work progresses). 69 On the relation between the morphological exponence of gender and number and the realization of nominal determination systems in Romance, see at least Stark (2007Stark ( , 2016; for a recent analysis of the alternation between bare nouns and partitive articles, Pinzin and Poletto (2021). 70 Both variants waññun@ and waLLun@ are found in Francavilla.
139 Loporcaro and Silvestri (2015, pp. 69-72). The suffix -a in the word st@ndEna is a residual of the neuter Latin suffix -A.
140 Mancarella (1998, pp. 89-92, 106-7, 147-48). In Mesagne, final -E and final -I are both realized as -i (Mancarella 1998, pp. 106-7). Thus, there is no suffix alternation between singular and plural on nouns originally ending in -E. In some such nouns, number alternations are realized through metaphonetic alternations of the root vowel, as shown in the examples. This sets a difference with the two other dialects of Salento (Botrugno and Cellino San Marco), where the alternation -E/-I was mantained.