Language contact and morphological competition: Plurals in central Sicily

This paper explores the effects of language contact in the nominal morphology of central Sicilian dialects. In particular, this study is concerned with the contact-induced changes related to the distribution of three plural formatives that give rise to competition between different inflectional classes with respect to a number of lexemes. It is shown that sociolinguistic factors such as speaker age account for the distribution of the competing plural forms and the high degree of variation. As a consequence, a slow and gradual change is leading to the disappearance of the plural form that has no equivalent in the contact language, that is, in Italian.


Introduction: Nominal inflectional classes in central Sicily
In central Sicilian dialects the realization of nominal plural by different formatives, as illustrated in (1), leads us to acknowledge the existence of at least as many inflectional classes (ICs). The most common plural ending across the island is -i (1a), which is typical of the ICs containing masculine nouns, such as carusu, carusi 'boy, boys', and feminine nouns, such as carusa, carusi 'girl, girls'. The suffix -a (1b), which is found for example in the paradigms of jitu, jita 'finger, fingers' and limuni, limuna 'lemon, lemons' (see Sornicola 2010), is etymologically connected to the plural formative of the Latin second declension neuter nouns. The suffix -ura (1c), from Lat. -ORA, is nowadays much rarer than the first two suffixes and is characteristic of the outcomes of the Latin third declension neuter nouns (e.g. Lat. corpus, corpora 'body, bodies'), together with fourth declension nouns such as jocu, jocura 'game, games' (Rohlfs 1968: §370).
(1) PL a. -i b. -a c. -ura As pointed out by Sornicola (2010) and Retaro (2013), the distribution of these three plural endings in central Sicilian dialects is rather complex. The inflectional forms of several lexemes seem to allow for assignment to two or even three ICs, yielding an emblematic case of overabundance (in the sense of Thornton 2011Thornton , 2013Thornton , 2019; only in very few cases is the IC assignment semantically or lexically motivated. In addition, the high degree of inter-and intra-speaker variation makes it difficult to detect covariation with precise extra-linguistic factors. In this study, I investigate the competition in nominal plural formation in the central Sicilian dialect of Mussomeli, focusing on the -ura plurals. I show that the complex situation that characterizes this dialect is the result of an intricate interplay between IC assignment, sociolinguistic variation, and on-going language change in a situation in which Sicilian is in constant contact with Italian, which is the dominant language in terms of prestige and written production. In fact, the three plural forms compete with each other only with respect to a limited number of lexemes and are neither in complementary distribution nor in free variation, prompting the need to resort to sociolinguistic variables. The results of a questionnaire, supported by a statistical analysis, clearly show that the only predictor of the distribution of -ura plurals is the age of the speakers. The -i plurals seem to prevail in terms of productivity, especially among young speakers, but the -a plurals are still frequent and healthy, especially with specific lexemes, as discussed in Sornicola (2010) for the north-eastern Sicilian dialect of Mistretta.
Our results show an on-going change that is leading to the gradual disappearance of the -ura plural and its hosting IC, and that can be attributed to the contact with Italian. In Italian there is no equivalent of -ura, while the ending -i is the most frequent plural suffix for masculine nouns. The survival and resistance of the -a plurals in Sicilian can in turn be explained in terms of reinforcement. Italian has an IC with plural in -a. Even if this class is rather moderate in size (Acquaviva 2008: Ch. 5;Thornton 2013;Loporcaro 2018: §4.2, §7.1), it is apparently sufficient to act as reinforcer or stabilizer of an already existing feature in the contact language. Interestingly, therefore, in this contact situation, Italian is simultaneously inducing the demise of the -ura plurals and the consequent weakening of the corresponding inflectional class, the increase in productivity of the -i plural ending, and the stabilization of the -a plural form.

Gender and inflectional classes in Sicilian
To pave the way for our discussion of Sicilian -ura plurals, we first have to introduce the basic properties of the nominal morphology of this language, starting with gender and number. Like Italian, Sicilian displays a binary gender contrast: masculine vs feminine. Unlike standard Italian, however, Sicilian has a convergent gender system in the plural, with the same plural ending -i for both masculine and feminine nouns. 1 In addition, several masculine nouns belong to the IC whose distinctive plural ending is -a, which is a residual of the Latin neuter nouns of the second declension, and which is still rather widespread in central and southern Italian dialects (Rohlfs 1968: §370, §384). IC assignment of nouns in Sicilian can be described as in Table 1.
Gender correlates with the ICs in a complex but systematic way. IC 1 (-u/-i) only contains masculine nouns. Most nouns pertaining to IC 2 (-a/-i) are feminine, even though this IC also comprises a handful of masculine nouns of Greek origin which end in -eta or -ista in the singular (e.g. u pueta, i pueti 'the poet, the poets', u musicista, i musicisti 'the musician, musicians')the same nouns that belong to an independent class in Italian (see, e.g., D'Achille & Thornton 2003). IC 3 contains invariable nouns, both masculine and feminine, that keep the same form both in the singular and in the plural. Within this IC, three subgroups can be identified based on a diachronic analysis. The first group (cf. -i/-i) is the largest and includes nouns with the plural in -i; in most cases the same form in the singular is the result of the phonological change that raised e to i in the singular in the diachronic development of the Sicilian vowel system. The subgroup a of IC 3 shows the extension of the -i plural ending across ICs, which is thus clearly independent of the singular form. The second group consists of feminine and masculine monosyllabic and oxytone nouns that remain invariable in the plural (cf. subgroup b). Finally, the third group of nouns belonging to IC 3 (cf. subgroup c: -u/-u) are, in diachronic terms, the few residuals of the Latin fourth declension which end in -u in the singular and are invariable in the plural (Rohlfs 1968: §367), such as a soru, i soru 'the sister, the sisters', a ficu, i ficu 'the fig, the figs', and a manu, i manu 'the hand, the hands' (alongside with the 'regularized' plural i mani in some modern Sicilian dialects, which follows the model of IC 1 even though it is a feminine noun). IC 4 and IC 5, with their characteristic plural in -a, exclusively contain masculine nouns. From a semantic viewpoint, the nouns belonging to these two classes tend to denote inanimate objects and to retain a collective meaning, but in their development from Latin to Sicilian this plural ending has also been extended so as to include animate and human referents (Rohlfs 1968: §370, §384;Sornicola 2010). 2 A number of masculine nouns ending in -u in the singular takeor can takethe formative -ura in the plural (cf. IC 6). Whether these nouns must be grouped together in an independent IC or whether they should rather be considered as a subclass or exceptions to IC 4, as suggested in traditional grammars of Sicilian (see Pitrè [1875Pitrè [ ] 1985, is not obvious. We will return to this issue in Section 1.2. As we can see in Table 1, there is a partial correlation between gender and ICs: nouns in IC 1, 4, 5 and 6 are systematically masculine, while IC 2 mostly includes feminine nouns, with the few exceptions mentioned above; nouns in IC 3 comprise both masculine and feminine nouns. The gender of the noun is more systematically and transparently expressed by the form of the determiner, at least in the singular (a for the feminine and u for the masculine). 3 Gender agreement is also revealing in this respect: the ICs of adjectives have the same formatives as those of IC 1-IC 3 for nouns. A first class of adjectives follows IC 1 and IC 2: nicu (m.sg.), nica (f.sg.), nichi (m./f.pl.) 'small', while a second one adopts the model of IC 3, group a: ranni (sg.), ranni (pl.) 'big' (with some exceptionsmostly loanwords -following IC3, group b, e.g. blu (sg./pl.) 'blue'. Nouns belonging to ICs 4, 5 and 6 do not show any special property with respect to gender and gender agreement, triggering masculine agreement in the singular and genderunspecified agreement in the plural, both with determiners and with adjectives. See the examples in (2)-(4): (2) a. A long-standing debate exists as to whether -a plurals instantiate a third genderthe neuter genderin Italian, and especially in Italo-Romance varieties where this ending is more widespread and productive (see Hall 1965;Maiden 1995;Acquaviva 2002Acquaviva , 2008Faraoni 2012;Rovai 2012;Faraoni, Gardani & Loporcaro 2013;Gardani 2013;Loporcaro, Nolè & Paciaroni 2013;Thornton 2013;Loporcaro, Faraoni & Gardani 2014;Ledgeway 2009;Loporcaro 2018, among others). While there is broad consensus that the Italian IC involving residuals of the Latin neuter plural ending -a (e.g. il braccio (m.sg.), le braccia (f.pl.) 'the arm, the arms') does not constitute a third distinct gender of its own (see, e.g., D'Achille & Thornton 2003), evidence has been adduced and discussed in favour of the hypothesis that in association with the -a and the -ora plural, Old Italian, to be understood as Old Tuscan, displayed a third gender with a distinct agreement pattern (e.g. lo prato (m.sg.), le pratora (f.pl.)) (see Gardani 2013: 330-333 andGardani 2014). 4

The -ura plurals
The Sicilian -ura plurals etymologically derive from the Latin neuter nouns of the third declension (Lat. corpus, corpora 'body, bodies'), 5 with extension by analogy to nouns from other declensions. In the transition from Latin to Romance, this plural ending persisted for a few centuries in several Italo-Romance varieties (e.g. as -ora in Old Italian), but it gradually disappeared (see Aebischer 1933 andGardani 2013). In southern Italy this plural formative is still present and, with respect to Sicily, Rohlfs (1968: § 370) mentioned the province of Caltanissetta, in central Sicily, as an area where -ura plurals were much more frequent.
The type -ORA also continues in Romanian as -uri, which regularly marks the plural of a specific IC. Nouns with this ending in the plural have been associated with a neuter gender and to ICs that displays a distinct agreement patterncoinciding with the masculine in the singular and with the feminine in the plural (see, e.g., Corbett 1991: 151;Maiden 2016, Loporcaro 2018. With regard to the -ORA type plurals, however, the Sicilian and Romanian facts are different in two major respects. First, as already mentioned, the Sicilian -ura plurals show the same gender agreement pattern as other ICs (cf. 4b). Second, in Sicilianand in southern Italy more generallythe -ORA types, even if to a certain extent still present, are recessive to the point of extinction, while they are robustly productive in Romanian. Moreover, as pointed out by Sornicola (2010: 548), in Romanian the suffix -uri regularly marks the inflectional paradigms of the plural of certain lexical classes that can be identified on semantic grounds (see also Maiden 2016). In the southern Italian dialects, however, the inflection in -ora, even if frequent, is not regularly associated to any specific lexical properties. In other words, Sicilian -ura plurals do not share any formal or semantic properties. 6 As for the Sicilian -ura suffix, I maintain that it is specific of an independent IC, that is IC 6 in Table 1. Elements belonging to this class, however, are rather unstable and subject to variation: several lexemes allow for a second or even a third option, alternating between the selection of the -ura suffix and of other plural endings (cf. §4). Nevertheless, insofar as these lexemes select the same set of inflectional realizations, however (cf. IC 1 and IC 2 in Table 1), Sicilian nouns present a lower number of distinct sets of agreement patterns (cf. (2)-(4) above) and, thus, offer no evidence for the identification of a third gender in relation to -a and -ura plurals. 5 It is important to remark that the plural ending of these Latin neuter nouns from the third declension was -a, and that -ora was never a plural ending in classical Latin; -or-was part of the stem in the genitive, dative, and ablative case cells of the paradigm. Only at a later stage, presumably after the loss of case distinctions in late Latin or in Romance, is -ora reanalysed as a plural ending. 6 Sornicola (2010) nevertheless identifies two properties that are common to allor almost allnouns admitting the -ura plural: they are all masculine and disyllabic in the singular. She attributes an important role to the prosodic nature of these lexemes, putting forward the hypothesis that alternation between the selection of the -ura suffix and other plural forms may depend on syntactic-prosodic factors.
presenting the same endings for the singular and for the plural, they constitute an independent IC. 7 This IC has been progressively eroded over the last centuries and is nowadays limited to a small number of members. Because of the reduced size and of the lack of productivity, the plural alternations affecting the lexemes of this class can be described as a case of IC instability and oscillation: there appear to be numerous factors governing the alternations in the IC assignment and the consequent realization of the plural, but none is phonological or morphological in nature. The three suffixes in (1) realize the same morphosyntactic value and are in competition with respect to a limited number of lexemes, indicating that this kind of variation is to a remarkably large extent a lexical matter.
With respect to this limited set of lexemes, we observe that the plural forms are neither in complementary distribution nor in free variation. Therefore, we need to resort to sociolinguistic variables such as the age of speakers. This sociolinguistic variation, in turn, reflects a change in progress which is the result of a long-standing contact situation between Sicilian and Italian, and which is leading to a partial resolution of the competition resulting in the loss of the IC 6 and the reassignment of IC 6 nouns to other ICs, namely IC 1 and/or IC 4.

The empirical investigation: Questionnaire and tasks
In order to investigate plural formation and, in particular, the distribution of the -ura plurals, in central Sicilian, I designed a questionnaire which was administered to native speakers from the village of Mussomeli. In this section, the methodology and the materials used in this empirical investigation, as well as the participants who took part in this research, are described.

Method and materials
The questionnaire consisted of 44 lexemes, which were selected on the basis of the following two criteria: (a) the attestation of the plural in -ura from the relevant literature (in particular, Sornicola 2010 andRetaro 2013), and (b) personally collected data and consultations with elderly speakers. Table 2 reports the 44 lexemes used in the questionnaire. 8 Some observations about these lexemes and their -ura plural are in order.
All nouns in the list are masculine. Moreover, they are mostly nouns that are disyllabic in the singular: the only exceptions are jinùacchiu 'knee', tabbutu 'coffin', viddricu 'navel', trappitu 'olive press' for which, even if the -ura plural is attested, it turned out that the preferred plural forms end in -a. These properties are in line with what has been observed by Sornicola (2010): only masculine and disyllabic nouns allow the -ura plural (cf. fn. 6). It should also be noted that I am here using -ura in a somewhat abstract way: in fact, in the dialect of Mussomeli the suffix -ura is more often realized as -ira or -ara (sometimes pronounced as [əra]), but I take these to be phonetic variants.
As reflected by the English translations in Table 2, some lexemes are polysemous (cf. cùarpu, furnu, ponti, puntu). The different meanings were factored in in the questionnaire, but such lexical differences did not show any effects on plural formation.

Participants and tasks
The questionnaire was administered to 45 native speakers divided into three age groups, as illustrated in Table 3. It is important to emphasize that the questionnaire was jùacura (pl.) but joca (pl.). Note that with the plural ending -ura, the second element of the diphthong is often very weakly articulated and almost imperceptible (e.g. jù(a)cura, fù(a)ssura).
administered orally: the lexemes and the alternative plural forms were read out loud by another native speaker from the same village (myself or a collaborator). This was necessary because Sicilian does not have a standard orthography and is barely used in written form: native speakers might have had trouble with a task that involved reading written words. The participants were asked to perform two tasks. In order to obtain reliable results, production data were elicited first. The first task was indeed an elicitation task, whereby speakers were asked to provide the plural form of the relevant lexemes (semispontaneously, i.e. elicited without having presented the alternatives). Since speakers may be unaware of variation in their speech, their passive competence was subsequently tested by eliciting grammaticality judgments in a rating task: speakers had to assign a point to the alternative plurals on a scale from 0 to 3, where the lowest score 0 corresponded to impossible and unacceptable forms, whereas 3 qualified alternatives that were as good as the form produced in the first task − only one answer, typically the first, was recorded for the first task.

Results
In this section, the results of the two tasks are reported. The results of the second task are further supported by a statistical analysis. As already mentioned, all data point to the significance of age to understand the distribution of the -ura plurals.

The production task
The average percentage of the plural forms produced in this task is provided in Figure 1, where the results are given per age group.
As we can see from Figure 1, only speakers from Group 3 frequently produced plurals in -ura under elicitation (group3 = 50%). This contrasts with the average scores of the other two groups, with much smaller percentages (group1 = 3% and group2 = 7%). At first sight, the score 50% for the third group may seem a rather limited result; however, it must be noted that, in this task, only the first option produced by the speaker was recordedaccepted alternative forms were only taken into account in the second task. This means that, even in the remainder 50% of the cases, the -ura plural was not invariably excluded: it was simply not given as first choice. Figure 1 also shows a striking disproportion with respect to the other plural forms, opposing the first two groups, which exhibit very similar results, to the third group (i.e. -a = 44% in group1, 43% in group2, but 22% in group3; -i = 53% in group1, 50% in group2, but 28% in group3). Recall, however, that the lexemes were selected on the basis of the -ura suffix, which explains why both -i and -a plurals received much lower scores by the speakers of Group 3. For Group 1 and Group 2, -ura plurals were not readily available, so they produced one of the two alternative forms. The results of this production task support the hypothesis that the most relevant factor for the production -ura plurals is age. Within Group 3, there was nevertheless some individual variation signalling that the correlation with age was not absolute and systematic. Let us consider, for example, the speakers from Group 3 who produced the highest and lowest number of -ura plurals: these data are illustrated in Figure 2 and Figure 3, respectively. 9 Despite this variation within Group 3, it is however undeniable that the age of the speakers is the major predictor of the distribution of -ura plurals. This conclusion is supported by the results in Figure 1 and is further confirmed by the results of the second task: the rating task.

The rating task
In Figure 4, the average (raw) scores for each plural form across age groups are reported. Recall that for this task, speakers had to express acceptability judgments rating each plural alternative with a score between 0 and 3.
If we look at the first set of columns corresponding to the average scores for -ura plurals across age groups, we immediately notice the contrast between Group 1 3% 7% 50% 44% 43% 22% 53% 50% 28% 0% 10% 20% 30% 40% 50% 60% GROUP 1 GROUP 2 GROUP 3 ura a i Figure 1: Average % of elicited plural forms across age groups 9 Below the columns in Figure 2 and Figure 3, the ID code for the speaker is reported: it consists of the speaker's initials, followed by their gender (m. or f.) and by the age. The potential impact of the speakers' gender on the results was also taken into account, but it did not prove significant.
(average score: 0.44) and Group 2 (average score: 0.86) on the one hand, and Group 3 (average score: 2.12) on the other. We also see that, despite this contrast, -ura plurals are not totally rejected by the speakers of Group 1 and Group 2, showing that they belongalbeit rather limitedlyto their passive competence. Once again, the lower scores assigned to -a and -i plurals by the speakers of the third group should not be interpreted as generally related to a more limited use of these plural formatives, but rather to the fact that the lexemes were chosen on the basis of the availability of the -ura plural  form. The higher scores of -i plural forms with respect to the alternative -a forms in all groups were also expected and will be commented on below (cf. also §4). I tested the results of the rating task through a statistical analysis. The data were modelled by means of a mixed model, where the acceptability scores were specified as a dependent variable, and the predictors were age and the plural forms: age was specified in years, hence as a continuous predictor, 10 while plural types was specified as a categorial predictor with three possible values: ura, a and i. To specify the contrasts, I used the Helmert coding, which compares each level of a categorical variable to the mean of the subsequent levels. The following contrasts were set up and analysed: The first contrast was between -ura and the mean of the other two possible forms (-a/-i); the second contrast was between -a and -i plural forms. The analysis included 5703 observations, 45 participants, and 44 items (i.e. the lexemes). 11 The maximal random structure justified by the data (assessed via model comparison) included by-participant and by-item intercepts, and by-item slope for plural form.   Figures 2 and 3), I decided to include age as a continuous variable, that is, specified in years rather than in groups. The groups used for the selection of the participants and for the results in Figure 4 which would constitute a categorical variable with three levelshave thus been neglected in the statistical analysis. 11 If we calculate 44 items by 3 levels (i.e. the plural types) by 45 participants, the expected number of observations is actually 5940. Some observations, however, were missing because some speakers did not know the meaning of the corresponding lexemes and gave no answers. This happened especially with words with obsolete referents (e.g. fìau 'fief', lemmu 'earthenware bowl', scifu 'trough').
Let us first consider the two contrasts in (5) independently of age. The test showed a main effect for contrast A, in that the difference between -ura and the other two forms (-a and -i) is statistically highly significant: independently of the age, -ura plurals received significantly lower scores than the other two plural formatives (Estimate = −2.595, Std. Error = 1.287, t value = −20.172, p<.001). The second contrast (contrast B) indicates a significantly higher acceptability of -i plurals as opposed to -a plurals: in general, and irrespective of age, speakers assigned to -i plurals higher scores than to -a plurals and this difference is, although modest in size, statistically significant (Estimate = −0.465, Std. Error = 0.221, t value = −2.102, p = .0404).
Let us now take the variable of age into consideration in our analysis. Clearly, age alone does not have any significant effects if we do not differentiate between the three types of plural forms. This amounts to saying that age does not predict per se the distribution of the acceptability scores: young and old speakers did not give different scores just because of their age. However, the interaction between age and contrast A proves fundamental: in the contrast between -ura and -a/-i, -ura plurals received significantly higher scores by older participants. In other words, the higher the age, the higher the scores assigned to -ura in comparison to the scores assigned to the other two plural types. This interaction is statistically very significant (Estimate = 0.03805, Std. Error = 0.0011, t value = 34.222, p <.001). Conversely, the interaction between age and contrast B does not have any significant effects (Estimate = −0.0014, Std. Error = 0.0012, t value = −1.130, p = .2586), showing that the differences in the scores assigned to -i and -a plurals do not correlate with age.
The interaction between age and the different plural forms is plotted in Figure 5, where the lines represent the estimated effects of age (a continuous variable) for the different plural forms; while the dots (in overlay) represent the average scores for each form aggregated by participant.
The blue line represents the effects of age on the -ura plural forms. As we can see, with the increase of the age, the scores also increase (cf. the positive Estimate = 0.03805). The green line corresponding to -i plurals stays constantly higher than the red line representing -a plurals (recall that -i plural forms scored higher than the alternative -a forms in all groups), but age does not have any significant effects on this difference.

Canonical and non-canonical overabundance
The results of the rating task have highlighted another property of the process of plural formation in central Sicily, namely, overabundance. Overabundance can be defined "as the situation in which two (or more) inflectional forms are available to realize the same cell in an inflectional paradigm" (Thornton 2019: 223). The forms must be synonymous and realize the same set of morpho-syntactic properties (see also Thornton 2011Thornton , 2013. Indeed, almost all lexemes in Table 1 allow for two or even three plural forms, oscillating between IC 6, IC 1 and IC 4. Only in very few cases is the IC assignment semantically motivated. As already known from the literature (Rohlfs 1968: §370, §384;Sornicola 2010), IC 4 and IC 5 with plural in -a tend to be associated with collective semantics; indeed, in a few cases our speakers attributed to -a plurals a semantic value of indistinct or generic plurality, possibly comparable to a collective meaning (e.g. tanti tabbuta 'many coffins' vs tri tabbuti 'three coffins'). These semantic distinctions, however, were rather sporadic and, as we saw in the previous section, age turned out to be the most significant predictor for the distribution of the plural types.
In the majority of the cases, different scores were assigned to the different plural types of the same lexeme, signalling a situation of non-canonical overabundance. Thornton (2011) observes that, for overabundance to be canonical, the two or more forms that compete to realize the same cell should exhibit the same frequency. In her study of Italian verb morphology, Thornton considered frequency in corpora as the relevant criterion to define canonicity, proposing that frequency relations that are progressively more distant than an ideal 1:1 ratio should be seen as less and less canonical (see also Thornton 2019 for the discussion of this and other criteria).
If we take the acceptability scores of our rating task as the main criterion for the assessment of overabundance canonicity, we have to conclude that we are mostly dealing with non-canonical overabundance, given that speakers assigned different scores to the different plural forms of the same lexemes. There are, however, a few cases of canonical overabundance, although limited to individual variation, that is, to a small percentage of the participants. Indeed, with a handful of lexemes all three plural forms received the highest score. These lexemes are nidu 'nest' (nidura, nida, nidi 'nests'), jùacu 'game' ( jù(a)cura, joca, jùachi 'games'), fùassu 'hole' ( fù(a)ssura, fossa, fùassi 'holes'), furnu 'oven/bakery' ( furnura, furna, furni 'ovens/bakeries'), spicchiu 'clove/slice' (spicchiura, spicchia, spicchi 'cloves/slices'). The percentage of speakers who assigned the score 3 to all three plural forms of these lexemes are illustrated in Figure 6: Most of the speakers expressing full acceptability for all three plurals belonged to Group 3, but a few speakers were from Group 2. Building on these findings, we can conclude that nominal plurals in central Sicily offer an interesting situation of overabundance, where its canonicityin this case measurable through the acceptability of the formsgreatly depends on individual variation.
The fact that canonical overabundance is mostly related to the older speakers from Group 3 is somehow expected, since these are the only speakers who actively produce -ura plurals. However, how can we explain the fact that the same speakers also accept the other two plural forms as equally good alternatives? In addressing this question, I would like to suggest that frequency plays a role here. The lexemes in Figure 6 are rather common words that can be expected to be used on a daily basis irrespective of age, both in the singular and in the plural. This means that older speakers are easilyand presumably frequentlyexposed to the new plural formations in -i and in -a, which they now fully accept and might even use, probably under the influence of younger generations. This explanation predicts a different scenario for those lexemes that are obsolete or that, even if common in the singular, are infrequent in the plural: the 20% 16% 13% 13% 13% 0% 5% 10% 15% 20% 25% nidu jùacu fùassu furnu spicchiu Speaker percentage Figure 6: Percentage of speakers who gave 3 to all three plural forms of given lexemes  Table 4. First of all, it must be noted that the -ura plurals that received the highest percentage in the production task (Task 1) and the highest scores in the rating task (Task 2) by Group 3 speakers correspond to lexemes that are generally infrequent, either because they are related to objects pertaining to the old life and activities (e.g. fìau 'fief', scifu 'trough') or because they are mainly used in the singular (e.g. sciumi 'river', vinu 'wine', puzzu 'well'). 12 Secondly, in Task 2 the alternative plurals of the lexemes in Table 4 received much lower scores than the other lexemes (cf. Figure 4). This is particularly evident for those lexemes that are (more) rarely used in the plural (e.g. vinu 'wine'), and for those that are no longer in use such as fìau 'fief' and scifu 'trough'. Indeed, several younger speakers did not know the meaning of the latter lexemes. In this case, we can assume that no influence from speakers belonging to younger generations has been exerted.

Inflectional classes in competition and the role of language contact
The results of our study confirm the situation of inter-and intra-speaker variation observed in Retaro (2013), which is also characterized by a great degree of overabundance. Despite this variation, age turned out to be the best predictor for the distribution of the -ura plurals in the dialect of Mussomeli. Only the older speakers from Group 3 produced plurals in -ura under elicitation in Task 1 (average scores: group1 = 3%; group2 = 7%; group3 = 50%), while speakers from Group 1 and Group 2 only have a limited passive competence of -ura plurals that emerged in the rating task and that appears to depend on extra-linguistic factors connected to the closeness to the speech of the elderly people (e.g. grandchildren to grandparents), as acknowledged by the speakers themselves when asked about the use of these plural forms.
We observe that the IC 6, which is characterized by the -ura plural suffix, is gradually disappearing from the speech of the younger generations, and its nouns are being reassigned to an alternative IC with singular in -u (either IC 1 or IC 4, or to both ICs in some cases). We are thus clearly dealing with a situation of ICs in competition, which will presumably result in the complete loss of IC 6. In this process of reassignment, the plural ending -i (IC 1) and -a (IC 4) seem to compete on equal terms. 13 In general, the -i plurals seem to prevail in terms of productivity: they are very often provided in the elicited answers (Task 1), especially by speakers in Group 1 and Group 2, and were generally accepted as good alternatives by all speakers. More importantly, the distribution of these two plural types does not correlate with age. Even though in several cases the alternative 12 A form of feudal system or latifundium was present in Sicily until the 1950s. Note also that the plural of sciumi 'river' is not per se infrequent, but there are very few rivers in Sicily and even less in central Sicily, possibly explaining the infrequency of the plural. 13 It is important to note that the competition between -i and -a exists beyond the set of lexemes considered in this study, with -i undermining the plurals forms that typically ended in -a. Under the influence of Italian, the ending -i is gaining ground (e.g. u libbru, i libbra but also i libbri 'the book, the books', see Ruffino 2001).
-i plurals (with the same lexemes) were sometimes preferred or rated with slightly higher scores, the -a plurals are still much alive in all speaker groups. With a few lexemes, -a plurals even consistently obtain higher scores (e.g. ligna 'wood', ossa 'bones', gruppa 'knots', jinocchia 'knees', trappita 'olive presses', tabbuta 'coffins'). Recall once again, moreover, that the lexemes were selected on the basis of the fact that they have been recorded as displaying the plural in -ura. In fact, -a is much more productive with other (groups of) lexemes, as shown by Sornicola (2010). The differences in the production and acceptability of the three plural types can be used as diagnostics of a change that is currently taking place and that is leading to the gradual disappearance of the IC containing -ura forms. As already pointed out, the -i plural form received slightly higher scores, but the -a plurals are nevertheless still frequent and healthy. Our study thus provides no ultimate evidence for a possible resolution, prevalence or partial 'victory' of one IC over the other in this competition. How can we account for this phenomenon of language change? I propose that this change has been triggered by language contact and, more precisely, by the influence of Italian. The inflectional inventory underlying plural formation in central Sicilian has been reduced over time: bilingual speakers have decreasingly used the IC that does not find an equivalent in Italian, that is, IC 6. 14 In Italian, by contrast, the ending -i is the most frequent plural-formation suffix for masculine nouns. So, an increase in productivity of the same suffix in Sicilian is expected as a phenomenon of contact-induced change.
This account applies to IC 6, but how do we explain the survival and resistance of IC 4 and IC 5 with -a plural forms in Sicilian? I argue that this is also due to language contact: Italian has an IC of nouns with plural in -a; despite that fact that this IC is rather moderate in size (see Acquaviva 2008: Ch. 5; Thornton 2013; Loporcaro 2018: §4.2, §7.1), it is apparently sufficient to act as a stabilizing factor maintaining an already existing feature in Sicilian and preventing changes.
The study of the effects of language contact on morphology have mainly focused on phenomena of morphological borrowing (see Gardani 2008Gardani , 2020aGardani , 2020bJohanson & Robbeets 2012;Vanhove et al. 2012;Gardani et al. 2015; see also Matras 2009). Interestingly, the Sicilian case investigated here does not involve a direct borrowing from one language to another, but shows that contact can also have other types of effects on morphological systems and paradigms: the demise (-ura), the increase in productivity (-i), and the reinforcement or stabilization of a plural suffix (-a) and of its corresponding IC.
A diachronic investigation of the -ura plurals and of the competition with the other ICs would certainly cast important light on the situation of overabundance and on the pace of the contact-induced change. Such an investigation lies outside the scope of this paper; however, I would like to close this section showing that the morphological competition yielding overabundance was already present in the Sicilian stories collected by Giuseppe Pitrè towards the end of the 19th century (see Pitrè [1875Pitrè [ ] 1985). The -ura plurals are found, along with the alternative forms, not only within the same stories, which were narrated by Sicilian native speakers from different areas of the island, but also in the footnotes and in the glossary added by Pitrè. Here are some examples from the glossary: (6) a. focu / fochi, focura, m., 'fire / fires' b. jocu / jochi, jocura, m., 'game / games' c. locu / lochi, locura, m., 'place / places' d. sangu / sangura, m., 'blood / bloods' e. sonu / soni, sonura, m., 'sound / sounds' f. tempu / tempi, m., 'time / times' ∥ tempura, plur. of tempu g. vinu / vini, vinura, m., 'wine / wines' h. voscu / vòscura, voschi, vosca, m., 'wood / woods' Interestingly, some of these -ura plurals have disappeared and fallen in disuse in central Sicilyor at least I was not able to find any attestations of themsuch as locura 'places', sangura 'bloods', sonura 'sounds' and tempura 'times'. From their use in the stories, it seems that some were already limited to fixed expressions, for example cuitàrisi li sangura (but also lu sangu in the singular) 'to calm down', and li quattru tempura 'the ember days' (quatuor tempora in Latin). Other -ura plurals occurring in the stories, but not listed in the glossary, include jardinura 'gardens', corpura 'blows', burgiura '(hay) bales', marmura 'marbles' (although both marmu and marmuru are attested in the singular).

Conclusions and final remarks
In the domain of plural formation, central Sicilian displays an intricate case of readjustment of ICs. We observe that competing plural forms appear not to be distributed in a systematic way and that the degree of variation is high. In order to explain these facts, we need to resort to sociolinguistic factors: as a matter of fact, the use of the -ura plural differs within the same community according to age. The different productivity and acceptability of the plural forms are indicative of a slow and gradual change that is leading to the disappearance of the -ura IC. The competition between the plural forms associated with the other two ICs does not correlate with age, although, in general, the -i suffix received slightly higher scores than the -a plurals, at least with respect to the set of lexemes considered in this study. Interestingly, the gradual loss of -ura plurals is not being followed by a systematic reassignment to a specific IC: either is used to replace -ura. In fact, the replacing plural types -i and -a co-exist in several cases as inflectional doublets for the same lexemes, and almost always without explanation of what the difference between the variant forms might be. The lack of a one-way reassignment (or merger) of the -ura plurals to a single IC further seems to confirm the hypothesis that we are dealing neither with an independent (neuter) gender nor with a vigorous IC (for a case of systematic reassignment, see Loporcaro & Pedrazzoli 2016).
In conclusion, the -ura plurals may be said to constitute a special case of Sicilian nominal morphology: they are 'lexical plurals' lying outside the normal and productive rules of plural formation for Sicilian. In other parts of Sicily or southern Italy, the -ura IC is vanishing. The situation observable in central Sicily, in which a handful of remnants of the -ura type persist in the speech of the elderly, compels an analysis which qualifies them as crystallized plural forms that are no longer productive. This analysis may be reminiscent of Acquaviva's (2002: Ch.5, 2009) treatment of -a plurals in Italian. However, while Italian -a plurals display formal, syntactic, and semantic characteristics which identify them as distinct lexical items that are inherently plural (see also Ojeda 1995), 15 Sicilian -ura plurals share neither formal nor semantic properties.
The findings and the results of this study, thus, attest the present situation with respect to the -ura IC, its current morphological and lexical status, and the ongoing competition with other plural formatives. This paper has brought up the varying and multifaceted effects of language contact in the nominal morphology. Given the genealogical proximity between the two languages, Italian and Sicilian, not much scholarly effort has been previously devoted to a full comparison or to a contrastive analysis; however, the influence of Italian on Sicilianalso at the level of morphologybecomes clear when one scrutinizes the primary data more carefully and especially when one takes into account the consequences and the directions of language change under contact.