Unstable Boundaries: Phonological Change and Morphosyntactic Ambiguity in Contemporary Sardinian

Abstract

This paper investigates ongoing phonological changes in Campidanese Sardinian and its morphosyntactic repercussions, focusing on the weakening of word-final codas under increasing pressure from Italian. Sardinian preserves the Latin nominal and verbal endings -s and -t, whose interaction with external sandhi processes traditionally sustains crucial distinctions of person and number. Through a comparison between a conservative variety (Tertenia) and an innovative one (Pula), the study shows that total regressive assimilation of -s and -t is becoming generalized in innovative areas, neutralizing the contrast between second- and third-person singulars. The decline of vowel epenthesis in contexts involving heterosyllabic clusters further destabilizes the system, occasionally generating ambiguity in clitic number and verbal person marking. A Strict CV analysis demonstrates that epenthesis-less outputs are structurally well-formed only if final codas are assumed to be absent at the underlying level, pointing to a deeper restructuring of phonological representations. Overall, the data document a shift from a morphologically transparent system toward one increasingly aligned with Italian phonotactics, with significant consequences for morphosyntactic disambiguation.

1. Introduction

Minority languages, typically under pressure from a dominant language, may undergo a restructuring of their phonological system, with direct repercussions at the morphosyntactic level. Sardinian varieties provide ideal case studies in this respect. They exhibit a conservative morphological profile, notably preserving the Latin morpheme -s, which functions both as a plural marker (e.g., cani ‘dog’ vs. canis ‘dogs’) and as the second person singular verbal ending (e.g., cántas ‘(you) sing’). Within the Romance continuum, these varieties stand out—along with only those of the Lausberg area (Rohlfs, 1966, pp. 434–435)—for retaining reflexes of Latin -t as the third person singular verbal ending (e.g., cántat ‘(s/he) sings’). The persistence of these final codas gives rise to a wide range of partial and total assimilation processes and, in specific structural contexts, to the insertion of a vocalic epenthesis. The different realizations are conditioned by the phonological nature of the segment or cluster involved and by the particular Sardinian variety in question. The presence (or absence) of specific external-sandhi phenomena is crucial for identifying person (e.g., second-person singular vs. third-person singular) or number, in the case of clitics (e.g., ddas vs. dda). Under increasing influence from Italian, however, some Sardinian varieties are progressively abandoning these phonological features in favor of a system more closely aligned with that of the majority language. The gradual erosion of sandhi processes directly undermines the phonetic environments that have long supported the language’s distinctive and conservative morphological patterns.

The varieties examined in this study are spoken in south-central Sardinia and belong to the Campidanese group of Sardinian varieties.1 Within the Campidanese domain, it is well known that some areas are more conservative than others. In particular, the varieties spoken in the Ogliastra region are more conservative than the Campidanese varieties spoken further south (Virdis, 1978, p. 47) and are considered to represent an earlier stage relative to the more southern Campidanese varieties (Virdis, 2018, p. 18).2 To facilitate the discussion, we will systematically compare the behaviour of a conservative variety (Tertenia, in the central-eastern part of the island, in the Ogliastra region) with that of an innovative variety (Pula, at the southernmost tip of the island, in the Sulcis region) in order to highlight the divergent traits of the more innovative varieties.

In what follows, we will examine the sandhi phenomena affecting final codas and focus on the morphosyntactic implications of the failure to apply the phonological rules that govern these sandhi processes.

2. External Sandhi Phenomena Affecting Morphemes

Campidanese Sardinian displays a wide range of external sandhi phenomena (Contini, 1986; Lai, 2022b): various types of weakening (voicing, spirantization, elision), total and partial assimilations, fortition processes, and vowel insertions, each of which shows considerable diatopic variation. In what follows, we will focus exclusively on the two types of sandhi phenomena whose failure to apply also has repercussions at the morphosyntactic level.

2.1. Assimilations

The first phenomenon examined here concerns the range of assimilation processes attested in Campidanese varieties involving -s as a plural marker and the verbal endings -s and -t. The assimilations affecting these word-final codas are consistently regressive and may be either partial or total.

With respect to the plural morpheme -s, it is well established that in Campidanese varieties, when it is followed by the voiceless stops /p, t, k/, the final coda is preserved, e.g., /is palas/ → [is ˈpalaza] ‘the shoulders’, /is teulas/ → [is ˈteulas] ‘the roof tiles’, /duas kambas/ → [duas ˈkambaza] ‘two legs’.3 When the sibilant is followed by a voiced consonant, it undergoes either voicing or rhotacism, depending on the area, e.g., /is bɔis/ → [iz ˈβɔizi] or [ir ˈβɔizi] ‘the oxen’.4

Increasingly widespread, however, is the solution whereby the plural marker assimilates totally to the following consonant, with the exception of /p, t, k/. Focusing on voiced stops, map 5 in Cossu (2013) shows that realizations with total assimilation are now predominant across the Campidanese domain, e.g., [i ˈbːɔizi] ‘the oxen’ (Cossu, 2013, p. 83).5 Total assimilation is also highly advanced in the case of the voiceless labiodental fricative (Contini, 1986, p. 533; Cossu, 2013, p. 94, map 22). Similarly, with laterals and nasals, most Campidanese varieties—excluding more conservative ones such as those of Ogliastra—now tend to display total assimilation of the plural -s, e.g., [i ˈlːɔɣuzu] ‘the places’, [i ˈmːanuzu] ‘the hands’, [i nːeˈβɔðizi] ‘the grandchildren’ (Wagner, 1941/1984, p. 302; Contini, 1986, pp. 531–533).

As for -s as a verbal ending in Campidanese varieties, the main studies addressing this issue are Pisano (2010), Molinu and Pisano (2016), and Lai (2022b). Lai (2022b, p. 99) reports data from the conservative variety spoken in Tertenia, whereas Pisano (2010) and Molinu and Pisano (2016) focus on a more innovative Campidanese variety, that spoken in Pula.6 For the sake of clarity, we first examine the behavior of the conservative varieties described in Lai (2022b), supplemented with new data from personal fieldwork, and then turn to the innovative variety of Pula on the basis of Pisano (2010).

In the Tertenia variety, the -s coda never undergoes total assimilation, except when followed by a voiceless palato-alveolar affricate, for obvious articulatory reasons.7 In all other contexts, the coda is always realized either as [s] or as [r]. More specifically, before [-voice] segments it surfaces as [s] (data in (1)a–e), whereas before [+voice] segments, /s/ undergoes rhotacism (data in (2), (3), and (4)).8

(1)
a.	/papas pani/ [ˈpapːas ˈpani] ‘(you) eat bread’
b.	/papas tamata/ [ˈpapːas taˈmːatːa] ‘(you) eat tomatoes’
c.	/papas kasu/ [ˈpapːas ˈkazu] ‘(you) eat cheese’
d.	/papas fiɡu/ [ˈpapːas ˈfiɣu] ‘(you) eat figs’
e	/papas sali/ [ˈpapːa ˈsːali] ‘(you) eat salt’
f.	/papas ʧeresia/ [ˈpapːa ʃːeˈrezia] ‘(you) eat cherries’

(2)
a.	/tɛnis baska mɛda/ [ˈtɛnir ˈbaska ˈmːɛða] ‘(you) are very hot’ (lit. ‘(you) have much heat’)
b.	/tɛnis disiʎu/ [ˈtɛnir diˈziʎːu] ‘(you) feel longing/nostalgia (for something)’
c.	/tɛnis ɡana mala/ [ˈtɛnir ˈɡana ˈmːala] ‘(you) feel nauseous’ (lit. ‘(you) have bad desire’)

(3)
a.	/tɛnis bakas/ [ˈtɛnir ˈβakːaza] ‘(you) have/own cows’
b.	/tɛnis dinari mɛda/ [ˈtɛnir ðiˈnari ˈmːɛða] ‘(you) have a lot of money’
c.	/pɔrtas ɡanɡas fortis/ [ˈpɔrtar ˈɣaŋɡas ˈfɔrtizi] ‘(you) have a strong voice’ (lit. ‘(you) carry a strong voice’)
d.	/tɛnis ʤenuɡus sanus/ [ˈtɛnir ʤeˈnuɣus ˈsanuzu] ‘your knees are fine’ (lit. ‘(you) have healthy knees’)

(4)
a.	/papas latua/ [ˈpapːar ˈlatːua] ‘(you) eat lettuce’
b.	/papas mendula/ [ˈpapːar ˈmendula] ‘(you) eat almonds’
c.	/papas nuʤi/ [ˈpapːar ˈnuʤːi] ‘(you) eat walnuts’

The data reported in Pisano (2010, pp. 184–185) clearly show that the situation in Pula diverges substantially from that of the conservative varieties. When followed by any consonant, and regardless of the voicing of the segment involved, only total assimilation is attested. For example: /papas pani/ → [ˈpapːa ˈpːani] ‘(you) eat bread’, /kastias totu/ → [ˈkastia ˈtːotːu] ‘(you) look at everything’, /papas kasu/ → [ˈpapːa ˈkːazu] ‘(you) eat cheese’, /tɛnis bɔis/ → [ˈtɛni ˈbːɔizi] ‘(you) have/own oxen’, /fais ʤentilːa/ → [fai ʤːenˈtilːa] ‘(you) make/cook lentils’, /fais fɛsta/ → [fai fːɛsta] ‘(you) celebrate/have a party’, /papas mɛla/ → [ˈpapːa ˈmːɛla] ‘(you) eat apples’.9

Let us now turn to the final coda -t (the third person singular verbal ending), which systematically undergoes total assimilation in all Sardinian varieties, irrespective of the type of following consonant. By convention, the phoneme /t/ is retained as the underlying representation of the verbal ending, although this assumption strictly holds only for Nuorese varieties, which have not undergone diachronic lenition. Let us therefore consider some examples illustrating the treatment of -t in the more conservative Campidanese varieties:

(5)
a.	/papat pani/ [ˈpapːa ˈpːani] ‘(s/he) eats bread’
b.	/papat tamata/ [ˈpapːa tːaˈmːatːa] ‘(s/he) eats tomatoes’
c.	/papat kasu/ [ˈpapːa ˈkːazu] ‘(s/he) eats cheese’
d.	/papat fiɡu/ [ˈpapːa ˈfːiɣu] ‘(s/he) eats figs’
e	/papat sali/ [ˈpapːa ˈsːali] ‘(s/he) eats salt’
f.	/papat ʧeresia/ [ˈpapːa ʧːeˈrezia] ‘(s/he) eats cherries’

(6)
a.	/tɛnit baska mɛda/ [ˈtɛni ˈbːaska ˈmːɛða] ‘(lit. ‘(s/he) has much heat’)
b.	/tɛnit disiʎu/ [ˈtɛni dːiˈziʎːu] ‘(s/he) feels longing/nostalgia (for something)’
c.	/tɛnit ɡana mala/ [ˈtɛni ˈɡːana ˈmːala] ‘(s/he) feels nauseous’

(7)
a.	/tɛnit bakas/ [ˈtɛni ˈβakːaza] ‘(s/he) has/owns cows’
b.	/tɛnit dinari mɛda/ [ˈtɛni ðiˈnari ˈmːɛða] ‘(s/he) has a lot of money’
c.	/pɔrtat ɡanɡas fortis/ [ˈpɔrta ˈɣaŋɡas ˈfɔrtizi] ‘(s/he) has a strong voice’ (lit. ‘(s/he) carries a strong voice’)
d.	/tɛnit ʤenuɡus sanus/ [ˈtɛni ʤːeˈnuɣus ˈsanuzu] ‘his/her knees are fine’ (lit. ‘(s/he) has healthy knees’)

(8)
a.	/papat latua/ [ˈpapːa ˈlːatːua] ‘(s/he) eats lettuce’
b.	/papat mendula/ [ˈpapːa ˈmːendula] ‘(s/he) eats almonds’
c.	/papat nuʤi/ [ˈpapːa ˈnːuʤːi] ‘(s/he) eats walnuts’

As can be seen from the data, -t is never overtly realized.10 The third-person singular morpheme is identifiable only through gemination of the initial segment of the following word.

The situation is analogous in innovative varieties such as Pula. The data reported in Pisano (2010, pp. 184–185) and partially reproduced in Molinu and Pisano (2016) show total assimilation before all consonants: /papat pani/ → [ˈpapːa ˈpːani] ‘(s/he) eats bread’, /kastiat totu/ → [ˈkastia ˈtːotːu] ‘(s/he) looks at everything’, /papat kasu/ → [ˈpapːa ˈkːazu] ‘(s/he) eats cheese’, /tɛnit bɔis/ → [ˈtɛni ˈbːɔizi] ‘(s/he) has/owns oxen’, /fait ʤentilːa/ → [fai ʤːenˈtilːa] ‘(s/he) makes/cooks lentils’, /fait fɛsta/ → [fai fːɛsta] ‘(s/he) celebrates/has a party’, /papat mɛla/ → [ˈpapːa ˈmːɛla] ‘(s/he) eats apples’.11

In short, in the more innovative varieties, -s and -t assimilate to the following consonant in exactly the same way (Pisano, 2010, p. 184). While conservative varieties clearly distinguish the second person singular from the third person singular, this distinction has been neutralized in innovative varieties: total assimilation now affects both verbal endings.

2.2. Vowel Insertion

A distinctive feature of Campidanese Sardinian varieties—absent in Logudorese—is their tolerance of heterosyllabic clusters in word-initial position, specifically sequences of /s/ + consonant (sC clusters) and initial geminates (Lai, 2015b). In external sandhi contexts, the co-occurrence of such heterosyllabic clusters with the final codas discussed above gives rise to structurally ill-formed configurations. Campidanese varieties resolve these configurations through vowel insertion, as illustrated by forms with initial sC clusters such as /is spɔnʤas/ → [iz [i] sˈpɔnʤaza] ‘the sponges’, /is stɔɡumus/ → [iz [i] sˈtɔɣumːuzu] ‘the stomachs’, /duas skalas/ → [ˈduaz [a] sˈkalaza] ‘two stairs’, or by forms with initial geminates such as /is tʧinemːas/ → [iz [i] ˈʧːinemːaza] ‘the cinemas’ (Lai, 2015b, 2022a, p. 613, 2022b, pp. 51–70).12

It should be noted that this type of phonological repair is not limited to the plural marker, as in the examples above, but occurs whenever a heterosyllabic cluster is preceded by a word-final coda, including verbal endings. For example: /sɛs stetiu/ → [sɛz [i] sˈtetiu] ‘(you) have been’, /iat skritu/ → [ˈiað [a] sˈkritu] ‘(s/he) had written’, /depeis ʃʃiri/ → [deˈpːeiz [i] ˈʃːiri] ‘(you) (pl.) must know’. The phenomenon is fully productive, and if the epenthetic vowel is not inserted, the resulting string is ungrammatical: *[is sˈpɔnʤaza], *[is sˈtɔɣumːuzu], *[is sˈkɔlaza], *[is ˈʧːinemːaza].13

Nevertheless, in recent years one increasingly hears realizations in which the epenthetic vowel is not inserted, even though the structural configuration—namely, a final coda followed by a heterosyllabic cluster—would require insertion. Thus, forms such as *[is sˈkalaza] instead of [iz [i] sˈkalaza] ‘the stairs’ or *[deˈpːei ˈʃːiri] instead of [deˈpːeiz [i] ˈʃːiri] ‘(you) (pl.) must know’ are becoming more common. Productions of this type are judged ungrammatical by native speakers and point to a phonological rule that is either in regression or was never fully acquired.

However, whereas in conservative Campidanese varieties such forms are found only among less fluent speakers, typically the younger ones, in more innovative varieties, such as Pula, the variant without vowel insertion is beginning to be perceived as fully acceptable (Simone Pisano, p.c.). This development indicates that some varieties are undergoing a restructuring of their phonological system, possibly related to the fact that Sardinian is no longer the mother tongue for a number of southern speakers, whose phonological systems may therefore be more closely aligned with Italian than with Sardinian.

It should be emphasized that many productions in which the epenthetic vowel ought to be inserted but is absent do not compromise comprehension. For instance, in *[is sˈkalaza] ‘the stairs’, the number is still recoverable through the presence of the plural article is, which remains clearly identifiable, as does the plural morpheme -s on the noun, followed by the paragogic vowel [a].

Similarly, in *[deˈpːei ˈʃːiri] instead of [deˈpːeiz [i] ˈʃːiri] ‘(you) (pl.) must know’, although the epenthetic vowel separating the final coda from the initial geminate /ʃʃ/ is absent, the fact that the form corresponds to the second person plural verbal ending is still recoverable from the remaining portion of the morpheme -eis.

There are, however, cases in which the absence of vowel insertion may give rise to ambiguity with respect to person and number. As far as person is concerned, native speakers clearly distinguish the second from the third person precisely by virtue of epenthesis, which makes the verbal morpheme of the second and third person singular overt. Compare, for instance, [az [i] ʃːiˈðau a ˈmːaria] as scidau a Maria ‘(you) have woken Maria up’ vs. [að [i] ʃːiˈðau a ˈmːaria] at scidau a Maria ‘(s/he) has woken Maria up’.

However, productions in which the second- and third-person singular receive identical realizations—due precisely to the absence of vowel insertion and concomitant total assimilation in both cases—are becoming increasingly widespread. Sentences such as [ˈparið [i] streˈkːendi ˈɣɔza] parit strechendi cosa ‘(s/he) seems to be squashing something’ vs. [ˈpariz [i] streˈkːendi ˈɣɔza] paris strechendi cosa ‘(you) seem to be squashing something’ are now more and more frequently realized without epenthesis and without the final coda: [ˈpari streˈkːendi ˈɣɔza]. This neutralization makes it impossible to distinguish between second and third person and forces speakers to overtly express the subject in a language that is otherwise pro-drop.

With regard to number, a now common case among speakers with more limited exposure to Sardinian concerns the pronunciation of plural clitics followed by heterosyllabic clusters. It frequently occurs that the third person plural accusative proclitics ddas and ddus, as well as the third person plural dative proclitic ddis, fail to display epenthetic vowel insertion when followed by a verb beginning with an sC cluster or an initial geminate.

If we take a verb beginning with a heterosyllabic cluster, such as scavulai ‘to leave alone’, and precede it with one of the plural clitics mentioned above, a native speaker would produce the following realization: [ɖːaz [a] sˈkavːulaza] ddas scavulas ‘(you) leave them (fem.) alone’. Among less fluent speakers, however, realizations without epenthesis are increasingly attested, as in [ɖːa sˈkavːulaza]. This form is not ungrammatical, but it no longer means ‘(you) leave them alone’; rather, it is interpreted as ‘(you) leave her alone’. In other words, the hearer understands the speaker to be using the singular clitic dda [ɖːa], not the plural ddas [ɖːas].

In cases such as these, the presence or absence of the epenthetic vowel is what distinguishes singular from plural clitics. If the phonological rule of vowel insertion fails to apply, it becomes unclear whether the clitic is singular or plural.

3. Weakening of Final Codas in Campidanese

The total assimilation of final codas and the absence of epenthesis in the presence of heterosyllabic clusters can both be attributed to a weakening of final codas. While total assimilation still allows us to assume that the original syllabic structure is preserved, the lack of epenthesis suggests that final codas have become so weakened that they may disappear structurally as well.

If we analyze the available data within a Government Phonology framework such as Strict CV and examine them in light of the Empty Category Principle, we find further evidence that final codas are gradually disappearing in certain Campidanese varieties.

Strict CV (Lowenstamm, 1996; Scheer, 2004, among others) is characterized by a radical reassessment of traditional assumptions about syllable structure and segmental organization. Within this framework, universal syllable structure is reduced to a strict alternation of consonantal and vocalic positions. As a consequence, canonical syllable structures—and, more generally, any configuration that does not involve an open syllable—are reanalyzed as sequences of alternating C and V positions.

Compared to Standard Government Phonology (Charette, 1990; Harris, 1990; Kaye et al., 1985, 1990), phonological representations shift from branching to flat structures, as illustrated in Figure 1.

One consequence of this theoretical approach is the proliferation of empty vocalic positions: consonant clusters and geminates are separated by an empty nucleus, and codas (both internal and final) are likewise followed by an empty vocalic position.

To regulate the distribution of empty nuclei, the Empty Category Principle (ECP) has been adopted since the earliest formulations of Standard Government Phonology. The ECP states that a position may remain phonetically unrealized if it is properly governed (Kaye et al., 1990, p. 220). In Strict CV, this principle is reformulated as follows: an empty nucleus may remain unexpressed if it is governed or if it occurs in word-final position (Scheer, 2004, p. 67). Only a phonetically realized nucleus can govern an empty one; consequently, a sequence of two empty vocalic positions is excluded for reasons of government. An empty vocalic position must therefore be adjacent to a realized vocalic position, ensuring that the alternation involves, at a minimum, one filled nucleus and one empty nucleus.

In light of the above, within Strict CV, a word-final coda followed by a heterosyllabic cluster would receive the representation shown in (2), illustrated here with the example /is skalas/ → [iz [i] sˈkalaza] ‘the stairs’. The representation of the sC cluster in Strict CV reproduces the traditional branching coda–onset configuration (Kaye, 1992, p. 13).

As already noted, however, realizations without epenthesis are becoming increasingly frequent. The example represented in Figure 2 is now produced by a number of speakers as *[is sˈkalaza], a form that older speakers—especially those from more conservative areas—would judge ungrammatical.14

If we attempt to render this epenthesis-less realization within the Strict CV framework, we obtain the representation shown in Figure 3:

Figure 3 displays structural well-formedness issues due to the presence of adjacent empty vocalic positions: the nucleus in V₃ of the sC cluster and the preceding nucleus in V₂. This illicit configuration arises because the word-final coda contains an empty nucleus and precedes a heterosyllabic cluster, which in turn contains a second empty nucleus. The nucleus in V₃ is properly governed by the filled nucleus in V₄, but the nucleus associated with the final coda (V₂) cannot be governed in any way, since there is no vocalic position available to govern it. For this reason, native speakers repair this dysfunctional configuration by inserting an epenthetic vowel in V₂ (see Figure 2).

It is also possible, however, that for some speakers these realizations are not merely the result of a failure to apply the epenthesis rule, as in Figure 3, but rather reflect an underlying phonological form in which the coda is no longer present—at least in phonosyntactic contexts—as represented in Figure 4.

The representation in Figure 4 does not display any structural well-formedness issues: the only empty nucleus, namely that associated with the cluster, is properly governed. Once the word-final coda is lost at the phonological level in phonosyntactic contexts, the empty nucleus that the coda carried with it (the nucleus in V₂)—which in Figure 2 could not be governed—is no longer present.15

The well-formedness of the structure in Figure 4 is not the only element suggesting that final codas in phonosyntax have become so weakened that they may no longer be present, even at the underlying level, in the grammar of many speakers.

If we return to the examples concerning the behavior of the sibilant in phonosyntactic contexts in the more innovative areas (cf. Section 2.1), it is worth recalling that the surface realization of the sibilant has become largely residual. It is therefore evident that, at least in some areas and among speakers with more limited competence in Sardinian, final codas—especially in phonosyntactic contexts—are in regression.

The most emblematic case, however, concerns the phonological behavior of new speakers of Sardinian who, due to the interruption of intergenerational transmission, are likely to constitute the largest group of Sardinian speakers in the future (Lai, 2025; Mura, 2025; Pisano, 2025). Among these speakers, final codas tend to be destabilized not only in phonosyntax but also in absolute word-final position. Indeed, one well-documented feature of their production is the tendency not to apply the paragogic vowel when the coda is not followed by any additional segment (Lai, 2025, p. 161). For example, a word such as froris ‘flowers’ is typically pronounced by new speakers as [ˈfrɔris] instead of [ˈfrɔrizi], resulting in a realization that is not licit in Sardinian phonology. By contrast, among native speakers, the rule of paragogic vowel insertion contributes to the preservation of the coda.

4. Conclusions

In light of the data discussed above, it emerges that in the more innovative Campidanese varieties, a systematic process of weakening of word-final codas is underway, affecting the phonological level and having direct repercussions on morphosyntax. The generalized total assimilation of the morphemes -s and -t in phonosyntactic contexts and the regression of epenthesis in the presence of heterosyllabic clusters indicate that final codas are no longer, for a growing segment of speakers, a structurally robust position.

The Strict CV analysis shows that epenthesis-less productions are well formed only if one assumes the disappearance of the coda at the underlying level as well, which points to a profound restructuring of the phonological grammar. This restructuring has significant morphosyntactic consequences: in contexts where distinctions of person or number depended precisely on the presence of the coda and on the sandhi processes associated with it, their erosion generates ambiguity (second vs. third person singular; singular vs. plural clitics) and tends to favor compensatory strategies, such as overt subject expression in a traditionally pro-drop language.

The overall picture fits coherently within the broader process of the language shifting toward Italian, which does not merely involve lexical or functional replacement but affects the phonological mechanisms that historically sustained the morphological identity of Sardinian. We are therefore witnessing a transition from a conservative system, grounded in the salience of final codas, to one reorganized according to phonotactic and morphosyntactic parameters increasingly aligned with those of the dominant language.