Jump to ContentJump to Main Navigation
The Oxford Guide to the Romance Languages$

Adam Ledgeway and Martin Maiden

Print publication date: 2016

Print ISBN-13: 9780199677108

Published to Oxford Scholarship Online: August 2016

DOI: 10.1093/acprof:oso/9780199677108.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2019. All Rights Reserved. An individual user may print out a PDF of a single chapter of a monograph in OSO for personal use.  Subscriber: null; date: 17 September 2019



(p.270) Chapter 17 Sardinian
The Oxford Guide to the Romance Languages

Guido Mensching

Eva-Maria Remberger

Oxford University Press

Abstract and Keywords

After a brief outline of the principal episodes in the external history and attempts at standardization, the chapter surveys the principal aspects of phonology, morphology and syntax of Sardinian and its dialects (notably Logudorese, Nuorese, and Campidanese). It covers: vowels systems; consonant systems; sandhi phenomena; stress and other suprasegmental features (including paragogic vowels); nominal and verbal inflectional morphology; word formation; the syntax of the nominal group and of the verbal group; auxiliary selection; the syntax of the clause and the basic properties of the clause; valency-changing operations; finite subordination; non-finite constructions (inlcuding finite marking on infinitives); information structure; illocutionary force.

Keywords:   Sardinian, Logudorese, Nuorese, Campidanese, phonology, morphology, syntax

17.1 Introduction

Sardinian is the autochthonous development from Latin in Sardinia following the Roman conquest during the first Punic war.1 In the first texts of the eleventh century it is already a language distinct from the dialects of Italy. The inherited lexicon shows some unique conservations (Porru 1942) such as domo ‘house’ (< DOMO) and mannu ‘big’ (< MAGNUM), as well as others common to other Romance varieties but non-existent or rare in Italy including the reflexes of SCIRE ‘to know’ (cf. Ro. a şti) and UERUEX ‘sheep’ (cf. Ro. berbec ‘ram’, but also Fr. brebis ‘ewe’). There are, however, also some striking parallels to southern Italian dialects (cf. Rohlfs 1937a, and Ch. 16).


Map 17.1 The Sardinian dialects (modified version of Virdis 1988:905)

Although Sardinian should be considered as a language in its own right (in the sense of Abstandsprache; cf. Kloss 1987), there is no common standard. Thus, the notion ‘language’ must be understood in the sense of a bundle of varieties usually classified into northern/central Logudorese and southern Campidanese varieties, with a border zone of overlapping isoglosses in the central part of the island, the western area of which is sometimes called Arborense (cf. Map 17.1). The eastern half of the Logudorese-speaking territory includes the subvariety around Nuoro, often considered a variety in its own right, termed central Sardinian. It is held to be the most conservative variety, a judgement mostly due to the lack of voicing of intervocalic stops and the conservation of Latin velars (§17.2.2). Since Arborense is not a uniform variety, sharing features of both Logudorese and Campidanese, we shall not consider it here, but rather refer to three main varieties, Nuorese, Logudorese, and Campidanese. Map 17.1 also shows two northern varieties, Sassarese and Gallurese, which have their origin in the Italian dialects imported from Pisa and Genoa from the eleventh to the fourteenth centuries (cf. Ch. 14). They will not be considered here, nor will other languages spoken in Sardinia (Catalan in Alghero and Ligurian varieties on the islands of Sant’Antioco and San Pietro).

Scholars from the nineteenth century onwards (e.g. Gröber 1884:210f.) claimed that, because of the early conquest of Sardinia, Sardinian conserved elements from archaic Latin. Yet the survey in Mensching (2004b) shows that this hypothesis cannot be maintained, since most of the relevant features (e.g. lack of palatalization of /k/ and /ɡ/ before /e/ and /i/) were still common in Latin during the first centuries AD. It thus seems that the Latin of Sardinia was not significantly different from that spoken elsewhere until at least the end of the third century AD. The striking conservative elements of Sardinian are not sufficient to classify it as archaic, since it also shows numerous innovations.

Some of the changes are due to language contact. Pre-Roman languages are mostly reflected in toponymy and some lexical items. The Paleosardinian substratum, of uncertain origin, may also be responsible for some phonological phenomena and, in part, the phonological make-up of Sardinian words (e.g. syllable structure, repetitive vowel patterns such as í-i-i; cf. Serra 1960, and §17.2.1).2 After the fall of the western Roman empire, the Vandals were present on the island only for around 80 years (455-533): the almost total lack of direct Germanic influences in Sardinian is a distinctive feature with respect to Italian and most other Romance languages.3 Sardinia belonged to the Byzantine Empire from the seventh century, but during the conflicts against the Saracens, neglect by Byzantium led to the development of independent political structures in the shape of the Judicates of Cagliari, Torres, Arborea, and Gallura. The threat posed by the Saracens was overcome at the beginning of the eleventh century with the aid of Pisa and Genoa, who gained special privileges, so that the Judicates (except Arborea) lost their autonomy. Whereas Byzantine Greek left only a few traces, the contribution of old Italian varieties, in (p.271) (p.272) particular Tuscan, is considerable in the area of vocabulary (e.g. Log. betzu, Cpd. beciu ‘old’; cf. Tsc.-It. vecchio), including some functional elements (e.g. indefinites). The most numerous superstratum elements stem from Catalan and (Castilian) Spanish, since Sardinia passed to the Kingdom of Aragon (where Catalan was official) at the beginning of the fourteenth century, becoming part of the united Spanish Kingdom from 1479. In 1718 Sardinia passed to Piedmont, and in 1861 to united Italy. Since then the political history of Sardinian has followed that of Italy, including its linguistic policies, leading to diglossia and a new wave of italianisms (cf. Rindler-Schjerve 1987).

At the end of the last millennium, a regional law on ‘promotion and support of culture and language in Sardinia’ (Regione Autonoma della Sardegna 1997) and a national law on ‘regulations for the protection of historical linguistic minorities’ (Parlamento Italiano 1999) created a legal basis for official bilingualism. From 1999 to 2001 an expert committee was established in order to elaborate a standardized orthography and grammar.4 The resulting system, the Limba Sarda Unificada (‘Unified Sardinian Language’), strongly favoured Logudorese and became a matter of highly polemical debate, leading to a second committee and a revised proposal, the Limba Sarda Comuna (LSC, ‘Common Sardinian Language’),5 but also to several counter-proposals. For this reason and without a systematic presence in the school system, official bilingualism in Sardinia is still far from being a reality.

17.2 Phonology


Map 17.2 The main isoglosses of Sardinia (modified version of Virdis 1988:908)

Phonetics and phonology is the main area of variation among the Sardinian dialects. Comprehensive studies are Wagner (1941), from a diachronic perspective, and the phonetic geolinguistic atlas of Sardinia by Contini (1987). For the Campidanian area, important sources are Virdis (1978) and Bolognesi (1998). Furthermore, Virdis (1988) describes the Sardinian-speaking area offering a geolinguistic profile (cf. Map 17.1) based on several decisive isoglosses referring mainly to phonological criteria (cf. Map 17.2).

17.2.1 Vowel system

The Classical Latin vowel system (cf. §25.1) was reduced to five vowels in the so-called ‘Sardinian vowel system’ (cf. also §, the result of quantitative neutralization without change in quality: Sardinian originally has three degrees of height with no distinction between close mid and open mid front and back vowels. The vowels e and o are usually pronounced open as [ɛ] and [ɔ], but are phonetically close in the context of metaphonic raising to [e] and [o] before a high vowel in the following syllable ([i] < Lat. I, [u] < Lat. U, but also early glides). Thus, in Logudorese, we find tempus [ˈtempus] ‘time’ < TĔMPUS vs tempos [ˈtɛmpɔs] ‘times’ < TĔMPOS and bonu [ˈbonu] ‘good.MSG’ < BŎNU(M) vs [ˈbɔnɔs] ‘good.MP’ < BŎNOS.

In Campidanese, metaphony led to the development of open mid front and back vowels now in systematic opposition with the corresponding close mid front and back vowels because of final vowel neutralization: when metaphony was no longer active, the atonic vowel system was further reduced to three vowels /i/, /u/, and /a/ by word-final vowel raising, e.g. Cpd. bonus [ˈbɔnus] ‘good.MPL’ < BŎNOS vs Log. bonos [ˈbɔnɔs] and Cpd. beni [ˈbɛni] ‘well’ < BĔNE vs Log. bene [ˈbɛnɛ] ‘well’ (cf. Map 17.2, isogloss 2). Consequently, we now have minimal pairs like beni /ˈbeni/ ‘come.IMP.2SG’ < UĔNĪ and beni /ˈbɛni/ ‘well’ < BĔNE, and ollu /ˈɔllu/ ‘oil’ < OLEUM and ollu /ˈollu/ ‘I want’ < *ˈvɔljo. Thus, /e o/ vs /ɛ ɔ/ has become part of the phonological system in these varieties.6

In contrast to many other Romance languages, there is no spontaneous diphthongization (cf. §38.1). The old Latin diphthongs were monophthongized (AE, OE > [ɛ]; AU > [a], rarely [o]). Apparent diphthongs are secondary, for instance, due to consonant deletion, e.g. Cpd. meigina ‘cure’ < MEDICĪNAM.

Table 17.1 The Sardinian vowel system: Logudorese/Nuorese


Table 17.2 The Sardinian vowel system: Campidanese


The two resulting Sardinian vowel systems are summarized in Tables 17.1 and 17.2.

Some Campidanese dialects have nasals vowels parallel to all seven oral vowels ([ĩ, ẽ, ɛ˜͂, ã, ɔ͂˜, õ, ũ]) whenever elision of a posttonic nasal consonant took place and the surrounding vowels were nasalized, e.g. LŪNA> [ˈlũã] ‘moon’ (Sanluri).7 Furthermore, in dialects of the Sarrabus area (Böhne 1950) both intervocalic /l/ and /n/ are substituted by a glottal stop,8 but the nasal leaves its trace by nasalizing the preceding vowel, yielding minimal pairs such as [ˈmaɁu] < (p.273) (p.274) MALU(M) ‘bad’ vs [ˈmãɁu] < MANU(M) ‘hand’. The phonological status of these nasal vowels is controversial.

Vowel assimilation seems also to be quite common, especially in Campidanese, e.g. [pis.ti.ˈna.ɣa] ‘carrot’ < PASTINACAM or [ˈan.ʤu.lu] ‘angel’ < ANGELUM. Also, [ɛ] in hiatus is raised to [e] or [i], e.g. MEUM > miu ‘my.MSG’, and /o/ to /u/ after a labial or before /nd/, e.g. tùndiri < TONDERE ‘shear.INF’. Present in all dialects is raising of mid vowels in pretonic position as in cuntentu < CONTENTUM ‘satisfied’ and grinucu < GENUCULUM ‘knee’.

All varieties show the so-called paragogic vowel, involving a phonological process by which the vowel of a final syllable in a word ending in a consonant is copied and attached to the final consonant giving rise to a further CV syllable, but only prepausally or at the end of an utterance. Thus we have Log. [is.ˈtran.ʣo.zo] / Cpd. [is.ˈtran.ʤu.zu] ‘strangers’ and Cpd. perdas [ˈpɛr.da.za] ‘stones’, timèis [ti.ˈme.i.zi] ‘you.PL fear’. Other (non-copied) paragogical (epenthetic) vowels are found in (mostly) monosyllabic words, cf. tui < TU ‘you.SG’, innoi < IN HOC ‘here’, cras > /ˈkra.zi/ ‘tomorrow’ and also verbal forms such as funti ‘they are’, asi ‘you.SG have’.

A phonological process found mainly in Campidanese yields prosthetic vowels before initial /r/, e.g. a in arriu ‘river’, arriri ‘laugh’ (cf. Map 17.2, isogloss 9). In Logudorese and Nuorese prosthetic vowels appear before /s/+consonant clusters, e.g. Nuo. /isˈkɔla/ vs Cpd. /ˈskɔla/ < SCHOLAM ‘school’ (cf. Map 17.2, isogloss 8).

17.2.2 Consonant system

Table 17.3 Lenition processes


The consonant system and the phonetic realizations of consonants vary greatly among the dialects. The most relevant diachronic (but also synchronic, cf. §17.2.3) phonological process is lenition of the (etymologically simple) plosives /p, t, k, b, d, ɡ/ and voiceless fricatives /f, s/. The rule is given in Table 17.3.

Thus we have the following results:



Table 17.4 Nuorese (from Jones 1988b:319)


The Nuorese area is the most conservative, since it mostly preserves intervocalic voiceless plosives.9 Both Nuorese and Logudorese also maintain velar plosives before front vowels, whereas Campidanese has palatalized them (cf. Map 17.2, isogloss 1): Nuo./Log. [ˈkentu], but Cpd. [ˈʧentu] < CENTUM ‘hundred’, Nuo./Log. [ɡiˈrare], but Cpd. [ʤiˈrai] < GYRARE ‘turn.INF’. In Nuorese these affricates only appear in loans (mostly from Italian, but also from Catalan and Spanish). The same is true of the palatal nasal /ɲ/ and the approximant /w/. Some Nuorese dialects (e.g. Bitti and Lula) distinguish (Latin) /b/ and /v/ (Wagner 1941:97; Jones 1997:318), elsewhere both appear as /b/. The interdental fricative /θ‎/, particular to Nuorese (cf. Map 17.2, isogloss 4) and absent from Logudorese and Campidanese, can have different origins (Wagner 1941:109f.), e.g. /ˈθ‎iu/ ‘uncle’ < Grk. theios, Lat. [kj] as in /ˈfaθ‎o/ < FACIO ‘I do’, and Lat. [tj] as in /ˈpuθ‎u/ < PUTEUM ‘well’ (cf. Map 17.2, isogloss 4). In some Nuorese subdialects, word-initial F- is deleted in absolute initial position (cf. Virdis 1988:908). Nuorese (with Logudorese) has an affricate /ʣ/, which mostly stems from Latin L+j, e.g. FILIUM > [ˈfiʣu] ‘son’, but is also found in (mostly Italian) loans e.g. [ˈʣɛntɛ] instead of [ˈʤente] ‘people’. The affricate /ts/ is not hereditary in most of the Nuorese territory, deriving mostly from It. /ʧ/ (OIt. cittade > [ʦitˈtaðɛ] ‘city’). The Nuorese consonant system (as in the dialect of Lula) is given in Table 17.4.

Table 17.5 Campidanese


The Logudorese system has lenition of voiceless intervocalic plosives (but [ß ð ɣ] are not phonemic), but lacks velar palatalization like Nuorese (palatals only appear in loans); Logudorese does not have /θ‎/, since the Latin contexts /kj/- and /tj/- produced other results, like /tt/ in /ˈfatto/ < FACIO ‘I do’ (cf. Map 17.2, isogloss 4a) (Cpd. /ˈfadʣu/, /ˈfatʦu/), /ˈputtu/ < PUTEUM ‘well (N)’ (Cpd. /ˈputʦu/, /ˈpuʧu/). In Campidanese (Table 17.5), Latin /k ɡ/ were palatalized to /ʧ ʤ/ and lenition took place for /p t k b d ɡ f s/ often up to complete deletion of the consonant.

(p.275) In most dialects /w/ is present only in loans, since no diphthongization took place and the Latin nexus /kw/ and /ɡw/, which contained the approximant, had a different development, namely /b/, as in QUATTUOR > battor ‘four’; in Campidanese, however, Lat. /kw/ is mostly maintained, e.g. [ˈkwatt(u)ru]. All dialects have retroflex voiced plosives which originate from Latin -LL-, e.g. UILLAM > bidda [ˈbiɖɖa] ‘village’, PULLUM > puddu [ˈpuɖɖu] ‘rooster’. Furthermore, intervocalic /l/ is subject to phonological processes not only word-internally, but also across word boundaries, giving rise to [ß w ʁ Ɂ ɡw L Ø] depending on dialect area (cf. also Frigeni 2005:21; Molinu 2009).

The status of Sardinian geminates is unclear (Bolognesi 1998:158-65; Ladd and Scobbie 2003). As we have seen, the former Latin opposition LL~L changed to /l/~/ɖ/, which is not one of phonological length. Furthermore, even in the most conservative Nuorese dialects, where the voiceless intervocalic plosives have been preserved, and an opposition /pp/~/p/, /tt/~/t/, /kk/~/k/ is in principle present, the status of a double vs a simple consonant remains unclear for the voiced plosives, since simple voiced plosives developed into /ß ð ɣ/ with the former length opposition now being /bb/~/ß/, /dd/~/ð/, /ɡɡ/~/ɣ/.10 In Logudorese/Campidanese, where voiceless plosives have also been subject to lenition, at least for plosives the length opposition has been lost. Moreover, even in Nuorese intervocalic plosives seem to have the same length, regardless of their (p.276) origin: /ˈlatɛ/ <ˈlakte ‘milk’ and /ˈlatu/ <LATUS ‘side’ (whereas we have Cpd. /ˈlati/, Log. /ˈlatɛ/, but Cpd./Log. /ˈlaðu/; Jones 1988b:321). As for obstruents, former geminates in all Sardinian varieties are characterized by the fact that they do not undergo lenition as do most simple intervocalic consonants—but they do not seem to be phonologically long, with the exception perhaps of Nuorese voiceless plosives. The only geminates which seem to show a phonemic length contrast in most Sardinian varieties are the sonorants /nn/ (cf. manu ~ mannu; cf. Virdis 1978:55) and /rr/.11 Nevertheless phonetic strengthening is common in all dialects, as in Cpd. (Senorbì) [tɛˈllɛffɔnu] ‘telephone’ < It. telefono (Molinu 2009:151), or as a consequence of metathesis, which often leads to consonant reinforcement at the former position of the dislocated vibrant, e.g. drommire ‘to sleep’ > DORMIRE.

17.2.3 Sandhi phenomena

Lenition of word-initial consonants also takes place synchronically in intervocalic contexts—including where the initial consonant is followed by a glide or [r]. Whereas in Nuorese voiceless intervocalic plosives are preserved also across word boundaries, e.g. [su ˈpanɛ] ‘the bread’, in Logudorese we find [su ˈßanɛ] and in Campidanese [su ˈßani]. In all Sardinian dialects, /b d ɡ/ undergo lenition in intervocalic phonosyntactic contexts, thus Nuo. [su ˈßɔɛ] ‘the ox’, [saˈ ðɔmɔ] ‘the house’, [su ˈɣenneru] ‘the son-in-law’, Log. [su ˈɔɛ], [sa ˈɔmɔ], [su ˈenneru], and Cpd. [su ˈɔi], [sa ˈɔmu], [su ˈenneru]. The same holds for the development of /f s/ with [ˈfoku] ‘fire’ in [su ˈvoku] ‘the fire’ and Nuo./Log. [ˈsɔlɛ]/Cpd. [ˈsɔli] ‘sun’ in [su ˈzɔlɛ]/[su ˈzɔli] ‘the sun’. Further phonosyntactic lenition processes in the Campidanese area include, e.g. [ˈʧelu] ‘sky’ vs [su ˈʒelu] ‘the sky’.

As for phonosyntactic doubling (cf. §40.3.1), in Sardinian it is not a phonological rule (cf. also Loporcaro 1997a,b). It involves a (diachronic) lexical assimilation process between word-final and word-initial consonants which triggers phonosyntactic strengthening. Since word-internal ‘geminates’ seem to have the same length as word-initial consonants altered by the sandhi rule discussed here (Contini 1986:530), they cannot be geminates proper, namely, phonologically long consonants. In the phonetic transcription, nevertheless, we shall represent these consonants by a double notation.

Typical lexical triggers of phonosyntactic strengthening are the conjunction e < ET ‘and’ and the preposition a < AD ‘to, at’, as well as the interrogative particle a < AUT (cf. §17.4.3):



Furthermore, since Sardinian has many inflected words ending in a consonant, especially -t in verbal inflection, phonosyntactic strengthening is found in exactly those contexts where assimilation processes can take place, e.g. fit pastore [fippasˈtɔre] lit. ‘he.was shepherd’.

Final -s and -r in syllable contact with an initial consonant are either assimilated, when the initial consonant is /n/ or /l/ (sos luches [sɔl ˈlukɛzɛ] ‘the lights’ and battor litros [ˈbattɔl ˈlitrɔzɔ] ‘four litres’), or neutralized to [s], e.g. before /p t k/ (duas caras [duas ˈkaraza] ‘two faces’, but battor caras [ˈbattɔs ˈkaraza] ‘four faces’) or to [r], e.g. before most other consonants (battor domos [ˈbattɔr ˈðɔmɔzɔ] ‘four houses’, but sas domos [sar ˈðɔmɔzɔ] ‘the houses’); in intervocalic context /s/ is voiced phonosyntactically (see also Contini 1986, and §

17.2.4 Suprasegmental features

Most Sardinian words are stressed on the penultimate syllable. Stress on the last syllable is generally avoided, leading to the addition of paragogic vowels on monosyllabic words, even for loans: [kafˈfɛi] ‘coffee’ (< It. caffè), [gatˈtɔu] ‘cake’ (< Fr. gâteau)—an exception being vocatives with apocope (but only in Logudorese, see §17.3.1). Campidanese also creates new oxytones by nasalization and contraction as in LANAM > lãã > ‘wool’, vowel contraction and/or consonant deletion, e.g. ‘bean’ < FABAM. In imperatives clitics can attract the stress (Cpd. nàra! ‘tell.IMP.2SG!’ but nara·mì! ‘tell.IMP.2SG=me!’; Nuo. daze·mì·las ‘give.IMP.2SG=me=them’), with much dialectal variation.12

The paragogic vowel (cf. §17.2.1) has consequences for syllable structure, since in absolute final position paroxytones become proparoxytones, e.g. amigos is realized as [a.ˈmi.ɣɔ.zɔ]. Furthermore, in Sardinian, proparoxytones are more often maintained than in other Romance languages (Virdis 1978:29f.); in the verbal paradigm even infinitives in -ĒRE (and occasionally in -IRE and -ARE) tended to change their stress pattern, thereby joining the third (p.277) conjugation, whose infinitives are stressed on the stem (cf. §17.3.2). Geisler (1994) claims that the initial syllable of a word is strengthened due to a ‘general levelling of accentual stress’ in syllables with primary and secondary accent and, as a consequence, the weakening of obstruction in syllable contact. According to him, this is why, especially in Campidanese, metathesis is progressive (e.g. càpra > cràpa) and accentual shift occurs in the verbal forms.

Metathesis of [r] (also from /l/) is common in all dialects; however, the output is different in Logudorese/Nuorese and Campidanese. In the latter, /r/ in the coda seems to be banned (Frigeni 2005:17), which is the main reason for metatheses such as FIRMARE > frimmai ‘to stop’ (Lörinczi 1971:425-8; cf. also Virdis 1978:75f.; Lai 2013). Consequently, the remaining consonant of the cluster dissolved by metathesis is phonetically reinforced. In Nuorese the result of metathesis can also be a closed syllable, e.g. PETRAM > perda ‘stone’ (besides preda/preta).

Metathesis also produces marked syllable onsets in word-initial position, like [sr], [mr], [ʧr], [ʤr], and also [zr], [ʒr] in word-initial position, cf. [ˈsroɣu] < SOCRUM ‘father-in-law’, [ˈmruɣu] < MUTULUM ‘pile’, [ʤrɛnˈnaʣu] < IANUARIUM ‘January’, in Campidanese (cf. Geisler 1994:115-17).

Studies on the intonation of Sardinian focus on interrogatives (cf. Contini 1976; Schirru 1982). The intonation of Nuorese is studied by Lai (2002). Recently, a group of researchers (the Grup d’Estudis de Prosòdia directed by Pilar Prieto) has proposed a preliminary description of Sardinian prosody and intonation in the ToBI system (cf. Vanrell et al. 2015). Their findings suggest that the most general contrasts found in other Romance languages are also used in Sardinian.

17.3 Morphology

The view that Sardinian morphology is conservative (Wagner 1997:290) can only be upheld in some respects, including the -us ending of originally neuter nouns (now an allomorph of the singular ending of the second declension), the conservation of the ending -t for the third person singular and of the Latin imperfect subjunctive. Rather, Sardinian morphology is structurally highly innovative, since many morphological patterns were replaced by analytic structures (cf. §§17.4.2, 46.1). We may also mention the absence of synthetically formed adverbs and the analytic formation of ordinal numbers (article + de ‘of’ + cardinal number), e.g. Log./Nuo. su de bator, Cpd. su de cuatru lit. ‘the of four’ (‘the fourth’; cf. Wolf 1983).

17.3.1 Nominal inflection

Table 17.6 Inflectional classes of nouns (Logudorese/Nuorese)


Sardinian nouns come in four classes which can be defined on the basis of plural morphology, at least in Nuorese and Logudorese, i.e. class I with plurals in -os, class II with plurals in -as, and classes III and IV with plurals in -es and -is, respectively (see Table 17.6).13

Class I continues the Latin first declension. Based on the singular morphology, class II is divided into three subclasses: IIa, ending with -u (most frequent or ‘regular’, from the Latin second declension); IIb, ending in -o (containing words of various origins); and IIc, ending in -us (with Latin words of the third declension that were originally neuter, e.g. tempus ‘time’, corpus ‘body’; Cpd. cinus ‘ash’). Class III is mostly based on masculine and femininine words of the Latin third declension, and has, as in Latin, no default gender. This class, apart from Latin nouns coming from the masculine and feminine accusative singular forms, also contains nouns to which a paragogical -e was added, such as some nouns ending in -EN that were neuter in Latin (e.g. FLUMEN > frùmene ‘river’, SEMEN > sèmene ‘seed’). Class IV mostly contains elements of old Tuscan, and Catalan, origin, in particular the suffix -eri (cf. §17.3.3).

Due to the neutralization of final -i/-e and -o/-u in Campidanese (cf. §17.2.1), in this variety, classes IIa and IIb as (p.278) well as III and IV show the same endings, respectively (IIa,b: SG -[u], PL -[us]; III, IV: SG -[i], PL -[is]). However, the relevant (sub)classes can still be distinguished by (lack of) metaphony, as pointed out by Bolognesi (1998), such as s[o]nu/s[ɔ]nus ‘sound(s)’ (IIa) vs c[ɔ]ru/c[ɔ]rus ‘heart(s)’ (IIb); m[ɛ]ri/m[ɛ]ris ‘master(s), owner(s)’ (III) vs barb[e]ri/barb[e]ris ‘barber(s)’ (IV).

Gender appears reduced to masculine and feminine. Nouns almost exclusively stem from the Latin accusative, with some very rare exceptions, such as Log. sorre/Cpd. sorri <SOROR ‘sister’ (Wagner (1960-64, II:428f.) and mere/meri ‘master, owner’ (< MAIOR, DES II:108) from the Latin nominative, as well the Latinism Deus ‘God’. Some fossilized remainders of cases other than nominative and accusative include the genitive in the names of the days Monday, Tuesday, Wednesday (lunis, martis, mèrcuris), the vocative in some person names (such as Antoni, Nuo. Dumìniche), and the ablative in Nuo./Log. domo ‘house’. Note that Logudorese Sardinian developed a productive vocative formed from truncating all material following the accented vowel, e.g. Antoni > Antò, Maria > Marì (cf. Floricic 2002; Cabré and Vanrell 2013; Vanrell et al. 2015; cf. also §§, Some neuter plural nouns are conserved as feminine singular, showing a collective meaning (e.g. frùttora ‘fruit’, linna ‘wood’, bestimenta ‘clothes’) (cf. §42.4). Sardinian has also developed a productive system of using the singular form to denote an indefinite number of small items, mostly of fruits, vegetables, insects, and small animals, the result being a collective reference, e.g. Nuo. patata ‘potatoes’, musca ‘flies’, pira ‘pears’, preducu ‘lice’ (characterized by Mensching 2005:96-8 as object mass nouns).

There are class I and IIa,b adjectives (bonu/bona ‘good’; Log. galloffo, Cpd. galloffu/galloffa ‘miscreant’, with -u/-o signalling masculine and -a for feminine) and class III adjectives (Nuo./Log. forte, Cpd. forti ‘strong’, for both genders). Comparatives and superlatives are formed analytically using prus (< PLUS ‘more’; Log. also plus/pjus), with the exception of bonu ‘good’ (Log./Nuo. mezus, Log. me(n)gius, Cpd. mellus ‘better’) and malu ‘bad’ (pe(j)us ‘worse’). Elative formation in -issimu is not indigenous. Forms such as bellìssimu, often found in poetry, are italianizing. Intensification is expressed by the quantifier meda (Nuo. also meta) ‘much’ or by reduplication (bellu bellu ‘very beautiful’).

The definite article derives from IPSE ‘himself’. In all dialects, it is su (M) and sa (F) in the singular. The plural is sos/sas in Nuorese/Logudorese, and mostly is for both genders in Campidanese (cf. Map 17.2 isogloss 7 and Map 17.3). The indefinite article is unu/una. Demonstratives show a threefold distinction as in Latin (custu ‘this (proximal)’, cussu ‘that (medial)’, and cuddu ‘that (distal)’; CF. §54.1.4). Most interrogatives are similar to those of other Romance languages (Log./Nuo. chie, Cpd. chini ‘who’, cale ‘which’, cando/candu ‘when’, cantu ‘how many’), with the excpetion of Log./Nuo. ite, Cpd. ita ‘what’ (supposedly < QUID DEU(S) ‘what god’, DES I 349-50). Possessives are mostly derived from the Latin equivalents (e.g. Nuo. meu ‘my’, tuo ‘your.SG’, suo ‘his, her, its’, nostru ‘our’, bostru/brostu ‘your.PL’), except for the third person plural Log./Nuo. issoro, Cpd. insoru ‘their’ (< GEN. IPSORUM ‘of the very same.PL’; cf. Loporcaro 2001).

Table 17.7 Object clitics


The personal pronoun system is similar to that of most Romance languages, showing clitic as well as full pronouns. The latter exist for subject and oblique functions. The subject forms are (according to the notation of the LSC) deo ‘I’, tue ‘you.SG’, issu ‘he, it’ (or isse for persons only)/issa ‘she’, nois ‘we’, bois ‘you.PL’, issos/issas ‘they’ (Regione Autonoma della Sardegna 2006), but Campidanese mostly has nosu for the first and bosàt(u)rus for the second person plural. These pronouns are also used as obliques, except for the first and second persons singular, which are Nuo./Log. me(ne) and te(ne), Cpd. mei and tei. In Logudorese and Nuorese (and partially in Campidanese), the prepositions a ‘to’ and cun (also: chin) ‘with’ select diverging forms, such as Nuo. a mime, a tibe, but chinmecus, chintecus (from MECUM, TECUM, cf. OIt. meco, teco, Sp. conmigo, contigo). Sardinian object clitics are presented in Table 17.7 (the first and second person pronouns also serve as reflexives).

Sardinian has three types of clitic adverb, traditionally listed among the pronouns: Nuo./Log. bi; Nuo./Log. nde (in some places: ne)/Cpd. ndi; Nuo./Log. (n)che/Cpd. (n)ci. The clitics bi (restricted to Log./Nuo.) and (n)che/nci are used for locatives, roughly meaning ‘(to/from) there’, whereas nde/ndi substitute prepositional phrases introduced by the preposition de, similar to Italian ne and French en. Nuo./Log. bi mostly indicates location and goal, and (n)che usually indicates source. Campidanese uses (n)ci for all three locative functions.

(p.279) 17.3.2 Verbal inflection

Table 17.8 Infinitives according to verb classes


The Sardinian verb presents three conjugation classes continuing the Latin first, third, and fourth conjugations.14 They show the patterns in the infinitive illustrated in Table 17.8.

Original second conjugation verbs have shifted stress to the stem and thereby entered class II, which also hosts some verbs originally stemming from the Latin fourth conjugation such as Nuo./Log. bènner/Cpd. bènn(ir)i ‘to come’. The infinitives of verbs of class II in Logudorese/Nuorese are often spelled with final -e (e.g. pònnere ‘to put’). Since this -e does not usually appear unless before a pause and the -r is affected by the neutralization process described in §17.2.4, it must be considered as paragogical and should thus not be represented graphematically (cf. Pittau 1972:98f.; Molinu 1998:132; Mensching 2004a). The infinitive can be inflected for person and number (cf. §, which is a rather rare feature in Romance and can only be found today in Portuguese and Galician (Loporcaro 1986; Mensching 2000). The forms of the inflected infinitive are identical to those of the imperfect subjunctive (see § for the syntax of the inflected infinitive).


Map 17.3 Approximate geolinguistic distribution of morphosyntactic phenomena in Sardinian

The other non-finite forms are the gerund and the past participle. The latter is formed regularly for classes I and III: istima(d)u ‘loved’, parti(d)u ‘departed’ (Nuo. also ‑atu/‑itu). Class II contains some verbs with a regular past participle pattern STÈM+i+du (Nuo. also ‑tu), e.g. tìmi(d)u / tìmitu ‘feared’, but most participles of this class are irregular. Some can be derived from the corresponding Latin past participles (e.g. fatu ‘made’, cotu ‘cooked’, postu ‘put’), but many are specific Sardinian formations, often based on the old synthetic perfect, e.g. àpi(d)u ‘had’, crèti(d)u ‘believed’, including (for Nuo./Log.) some interesting forms that show -f-, such as Log. pàrfidu ‘seemed’, bàlfidu (from baler ‘to be worth’), chèrfidu ‘wanted’. The gerund, perhaps better characterized as a present participle (at least it also has this function in abba buddinde ‘boiling water’ and the progressive periphrases discussed in §17.4.2), is formed with -a-nde, -i-nde, -e-nde in Nuo. (see Map 17.3); in the other varieties all three classes have the same ending in most places (-ende in Log., -endi or -endu in Cpd.).

The personal endings of the verb are usually distinct for each person (e.g. in Nuo.: -o, -s, -t, -mus-, -tes/(d)es, -n; note the conservation of final -t in the third person singular; in Campidanese -t is also preserved in the third person plural, see Map 17.3). Interestingly, in Campidanese, original perfect endings have entered the second person singular and plural of the imperfect indicative (e.g. of partìri ‘to leave’: partìast(a)/partèstis).

Table 17.9 Synthetic tenses/moods (2SG) (cantare/cantai ‘to sing’)


The tense/mood system for the synthetic verb forms is exemplified in Table 17.9 for the three basic varieties (the analytic tenses/moods—the indicative and subjunctive of the compound perfect and pluperfect, the future I and II and the present and past conditional—will be discussed in §17.4.2).

The synthetic Latin perfect is mostly lost (except for a very few small areas, see Map 17.3). The preservation of the Latin imperfect subjunctive in Nuorese/Logudorese is noteworthy, although generally only àer ‘have’ and èsser ‘be’ preserve these forms. With other verbs, it seems to have survived (with various modifications) only in the southern (Barbagia) part of Nuorese and in several Logudorese varieties.15 The Nuorese/Logudorese areas that have lost the imperfect subjunctive of verbs other than èsser and àer use periphrastic forms, i.e. IPFV.SBJV of èsser/àer + participle, i.e. the original Sardinian pluperfect subjunctive, for the imperfect subjunctive. Campidanese has forms like cantèssit that ultimately derive from the Latin pluperfect subjunctive (but are probably taken from Catalan).

Table 17.10 The L-pattern (Campidanese) for the verb pònni(ri) ‘to put’


Like all Romance languages, some Sardinian verbs show alternations which affect the first person singular present indicative and all persons of the present subjunctive, thus leading to the common Romance L-pattern (Maiden 2005; 2009a:47). The L-pattern is exclusively caused by original yod (cf. Table 17.10).

The L-pattern extends to some other verbs by analogy, and often old synthetic perfect stems (see above for the past participle) ‘intrude’ into the present tense, occupying the L-pattern cells, such as the stem ap- for àer/ai ‘have’ or the perfect stems with the -f- theme: parfo ‘I seem’ or balfo ‘I am worth’ (for further discussion of the L-pattern in northern Logudorese, see Loporcaro 2012b:17-20; also cf. §43.2.3).

We find suppletion in a small number of verbs. One example is the Italianism andare/andai ‘to go’. Some forms (p.280) (p.281) of UADERE ‘to go’ are locally preserved in Nuorese and the Barbagia region for the first person singular present indicative, and are more widespread in the second and third persons of the singular, whereas the plural always seems to present forms of andare. Elsewhere this verb is mostly a regular verb without suppletion (cf. Loporcaro 2012b:13). The verb ‘to be’ (èsser/èssi(ri)), apart from the kind of suppletive variation known from other Romance languages (E-forms besides S-forms), has striking imperfect forms, which show the F-theme of the Latin perfect (FUI), cf. the LSC version (Regione Autonoma della Sardegna 2006), essentially corresponding to Log. fia, fias, fiat, fiamus, fiais, fiant. The -a- is not present everywhere, and the first person singular has local variants with -p-, so we get e.g. in Nuoro: fipo, fis, fit, fimus, fizis, fin. In Campidanese, the first person singular shows many variants (e.g. femu, fia, fudiu).16 Finally, it is worth mentioning that in Campidanese the F-theme can also be found in the third person plural of the present tense of èssi(ri), so we get funt(i) alongside sunt(i). The verb nàrrer/nàrri ‘to say, tell’ (< NARRARE), is mostly conjugated regularly following the first conjugation, cf. the present indicative paradigm according to LSC: naro, naras, narat, naramus, narades, narant. This verb has developed short forms without -r- (e.g. in Cpd.: nau, nas, nat, naus, nais, nant(a)), which locally appear as alternatives to the full forms.17

17.3.3 Word formation

The first comprehensive work on Sardinian word formation is Wagner (1952). Pinto’s (2011) work is also historical. Unless otherwise stated, in what follows examples are taken from Wagner (1952) and Pinto (2011), with some additions from various dictionaries (DES, Puddu 2000; Rubattu 2001-4; Farina 1987; 1989).

According to Pinto (2011:18), characteristic of Sardinian is its poverty of prefixes. Only a small part of the vocabulary shows prefixes, and the few formations that can be found usually involve only three prefixes: (i) a-, (ii) in-, and (iii) Nuo./Log. is-/Camp s-, all of them verbal. A (< AD) provides the verb with a reiterative/intensive value or it describes (re-)entering into a state, as in Log. abbumbare ‘to get drunk’ from bumbare ‘to drink’. Many examples given by Pinto (2011:113) can be described as causative, e.g. Nuo./Log. allù(gh)er, Cpd. allùiri ‘to light’ from lù(gh)er/lùxi(ri) ‘to shine’; Log. appasare ‘to calm down’ < pasare ‘to rest’. These changes in Aktionsart logically go along with a change in valency. The prefix in- sometimes shows intensifying function, such as Log. impasare ‘to make a pause’ < pasare, but most of Pinto’s examples are strongly lexicalized and no longer transparent (e.g. Log. basare ‘to kiss’ > imbasare ‘to match’). The prefix is-/s- continues (DE)EX- (Wagner 1952:136) and mostly seems to have an adversative meaning, as in Log. iscuncordare, Cpd. scuncordai ‘to disagree’ from cuncordare, -ai ‘to agree’. While this verb may have been built on an Italian model (cf. It. concordare > sconcordare), there are also clearly independent formations such as Cpd. crosai ‘to seal’ > scrosai ‘to unseal’.

These three prefixes have mostly been used for parasynthetic verb formation (cf. Pinto 2005, and also §28.4.2), a pattern which has given rise to an impressive number of mainly N-to-V-formations (Pinto 2011:135 reports around 500 in the DES), e.g. ammustare/ammustai ‘to press grapes’ (< mustu ‘must’), isconcare/sconcai ‘to behead’ (< conca ‘head’).

The most important word formation process is suffixation. Pinto (2011:56) has identified around 60 nominal and adjectival suffixes. The most frequent for N-to-N derivations are ‑ale/ali (< -ALIS): Nuo./Log. fundale, Cpd. fundali ‘bottom (of a valley or the sea), dregs’ (< fundu ‘depth’), -arju/-arzu/‑argiu (< -ARIUS): Nuo. berbecarju, Log. berbegarzu, Cpd. brebegargiu ‘shepherd’ (<berbeche/berbeghe/brebei ‘sheep’), -eri (< OCat. ‑er(i), OTusc. ‑eri): Nuo./Log. castanzeri, Cpd. castangeri ‘chestnut seller’ (< castanza/castangia ‘chestnut’), -ile/-ili (< -ILIS): Nuo. fochile, Log. foghile, Cpd. foxili ‘fire place’ (< focu/fogu ‘fire’); for V-to-N derivations -ATA: torrada (Nuo. also torrata) ‘return’ (< torrare/torrai ‘to return’), -TOR: Nuo. messatore, Nuo./Log. messadore, Cpd. messadori ‘mower’ (< messare/messai ‘to mow’), -TORIUS: Nuo. filadorju, Log. (p.282) filadorzu, Cpd. filadroxu ‘spinning place/wheel’ (< filare/filai ‘to spin’), -ŪRA: ingranidura ‘seeding’ (< ingranire,-iri ‘to seed’); for N-to-A derivations -ŌSUS: Nuo. pretosu/predosu, Log. pedrosu, Cpd. perdosu ‘stony’ (< preta/preda/pedra/perda ‘stone’), -UTUS: Nuo./Log. pitudu, Cpd. pitzudu ‘beaked’ (< pitu/pitzu ‘beak’).

There are many other suffixes and some that are mostly present in loans, in particular in learnèd words and italianisms (e.g. the V-to-N-suffixes -tzione and -àntzia, ‑èntzia mentioned in Regione Autonoma della Sardegna 2006). In the standardization debate, the use of the autochthonous suffixes is sometimes suggested: Cpd. standardisadura instead of standardizatzioni ‘standardization’ (CSNC 2009), elaboradore or carculadore ‘computer’, Cpd. computadori/carculadori/ordinadori) (cf. Regione Autonoma della Sardegna 2006:56; CSNC 2009:182).

Wagner (1952:3) had also referred to the poverty of affective suffixes. Diminutives are formed by the still very productive -eddu, -a (-ELLUS), cf. domedda (< domo/domu ‘house’), which can also attach to adjectives, e.g. Nuo./Log. minoreddu (< minore ‘small’), with the variant Nuo. -icheddu, Log. -igheddu, Cpd. -ixeddu. The Latin suffix -ONE had some productivity for augmentative/pejorative formations in earlier times, but now seems to have been supplanted by Log./Nuo. -atzu, Cpd. -aciu of Italian origin.

As in all Romance languages (cf. Ch. 29), compounding in a strict sense (i.e. excluding syntagmatic compounds, e.g. those containing prepositions, see Pinto 2011:34-7) is much lesss productive than derivation (cf. also §29.2.7). There is a small series of mostly strongly lexicalized left-headed compounds, such as N+N: Log. murufossu (‘border of a ditch or field’, DES II,142) < muru ‘wall’ and fossu ‘ditch’, or Cpd. lanarrangiu ‘spider web’ < lana ‘wool’ and arrangiu ‘spider’; N+A: Log. gattagreste ‘marten’ < gattu ‘cat’ and agreste ‘wild’.

There is a N+A noun formation pattern with a juncture vowel /i/ (Pinto 2011:41 discusses its status with a few examples extracted from DES; some more examples can be retrieved from Wagner 1952:145f.). They mostly serve to name animals, and the first part consists of a body part, e.g. for birds ala ‘wing’: Arborense alimannu ‘lark’ (mannu ‘big’); Log. aliarza ‘finch’ (Nuo. also aliarja) (barzu/barju ‘multicoloured’), Log. alibintu ‘finch’ (pintu ‘coloured’). The juncture vowel is much more common and popular in a right-headed N+A pattern, which leads to adjectives that mostly characterize persons and show a lesser degree of opacity than the other patterns mentioned so far, including fully transparent words such as those quoted by Wagner (1952:144) from Nuorese folk songs, e.g. pilibrundu ‘blond haired’ (< pilu ‘hair’ + brundu ‘blond’), coriduru ‘hard-hearted’ (< coro ‘heart’ + duru ‘hard’).

As in other Romance languages, one of the most productive types of compounding is the exocentric V+N-to-N type, which forms masculine singular nouns. The nominal element often presents plural morphology, as in Cpd. stasibois ‘rest-harrow’ (literally ‘fatigue+oxen’) or Log. fagheganneddos ‘kind of winder for spinning’ (< fagher ‘to make’ + caneddu ‘rolling spin’), Cpd. faiganeddus. Whereas the verbs involved in this pattern are mostly transitive, we can find some examples built from intransitive verbs, such as lampaluche/-lughe/-luxi ‘flashing’ (<lampare/lampai ‘to flash, to glitter’+ luche/lughe/luxi ‘light’). This word formation process is still productive (cf. Lepori 2001:144): Cpd. bogapruini ‘vacuum cleaner’, bogaludu ‘mud wing’ (bogai ‘remove’, pruini ‘dust’, ludu ‘mud’); also cf. sciacuapannus ‘washing machine’ (sciacuai ‘rinse’, pannus ‘clothes’), sciacuastrexu ‘dishwasher’ (strexu ‘recipient, dish’) (CSNC 2009:182).

Verbs are often formed from nouns by what can be regarded as conversion or null suffixation. Thus, the verb stem of arcare/arcai ‘to curve’, arc-, corresponds to the nominal stem of arcu ‘bow’ (but cf. Pinto 2011:23, 99-102). With very few exceptions belonging to the third conjugation class, this process builds verbs of class I, e.g. Log. landare ‘to collect acorns’ from lande ‘acorn’, linnare/linnai ‘to collect wood’ from linna ‘wood’. Many Sardinian nouns are also the result of V-to-N-conversion, in the sense that the verbal stem is transformed into a nominal stem (mostly belonging to class IIa); cf. ammentu ‘memory’ < ammentare, -ai ‘to remember’, Log. chistionu ‘talking, quarrel’ < chistionare ‘to talk, to discuss’, Nuo. issèperu, Log. issèberu, Cpd. iscèberu ‘choice’ < isseperare/isseberare/sceberai ‘to choose’, Log./Cpd. accorru ‘enclosure, corral’ (< accorrare,-ai) for ‘to gather or shut in the cattle’.18

17.4 Syntax

Since Michael Jones’s (1993) seminal study, Sardinian has become one of the most quoted Romance languages in modern syntactic theory. Although Jones analyses data exclusively from Lula, a Nuorese village, it has become commonplace to consider Sardinian syntax as quite homogeneous and less subject to variation than other areas of the linguistic system (e.g. phonology). However, recent studies, such as Bentley (2011), show that there is considerable microvariation.

(p.283) 17.4.1 Nominal group

Within the Sardinian noun phrase (cf. Jones 1993:ch.2),19 most prenominal elements (articles, demonstratives, wh-determiners, and numerals) show agreement with the head noun (3a).20 The quantifiers cada or Nuo./Log. donzi/Cpd. dogna ‘each’, Nuo./Log. carchi/Cpd. calincunu,-a ‘some’ are also prenominal. With the exception of calincunu, these do not inflect (3c,d). The item totu ‘all’ is usually uninflected too, and needs a determiner which it precedes (3b):



Attributive adjectives are rare in prenominal position (cf. 4a), only very few items taking this position, e.g. bellu ‘beautiful’, bonu ‘good’, bravu ‘good, well-behaved’, santu ‘holy’, pòveru ‘poor’. All of these are Italian loans which were apparently borrowed together with their syntax. All other attributive adjectives, as well as possessives, are obligatorily postnominal:



Adjectives, including possessives, show concord in gender and number with the head noun (thus: sos libros bellos lit. ‘the.MPL books.M beautiful.MPL’, una mesa manna ‘a.FSG big.FSG table.FSG’, tres mesas mannas lit. ‘three tables.FPL big.FPL’). The properties of the quantifier meda (M/F, Nuo. also meta) ‘many, much’ are subject to diatopic variation; the postnominal position is generally preferred, and, especially in this case, it can lack number agreement.

Apart from transparent partitive constructions such as una de sas berbeches (‘one of the sheep’), un’azicu de (custu) vinu ‘a bit of (that) wine’, the literature mentions the Campidanese quantifier element unos/-as cantu (where cantu is not inflected) with an optional de (e.g. unas cantu (de) melas ‘some apples’). Here, the partitive semantics is lost (similar to Fr. beaucoup de ‘a lot of’). Sardinian seems not to have an independent partitive construction for indefinite mass nouns or plurals (unlike (central-northern) Italian and French, cf. dell’acqua/de l’eau ‘some water’). The cases mentioned in the literature (Wagner 1997:328; Blasco Ferrer 1984a:84f.) are mostly of clitic right-dislocation, where bare nouns are preceded by de ‘of’ (see § In contrast, a non-dislocated indefinite direct object of a singular mass or plural noun is bare and not introduced by de: Apo mandicatu petha ‘I ate meat’ (Jones 1993:217).

Bare nouns cannot appear in subject position. However, Sardinian allows even singulars of count nouns to be bare in some non-subject environments. Apart from more or less idiomatic locative expressions such as in/a domo ‘(at) home’, in mesa ‘on the table’, such cases include complements of verbs of possession in negated or interrogative contexts, e.g. Maria non juchet bonete ‘Maria has no hat on’; Bonete juchìas? ‘Did you have a hat on?’ (note that Spanish and Romanian behave similarly).

Names of persons and towns regularly lack articles, as do kinship terms. For the latter, the words designating parents and grandparents, at least in most varieties, also lack a possessive pronoun when they belong to the speaker (i.e. babbu = ‘my father’). Finally, the use of the definite article with a kinship term usually forces a third person singular possessor reading (sa sorre = ‘his/her sister’). In such cases, the possessive pronoun is even judged ungrammatical by many speakers.

As in other Romance languages, there is no dedicated device to express specificity. However, quantifiers and other indefinites can be classified with respect to their function as or to their co-occurrence with specific and non-specific expressions: charchi and cali(n)cunu ‘some’ show a strong preference for [–specific] NPs, whereas the indefinite article as well as unos(‑as), cantos(-as) (de), argunos(-as), paritzos(-as) (only Log.) and diversos(-as), also meaning ‘some’ or ‘several’, are mostly restricted to [+specific] contexts. As shown in Mensching (2005), many of these indefinites were taken from Italian varieties and some from Spanish.

Specificity is known to play a role in the Spanish prepositional accusative construction (Torrego 1998; von Heusinger and Kaiser 2003). Sardinian also has differential object marking (DOM; cf. Bossong 1982),21 whose surface (p.284) properties are in fact similar to Spanish: whereas indirect objects are regularly preceded by the preposition a, direct objects often bear this preposition but only when they refer to persons, at least as a rough approximation. However, in contrast to Spanish, in Sardinian DOM usually occurs only with definite nominal expressions. They can either be inherently definite such as personal pronouns, proper names, or kinship terms, or become definite by the use the definite article. DOM is obligatory with inherently definite expressions and optional with others:



The exact properties and diatopic variation of cases such as (5b) require further investigation. Note that with proper names animacy or even a [+human] feature is not decisive, since the construction at issue also appears with place names (5a). Whereas the nominal expressions in (5a) are all specific, this is not the case with interrogative pronouns and negative indefinites, which also obligatorily trigger DOM when applied to humans (a chie ‘whom’, a nemos ‘nobody’). In addition, other indefinite NPs never show DOM, even when they are specific. It thus seems that definiteness and (with the exception of proper names) a [+human] feature are relevant, but specificity is not.

17.4.2 Verbal group

In the verb group, argument structure (valency, categorical selection, and thematic roles) and the event situation (lexical and syntactic aspect) are encoded. (Di)transitive and unergative verbs on the one hand and unaccusative verbs on the other show different syntactic behaviour with respect to auxiliary selection in the perfect/pluperfect. Unaccusative verbs select BE and (di)transitive and unergative verbs select HAVE:22



A typical causative/ergative alternation in Sardinian can be obtained by the reflexivization of a causative transitive verb, giving rise to an ergative/unaccusative alternation and with selection of BE in the latter case (and participle agreement, see below; Jones 1993:99):



However, some verbs also show a causative/ergative alternation without reflexivization, e.g. issire ‘to take out something’ (with HAVE) and ‘to go out’ (with BE), or a causative/unergative alternation like buddire ‘to boil something’ and ‘to boil’ and imparare ‘to learn’ and ‘to teach’ (both with HAVE). Other verbs can be used as unaccusatives with or without a reflexive clitic with no noticeable change in meaning, e.g. assimidzare/s’assimidzare ‘to resemble’ and mòrrer/si mòrrer ‘to die’ (with BE), but mòrrer is also transitive ‘to kill’ (with HAVE), hence another example of the causative/ergative alternation (cf. Jones 1993:99f., 123; Puddu 2000, s.v.).

Sardinian encodes most of its principal tense, aspect, and mood oppositions, not by synthetic inflection, but analytically. Not only are the perfect and the pluperfect expressed by analytic forms, but the conditional and the future also consist of periphrastic constructions involving a finite auxiliary and a main verb in a non-finite verbal form (cf. Iliescu and Mourin 1991:448; Jones 1988a):



(p.285) The future with dèper (only used in the simple future of the verbs èsser ‘to be’ and àer ‘to have’) is typically Logudorese/Nuorese, but less widespread than the formation with àer. The paradigm of the auxiliary dèper is not the same as dèper in the sense of ‘must’ (e.g. 2SG des vs depes; cf. Blasco Ferrer 2002:86; for the HAVE TO type, see also Bentley 2009). (8f) is the new analytically formed conditional which—in contrast to most other Romance languages (§—remains analytical, representing a future-in-the-past construction, with the future auxiliary dèper ‘must’ in the past giving rise to the meaning of the conditional I or the perfective conditional II. In Campidanese, the conditional is formed with the imperfect tense of ài ‘have’ (Blasco Ferrer 1986:123f.), hence emu/iast/iat/emus/estis/iant a cantai. Most of the auxiliaries used in the Sardinian verb phrase are morphophonologically reduced forms of the verbs they stem from (cf. §

There is a tendency in several varieties across the whole island to continue to express immediate past/past action with effect on the present by means of the analytic perfect and more remote past/past action without effect on the present (earlier expressed by the synthetic perfect) by means of the pluperfect, see Mensching (2012), where a sample sentence from Ploaghe, one of the few localities that still have the synthetic perfect (cf. Map 17.3), is contrasted with varieties that have lost the synthetic forms:



Iliescu and Mourin (1991) also list three surcomposé tenses (aio àpiu cantau ‘I had had sung’, des àer àpiu cantau ‘you will have had sung’, dio àer àppiu fatu ‘I would have had done’; cf. 22; cf. also Pisano 2010b, and §58.3.4). These periphrases consist of an auxiliary, HAVE or BE, and two past participles, one representing another auxiliary and the other one the full verb and appear typically (but not necessarily) in counterfactual conditionals or other irrealis contexts (§

The present tense too can be said to be periphrastic, at least in the case of a specific reference situation where the progressive form built on the verb BE and the gerund/present participle is chosen in many more contexts than in Italian:



This progressive periphrasis can be also used with stative verbs such as ‘to understand’ and in the imperfect, giving rise to a progressive in the past (cf. Jones 1993:83f; also cf. Blasco Ferrer and Contini 1988:844).24

Argument structure is co-responsible for auxiliary choice, and if the auxiliary is BE, participle agreement with the surface subject is obligatory (cf. §49.3). However, in contrast to Italian, reflexive constructions with the reflexive pronoun representing the indirect object in the dative (reciprocal, reflexive, dative of interest) select HAVE and do not exhibit participle agreement (cf. Loporcaro 1998b:46, 54):



Yet, not only argument structure but also the referential properties of the arguments are decisive for auxiliary selection, witness the following existential constructions (cf. also §52.4) which exhibit definiteness effects (Jones 1993:114f.):



(p.286) In existential constructions proper, as in (12b), the argument coming into existence, the pivot, can only be indefinite (it must be ‘brand new’; cf. Bentley 2004b). In this case, the position of the pivot is postverbal, auxiliary selection is HAVE, and there is no agreement. (12a), with a strong, definite argument, is not an existential but a locative construction (cf. Remberger 2009) and in this case the argument can either be pre- or postverbal, auxiliary selection is canonical, i.e. BE for unaccusatives, and there is always agreement (cf. La Fauci and Loporcaro 1997; Bentley 2004b; 2011). Impersonal presentational constructions with a verb in the perfect,25 like b’at vènnitu tres pitzinnas ‘three girls have arrived’, show the same properties as the existential construction proper (auxiliary HAVE; no agreement; postverbal subject), whereas canonical clauses in the perfect, like bi sun vènnitas sas pitzinnas ‘the girls have arrived there’, correlate with the locative constructions (auxiliary BE, if unaccusative; agreement; also possibly preverbal subject).

Participle agreement, which is obligatory in conjunction with auxiliary BE, additionally appears in the context of accusative clitics, where it is obligatory in the third person, but not in the first and second person nor with the partitive clitic n(d)e (cf. Loporcaro 1998b:42).



17.4.3 Clause syntax Basic properties of the clause

Sardinian is a null subject language with several constructions where postverbal subjects appear more natural than preverbal subjects (for information structure, see §, and for general discussion, Ch. 34).26 Besides the existential constructions mentioned above, unaccusative verbs in general show postverbal subjects (cf. 32b; also Giurgea and Remberger 2012b). Inversion is typical for interrogatives (§§, as well as for focus fronting (§§34.5.2-4), a phenomenon particular to Sardinian. However, Sardinian has never been argued to display V2 effects in its earliest stages, having had a preference for V1-clauses (cf. Lombardi 2007b; Wolfe 2013a,b). Nevertheless, in transitive sentences the order of principal sentential constituents is now SVO.

The clitic pronouns mentioned in §17.3.1 usually occur immediately before the finite verb (cf. 31). With gerunds and imperatives, they are enclitic (biende·la ‘seeing=her’, nara·lis ‘tell=them!’; for pro- vs enclisis cf. §48.3). In contrast, infinitives show proclisis (pro la bier lit. ‘for her= see.INF (= in order to see her)’; cf. § Clitic pronouns form clusters in the order dative–accusative (e.g. Nuo./Log. mi lu, ti lu; Cpd. mi ddu, ti ddu ‘me/you.SG it’). When two third person clitics are combined, there are morphophonological modifications (e.g. Nuo. li+lu > [liu] ‘him it’) or the dative pronoun is replaced by another clitic (e.g. bi in Lula; si in Cpd.) The clitic adverbs precede third person clitics but follow first and second person clitics.

Sentential negation is encoded by non (with a variant no mostly used before vowels, cf. also Floricic 2012). This element precedes the finite verb and, if present, proclitics. Items such as Log./Nuo. nemos, niune/neune, Cpd. nemus, nisciunu(s) (‘nobody’), nudda ‘nothing’, mai ‘(n)ever’, prus ‘more’, among others, still require the element no(n) (negative concord). However, if these items appear preverbally, no(n) is omitted:


Sardinian Valency-changing operations

Besides the canonical passive,27 formed with auxiliary be, Sardinian has a want-passive similar to that of several Italian dialects (cf. Jones 1993:124f; Remberger 2006b; §§, 60.5.2): (p.287)



Another type of valency-changing operation concerns reflexive constructions (cf. §§17.4.2, 60.4.1). The following constructions involving a reflexive can be found (cf. Burzio 1986):

  1. 1. unaccusative reflexives, which belong to the causative/unergative alternation like si abbèrrer ‘self= open.INF’, si brujare ‘self= burn.INF’ (cf. 7b vs 7a);

  2. 2. inherent reflexives, e.g. s’irballare ‘self= be.mistaken.INF’, si pentire ‘self= repent.INF’, which are inherently unaccusative and do not take part in a valency changing operation;

  3. 3. reflexive constructions proper where the reflexive clitic indicates a reduction of the argument structure of a transitive verb like lavare ‘wash.INF’ giving rise to an unaccusative verb such as si lavare ‘self= wash.INF’ (cf. 16a; Loporcaro 1998b:45);

  4. 4. impersonal constructions with si where any class of verb can be involved, giving rise to an impersonal construction with an arbitrary subject; this is not a valency-changing operation proper (cf. 16b);

  5. 5. passive-medial constructions involving si where the underlying object is promoted to subject, triggering obligatory subject–verb (and participle) agreement, whereas the underlying subject is suppressed; this construction can have a modal interpretation (cf. 16c; Jones 1993:127):



As for causative verbs, as in lu faco/lasso travallare ‘I make/let him work’ (Nuo.; Jones 1993:271), these involve not so much a valency-changing operation but complex predicate formation (cf. §61.3.3). Since clitic climbing (cf. § is obligatory in these constructions, causative constructions are complex but monoclausal, comprising the argument structure of both the causative verb (responsible for the thematic role of the Agent) and the infinitival verb, whose thematic subject becomes the object of the causative verb. As in modal verb constructions, auxiliary selection in causative constructions is only dependent on the full verb in the infinitive. Finite subordination28

Adverbial subordinate clauses are formed with conjunctions such as Log./Nuo. cando, Cpd. candu ‘when’, ca ‘because’, Log./Nuo. pro chi/Cpd. po chi ‘so that’, manca(r)i ‘although’. There are many more comprising other elements plus chi ‘that’, e.g. in su mentres/-is chi ‘while’, apenas chi ‘as soon as’, apustis chi ‘after’.

In Logudorese and most of Nuorese territory, complement clauses are generally introduced by the complementizer chi:



Volitional verbs such as cherrer ‘want’ and some other predicates which introduce non-veridical contexts select the subjunctive in the complement clause, as do directive predicates. In Campidanese there is a widespread system with two complementizers (cf. §, chi (or ci) and ca, the northern extension of which enters the Arborense zone in the west and penetrates Nuorese territory at Fonni and Dorgali; see the tentative isogloss in Map 17.3. Roughly speaking, ca is used after verba dicendi, sentiendi, and putandi, whereas chi (ci) is used after verba timendi and volitional verbs. Although in some places the system seems recessive with speakers confusing both complementizers, it is generally well preserved and aligned with mood (ca with indicative, chi/ci with subjunctive). The system is still absolutely stable in many places. Thus with predicates that can select either the indicative or the subjunctive (cf. Jones 1993:253-60), the correlation between indicative and ca (p.288) and subjunctive and chi is clearly part of the grammatical competence of the speakers:



Chi (ci) is also the complementizer for relative clauses, roughly corresponding to English relative ‘that’. In a Campidanese area partially overlapping with the ca/chi distinction for complement clauses, ca and chi are used as complementizers for relative clauses, but here no mood distinction is involved. In these places, ca appears in appositive relative clauses, whereas chi is used in restrictive relative clauses:



For functions other than subjects and objects, i.e. after prepositions (including dative a ‘to’) the pronouns Log. cale(s), Cpd. cali(s) preceded by the definite article can be used in formal style, probably under Italian influence. The autochthonous construction uses chi while repeating the associate in the relative clause in the shape of a clitic (dative) or a full pronoun with the relevant preposition, cf. Jones’s (1993:294) example (Nuo.) sa pitzinna chi so issitu chin issa, literally ‘the girl that I went out with her’ (cf. §64.2.2).

For indirect constituent questions, a wh-word is located at the left periphery of the embedded clause, and for indirect yes/no questions the complementizer si ‘if, whether’ is used, with no effect on word order. The complementizer si is also used in conditional clauses, but in Campidanese chi (ci) is preferred, at least in the zone where the ca/chi distinction for complement clauses exists.

The sequence of tenses in conditional clauses varies greatly. In Nuorese/Logudorese speculative/counterfactual conditional sentences can contain either conditional forms or the imperfect or pluperfect indicative both in matrix and in subordinate clause. Pittau (1991:151) also mentions the use of the imperfect subjunctive in the protasis (only with èsser ‘be’ and àer ‘have’) and the conditional in the apodosis. Blasco Ferrer (1986:202) includes Campidanese and presents the imperfect (Nuo.-Log.: pluperfect, cf. §17.3.2) subjunctive in the protasis with the conditional in the apodosis as the most frequent pairing. The conditional and the imperfect indicative in both clauses is mentioned as an alternative. According to Mensching (2012), a frequent option for a condition that can be fulfilled, in particular for Campidanese, is […COND…[si/chi…IPFV]]:



Nuorese/Logudorese sometimes employ the conditional in the main clause and the imperfect subjunctive in the si clause:



The most widespread option for unreal conditions seems to be pluperfect in both clauses. A less frequent option is the pluperfect subjunctive in the si clause with either pluperfect indicative or conditional II in the matrix clause. Typically Nuorese, but also documented in (northern) Logudorese, is a kind of ‘surcomposé’ pluperfect, which is usually found in the protasis but is accepted by some speakers in the apodosis (VIVALDI; cf. also Jones 1993:83; Pisano 2010b):



(p.289) Non-finite constructions

Modal and aspectual periphrases are obligatorily restructuring in Sardinian, i.e. clitic climbing is not optional, with clitics standing proclitic to the finite verb (cf. Jones 1993:142; Remberger 2008):



As a consequence, auxiliary selection is solely dependent on the main verb, since the verb complex is monoclausal, never on the modal verb alone. Thus we get HAVE selection with the transitive verb preguntare ‘to ask’ and BE selection with the reflexive, unaccusative si frimmare ‘to stop’ (Sa-Limba 1999-2012):



In this sense, modal constructions are similar to causative and permissive constructions, in which clitics also appear in proclitic position before the finite verb.29 The difference is that in causative constructions the clitic represents the logical subject of the embedded clause and is case-marked by the causative main verb and not by the infinitival verb (as it is in the monoclausal modal construction). In periphrases involving a gerund, as in the aforementioned progressive form, clitic climbing seems optional, cf. Log. Juanne los fit chirkande/fit chirkande·los ‘Juanne was looking for them’.

Unlike the modal and aspectual periphrases, other constructions are biclausal:



Here clitic climbing is impossible, since main and subordinate clause are clearly separated by the element a ‘to’, which marks the left edge of the embedded clause. In fact, infinitives in biclausal structures are always introduced by a prepositional element (Jones 1993:260-63). In adverbial infinitival clauses these are elements such as pro (Cpd. po) ‘in order to’, chentza (or chene, chena, sena and many other local variants) ‘without’. The preposition de ‘of’ forms compound expressions that also introduce infinitival clauses, such as prima de/innantis de ‘before’ and also appears in infinitival complements to adjectives (e.g. cuntentu de ‘glad to’). Finally, verbs that lexically select a preposition for NP objects present the same preposition when introducing an infinitival clause (such as servire ‘to serve’ + a ‘to’, but si pentire ‘to repent’ + de ‘of’). In other cases, when there is no lexically or otherwise motivated preposition, the elements a ‘to’ or de ‘of’ serve as a kind of default (cf. 25, where the infinitive clause acts as a direct object of the verb provare, which does not select for any preposition; see also §63.2). Similarly a ‘to’ (or de ‘of’) is obligatory in subject clauses: Nuo. A faveddare su sardu no est diffitzile ‘To speak Sardinian is not difficult.’ Subject clauses clearly seem to favour the element a ‘to’, whereas many transitive verbs (e.g. of saying and believing, but also verbs meaning ‘to finish’) prefer de ‘of’ (Jones 1993:260; Mensching 2004a:42).

Like Spanish, Catalan, Gascon, and (in some cases) Romanian as well as old Italian and most modern Italian dialects, Sardinian allows an overt postverbal subject in infinitive clauses (in the nominative case, which can be seen in the case of pronouns) (Blasco Ferrer 1986:159; Jones 1992), the so-called personal infinitive (cf. §



In Nuorese/Logudorese the infinitive can even be inflected in such cases (cf. §17.3.2; Pittau 1972:93f.; Jones 2000):



Example (27b) shows that the infinitive can be inflected even when there is no overt subject. The forms of the inflected infinitive correspond to the imperfect subjunctive. Varieties that have lost the imperfect subjunctive with verbs other than èsser ‘be’ and àer ‘have’ (cf. § may (p.290) still use the corresponding forms of all verbs as the inflected infinitive.30

Finally, perception constructions employ the gerund and not the infinitive (Amus intesu sos tenores cantande/**cantare ‘We heard the tenors singing’ (cf. Pittau 1972:139; Jones 1993:285; cf. also §61.2), as in other Romance varieties such as Catalan and Romanian. Information structure

In Sardinian the (discourse) topical subject can be overt, but is mostly omitted (28a). One possibility to mark an all-new context is inversion (28b):



The default focus of a sentence is on the rightmost part of the sentence, i.e. it is (part of) the last constituent (Jones 1993; 2013). Thus, in (28a) the focus can either be the whole VP or the NP unu libru. Since there is no overt topic in (28b), the whole sentence can be the focus (i.e. a thetic expression) or the focus can be Maria (for a question such as ‘Who has arrived?’). However, for cases such as those in (28a), Sardinian often fronts the focused constituent by a special fronting operation. When present, the subject has to appear postverbally:



This structure resembles, but is not identical to, the focusing operation in Italian identified e.g. by Rizzi (1997), because the latter (and similar constructions in other Romance languages) are mostly reserved for contrastive focus (cf. §34.5.2). This structure is ungrammatical in negated sentences and can apply to almost all constituent types, including adjectival phrases, participles, and even the entire VP (Jones 1993; Bentley 2009; Remberger 2010; Mensching and Remberger 2010a,b; Egerland 2011; Jones 2013):



It is sometimes difficult to consider the fronted element as a focus, in particular in cases such as (30a) (see further Jones 2013). However, we shall continue to use the term focus fronting.

To make a non-subject constituent the topic of a sentence, Sardinan has clitic left-dislocation (ClLD) and, for backgrounding of a constituent, clitic right-dislocation (ClRD), which consist in displacing a constituent to the left or right edge of the clause while inserting a coreferent clitic pronoun (for objects). The following example of the interaction of ClLD with focus fronting shows how the ClLD constituent must precede the focus fronted constituent (Lörinczi 1999:104; cf. §34.4):



Here only the topicalized object, not the topicalized subject, is repeated in via the clitic (dda) (ti is an ethic dative). The participle agrees with the dislocated object (and with the clitic). With ClRD and ClLD participle agreement is also obligatory in the case of dative clitics (reflexive or not) and first and second person clitics (cf. Loporcaro 1998b:55f.). There is no participle agreement with the focus fronted object (30b).

Bare NPs in object position (cf. §17.4.1) as well as nouns modified by quantifiers or adjectives can also become topics, showing a special variant of ClLD and ClRD, in which the preposition de ‘of’ is inserted before the NP/noun (obligatorily with ClRD and optionally with ClLD), and the resulting PP is mirrored by the clitic nde/(n)ci (Mensching 2008a):



(p.291) For this construction, which seems to exist in all languages that have a clitic of the en/ne-type, see Jones (1993:17), Mathieu (2004), Mensching (2005; 2008a). The left-dislocated items in example (32) can be interpreted as hanging topics when de is absent. Illocutionary force

Sardinian examples for the basic sentence types (in the sense of sentence force, cf. §53.1) are given in (33) (Jones 1993:25-7):



Constituent questions are built by fronting the interrogative pronoun (33b). Positive imperatives have their own inflectional forms for the second person, whereas negative and other imperatives are served by the paradigm of the subjunctive (33c,d); in the negative imperative, clitics are in proclitic position. Exclamatives usually involve the same type of pronoun as interrogatives (cf. 33e, with the insertion of chi), but do not show inversion effects. Optatives contain verbal forms in the subjunctive (33f).

Polar questions are often characterized by predicate fronting or argument fronting, which leads Jones (1993:24) to consider fronting as one of the typical Sardinian question formating processes (Pittau 1991:143):



However, fronting is also common in declaratives (cf. §, and what makes it appear so frequent in questions may be the fact that question focus can also be placed on single constituents and even the predicate, which in the latter case often creates a verum-focus interpretation (cf. also Giurgea and Remberger 2012a; for another view, see Jones 2013).

Available only in Nuorese/Logudorese is the formation of polar questions by the question particle a (< AUT ‘or’), which sometimes implies the illocutionary force of a request (cf. 2). Another strategy, found in other Romance languages (cf. §53.3.3), consists in using the wh-item ite ‘what’ as a question particle (cf. Mensching 2012; note that there is no pause between ite and the rest of the sentence) (VIVALDI):



There are particles which serve to modify a basic sentence type by specifying aspects of its illocutionary force, e.g. ge in declaratives, which adds an affirmative flavour (cf. §53.2.2), and ello in (polar and constituent) interrogatives, which connects the question to the situational context and often gives it a rhetorical interpretation (Hinzelin and Remberger 2009; cf. §53.3.3).

The fronting operations mentioned above and described in § add emphasis, mirativity, or verum-focus interpretation also to declaratives so that speakers often interpret them as illocutionary acts of exclamation (note the exclamation mark):




(1) For a general introduction see Atzori (1982), Contini and Tuttle (1982), Tagliavini (1969:388-93), Jones (1988b), Blasco Ferrer (1995a; 2000), Jones (1997), Bossong (2008).

(2) For the substratum, see Blasco Ferrer (2002:35-8) and Paulis (2008). Phoenician/Punic and Greek elements are negligible (cf. Wagner 1997:150-65).

(3) For the superstratum influences mentioned, see Wagner (1997:62-4,165-74,184-232), Blasco Ferrer (2002:189-212).

(4) Earlier attempts were Porru (1811), Spano (1840), Corraine (1990; 1992), Blasco Ferrer (1986), Bolognesi (1999). Cf. also the overviews in Blasco Ferrer (1986:62 n.110; 2002:109-12), Jones (1993:8-12), Mensching and Grimaldi (2005), Stolfo (2009) and the documentation available at Condaghes <http://www.condaghes.com/limbasarda.asp>.

(5) The LSC permits the use of local varieties in a standardized spelling (a practice followed here, with some modifications, to note Sardinian examples graphematically), but also proposes a standard spelling and grammar. The standard version, adopted on an experimental basis by the Regione Autonoma della Sardegna in 2006, tries to reflect the most widespread variant.

(6) Bolognesi (1998) and Frigeni (2002) claim that in Campidanese the vowel system remains underlyingly a 3-height system.

(7) Cf. Wagner (1941:62-4), Contini and Boë (1972:192, 166), Virdis (1978:53), Bolognesi (1998:26f.).

(8) In some varieties of the Barbagia also /k/; cf. Wagner (1941:71), Contini (1971), Virdis (1978:41), Wolf (1985).

(9) They are not all intact in all subvarieties; e.g. Nuoro has /t/ > /d/ (UITAM > /ˈbida/ ‘life’) or even further reduced to Ø in masculine participles (Pittau 1972:109).

(10) Cf. Wagner (1941:195, fn.1), Lüdtke (1953:413), Virdis (1988:904), Jones (1988b:321), Frigeni (2005:20).

(11) Double consonants are usually not represented in the LSC standard and other graphemic systems (cf. §17.1), with some exceptions (b l m n r s). The grapheme dd (spelled dh in other systems) represents [ɖ]).

(12) Cf. Blasco Ferrer (1988a); Bolognesi (1998); Loporcaro (2000); Mensching (2004a:78); Kim and Repetti (2013:268); also cf. §§17.3.2, 40.4, 45.3.1.

(13) The table is adapted from Pittau (1972:67). For the section on nominal morphology, we refer in particular to Wagner (1997:316f.; 1938-9:98-107, 110f.); Pittau (1972:16, 68f., 145; 1991:78, 143); Blasco Ferrer (1986:82-92, 107-9; 2002:81-3); Molinu (1989; 1998); Jones (1993:31-4); Mensching (2004a:26-31, 44-5, 53-6).

(14) Overviews of Sardinian verb inflection are: Wagner (1938-9); Blasco Ferrer (1984a; 1986); Iliescu and Mourin (1991); Pisano (2004-6; 2008; 2010a,b). Two Logudorese varieties (Buddusò and Bonorva) are exhaustively described in Molinu (1989) and Loporcaro (2012b), respectively. We have also included some information and examples from Pittau (1972), Lepori (2001), Mensching (2004a).

(15) For original Latin proparoxytone stress in verbs of class II, see Map 17.3.

(16) As all Campidanese verbs in the imperfect tense, the second person has Latin perfect endings (e.g. singular: fìast(a), fusti, fudiàst(a); plural: festis, fistis, fustis, fudistis).

(17) Cf. also the hearsay marker nachi < na(ra)t chi/nanchi < na(ra)n chi (Cruschina and Remberger 2008).

(18) Pinto (2011:137-44) treats these cases as backformations.

(19) §17.4.1 is mostly based on this chapter, with additions from Blasco Ferrer (1986) and Mensching (2004a; 2005; 2012).

(20) Only ‘one’ and ‘two’ among the cardinal numbers.

(21) Cf. Jones (1993:65-8), Floricic (2003), Mensching (2005), Mardale (2008).

(22) Cf. Remberger (2006a), and also §§49.3, 50.2; for the passive, see §

(23) The participle cannot be doubled if BE or HAVE are used as main verbs (cf. Pisano 2010b:129).

(24) Cf. Casti (2012) for a detailed overview of other periphrastic constructions.

(25) These impersonal constructions are possible not only with intransitive verbs, i.e. unaccusatives, but also with unergatives as in (18a), but not with transitives.

(26) For the basic properties of the clause, we refer to Jones (1993) with some additions from Blasco Ferrer (1986).

(27) As in many mostly spoken varieties, canonical passive constructions are not very frequent even in Sardinian (cf. Loporcaro 1998b:43 fn.7 and references therein).

(28) For this section, see Blasco Ferrer (1986:195-202); Jones (1993:251f., 249, 291-3, 296, 305-8); Manzini and Savoia (2005, I:452-69); Mensching (2004a:55, 72, 84); Damonte (2006b); Mensching (2012).

(29) For the following discussion of the clausal properties of the infinitive construction and the behaviour of clitics, see Jones (1993:137, 143, 271).

(30) For discussion of personal and inflected infinitives from a pan-Romance view, see Mensching (2000); for detailed description of the Sardinian structures, see Jones (1992; 1993:270-82).