The evolution of Lat. ILLUM in old Veronese: apocope and related phenomena* I testi in Veronese antico presentano una notevole varietà di esiti dal lat. illum dovuta all’interazione di numerosi processi fonologici. In questo contributo mi soffermerò principalmente sull’apocope, discutendo poi alcuni fenomeni ad essa collegati: la prostesi e la degeminazione. Nel par. 2 discuterò le condizioni di estensione della regola di apocope, che, inizialmente limitata alle parole toniche, si estende al clitico lo a partire dai casi di enclisi o quando lo si trova in coda di un nesso clitico (es. me lo → mel’). Mostrerò come tale estensione del dominio dell’apocope sia facilmente spiegabile nel quadro della teoria prosodica di Selkirk 1995, mentre una gerarchia prosodica esaustiva e simmetrica come quella proposta da Nespor/ Vogel 1986 fatichi a cogliere alcune rilevanti asimmetrie fra enclisi e proclisi. Alla luce di tale impostazione nel par. 3 discuterò una proposta sull’origine della prostesi (l → el) proponendo che la prostesi intervenga per prevenire il disallineamento fra struttura prosodica e sintattica provocato dall’apocope di lo: qualora essa si applichi ad un pronome proclitico (non in coda di nesso), questo diventa infatti prosodicamente enclitico alla parola precedente: *tuti-l diso → tuti el diso (‘tutti lo dicono’). Infine, nel par. 4 discuterò alcune alternanze dovute alla presenza sporadica di forme con ll, per es. elo vs ello. Cercherò di argomentare che le grafie con ll rappresentassero una consonante geminata osservando che dopo ll l’apocope non è mai ammessa. 1. Introduction Old Veronese 1 exhibits a rich set of pronominal forms deriving from Latin illum via several phonological rules like apocope, degemination and prosthesis. In this article I will mainly focus on apocope, i. e. loss of final vowel, showing how a principled analysis of apocope can shed light on further related aspects like the distribution of prosthesis and degemination of -ll-. The facts reported below are not expected to challenge previous descriptions of the morphology of personal pronouns and definite articles; rather, the present work aims to improve our knowledge of the nature of some phonological processes and their impact on morphology. I will start by revising Vanelli’s 1992, 1998 analysis of apocope in the light of Selkirk’s 1995 account of the prosodic status of clitics. I will show that Selkirk’s * I would like to thank Mair Parry, Paola Benincà and the anonymous reviewer of this paper for having provided many helpful comments, which led to a much improved version. This work was supported by a Marie Curie fellowship on «Pronouns of medieval Italian dialects» (call 2009, n. 253332) and a FIRB grant (call 2008, n. RBFR08KR5A) on «A grammatical survey of Italian dialects: fieldwork, data management, and linguistic analysis». 1 I will refer here to the dialect spoken in Verona (Veneto region, North-Eastern Italy), as attested in 13 th -14 th century documents. Vox Romanica 70 (2011): 63-78 Diego Pescarini asymmetric account captures satisfactorily some enclitic vs proclitic asymmetries that otherwise remain unaccounted for under Nespor/ Vogel’s 1986 prosodic analysis of clitic elements. Then, in section 3, I will argue that the apocopated form l undergoes prosthesis (e. g. l → el) in order to prevent a syntax/ prosody misalignment (McCarthy/ Prince 1993) due to the syllabification of a proclitic element with the preceding prosodic word. Lastly, in section 4, I will address the phonological status of -llin strong forms like ello, ell’ (‘he’) and its feminine/ plural counterparts, arguing that in 13 th century Veronese ll still counts as a geminate and so prevents the following -o from undergoing apocope. The examples cited in this article are from Giacomino da Verona’s De Ierusalem Celesti and De Babilonia Civitate Infernali (ed. by Contini 1950/ I: 627-52). As the present work aims to give a synchronic analysis by reconstructing the linguistic competence of a 13 th century speaker, I prefer to provide here a comprehensive set of intra-textual variants, rather than a constellation of examples from various texts. Nevertheless, every generalization results from the analysis of a vast corpus, including the Old Veronese texts that are part of the Opera del Vocabolario Italiano 2 digital database and the documents collected by Bertoletti 2005. General differences in the distribution of apocope have been observed within such a corpus by Stussi 1992 and Bertoletti 2005: 116; nevertheless the impact of apocope and related processes on the morphology of pronominal forms and definite articles is quite stable across different textual types and chronological stages (see also Bertoletti 2005: 212-14, 226-27): on the one hand, reflexes of Lat. illum commonly undergo apocope even in prose and in later texts and, on the other hand, apocope is always optional, even in poetry. Quantitative differences between poetry and prose, in particular after non sonorant consonants, might therefore depend on stylistic and metrical factors, but the contextual restrictions, the nature, the origin and the evolution of the rule can be clearly detected independently from the format of the source. 2. Apocope Northern vernaculars, including Tuscan varieties, exhibit cases of apocope, namely loss of final vowels. The process originated after single postonic sonorants, which were reanalysed as codas of the preceding syllable (e. g. co.re → cor.e - ‘heart’). In Tuscan and central Veneto vernaculars apocope is allowed only in this original context and targets only -o and -e (when the latter is not a feminine plural ending), while in Gallo-italic varieties apocope has extended to other phonological contexts (i. e. after any kind of consonant) and to any final vowel but -a. 64 2 http: / / www.lib.uchicago.edu/ efts/ ARTFL/ projects/ OVI/ . I refer the interested reader to the OVI website for the list of the documents in old Veronese contained in the database. The evolution of Lat. illum in old Veronese Old Veronese exhibits an intermediate pattern: it frequently exhibits apocope after single sonorants - even in those texts wherein apocope is on average less frequent - and, furthermore, allows apocope after etymologically double sonorants (e. g. fradel ‘brother’) and after non-sonorant consonants (Rohlfs 1966: §143, §146; Riva 1953: §84, Bertoletti 2005: 116-37). In all northern vernaculars, including Fiorentino, apocope can target also reflexes of Lat. illum, in particular the subject pronoun elo ( → el ‘he’) and the clitic item lo ( → l ‘him’/ ‘the’), which expresses both the 3p.m.sg accusative pronoun and the m.sg definite article. As noticed by Vanelli 1992, 1998 3 , the extension of apocope - i. e. a word-level rule - to the clitic element lo originates from a narrow set of contexts: when it is enclitic (1a), or when lo follows an unstressed element like a dative clitic pronoun (1b) or a negative marker (1c): (1a) batando-l molto forto. (Babilonia 83) beating-him.cl very hard ‘beating it very hard’ (1b) ve’l poës cuitar (Ierusalem 238) to.you.pl.cl it.cl can tell ‘he can tell it to you’ (1c) cor no’l po’ pensar (Ierusalem 240) heart not it.cl can think ‘the heart cannot think it’ In Giacomino da Verona’s poems, for instance, in these contexts apocope is almost mandatory: there is a sole example of lo after a (benefactive) dative clitic pronoun (Babilonia 294: tu me lo concostasi ‘you conquered it for me’), otherwise the accusative clitic normally undergoes apocope, e. g. mel (‘to-me it’), tel (‘to-you it’), gel (‘to-him/ her/ them it’), no l (‘not it’), etc. After lexical words, on the other hand, apocope normally does not take place: (2) en un’aqua lo meto k’è de sì gran fredura (Babilonia 113) in a’water him.cl put.they that is of such great coldness ‘they put him in some water which is so cold’ Given this asymmetry, we can therefore suggest that apocope was originally allowed only when the element preceding lo was its phonological host, as in (1a), or another unstressed element, like in (1b) and (1c) 4 . 65 3 See also Renzi 1993, Renzi/ Vanelli 1993, Formentin 1996. Vanelli mainly deals with Friulian data. As Friulian is more conservative than other northern dialects, it allows Vanelli to trace the morphology of definite articles and object clitics back to its original stage. Although such a diachronic development cannot be documented in old Veronese, the synchronic distribution shown by 13th-century Veronese is symptomatic of the same evolution. 4 With respect to the conditions triggering apocope, Old Veronese is therefore representative of an intermediate chronological stage between Friulian on the one hand and Fiorentino on the Diego Pescarini The asymmetric distribution in (1) and (2), typical of northern vernaculars, can be captured by constraining the apocope rule within a specific prosodic domain, which, according to Vanelli 1992, 1998: 194, Pescarini 2009, is the Clitic Group (abbreviated as C). The Clitic Group is defined by Nespor/ Vogel 1986: 150, Hayes 1989 as the prosodic constituent that dominates one or more Prosodic Words (PrW) and is dominated by the phonological phrase (PPh): (3) Given this definition, apocope takes place if the element preceding the clitic pronoun is part of the same clitic group, as represented below (where X stands for every kind of element, either lexical or functional): (4) Apocope is therefore triggered if X is either a verb (e. g. batando-l, as in (1a)) or another clitic element (e. g. ve’l poës, no’l po’, as in (1b) and (1c) respectively). In the latter case both clitic elements are followed by their host verb, as shown in the diagram (5), which can be regarded as a variant of the basic condition in (4). (5) The prosodic domain in (4), however, is rather problematic for both theoretical and empirical reasons. First, it is rather uncommon that a phonological rule targets the second element within a prosodic constituent. Second, the condition in (4) can- 66 other (Vanelli 1992, 1998). 14th century Friulian differs from Fiorentino and Veronese as apocope is allowed only after prepositions and other clitic pronouns, while apocope began to be allowed after other functional word (like conjunctions) from the end of the 14th century. In Fiorentino, on the contrary, apocope is almost always allowed from its early documented stage. PPh C C PrW PrW PrW porta-lo tu (Modern Italian) bring-it you ‘bring it’ C PrW PrW X l(o) (apocope: prosodic domain) C PrW PrW PrW ve/ no l V The evolution of Lat. illum in old Veronese not follow - neither logically nor chronologically - from the original rule of apocope, namely a word-level rule targeting lexical elements even if they are the sole item within C, as exemplified below: (6) Moreover, the account in (4) is falsified by observing sequences formed by an infinitive and the enclitic lo, like the one represented in (7). As both elements are candidates for apocope 5 and, according to (4), one would expect apocope to target the second element of the clitic group, as shown below: (7) However, as marked by the asterisk, this pattern is never attested. On the contrary, it is the verb that normally undergoes apocope, e. g. far(-o)lo, while the final vowel of the clitic does not drop: (8) far-lo (Old Veronese) make-it ‘make it’ In order to account for these problems I will revise the analysis of apocope by assuming an alternative prosodic hierarchy suggested by Selkirk 1995. Selkirk departs from Nespor/ Vogel 1986 in claiming that Prosodic Words (PrW) are immediately dominated by the Prosodic Phrase (PPh) without any intermediate C node. The comparison between Nespor and Vogel’s and Selkirk’s hierarchies is provided below: 67 5 Not all infinitive verbs are subject to apocope: if the etymological penultimate syllable is unstressed, its nucleus is dropped (syncope) and, as a consequence, the final vowel cannot undergo apocope because it is no more preceded by a single intervocalic sonorant, e. g. mittere metro → metr(*o). See Bertoletti 2005: 130-33 for an in-depth analysis of the interaction of syncope, apocope, and cliticization. C PrW far(o) (Old Veronese) to-make ‘to make’ C PrW PrW faro l*(-o) (Old Veronese) Diego Pescarini (9a) (9b) In rejecting the C node, Selkirk argues that clitics are generated as extrametrical syllables (i. e. syllables not governed by a metrical foot), daughter to a recursive PrW: (10) In what follows I will argue that the asymmetrical configuration above captures the behavior of Italo-Romance clitics in a better way than the symmetrical model proposed by Nespor/ Vogel 1986, wherein clitic and non-clitic elements are daughters to the same prosodic constituents. In particular, the hierarchy in (10) can capture the historical evolution of the process, which originally targeted the embedded PrW (namely, the lexical element) and later on expanded to the upper PrW, which contains the clitic element. Such an explanation accounts straightforwardly for the noted asymmetry between proclitic lo, which cannot originally undergo apocope (see (2)), and enclitic lo, whose final vowel can drop (see (1)). (11a) (11b) If every clitic - both proclitic and enclitic - was directly dominated by a Prosodic Word 6 (as assumed by Nespor/ Vogel 1986), apocope would be allowed symmetri- 68 6 Nespor/ Vogel 1986: 147-48 argue that «the segmental rule of Intervocalic s-Voicing, which has the phonological word as its domain of application . . ., does not apply across the juncture between a clitic and a word, as exemplified in (5) below. (5a) lo [s]aluto *[z] ‘(I) greet him’ (5b) essendo[s]i salutati *[z] ‘having greeted each other’ . . . A second rule, Raddoppiamento Sintattico, whose domain of application is larger than the word . . ., also applies to a clitic . . . Selkirk 1995 PPh PrW Nespor/ Vogel 1986 PPh C PrW PrW σ PrW clitic element lexical element l*(o) dona dona-l(o) The evolution of Lat. illum in old Veronese cally 7 and, as a consequence, every clitic pronoun would be subject to apocope independently from its position, as schematized in (12): (12) Furthermore, Selkirk’s asymmetrical analysis accounts for sequences formed by an infinitive - ending with -ro - followed by lo. As previously observed, both these elements are candidates for apocope, which would apply symmetrically if they were PrW daughter to the same prosodic categories, see (7). The data, on the contrary, show that apocope applies asymmetrically, as apocope targets the verb (e. g. faro → far ‘to make’), while the following clitic cannot undergo apocope: (13) The patterns above follow from Selkirk’s asymmetric configuration, wherein apocope applies cyclically targeting the most embedded PrW first. If apocope of the embedded (lexical) word takes place (e. g. faro → far), it prevents apocope from targeting the outer PrW - the one ending with the clitic lo - because the dominating PrW does not exhibit the appropriate conditions for apocope: as a matter of fact, after the first cycle of apocope, the final vowel of the outer PrW is preceded by a consonant cluster, which blocks apocope, e. g. farl(*o). 69 (6a) da[m: ]i ( da mi) ‘give me’ (6b) amò[l: ]o ( amò lo) ‘(he/ she) loved him’ These facts are a positive indication, according to the clitichood test under examination, that Italian clitic pronouns are independent phonological words.» In my opinion, these arguments are rather flawed, as both show that there is a word-boundary between the clitic and the preceding/ following word, but it does not entail that the clitic itself has a PrW status. Furthermore, it is worth noting that s-voicing does not apply even across word-internal boundaries, as shown in Nespor/ Vogel 1986: 127: (i) anti[s]ociale *[z] ‘antisocial’ 7 Counterexamples, i. e. cases wherein apocope is allowed, will be discussed in the next section. C PrW PrW l*(o) vede wrong prediction PrW PrW σ far(*o) l*(o) → [farlo] vs *[farol] Diego Pescarini If, on the other hand, inner apocope cannot take place, outer apocope is free to apply, so that the final vowel of the enclitic pronoun is dropped, as shown by sequences formed by a gerund and a following proclitic: (14) Incidentally apocope of the inner PrW is blocked when the infinitive has undergone syncope of the penultimate unstressed syllable as in the case of metro mittere (‘put’, cf. Bertoletti 2005: 131). In these cases apocope cannot take place as the final vowel is not in the appropriate context and, consequently, the following enclitic might undergo apocope (e. g. metro-lo → metro-l ‘to put it’). Unfortunately, however, sequences of a syncopated verb followed by a 3p.sg.m pronoun are not attested in old Veronese. Selkirk’s proposal, lastly, turns out to provide a promising account of the prosodic status of clitic clusters. In fact, when two proclitics co-occur, the latter always undergoes apocope, as shown in (1b-c), repeated here as (15a-b): (15a) ve’l poës cuitar (Ierusalem 238) to.you.pl.cl it.cl can tell ‘he can tell it to you’ (15b) cor no’l po’ pensar (Ierusalem 240) heart not it.cl can think ‘the heart cannot think it’ On the basis of the assumption that apocope targets the right edge of PrWs, we can therefore claim that a combination of two proclitic elements form an autonomous PrW, as schematized by the following diagram in (16). (16) It has already been suggested that clitic elements, though originating as extrametrical syllables, are then reorganized under a metrical foot when clustered together. This accounts for Italo-romance varieties, like Neapolitan, that assign stress to enclitic elements when they occupy the penultimate position of the outer PrW (Peperkamp 1997, Loporcaro 2000). According to (16), northern vernaculars like Old Veronese seem to witness a further evolution, as (pro)clitic clusters behave like an autonomous PrW which can finally undergo word-level processes like apocope. 70 PrW PrW → PrW batando l(-o) batando-l PPh PrW PrW ve-l(-o) poës to.you-it can The evolution of Lat. illum in old Veronese 3. Prosthesis In northern and Tuscan vernaculars, the apocopated clitic l can undergo a further process of prosthesis, giving rise to the forms il/ el (the quality of the prosthetic vowel is subject to cross-linguistic variation, see Renzi 1993, Sampson 2009: 15-8). The diffusion of prosthesis varies across medieval vernaculars: in Old Florentine the distribution of il is rather free, Old Friulian does not exhibit any prosthetic form (Vanelli 1992, 1998), while Old Veronese shows traces of el since its early attestations, as in the following example from a 13 th century document: (17) co(n)tra el sindico d(e) Valeço (Doc. Ver. 1265-1267, Stussi 1992: 264) against the mayor of V. ‘against the mayor of V.’ Giacomino does not exhibit prosthetic articles, but the form el is sporadically used as an object clitic, like in the following examples: (18a) la scriptura el diso (Ierusalem 63, 196) the scripture it.cl says ‘so scripture says’ (18b) li sancti tuti el diso (Babilonia 34) the saints all it.cl say ‘all the saints say that’ (18c) enanço k’eli el meta en logo de calura. (Babilonia 115) before that’they it.cl put in place of warm ‘before they put him in a hot place’ Prosthesis can be intuitively regarded as a strategy repairing marked syllabic configurations, which arise as a consequence of vowel deletion processes like apocope. In Gallo-italic dialects, wherein apocope determined a systematic loss of final vowels for -a, object clitics are reduced to single consonants like m ( me, ‘me’), t ( te, ‘you.sg’), l ( illum, ‘him’), etc., which syllabify with the preceding subject clitic (which is often expressed by a vocalic segment, see Vanelli 1984, 1998: 91- 104) or the following verb, if it begins with a vowel. Otherwise, when neither of these conditions holds, the clitic is syllabified by means of a prosthetic vowel. For instance, in modern Torinese, a prosthetic vowel a is inserted when the object clitic follows the 2 nd person subject clitic, which is not expressed by a vocalic exponent: e. g. *it m (‘you to-me’) → it am. (19) It am das an pum. (Torinese, Vanelli 1984, 1998: 103) You.cl to-me.cl give an apple ‘You give me an apple’ However, in other Italian vernaculars like Old Veronese prosthesis cannot follow from syllabic constraints as apocope is allowed only after a vowel and, as a conse- 71 Diego Pescarini quence, l is free to syllabify with the preceding vowel. Vanelli 1992, 1998: 196-97 suggests that in these cases prosthesis of clitic elements originates from the tension between apocope and the co-occurring rule of prevocalic elision: «[l]a nostra ipotesi è che l’introduzione di il nel lessico delle lingue in questione si possa interpretare come un dispositivo messo in atto per ‹sanare› questa ambiguità tra l’unicità del contenuto segmentale coinvolto nel processo e la duplicità delle regole dovuta alle diverse condizioni prosodico-contestuali.» In this section I will discuss an alternative explanation, arguing that in these vernaculars prosthesis is triggered by an alignment condition (McCarthy/ Prince 1993) in order to prevent the proclitic pronoun/ article l from becoming prosodically enclitic to the preceding word. As a matter of fact, prosthesis is originally attested in contexts wherein apocope has exceptionally targeted single proclitic elements that follow stressed elements like complex prepositions (as in (17)), lexical words (as in (18a) and (18b)) or strong subject pronoun (like in (18c)). In these cases lo occupies a PrW-initial position, represented in (20) and, therefore, apocope is not expected because it is usually allowed only in PrW-final position: (20) The prosodic configuration resulting from such a non-canonic apocope is shown in (21), wherein the proclitic lo syllabifies with the preceding lexical words and thus ends up becoming part of the preceding Prosodic Phrase: (21) Given the distribution of the form el, we can therefore claim that prosthesis takes place in order to avoid this very configuration, which exhibits a syntax/ prosody misalignment due to the position of l. Since the proclitic syllabifies with the preceding element, the right edge of the first PPh is not aligned with the right edge of the corresponding syntactic phrase, we can therefore suppose that such a misalignment violates a generalized alignment constraint that, following McCarthy/ 72 PPh PPh PrW PrW σ PrW σ PrW la scriptura l(o) diso ‘the scripture it says’ PPh PPh PrW σ PrW PrW la scriptura-l diso ‘the scripture-it says’ The evolution of Lat. illum in old Veronese Prince 1993, can be expressed as follows (read: the Right edge of a syntactic phrase (XP) coincides with the Right edge of a prosodic phrase (PPh)): (22) Align (XP, R, PPh, R) Since the configuration (21) produces a violation of (22), prosthesis is therefore triggered to re-align the prosodic and syntactic boundaries. In other words, while non-canonical apocope makes the proclitic l become phonologically enclitic to the preceding element, as shown in (23a), prosthesis ends up repairing such a misalignment, bringing the accusative pronoun back to its original constituent, cf. (23b): (23a) [la scriptura] PPh [lo diso] PPh → [la scriptura l] PPh [diso] PPh [the scripture] PPh [it.cl says] PPh [the scripture it.cl] PPh [says] PPh (23b) [la scriptura l]PPh [diso]PPh → [la scriptura]PPh [el diso]PPh [the scripture it.cl]PPh [says]PPh [the scripture]PPh [it.cl says]PPh This hypothesis rightly predicts the absence of prosthesis in enclisis, ruling out cases like (24), where apocope cannot cause any misalignment because the object clitic is on the right edge of the Prosodic Word (see also Loporcaro 1997). (24) batando-l → batando-(*e)l beating-it.cl beating-it.cl Second, this analysis accounts straightforwardly for the absence of prosthesis after unstressed elements like negation markers: although it is a prototypical environment triggering apocope - see (1c) - prosthesis is in fact rarely found after negation because the negative marker avoids the misalignment of l as it occupies the rightmost position within the PPh: (25) [no (*e)l pò] PPh not it.cl can ‘he cannot (do) it’ In conclusion, while previous studies - Vanelli 1992 a.o. - have ascertained that prosthesis is a consequence of apocope, in this section I have discussed a possible explanation concerning the nature of the process, arguing that prosthesis is a repair strategy triggered when apocope exceptionally produces a violation of the alignment constraint in (22). 4. A remark on degemination Old Veronese displays a puzzling alternation between subject forms with and without geminate l (e. g. ello, ella, ell’ vs elo, ela, el’), which, in some cases, co-occur within the same sentence. 73 Diego Pescarini (26) E quand ell’è al caldo, al fredo el voravo esro (Babilonia 116) And when he is at.the warm, at.the cold he would be ‘And when he is in the warm, he wants to be in the cold’. First of all, such alternations do not seem to correlate directly with any syntactic parameter like clause typing (e. g. matrix vs subordinate) or the position of the pronoun in the clause (e. g. dislocated, focalized, etc.). As a consequence, l/ ll does not seem to express a contrast between a strong and a weak pronominal series, nor can monosyllabic pronouns (e. g. el) be considered clitic elements, as they can be separated from the inflected verb (pace Bertoletti 2005: 223): (27) quand el tanti dïavoli se vé corir da provo, (Babilonia, 193) when he many devils to.himself.cl sees run of near ‘when he sees many devils running close to him’ Second, it is worth noting that the phonological status of -llin medieval vernaculars is rather controversial, as several scholars - since Stussi 1965 - claim that ll is a complex orthographical sign standing for / l/ . Also Bertoletti 2005: 200, who convincingly argues that 13 th -14 th century Veronese still exhibited postonic geminates 8 , writes that «[p]er quanto riguarda la laterale, i raddoppiamenti arbitrari caratteristici di questa lettera e dovuti probabilmente ad un fatto d’esecuzione grafica impediscono di trarre alcun elemento di prova dalle forme che la contengono». If we focus on the early documents, however, the etymological conditions seem respected, since ll can be normally found only as a reflex of ll, cf. castel(l)o, fradel(l)o, while *cello caelum ‘sky’, *angello angelum ‘angel’. In what follows I will argue that such an intuition is confirmed by observing the distribution of apocope. As noticed by Zamboni 1976, apocope was originally sensitive to the length of the preceding consonant: in Venetian apocope is therefore mandatory after -l-, e. g. miel vs *miele mele (‘honey’), while it is not allowed after -ll-, e. g. cae 9 vs cal* calle (‘narrow street’). Such a synchronic distribution follows from the chronological order of the phonological processes, assuming that apocope took place before degemination. Veronese slightly differs from Venetian as it does not show such a conservative pattern: apocope is in fact found with degeminated words as well, e. g. fradel ‘broth- 74 8 The evolution of Old Veronese is therefore consistent with the pattern of gemination found in modern northern dialects, in which either the original postonic geminates have been maintained (e. g. in several Apenninic dialects, cf. Loporcaro et al. 2005) or gemination has been generalized to every postonic consonant (cf. Loporcaro et al. 2006 on the Alpine dialect of Soglio) with the remarkable exception of postonic liquids, which still display length contrasts (Loporcaro et al. 2006: 601). Early Old Veronese seems consistent with such an evolution (cf. Bertoletti 2005: 191-200) and, also in the case of liquids, the orthography is more or less consistent with the etymological conditions, i. e. ll ll/ l; l l. 9 In modern Venetian intervocalic / l/ is pronounced as a non-syllabic [e]. The evolution of Lat. illum in old Veronese er’, caval ‘horse’, castel ‘castle’ (vs Venetian fradeo, cavao, casteo). This means that in an early diachronic stage of Veronese both degemination and apocope were active rules and, furthermore, degemination fed apocope by increasing the number of words subject to the latter, e. g. castello castelo castel. The early documents seem representative of this chronological stage, as the orthographic variants are compatible with the optional application of both rules. We can therefore find forms 1) without degemination and without apocope, e. g. castello; 2) with degemination and without apocope, e. g. castelo; 3) with degemination and apocope, e. g. castel. As a matter of fact, the fourth combination (namely, words with ll and with apocope) is never attested. In Giacomino and other Old Veronese documents, for instance, words that are written with geminate ll are always followed by -o, e. g. cristallo (‘crystal’), millo (‘thousand’), cortello (‘knife’), quello (‘that.m’), flagello (‘disaster’), cativello (diminutive of ‘bad’), bello (‘nice’), while they normally display apocope if the preceding sonorant is written as a single consonant (e. g. cristal, cortel, flagel, bel, etc.). Such an asymmetry is consistent with the historical conditions on apocope, as reconstructed by Zamboni 1976, while it remains unaccounted for under a mere orthographical explanation: if ll was just a complex sign for / l/ , nothing would prevent it from occurring in word-final position (e. g. *castell). The same argument can be extended to the morphology of personal pronouns, which, unlike lexical words, exhibit an orthogonal morpho-phonological process of prevocalic elision. Apocope and elision do therefore determine the same output (i. e. deletion of the final vowel), but are sensitive to different contexts. Crucially, in preconsonantal context -o can drop only as a consequence of apocope. Before consonants, Old Veronese displays only three m.sg subject forms: ello, elo, el 10 , as shown by the examples in (28), while the fourth allomorph, ell, is found only before vowels 11 , as in (29): (28a) li castegi e le roche k’ello lagà l’altrer (Babilonia 270) the castles and the fortresses that he left the other-yesterday ‘the castles and fortresses he left the day before yesterday’ (28b) La clarità è tanta k’elo reten en si (Ierusalem 83) the brightness is much that he keeps in himself 75 10 I am not considering here the etymologically oblique form luy ( illui), which is attested in subject position since the 14 th century (Bertoletti 2005: 223), and the aphaeretic subject pronoun lo ( illum), whose distribution in Old Veronese - and in northern vernaculars in general - has been addressed by Bertoletti 2009: 51. 11 Furthermore, the texts from Lombardia, Veneto, Emilia, and Liguria stored in the OVI database contain 440 instances of ell. 430 out of 440 precede be/ have. Such a distribution deserves an in-depth analysis taking into consideration, among other factors, the type of the verb (auxiliary vs lexical) and the accentual status of its first syllable. Diego Pescarini (28c) enançi k’el mora. (Babilonia 23) before that he dies ‘before he dies’ (29) Ell’è vero e certo (Ierusalem 207) It is true and sure ‘It is true and sure’ Thus, the distribution of the four allomorphs is not completely free as none of the 41 occurrences of ell attested in the OVI collection of Old Veronese texts is before a consonant, as summarised in the following table (where √ stands for an attested form and * for a non-attested one). (30) elo el ello ell before C √ √ √ * Again, such a gap cannot follow from the hypothesis that ll is an orthographic sign standing for / l/ . Rather, this asymmetry means that ello, unlike elo, still displayed a geminate sonorant preventing the following vowel from undergoing apocope, in accordance with Zamboni 1976. 5. Conclusions In this article I have explored some phonological processes responsible for the morphology of the 3.m.sg. pronoun in old Veronese (and, tangentially, of the m.sg definite article). Firstly, on the basis of Selkirk 1995, I argued that apocope extends to clitic pronouns when they occupy a final position within a recursive Prosodic Word. This led to a revision of Vanelli’s 1992, 1998 analysis in order to account for some enclitic vs proclitic asymmetries in the distribution of apocope. Second, I claimed that prosthesis aims to re-align a proclitic object after apocope has made it become prosodically enclitic to the preceding element: if apocope takes place on the left edge of a Prosodic Phrase the proclitic pronoun l ends up becoming prosodically enclitic to the preceding element. Then prosthesis repairs such a misalignment by bringing the object clitic back to its original Prosodic Phrase, as shown in (31b). (31a) [la scriptura] PPh [lo diso] PPh → [la scriptura l] PPh [diso] PPh [the scripture] PPh [it.cl says] PPh [the scripture it.cl] PPh [says] PPh (31b) [la scriptura l] PPh [diso] PPh → [la scriptura] PPh [el diso] PPh [the scripture it.cl] PPh [says] PPh [the scripture] PPh [it.cl says] PPh 76 The evolution of Lat. illum in old Veronese Thirdly, I argued that forms with ll, e. g. ell(o), were not trivial orthographic variants of el(o), but that old Veronese still shows traces of geminate sonorants. In particular, I have shown that after ll apocope of -o is not allowed and, as a consequence, the nominative form ell cannot be found before a consonant. Padua/ Bristol Diego Pescarini Bibliography Bertoletti, N. 2005: Testi veronesi dell’età scaligera, Padova Bertoletti, N. 2009: Veronese antico. Nuovi tesi e vecchie discussioni, Padova Contini, G. (ed.) 1950: Poeti del Duecento, Napoli Formentin, V. 1996: «Alcune considerazioni e un’ipotesi sull’articolo determinativo in area italoromanza», in: L. Lugnani/ M. Santagata/ A. Stussi (ed.), Studi offerti a Luigi Blasucci dai colleghi e dagli allievi pisani, Lucca: 257-72 Formentin, V. 2002: «Antico padovano ‹gi› da illi: condizioni italo-romanze di una forma veneta» Lingua e Stile 37/ 1: 3-28 Hayes, B. 1989: «The Prosodic Hierarchy in meter», in: P. Kiparsky/ G. Youmans (ed.), Rhythm and Meter, Orlando (FL): 201-60 Loporcaro, M. 1997: «On Vowel epenthesis in Alguer Catalan», in: P. M. Bertinetto/ L. Gaeta/ G. J. Jatchev/ D. Michaels (ed.), Certamen Phonologicum III, Turin: 205-27 Loporcaro, M. 2000: «Stress stability under cliticization and the prosodic status of Romance clitics», in: L. Repetti (ed.), Phonological Theory and the Dialects of Italy, Amsterdam: 137-68 Loporcaro, M./ Paciaroni, T./ Schmid, S. 2005: «Consonanti geminate in un dialetto lombardo alpino», in: P. Cosi (ed.), Misura dei parametri. Aspetti tecnologici ed implicazioni nei modelli linguistici, Brescia: 579-618 Loporcaro, M./ Delucchi, R./ Nocchi, N./ Paciaroni, T./ Schmid, S. 2006: «La durata consonantica nel dialetto di Lizzano in Belvedere (Bologna)», in: R. Savy/ C. Crocco (ed.), Analisi prosodica. Teorie, modelli e sistemi di annotazione, Torriana (RN): 491-551 Martinet, A. 1955: Économie des changements phonétiques, Bern McCarthy, J./ Prince, A. 1993. «Generalized alignment», Yearbook of Morphology 1993: 79-154 [also in: Rutgers Optimality Archive 7: http: / / roa.rutgers.edu] Nespor, M./ Vogel, I. 1986: Prosodic phonology, Dordrecht Peperkamp, S. 1997: Prosodic Words, Den Haag Pescarini, D. 2009: «Note sulla morfologia dei clitici oggetto del Veronese antico», LabRomAn 3: 73-80 Renzi, L. 1993: «Da dove viene l’articolo il», in: J. Kramer/ G. A. Plangg (ed.), Verbum Romanicum. Festschrift für Maria Iliescu, Hamburg: 215-30 Renzi, L./ Vanelli, L. 1993: «Storia e struttura dell’articolo italiano il», in: Actes du XX e Congrès International de Linguistique et Philologie Romanes, vol. 3, Zürich: 293-305 Riva, F. 1953-1954: «Storia dell’antico dialetto di Verona», in: Atti dell’Accademia di Agricoltura, Scienze e Lettere di Verona 6/ 3-4 Rohlfs, G. 1966: Grammatica storica della lingua italiana e dei suoi dialetti, vol. 1, Fonetica, Torino Sampson, R. 2009: Vowel Prosthesis in Romance, Oxford/ New York Selkirk, E. 1995: «The prosodic structure of function words», in: J. Beckman/ L. Walsh Dickey/ S. Urbanczyk (ed.), Papers in Optimality Theory, Amherst (MA): 439-70 [also in: J. L. Morgan/ K. Demuth (ed.), Signal to Syntax: Bootstrapping from Speech to Grammar in Early Acquisition, Mahwah (NJ): 187-214] Stussi, A. 1965: «Sui fonemi del dialetto veneziano antico», ID 28: 125-42 Stussi, A. 1992: «Testi in volgare veronese del Duecento», Italianistica 21: 247-67 77 Diego Pescarini Vanelli, L. 1984: «Pronomi e fenomeni di prostesi vocalica nei dialetti italiani settentrionali», RLiR 48: 281-95 [republished as Ch. 3 of Vanelli 1998] Vanelli, L. 1992: «Da lo a il: storia dell’articolo definito maschile singolare in italiano e nei dialetti settentrionali», RID 16: 29-66 [republished as Ch. 8 of Vanelli 1998] Vanelli, L. 1996: «Convergenze e divergenze nella storia del pronome e dell’articolo: esiti di illu(m) nei dialetti italiani settentrionali», in: P. Benincà/ G. Cinque/ T. De Mauro/ N. Vincent (ed.), Italiano e dialetti nel tempo. Saggi di grammatica per Giulio C. Lepschy, Roma: 369-86 [republished as Ch. 10 of Vanelli 1998] Vanelli, L. 1998: I dialetti italiani settentrionali nel panorama romanzo, Roma Zamboni, A. 1976: «Alcune osservazioni sull’evoluzione delle geminate romanze», in: R. Simone/ V. Ugo/ G. Ruggiero (ed.), Studi di fonetica e fonologia, Roma: 325-36 78