Chapter 2. Velar allomorphy in Ibero-Romance

Paul O&#39;Neill

This article examines the origin and spread of the verbs which display a velar allomorph in Spanish and, in particular, the paradigmatic pattern which this allo-morphy exhibits. I draw upon Maiden's studies (1992, 2001b) of velar allomorphy in Italian and Spanish which define the different cells that display this allomorphy as an autonomous morphological structure, or morphome (Aronoff 1994) and I argue that the theoretical construct of the morphome offers a more convincing, rigorous and theoretically motivated account for the origin and spread of velar allomorphy in Spanish, as opposed to recourse to the word 'analogy'. However, I maintain that Maiden's conclusions for the spread of the velar in Italian are not entirely valid for Spanish since they predict the wrong result. I argue that the explanations offered by Maiden need to be nuanced, along with the concept of the morphome itself. More specifically, I argue that in Spanish the velar allomorph should not be considered as part of the lexical root but rather a semantically vacuous formative which is an allomorph of the verbal 'endings'. I present evidence from within the history of Spanish to support this hypothesis. Resumen Este trabajo examina el origen y la extensión de aquellos verbos que presentan un alomorfo velar en el verbo español (tengo, pongo, salgo) y se hace especial hincapié en la distribución paradigmática de este alomorfo. Hacemos referencia a los estu-dios de Maiden (1992, 2001b) sobre el español y el italiano en los que se define el conjunto de celdas paradigmáticas en las que aparece el alomorfo velar como una estructura puramente morfológica, o un morfoma (Aronoff 1994). Sostenemos que la noción teórica del morfoma ofrece una explicación más convincente y rigurosa para el origen y la extensión del alomorfo velar; además esta explicación tiene una justificación más teórica y no simplemente recurre a la analogía. Defendemos, sin embargo, que las conclusiones de Maiden para el español no son de todo válidas ya que producen resultados equivocados y proponemos que se ha de matizar la tesis de Maiden junto con la del morfoma. Específicamente, defendemos que en el español el alomorfo velar no se analizaba como un alomorfo de la raíz léxica sino de las terminaciones verbales y a este elemento no se le atribuía semántica alguna. Presentamos datos diacrónicos del español para apoyar nuestra hipótesis.

In modern Catalan, there are a few verbs that display stem-vowel alternation in the present tense: a high stressed vowel alternates with a mid unstressed vowel, as in tinc-tenim, surt-sortim. In Barcelona Catalan, this pattern is residual, (and for the rounded vowels purely orthographic). In Valencian, it is more widespread, with more than 20 verb stems showing the alternation. In medieval Catalan, it was more widely attested. I review the origin (mainly phonological) and development (extensively analogical) of this alternation. I show how this pattern relates to a similar phenomenon in Spanish (e.g. pido-pedimos, duermo-dormimos-durmamos) — a curious resemblance, since for the most part the conjugation patterns of the two languages are not at all similar. A substantial part of the paper focuses on establishing, for each language, the exact contexts in which metaphony, in both stressed and unstressed syllables was a regular sound change. The other major focus is to reconstruct in detail, for each paradigm model, on the basis of corpus evidence together with principles of morphological change (from Autonomous Morphology and Natural Morphology), the processes by which inherited stem allomorphs came to be redistributed or eliminated.

This article examines the concept of defectivity in the verbal system of Spanish, Portuguese and Catalan. Building on previous studies on defective verbs in Spanish and Portuguese I investigate why there are no such defective verbs in Catalan. I conclude that the structure of the verbal paradigm in Catalan non-first conjugation verbs is radically different from that of the other languages; Catalan verbs constitute paradigms which correspond to regular patterns of inflection whereas Spanish and Portuguese display non-predictable types of root allomorphy which require all non-first conjugation verbs to have a memorised form for the rhizotonic forms of the verb. Theoretically, this type of defectivity poses problems for models of inflectional morphology and suggests that the patterns of frequent verbs can become general rules for all verbs of a particular conjugation.

Paul O’Neill Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes. 1.1 Introduction Velar allomorphy refers to the inflectional forms of a lexeme which display a velar occlusive consonantal phoneme, either voiced /g/ or voiceless /k/, before the desinential endings of person and number and/or tense and mood; this allomorphy displays a particular paradigmatic patterning. The classical example of such velar allomorphy is the Spanish verb caer ‘fall’ in which the velar element occurs exclusively in the 1 SG present indicative and all of the present subjunctive. I follow Maiden (2004) and refer to this particular paradigmatic patterning as the L-pattern1, for reasons that will become apparent; the verbs which display this allomorphy are termed velar verbs. (1) 1SG 2SG 3SG 1PL 2PL 3PL 1SG 2SG 3SG 1PL 2PL 3PL Present Indic caigo caes cae caemos caéis caen Imp. Subj. (ra) cayera cayeras cayera cayéramos cayerais cayeran Imperative Present Subj. caiga caigas caiga caigamos caigáis caigan Imp. Subj. (se) cayese cayeses cayese cayésemos cayeseis cayesen infinitive Future caeré caerás caerá caeremos caeréis caerán Preterite caí caíste cayó caímos caísteis cayeron Gerund Conditional caería caerías caería caeríamos caeríais caerían Imp. Indic. caía caías caía caíamos caíais caían Participle The term ‘L-pattern’ has no semantic or phonological significance but rather was adopted by Maiden because in conventional representations of the verb paradigm on paper, the cells which comprise the L-pattern resemble a rotated letter 'L' 1 2SG 2PL cae caed caer cayendo caído Velar allomorphy has been of special interest to Romance linguists since with the exception of French, Rhaeto-Romance and the majority of Sardinian varieties, this velar allomorphy is attested in all other Romance languages. The different languages, however, vary as to (a) the paradigmatic distribution of this allomorphy and (b) the number of lexemes which display such allomorphy. Ibero-Romance is of particular interest in this respect since Portuguese displays only one velar verb, dizer ‘say’, whose allomorphy and paradigmatic distribution (the L-pattern) is entirely predictable on the basis of regular sound change. Modern Spanish contains approximately 155 velar verbal roots, excluding their derivatives, which display a voiceless velar allomorph in the L-pattern2 and 11 verbal roots 3, again excluding derivatives, with a voiced velar allomorph. These verbs have been of particular interest to historical linguists since the majority of velar allomorphy is non-etymological. In fact, of the voiced velar allomorphy the only lexeme whose allomorphy is etymologically expected is decir ‘say’. Catalan, along with Occitan, are the most bountiful languages with regards velar verbs since such verbs tend to be some of the most frequently used verbs in the language; according to Wheeler’s calculations (2011:184) for the Barcelona standard variety of Catalan there are 2 The full list of these verbs can be found in the Appendix. The majority of these verbs are denominal or deadjectival verbalizations e.g. humedecer ‘make humid’, entontecer ‘make stupid’, anochecer ‘to become dark/nightime’. This list of verbs also includes some frequent verbs not of this type, e.g. apetecer ‘to feel like’ parecer ‘seem’ and derivates thereof aparecer ‘appear’ etc. 3 decir ‘say’, hacer ‘do’, salir ‘go out’, valer ‘be worth’, poner ‘put’, venir ‘come’, tener ‘have’, caer ‘fall’, traer ‘bring’, roer ‘gnaw’, oír ‘hear’, plus their derivatives. Paul O’Neill approximately 80 verb roots 4 (316 lexemes) which display velar allomorphy, voiced and voiceless. In Catalan the distribution of this allomorphy extends beyond the L-pattern cells; it is also present in the synthetic preterite, the imperfect subjunctive and the gerund. This is illustrated in (2) for the verb moure ‘move’. (2) 1SG 2SG 3SG 1PL 2PL 3PL 1SG 2SG 3SG 1PL 2PL 3PL 2SG 2PL Present Indic Present Subj. Future moc mogui mouré mous moguis mouràs mou mogui mourà movem moguem mourem moveu mogueu moureu mouen moguin mouran Imp. Indic. Imp. Subj. (se) movia mogués movies moguessis movia mogués movíem moguéssim movíeu moguéssiu movien moguessin Imperative Participle Gerund mou mogut movent moveu Condicional mouria mouries mouria mouríem mouríeu mourien Preterite moguí mogueres mogué moguérem moguéreu mogueren Infinitive moure Additionally there are also 19 or 20 roots (101 to 113 lexemes taking into account derived forms) which display velar allomorphy in all the cells above with the exception of the past participle5 in which there occurs an irregular rhizotonic participle. According to Wheeler 4 Of these 80 roots, six roots vary as to whether they display the velar allomorph in all the cells; four roots display it only in the 1SG present indicative and all the forms of the present subjunctive and seven in the synthetic preterite, imperfect subjunctive and past participle only. Note also that the verb córrer, in the Barcelona standard, never exhibits a velar in the 1sg present indicative and the rhizotonic forms of the present subjunctive but can display one variably in the 1PL and 2PL present subjunctive only. In the other synthetic preterite, the imperfect subjunctive and the past participle, however, this verb consistently displays velar allomorphy. The verb escriure consistently inflects with the velar consonant in the L-pattern cells but in the other cells there is variation as to whether it possesses the velar allomorph or not. This, however, is not the case with the past participle in which the velar does not occur. 5 cf.. prendre ‘to take’, 1SG present indicative prenc, 1SG present subjunctive, prengui, 1SG preterite prenguí, 1SG imperfect subjunctive prengués, past participle pres. (2011:184) there are no verbs in which the velar allomorph is expected according to its modern distribution. The linguistic factors which determine the distribution of velar allomorphy in Romance has been a contentions matter for linguistic theory since some have proposed that the allomorphy is a matter of phonology (St. Clair and Park 1974, Bermúdez-Otero 2013, for Spanish; Burzio 2004 for Italian; Hualde 1992:410, Mascaró 1983, for Catalan) others have claimed that it is a matter of pure morphology (Maiden 2001, 2004, O’Neill 2011a:204-243, forthcoming(a), for Spanish; Maiden 2001, 2004, 2009, Pirrelli 2000:79f.;178-84, Pirrelli and Battista 2002, for Italian; Wheeler 2011, for Catalan) and others (Bybee 1985, for Spanish; Pérez-Saldanya 1995, for Catalan) have highlighted the importance of semantic notions of markedness and explained the paradigmatic distribution by recourse to the 1SG present indicative being classed as unmarked and forming the base upon which all the other more marked velar forms are derived. Moreover, both within and across the different theories as to what grammatical component is responsible for the distribution of the velar element, there is no consensus as to the morphological parsing of the velar consonant: some consider it to constitute part of the allomorphic root or stem (Maiden 2001, 2004, for Spanish; Roca Pons 1968:231, Mascaró 1983:155-194, Badia 1994:574-578, for Catalan) others an infix (Malkiel 1974 for Spanish; Perea 2002:597-601, Pérez Saldanya 1998:73-75, Viaplana 1984, Wheeler 1993:196 for Catalan), and others part of the desinential endings (Martín Vegas 2007:168, O’Neill 2011a:216-220, 2015), for Spanish). In what follows I summarise the reasons for considering the velar element to be morphologically parsed as part of the desinential endings in old Spanish and, for the allomorphy Paul O’Neill not being phonologically determined but morphologically determined. On the basis of previous studies (Maiden 2001, 2004, 2009; O’Neill 2011a, 2015, forthcoming(a)) I contend that the cells which display velar allomorphy constitute a morphome, termed the L-pattern morphome for Spanish, Asturian, Aragonese and the Catalan velar morphome for Catalan (Wheeler 2011:183). Morphomes are to be understood as a semantically heterogeneous collection of cells which historically were the locus of a large amount of allomorphy, usually on account of purely phonological factors. Once the phonological rules ceased to be a reality for the language, however, the allomorphy persisted in these cells and the paradigmatic distribution of the allomorphy itself became a grammatical reality for the language and a determinant of the distribution of the allomorphy. Thus, synchronically, the morphome could be construed as a grammatical generalisation regarding the distribution of a common morphological formative, be it the lexical root or the desinential endings (see also the notion of a template, Aske (1995)). With regards the morphome and velar allomorphy, I present evidence from a number of IberoRomance varieties which suggest that there can be diachronic and diatopic variation as to the parsing of the velar element. At times, it can be considered to be part of the lexical root, other times the evidence suggests it has been parsed as part of the desinence. Such variation and change as to the morphological parsing of this element is apparent when the L-pattern morphome enters into conflict with another morphome, the N-pattern morphome since the paradigmatic domains of both morphomes overlap and can be in competition. I contribute to the discussion of morphomes in diachrony in that, on the basis of comparative Ibero-Romance data I show that morphomes not only tend to converge upon a particular phonological exponent/characteristic (see Maiden 2004 and the principle of convergence 6) but also upon a particular transparent morphological structure or typology, and that this is subject to diachronic and diatopic change. Thus, in old Spanish I claim that velar allomorphy converged upon a morphomic structure whereby a root, common to other members of the paradigm, formed the basis of the L-pattern forms and this root underwent affixation of velar desinences (-go, -ga, -gas, -ga, -gamos, gades, -gan). Thus velar allomorphy formed part of the inflectional desinences and was the marker of a particular verbal class, a type of thematic vowel indicating the inflectional forms of a sub-paradigm, much the same way as in Modern Catalan (Perea 2002:597-601, Pérez Saldanya 1998:73-75, Viaplana 1984; Wheeler 1993). In later developments of the language, however, I claim that there was a change whereby the L-pattern morphome came to be a generalisation not about inflectional formatives but the distribution of lexical root allomorphy. I present data from other varieties of Ibero-Romance to support my argument that there can be variation and change regarding the morphological parsing of the velar element and variation and change as to what type of morphological structures morphomes can converge upon. Morphomes, I argue, reveal the ‘deep’ morphological structures (see Malkiel 1974:315) around which the verbal morphology of the different varieties of Ibero-Romance are structured. Diachronically, morphomes can condition morphological change and I argue that the 6 Maiden (2001, 2004) makes reference to an interesting characteristic of morphomes which he defines as their tendency towards convergence which refers to the historically attested fact that morphomes ‘tend over time to acquire certain common phonological characteristics across all verbs in which they occur.’(Maiden 2004). Paul O’Neill recognition of such purely morphological structures, coupled with an appreciation of the variation in the morphological parsing of the velar element can facilitate explanations for poorly justified instances of morphological change in the Ibero-Romance verb. The organisation of this article is the following: in 2. I briefly describe the origins of velar allomorphy in Ibero-Romance, in 3. I summarise the arguments for believing that the velar element forms part of the desinential endings in old Spanish and that it is morphologically conditioned; in 4. I present evidence which prompts the conclusion that in further developments in Spanish the velar element was analysed as part of the root; in 5. I introduce the N-pattern morphome; in 6. I present data from central Ibero-Romance varieties of how the L-pattern and N-pattern morphomes interact in support of the hypothesis that there can be variation and change as to the parsing of the velar element and that morphomes can converge on different morphological structures and phonological exponences. In 7. I advance an argument as to why an appreciation of the ideas developed in this article can help explain some seemingly conflicting instances of morphological change in the history of the Spanish verb. Finally, I offer some tentative views on the lack of velar verbs in Portuguese. 2 Brief explanation as to the origin of velar allomorphy in Ibero-Romance The source of the velar allomorphy is phonological since it is due to the absence of palatalisation of velar consonants before front vowels in opposition to its presence before back vowels. In the transition from Latin to Romance velar consonants in the Iberian Peninsula underwent a process of palatalisation and affrication exclusively before front vowels; this caused a phonemic split of both /k/ and /g/. In old Spanish, before back vowels these consonants remained velar (although /k/ was subject to voicing and /g/ to become an approximant, or was lost7); before front vowels velar occlusives were supposedly first palatalised and then pronounced as the affricates [d͡z] and [t͡z], depending on the voicing contexts (VĪCĪNU > vezino ‘neighbour’ vs. LOCĀLE > logar ‘place’, Penny 2002:66). Within the verb morphology, the verbs which met the criteria to be sensitive to these different phonetic contexts8 preserved their root-final velar consonants exclusively in the L-pattern cells since the desinences of these cells contained back vowels. The phonological outcome of the root final velar consonant in the non-L-pattern cells was not uniform but depended upon the phonetic context. When the velar consonant was preceded by a vowel (or by a rhotic consonant) it was subject to the normal process of affrication or voicing (see (3) and (4)). (3) examples from Latin DIC ‘I say’, REDUC ‘I reduce’ and ADDUC ‘I adduce’. 9 c.f. NEGĀRE > negar ‘deny’, LIGĀRE > liar ‘tie’ (Penny 2002:68) These verbs were those that (a) were not of the Latin 1st conjugation (b) did not contain etymological [j] and (c) whose roots ended in a velar consonant. 9 Other Latin verbs which could have shown this development were: CŎQ(U) ‘ cook’. 7 8 Paul O’Neill Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive 1SG digo diga redugo reduga adugo aduga 2SG dizes digas reduzes redugas aduzes adugas 3SG diz(e) 1PL dezimos diga digamos reduz(e) reduzemos reduga redugamos aduz(e) aduzemos aduga adugamos 2PL dezides digades reduzedes redugredes aduzedes adugades 3PL dizen digan reduzen redugan aduzen adugan (4) Examples from Latin SPARG > espargo ‘I scatter’ and ĔRIG > yergo ‘I erect’ Indicative Subjunctive Indicative Subjunctive 1SG espargo esparga yergo yerga 2SG esparzes espargas yerzes yergas 3SG esparz(e) esparga yerz(e) yerga 1PL esparzemos espargamos yerzemos yergamos 2PL esparzedes espargades yerzedes yergades 3PL esparzen espargan yerzen yergan When the velar consonant was preceded by a nasal consonant the outcome was /ɲ/ (see (5)); when preceded by a voiceless consonant (which are principally those verbs which were formed with the Latin ingressive infix –SC- /sk/-) the outcome was the voiceless affricate /ts/ (see (6) & Penny 2002:178f.). (5) Examples from Latin TANG ‘I touch’, CING ‘I surround’ and FRANG ‘I break’. 10 Other Latin verbs which could have shown this development are PING ‘paint’, FING ‘make up, touch’, *RINGO ‘snarl’, STRING ‘draw tight’, TING ‘wet, dye’. Also, attested in old Spanish with the velar allomorph are the verbs FRANG > frango ‘I break’ and IUNGO > ungo ‘I join, yoke’. As Martín Vegas (2007:170-171) has pointed out, in old Castilian, these latter two contained competing allomorphs in the non L-pattern cells: those ending in a palatal nasal and also ones ending in /nd͡z/; thus the forms uñe, uñes from the verb uñir ‘to yoke’ and frañe, frañes frañir ‘to break’ alternated with forms unze, unzes, unzir and franze, franzes, franzir respectively. This is due to the various results of the Latin consonantal sequence NG + front vowel, which could either result in 10 Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive 1SG tango tanga cingo cinga frango franga 2SG tañes tangas ciñes cingas frañes frangas 3SG tañe tanga ciñe cinga frañe franga 1PL tañemos tangamos ciñemos cingamos frañemos frangamos 2PL tañedes tangades ciñedes cingades frañedes frangades 3PL tañen tangan ciñen cingan frañen frangan (6) Examples Classical Latin COGN SC ‘I know’ and late[?] Latin *meresko ‘I deserve’ Indicative Subjunctive Indicative Subjunctive 1pl conosco conosca meresco meresca 2pl cono(s)çes conoscas mere(s)çes merescas 3pl cono(s)çe conosca mere(s)çe meresca 1pl cono(s)çemos conoscamos mere(s)çemos merescamos 2pl cono(s)çedes conoscades mere(s)çedes merescades 3pl cono(s)çen conoscan mere(s)çen merescan This velar allomorphy and its particular paradigmatic patterning underwent two opposing types of development: (a) The velar allomorphy was eradicated from those verbs in which it was morphological. The motivation behind this development is supposedly to restore the relationship between form and meaning which was eradicated by sound change. This development is attested in old Spanish for a number of verbs in which the most frequently occurring form, usually 3 SG present indicative, replaces the allomorphy in the L-pattern11 e.g., TANG > tango >> taño ‘I play’, FRANG > /ɲ/ (c.f. UNGULA > uña ‘nail’), /nd͡z/ (c.f. *RINGELLA > renzilla ‘quarrel’); in the case of QUINGĔNTI > quinientos ‘five hundred’ the result can even be /n/. For the verb uñir/unzir the most frequent forms were those ending in the sound /nd͡z/ from which the modern verb uncir ‘yoke’. 11 There are examples in which it is the L-pattern allomorph which becomes dominant across the paradigm, e.g., erguir ‘to erect’ < ERIGO and muñir ‘to summon’ < MONE . Paul O’Neill frango >> fraño ‘I break’, CING > cingo >> ciño TING > tengo >> tiño ‘I dye’, SPARG > espargo >> esparzo ‘I scatter’, RING (R) > ringo >> riño ‘I scold’ (b) The velar allomorphy was extended to other verbs in which it was not etymologically expected (salgo, valgo, pongo, vengo, firgo, caigo, traigo, oigo, which all persist in the modern Spanish and suelgo, muelgo, duelga, calga, fuigo, fiergo, remango, tuelgo, distruigo, restituigo, falgo < fallir which are only attested in old Spanish.) It should be noted that the overwhelming majority of verbs which displayed non-etymological velar allomorphy are those for which, given the effects of desinential yod in the verbal paradigm (O’Neill 2012), ought to have shown a particular type of allomorphy, usually characterised by a palatal consonant, exclusively in the L-pattern cells. Examples of this allomorphy are given in (7) for the present tense paradigms of proto Ibero-Romance reflexes of FACERE12 ‘do’, TENĒRE ‘have’, SALĪRE ‘go out’ AUDĪRE ‘hear’, UENĪRE ‘come’ UALĒRE ‘be worth’, TRAHERE ‘bring’, CADERE13 ‘fall’ (c.f. Spanish hago, tengo, salgo, oigo, valgo, traigo, caigo) 12 The majority of scholars of Spanish (Menéndez Pidal 1941:292-293, Penny 2002:174 amongst others) are of the opinion that the L-pattern reflexes of Latin FACI ‘I do’ derive from a form a preliteraty form [fako] in which the yod was lost at a preliterary stage of the Language. This view is due to the first ever attested Castilian L-pattern forms which display no effects of yod. See Malkiel (1974) for a refutation of this claim and a full discussion of the effects of yod on verbs in Ibero-Romance (see also O’Neill 2012). 13 For TRAHERE and CADERE the presence of the yod must have been analogical since their Latin etyma did not contain a dessinential yod. The insertion of a non-etymological yod must also be supposed in a number of other ERE verbs whose root allomorph ended in /d/ (e.g., RĀDO ‘I scratch’, CRED ‘I believe’, R D ‘I gnaw’, UAD ‘I wander’). For a discussion of the possible motivations of this analogical yod see O’Neill 2012 and for an alternative view see Lloyd 1987: 296-297; Malkiel 1973-74:333; Penny 2002:174-175. Likewise the reflexes of the Latin verb PET ‘aim at’ and PON ‘put’ in Portuguese (peço, ponho) both suggest a proto-form *[petjo] and * (7) Indicative [ou̯ʝo] Subjunctive [ou̯ʝa] Indicative [fat͡so14] Subjunctive [fat͡sa] Indicative [βeɲo] Subjunctive [βeɲa] Indicative [tɾaʝo] Subjunctive [tɾaʝa] 2pl 3pl [ou̯es] [ou̯ʝas] [fad͡zes] [fat͡sas] [βɛnes] [βeɲas] [tɾaes] [tɾaʝas] [ou̯e] [ou̯ʝa] [fad͡ze] [fat͡sa] [βɛne] [βeɲa] [tɾae] [tɾaʝa] 1pl [ou̯emos] [ou̯ʝamos] [fad͡zemos] [fat͡samos] [βe i os] [βeɲamos] [tɾaemos] [tɾaʝamos] 2pl [ou̯etes] [ou̯en] [ou̯ʝates] [ou̯ʝan] [fad͡zetes] [fad͡zen] [fat͡sates] [fat͡san] [βe ites] [βɛnen] [βeɲates] [βeɲan] [tɾaetes] [tɾaen] [tɾaʝates] [tɾaʝan] Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive 1pl [teɲo] [teɲa] [saʎo] [saʎa] [βaʎo] [βaʎa] [kaʝo] [kaʝa] 2pl [tɛnes] [teɲas] [sales] [saʎas] [βales] [βaʎas] [kaes] [kaʝas] 3pl [tɛne] [teɲa] [sale] [saʎa] [βale] [βaʎa] [kae] [kaʝa] 1pl [tenemos] [teɲamos] [salimos] [saʎamos] [βale os] [βaʎamos] [kaemos] [kaʝamos] 2pl 3pl [tenetes] [teɲates] [salites] [saʎates] [βaletes] [βaʎates] [kaetes] [kaʝates] [tɛnen] [teɲan] [salen] [saʎan] [βale ] [βaʎan] [kaen] [kaʝan] 1pl 3pl 14 Portuguese maintained a distinction between the results of /k/ + [j] and /k/ + front vowel, the first producing a voiceless sibilant the second a voiced one ( FACI > faço, DĪCIS > dizes (Williams 1962: 79 & 67); Spanish however underwent secondary voicing upon the results of /k/ + [j], thus ĒRĪCIU > erizo, DĪCIT > dize (Penny 2002: 63 & 66). The reasons for this are unsure and much debated (c.f. Malkiel (1971, 1993) for a full discussion. Trusting Malkiel’s conclusions in what follows I shall assume that in pre-literary Castilian /k/ + [j] > /ts/. Paul O’Neill 3. Velar allomorphy in old Spanish: a purely morphological account based on allomorphic endings A purely morphological account of velar allomorphy in Spanish supposes that the cells of the Lpattern constitute a morphome and thus the velar allomorphy is not licensed phonologically or semantically but rather there exists a purely morphological generalisation, that speakers are aware of, that a particular morphological formative, for example the root caig- of the verb caer ‘fall’ in (1), occurs in a specific set of cells. Such a conclusion is problematic since the cells of the L-pattern can be reduced to a single phonological generalisation: the root of the verb is always followed by a non-front vowel. Such a correlation between an allomorph and a particular phonetic environment invites a phonologically motivated interpretation of the allomorphy. Moreover, Bybee (1985) has argued that the distribution of the velar allomorph in Spanish could be determined by notions of markedness whereby the 1 SG present indicative is considered to be the basic form from which all the other forms are derived (see Pérez Saldanya (1995) for similar arguments for Catalan) On the matter of the L-pattern forms being derived from the 1SG forms O’Neill (2011a: 204246, and forthcoming(a)) has presented diachronic evidence from Spanish and other Romance varieties which contradict such a view (also see Wheeler 2011, for Catalan). The conclusion of these authors is that there does not exist any coherent semantic set of features nor any morphological basic-derived relationship relating to frequency/markedness from which to derive the particular paradigmatic distribution of the velar allomorphs in Spanish. On the topic of the velar allomorphy being derived phonologically, O’Neill (2011a: 204-246, and forthcoming(a)) has advanced diachronic, synchronic, comparative and psycholinguistic evidence (Bybee and Pardo, 1981) which prompts the conclusion that the correlation between the velar allomorph and non-back vowels is an historical accident and, in no way, does the former determine the latter. As for the diachronic evidence, the author draws upon conclusions made by Martín Vegas (2007) which differs from other previous analyses of the spread of velar allomorphy in Spanish (Maiden 1992, 2001; Malkiel 1974; Menéndez Pidal 1941:292-293; Penny 2002:174-180) in claiming that historically the extension of the velar was an extension of the desinences -go, -ga, -gas, gamos, -gades, -gan. For those verbs which possessed or adopted such desinences it is impossible to claim that the velar consonant is phonologically licensed since, to take poner ‘put’, it is not the case that there are two allomorphs pon- and pong- whose distribution is determined phonologically; rather there is a lexeme PONER, with the root pon, which is marked as having the desinences -go, -ga, -gas, gamos, -gades, -gan in the L-pattern cells. The argument is the following: after the analogical changes whereby [fat͡so] > fago15 there were a number of high-frequency verbs (decir ‘say’, facer ‘do’ and derivatives of DUCERE ‘lead’ (see (3)) which in the L-pattern cells ended in a velar consonant + /a/ + markers of person and number ( -go, -ga, -gas, gamos, -gades, -gan). These desinences coincided with the large number of inchoative verbs which through regular sound change displayed the same terminal 15 The majority of scholars of Spanish (Menéndez Pidal 1941:292-293, Penny 2002:174 amongst others) are of the opinion that the L-pattern reflexes of Latin FACI ‘I do’ derive from a form a preliteraty form [fako] in which the yod was lost at a preliterary stage of the Language. This view is due to the first ever attested Castilian L-pattern forms which display no effects of yod. See Malkiel (1974) for a refutation of this claim and a full discussion of the effects of yod on verbs in Ibero-Romance (see also the footnote above). Paul O’Neill elements, although in this case the velar consonant was voiceless (c.f. meresco, floresco, consoco, nasco, aparesco of the type in (6)). Given such a situation the suggestion is that the old Castilian forms nasco, aparesco, digo, fago, redugo, frango, cingo were analysed as composed of a root nas-, apares-, di-, fa-, redu-, franm cin- plus the desinences –go, -ga, -gas, -gamos, -gades, -gan for the L-pattern cells. Martín Vegas (2002:168) notes that ‘con un análisis de este tipo no hablaríamos de una inserción de –gen determinados verbos, sino de la extension de un esquema/modelo de flexion que carecteriza porque la 1ª persona del pres.[ente] ind.[icativo] y el pres.[ente] subj.[untivo] terminan en –go, ga, -gas…’. O’Neill (2011a, 2015) differs from this author in that he extends this analysis to motivate the velar element in some of the earliest attested velar verbs of Spanish, vengo, tengo, salgo. With specific reference to the Castilian proto-forms [teɲo] < TENE VENI ‘I come’, [saʎo] < SALI ‘I have’ [βeɲo], ‘I leave’, if these were analysed as having the roots [teɲ]- [βeɲ]-, [saʎ]-, and were made to coincide with the morphological model above then such protoforms would, through regular sound change, produce the attested words tengo, vengo, salgo16, since the palatal consonants would be located in the coda of the syllable and would become depalatalised (c.f. DOMINE > [doɲe] > don vs. DOMINA > doña, GALLU > gallo vs. GALLICU > *[gaʎgo] > galgo and also PIGNORA > *[peɲra] > pendra > prenda, (Penny 2002) and COLLOC > [kweʎgo] > cuelgo (Malkiel 1974:322)). 16 Possible evidence for this analysis can be obtained from varieties of Aragonese which usually display a more extensive extension of the velar element. The phonological rules of these varieties differ regarding whether or not certain palatal consonants are allowed in the coda of the syllable or not but in Ansotano (Barcos 2007:90) the palatal lateral, [ʎ], is not allowed in coda positions (c.f. BELLU > bel, BALLE > bal (Barcos 2007:37), and thus the verbs whose lexical root terminates in <ll> = [ʎ] display this throughout the paradigm with the exception of the lexemes which display the velar element ( in bullir ‘to boil’, cullir ‘to catch’ and derivatives thereof) which in the L-pattern cells display the forms bulg-, culg-. In other varieties there is variation as to whether cullur diplays the velar element (in Belsetán, c.f. Lozano Sierra & Saludas Bernad 2005:86) and here the forms cullo alternate with culgo. O’Neill (2011a, 2015) maintains that this analysis of the spread of velar allomorphy in Spanish is preferable to previous accounts due to (a) its ability to explain the historical facts in a more convincing and less problematic way17 (b) its ability to give an explanation to the hybrid verbs forms of Medieval Spanish of the type plazgo ‘I please’, yazgo ‘I lie’, reduzgo ‘I reduce’ (c) its ability to motivate the irregular sound change /s/ > /t͡s/ which took place in the class of verbs which contained the Latin inceptive reflex –SK. In what follows these arguments will be briefly summarised 4. Evidence for the velar element being morphologically parsed as belonging to the inflectional endings in old Spanish The hypothesis that the extension of the velar consonant to other verbs in Spanish constituted an extension of desinences and not of velar allomorphs can provide an explanation to the hybrid types of forms of Medieval Spanish such as plazgo ‘I please’, yazgo ‘I lie’, reduzgo ‘I reduce’ (Modern Spanish plazco, yazco, reduzco),. The first set of these words, in the earliest testimonies, displayed the alternation of the root final velar /g/ with the sound /d͡z/: plago vs. plaze; yago vs. yaze and regugo vs. reduze. However, the velar L-pattern forms came to be characterised by the sound [d͡zg]: plazgo vs. plaze; yazgo vs. yaze and reduzgo vs. reduze. If in 17 This alternative analysis (a) obviates the problems, first pointed out by Malkiel, inherent with explaining the origin of the velar forms in tengo and vengo via hesitation between the sounds /ɲ/ and / ɡ/ in verbs such as tango – taño. This alternation in this class of verbs, supposedly through hypercorrection, was paradoxically levelled in favour of / ɡ/ sequence in those verbs in which the velar was not etymological but in favour of /ɲ/ in verbs which it was (c.f. Menéndez Pidal 1941, Maiden 1992, 2001); also (b) this explanation can, in a straightforward way, explain why in verbs of the type caigo, traigo, oigo, the velar element is preceded by the glide [j], without the tenuous references of Malkiel (1974:335-336) to the double etymology of the forms of traer and the blending of the forms trayo and trago into traigo which acted as a catalyst for all other forms in –ig-. The explanation would be that these verbs were analysed as having the roots [kaj], [traj], [oj] to which the endings [go], [gas], [ga]… were concatenated. (cf. O’Neill 2011, forthcoming(a) for a full discussion) Paul O’Neill these original forms the velar element were analysed as pertaining to the lexical root (plag-o, yag-o. redug-o) then there is no way to explain the Medieval forms with the sequence –zgosince if one were to advocate an explanation based on analogical levelling of the root in favour of the most common form (the 3SG, present indicative forms: plaz-e, yaz-e, reduz-e) then the predicted outcome would be the forms plazo, yazo, reduzo. If, however, the velar consonant is not considered to be part of the root but part of the ending (or a post radical formative which is neither an ending nor a root), then the morphological segmentation is pla-go, ya-.go, redu-go and thus one can explain the medieval forms in terms of levelling in favour of the most common form of the root: pla-go > plaz-go under the influence of plaz-e (O’Neill 2011a, 2015 and Martín Vegas 2007) Another advantage of considering the velar consonant as being parsed as part of the desinence in old Spanish is that it can account for the irregular sound change that took place in the class of verbs which contained the Latin inceptive reflex –SK. As Penny (2002:180) notes, according to regular sound change, in the L-pattern cells this suffix was realised as /sk/ (c.f. MUSKA > mosca ‘fly’) due to it always being followed by a back vowel; in the non L-pattern cells ͡ (c.f. PISCĒS however, in which it was followed by a front vowel, the regular outcome was /ts/, > peçes > peces ‘fishes’); thus, old Spanish present tense reflexes of the Latin verb COGN SCERE ‘know’, as shown in (8) along with the modern forms, display different allomorphs in the L-pattern than in the rest of the present tense (and the rest of the paradigm). (8) Old Spanish Indicative Subjunctive Modern Spanish Indicative Subjunctive 1SG conosco conosca conozco conozca 2SG conoçes conoscas conoces conozcas 3SG conoç(e) conosca conoce conozca 1PL conoçemos conoscamos conocemos conozcamos 2PL conoçedes conoscades conocéis conozcáis 3PL conoçen conoscan conocen conozcan Note that in old Spanish the graph <ç> represented the affricate phoneme /ts/ and in Modern Spanish the graphs <z> and <c> are graphemes of the same sound /θ/ or /s/ depending on the variety. Towards the end of the fifteenth century, however, the root final sibilant /s/ was replaced by the sound /t͡s/ (perhaps at this time modified to the dental sibilant /s̪/), which was characteristic of the majority of the forms of the paradigm (Penny 2002:108). As with the examples such as yazgo, plazgo above, this change cannot correspond to a process of levelling in favour of the lexical root of the majority of forms if the velar element is included within the root allomorph. This is because, if this were the case the verbs would have two allomorphs: [konot͡s]- and [konosk]-, and levelling would suppose the ousting of the latter, leaving behind a paradigm, well attested in the Asturian and Galician varieties, in which the root is invariable [konoθ] 18 and there is no velar: thus, conozo, conoces, conoce..; conoza conozas, conoza... I maintain that in Castilian the process of levelling did take place but the root allomorphs were not [konot͡s]- and [konosk]- rather [konot͡s]- and [konos]-, and the velar element was included as part of the ending. Accordingly the effect of levelling produces the attested modern Castilian forms in (8) in which the graphs <z> and <c> are graphemes of the same sound /θ/. However, in its origins the velar element must have been considered part of the lexical root, hence the development whereby FACIŌ > fago and IACIŌ > iago and not the expected [fatsko] and [ʝasko]. The attested forms are a matter of proportional analogy with the roots of the present 18 in accordance with the regular latter sound change /ts/ > /θ/. Paul O’Neill tense forms of verbs of the type [dik-o] – [did͡z-e]. Such verbs, which had an alternation between a root allomorph ending in /g/ in the L-pattern and [d͡z] elsewhere, had a high type and token frequency: including decir ‘say’, reflexes of verbs derived from DŪCERE (see (3) REDŪC ‘I reduce’ and ADDŪC espargo ‘I scatter’ and ERIG ‘I adduce’) and also verbs of the type in (4) (SPARG > > yergo ‘I erect’) 19. Moreover, it is far from clear, for modern Spanish, that velar allomorphy is a function of the endings of verbs and not their roots. This uncertainty is prompted by the fact that, with the exception of verbs of the type merecer 20 ‘be worth’ which display the voiceless velar consonant, velar allomorphy in Spanish has, from the Golden Age onwards, become reduced in number and is now limited to a small number of verbs which generally display other cases of allomorphy within their paradigms. There is no evidence that the velar allomorphy of the type salgo, valgo, tengo, vengo, traigo for salir ‘go out’, valer ‘be worth’, tener ‘have’, venir ‘come’ and traer ‘bring’ is the result of the concatenation of a particular type of velar allomorphic endings, as opposed to another root allomorph for these forms. That is, it is not clear that the morphological segmentation of salgo is sal-go and not salg-o, especially when one considers the amount of root allomorphy which such forms display elsewhere (cf. the allomorphs saldré, valdré, tiene, viene, trajera). I therefore assume the hypothesis that there can be variation and change as to the morphological parsing of the velar element. In what follows I shall provide evidence in support of this hypothesis; the majority of the evidence will involve how velar allomorphy interacts with 19 Moreover, this hypothesis gains strength when one takes into account the relative productivity that the alternation [g] – [d͡z] must have had in the late Latin/early Romance of the Iberian Peninsula on account of the number of semantically related derivatives [nɔd͡ze] ‘nut’ - [nokale] ‘nut tree’ (see Martín Vegas 2007:117-173). Much of these derivatives, however, in modern Spanish have become semantically opaque, thus: paz ‘peace’, pagar ‘to pay’, hoz ‘sickle’ – ahogar ‘to drown’ (Martín Vegas 2007:163-166) 20 Other verbs of this type are vencer, cocer, escocer, ejercer, torcer diphthongised roots in central Ibero-Romance varieties. Firstly, however, it is necessary to introduce another morphome of Ibero-Romance, the N-pattern morphome, since the allomorphy determined by this morphome, that of diphthongised stems, overlaps with the L-pattern morphome and the scope of each morphome (i.e. whether they can be conceived as generalisations regarding the endings or roots/stems of verbs) is revealing as to the morphological parsing of the velar element. 5. The N-pattern morphome The label ‘N-pattern’21 refers to a pattern of alternation, recurrent across the Romance languages, whereby the forms of the first, second and third persons singular and third person plural of the present indicative and of the present subjunctive, and the second person singular of the imperative, share a distinctive common form, usually a common root allomorph.22 This is illustrated in (9) with examples from the Italian verbs morire ‘die’, udire ‘hear’, dovere ‘ought to’. (9) Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive 1SG 2SG 3SG muoio muori muore muoia muoia muoia odo odi ode oda oda oda devo devi deve deva devas deva 1PL 2PL moriamo morite moriamo moriate udiamo udite udiamo udiate dobbiamo dovete dobbiamo dobbiate 3PL muoiono muoiano odono odano devono devano 21 This label is arbitrary and has neither semantic nor phonological significance. It is due to the perception that the pattern, as conventionally set out on paper, resembles the form of the letter ‘N’ in Morse Code. 22 In subsequent publications Maiden (e.g. 2011) has stated that it is not merely a common allomorph but rather common characteristics of various types which are not found elsewhere in the paradigm Paul O’Neill 2SG Imperative muori Imperative odi Imperative - 1SG Imp. indic morivo Imp. indic udivo Imp. indic dovevo The cells of the N-pattern also share the property of being rhizotonic. Maiden, however, considers the allomorphy in these stems to be neither phonologically derivable or conditioned (however cf. Anderson 2008, 2010 and 2011 for an alternative view for the Rumantsch variety of Surmiran, and Maiden 2011 for a response). Such a conclusion is problematic for Spanish since, in Modern Spanish, the only type of unambiguous N-pattern allomorphy is that of diphthongised roots, as shown in (10) for perder ‘loose’ and mover ‘move’, and this allomorphy is largely considered to be a matter of phonological conditioning from an underlying invariant form (Harris 1969, 1977, 1978, 1985, Schuldberg 1984, García-Bellido 1986, Carreira 1991)). (10) 1sg Present Indic pierdo Present Subj. pierda 2sg 3sg 1pl 2pl 3pl pierdes pierde perdemos perdéis pierden pierdas pierda perdamos perdáis pierdan 1SG 2SG 3SG 1PL 2PL 3PL Present Indic muevo mueves mueve movemos movéis mueven Present Subj. mueva muevas mueva movamos mováis muevan O’Neill (2011a:246-279, forthcoming(b)), however, has advanced diachronic, synchronic, psycholinguistic and cross-linguistic evidence to suggest that the diphthong-monophthong alternation which exists in a great many verbs in Spanish and which is correlated with word stress is not a matter of an invariant underlying form from which the different allomorphs are predictable via a phonological stress conditioned rule (see also Bermudez Orero 2013). Rather, the different allomorphs must be considered to be lexically stored. Furthermore, he has argued (contra assumptions of Bermúdez-Otero, 2013) that the selection of the respective allomorphs is not a matter of phonologically conditioned allomorphy, whereby rhizotonic stress selects for the diphthongized allomorph, since word stress itself in the Spanish verb cannot be predicted on the basis of phonological criteria (contra Roca 1990). The most straightforward solution is to admit that the different stored word forms are merely lexically marked as being stressed and that stress in the Spanish verb is morphologically determined. The N-pattern therefore is a morphome for Spanish and I would claim that this also potentially holds for all Ibero-Romance varieties. As to whether the morphome is to be considered a generalisation over roots or desinences, there is comparative evidence to suggest that both options are possible. Thus, the verbs of what is traditionally termed the Catalan IIIa conjugation display the increment –eix23, exclusively in the N-pattern cells as illustrated in (11) for the Catalan verbs servir ‘serve’. (11) Present Indicative serveixo Present Subjunctive serveixi serveixes serveix serveixis serveixi servim servim 2PL serviu serviu 3PL serveixen serveixin 1SG 2SG 3SG 1PL Etymologically this formative derives from the Latin inchoative or ingressive affix -SC- e.g., FL RET ‘it flowers’ vs. FL RĒSCET ‘it’s coming into bloom’. In Latin, this augment was incompatible with perfective meaning, and thus could only appear in imperfective verb-forms. In a number of Romance varieties, however, what had then become a unmotivated distribution of allomorphy was incorporated into an N-pattern distribution 23 Paul O’Neill In Modern Catalan, this increment, as it names suggests, is generally considered to be a regular formative which is concatenated to the verbal root 24 and which constitutes a regular model of verbal inflection for a group of verbs (Pilar Perea 2002:600). Thus, the morphological parsing is serv + eix. There are other verbs, however, for which the N-pattern distribution implies a different selection of a lexical allomorph; witness the present tense paradigm for the verb anar ‘go’ in (12) (cf. the present indicative form of the verb ir in old Spanish’: vo, vas, va, imos, ides, van)25. (12) Catalan Indicative Subjunctive vaig vagi vas vagis va anem vagi anem aneu aneu van vagin The foregoing suggests that there can be variation as to what morphological formatives morphomes are ‘sensitive’ to with regards to formal identity; that is, morphomes could be considered to be generalisations regarding the distribution of a certain root allomorph or of a particular desinential formative. This conclusion becomes interesting when one takes into account that (a) there can be variation as to the morphological pasting of the velar element (b) the N-pattern and L-pattern morphomes overlap. 24 Mascaró (1986) has even suggested that it is an allomorph of the thematic vowel Historically, these verbs are cases of incursive suppletion (used in the sense of Corbett 2007). Incursive suppletion is where two or more different lexemes are integrated into a paradigm of a single lexeme, usually in mutually exclusive cells. Thus, Catalan contains reflexes of Latin UĀDERE 'to wander’, surviving exclusively in the N-pattern cells but reflexes of AMBULĀRE 'walk' exclusively in the remainder of the paradigm. In Old Spanish the reflexes were of the paradigms of UĀDERE and ĪRE ‘to go’. 25 The L-pattern morphome involves the 1SG present indicative and all the present subjunctive cells, whilst the N-pattern involves the 1SG, 2SG, 3SG and 3PL present indicative and subjunctive as well as the 2SG imperative. This is illustrated in (13) in which the light grey shading denotes cells which are exclusively part of the N-pattern, the dark grey shading marks cells which are exclusively part of the L-pattern, the diagonal lines designate those cells which are common to both patterns and the clear cells mark those forms which do not form part of either morphome. (13) Present Indic Present Subj. 1SG 2SG 3SG 1PL 2PL 3PL Imperative 2SG 2PL Given the assumption that morphomes can determine the distribution of a morphological formative (be it a root, infix or ending), this overlap of morphomes poses problems for cases in which diphthongised roots co-occur with velar allomorphs since usually the occurrence of a particular formative (e.g. diphthongised stem or velar element) in one cell of the morphome implies its presence26 in all the cells of the morphome (Maiden 2004). Thus, the cells marked with diagonal lines are ‘torn’ between respecting the formal coherence of identical allomorphy 26 This is however a tendency and diagnostic of morphomes rather than a defining principle since it is not always the case that a particular allomorph needs to be present in all cells of the morphome (see se vs. sepa from Spanish saber ‘know’). See also arguments in Wheeler (2011) Paul O’Neill of the N-pattern or the L-pattern. This matter is further complicated depending on the morphological segmentation of the velar element. In the following sections I will provide evidence from different varieties of central IberoRomance which show a number of solutions to this clash of morphomes. The data supports the hypothesis that there can be variation as to how the velar element can be morphologically parsed and suggests that morphomes can not only converge upon one particular type of phonological form (Maiden 2004) but also on a particular morphological structure or type. 5. The clash of the morphomes The first hypothetical solution to the clash of the morphomes is that when morphomes overlap, the tendencies of coherence of each particular morphome can be focused upon a particular type of phonological and/or morphological exponent. Each morphome is sensitive to a different morphological formative: the N-pattern to the lexical root characterised by diphthongs and the L-pattern to velar endings. Such is the situation that I envisage may have held in old Spanish since diphthongised roots were compatible with velar allomorphy and, even though there was variation as to which lexemes diphthongised and which presented velar allomorphy, the distribution of the former was the N-pattern and the latter the L-pattern cells (see tuele, tuelga, tolgamos, tolgades for the verb toler ‘take’). Such a state of affairs also holds, in my opinion, in the Asturian variety of Cándamo (Díaz González 1986: 81) as attested by the reflex of DOLĒRE > doler – duler in (14) and also in the Ansotano variety of Aragonese (Barcos 2007:90), as illustrated by the reflexes of TORQUĒRE > torzer and TENĒRE > in (15) (14) 1SG Present Indicative duelgo Present Subjunctive duelga 2SG 3SG duelis duel duelgas duelga 1PL dulemos dulgamos 2PL duleis dulgais 3PL duelin duelgan (15) 1SG Present Indicative tuerzco Present Subjunctive tuerzcai Present Indicative tiengo Present Subjunctive tiengai 2SG tuerces tuerzcas tienes tiengas 3SG 1PL tuerce torcemos tuerzca torzcamos tiene tenemos tienga tengamos 2PL torcez torzaz tenez tengaz 3PL tuercen tuerzcan tienen tiengan These data are compatible with the assumption that the N-pattern is the domain of diphthongisation and of morphological roots and the L-pattern that of velar allomorphy and desinential endings. Any phonologist, however, could claim that given the fact that diphthongization is coterminous with stress and velar allomorph with the appearance of a non-back vowel, then the allomorphy is phonologically licensed. This, of course, is a possibility, but, as demonstrated for Spanish (O’Neill 2011a:211-246 and forthcoming(a)) the mere coincidence of an allomorph and a phonological environment does not necessarily imply that there exists a casual relationship Paul O’Neill between both. The general assumption in modern phonological theory is, however, that such recurrent phonological correlations do imply a casual relationship; Burzio (2004) on the topic of velar allomorphy in Italian states that ‘whatever identity relations have a statistical presence in the data, also have, ipso facto, a grammatical status, expressible as a faithfulness constraint in the O[ptimality] T[heory] formalism’. I question this assumption and follow Eddington (2004:3) who has points out that ‘a detailed, rigorous, or sophisticated description of a linguistic phenomenon does not necessarily indicate that the phenomenon has any relevance to linguistic cognition’ (see also Goyvaerts 1978; Lass 1976; Morin 1988; Skousen 1989). In fact, the assumption that a recurrent phonological correlation between a phonological and morphological element is tantamount to phonologically conditioned allomorphy constitutes the promotion of linguistic description to the status of psychological explanation and in doing so, creates a circularity in the argumentation since the data/ observable facts which need to be explained are equated with the very explanation of the data (see Eddington 2004:14; Higginbotham 1991:555; Ohala 1990:159; Sampson 2001:124). Given the lack of any independent evidence to suggest that the allomorphy discussed in this section is phonologically conditioned and on the basis of the comparative Spanish data I entertain the hypothesis that the allomorphy is morphologically conditioned. From such a perspective the evidence from the paradigms above indicate that the root of the L-pattern is created by the concatenation of a velar ending to the root of the N-pattern. Independent proof in support of such an hypothesis can be found in other velar verbs from the Ansotano variety which do not display diphthongised roots but for which the L-pattern forms seem to be constructed on the N-pattern roots (trai - traigo, fui -. fuigo, bei – beigo from the verbs ‘bring’, ‘flee’ and ‘see’). Further evidence for such a conclusion, not involving diphthongising verbs, comes from the Aragonese locality spoken around the valley of Benasque (see Saura Rami 2003:233-271). The speech of this area is considered to be transitional between Aragonese and Catalan varieties and the different localities exhibit a great deal of morphological variation. With respect to velar allomorphy, the distribution of this allomorphy is, as with other varieties of Aragonese and as opposed to Catalan, exclusively in the L-pattern cells. Of particular interest to the present discussion is that the L-pattern forms seem to be built via the concatenation of a velar consonant to the root of the 3SG present indicative. This is illustrated in (16) for the verbs veure ‘live’, creure ‘believe’, beure ‘drink’. Note that the paradigms given here do not correspond those given by the author (ibid:233-271) since the author tries to incorporate all the variation in his paradigms. The verb forms below correspond to those from Zone B27 based on the supplementary information regarding the root allomorphs and desinences in this zone (ibid:215-222). (16) Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive viuga ~ viva creugo ~ credo creus28 creuga - creda beugo ~ bebo beuga ~ beba 2pl. viugo ~ vivo vius beus 3pl. viu 1pl. vivim 2pl. vivits 3pl. viven creuguas credas creugua creda creguan ~ credam creguats ~ credats creguan ~ credan beuguas ~ bebas beugua beba beugam ~ bebam beuguats ~ bebats beuguan ~ beban 1pl. viuguas ~ vivas viuga ~ viva viuga ~ vivam viuguats ~ vivats viuguan ~ vivan creu credem credets creden beu bebem bebets beben Here, in the L-pattern, the etymological roots viv-, cred-, and beb- alternate with the velar forms viug-, creug- beug-29. The forms viu, creu, beu are etymologically expected and attested in 3SG forms of the present indicative via the vocalisation of a final labial, dental or voiceless velar consonant after apocope had taken place (esriu < scribe < SCRIBIT, c.f. dèu < DECE, pèu < PEDE, nèu < *NEVE, perdiu < PERDICE, cllau < CLAVE, ibid:105). This sound change, although not explicitly stated, must have also held in coda final position (SCRIBERE > scribre > escriure). Apocope did not always occur with final /o/ (c.f güello, sapo, bllanco, and 27 See (Saura Rami 2003: 28) for a list of these localities. For the 2SG and 3SG present indicative the variants credes, crede respectively are given in parenthesis. Presumably this notation means that these forms are attested but very frequent. 29 This is also the case with the verbs riure ‘laugh’, chaure ‘lie’, deure ‘owe’, escriure ‘write’ in which the 1SG present indicative forms have the respective alternants: rido – riugo, chaso – chaugo – chasco, debo – deugo, escribo – escriugo. 28 Paul O’Neill derivations in –ano, -iello, and uelo (ibid:6430) and thus, the alternative morphological forms for the 1SG present indicative have come about either by way of the concatenation of the endings – go, -ga, -gas, -gam, -gats, -gan, to the form of the 3SG viu + go > viugo, or, assuming the validity of the vocalization of the consonants in coda position, via the concatenation of the aforementioned endings to the root allomorph of the 1SG (vid-, cred, beb-) producing the forms vidgo > viugo, credgo > creugo, bebgo > beugo. Upon the assumption that the allomorphy is not phonologically determined, the varieties analysed in this section (those of (14)and (15)) suggest that the velar element forms part of the desinential ending and the consistency of the velar consonant is determined by the L-pattern, whilst the distribution of the diphthongized allomorphs is determined by the N-pattern. Such evidence not only shows that morphomes can be sensitive to different morphological formatives but it also confirms their tendency towards convergence on a particular phonological form. In the varieties analysed here the N-pattern is relative to roots and the L-pattern to desinences, therefore, there is no conflict and incompatibility between the two types of allomorphy. As stated, I assume such a state of affairs held for Old Spanish as attested by lexemes in which velar allomorphy and diphthongization coincided either due to regular sound change (cuego/cuezo ‘I cook’) or analogy (duelgo ‘I hurt’, muelgo ‘I grind’, tuelgo ‘take away’, suelgo ‘I am accustomed to’, fiergo ‘I harm’); modern Spanish has, in all cases, preferred the diphthongized forms and eradicated the velar allomorphs. In fact, it is a peculiarity of Castilian, in comparison with its central Ibero-romance neighbours, that there are no verbs in which 30 Note however that in the plural form the /o/ is lost: sapo-saps, bllanco/bllancs, casuelo/casuels. Also, final /o/ is sporadically lost under the influence of neighbouring Catalan varieties (llop < LUPU, caball < CABALLU, llech < LECTU (ibid: 63). diphthongization coincides with velar allomorphy. This, I claim, is because in Castilian velar allomorphy is no longer exclusively a function of the endings of verbs but could be viewed as an allomorph of the root, at least for verbs with voiced velar allomorphy. That is, the observed mutually exclusive relationship between velar allomorphy and diphthongs suggests that speakers analysed them as alternative manifestations of root allomorphy; not one type of allomorphy pertaining to roots and the other to endings. In such a case there occurred a clash of morphomes and a competition over forms: the velar root duelg- of the verb doler was present in all the L-pattern cells but the root duel- was only present in a subsection of the N-pattern cells. Given the type frequency of diphthongising verbs, together with the inclusion of the most frequent and least marked form of the paradigm, 3 SG present indicative, in the N-pattern, it is not surprising that this pattern won out in Spanish and thus duelga > duela. In neighbouring Romance varieties, especially those to the East, which contained more lexemes displaying velar allomorphy, this solution was not dominant and what could occur is that both patterns could exist but one could take precedence over the other: The L-pattern could dominate the N-pattern (L>>N, to be read as the L-dominates-N-pattern) or vice versa, the N-pattern could dominate the L-pattern. In the following section I will only analyse the cases in which the L-pattern dominates the N-pattern (L>>N pattern) since the converse is not attested for velar allomorphy31, although it is arguably attested for Spanish for vocalic allomorphy (O’Neill 2011a:339-349, 2011b). 5.1 The creation of the L>>N-pattern 31 Such a pattern would present a diphthongised roots in the N-pattern (duelo, duele, duela) and the velar allomorphy would be relegated only to the 1pl & 2pl L>>N (dolgamos, dolgáis) present subjunctive, the only cells of the L-pattern which do not overlap with the N-pattern. Paul O’Neill The tendency of morphomes to converge upon a particular phonological and/or morphological exponent was illustrated in the previous section. In these localities, velar allomorphy was applied to endings and was distributed according to the L-pattern, diphthongised stems was a question of allomorphic roots and distributed according to the N-pattern. Such a state of affairs does not hold, however, in the Aragonese varieties of the Alta Ribagorza (Haensch 2003:141143), since when diphthongisation and velar allomorphy co-occur, the velar is analysed as part of the lexemic root together with the diphthong and thus, in accordance with the tendency towards convergence in the L-pattern, the diphthong spreads to all the other cells of the Lpattern in which it is not present: the arhizotonic 1PL and 2PL present subjunctive forms. This is illustrated by the verbs tórse ‘twist’ and tínrre ‘have’ in (17) (Haensch 2003:128; 121-122). (17) Indicative Subjunctive Indicative Subjunctive 2SG twérsko twérses twérska twérskas tjéngo tjénes tjénga tjéngas 3SG twérse twérska tjéne tjénga 1PL torsém twerskám tením tjengám 2PL torséts twerskátz teníts tjengáz 3PL twérsen twérskan tjénen tjéngan 1SG In this variety, then, the L-pattern dominates the N-pattern, and the latter is reduced to the 2SG, 3SG and 3PL present indicative (the L>>N-pattern). Such domination of the N-pattern by the Lpattern is only valid for those lexemes in which velar allomorphy and diphthongization coincide. Lexemes which only display diphthongized allomorphs, distribute this allomorphy according to the N-pattern only as illustrated by poder ‘be able’ (Haensch 2003:130) and dormir ‘sleep’ (ibid.:145) in (18). (18) Indicative Subjunctive Indicative Subjunctive 1SG pwédo pwéda dwérmo dwérma 2SG pwéts pwédas dwérmes dwérmas 3SG pwéde pwéda dwérme dwérma 1PL podém podám dormim dormám 2PL podéts podátz dormíts dormátz 3PL pwéden pwédan dwermen dwérman In this variety32, therefore, the N-pattern still retains its original, and in many cases etymological, distribution associated with the class of lexemes which only display diphthongized allomorphs; however, when this type of allomorphy coincides with velar allomorphy there occurs a conflict between the ‘coherent’ tendency for there to appear the same allomorph over both the N-pattern cells and L-pattern cells. These data demonstrate that different morphomes can be correlated with different classes of lexemes depending on the type of allomorphy which they exhibit. Thus in this variety velar allomorphs are distributed according to the L-pattern, diphthongised allomorphs according to the N-pattern but when both coincide the result is a L>>N-pattern. Such a pattern correlated with diphthongisation is the first step, to my mind, to the creation of a new morphome, the L+N-pattern which is the merger of both the L-pattern and the N-pattern and which exerts the tendency that all cells in this pattern must share the same morphological formative. The L>>N-pattern constitutes the first step since it establishes a formal concurrence between a diphthongised root and all the cells of both the N-pattern and the L-pattern. Thus, in the Aragonese variety spoken around the valley of Benasque (Saura Rami 2003), it seems that there has been a reanalysis as to the paradigmatic distribution of diphthongised roots, based, most probably, on the coocurrence of diphthongized roots and velar allomorphy in the Lpattern. There is much variation between varieties in this valley and, in particular, with regard to the interaction between velar allomorphy, diphthongisation and word-stress (see Saura Rami 2003:221-240). However, in all zones apart from what the author designates ‘Zone A’, 32 Such a situation also holds in the Cantabrian valle del Pas (Penny 1970), although in this variety there is variation as to whether the diphthong appears in the 1PL & 2PL present subjunctive with velar verbs only (cozcámos ~ cuezcámos, cozcáis ~ cuezcáis. ibid:126). Paul O’Neill diphthongized allomorphy is associated with the N+L-pattern33. This is illustrated by the verbs serner ‘sieve’ and dormir ‘sleep’ in (19). (19) Indicative Subjunctive Indicative Subjunctive 1SG sierno siérne dwérmo dwérme 2SG siernes siérnas dwérmes dwérmas 3SG 1PL sierne sernem siérna siernám dwérme dormem dwérma dwermám 2PL sernets siernáts dormets dwermáts 3PL siernen siernan dwérmen dwérman For this class of diphthongizing verbs, unlike those in which diphthongisation and velar allomorphy coincide, it cannot be stated, as with the examples from (17), that the L-pattern dominates the N-patterns and that this latter pattern is reduced to the L>>N-pattern. Rather, there has been a merger between the N-pattern and the L-pattern, in which all cells of these morphomes display identical root allomorphy. 6. Conclusions and implications In the foregoing I have, on the basis of previous studies (O’Neill 2011a, forthcoming a&b, Wheeler 2011), assumed that the distribution of velar allomorphy and diphthongization in IberoRomance is morphologically determined by reference to what have been termed morphomes (Aronoff 1994, Maiden 2004). I have presented data from verbs displaying such allomorphy from different varieties of Ibero-Romance and at different diachronic stages. The comparative results prompt the conclusion that there can be variation and change as to (a) how the velar element is morphologically parsed and also (b) what morphological formatives morphomes can 33 Similar examples exist for the Aragonese variety of Nerín and Sercué (Quintana i Font 2007:34-37) be construed as being generalisations about (roots or desinences or formatives which may be classed as neither). The generalisations which emerge from the data is that there is a tendency not only for morphomes to be correlated with a particular phonological type of allomorphy (velar allomorphy, diphthongized stems or a combination of both) but also with a particular morphological type (root allomorphy, desinential allomorphy) or structure which tends to become transparent and associated with a particular abstract paradigmatic patterning and also a particular phonological form. Thus, in old Castilian the N-pattern was a generalisation about diphthongized roots (duel(e), cuez(e)), the L-pattern about velar endings and both were compatible (tuelga, duelga, cuezga) producing a tendency towards a structure whereby the root of the N-pattern formed the basis for the inflectional velar forms of the L-pattern via the concatenation of velar endings. I maintain that an appreciation of such morphological structures combined with a recognition of morphomes towards converging on a particular phonological and morphological structure can provide a more convincing explanation for hitherto poorly justified morphological changes. I will illustrate this with reference to the history of velar allomorphy in Spanish and relate my conclusions to the lack of velar allomorphy in Portuguese. 6.1 Velar allomorphy in Spanish revisited As detailed at the end of section 2., in the history of velar allomorphy, Spanish underwent a paradoxical development whereby a number of verbs which displayed an etymological velar consonant came to lose this allomorph via classical analogical levelling whereby their verbal paradigm came to have only one root allomorph (TANG > frango >> fraño ‘I break’, CING SPARG > tango >> taño ‘I play’, FRANG > cingo >> ciño TING > tengo >> tiño ‘I dye’, > espargo >> esparzo ‘I scatter’, RING (R) > ringo >> riño ‘I scold’); other verbs, Paul O’Neill however, came to acquire a non-etymological velar allomorph precisely in the L-pattern cells (tengo, salgo, caigo, fuigo, pongo, suelgo, muelgo, duelga, calga, fuigo, fiergo, remango, tuelgo, distruigo, restituigo,). My explanation for the loss of the etymological velars is that they did not and could not correspond to the general morphological structure for L-pattern forms. The data from old Spanish suggests that the tendency was to have L-pattern forms which were built on the root of the N-pattern, this is largely due to verbs of type (6) which contained the Latin ingressive suffix; these verbs had a high type frequency and constituted a productive way in which to form adjectival and nominal verbal derivations, often via parasynthesis (enorgullecer ‘make proud’, ensordecer ‘make deaf). Their morphological structure in the L-pattern (N-pattern stem + velar endings) also coincided with a number of voiced velar verbs which had a high token frequency: cae – caiga, trae – traiga, o(y)e – oiga, sale – salga, vale – valga, pone – ponga. There were, of course, exceptions to this rule, notably dice – diga, tiene – tenga, viene – venga, but despite their token frequency such alternations were limited to these verbs. Therefore, upon the assumption that the most prominent morphological structure for velar verbs in the L-pattern was for them to consist of a N-pattern root to which velar desinences were concatenated, and given the tendency of morphomes to converge upon a certain type of transparent morphological structure and phonological exponent then the levelling of allomorphy in verbs of the type in (5), e.g. cingo > ciño, tango > taño, plango > plaño is relatively straightforward. These verbs posed problems for the L-pattern since even though their endings did conform to the tendency towards velar allomorphy, their morphological structure did not conform to the pattern detailed above since the forms cin-go, fran-go, tan-go, plan-go did not correspond to the preferred structure whereby the stem of the L-pattern was the root of the N-pattern (ciñ-, frañ-, tañ-, plan-) + the velar endings. Moreover, such forms could not come about via levelling of the root allomorph since the desired forms *ciñgo, *frañgo, *tañgo, *plañgo were phonologically impossible given the restriction against palatal consonants in coda position (DOMINE > [doɲe] > don vs. DOMINA > doña, GALLU > gallo vs. GALLICU > *[gaʎgo] > galgo and also PIGNORA > *[peɲra] > pendra > prenda, (Penny 2002) and COLLOC > [kweʎgo] > cuelgo (Malkiel 1974:322)). The verbs of the type in (3), however, which also displayed a ‘non preferred’ structure in the L-pattern cells, i.e.. the root upon which the velar endings were concatenated was not the same as that of the N-pattern (redu-go vs. reduc(e), adugo vs. aduc(e)) could readily undergo levelling since they did not contain a final palatal consonant: redugo > reduzgo > reduzco; adugo >aduzgo > aduzco34. Since such an option was not available to the verbs with root final palatal nasal consonants the optimal option was for these verbs not to be included in the group of verbs with velar endings thus cin-go > ciñ-o, tango > tañ-o. (see also Pérez Saldanya 1995:420) An explanation based on a tendency towards a convergence of morphological transparency or a particular morphological typology could also perhaps explain for the lack of velar verbs in Portuguese. The L-pattern in Portuguese stands out amongst the Ibero-Romance languages in that it allows various types of allomorphy and has not undergone convergence upon one particular phonological type: witness the modern forms of the Portuguese verbs fazer ‘do’, medir ‘measure’, valer ‘be worth’, vir ‘come’, ver ‘see’, caber ‘fit’ in (20). 34 It must be noted, however, that such a development is not attested for espargo. Paul O’Neill (20) Indicative Subjunctive Indicative Subjunctive Indicative Subjunctive 1SG 2SG faço fazes faça faças meço medes meça meças valho vales valha valhas 3SG faz faça mede meça vale valha 1PL fazemos façamos medimos meçamos valemos valhamos 2PL fazeis façais medis meçais 3PL fazem Indicative façam Subjunctive medem Indicative meçam Subjunctive valeis valem Indicative valhais valham Subjunctive 1SG venho venha vejo veja caibo caiba 2SG vens venhas vês vejas cabes caibas 3SG vem venha vê veja cabe caiba 1PL vimos venhamos vemos vejamos cabemos caibamos 2PL vindes venhais vedes vejais cabeis caibais 3PL vêm venham vêem vejam cabem caibam These allomorphs and their distribution are phonologically predictable given the effects of desinential yod35 and therefore represent old morphological alternations. Old Portuguese also had velar allomorphy in the L-pattern from the class of Latin ingressive verbs of the type (6): old Portuguese paresco paresca, cognosc, cognosca; these verbs, however, were levelled in favour of the majority root of the paradigm which lacked the velar: paresco > pareço; congosco > cognoço . What I propose is that even though Portuguese speakers could have analysed these verbs either as (a) having the morphological structure root + velar endings or (b) having a different root Thus /k/ + [j] is attested in FACI > faço, and also the old form iaço < IACE ; /t/ + [j] in METI > meço; /l/ + [j] in valho < VALE ;35 /n/ + [j] in tenho < TENE ; /d/ +[j] in vejo < VIDE ; /p/ + [j] in CAPI > caibo. Old Spanish contains remnants of the latter development only for /p/ + [j] in CAPI > quepo and /d/ + [j] (veya < VIDEAM, seya < SEDEAM, oya < AUDIAM, riya < RĪDEAM) and /g/ + [j] (fuya < FUGIAM) 35 allomorph from the rest of the paradigm which only differed in one formative (a velar consonant, which was attached to the root) both these analyses contrasted with other L-pattern forms which clearly showed different radical types of root allomorphy (see (20)). Thus Portuguese, I claim, may have preferred a pattern whereby the L-pattern forms possessed allomorphs which were different from the rest of the paradigm and not formed via the concatenation of elements to the root of another form. Such a preference in old Portuguese may have encouraged speakers to reject the morphological type displayed by the L pattern allomorphy of ingressive verbs, cognosco, paresco, which either differed so little from the allomorphs of the rest of the paradigm or else had the structure cognos-co, pares-co. Thus, in old Portuguese there could have been different morphological models/types for L-pattern forms and these were reduced to one. Support for this hypothesis comes from the development of the verb poder ‘be able’ in Portuguese. In Latin, this verb displayed two root allomorphs POSS- and POT- which, as demonstrated in (21), a partial paradigm of the Latin verb POSSE to be able , had a rather random distribution which was shared by no other verbs in the language. (21) INDICATIVE Present Imperfect Pluperfect 1SG. 2SG. 3SG. 1PL. 2PL. 3PL. SUBJUNCTIVE 1SG. 2SG. POSSUM POTES POTEST POSSUMUS POTESTIS POSSUNT Present POSSIM POSSĪS POTERAM POTE‘ĀS POTERAT POTE‘ĀMUS POTE‘ĀTIS POTERANT Imperfect POSSEM POSSĒS POTUERAM POTUE‘ĀS POTUERAT POTUE‘ĀMUS POTUE‘ĀTIS POTUERANT Pluperfect POTUISSEM POTUISSĒS Paul O’Neill 3SG. 1PL. 2PL. 3PL. POSSIT POSSĪMUS POSSĪTIS POSSINT Present infinitive POSSE POSSET POSSĒMUS POSSĒTIS POSSENT Perfect infintive POTUISSE POTUISSET POTUISSĒMUS POTUISSĒTIS POTUISSENT Participle POTĒNS In modern Portuguese this original distribution is lost and, in its place, the suppletive allomorph poss- is distributed according to the L-pattern as demonstrated in present tense forms of the modern Portuguese36 verb poder in (22), (22) Indicative Subjunctive 1pl. posso possa 2pl. podes possas 3pl. pode possa 1pl. podemos possamos 2pl. podeis possais 3pl. podem possam This development of incursive suppletion (used in the sense of Corbett 2007) in the L-pattern is exclusive to Portuguese, as opposed to the incursive suppletion in the N-pattern with the verb ‘to go’ attested in Portuguese, Old Spanish (see footnote 25) and Catalan (see (12) and footnote 25). This suppletion suggests, to my mind, (a) a preference for L-pattern allomorphy and (b) that this allomorphy be root allomorphy which can be radically different from the other forms of the lexeme. Such preferences of Portuguese, as compared to Spanish, can, in my opinion, be traced back to the different treatments of dessinential yod within the two languages and, specifically, in –er verbs and –ir verbs (see O’Neill 2011c). 36 Similar developments are also attested in other Romance languages (Old Tuscan, Cascinagrossa (Piedmont), Romansh (Savognin)) In this article, I have argued for the recognition of (a) autonomous morphological structures, or morphomes, in the Ibero-Romance verb and (b) variation as to the morphological parsing of velar allomorphy. With regards morphomes I have presented data which prompts the conclusion that the allomorphy in these cells not only tend to converge upon one particular phonological form (Maiden 2004) but also tends to converge upon a particular morphological structure or type. An appreciation of this quality of morphomes, together with an acknowledgement that velar allomorphy can correspond to either root allomorphy or desinential allomorphy can facilitate more convincing explanations of instances of change in the class of velar verbs in Ibero-Romance, especially when this allomorphy coincides with diphthongisation. Paul O’Neill Appendix Lexemes which display voiceless velar allomorphy according to the L-pattern (derivatives not included) . Note that this list has been compiled on the basis of a Bosque and Pérez Fernández´s reverse Spanish dictionary (1987). Verbs were excluded if they were catalogued in the online dictionary of the Real Academia Española as in disuse or antiquated, and were not given a model of conjugation, e.g. adherecer. 155 in total placer , nacer , pacer, yacer , acaecer, escaecer, embebecer, embobecer, embarbecer, herbecer, padecer, agradecer , obedecer, humedecer, empecer, palidecer, aridecer, languidecer, lividecer, encandecer, ablandecer , resplandecer, engrandecer, ensandecer, enardecer, atardecer, verdecer, engordecer, ensordecer, enzurdecer, enmudecer, recrudecer, enrudecer, anochecer, mohecer, entibiecer, embermejecer, vejecer, hojecer, enrojecer, calecer, escalecer, enmalecer, enralecer, fortalecer, prevalecer, convalecer, establecer, ennoblecer, encrudelecer, encruelecer, envilecer, encallecer, fallecer, tallecer, embellecer, amarillecer, amollecer, pimpollecer, , enorgullecer, entullecer, arbolecer, herbolecer, adolecer, entumecer, adormecer, emplumecer, encanecer, amanecer, hermanecer, permanecer, desvanecer, lozanecer, fenecer, pertenecer, juvenecer, orinecer, enruinecer, onecer, adonecer, embarnecer, encarnecer, escarnecer, ensarnecer, guarnecer, enternecer, fornecer, empequeñecer, retoñecer, empecer, entorpecer, carecer, aclarecer, amarecer, parecer , aparecer, enrarecer, guarecer, entenebrecer, ensombrecer, empobrecer, crecer, podrecer, merecer, perecer, ofrecer, magrecer, negrecer, enmugrecer, alborecer, arborecer, encorecer, fosforecer, florecer, amorecer, esmorecer, favorecer, empavorecer, enfervorecer, aterecer , aborrecer, enlustrecer, obscurecer, escurecer, oscurecer, escurecer, endurecer, enfurecer, apetecer, enaltecer, endentecer, lentecer, enllentecer, acontecer, entontecer, enfortecer, amortecer, abastecer, plastecer, entestecer, entristecer, robustecer, emputecer, embrutecer, frutecer, colicuecer, envaguecer, enceguecer, lobreguecer, enflaquecer, enriquecer, blanquecer, enfranquecer, enronquecer, encloquecer, enloquecer, embosquecer, abravecer, engravecer, altivecer, encalvecer, emplebeyecer, conocer, pubescer, evanescer, fosforescer. Referenecs Anderson, S. (2008). Phonologically conditioned allomorphy in the morphology of Surmiran (Rumantsch). Word Structure 1:109–134. Anderson, S. (2010). Failing one’s obligations: Defectiveness in Rumantsch reﬂexes of DĒBĒRE. In M. Baerman, G. Corbet, & D. Brown (eds), Defective Paradigms: Missing Forms and what they tell us (pp. 19-34). Oxford: British Academy/OUP. Anderson, S. (2011). Stress-conditioned allomorphy in Surmiran (Rumantsch) . In M. Maiden, J.C. Smith, M. Goldbach, & M-O. Hinzelin (eds), Morphological Autonomy: Perspectives from Romance Inflectional Morphology. Oxford: OUP. Aronoff, M. (1994). Morphology by itself. Cambridge, MA: MIT Press. Aske, Janice M. 1995. Verbal suppletion: an analysis of Italian, French and Spanish to go. Linguistics 33. 403-432. Badia, Antoni M. (1994). Gramàtica de la llengua catalana. Barcelona: Enciclopèdia Catalana Barcos, M. Á. (2007). El aragonés ansotano: estudio lingüístico de Ansó y Fago. Zaragoza: Gara d’Edizions. Bermúdez-Otero, R. (2013).The Spanish lexicon stores stems with theme vowels, not roots with inflectional class features. Probus 25(1): 3-103. Burzio, L. (2004). Paradigmatic and syntagmatic relations in Italian verbal inflection. In J. Auger, J. Clancy Clements, & B. Vance (eds), Contemporary Approaches to Romance Linguistics (pp. 17-44). Amsterdam: Benjamins. Bybee, J. (1985). Morphology: A study of the relation between meaning and form. Amsterdam: Benjamins. Paul O’Neill Bybee, J., & Pardo, E. (1981). On lexical and morphological conditioning of alternations: a nonce-probe experiment with Spanish verbs. Linguistics 19:937-968. Carreira, M. (1991). The alternating diphthongs of Spanish: A paradox resolved. In H. Campos, & F. Martínez-Gil (eds), Current Studies in Spanish Linguistics (pp. 407-445). Washington, D.C: Georgetown University Press. Corbett, G. (2007). Canonical typology, suppletion and possible words. Language 87:8-42. Díaz González, O. (1986). El habla de Candamo: aspectos morfosintácticos y vocabulario. Oviedo: Universidad de Oviedo. Eddington, D. (2004). Spanish Phonology and Morphology: Experimental and Quantitative Perspectives. Amsterdam: John Benjamins. García-Bellido, P. (1986). Lexical diphthongization and high-mid alternations in Spanish: An autosegmental account. Linguistic Analysis 16:61-92. Goyvaerts, D. (1978). Aspects of post-SPE phonology Didier L. Goyvaerts. Ghent: E. StoryScientia. Haensch, G. (2003). Las hablas de la Alta Ribagorza. Zaragoza: Instituto de Fernando el Católico. Harris, J. (1969). Spanish Phonology. Cambridge, MA: MIT Press. Harris, J. (1977). Remarks on diphthongization in Spanish. Lingua 41:261-305. Harris, J. (1978). Two theories of non-automatic morphophonological alternations. Language 54:41-60. Harris, J. (1985). Spanish Diphthongization and stress: a paradox resolved. Phonology Yearbook 2:31-45. Higginbotham, J. (1991). Remarks on the metaphysics of linguistics. Linguistics and Philosophy 14:555-566. Hualde, José Ignacio (1992). Catalan. London, New York: Routledge. Lass, R. (1976). On generative taxonomy and whether formalisms 'explain'. Studia Linguistica 30:139-154. Lloyd, P. (1987). From Latin to Spanish. Vol. I: Historical Phonology and Morphology of the Spanish Language. Philadelphia: American Philosophical Society. Lozano Sierra, C., & Saludas Bernad, Á. L. (2007). Aspectos morfosintácticos del Belsetán: aragonés del valle de Bielsa . Zaragoza: Gara d'Edizions. Malkiel, Y. (1971). Derivational transparency as an occasional co-determinant of sound change: a new casual ingredient in the distribution of -ç- and -z- in ancient Hispano-Romance. Romance Philology 25:1-52. Malkiel, Y. (1974). New problems in Romance interfixation. Romance Philology 27:304-355. Maiden, M. (1992). Irregularity as a determinant of morphological change. Journal of Linguistics 28:285-312. Malkiel, Y. (1993). The problem of 'the Old Spanish sibilants': three consecutive new-style explanations. Bulletin of Hispanic Studies 70 (2), 201-211. Maiden, M. (2001). Di nuovo sulle alternanze velari nel verbo italiano e spagnolo. Cuadernos de filología italiana 8:39-61. Maiden, M. (2004). Morphological autonomy and diachrony. Yearbook of Morphology, 137-175. Maiden, M. (2009). 'From pure phonology to pure morphology. The reshaping of the Romance verb. Recherches linguistiques de Vincennes 38:45-82. Paul O’Neill Maiden, M. (2011c). Morphomes and ‘stress- conditioned allomorphy’ in Romansh. In M. Maiden, J.C. Smith, M. Goldbach and M-O. Hinzelin, Morphological Autonomy: Perspectives from Romance Inflectional Morphology. Oxford: OUP. Martín Vegas, A. R. (2007). Morfofonología histórica del español: estudio de las alternancias /jé/ - /e/, /wé/ - /o/ y /Ø/ - /g/. Munich: Lincom Europa. Mascaró, Joan (1983). La fonologia catalana i el cicle fonològic. Bellaterra: Publicacions de la Universitat Autònoma de Barcelona. Menéndez Pidal, R. (1941:1949). Manual de gramática histórica española. Madrid: EspasaCalpe. Morin, Y. C. (1988). Morphological conditioning in phonologically transparent processes: evidence from the evolution of vowel reduction in Vinzelles Occitan. Canadian Journal of Linguistics 33:431-442. Ohala, J. (1990). There is no interface between phonology and phonetics: a personal view. Journal of Phonetics 18:153-171. O'Neill, P (2015). Velar Allomorphy in the Spanish Verb: Phonologically or Morphologically Conditioned Allomorphy? Available online at http://www.sheffield.ac.uk/hispanic/staff/profiles/pon O’Neill, P. (2011a). The Ibero-Romance Verb: allomorphy and the notion of the morphome. Unpublished DPhil thesis. University of Oxford. Available online at http://www.sheffield.ac.uk/hispanic/staff/profiles/pon O´Neill, P (2011b) Alternancias vocálicas en el presente de los verbos en -ir. Un análisis desde la Morfología Autónoma. Revista de la Historia de la Lengua Española. O'Neill (2011c). New Perspectives on the Effects of Yod in Ibero-Romance, Bulletin of Spanish Studies: Hispanic Studies and Researches on Spain, Portugal and Latin America, 89:5, 665-697 O´Neill, P (Under consideration (a)). The Origin and Spread of Velar Allomorphy in the Spanish Verb. Available online at http://www.sheffield.ac.uk/hispanic/staff/profiles/pon O´Neill, P (Under consideration (b)). Diphthongisation and stress in the Spanish verb: examples of morphological conditioning. Available online at http://www.sheffield.ac.uk/hispanic/staff/profiles/pon Penny, R. (1970). El habla pasíega : ensayo de dialectología montañesa. London: Tamesis Penny, R. (2002). A History of the Spanish Language (2nd ed.). Cambridge: CUP. Perea. Mª, P. (2002). Flexió verbal regular. In J. Solà, M.-R. Lloret, J. Mascaró, & M. Pérez Saldanya (eds), Gramàtica del català contemporani. (Vol. I, pp. 583-646). Barcelona: Empúries. Pérez Saldanya, M. (1995). Un problema de morfología verbal románica: les formes verbals velaritzades. In Mª T. Echenique, M. Aleza, Mª J. Martínez (eds), Actas del I Congreso de Historia de la lengua española en América y España. (pp. 409-420). Valencia: Tirant lo Blanch. Pérez Saldanya, M. (1995). Del llatí al català: morfosintaxi verbal històrica. Valencia: Universitat de València. Pirrelli, V. (2000). Paradigmi in morfologia. Un approccio interdisciplinare alla flessione verbale dell’italiano. Pisa, Rome: Istituti Editoriali e Poligrafici Internazionali. Pirrelli, V., & Battista. (2000). The paradigmatic dimension of stem allomorphy in Italian verb inflection. Rivista di linguistica 12 (2), 307-380. Paul O’Neill Quintana I Font, A. (2007). El aragonés nuclear de Nerín y Sercué. Zaragoza: Garza d’Edizions. Roca, I. (1990). Morphology and verbal stress in Spanish. Probus 2 (3), 321-350. Roca Pons, Josep (1968). “Morfologia verbal catalana”. Estudis Romànics, XII: 227-254. Sampson, G. (2001). Empirical Linguistics. London: Continuum. Saura Rami, J. A. (2003). Elementos de fonética y morfosintaxis benasquesas. Zaragoza: Garza d’Edizions. Schuldberg, H. K. (1984). Diphthongization in Spanish verbs. Hispanic Linguistics 1:215-228. Skousen, R. (1989). Analogical Modeling of Language. Dordrecht: Kluwer Academic Press. St. Clair, R., & Park, C. (1974). The irregular present tense verbs of Spanish . Linguistics, 73-100. Viaplana, Joaquim (1984). “La flexió verbal regular del valencià”. Miscel·lània Sanchos Guarner. Vol. I. València: Universitat de València, pp. 391-407. Williams, E. B. (1962). From Latin to Portuguese. Historical Phonology and Morphology of the Portuguese Language (2nd ed.). Philadelphia: University of Pennsylvania Press. Wheeler, Max W. (1993). “Changing inflection: verbs in North West Catalan”. Mackenzie, D; Michael, I. (eds.). Hispanic Linguistic Studies in Honour of F. W. Hodcroft, Llangrannog: Dolphin, pp. 171-206

RELATED PAPERS

RELATED TOPICS

Log In

Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes

Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes

Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes

Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes

Velar Allomorphy in Ibero-Romance: roots, endings and clashes of morphomes

Related Papers

RELATED PAPERS

RELATED TOPICS