The emergence of structure in inflection: perfect roots in irregular Galician verbs

MÉMOIRES DE LA SOCIÉTÉ DE LINGUISTIQUE DE PARIS ___ NOUVELLE SÉRIE TOME XXII MORPHOLOGIE FLEXIONNELLE ET DIALECTOLOGIE ROMANE: TYPOLOGIE(S) ET MODÉLISATION(S) PEETERS 2014 96860_Memoires_SLP_XXII_VWK.indd 596860_Memoires_SLP_XXII_VWK.indd 5 14/01/15 11:0414/01/15 11:04 SOMMAIRE INTRODUCTION ......................................................................................... 1 1. MODÉLISATION DE DONNÉES DIALECTALES ET TYPOLOGIE LEDGEWAY Adam. La morphologie flexionnelle de l’impératif des dialectes de l’Italie méridionale : la distribution de la métaphonie 13 MAIDEN Martin. Le rôle de la synonymie lexicale dans la formation du pluriel flexionnel en daco-roman .............................................. 35 LOPORCARO Michele. Contre le principe de maximisation du thème : le témoignage de la flexion verbale du sarde ................................ 51 2. DIATOPIE ET RETOUR SUR LES DONNÉES GÉOLINGUISTIQUES DUBERT-GARCÍA Francisco. Splinter morphologique et analogie dans les dialectes galiciens et portugais ................................................. 69 ESHER Louise. Allières et le verbe gascon, de l’ALG au modèle PFM 89 3. BASES DE DONNÉES ET RELATIONS IMPLICATIVES BONAMI Olivier et LUÍS Ana R. Sur la morphologie implicative dans la conjugaison du portugais : une étude quantitative .................... 109 GUERRERO Aurélie. Distribution morphomique dans la flexion verbale du catalan central ............................................................................ 153 VARIA : ÉTUDE COMPLÉMENTAIRE DUBERT-GARCÍA Francisco. The emergence of structure in inflection: perfect roots in irregular Galician verbs ........................................ 185 96860.indb 20996860.indb 209 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION: PERFECT ROOTS IN IRREGULAR GALICIAN VERBS1 Abstract Maiden (2001) identified the properties of coherenceand convergencein the perfect stems of Galician, Portuguese and Spanish irregular verbs. Coherence is responsible for analogical levelling. Convergence is responsible for the tendency shown by the irregular perfect roots of different lexemes to share morphological similarities. In this paper, I will show that the form of irregular roots, and the properties of convergence and coherence, would be fanciful without the notions of lexicalconnection, lexical strength, relevance, or typeand tokenfrequency, proposed by the Exemplar Model of morphology. The changes undergone by roots also show how morphological structure is a property emerging from storage. Key words Galician, Exemplar Morphology, verbal inflection 1. Introduction Galician is a language made up of Romance dialects that evolved in the northwest of the Iberian Peninsula. In Hispanic linguistics the northern dialects are known as dialectosconstitutivos because they have developed in situ since Romanization began. Early varieties of these dialects that were carried southward during the historical process known as the Reconquista (between the 8th century and 15th century) gave rise to varieties known as dialectos consecutivos:this is the point of origin for central Castilian, Andalusian Spanish, Portuguese and southern varieties of Catalan. The northern dialects, which descend directly from Latin, contain greater variation than the southern ones since the latter have undergone more dialect levelling. Northern variation displays the potential for linguistic change since it is possible to observe how an item (e.g. Lat. CANTĀTE, 2PL.IMP of CANTĀRE ‘to sing’) gave rise through phonological and morphological changes to several forms (Gal. cantade,cantande,cantai,cantá). Variation also allows us to test the explanatory power of linguistic theories. My goal in this paper is to 1. I wish to thank Rosario Álvarez, Maria Goldbach, Ramón Mariño, and Paul O’Neill for their comments on previous drafts of this paper. This work has been carried out with funding from the Ministerio de Ciencia e Innovación of the Spanish Government, under the project “Cambio lingüístico en el gallego actual” (FFI2012-33845). 96860.indb 18596860.indb 185 14/01/15 11:0614/01/15 11:06 186 FRANCISCO DUBERT-GARCÍA explore how changes in irregular perfect roots2 in Galician support Exemplar Morphology and its view of grammar as structure emerging from the storage of linguistic items. By perfectroots I mean the Galician root forms seen in the Past Perfect Indicative, the Pluperfect Indicative and the Past Subjunctive, which share a set of special irregularities. Maiden (2001) calls these tenses PYTA (a Spanish acronym for PerfectoyTiemposAfines), while in the Latin tradition maintained in grammars in Romance languages these tenses make up the perfectum, as opposed to the infectum, which groups together the remaining tenses, formed from verbs’ presentroots. In this paper I will analyse formal changes that were either generated in the course of the historical development of Galician or else are the product of ongoing variation. Sometimes variants compete with each other within a dialect (fixemos vs. figuemos‘we made’); other times, older forms make a come-back and prevail over innovations (puxemosvs. puñemos‘we put’). It is beyond our scope to consider each and every variant in sociolinguistic or geolinguistic detail: while I will focus on formal analysis, it should not be forgotten that the language’s formal elements are one thing, but their historical fortune another. In §2 I will set out the Galician data and their problems. In §3 I will discuss the basic theoretical tools of the Exemplar Model developed by J. Bybee and use them to explain the Galician data. I will present some conclusions in §4. 2. Verbs with irregular perfect stems Eighteen Galician verbs, together with their derivatives, have perfect stems which present certain irregularities, namely dar,ver,dicir,facer,querer, traer,caber,haber,saber,pracer,poder,ter,estar,ser,ir,poñer/pór and vir. A subset of these, the fourteen verbs listed in (1), are among the twenty most frequently used verbs in the Galician language (Bugarín 2007)3: 1. ser‘to be’ (copula) 1.9444% [of the tokens in Bugarín’s corpus] ter‘to have’ 0.9011%4 ir‘to go’ > 0.6348% haber‘to be, exist etc.’ (etymologically ‘to have’) > 0.5841% estar‘to be’ > 0.5635% 2. I will use stem, root, (word) form, morphosyntacticword,morphosyntacticrepresentation, and lexeme strictly in the sense of Matthews (1991). 3. Bugarín’s (2007) corpus contains 2,064,874 tokens obtained from written and oral sources. Unfortunately, because of the way it was conceived, it does not represent the language used by speakers in natural interactions. Due to how entries are organised, it is also not always possible to identify the morphosyntactic representation of the forms presented. Thus figures given from this source should be treated with caution as a rough approximation. 4. Figures include the verbs indicated and others derived from them. 96860.indb 18696860.indb 186 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 187 facer‘to make, to do’ > 0.5484% dicir‘to say’ > 0.54% poder‘to be able, can, may’ > 0.3651% dar‘to give’ > 0.3236% vir‘to come’ > 0.2692% saber‘to know, to taste’ > 0.2528% ver‘to see’ > 0.2519% poñer~pór‘to put’ > 0.2253% querer‘to want’ > 0.2041% In this paper, I shall ignore ser, ir, ver, and dar, since their perfect stems behave differently from the others in (1) (Álvarez & Xove 2002), but shall include traer‘to bring’ (0.0536%). Thus I will focus on dicir,facer,querer, traer,caber ‘to fit’,saber,poder,ter,estar,poñer, and vir and their derivatives; for most dialects it is necessary to add andar ‘to walk’ (0.1082%), which sometimes has an irregular root [andib] ~ [andub]. 2.1. General properties of the irregular Galician perfect stems Compare the conjugation of regular bater‘to beat’ and irregular querer ‘to want’ in (2), where the forms under the column IND.PRS show the present stem and the forms under IND.PST.PRF represent the perfect stem: 2. irregular: querer  regular: bater IND.PRS IND.PST.PRF IND.PRS IND.PST.PRF 1SG ˈkɛɾ-o ˈkiʃ-e-ŋ ˈbat-o bat-ˈi-ŋ 2SG ˈkɛɾ-e-s kiʃ-ˈɛ-tʃes ˈbat-e-s bat-ˈi-tʃes 3SG ˈkɛɾ-e ˈkiʃ-o ˈbat-e bat-ˈe-w 1 PL keɾ-ˈe-mos kiʃ-ˈɛ-mos bat-ˈe-mos bat-ˈe-mos 2PL keɾ-ˈe-des kiʃ-ˈɛ-stes bat-ˈe-des bat-ˈe-stes 3PL ˈkɛɾ-e-ŋ kiʃ-ˈɛ-ɾoŋ ˈbat-e-ŋ bat-ˈe-ɾoŋ Note that the irregular perfect stems display a special stress pattern in the 1SG/3SG.IND.PST.PRF: in these forms the root is stressed: quíx-en,quíx-o. In the regular perfect, the root is unstressed throughout: bat-ín,bat-éu. In the irregular forms, the 3SG.IND.PST.PRF ends in [o], which Santamarina (1974) and Álvarez & Xove (2002) treat as an allomorph of the personnumber suffix (PN) in complementary distribution with regular [w]. The allomorph [o] appears as a syllabic nucleus immediately following a stressed root ending in a consonant: quíx-o; [w] (spelt <u>), on the other hand, follows regular stems ending in a vowel: cant-ó]-u,com-é]-u,part-í]-u. The vowel preceding [w] is the thematic vowel (TV). (A reference list of abbreviations used in this paper precedes the bibliography.) The irregular forms of the 1SG.IND.PST.PRF have the vowel [e] after the stressed root: quíx-e-n. Except for the 3SG.IND.PST.PRF, in all other forms of the perfect group of tenses the root is unstressed and followed by [ɛ] which 96860.indb 18796860.indb 187 14/01/15 11:0614/01/15 11:06 188 FRANCISCO DUBERT-GARCÍA appears both in stressed (e.g. quix-é-mos) and unstressed pretonic syllables (e.g. quix-e-rá-mos). Both [e] and [ɛ] are considered allomorphs of the TV (Santamarina 1974, Álvarez & Xove 2002, Fernández Rei 1998) which appears no matter what conjugation the verb is assigned to in the present tenses: thus estar belongs to the 1 st conjugation in the present, saber to the 2 nd and dicir to the 3 rd , yet all three take [e] and [ɛ] as their VT in the perfect. One last point that should be noted at this stage is that these irregularperfect verbs have different roots in the present and the perfect stems. Their perfect stems in standard Galician are illustrated in (3): 3.  Present Perfect 1 PL.IND.PRS 1PL.IND.PST.PRF  dicir [diˈθi-mos] [diˈʃɛ-mos]  facer [faˈθe-mos] [fiˈʃɛ-mos]  querer [keˈɾe-mos] [kiˈʃɛ-mos]  pór/poñer [ˈpo-mos/poˈɲe-mos] [puˈʃɛ-mos]  traer [tɾaˈe-mos] [tɾowˈʃɛ-mos]  estar [esˈta-mos] [estiˈbɛ-mos]  caber [kaˈbe-mos] [kowˈbɛ-mos]  haber [aˈbe-mos] [owˈbɛ-mos]  saber [saˈbe-mos] [sowˈbɛ-mos]  pracer [pɾaˈθe-mos] [pɾowˈɡɛ-mos]5  poder [poˈde-mos] [pujˈdɛ-mos]  ter [ˈte-mos] [tiˈbɛ-mos]  vir [ˈbi-mos] [biˈɲɛ-mos] The rest of the exponents of the irregular forms are fully predictable, identical to those of the regular forms. To sum up, irregular perfectforms in Galician show specific exponents that are in some way identifiable,predictableandparadigmatic. The place of stress in the 1SG/3SG.IND.PST.PRF, the PN [o] in the 3SG.IND.PST.PRF, and the TVs [ɛ] ~ [e] are all identifiable elements that constitute the paradigm of this set of verbs. 2.2. Perfect roots Maiden (2001) states that the distribution and form of irregular perfect roots are not fully predictable from phonological features, morphosyntactic 5. Pracer ‘to please’, a less frequent irregular verb. In western dialects there is no /ɡ/ phoneme, which in the standard variety and eastern dialects represents the allophones [ɡ] and [ɰ] in complementary distribution (Álvarez and Xove 2002, p. 40); the lexical place of /ɡ/ is occupied by /h/: eastern /faɡo/, 1SG.IND.PRS of facer, western /faho/. Hence wherever the reader finds /ɡ/ we should understand that this consonant’s place may be occupied by /h/ and vice versa. The phenomenon of using [h] is known as gheada in Galician grammars (Fernández Rei 1990, p. 163-188). 96860.indb 18896860.indb 188 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 189 properties or semantic values. Perfect roots appear in the IND.PST.PRF (e.g. quix-en), the IND.PST.PPRF (quix-era) and the SBJV.PST (quix-ese). Although the property PST appears in all the morphosyntactic representations, its presence is not a sufficient condition: the Imperfect Indicative also has PST in its morphosyntactic representation (IND.PST.IPRF) but takes the present root: quer-ía.6 The semantic feature ‘perfective’ is not responsible for the selection of the irregular root: as (4) shows, SBJV.PST may lack this feature: 4. semañáofixese,sabíao ‘If he did it tomorrow, I would know (it)’  seagoraofixese,sabíao‘If he did it today, I would know (it)’ Neither is the semantic feature ‘past’ sufficient: the Conditional forms in (5) indicate a future event within the past and they use present roots: 5. Anosdespoishaberíacartossuficientes ‘Some years later, there would have been money enough’  Onteviríaseodeixasen “He would have come yesterday, if they had let him” The Portuguese Future Subjunctive(which has fallen into disuse in Galician) does not refer to the past either, and uses irregular perfect roots: Seo fizeresbem,nãoterásdearrepender-te ‘If you do it right, you will not have to be sorry’. Irregular roots have familyresemblances: they share similarities, although there is no single feature common to all the forms.7 Galician grammarians have always been aware of this (Álvarez and Xove 2002, p. 255; Álvarez, Regueira, and Monteagudo 1986, p. 318). Thus in the stems in (3) the last vowel of the verb’s root (the rootvowel, RV) is high: pór/poñer,[puʃ]; querer,[kiʃ]; vir,[biɲ]. Diphthongs may also appear: [ow] in [owb], [sowb], [tɾowʃ], and [uj] in poder[pujd]. In the dialects other combinations occur: e.g. traermay have the perfect root forms [tɾowʃ] ~ [tɾuʃ] ~ [tɾojʃ]; saber, [sowp] ~ [sup]; haber,[owb] ~ [ub]; poder,[pud] ~ [pujd] ~ [pojd]. In some verbs, the last consonant of the root (the rootconsonant, RC) is often a postalveolar fricative: pór/poñer [puʃ], traer[tɾowʃ], but we can also find [ɲ] (viñ-emos), [b] (soub-emos). In the dialects, the perfect root of traer may be [tɾowɡ] ~ [tɾowʃ] ~ [tɾu]; pór/poñer, [puʃ] ~ [puɲ]; saber, [sowp] ~ [sowb]. The forms listed in (6) exemplify these family resemblances: 6. The Imperfect Indicative may be interpreted either as a past imperfective or as a present in the past. In any case it is a form used to express the past. 7. Ladefoged and Maddieson apply the notion of familyresemblancesto rhotics: “Each member of the rhotic class resembles some other member with respect to some property, but it is not the same property that constitutes the resemblance for all members of the class” (1996, p. 245). 96860.indb 18996860.indb 189 14/01/15 11:0614/01/15 11:06 190 FRANCISCO DUBERT-GARCÍA 6. pór/poñer :[puʃ], [puɲ]  traer :[tɾowʃ], [tɾuʃ], [tɾuj], [tɾu], [tɾojʃ], [tɾowɡ], [tɾojɡ]  saber :[sowp], [sowb], [sup]  estar : [estub], [estib] etc. These facts led Maiden (2001) to detect two properties of irregular perfect roots. The first of these he calls coherence: ‘Coherence’ means that the mutually implicational relationship between preterites and non-present subjunctives has stood indissoluble throughout history. […] Nothing has sundered the root of the preterite from that of the non-present subjunctives. Any analogical change affecting a PYTA [i.e. perfect] root affects that root throughout the paradigm of the relevant verb. […] Replacement of the PYTA root by its non-PYTA counterpart never creates ‘mixed’ systems (Maiden 2001, p. 444) Coherenceexplains the relationship among the perfect roots within each lexeme: every perfect stem of a lexeme shares one root throughout its history; and if a phonological change produces a different root in one element of the paradigm, the other roots may be reconstructed by analogy. For example, in the roots pór/poñer in (7) we see phonological changes in the form corresponding to 1SG.IND.PST.PRF that ended up being transmitted to the whole class: 7. Latin Proto-Romance Old Galician Modern Galician PŎSU-Ī *[poz]-i [puʒ]-i8 [puʃ]-en PŎSU-ISTĪ *[poz]-esti [poz]-este [puʃ]-eches PŎSU-IT *[pos] [pos] [puʃ]-o PŎSU-imus *[poz]-emos [poz]-emos [puʃ]-emos The final [i] in the 1SG.IND.PST.PRF *[ˈpozi] in Proto-Romance raised the RV from [o] to [u], giving [ˈpuz i];9 next, the RC, an apico-alveolar fricative [z ], is palatalized by the final [i], giving [ˈpuʒi] in Old Galician; the new root [puʒ] spread by paradigmatic levelling to all the perfect tenses and formsof pór/poñer ; while a late (or simultaneous) devoicing of the coronal fricatives created the [puʃ] of Modern Galician: pux-en,pux-eches,pux-o (Santamarina 1974, p. 45). 8. Martínez Mosquera (2000/2001, p. 109) finds the following roots for the 1SG.IND.PST. PRF in old Galician: [poʒ], [puʒ], [puʃ], [pos] and [pus], sometimes followed by [e] as TV and sometimes by [i]. See Varela (1999) for a chronology. 9. PŎSUĪ had a stressed Ŏ, [ɔ] in Proto-Romance (Santamarina 1974 suggests PŌSUĪ as the etymon). Penny (2002, p. 226) and Ferreiro (1995 p. 37 and 333) speak of a first harmony caused by the U and a second one by the ī. Williams (1938, p. 236) speaks of the reduction of [ɔw] in Vulgar Latin, with subsequent raising of [o] by umlaut: *[pɔzwi] > *[pɔwzi] > [poz i] > [puz i]. 96860.indb 19096860.indb 190 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 191 The second property Maiden describes is convergence: ‘Convergence’ means that PYTA forms assume a common phonological shape, resembling each other more and more, and the individual lexical verb to which each root belongs, less and less. Moreover, convergence is peculiar to PYTA, in that the relevant verbs show no sign of convergence in any tense-forms outside PYTA. (Maiden 2001, p. 447). This property, which establishes formal links among the irregular forms of the different lexemes, is related to family resemblances. Xavier Varela (1999), using the term paradigmaticsolidarities, studies this phenomenon in the sigmaticverbs(those which present schemata close to that of 7).Convergence explains why the perfect roots of different lexemes tend to be more similar to each other, and less similar to the present root forms of their lexemes. Thus, the Romance outcome of Latin FĒCĪ was [ˈfizi], with a laminoalveolar [z] (Varela 1999, p. 1051); in some dialects, the RC [z] regularly evolved to [θ], giving rise to [fiθ] in the perfect. However, in other dialects, the perfect stem of old fazerdeveloped a root that ended in a postalveolar fricative, [fiʃ] (Santamarina 1974, p. 42 and Varela 1999, p. 1053). This change is attributable to convergence : the perfect rootof old fazer(presentday facer) changed by analogy with the verbal forms that spontaneously developed a postalveolar fricative: [ˈfizi] > [ˈfiʒi], like [ˈpuʒi], [ˈdiʃi].10 The new root [fiʒ] convergeswith the perfect roots of querer[kiʒ], pór/poñer [puʒ], with postalveolar RC, and divergesfrom the present root of fazer, with lamino-alveolar RC : faço,fazer,fazía. The perfect rootof traeralso illustrates convergence. The root in Standard Galician is [tɾowʃ]; in the dialects, we find [tɾowɡ] (compare the maps 381, 382 and 386 in ALGaI). However, there also exist variants that alternate different RVs, recombined with other RCs: [tɾowʃ], [tɾowɡ], [tɾojʃ], [tɾojɡ], [tɾuʃ], [tɾuɡ]… These variants of RVs occur because dialects take different elements of the formal possibilities detected in the perfect roots, which constitute the elements of the family resemblances. These formal elements move the perfect root forms of traer away from their present root forms, showing the power of convergence. 2.3. Stubborn and unruly forms Although Maiden (2001) correctly attributes the properties of coherence and convergence to the irregular roots, these only represent two general tendencies, whose explanatory power may be limited if there are exceptional, but systematic, word forms that fail to obey them. Stubbornforms. Some dialects present roots resisting the coherence principle in the 3SG.IND.PST.PRF. In Portuguese, more conservative than Galician 10. The forms of old fazer had a fricative lamino-alveolar RC (not an apico-alevolar one; Varela 1999); lamino-alveolar fricatives were not affected by palatalization. 96860.indb 19196860.indb 191 14/01/15 11:0614/01/15 11:06 198 FRANCISCO DUBERT-GARCÍA In Figure 1 we see how the schema that segregates the root [bat] from bat-o,bat-es,bat-a,bat-ín is generated. We also see the emergence of the schema {[o] ⇔ V[1SG.IND.PRS]}, which appears in bat-o,com-o,vex-o; the schema {[a] ⇔ V[SBJV.PRS} in bat-a,com-a,vex-a; the schema {[des] ⇔ V[2PL]} in bate-des,come-des,ve-des,canta-des, parti-des, etc. It is in this sense that we can say that exponents like [des] and roots like [bat] emerge from stored material. The schema {[o] ⇔ V[1SG.IND.PRS]} is abstract: it is obtained by comparing cant-o,vend-o,coll-o,sig-o,pid-o; it works in the three Galician conjugations. It may become more concrete if we find more commonalities: the word forms with {[o] ⇔ V[1SG.IND.PRS]} are stressed on the penultimate syllable, so that we could add this feature, {[o] ⇔ V[1SG.IND.PRS] ⇔ [...σσ]}; it appears linked to the root, without any intervening TV; these commonalities put aside the differences (such as which roots it links to, information about the verb’s conjugation, the lexeme’s meaning, etc.). The schema is used when an infrequent verb, which must be constructed on the fly, is conjugated: zarp-o, 1SG.IND.PRS of zarpar ‘to set sail’ (the least used verb in the Bugarín 2007 corpus). Geert Booij (2010, p. 41-42) highlights four properties of schemata: they a) have variables, elements that may be filled by different units; b) may embed within each another, producing a hierarchy in which the more abstract Figure 1 96860.indb 19896860.indb 198 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 199 ones encompass the more concrete ones; c) represent knowledge in each one of these levels of abstraction; and d) are active elements that speakers use to produce or analyse new utterances. Schemata also have diverse degrees of lexical strength, depending on the frequency of the items containing them, the token frequency, and the frequency of each schema, the type frequency: “The strength of lexical representations of individual items is in part a reflection of token frequency, while the strength [...] of lexical associations or schemas is built up by the type frequency” (Bybee 1995, p. 452). A strong schema may spread to items that did not contain it before. In Galician, [des] is the default exponent of 2PL in verbs: [de] is only used in the 2PL. IMP (cantade) and [stes] only in the 2PL.IND.PRT.PRF (cantastes); [des] is used elsewhere. In a lot of dialects, [des] spread to the forms of the 2PL.IND.PST.PFR, where 2PL was initially marked with [stes]; the spread of [des] created a hyper-characterized schema {[stedes] ⇔ V[2PL.IND.PRT.PRF]} (collestes> colléstedes) by virtue of the lexical strength of {[des] ⇔ V[2PL]}. In order to understand how {PÓR/POÑER ⇔ V[1SG.IND.PRS/SBJV.PRS] ⇔ [poɲV]} spread to other present forms in Paradigm B of (11) (pó-mos > poñé-mos), we should note that: A) The root [poɲ] appears in the 1SG.IND.PRS and in the SBJV.PRS; in Bugarín’s corpus (2007, p. 830) the forms corresponding to these properties constitute 7.989% of the uses of pór/poñer and 12.388% of its uses in the present-root forms. The forms of 1SG.IND.PRS figure among the most frequent, and, in so doing serve as a model (Bybee 1995). Thanks to this frequency, a schema emerges that links PÓR/POÑER with [poɲV], a sequence that may spread to the rest of the present-root forms. The new schema of dialect-type B in (11) may be represented as {PÓR/POÑER ⇔ V[INFECTUM] ⇔ [poɲV]},13 meaning that [poɲ] and the variable [V] are no longer linked only to the properties 1SG.IND.PRS and SBJV.PRS and the exponents [o] and [a], and become linked to a larger set: all the present-root tense exponents, NP [o], TAM [a] and the allomorphs of the TV (poñ-o,poñ-emos,poñ-ía). The forms corresponding to the 2SG.IND.PRS, 3SG.IND.PRS and 2SG.IMP resist, owing to their individual token frequency. B) To the lexical strength of {PÓR/POÑER ⇔ V[INFECTUM] ⇔ [poɲV]} there must be added the lexical strength of the inflectional schemata of regular verbs. In these verbs, we find a TV between the root and the TAM and NP affixes: com-e-mos,com-e-re-mos. In these schemata, the TV is associated with the [V] which follows the regular root. So in PÓR/POÑER not only does the root [poɲ] spread, but the word forms also take all the affixes and prosodic schemata corresponding to regular verbs, as shown in (13): 13. The notations V[INFECTUM] and V[PERFECTUM] will be used here as cover terms for the sets of morphosyntactic words realized through the corresponding root, e.g. {V[INFECTUM] ⇔ [poɲ]} refers to all the verbal forms of pór/poñerwith the root [poɲ]: the 1SG.IND.PRS, 1PL. IND.PRS, 2PL.IND.PRS, SBJV.PRS, etc. Thus, INFECTUM and PERFECTUM are variables. 96860.indb 19996860.indb 199 14/01/15 11:0614/01/15 11:06 200 FRANCISCO DUBERT-GARCÍA 13. {V[1PL.IND.PRS] ⇔ [-ˈV-mos]} old [ˈpo-mos] > new [poɲ-ˈe-mos], like [komV] > [kom-ˈe-mos] { V[1SG.IND.PST.IPRF] ⇔ [-ˈi-a]} old [ˈpuɲ-a] > new [poɲ-ˈi-a], like [komV] > [kom-ˈi-a] So, the TV’s introduction is also a requirement of the regular verbs’ schemata, those with the higher frequency type. In these schemata, the TAM and the PN are preceded by the TV, and the TV is preceded by a regular root. The innovation poñ-e-mos results from the spread of the root’s allomorph [poɲ], based on its token frequency, and the spread of the use of the regular prosodic patterns and affixes, based on their type frequency. The restrictions and rules that guide this process are not stored outside of the forms, but emerge from the stored material. Figure 2 represents part of the network that links the forms of pór/poñer in Paradigm A in (11) before levelling; figure 3 represents parts of the network in Paradigm B in (11) after levelling; note the autonomy of [ˈpɔs] 2SG.IND.PRS and [ˈpɔŋ]3SG.IND.PRS. Figure 2 Figure 3 3.2. Regularizations in the perfect root and emerging structure We have seen in (9) and (11) that facer and pór/poñer present evolutions opposite to convergence, since they take elements of their present tenses to create their perfect forms. In (9) we have seen how [fih] spread to all the perfect, adding coherence. In (11), the root [puɲ] encounters resistance in the 1SG and 3SG of the IND.PST.PRF. But the most interesting fact is that, in order to create [puɲ] (the perfect root of pór/poñer), only the present RC [ɲ] was taken, while the present RV [o] was ignored; and in [fih] (perfect 96860.indb 20096860.indb 200 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 201 rootof facer), the present RC [h] was adopted, while the present RV [a] was ignored. So, the levelling process only took into account the present RCs [ɲ] and [h], not the present RVs [o] and [a]. The change did not affect the irregular allomorphs of the TV and stress patterns: the usual characteristics of irregular perfect forms are maintained. Figure 4 shows the lexical connections in three evolutionary stages of the pór/poñer’s perfect forms.In Old Galician, [ˈpuʒi] 1SG.IND.PST.PRF was the outcome of phonological evolution; the perfect paradigm of pór/poñerwas not coherent ([ˈpuʒ-i] vs. [ˈpoz-o]), but was convergent: the present and perfectroots were different. Coherence was recovered and convergence reinforced in Common Galician when the paradigm was regularized through the spread of [ˈpuʒ] to the rest of the perfect paradigm. However, the innovative system that is now being created through the spread of the RC [ɲ] reduces coherence, since the change leaves [ˈpuʃeŋ] 1SG.IND.PST.PRF and [ˈpuʃo] 3SG.IND.PST.PRF unaffected; and the innovative system violates convergence, since the perfect abandons the RC [ʃ] and takes [ɲ] from the present root. Figure 4 In order for [fih] and [puɲ] to appear in the perfect, it is necessary for schemata like {FACER ⇔ V ⇔ [fVh]} and {PÓR/POÑER ⇔ V ⇔ [pVɲ]} to emerge; in these cases, the phonological representations are more abstract, [fVh] and [pVɲ], with a variable [V] that only tells us that the roots must contain a RV. The RCs [h] in [fih-ɛmos] and [ɲ] in [puɲ-ɛmos] agree with those of the present root, but the high RVs [i] in [fih] and [u] in [puɲ] are taken from the batch of perfect formal resources. In the case of pór/poñer, {V ⇔ [pVɲ]} is more abstract than {V[PERFECTUM] ⇔ [puɲ]} or {V[INFECTUM] ⇔ [poɲ]}; both {V[PERFECTUM] ⇔ [puɲ]} and 96860.indb 20196860.indb 201 14/01/15 11:0614/01/15 11:06 202 FRANCISCO DUBERT-GARCÍA {V[INFECTUM] ⇔ [poɲ]} inherit the basic form of their roots from {V ⇔ [pVɲ]}: {PÓR/POÑER ⇔ V ⇔ [pVɲ]} {PÓR/POÑER ⇔ V[INFECTUM] ⇔ [poɲ]} {PÓR/POÑER ⇔ V[PERFECTUM] ⇔ [puɲ]}  poñ-o,poñ-emos,poñ-erei puñ-emos,puñ-eses In order to explain puñ-eches,puñ-estes,puñ-ese we need another schema to preserve the RV [u] in the perfect. The schema that explains why RV [u] remains in [puɲ] is {V[PERFECTUM] ⇔ [u]RV}, which also appears in many dialectal forms: trux-emos,tru-emos,hub-emos,sup-emos. This is an instantiation of another more abstract schema, {V[PERFECTUM] ⇔ [+high]RV}, which dominates and covers not only trux-emos,tru-emos,hub-emos, but fix-emos,quix-emos,dix-emos, instantiations of {V[PERFECTUM] ⇔ [i]RV}, also responsible for the RV of [fih-ɛmos]. There are even some verbs that alternate between both schemata: tiv-emos ~ tuv-emos, estiv-emos~estuvemos, andiv-emos~anduv-emos. In other verbs, {V[PERFECTUM] ⇔ [u]RV} may alternate with {V[PERFECTUM] ⇔ [ow]RV}:14 troux-emos ~ trux-emos, soupemos~soubemos, houbemos~hubemos, etc.: {V[PERFECTUM] ⇔ [+high]RV} { V[PERFECTUM] ⇔ [i]RV} {V[PERFECTUM] ⇔ [u]RV}  fix-emos,quix-emos pux-emos,trux-emos The root [puɲ] shows the conflation of various hierarchical schemata: { PÓR/POÑER ⇔ V ⇔ [p V ɲ]} | | | { PÓR/POÑER ⇔ V[PERFECTUM] ⇔ [p u ɲ - ɛ - mos]} | { V[PERFECTUM] ⇔ [u]} | { V[PERFECTUM] ⇔ [+high]} Schemata are not rules with inputs and outputs, but elements that emerge and give structure to the stored individual representations. The schema {V[PERFECTUM] ⇔ [+high]} emerges from the storage of fixen,quixen, puxen,truxenetc. It has psychological reality since it is a guiding element in linguistic change. Levelling creates forms like puñ-emos because speakers extract schemata like {PÓR/POÑER ⇔V[PERFECTUM] ⇔ [u]} and {PÓR/ POÑER ⇔ V ⇔ [pVɲ]} from the stored material. 14. The forms containing diphthongs as RVs (troix-,puid-,poid-) are also subject to {V[PERFECTUM] ⇔ [+high]RV}, since semivowels are asyllabic realizations of high vowels. 96860.indb 20296860.indb 202 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 203 Figure 5 shows a proposal for the representation of the lexical connections of Paradigm D for pór/poñer in (11): [puʃ-eŋ]and [puʃ-o]are autonomous, since they maintain their own RC [ʃ], but use the RV [u], characteristic of perfect roots. Figure 5 However, as already mentioned, the ALGa shows the birth of the new Paradigms E and F of pór/poñershown in (14): 14. Paradigma E Paradigma F 1 SG.IND.PST.PRF [puɲ]-en [puɲ]-en 2SG.IND.PST.PRF [puɲ]-eches [puɲ]-eches 3SG.IND.PST.PRF [puʃ]-o [puɲ]-o 1 PL.IND.PST.PRF [puɲ]-emos [puɲ]-emos These paradigms originate from the strength of the lexical connections of [puɲ], based on type frequency. With the spread of [puɲ] to the 1/3SG of the IND.PST.PRF coherence is restored. The form of 1SG is regularized before that of 3SG; this is also happening with vir, where the creation of a coherent form viñ-en in the 1SG is not uncommon; in those dialects where viñ-en replaces vin, only the stubborn and autonomous forms of the 3SG (veu,viu, vén) remain. We can understand the preservation of irregularity in the 3SG puxo i n connection with the fact that, in the Bugarín (2007) corpus, the use index of 3SG.IND.PST.PRF is 40.27% of all perfect forms, but the use index of 1SG is only 13.11%; while the 3SG.IND.PST.PRF of virhas 42.57% and the 1SG only has 8.05%. Use of the 3SG is so frequent that its forms acquire autonomy. Paradigm F in (14) is a consequence of the strength of type frequency: in Paradigm F, complete coherence is restored. But frequency does not explain everything; it must interact with other linguistic factors: 96860.indb 20396860.indb 203 14/01/15 11:0614/01/15 11:06 204 FRANCISCO DUBERT-GARCÍA vir facer pór/poñer 1SG.IND.PST.PRF 8.05% 6.97% 13.11% 3SG.IND.PST.PRF 42.57% 24.94% 40.27% 3PL.IND.PST.PRF 13.16% 10.34% 16.02% Use indices of some perfect forms (Bugarín 2007) The 3PL forms have use indices intermediate between those of 1SG and 3SG. These indices should guarantee autonomy and resistance against regularization, yet the 3PL forms behave the same way as the other perfect forms. The only explanation is that the location of stress is significant and helps to group these forms with the others.15 Changes like [fiʃ] ~ [fiθ] > [fih] under the influence of [fah] or those like [puʃ] > [puɲ] under the influence of [poɲ] show that the root and its parts, RVs and RCs, are also emergent units. However, RVs and RCs do not only constitute a part of the root: they are elements with their own personality and may operate with some degree of autonomy. The root [puʃ] in Paradigm C in (11) points to IND.PST.PRF, IND.PST.PPRF and SBJV.PST enbloc. Nevertheless, the root [puʃ] in Paradigm D in (11) may be analysed as containing the RV [u], an index of the perfect,plus the RC [ʃ], an index of 1/3SG.IND.PST. PRF. In Paradigm E in (14), the RC [ʃ] only marks the 3SG.IND.PST.PRF. Finally, in Paradigm F in (14), the RV [u] is the only mark of the perfect, since the RC [ɲ] is shared with the present root. The uses of the RC [ʃ] to mark 1/3SG.IND.PST.PRF in Paradigm D in (11) and to mark 3SG.IND.PST.PRF in Paradigm E in (14) result in an increment in the root’s structure, with the creation of a morphological contrast inside a previously unanalysable unit; this phenomenon Jespersen called secretion: By secretion I understand the phenomenon that one portion of an indivisible word comes to acquire a grammatical signification which it had not at first, and is then felt as something added to the word itself. Secretion thus is a consequence of a ʻmetanalysisʼ...; it shows its full force when the element thus secreted comes to be added to other words not originally possessing this element. (Jespersen 1922, p. 384) We may represent secretion as in (15): 15. [p u ʃ] > [p u ʃ] | | RV RC]R 15. Nevertheless, in the present-root tensesof pór/poñer some dialects that have extended the use of the root [poɲ] and created Paradigm B of (10) keep [pɔŋ] forms without the analogical palatal nasal (and, therefore, with neither TV nor PN) in the 3 PL . IND . PRS , rather than creating form [ˈpɔɲ-e-ŋ] by analogy; the same is true of the 3PL.IND.PRS of ter and vir, which tend to preserve the old forms [ˈtɛŋ] and [ˈbɛŋ], respectively, in the same dialects and do not create the analogical [ˈtɛɲ-e-ŋ] and [ˈbɛɲ-e-ŋ]. In these cases, frequency helped to preserve the more conservative forms. 96860.indb 20496860.indb 204 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 205 In order to understand why the expression of 1SG/3SG.IND.PST.PRF in pór/ poñeris regularized in Paradigms E and F of (14) in spite of being autonomous, we must present another notion developed by Bybee (1985), that of relevance. A category “is relevant to the verb to the extent that themeaning ofthecategorydirectlyaffectsthelexicalcontentoftheverb” (Bybee 1985, p. 15). When the meaning of the category is very relevant to the verb, its expression tends to affect the verb root; when it is less relevant, it can be expressed through affixes or syntactic procedures. Person agreement with the subject is not as relevant as tense, mood or aspect, “since it refers to an argument of the verb, and not to the action or state described by the verb itself”. Thus, one assumes that person agreement should not affect the form of the root. So the autonomy of the forms corresponding to 1SG/3SG.IND.PST. PRF in Paradigms E and F goes against relevance, since it marks person agreement. Regularization of the root eliminates the forms marked for person agreement. If this analysis is correct, the restoration of coherence in this case is the outcome of relevance. 4. Conclusions Martin Maiden (2001) identified two properties of perfect roots in Spanish irregular verbs that are also found in Galician and Portuguese irregular verbs: coherenceand convergence. Because of coherence, perfect roots tend to be identical within the paradigm of each lexeme; coherence is responsible for the analogical reconstruction of roots within the perfect paradigm. Because of convergence, the perfect forms of a given lexeme tend to share formal similarities with perfect forms of other lexemes. However, the form and distribution of perfect roots (and the properties of convergenceand coherence) would be fanciful and unexplainable without notions like lexicalconnec tion, lexicalstrength, relevance, and frequencyofuse, proposed in the Exemplar Model. The changes undergone by forms throughout history also show how linguistic structure is a property emerging from storage. Frequency of use is a basic element in Exemplar Morphology, since the form of the grammar depends in part on this parameter. Tendencies like coherence and convergence are expressions of the emergenceof patterns linked to frequency: The notion of emergent structure has become important in various branches of the sciences in the last two decades. The basic idea is that what may appear to be a coherent structure created according to some underlying design may in fact be the result of multiple applications or interactions of simple mechanisms that operate according to local principles and create the seemingly well-planned structure as a consequence. (Bybee and Hooper 2000, p. 10) Therefore, frequency of use, lexical connections, lexical strength, and relevance are some of the elements responsible for producing and maintaining coherence and convergence in Galician (and Portuguese and Spanish) verbs. 96860.indb 20596860.indb 205 14/01/15 11:0614/01/15 11:06 206 FRANCISCO DUBERT-GARCÍA Abbreviations 1 First person 2 Second person 3 Third person FUT Future IND Indicative IPRF Imperfect PL Plural PN Person and Number suffix PPRF Pluperfect PRF Perfect PRS Present PST Past PTCP Participle R Root RC Root consonant, last consonant of the root RV Root vowel, last vowel of the root SBJV Subjunctive SG Singular TAM Tense, Aspect and Mood suffix TV Thematic vowel V Verb V Vowel References ALGaI = Instituto da Lingua Galega (1990). AtlaslingüísticoGalego.Vol.1.Morfoloxíaverbal, A Coruña, ILG/Fundación Pedro Barrié de la Maza. ÁLVAREZ, Rosario and XOVE, Xosé (2002). Gramáticadalinguagalega, Vigo, Galaxia. ÁLVAREZ, Rosario, REGUEIRA, Xosé Luís and MONTEAGUDO, Henrique (1986). Gramáticagalega,Vigo, Galaxia. BABARRO, Xoán (2003). GalegodeAsturias :delimitación,caracterizaciónesituaciónsociolingüística, A Coruña, Fundación Pedro Barrié de la Maza. BOOIJ, Geert (2010). ConstructionMorphology, Oxford: OUP. BUGARÍN, María Xesús (2007). Ogalegofundamental.Dicionariodefrecuencias, A Coruña, Fundación Pedro Barrié de la Maza. BYBEE, Joan (1985). Morphology, Amsterdam, John Benjamins. BYBEE, Joan (1988). “Morphology as lexical organization”, Hammond, M. and Noonan, M. (eds.), Theoreticalmorphology,San Diego, Academic Press, 119-41. BYBEE, Joan (1995). “Regular morphology and the lexicon”, LanguageandCognitiveProcesses 10, 425-55. BYBEE, Joan (2001). Phonologyandlanguageuse, Cambridge, CUP. BYBEE, Joan (2010). Language,usageandcogntion, Cambridge, CUP. BYBEE, Joan and HOOPER, PAUL (eds.) (2001). Frequencyandtheemergenceoflinguisticstructure,Amsterdam, John Benjamins. 96860.indb 20696860.indb 206 14/01/15 11:0614/01/15 11:06 THE EMERGENCE OF STRUCTURE IN INFLECTION 207 BEVLINS, James P. (2006): “Word-based morphology”, JournalofLinguistics 42, 531-573. FERNÁNDEZ REI, Francisco (1990). Dialectoloxíadalinguagalega, Vigo, Xerais. FERNÁNDEZ REI, Francisco (1998). “Vocal temática e conxugación verbal. A expresión da clase mórfica en español, catalán e galego moderno”, Kremer, Dieter (ed.), HomenaxeaRamónLorenzo,Vol.2, Vigo, Galaxia, 591613. FERREIRO, Manuel (1995). Gramáticahistóricagalega, Santiago de Compostela, Laiovento. GLEASON, H. A. (1961). AnIntroductiontoDescriptiveLinguistics.RevisedEdition, New York, Holt, Rinehart and Winston. JESPERSEN, Otto (1922). Language.ItsNature,DevelopmentandOrigin, London, George Allen & Unwind Ltd. LADEFOGED, Peter and MADDIESON, Ian (1996). TheSoundsoftheWorld’sLanguages, Oxford, Blackwell. LANGACKER, R.W. (2000). “A dynamic usage-based model”, Barlow, M. and Kemmer, S. (eds.), Usagebasedmodelsoflanguage, Standford, CSLI: 1-63. MAIDEN, Martin (2001). “A strange affinity: ‘perfecto y tiempos afines’”, Bulletin ofHispanicStudies78, 441-464. MARTÍNEZ MOSQUERA, Guadalupe (2000/2001). Olexemadoverbopõer/poerna épocamedieval,Master thesis (unpublished), Santiago de Compostela, USC. MATTHEWS, Peter (1991). Morphology.SecondEdition,Cambridge, CUP. PENNY, Ralph (2002). AHistoryoftheSpanishLanguage, Cambridge, CUP. SANTAMARINA, Antón (1974). Elverbogallego, Santiago de Compostela, USC. VARELA, Xavier (1999). “A evolución dos pretéritos irregulares sigmáticos”, Álvarez, Rosario and Vilavedra, Dolores (eds.), Cinguidosporunhaarelacomún: HomenaxeóprofesorXesúsAlonsoMontero,Vol.1,Santiago de Compostela, USC, 1047-1072. WILLIAMS, Edwin (1967). FromLatintoPortuguese.SecondEdition, Philadelphia, University of Pennsylvania Press, Portuguese translation DoLatimaoPortuguês, Rio de Janeiro, Tempo Brasileiro, 1975. Francisco DUBERT-GARCÍA 96860.indb 20796860.indb 207 14/01/15 11:0614/01/15 11:06