(Extreme) Polymorphism in Occitan Verb Morphology

Abstract

Polymorphism has long been recognized as a crucial dimension of the nature of language. One of the merits of dialectology and dialectologists is emphasis on the inherently variable and polymorphic nature of linguistic systems, which are always in a state of relative equilibrium and stability. The most striking features of the Occitan data that will be discussed lies in the possibility of finding various forms in a given cell in certain paradigms; more strikingly, it will be shown that two or three (even four) paradigms for one and the same (tense) verb may coexist in the same variety. It will be argued that if polymorphism is the natural state of linguistic systems, it is also anti-economic from a cognitive and processing point of view. It follows that the diachronic evolution of languages tends to develop adaptive solutions to circumvent the potential drawbacks of extreme polymorphism: “natural selection” leads to the reduction or elimination of morphological proliferation. Of course, before reduction or elimination take place, a more or less extended period of time may elapse during which a preference for some paradigmatic options may arise.

1. Introduction

The very development of linguistics thus reveals that linguists should not only try to discover regularities as general as possible but also to fight, even more intensely, against the excessive, mechanical simplification of language phenomena.

(Mathesius [1911] 1964, p. 2)1

Polymorphism—or “overabundance” in recent theoretical works on variation in morphological systems—has long been recognized as a crucial dimension of the nature of language. Within transformational grammar, variation was either ignored or assigned the status of a performance deviancy with respect to the unity and stability of abstract underlying forms2. It is one of the merits of dialectology and dialectologists to have repeatedly emphasized the inherently variable and polymorphic nature of linguistic systems, which are always in a state of relative equilibrium and stability (cf. among others the many works of Jean Séguy, Jacques Allières, Jean-Louis Fossat, Jacques Boisgontier, Xavier Ravier etc., all members of what could be called the Toulouse School of Dialectology). Remember as well that in a work like Pathologie et thérapeutique verbale, Jules Gilliéron (1915) extensively investigates the way “concurrent” forms may “compete” in a given area and the way such conflicts may resolve in the long run—cf. as well his various contributions on the “mots en collision.” Additionally, the pioneering works of Mathesius emphasized as well the importance of “static oscillation,” i.e., variation in the speech properties of the individuals of a given community at a given time, variation which may be correlated or not with emphasis and expressivity (cf. Stankiewicz 1964).

The most striking features of the Occitan data that will be discussed in this contribution lie not only in the possibility of finding various forms for a given cell in certain paradigms3. Rather, it will be shown that in certain cases, two, three and sometimes even four (sub-)paradigms for one and the same verb may coexist in the same variety. The question of the ‘raison d’être’ of such complexity will be addressed as well as its resolution. Let us add that the Occitan data presented in this paper are taken from the Atlas Linguistique du Languedoc occidental (Ravier 1978–1993; henceforth ALLOc), as well as from fieldwork material gathered by Xavier Ravier, Jacques Boisgontier and Ernest Nègre with a linguistic questionnaire containing 2396 questions for each locality—the questionnaire is followed by a version of the parable of the prodigal son. These data which were initially transcribed following the “Romanists” notation, have been turned into the IPA transcription4: they offer an exceptional picture of the dialectological situation of the Occitan varieties spoken in Languedoc in the 70s. As recalled by Esher (2015, p. 251), the ALLOc provides comparable data for 132 localities across the départements Ariège (09), Aude (11), Aveyron (12), Corrèze (19), Dordogne (24), Gironde (33), Haute-Garonne (31), Lot (46), Lot-et-Garonne (47), Tarn (81), and Tarn-et-Garonne (82). Of course, the data recorded in the various atlases are not exempt from criticism and the dialectologists are perfectly aware of the methodological limitations of this kind of data. It must be pointed out, however, that the phonetic and morphological information reported in these enquêtes is rigorously noted and the enquêteurs systematically provide the reader with the free variants recorded in such or such variety. They also mention the existence of variants used; for example, before a pause in the final position, as opposed to those used before a vowel or a consonant. To take just one example, in the fieldwork realized in the dialect of Auzits (Aveyron), the enquêteur clearly indicates that such or such (coexistent) verb forms are used indifferently by the informant(s). When a possible or expected variant is suggested, one of the informants may point out that it was used by the preceding generation, that it is used in the neighbouring dialect or (s)he may exclude it completely. A given variant may as well be mentioned as more frequent than another one or prefered in a slow vs. rapid rate of speech, thus providing the enquêteur with a particularly rich set of (linguistic and metalinguistic) data5.

2. Polymorphism and Linguistic Variation

Polymorphism and variation are the natural state of linguistic systems (cf. Lope Blanch 1977, p. 597)6. Accustomed with our linguistic traditions and our historical norms, this is a truism we may be led to forget, even when we are linguists. However, linguistic systems are always in a relative state of (in)stability (cf. the notion of ‘static oscillation’ in Mathesius ([1911] 1964), if not because they are the result and the outcome of a tension between historical traditions and individual innovations. Needless to say, if polymorphism and variation are a rather banal situation among dialects, they are also a consubstantial aspect of varieties endowed with a shared norm as illustrated, for example, by the competing perfective forms apparse, apparve, apparì ‘(s)he appeared’ in Italian, which coexist without any difference from a semantic point of view (cf. Thornton 2011, 2012, etc.)—we can mention as well the pioneering work of Nencioni (1953–1954) dedicated to the coexisting third person plural Preterit forms in the history of Italian). From this point of view, it is not certain that we can follow Alvar (1966, p. 34) when he holds that polymorphism is a phenomenon typical of living speech and not of the literary standard: polymorphism is just easier to observe in living dialects than in fixed norms. Observe that Occitan itself has been and still is a literary language—cf. the Donatz Proensal or the Razós de trobar (Guessard 1858), which represents the first Romance Language grammar in Europe (cf. Bourciez 1910, p. 314, §261; Bec 1963, pp. 72–73, etc.); to quote just one example, the Occitan norm does not resort to clitic subjects; but of course subject clitics may be attested in such or such dialect.

The idea according to which multiplicity of forms should be viewed as the surface manifestation of an underlying unity was one of the credos of the transformational period of Generative grammar7. In his analysis of the French verb, Schane (1966, p. 747) takes French verb forms such as je vis, tu vis, il vit, nous vivons, vous vivez, ils vivent ‘I live, you live, he lives etc.’ to be the result of deletion/insertion process: “(…) throughout the plural the verb stem manifests itself phonetically as [viv] (with a stem-final consonant), whereas in the singular the stem is [vi] without the stem-final consonant. One could, of course, postulate two allomorphs for the morpheme ‘live’-namely |vi| and |viv|, these two allomorphs being distributed in the singular and in the plural, respectively. However, this solution is of limited interest since it merely restates the already observable facts and in no way explains why the verb stem should exhibit this difference at all. Instead, we shall represent a given morpheme by a single phonological representation, so that throughout the paradigm the same morpheme will always have the same phonological shape.” As can be seen, the very notion of allomorphy has no place in such a perspective.

In addition, Mondorf (2014, p. 209) recalls that “Traditional grammatical theory had, for a long time, relegated language variation to the realm of performance, thereby declaring it out of bounds for linguistic theory-building. And even in sociolinguistics, the cradle of linguistic variability, research on morphological or syntactic variation as opposed to phonological or lexical variation, was occasionally impaired by erroneous assumptions concerning an alleged uniformity of syntax, which was regarded as “the marker of cohesion in society” by Hudson (1980, pp. 44, 47).” Of course, the search for unity behind variation cannot be said to characterize Generative Grammar only, and we should remember the motto of Jakobson (1971, p. 588) according to which “Naive attempts to deal with variations without attacking the problem of invariants are condemned to failure.” However, observe first that the legitimate search for general principles underlying variation should not lead to some kind of reductionism whereby variation itself would lose any relevance or any raison d’être; and second, the question of ‘invariants’ should not be equated to that of ‘universals’.

Thus, we are not claiming that when possible and justified, the identification of unity behind diversity would be an erroneous task. We are just recalling that variation is not an ‘accident’ or a disturbance that would disrupt the best of the (phonetic and morphological) worlds; unity and stability should not be taken to be something else than a platonic ideal which is only met in exceptional conditions. As Haugen (1966, p. 931) points out, “The ideal case of minimal variation in form would be a hypothetical, ‘pure’ variety of a language having only one spelling and one pronunciation for every word, one word for every meaning, and one grammatical framework for all utterances.”

From this point of view, the statement of Bonami and Stump (2016, p. 469) according to which “It is natural to assume that each cell in a lexeme’s paradigm has at most one realization” may only be relevant within a given theoretical framework and on the basis of assumptions concerning language in general that are at least questionable. In French, any native speaker asked to provide the Imperative form of the verb attendre [atᾶdʁ] ‘to wait’ in the second person singular will hardly give anything else than attends! [atᾶ], and this is exactly what the normative grammars indicate. However, one only has to listen to any informal conversation to realize that this is not the only option, and in spontaneous speech it is not hard to find reduced forms such as [tᾶ]. Needless to say, this kind of reduced Imperative tends to acquire specific properties and only partially covers the same functional field as the full form (cf. Floricic and Molinu 2003, 2012 and the works cited therein). However, this is, of course, not a sufficient reason to exclude it from the set of realizations of the Imperative.

We shall show in the following lines that polymorphism and variation should not be viewed as a “fait de parole” (performance) as opposed to a “fait de langue” (competence), as argued by Alvar (1966, pp. 34–35)—we are of course aware of the fact that “fait de langue” and “competence” are not exactly the same thing. They should not be viewed either as a manifestation of the existence of ‘multiple grammars’ (cf. Guy 1997, 125sqq., who argues against the existence of ‘multiple grammars’ and against the relevance of the competence -performance distinction). Polymorphism and variation should be viewed as a systemic property of language, thus as a “fait de langue” (cf. Lope Blanch 1977; Kristol 2018, etc.). Once again, it is a commonplace experience that in many cases, several options may be equally offered to the speaker without any difference and without any particular connotation between them. However, it is also true—and this point will be illustrated in the following lines—that at some point in the historical development of a language, the “old” may coexist with the “new,” some preferences may manifest in favour of such or such option, for such or such configuration, and this state of oscillating and coexisting options may lead to the reduction or elimination of fragmentation and plurality (on variation, its nature and its fake, cf. Moravcsik 2019). As a matter of fact, we should not forget that the amount of complexity is at least limited by our cognitive and perceptive abilities, let alone our memory limitations, even though the polymorphic nature of linguistic systems shows that the human brain may accommodate with a certain amount of complexity.

Needless to say, it is not possible to discuss in this paper the general question of morphological complexity (for an overview, cf. Baerman et al. 2017). Of course, complexity may take very different forms, but it is hard to deny that multiplicity of coexistent options for one and same property or set of properties offers a major amount of complexity than the existence of a unique one. In the same way, it is hard to deny that the existence of various allomorphs for a given verb lexeme is an instance of complexity as compared to a situation in which such allomorphy is absent. Here again, we should not rely on the picture offered by the linguistic norm of standard languages to evaluate or describe the degree of complexity of its patterns. As pointed out by Paul Passy (1890, p. 56) quoting the French Celticist Henri Gaidoz (1883–1885, pp. 87–88), if French were analyzed or studied like an American or Oceanic language, we would identify conjugation forms like jlèm ‘I love him’, tulèm ‘you love him’, imèm ‘he loves me’ and we would thus conclude that French is a polysynthetic language.

2.1. Multiple Coexisting Paradigms

Recent works on “overabundance” have mainly concentrated on the existence of ‘doublets’ for the same slot or cell of a given (sub-)paradigm—and this is the (main) sense given to the term “overabundance” in recent morphological research. In other words, it has been held to be a typical example of (morphological) non-canonicity that the same cell of an inflectional paradigm may be filled by more than one form (cf. Thornton 2012, p. 183). For sure, in a (morphologically) ideal world, perfection would be achieved if, as brilliantly formulated by Meillet (1904, p. 462), “Tout le développement des formes grammaticales est dominé par un principe, qu’on ne formule pas le plus souvent dans toute sa rigueur et dans toute sa généralité, mais qui est le fondement essentiel de toutes les recherches faites dans ce domaine: Chaque fonction tend à être remplie par une forme unique, bien définie et toujours la même en toutes conditions dans une langue donnée” [“The whole development of grammatical forms is dominated by one principle, which is not usually formulated in all its rigour and generality, but which is the essential foundation of all research in this field: Each function tends to be fulfilled by by a form that is unique, well-defined, and always the same under all conditions in a given language”]. Such a one-to-one correspondence between form and content is far from being achieved in natural languages. What is more, formal variation and proliferation may be said to be the norm rather than the exception. If the kind of polymorphism such as that referred to earlier is widely attested among dialects and official languages, it seems that multiple coexisting (sub-)paradigms for the same verb are less easy to observe and much less work has been devoted to this kind of complexity offered by morphological systems. Of course, if linguists take as a base for their investigations the paradigms found in school grammars, they will hardly find the kind of variation that will be presented in the following lines (cf. Kristol 2018 for a criticism of the standpoint of those linguists who limit their investigations to data and phenomena of a given linguistic ‘norm’).

An example of such extreme polymorphism is given in (1a–c): in the dialect of Auzits (Aveyron) the Present Indicative and the Imperative of the verb dyrˈbi ‘to open’ in (1a) show that the speakers have at their disposal a double inflectional paradigm. The point to be stressed is that we are faced with pure variation: it is clear from the data gathered by Xavier Ravier that no semantic criterion may be put forward to account for the choice or preference of one paradigm over another:

(1a)
dyrˈbi ‘to open’ Auzits (12)
Present Indicative			Imperative
dyrˈbise	/	ˈdɥɛrbe
dyrˈbises	/	ˈdɥɛrbes	dyrˈbis	/	ˈdɥɛrp
dyrˈbis	/	ˈdɥɛrp		/
dyrbiˈsɛ͂n	/	dyrˈbɛ͂n		/
dyrbiˈsɛs	/	dyrˈbɛs	dyrbiˈsɛs	/	dyrˈbɛs
dyrˈbisu	/	ˈdɥɛrbu		/

(1b)				(1c)
durˈmi ‘to sleep’ Auzits				pɔrˈti ‘to go away’ Auzits
Present Indicative			Imperative.	Present Indicative			Imperative
durˈmise	/	ˈdwɔrme		ˈparte	/	pɔrˈtise
durˈmises	/	ˈdwɔrmes	ˈdwo͂n	ˈpartes	/	pɔrˈtises	ˈpar
durˈmis	/	ˈdwo͂n		ˈpar	/	pɔrˈtis
durmiˈsɛ͂n	/	durˈmɛ͂n		pɔrˈtɛ͂n	/	pɔrtiˈsɛ͂n
durmiˈsɛs	/	durˈmɛs	durˈmɛs	pɔrˈtɛs	/	pɔrtiˈsɛs	pɔrˈtɛs
durˈmisu	/	ˈdwɔrmu		ˈpartu	/	pɔrˈtisu

On the other hand, in the case of the verbs durˈmi ‘to sleep’ and pɔrˈti ‘to go away’ in (1b-c), competing (sub-)paradigms are restricted to the Present Indicative. The historical “raison d’être” of such a situation may be very difficult to identify. In the dialect of Auzits spoken in the Aveyron department (12)8, it can be argued that the emergence of competing paradigms has a purely morphological ‘raison d’être’: the propagation and success of the inchoative infix -isc- has given rise to the configuration illustrated in (1), where the same verb appears with or without its infixed variant (on the role and extension of this infix, cf. Ravier 1971). Of course, we are using here the expressions “inchoative infix” or “inchoative inflection” as does Ravier (1971), i.e., as a practical label (cf. as well Dalbera 1990; Casagrande 2012, etc.). However, it must be clear that (a) synchronically there is no aspectual value in the formations where this affix appears—should such a value have been available, it would have been mentioned by the enquêteurs, who regularly provide many metalinguistic information in their carnets d’enquêtes9; (b) it is far from being clear that we should talk about infixation. The same type of morphological object sometimes is refered to as ‘augment’ or as ‘interfix’ (cf. Floricic 2018, from which some of these data are borrowed, even though (Floricic 2018) is mainly concerned with Imperatives)10.

Let us add that the 2nd sg. Imperative has the same shape as the 3rd person sg., a feature which is common to Occitan and Ibero-Romance. Thus the 2nd sg. Imperative of the verb dyrˈbi ‘to open’ is either dyrˈbis or ˈdɥɛr^p ‘open!’—the unvoiced bilabial in final position is due to devoicing in that position. The (sub-)paradigms of the verb durˈmi ‘to sleep’ and pɔrˈti ‘to go away’ in (1b–c) show only one variant in the Imperative—the shortest one (cf. Companys 1964, p. 55). And such a property is in keeping with what we know about the Imperative: Imperatives are commonly expressed by the bare root or stem of a verb (cf. Maiden 2007; Aikhenvald 2010; Floricic 2011, 2021b, etc.). By the way the poor or reduced inflectional marking of the Imperative may be correlated with another fundamental property of the Imperatives: their (pragmatic) efficiency is linked with tone and they tend to maximize the beginning of the word (cf. Jespersen 1949, pp. 26–27), who rightly stresses the importance of the beginning of the word in word recognition, and the opposite value of the end of the word, hence the various deletion or reduction phenomena found in this position). In languages with rather developed focus systems, the Imperative may suspend any (focus) marking onto the other elements of the sentence, hence the conclusion that the Imperative is itself on focus (cf. Floricic 2000, 2009, 2020 and the references reported therein). No wonder, then, that this category may privilege short forms over longer ones. This is exactly what is illustrated in (1b–c), where the forms ˈdwo͂n ‘sleep!’ and ˈpar ‘go away!’ are preferred over durˈmis and pɔrˈtis.

2.2. Inchoative Inflexion and Polymorphism

A particularly complex situation holds in the dialect of Dun (Ariège) in (2), where in the Present Indicative the verb ˈdɛrβe/durˈβi ‘to open’ shows up to three coexistent paradigms:

(2)
Dun (09)
ˈdɛrβe/durˈβi ‘to open’				parˈti ‘to go away’			ˈkurre ‘to run’
Present Indicative			Imper.	Present Indicative		Imper.	Present Indicative		Imper.
ˈdɛrβi	durˈβisi	derˈβisi		ˈparti	parˈtisi		ˈkurri	kurˈrisi
ˈdɛrβes	durˈβises	derˈβises	ˈdɛr ˈdɛrp	ˈpartes	parˈtises	ˈpart	ˈkurres	kurˈrises	ˈkurr
ˈdɛrp	durˈβis	derˈβis		ˈpart	parˈtis		ˈkur kurr	kurˈris
derˈβɛ͂n	durβiˈsɛ͂n	derβiˈsɛ͂n		parˈtɛ͂n	partiˈsɛ͂n		kurˈrɛ͂n	kuriˈsɛ͂n
derˈβɛts	durβiˈsɛts	derβiˈsɛts	derβiˈsɛts	parˈtɛts	partiˈsɛts	parˈtɛts	kurˈrɛts	kuriˈsɛts	kurˈrɛts
ˈdɛrβen	durˈβise͂n	derˈβise͂n		ˈparte͂n	parˈtise͂n		ˈkurre͂n	kuˈrise͂n

The particular case of the verb ˈdɛrβe/durˈβi ‘to open’ could be due to both a morphological and a phonetic impulse. The quality of the pretonic vowel may reflect either the outcome of the preposition de of the complex form deoperire, or the initial vowel of the (simple) verb form, which surfaces as [u] when reduced. As pointed out by one of the reviewers, however, the stem ˈdɛrβ- may as well be the outcome of ˈdɥɛrβ-, with monophtongization of [ɥɛ]. And here too, the inchoative infix -isc- has given rise to the variants in -is- which extended as well to the verbs parˈti ‘to go away’ and ˈkurre ‘to run’. The data of this dialect seem to confirm what has been said about the Imperative: the short form is preferred over the longer.

Observe that in some dialects, the success of the inchoative infix -isc-/-esc- has been such as to impose “double” inchoative marking in the same verb. This is the case for example in the dialect of Meljac (12) with verbs like ˈkrei̯se ‘to grow up’ (cf. (3a)), where of course this “double” marking corresponds to two different strata: the second stratum applies to a form where the inchoative is no longer transparent. The paradigm of the verb ˈɛse ‘to be’ in the same dialect shows that this infix extends to the Present Subjunctive, hence a triple paradigm for the same verb11:

(3a)		(3b)
ˈkrei̯se ‘to grow up’ Meljac (12)		ˈɛse ‘to be’ Meljac (12)
Present Indicative		Subjunctive
ˈkrei̯si	krei̯ˈsisi	ke ˈsao	ke ˈsjao	ke ˈsjasko
ˈkrei̯ses	krei̯ˈsises	ke ˈsaos	ke ˈsjaos	ke ˈsjaskos
ˈkrei̯s	krei̯ˈsis	ke ˈsao	ke ˈsjao	ke ˈsjasko
krei̯ˈsɛ͂n	krei̯siˈsɛ͂n	ke ˈsjẽn	ke sjaˈẽn	ke sjasˈkẽn
krei̯ˈsɛs	krei̯siˈsɛs	ke ˈsjes	ke sjaˈes	ke sjasˈkes
ˈkrei̯su	krei̯ˈsisu	ke ˈsau	ke ˈsjau	ke ˈsjasku

From (sub-)paradigms like those in (1)–(2) and (3a), one may want to infer that the present indicative is particularly subject to polymorphism. However, as shown by the paradigm of the verb ˈɛse ‘to be’ in the present subjunctive in (3b), such a conclusion would be mistaken. Nor could we hold that such a proliferation is limited to auxiliaries (cf. the case of the verb ˈdɛrβe/durˈβi ‘to open’ in the dialect of Dun in (2)).

Interestingly, the Imperfect Subjunctive and the Preterit of the verb ˈfai̯de ‘to do’ may show the same kind of paradigmatic complexity, as illustrated in (4) with data from the dialect of Onet-l’Eglise (12): here too the inchoative infix -isc/-esc- has extended to the Present and Imperfect Subjunctive and in the Preterit, hence the coexisting paradigms found in the last two “tiroirs” (cf. Buckenmaier 1934, pp. 84–85, §§292–294; Ronjat 1937, p. 282, §635, etc.):

(4)
ˈfai̯de ‘to do’ (Onet-l’Eglise (12))
Pres. Ind.	Imper.	Pres. Subj.	Imperf. Subj.		Preterit
ˈfau̯		ke ˈfaske	ke fɔsˈkɛse	ˈfɛse	ˈfɛre	fɔsˈkɛre
ˈfas	ˈfai̯	ke ˈfaskos	ke fɔsˈkɛsos	ˈfɛsos	ˈfɛros	fɔsˈkɛros
ˈfɔ		ke ˈfasko	ke fɔsˈkɛso	ˈfɛso	ˈfɛt	fɔsˈkɛt
fɔˈzɛ͂n		ke fɔsˈkẽn	ke fɔsˈkɛsẽn	ˈfɛsẽn	ˈfɛrẽn	fɔsˈkɛrẽn
fɔˈzɛs	fɔˈzɛs	ke fɔsˈkes	ke fɔsˈkɛses	ˈfɛses	ˈfɛres	fɔs’kɛres
ˈfɔu̯		ke ˈfasku	ke fɔsˈkɛsu	ˈfɛsu	ˈfɛru	fɔsˈkɛru

Let’s add that the Imperfect Subjunctive and Preterit stem fɔsˈk- is found as well in the Gerund (cf. ẽ fɔsˈkẽn ‘doing’) and that the same stem occurs with verbs like ˈpui̯re ‘to be able to’ (cf. in the Preterit 1. pusˈkɛre/puˈɛre, etc. and in the Imperfect Subjunctive 3. ke pusˈkɛso/puˈɛso, 3. ke pusˈkɛt/puˈɛt, etc.). Needless to say, the -ɛr- Preterit forms are the outcome of the -ērunt/-ĕrunt Latin Perfects12, while the -ɛs- Imperfect Subjunctive forms are the outcome of Latin Pluperfect Subjunctive (cf. fēcissem, fēcissēs, fēcisset, etc.).

It is very difficult to predict which verbs and which paradigms may be affected by the extension of the inchoative infix. Buckenmaier (1934, p. 84) points out that a verb like beˈni ‘to come’ does not show any inchoative variant, and the same observation holds for many other dialects: this verb resists any contamination of the inchoative inflexion. It could be argued that (token or/and type) frequency is a crucial property of this kind of verb. Highly frequent verbs often are irregular from a phonetic point of view, and their cognitive salience sets them apart in many respects: they are learned rather than reproduced by rules, they escape phonetic laws, and they resist analogical change (cf. on frequency the many works of Witold Mańczak and his rehabilitation in the works of Joan Bybee, particularly (Bybee 2007)). On the other hand, the same principle should prevent other frequent verbs from falling under the attraction of the inchoative type: the verbs ˈfa ‘to do’ and ˈɛse ‘to be’ are such verbs. Nonetheless, the data illustrated in (3) and (4) show that the inchoative infix may extend to these verbs too.

In the dialect of Les Lèves-et-Thoumayragues (33) for example, the inchoative inflexion has penetrated into the verb ˈɛstrə ‘to be’ (Preterit, Present and Imperfect Subjunctive), the Present Subjunctive of the verb ou̯ˈɣer ‘to have’ (1. ˈk aski), the Preterit (1. fasˈkɛri/faˈɣɛri), the Present (1. ke ˈfaski) and Imperfect Subjunctive (1. ke fasˈkɛsi/faˈɣɛsi) of the verb ˈfɑ/ˈfai̯rə ‘to do’, and even the Present Subjunctive of the verb ˈdi ‘to say’ (cf. the variants ke i ˈdiski/ˈdiɣi ‘that I say’):

(5)
ˈɛstrə ‘to be’ Les Lèves-et-Thoumayragues (33)
Preterit				Pres. Subj.		Imperf. Subj.
siˈɣɛri	fyˈɣɛri	sisˈkɛri	fysˈkɛri	ke ˈsjɛski		ke sisˈkɛsi
ty siˈɣɛrə	ty fyˈɣɛrə	ty sisˈkɛrə	ty fysˈkɛrə	ke ty ˈsjɛskə		ke ty sisˈkɛsə
siˈɣe	fyˈɣe	sisˈke	fysˈke	ke ˈsjɛskə		ke sisˈke
siˈɣɛrə͂	fyˈɣɛrə͂	sisˈkɛrə͂	fysˈkɛrə͂	ke siˈɣjẽŋ	ke sisˈkɛ͂ŋ	ke sisˈkɛsə͂
vuzau̯ siˈɣɛrə	vuzau̯ fyˈɣɛrə	vuzau̯ sisˈkɛrə	vuzau̯ fysˈkɛrə	ke vuzau̯ sisˈkjɛ		ke vuzau̯ sisˈkɛsə
siˈɣɛrə͂	fyˈɣɛrə͂	sisˈkɛrə͂	fysˈkɛrə͂	ke ˈsjɛskə͂		ke sisˈkɛsə͂

In the dialect of Les Lèves-et-Thoumayragues (Aquitaine), the connections between the Present Perfect and the Present/Imperfect Subjunctive of the verb ˈɛstrə ‘to be’ can be seen in the selection of the stem used in the various tenses/moods: in the Present Perfect, the stem si-/sj- (Allières 1971, p. 142) and the stem fy-, to which the Perfect exponent -er- (<-ērunt/-ĕrunt) can be attached and may be lengthened with the infix -isc-/-esc- (cf. 1st sg. forms si-sˈk-ɛr-i, fy-sˈk-ɛr-i). Interestingly, however, in the Imperative only one of the two variants is reported: ˈdiɣə ‘say! (sg.)’/ˈdiɣei̯ ‘say! (pl.)’: no corresponding inchoative stem has been mentioned. Notice that the Imperative is known for showing (phonetically) aberrant patterns and for its resistance to analogical change.

Another striking case of extreme polymorphism can be observed in the paradigms of the verb aˈna ‘to go’ again in the dialect of Dun in the Ariège department:

(6)
aˈna ‘to go’ (Dun (09))
Pres. Ind.	Imper.	Pres. Subj.	Imperf. Subj.				Preterit
ˈbau̯		ˈãŋɡe	ãŋˈɡɛse	ãŋˈɡɛso	aˈnɛso		ãŋˈɡɛɡi	ãŋˈɡɛri
ˈbas	ˈbas	ˈãŋɡes	ãŋˈɡɛsos	ãŋˈɡɛses	aˈnɛses	aˈnɛsos	ãŋˈɡɛɡes	ãŋˈɡɛres
ˈba		ˈãŋɡe	ãŋˈɡɛso	ãŋˈɡɛse	aˈnɛse	aˈnɛso	ãŋˈɡɛk
aˈnãn		ãŋˈɡẽn	ãŋˈɡɛsẽn	ãŋˈɡɛsɔ͂n	aˈnɛsɔ͂n	aˈnɛsẽn	ãŋˈɡɛɣẽn	ãŋˈɡɛrẽn
aˈnats	aˈnats	ãŋˈɡets	ãŋˈɡɛsɔts	ãŋˈɡɛsets	aˈnɛsɔts	aˈnɛsets	ãŋˈɡɛɣets	ãŋˈɡɛrets
ˈbãn		ˈãŋɡẽn	ãŋˈɡɛsɔ͂n	ãŋˈɡɛsẽn	aˈnɛsɔ͂n	aˈnɛsẽn	ãŋˈɡɛɣẽn	ãŋˈɡɛrẽn

The extreme polymorphism found in the verb morphology of this dialect (up to 4 coexistent (sub-)paradigms in the Imperfect Subjunctive) is not easy to explain.

As far as the Present Indicative is concerned, no particular variation can be mentioned. The unique aspect to be pointed out is the correlation between the 2nd sg. of the Present Indicative and the same cell of the Imperative: as we already observed, the general pattern in Occitan is to have the 2nd sg. (Imperative) alongside the 3rd sg. (Present Indicative) (cf. Floricic 2018). From this point of view, the relation between Imperative and Present Indicative in this dialect is reminiscent of that of Italian, where the Imperative of monosyllabic verbs is taken from the 2nd sg. Present Indicative and does not continue the corresponding Latin Imperative (cf. (7)):

(7)
ˈdare ‘to give’		ˈfare ‘to do’			sˈtare ‘to stay’		anˈdare ‘to go’
Pres. Ind.	Imper.	Pres. Ind.		Imper.	Pres. Ind.	Imper.	Pres. Ind.		Imper.
ˈdɔ		ˈfatʧo	ˈfɔ		sˈtɔ		ˈvado	ˈvɔ
ˈdai̯	ˈdai̯ ˈda	ˈfai̯		ˈfai̯ ˈfa	sˈtai̯	sˈtai̯ sˈta	ˈvai̯		ˈvai̯ ˈva
ˈda		ˈfa			sˈta		ˈva
ˈdjamo	ˈdjamo	fatˈʧamo		fatˈʧamo	sˈtjamo	sˈtjamo	anˈdjamo		anˈdjamo
ˈdate	ˈdate	ˈfate		ˈfate	sˈtate	sˈtate	anˈdate		anˈdate
ˈdanno		ˈfanno			sˈtanno		ˈvanno

In other words, while in Italian the Imperative 2nd singular of the 1st conjugation in -are coincide with the 3rd person singular of the Present Indicative, the monosyllabic verbs in -are have their Imperative identical to the 2nd person singular of the Present Indicative (cf. Floricic and Molinu 2003, 2012; Floricic 2007, 2018, etc.)13.

However, the most interesting aspect of the verb (sub-)paradigms in (6) is the extension/generalization of the velar segment, which has again as its starting point the Latin Perfect forms (cf. Bourciez 1927). From the Preterit, it extends to the Subjunctive, as shown by forms such as the Present Subjunctive ˈãŋɡe and the Imperfect Subjunctive ãŋˈɡɛse/ãŋˈɡɛso. As a matter of fact, the velar stop of the stem ãŋˈɡ- can be said to go back to the Latin Perfectum: it is a well-known fact that the Preterit of various verbs played a crucial role in Occitan verb morphology and that it spread in many sectors of the verbal system (cf. Chabaneau 1876, p. 261). However synchronically, is it not clear to which extent we can hold that we are dealing with a ‘Preterit stem’ obtained adding the augment/-ɡ/to the stem an-.

The phonetic process at stake had already been explained by Millardet (1910, pp. 185–86). And if the forms with a velar consonant have had such a success in Occitan morphology, it is also due to the frequency and saliency of the Preterit of such verbs as ac <habuit (cf. Camproux 1962, p. 436), as has been repeatedly noticed (cf. infra)14.

From this point of view, the stem ãŋˈɡ- is not stranger than the stem beŋˈɡ- of the verb ˈbeŋɡe ‘to come’ (cf. Ronjat 1937, pp. 295–96, §645), and Allières (1971, pp. 146–47) points out the close link between the two verbs (for a discussion, cf. Floricic 2019, 2021a). The point to be stressed is that in the dialect of Dun, the spread of the velar segment in the verb paradigms of aˈna ‘to go’ has not yet eliminated the non-velar stems, hence the coexistence of both in the Imperfect Subjunctive. As for the polymorphism found in the Preterit (cf. the 1st person singular forms ãŋˈɡɛɡi/ãŋˈɡɛri ‘I went’), it is due—as has been already observed—to the evolution of the Latin Perfect forms in -ui: the velar stop arises as a strengthening process of the velar glide which devoices in final position in the zero 3rd person singular form (cf. Schulz-Gora 1906, pp. 96–97). And this form acts as a pivot on which the whole paradigm may be reshaped (cf. Bybee and Brewer 1980; Floricic 2021a and especially Bourciez 1927, pp. 116, 122–23): the form 3. ãŋˈɡɛk ‘(s)he went’, modelled on 3. bẽŋˈɡɛk ‘(s)he came’ and reanalyzed as a zero stem, takes on the various personal endings15. In other words, the basicness of the Preterit 3rd person singular can be seen in the fact that it has served as a base on which the first person singular has been reshaped; hence 1. ãŋˈɡɛɡi, 2. ãŋˈɡɛɡes, etc., where the second velar segment is of course secondary (cf. Sicre (1907–1908, p. 394) for similar data in the dialect of Foix in the Ariège area).

Interestingly, the variants of the ãŋˈɡɛri type are derived as well from the Perfect forms. However, in this case the starting point of the process is not the 3rd person singular form but the 3rd person plural one. As a matter of fact, the stem found in 1. ãŋˈɡɛri, 2. ãŋˈɡɛres, etc. can only be explained taking as base the forms in -ērunt/-ĕrunt, and this feature further illustrates the crucial role of the 3rd person forms (cf. Bourciez 1927, p. 114). As for the stem ˈãŋɡ- itself, it may be due to the analogy of the stem ˈbẽŋɡ- of the verb beˈni ‘to come’, as pointed out earlier (cf. Floricic 2021a, 2022).

As for the origin of the stem ãŋˈɡes-, it can be said to continue the Latin Subjunctive Pluperfect of the ambulāvissĕm type. And the vowel following the velar segment (cf. ãŋˈɡ-ɛ-so) can be held to come from the Preterit forms (cf. ãŋˈɡ-ɛ-k, etc.). The final vowel of the ãŋˈɡɛso variant may be said to have spread from the Present Indicative where it is the regular outcome of -a in unstressed syllable (it is attested as well thorough the paradigm of the Imperfect Indicative: aˈnaβo, aˈnaβos, aˈnaβo, aˈnaβɔ͂n, anaˈβɔts, aˈnaβɔ͂n). It must be emphasized that the Latin singular endings of the Pluperfect Subjunctive are only maintained in the second person: the 1st and 3rd person endings -m and -t are lost, hence the syncretism found in these cells (cf. ãŋˈɡɛso/ãŋˈɡɛse/aˈnɛso). Of course, this particular vocalism has spread from the singular to the plural, hence forms like ãŋˈɡɛsɔts/aˈnɛsɔts.

Another syncretism can be found, i.e., that between the cell 4 (1st plural) and the cell 6 (3rd plural). A syncretism which is probably due—in this case too—to the fate of the final nasal consonant of the Pluperfect Subjunctive forms (cf. ambulā-v-is-sē-mus, ambulā-v-is-se-nt): once the final consonant of the ambulā-v-is-se-nt type is lost, we are left with a final nasal which nasalizes the preceding vowel, hence ãŋˈɡɛsẽn/aˈnɛsẽn. In the same way, the shape of the 1st person plural can be accounted for holding that the -us final sequence has been lost (cf. in Romanian dormim from dormīmŭs), hence the nasalization of the post-tonic vowel.

2.3. Phonetic and Lexical Conditioning behind Extreme Polymorphism

Phonetic polymorphism probably is the easiest kind of polymorphism to observe in everyday speech, and this is probably the type of polymorphism that linguists have most readily recognized in earlier linguistic research (cf. Sturtevant 1917, p. 121; the question of free variants was discussed among others by Troubetzkoy). By the way, Allières takes phonetic polymorphism to be the ‘purest’ kind of polymorphism and the most palpable one, as widely illustrated by the fate of the plural definite article derived from illos/illas (cf. Straka 1979, 443sqq. on the phonetic scenario underlying such a polymorphism).

As a further example of pure variation within coexistent paradigms, we shall mention the case of the Conditional forms in the dialect of Clermont-le-Fort (31) in (8):

(8)
Clermont-le-Fort (31)
ˈɛstre ‘to be’			basˈti ‘to build’			ˈfa ‘to do’
Conditional		Future	Conditional		Future	Conditional		Future
siˈɔi̯	siˈrɔi̯	siˈrɛi̯	bastiˈjɔi̯	bastiˈrɔi̯	bastiˈrɛ	faˈjɔi̯	faˈrɔi̯	faˈrɛ
siˈɔs	siˈrɔs	siˈras	bastiˈjɔs	bastiˈrɔs	bastiˈras	faˈjɔs	faˈrɔs	faˈras
siˈo	siˈro	siˈra	bastiˈjɔ	bastiˈrɔ	bastiˈra	faˈjɔ	faˈrɔ	faˈra
siˈɔ͂n	siˈrɔ͂n	siˈrẽn	bastiˈjɔ͂n	bastiˈrɔ͂n	bastiˈrẽn	faˈjɔ͂n	faˈrɔ͂n	faˈrẽn
siˈɔts	siˈrɔts	siˈrets	bastiˈjɔts	bastiˈrɔts	bastiˈrets	faˈjɔts	faˈrɔts	faˈrets
siˈɔ͂n	siˈrɔ͂n	siˈrãn	bastiˈjɔ͂n	bastiˈrɔ͂n	bastiˈrãn	faˈjɔ͂n	faˈrɔ͂n	faˈrãn

The coexistent paradigms in (8) illustrate the idea put forth by Jacques Allières (1954) that, in many cases, polymorphism has a phonetic grounding. As is well known, the Romance Conditional continues a periphrastic construction made of the Infinitive of the lexical verb plus the Imperfect Indicative habēbam of the auxiliary habēre (cf. cantāre habēbam > cantarea > cantaría ‘(s)he would sing’ (Sp.))16. In the formation of Conditional in modern Italian, the auxiliary habēre is not used in the Imperfect Indicative but in the Preterit, hence cantāre habuī > canterei—the cantāre habēbam type is widely attested in Old Italian and in various dialects and dialectal areas in Italy (cf. Rohlfs 1968, pp. 339–42, §§593–594). The point to be stressed is that the same ending as that used in (7) appears in the Imperfect Indicative (cf. bastiˈsjɔi̯ ‘I built’, bastiˈsjɔs, bastiˈsjɔ, etc.; faˈȝɔi̯ ‘I did’, faˈȝɔs, faˈȝɔ, etc.)17. From this point of view, it can be said that the Conditional and the Future are built upon the same stem, except that the -r- exponent of the Future may simply be in complementary distribution with zero or a glide in the Conditional. It follows that a simple analysis would be to assume that one of the variants of the Conditional (co-existent) forms is obtained deleting the -r- morph of the Future/Conditional. In intervocalic contexts, this situation leads to a hiatus which would be solved inserting an anti-hiatic glide, hence forms like bastiˈjɔi̯, faˈjɔi̯, etc. However, this analysis does not hold in other dialects, where the glide cannot be said to have an anti-hiatic function. In the dialect of Meljac (12) for example, the 1st person singular Conditional form durmiˈrjɔ ‘I would sleep’ (as opposed to the Future form durmiˈrɔ ‘I will sleep’) cannot be said to owe its glide to any anti-hiatic constraint. What the dialect of Meljac shows is that -ˈrjɔ must be taken to be the primitive outcome of the Conditional: from there, a simplification process must have taken place, which reduces the -ˈrjɔ sequence to -ˈjɔ, hence the variants bastiˈrɔi̯ and bastiˈjɔi̯ from *bastiˈrjɔi̯ (cf. as well Esher 2015, p. 253)—see as well the case of arĕa > arja > aja in Tuscan. Mutatis mutandis, the same kind of analysis can account for the Conditional form faˈrɔi̯ and the Imperfect Indicative form faˈȝɔi̯ ‘I did’: faˈȝɔi̯ can be taken to presuppose a form like faˈȝjɔi̯ whose -ȝj- may have secondarily simplified in -ȝ- (cf. in Tuscan the evolution of gĕlu > ʤjɛlo > ʤɛlo). The data of the aveyronnais dialect of Sonnac in (9) confirm our analysis: as can be seen, Conditional forms in -ˈrjɔ and -ˈrɔ coexist within the same (sub-)paradigms:

(9)
Sonnac (12)
kãnˈta ‘to sing’			ˈbẽndre ‘to sell’			ˈfa ‘to do’
Conditional		Future	Conditional		Future	Conditional		Future
kãnta ˈrɔ	kãnta ˈrjo	kãntaˈrɛ	bẽnˈdrɔ	bẽnˈdrjɔ	bẽnˈdrɛ	faˈrjɔ	faˈrɔ	faˈrɛ
kãntaˈrɔs	kãntaˈrjɔs	kãntaˈras	bẽnˈdrɔs	bẽnˈdrjɔs	bẽnˈdras	faˈrjɔs	faˈrɔs	faˈras
kãntaˈrɔ	kãntaˈrjo	kãntaˈra	bẽnˈdrɔ	bẽnˈdrjo	bẽnˈdra	faˈrjɔ	faˈrɔ	faˈra
kãntaˈrɔ͂n	kãntaˈrjɔ͂n	kãntaˈrẽn	bẽnˈdrjɔ͂n	bẽnˈdrɔ͂n	bẽnˈdrẽn	faˈrjɔ͂n	faˈrɔ͂n	faˈrẽn
kãntaˈrɔts	kãntaˈrjɔtʃ	kãntaˈrets	bẽnˈdrjɔtʃ	bẽnˈdrɔts	bẽnˈdrets	faˈrjɔtʃ	faˈrɔtʃ	faˈrets
kãntaˈrɔ͂n	kãntaˈrjɔ͂n	kãntaˈrãn	bẽnˈdrjɔ͂n	bẽnˈdrɔ͂n	bẽnˈdrãn	faˈrjɔ͂n	faˈrɔ͂n	faˈrãn

Interestingly, in some dialects like that of Gaillac (81), within the same (sub-)paradigm variants in -ˈrjɔ and in -ˈjɔ can be found, as illustrated in (10), thus confirming our analysis:

(10)
Gaillac (81)
basˈti ‘to build’			aˈna ‘to go’			ˈfa ‘to do’
Conditional		Future	Conditional		Future	Conditional		Future
bastiˈjɔ	basti ˈrjo	bastiˈrɛi̯	aniˈrjɔ		aniˈrɛi̯	faˈrjɔ		faˈrɛi̯
bastiˈjɔs		bastiˈras	aniˈrjɔs		aniˈras	faˈrjɔs	faˈjos	faˈras
bastiˈjɔ		bastiˈra	aniˈjɔ		aniˈra	faˈrjɔ		faˈra
bastiˈjãn		bastiˈrẽn	aniˈrjãn		aniˈrẽn	faˈrjãn	faˈjãn	faˈrẽn
bastiˈjas		bastiˈres	aniˈrjas		aniˈres	faˈrjas		faˈres
bastiˈjɔu̯		bastiˈrɔu̯	aniˈjɔu̯	aniˈrjɔu̯	aniˈrɔu̯	faˈrjɔu̯	faˈjɔu̯	faˈrɔu̯

To sum up: the sequence -ˈrjɔ can be simplified by either eliminating the glide (hence -ˈrɔ), or by eliminating the vibrant (hence -ˈjɔ). In some cases, the reduction of the phonetic string of the Conditional may lead to a situation that could be said to be in some sense anti-iconic: in the dialect of Saint-Martin Lalande (11), for example, the verb durˈmi ‘to sleep’ has as Future form durmiˈra (cf. the Present Indicative forms durˈmisi, durˈmises, durˈmis, durmiˈsɛ͂n, durmiˈsɛts, durˈmisũn). On the other hand, the Conditional form is durmiˈo ‘I would sleep’: the amount of morphological complexity thus is less high in the Conditional than in the Present Indicative, where the verb forms are augmented via the aforementioned element -isc-. In any case, the pattern that shows up in the dialect of Saint-Martin Lalande offers an illustration of the formal asymmetry between the Future and the Conditional (cf. Esher 2015 for an indepth analysis).

Another instance of (phonetic) polymorphism may be illustrated in the paradigms of the Future of the verb aˈna ‘to go’ in several dialects. This verb has given rise to a large amount of discussions and hypotheses concerning its etymology and its evolution from Latin. It is of course not the place to re-open this old debate. As exemplified in (11) with data from the dialects of Sonnac and Saint-Matré, the nasal consonant of this form may delete intervocalically, hence the double paradigm found in the dialect of Sonnac. Notice that in the dialect of Saint-Matré, the reduced allomorph is limited to certain cells of the paradigm (1, 4, 5), thus showing that in this dialect the phenomenon at stake is just incipient:

(11)
Saint-Matre (46)		Sonnac (11)
Future		Future		Conditional
aniˈrɛ	aiˈrɛ	aniˈrɛ	aiˈrɛ	aniˈrɔ	aiˈro
aniˈras		aniˈras	aiˈras	aniˈrɔs	aiˈrɔs
aniˈra		aniˈra	aiˈra	aniˈro	aiˈro
aniˈrẽn	aiˈrẽn	aniˈrẽn	aiˈrẽn	aniˈrɔ͂n	aiˈrɔ͂n
aniˈres	aiˈres	aniˈrets	aiˈrets	aniˈrɔts	aiˈrɔts
aniˈrãn		aniˈrãn	aiˈrãn	aniˈrɔ͂n	aiˈrɔ͂n

It is probable that such kind of variation arises in contexts of tempo allegro, with effects that are all the more important that this verb belongs to the most frequent ones. However, once the two kinds of variants—a short one and a long one—have come to existence, they may extend to the whole paradigm and leave their ‘own life’.

A slightly different case to be mentioned in this respect is the case of the verb a’na ‘to go’ in the dialect of Caychax (09).

(12)
aˈna ‘to go’ (Caychax (09))
Pres. Ind.	Pres. Subj.	Future		Conditional
ˈbau̯	ˈbaŋgo	aniˈre	biˈre	aniˈrɔ	biˈrɔ
ˈbas	ˈbaŋgos	aniˈras	biˈras	aniˈrɔs	biˈrɔs
ˈba	ˈbaŋgo	aniˈra	biˈra	aniˈrɔ	biˈrɔ
aˈnãn	aˈnẽn	aniˈrẽn	biˈrẽn	aniˈrɔ͂n	biˈrɔ͂n
aˈnats	aˈnets	aniˈrets	biˈrets	aniˈrɔts	biˈrɔts
ˈbãn	ˈbaŋgũn	aniˈrãn	biˈrãn	aniˈrɔ͂n	biˈrɔ͂n

As shown in (12), the paradigms of the Present Indicative and of the Present Subjunctive show suppletive stems: ba- in the cells 1, 2, 3, 6 and ana- in the cells 4 and 5: the first one is the outcome of Lat. vadĕre, while the second could derive from ambulāre18. As recalled by Mańczak (2001, p. 23), since the XVIth Century at least 60 etymologies have been proposed to account for the particular shape of this verb in the Romance Languages. The idea according to which the unexpected phonetic evolution of this verb would be due to its high frequency, is taken to go back at least to Friedrich August Pott (1852, p. 315). As for the stem of the Present Subjunctive ˈbaŋɡo, its analysis is more complex but it can be argued to be modelled on the (Present and Imperfect) Subjunctive of the verbs beˈni ‘to come’/teˈni ‘to hold’ (cf. Massourre 2005, p. 266; 2012, p. 251): ˈbẽŋɡu and bẽŋˈɡɛs/ˈtẽŋɡo and tẽŋˈɡɛs, the Subjunctive stem being borrowed from the Preterit (cf. bẽŋˈɡɛɡi ‘I came’, tẽŋˈɡɛi/tẽŋˈɡɛɡi ‘I held’, etc.).

A rather similar situation holds in the Italo-Romance dialect of Sonnino (LT, point 682), where according to the AIS data, the 1st person singular of the Present Indicative ˈvaŋɡo ‘I go’ is clearly built upon ˈvɛŋɡo ‘I come’ (cf. Jaberg and Jud 1928, 1928–1940). Crocioni (1907, p. 63) mentions the same type of extension in the dialect of Sezze (LT): ˈvaŋɡo ‘I go’, sˈtɔŋɡo ‘I stay’, ˈtɔŋɡo ‘I give’, ˈsɔŋɡo ‘I am’ (cf. as well Markun 1932, p. 330; Rohlfs 1968, p. 260, §535; Floricic 2019 etc.). The morphological shape of the Future and Conditional forms is still more difficult to account for. A possible analysis is to assume that forms like aniˈre/aniˈrɔ and biˈre/biˈrɔ are the result of a blending process: the pretonic vowel of the aniˈre/aniˈrɔ can only be explained assuming some kind of lexical blend between ambulāre and īre (cf. Anglade 1921, p. 278; Bourciez 1927, p. 345, §297, 2° b), etc.), and the same observation holds for the forms biˈre/biˈrɔ, which can be explained assuming some kind of lexical blend between vādĕre and īre. Our interpretation seems to be confirmed by the data of the dialect of Les Lèves-et-Thoumayragues (33), as illustrated in (13):

(13)
ˈna ‘to go’ Les Lèves-et-Thoumayragues (33)
Present Indicative	Imperfect Indicative	Preterit	Future Indicative
ˈvɔu̯	ˈnavi	nãŋˈɡɛri	iˈrɛi̯	niˈrɛi̯
ty ˈva	ty ˈnavə	ty nãŋˈɡɛrə	ty iˈrɑ	ty niˈrɑ
ˈvai̯	ˈnavə	nãŋˈɡɛ	iˈrɑ	niˈrɑ
aˈnẽŋ	ˈnavə͂	nãŋˈɡɛrə͂	iˈrẽŋ	niˈrẽŋ
vuzau̯ aˈnɑ	vuzau̯ ˈnavə	vuzau̯ nãŋˈɡɛrə	vuzau̯ iˈre	vuzau̯ ˈnire
ˈvãŋ	ˈnavə͂	nãŋˈɡɛrə͂	iˈrãŋ	niˈrãŋ

As can be seen, the stem na- is obtained by aphaeresis of the verb aˈna and appears in the whole (sub-)paradigm of the Imperfect Indicative. The same stem occurs in the Preterit, where the nasal consonant preceding the velar stop is due to the analogical effect of the forms veŋˈɡɛri ‘I went’ and teŋˈɡɛri ‘I held’, on which nãŋˈɡɛri has been modelled on account of the semantic/formal link between them. The main point is that the future indicative is polymorphic, and the coexisting (sub-)paradigms only differ in the presence of the initial nasal consonant due to the fusion of (a)na and ire.

A last kind of pure polymorphism leading to the (co)existence of multiple paradigms may be identified and may be taken to be purely lexical: this type of polymorphism can be illustrated with data from the dialect of Sainte-Alauzie (46) in (14): the variants s asjeˈta/s aseˈta ‘to sit down’ have been taken to derive from seditare/asseditare < sĕditus/assĕdita (cf. Meunier 1927, p. 317), while the form se ˈsɛi̯re is derived from sĕdēre (>ser ‘to be’ in Spanish) (cf. Rohlfs 1970, p. 129, §442).

(14)
s asjeˈta, s aseˈta/se ˈsɛi̯re ‘to sit down’ (Sainte-Alauzie (46))
Pres. Ind.		Imper.
m aˈsɛti	me ˈsɛzi
t aˈsɛtos	te ˈsɛzes	aˈsjɛto ˈte	ˈsɛi̯te
s aˈsɛto	se ˈsɛi̯
nuz aseˈtãn	nui̯ seˈzɛ͂n
buz aseˈtas	βui seˈzɛs	asjeta ˈβus		sezɛ’βus
s aˈsɛtu	se ˈsɛzu

As can be seen in table (14), the shorter variant has an initial CV syllable (vs. the onsetless variants aˈsɛti, aˈsɛtos, etc.) which makes it more optimal from a phonological point of view. Interestingly, it can be observed that in various dialects, the verb s asjeˈta, s aseˈta ‘to sit down’ may show two coexisting paradigms, one with an initial vowel and one with an intial consonant:

(15)
se sjeˈta, s asjeˈta ‘to sit down’ (Labastide de Lordat (09))
Present Indicative		Imperative
m aˈsjɛti	me ˈsjɛti
t aˈsjɛtos	te ˈsjɛtos	ˈsjɛto te	aˈsjɛto te
s aˈsjɛto	se ˈsjɛto
nuz asjeˈtãn	nui̯ sjeˈtãn
buz asjeˈtats	bui̯ sjeˈtats	sjeˈtabbus		assjeˈtabbus
s aˈsjɛtɔ͂n	se ˈsjɛtɔ͂n

Systematic aphaeresis of the initial vowel of the verb s asjeˈta ‘to sit down’ has thus given rise to two competing paradigms. The point to be stressed is that this example could be very well analyzed as another type of polymorphism; for sure, the basis for this kind of alternation is phonetically driven. However, it can be hypothesized that it is the result of a process of reanalysis (or metanalysis, or rebracketing), whereby a reinterpretation of the boundaries of an expression has taken place.

Needless to say, lexical polymorphism is less easy to observe because a given lexical unit covers a functional/semantic niche which can hardly overlap in all its contours with another one (cf. the case of French patate/cartoufle/pomme de terre ‘potato’ discussed by Bertoldi 1939, pp. 121–23: among the concurrent designations, the one coming from the seas and the markets of ports cities (patata) and the one coming from the inland markets (cartoufle) were superseded by the victorious designation pomme de terre: from the agricultural milieu the later has passed into the common language (for a discussion of the geolinguistic diffusion of these terms, cf. Kawaguchi 2017 and the references therein))19.

Crucially in some cases, the speakers seem to have a preference for one form among the coexistent ones for reasons that have to do not only with phonetic optimization, but also with ‘prestige’ or fashion (cf. Gauchat 1902). In such situations of imbalance, in the long run the preferred form will probably eliminate its “rival” or a (lexical/functional) specialization will take place.

3. Why Polymorphism?

The question of the motivation behind the extreme proliferation of forms and paradigms just mentioned is a real challenge for morphology and morphological theories20. We already said that variation and variability are the natural state of languages and that their stability is something very relative: only the prism of a shared norm can give the (false) impression of unity, homogeneity and stability. As Makaev (1965, p. 35) puts it “(…) variation is a necessary characteristic not only of a certain level of language, but also of the language system itself.” In addition, we already observed that if the coexistence of interchangeable variants is easily observable in dialects, it is also a property of official norms. As a matter of fact, there is nothing strange in the fact that at some point, official languages show the same internal tensions and the same proliferation of forms that in the long run consecrate the selection of a given option21.

The question then arises as to the genesis, the maintenance and resolution of this kind of situation. Recall that according to Allières (1953–1954, p. 1) phonetic polymorphism is the “coexistence, dans le langage d’un sujet parlant, de deux ou plusieurs variantes phonétiques d’un même mot, utilisées concurremment pour exprimer le même concept, le choix de l’une ou de l’autre apparaissant comme absolument indépendant du conditionnement articulatoire—nous pensons par exemple au “tempo” d’élocution,- ou d’une recherche quelconque d’expressivité” [Polymorphism is the coexistence, in the language of a speaking subject, of two or more phonetic variants of the same word, used concurrently to express the same concept, the choice of one or the other appearing to be absolutely independent of articulatory conditioning—we are thinking, for example, of the ‘tempo’ of elocution—or of any search for expressivity”] (cf. as well Allières 1981). And the master of Allières, Jean Séguy (1973, p. 88), pointed out that “Le polymorphisme est un aspect du langage tout à fait réel, qu’il faut observer et étudier, qui devra livrer ses secrets, car il est en contradiction avec la loi d’économie. Tout ce qu’on en sait, c’est qu’il fonctionne comme prélude d’un changement à terme plus ou moins long: la coexistence de signes équivalents se réduit par élimination, et par une élimination aléatoire” [“Polymorphism is a real aspect of language, which must be observed and studied, and which will have to reveal its secrets, because it is in contradiction with the law of economy. All that is known about it is that it functions as a prelude to a more or less long-term change: the coexistence of equivalent signs is reduced by elimination, and by random elimination”]. In other words, polymorphism refers to free variation of forms which are equally available in the system of language, without any semantic or stylistic motivation behind the variation. Needless to say, although linguists have been reluctant to acknowledge the very possibility of free variation (cf. Goldberg 2019, 25sqq.), although polymorphism contradicts any principle of economy, it can be observed at any moment in the life of language, and it may be considered as a cue of transitional stages in language development and evolution. This is understandable on the grounds that polymorphism often is the manifestation of ongoing (phonetic, lexical, etc.) mutations (cf. Lope Blanch 1977, p. 599). By the way, Allières (op. cit., p. 2) rightly points out that in transition areas, polymorphism is more expected and frequent than in other areas, thus reflecting the multiple attraction some dialects may be subject to in these particular situations (cf. the case of the coexisting forms pɛt and pɛl (<pelle) in the dialect of Seilh, where the typically Gascon evolution -ll- > -t- is facing the pressure of Languedocian, where -ll- > -l-). If we add that old forms or features may coexist with innovative ones, we will see that the proliferation illustrated above is not something exceptional: rather, what should be viewed as exceptional and unrealistic is the erroneous picture offered by traditional grammars (and even by some linguists) that present forms and paradigms as something fixed, stable, unique and carved in stone for eternity. As Allières (1992, p. 181) points out, «L’exercice qui consiste à décliner des paradigmes nominaux ou verbaux hors contexte n’est évidemment qu’un dévoiement des fonctions langagières, propre à ces êtres aberrants que sont les grammairiens, professeur ou élèves….» [«The exercise of declining nominal or verbal paradigms out of context is obviously only a deviation of language functions, specific to these aberrant beings that are grammarians, teachers or students….»].

From this point of view and as pointed out by Francescato (1961), the opposition between a diachronic plane and a synchronic one should at best be viewed as an oversimplification: every linguistic system bears the traces of old patterns (cf. Darmesteter 1890, p. 29) and every linguistic system pre-announces—de facto or in potentia—future developments and evolutions. Of course, when concurrent items or paradigms are synchronically available, their equivalence or interchangeability is doomed to be lost due to the asymmetric nature of linguistic signs. For whatever reason a given item or a given (sub-)paradigm may be felt as more adapted to the context—to not say ‘to its environment’ –, more expressive than another one, more optimal from a phonological point of view (cf. the preference for CV syllables over CCVCC ones for example), etc., hence, its promotion and the progressive loss of the ‘concurrent’ one(s)22. From this point of view, polymorphism can be seen as the reflex of diachrony in the very synchrony (cf. Kristol 2018, p. 344). These considerations are not without echoing the analysis of Darmesteter (1890, p. 32), according to whom “Dans le langage comme dans la matière organisée, nous assistons à cette lutte pour l’existence, à cette concurrence vitale qui sacrifie des espèces à des espèces voisines, mieux armées pour le combat de la vie.” [“In language as in organized matter, we are witnessing this struggle for existence, this vital competition that sacrifices species to neighbouring species that are better armed for the struggle of life.”]

Needless to say, the biological metaphor—due among others to Schleicher, Bopp (1836),23 etc. and endowed by Darmesteter (1887) and (1890)—cannot be pushed too far. It gave rise to many criticisms, as illustrated, for example, by the important review of the work of Darmesteter La vie des mots (1887) writen by the French Romanist Gaston Paris in 1906 (Paris 1906) after the publication of Schleicher’s Essay on Language and Darwinism.

By the way, the very notion of polymorphism can also be found in the works of biologists and was used, among others, by the English naturalist Alfred Russel Wallace to refer to “the co-existence in the same locality of two or more distinct forms, not connected by intermediate gradations, and all of which are occasionally produced from common parents” (cf. Wallace [1871] 2009, p. 145).

From this point of view the revival of the “survival for the fittest”24 alluded to by Aronoff (2016) and Lindsay and Aronoff (2013) among others—and this holds as well for the notion of “competition” itself—should be considered with great caution. By the way, in his review of Darmesteter La vie des mots (1887), Michel Bréal (1887, pp. 187–88) pointed out that the expressions “life,” “organisms,” “struggle,” “competitors,” etc. should be used with the awareness that they are merely metaphors. In the case of the coexistent Conditional forms mentioned in (8)–(10), the complexity of the string ˈ(d)rjɔ is rooted in production constraints: it is understandable that sequences requiring great articulatory energy may be disfavoured over sequences requiring less or minimal articulatory energy (cf. Straka 1979). From this point of view, it must be borne in mind that phonetic optimization is a leading parameter in morphological as well as in syntactic innovations25.

We should not forget either what was one of the main ideas put forward by Antoine Meillet and his school but also by Otto Jespersen (op. cit., p. 12): “Language is activity, chiefly social activity undertaken in order to get into touch with other individuals and communicate to them one’s thoughts, feelings and will.” Of course, getting in touch with other individuals presupposes adaptation, (real time) optimization, distinctiveness, relevance and informativeness or, a contrario, (informational) worthlessness26, all constraints that may conflict with one another. Getting in touch with other individuals may also lead one to differentiate, to distinguish from others in one way or another, and such a differentiation process may manifest in particular (lexical, phonetic, etc.) options made available by the linguistic system. From this point of view, it is clear that hypercharacterization, overmarking, expressivity, etc. may lead to more complex patterns than expected by the economy principle27. Observe as well that even though the Occitan dialects presented in this contribution may offer up to four coexisting paradigms for one and the same verb, it is also true that in the individual speech acts and discourse events, the speaker generally selects one option among several choices at their disposal. Moreover, the selection of such or such a variant will depend on as many parameters as the tempo, the degree of formality, focus marking, the discourse situation and the phonetic environment of the selected item, etc.

4. Conclusions

It has been argued that if polymorphism is the natural state of linguistic systems, it is also anti-economic from a cognitive point of view: learning and memorizing full coexistent paradigms presupposes a high processing cost28. It follows that the diachronic evolution of languages may develop adaptive solutions to circumvent the potential drawbacks of extreme polymorphism: ‘(un)natural (?) selection’ leads to the reduction or elimination of morphological proliferation. Of course, before reduction or elimination take place, a more or less extended period of time may elapse during which a preference for some (coexisting) paradigmatic options may manifest. Such a preference can be illustrated in the Imperative forms in (1b–c), where the forms of only one paradigm are selected. In other dialects however, co-occurring Imperative forms are attested as well. However, where a given form has been preferred over another/other one(s), it may be taken to rely on phonetic maximization and pragmatic efficiency amenable to Jespersen’s principle according to which “the best is what with a minimum of effort on the part of the speaker produces a maximum of effect in the hearer.” Of course, we cannot exclude as well that morphological proliferation may be the result of some kind of linguistic insecurity: Occitan speakers faced with the pressure to speak and write in French since their childhood have lost, year after year, the mastering of Occitan. As recalled by Salminen (2007, p. 256) “There may well be over a million people able to speak Languedocian, but very few children learn the language, which makes the language severely endangered despite the large number of speakers and the continuing cultivation of the Occitan literary language which is essentially based on Languedocian.” However, the situation of the endangerment of the Occitan languages cannot be put forward, to account for polymorphism in (fully) living languages and dialects, nor can it be put forward to account for polymorphism in times when the Occitan dialects were the first languages of the speakers. By the way, we saw that polymorphism cannot be said to characterize exclusively living dialects: literary varieties may also show coexisting forms for the same cell of a given (sub-)paradigm. Of course, polymorphism is not limited to phonetics or phonology. In Italian, it is hard to find any stylistic/semantic difference between, say, Non aiutarlo/Non lo aiutare ‘Don’t help him’ (cf. also variants such as Spanish Se lo voy a decir quando lo vea/Voy a decírselo quando lo vea mentioned by Lope Blanch 1993, p. 104). But if both options are available synchronically in Italian, the first one was excluded in Old Italian and the coexistence of both patterns can be seen as the reflection of a diachronic scenario in which the second position clitic constraint has been relaxed (cf. Mussafia 1886, p. 261). From this point of view, polymorphism cannot be reduced to a lack of stability, to performance error, to linguistic attrition or to interference. It ought to be fully recognized as a fundamental aspect of language and language change.