The overall goal of this doctoral study was to investigate the different types of modal markers in Kirundi, to describe their syntactic properties and semantic values, and to establish their frequency in the corpus. To this end, we used both an onomasiological and a semasiological approach, in order to identify and describe the different modal markers. The study was based on a corpus of Kirundi, consisting of 1.9 million tokens, which was compiled according to authoritative standards in the domain, taking into account requirements in terms of balance and representativeness. The corpus comprised texts belonging to a number of different genres/topics and was made up of texts extending over a time span of nine decades. Three types of modal markers were identified. The most important is the affix oo , commonly labelled potential or conditional in Kirundi descriptions. It is the most frequent modal marker in the corpus, with the largest semantic scope and the greatest number of modal values. Therefore, it stands as the grammatical morpheme dedicated to the expression of modality in Kirundi. The second type of modal marker is a set of verbs, whose semantic values and frequency vary. Some of them have been grammaticalized to the extent that they act as auxiliaries. Others have not been grammaticalized to this extent. The third category consists of adverbs. This category expresses only one modal category: epistemic modality. This study showed the intertwining between modal and non-modal categories, by studying elements whose use overlaps with the domain of modality. On this matter, the situation is that directivity is attested as a post-modal value for markers expressing specifically deontic necessity. As regards volition, it is also a post-modal value attested on markers expressing necessity. As for evidentiality, it is closely associated with epistemic modality. This work showed also that within modal markers in Kirundi, intersubjectification and grammaticalization are linked. The study makes a contribution to the description of a semantic category that is generally under-described in Bantu languages. Moreover, through the method of corpus linguistics, it has used an original approach for a language without written tradition, i.e. the approach of studying and exemplifying linguistic phenomena by looking at natural communication, which is different from elicitation. This thesis has illustrated the advantages offered by such a method, as it allows facts to be expressed statistically. This work has also created the possibility of a diachronic analysis of the development of modal markers in Kirundi. Moreover, as this study dealt only with affirmative forms, an extension to the study of the relationship between polarity and modality in Kirundi would allow for a better understanding of the modal category in Kirundi.

The overall goal of this doctoral study was to investigate the different types of modal markers in Kirundi, to describe their syntactic properties and semantic values, and to establish their frequency in the corpus. To this end, we used both an onomasiological and a semasiological approach, in order to identify and describe the different modal markers. The study was based on a corpus of Kirundi, consisting of 1.9 million tokens, which was compiled according to authoritative standards in the domain, taking into account requirements in terms of balance and representativeness. The corpus comprised texts belonging to a number of different genres/topics and was made up of texts extending over a time span of nine decades. Three types of modal markers were identified. The most important is the affix oo , commonly labelled potential or conditional in Kirundi descriptions. It is the most frequent modal marker in the corpus, with the largest semantic scope and the greatest number of modal values. Therefore, it stands as the grammatical morpheme dedicated to the expression of modality in Kirundi. The second type of modal marker is a set of verbs, whose semantic values and frequency vary. Some of them have been grammaticalized to the extent that they act as auxiliaries. Others have not been grammaticalized to this extent. The third category consists of adverbs. This category expresses only one modal category: epistemic modality. This study showed the intertwin¬ing between modal and non-modal categories, by studying elements whose use overlaps with the domain of modality. On this matter, the situation is that directivity is attested as a post-modal value for markers expressing specifically deontic necessity. As regards volition, it is also a post-modal value attested on markers expressing necessity. As for evidentiality, it is closely associated with epistemic modality. This work showed also that within modal markers in Kirundi, intersubjectification and grammaticalization are linked. The study makes a contribution to the description of a semantic category that is generally under-described in Bantu languages. Moreover, through the method of corpus linguistics, it has used an original approach for a language without written tradition, i.e. the approach of studying and exemplifying linguistic phenomena by looking at natural communication, which is different from elicitation. This thesis has illustrated the advantages offered by such a method, as it allows facts to be expressed statistically. This work has also created the possibility of a diachronic analysis of the development of modal markers in Kirundi. Moreover, as this study dealt only with affirmative forms, an extension to the study of the relationship between polarity and modality in Kirundi would allow for a better understanding of the modal category in Kirundi.

Résumé : La présente étude concerne l'évolution sémantique et formelle d'un adverbe épistémique du kirundi, umeengo/umeenga, les deux formes constituant des variantes libres. Elle est une des premières études s'inscrivant en linguistique de corpus pour les langues bantoues, une démarche rarement adoptée pour cette famille de langues, de même qu'elle étudie une catégorie grammaticale généralement négligée dans les études linguistiques, c'est-à-dire l'adverbe. Nous décrivons les différents emplois de umeengo/umeenga : comme adverbe, mais aussi comme comparatif, et dans les emplois du verbe qui en émerge à l'étape suivante dans son évolution. Ses origines sont éclairées par un indice retrouvé en kinyarwanda, langue proche du kirundi, ainsi que dans la variante dialectale du kirundi parlée à l'ouest du Burundi, dans la région de l'Imbo. D'après ces éléments, il apparaît que cet adverbe est formé par univerbation, plus précisément par fusion d'un verbe conjugué, umenya (du verbe -meny-« connaître ») et du quotatif ngo « que ». Dans la suite du processus, à partir de cet adverbe épistémique, se forme un nouveau verbe, -meeng-« croire ; penser », différent de celui du départ, tant dans le sens qu'au niveau de la forme. Une telle évolution au sein de l'adverbe, qui naît d'un verbe et débouche lui-même à la naissance d'un autre verbe, est rare dans les langues du monde, de là l'intérêt de notre analyse. Dans cette étude, nous nous interrogeons sur la nature des processus à l'oeuvre. Nous montrons que les changements structurels et sémantiques par lesquels l'adverbe umeengo/umeenga s'est (Mould 1977). Dans notre étude sur la modalité en kirundi, langue parlée au Burundi, nous apportons également une contribution à ce sujet en consacrant un chapitre aux adverbes épistémiques (Mberamihigo 2014). Il s'agit sans doute de la première étude systématique de ce genre d'adverbes pour une langue bantoue. Pour le kirundi, si nous pouvons faire préalablement un état des lieux du traitement de l'adverbe, nous constatons que la brièveté de la manière dont cette classe est abordée témoigne de ce moindre intérêt, ou de l'évitement dont il fait l'objet de manière générale. Elle n'est pas traitée dans la grammaire du kirundi de Meeussen (1959), pas plus qu'elle n'est évoquée dans l'index de cette étude. Ntahokaja (1994) réserve à l'adverbe un traitement dans le cadre des classes de mots. Néanmoins, il ne donne le détail que sur quatre espèces : les adverbes de lieu, les adverbes de temps, les adverbes de manière et les adverbes d'affirmation et de négation. Quant à Cristini (2001), il leur consacre un chapitre, où il en décrit la formation, en grandes lignes. Il répertorie les 1 Les recherches pour cet article ont été effectuées dans le cadre du doctorat du premier auteur (2010-2014) sous la direction des deux autres auteurs. La thèse a été faite en cotutelle entre l'Université libre de Bruxelles et l'Université de Gand et a été financée par une bourse du Gouvernement du Burundi. Nous tenons à remercier deux collègues anonymes ainsi que ...

This article is a first systematic study of the expression of possibility in Kirundi, more specifically of its verbal markers. Possibility is traditionally seen as one of the core components of the semantic domain of modality. The onomasiological approach of this modal sub-domain has resulted in the identification of four verbal potential markers, i.e., the auxiliaries-bâsh-and-shóbor-, the semi-auxiliary-shóbok-, and the TAM affix-oo-. These four markers of possibility manifest different degrees of grammaticalization along the full verb > auxiliary > affix cline. Grammaticalization in the structural domain seems to be correlated with semantic change, both within and beyond the semantic domain of possibility. The related verbs-shóbor-and-shóbok-, which have no or little remaining lexical uses, cover the entire semantic domain of possibility in contrast to-bâsh-which has still clearly distinct lexical uses ('to be active, to be healthy') and only conveys participant-inherent possibility. The inflectional affix-oo-, covering the entire domain of possibility and having developed other modal and non-modal meanings, manifests the most advanced semantic generalization. Moreover,-oo-underwent the strongest subjectification within the semantic domain of possibility and even developed different intersubjective uses. Hence, the most grammaticalized marker of possibility in Kirundi not only underwent the strongest semantic generalization, but its meaning is also the most (inter)subjectified.