Towards a semantic typology of feelings. The linguistic encoding of emotions across languages, and what it tells us - TEL - Thèses en ligne Access content directly
Habilitation À Diriger Des Recherches Year : 2022

Towards a semantic typology of feelings. The linguistic encoding of emotions across languages, and what it tells us

Vers une typologie sémantique des sentiments. L’encodage linguistique des émotions à travers les langues, et ce qu’il nous apprend

Maïa Ponsonnet


As is customary with reports submitted for Habilitation à Diriger les Recherches in France, this document accounts for my research trajectory and results since I completed my Linguistics PhD, in 2013. The volume opens with an introduction where I present my intellectual and institutional itinerary. This is an opportunity to explain how I became a linguist – with some disciplinary ‘detours’ via Philosophy, Anthropology, and early experience with the language community I was to work with for my Linguistics PhD. Chapter 1 also explains how my current research theme, the linguistic encoding of emotions, originated in a philosophical interest for other minds and internal states, inspired by my Wittgensteinian training. In a nutshell, my current research question remains: how do humans manage to communicate about ‘private’ states – in Wittgenstein’s sense of ‘private’, which means ‘internal’? Chapter 2 begins with my contributions to linguistic description, the subfield of linguistics that grounds and anchors my scientific approach. Field-based description is the scientific tradition in which I was first trained during my PhD in Linguistics at the Australian National University. Although much of my more recent scientific results pertain to typology and linguistic anthropology, descriptive linguistics still forms an important part of my research activities and network. Even before the end of my PhD, I have always endeavored to contribute to the description of languages of northern Australia – where my field sites are located – with two goals in mind. On the one hand, I have produced the descriptions needed to understand my primary object of study, i.e. the linguistic encoding of emotions. In addition, I have collaborated with colleagues and students to address questions of general interest in the descriptivists’ community and beyond. The first sections of Chapter 2 offer brief presentations of the Australian languages I have investigated first hand: Dalabon and other languages of the Gunwinyguan family (Australian, non-Pama-Nyungan, centre-north of the continent); as well as Kriol, an English-based creole spoken in the Gunwinyguan region. The chapter then explains my contributions to the description of a range of linguistic phenomena including: possession classes, benefactive constructions, optional ergativity in Dalabon; reflexive and reciprocal marking, reduplication, modality, and subject elision in Kriol. The third Chapter presents my research concerning the semantic typology of the linguistic encoding of emotions. This work investigates the linguistic tools that humans use to communicate about emotional internal states, including both descriptive tools – mostly words – and expressive tools, i.e. those indexing speakers’ states, like interjections, prosody, etc. I aim to understand how these tools pervade language and influence linguistic communication more broadly, with a view to shed light upon language itself, as well as upon the cognitive status of emotions. With respect to language, studying the linguistic encoding of emotions tells us how humans communicate less visible experience, and how grammars realize expressive features. With respect to the cognitive status of emotions, it tells us about how human categorize them, i.e. which types of emotions are more often (or even universally) lexicalized or grammaticalized across the world, and how sensitive to cultural variation these preferences are. In this spirit, I have conducted ground-breaking studies concerning a number of understudied linguistic resources, both at the scale of the Australian continent (e.g. figurative language, emotion nouns, interjections), and across the world (reflexemes, i.e. reflexive constructions that lexicalize an emotional meaning; evaluative morphology). Considered altogether, these studies confirm that different linguistic resources specialize for different emotions. For instance, reflexemes (see above) often describe anger, while body-based emotion metaphors more often describe fear; interjections frequently express surprise, while morphology favors endearment, affection and compassion. Importantly, there is a clear contrast between descriptive resources and expressive resources. On the one hand, the descriptive resources I have investigated so far tend to encompass a broader, and ultimately comparable range of emotions. On the other hand, expressive resources diverge sharply from this descriptive range, as each of them focuses on a distinctive and smaller set of emotion categories. Although descriptive linguistic resources are semantically more homogenous, with some emotions more lexically productive than others. These include generic categories such as ‘feel good’ and ‘feel bad’ as well as some other, often ‘primary’ emotions such as anger. Future research is needed to confirm and refine our understanding of emotional specialization for each linguistic resource; and explain these patterns of specialization. The question addressed in Chapter 4 flows from the examination of linguistic diversity in the semantic domain of emotions: given the scope of variation, do the particular tools available to encode emotions in a given language influence the way its speakers talk about emotions? And perhaps even, by way of consequence, the way they communicate about, manage, or experience emotions? Phrased as above, this question sounds daunting: it asks nothing less than ‘is our emotional experience subject to linguistic relativity?’ However, the matter becomes more approachable once rephrased, as already suggested in my PhD thesis on the Dalabon language (Ponsonnet 2013, 2014): ‘can the particular grammatical architecture of a language influence the tools available to its speakers to express and describe emotions?’. The latter interrogation is much easier to investigate, because it does not require us to scrutinize speakers’ ‘representations and practices about emotions’, which are very hard to capture and quantify. Instead, we can now focus on correlations between grammar (for example, the presence of a particular construction) and linguistic resources in the domain of emotions (for example, the availability of evaluative morphology, or of certain metaphors). Expanding upon my PhD research, I have approached this question from two practical angles: the grammatical affordance of metaphors, discussed in Chapter 4 of the HDR report; and the case of language shift, presented in Chapter 5 (see below). My PhD thesis on the linguistic encoding of emotions in Dalabon highlighted some possible influence of grammar upon figurative representations of emotion in this language. Dalabon emotion metaphors are formed with body-part or emotions nouns. Both types of nouns are inalienable, and this appears to constrain the range of emotion metaphors they can participate in. To further test this hypothesis, in collaboration with a number of linguists working across the continent, I considered emotion metaphors and their grammar in a larger sample of Australian languages. This comparison brought further support to the hypothesis that grammar can influence metaphors. Warlpiri (Pama-Nyungan), for instance, where emotion and body-part nouns do not share all the grammatical constraints observable in Dalabon, also features emotion metaphors that are absent in Dalabon (Laughren & Ponsonnet 2020). Further, in languages with coverbs, like Gija (Jarragan, non-Pama-Nyungan), the emotions metaphors produced by the pairing between coverbs and generic verbs produce a completely different set of metaphors than the ones typically derived from emotions and body-part nouns across Australia (Kofod & Crane 2020). This supports the hypothesis that figurative representations of emotions do not simply echo speakers’ shared cultural representations, but are also influenced by some of the particular grammatical traits of individual languages. The above conclusion is further corroborated by the analysis of a language-shift situation, as reported in Chapter 5. In this study, I compared the tools available to describe and express emotions in Dalabon, with those available in Kriol, the English-based creole that has replaced Dalabon. The typological profile of Kriol contrasts sharply with that of Dalabon, primarily because Kriol (like most creoles) is to a large extent isolating, while Dalabon is highly polysynthetic. This results in differences in terms of emotional linguistic tools, and the range of emotion metaphors seems to be one of them. Body-based metaphors constructed via the incorporation of body-part nouns are pervasive in Dalabon, but this phenomenon is limited in Kriol. On the other hand, beyond figurative representations, adopting a language with a fundamentally different architecture does not, at this level of analysis, appear to have a strong impact on what speakers can say about emotions. Although the vast majority of Kriol words formally resemble their English etyma, many of them share semantic properties with Dalabon words. Sharp formal linguistic differences do not necessarily translate into significant semantic gaps, as illustrated with morphological diminutives, for instance. They are pervasive in Dalabon, and – unsurprisingly – absent in Kriol. However, this does not mean that Kriol speakers are left with no tools to fulfil the communicative roles played by diminutives in Dalabon. On the contrary, Kriol expands the semantic range and usage of alternative resources such as interjections and reduplication, which in this language occur in the contexts where morphological diminutives occur in Dalabon. It is useful to note that language-shift situations differ from translation situations, where there are indications that bilingual speakers are constrained by the alternative code they use. By contrast, when a new language is created, speakers are free to use forms as they see fit. Old meanings can be repackaged under new forms, and speakers use a variety of strategies to carve linguistic tools that will communicate what they want to communicate. In other words, when a whole group shifts to a new language, the new lexico-grammatical code adapts to what speakers have to say – not the opposite. Or at least this is true in the semantic domain of emotions – whether this property is shared with many other semantic domains, or whether this reflects the particular cognitive and social status of emotions, remains to be established. Chapter 6 details the implications and benefits of my work for the language communities I work with. This discussion flows directly from the presentation of the case study on Kriol and Dalabon presented in Chapter 5. The impulse for this research came from my observing Kriol speakers’ attachment and enthusiasm for their language. Beyond the question of emotions, an important message to take away from this study is that the impact of language shift on what speakers can say and express should be examined rather than taken for granted. It can be tempting for linguists to approach language shift as a loss, but this raises ideological issues. In the context of colonisation, language shift is often experienced by communities as a loss indeed, in fact as a theft, or worse. At the same time, many minority groups around the world suffer the negative consequences of naïve representations that equate shifting to a new language with ‘losing one’s culture’. In practice, as the above study shows, adopting a new language activates linguistic creativity, and descriptive linguists are in a good position to shed light upon this mode of cultural affirmation. The rest of Chapter 6 presents the results of my research in terms of language documentation, as well as engagement with speakers, community organisations, and the broad audience in general (e.g. curated language corpora, on line broad audience articles, etc.). Chapter 7 The report closes with a brief exposition of future research pathways, in Chapter 7. This is prefaced by a summary of my main results: • On the semantic typology of the linguistic encoding of emotions (Chapter 3): o Typological investigations of understudied linguistic resources: figurative representations, emotion nouns, reflexemes, interjections, evaluative morphology, as well as other expressive morphological features. o Expressive resources are more specialized semantically than descriptive resources, i.e. each of them focuses on a different set of emotions; while different descriptive resources tend to target the same suite of emotions. o Not all emotions attract the same amount of descriptive linguistic resources. Generic categories like ‘feel good/bad’ are better represented, along with some primary emotions such as anger. • On the influence of grammatical architecture on the linguistic encoding of emotions (Chapters 4 & 5): o There are indications that figurative language is sensitive to grammatical properties, i.e. the range of emotion metaphors that occurs in a language is partly constrained by the grammatical constructions available in this language. o Beyond this, the examination of a language-shift situation suggests that language is plastic enough for speakers to compensate differences in grammatical architecture. o In this context, notwithstanding the accumulation of subtle differences, at the level of semantic description and corpus comparison the impact of using a new language on the linguistic encoding of emotions is not clearly significant. The rest of Chapter 7 fleshes out these main points to derive research pathways, and explains my plans to pursue them.
Comme il est d’usage, le premier chapitre de ce rapport résume mon itinéraire, à la fois intellectuel et institutionnel. J’y explique notamment l’ancrage philosophique de mes questions de recherche sur l’encodage linguistiques des émotions. Fondamentalement inspiré par les écrits de Wittgenstein, mon travail explore le problème suivant : comment les communautés humaines utilisent-elles le langage pour communiquer des états « privés » (au sens de Wittgenstein). Le Chapitre 2 traite de mes contributions à la linguistique descriptive, qui constitue ma formation doctorale à l’Australian National University, et reste un pivot de mes activités et réseaux de recherche. Bien que mon travail s’inscrive aujourd’hui principalement en typologie ou en anthropologie linguistique, je continue de contribuer au travail de description des langues, notamment lorsque ces descriptions éclairent mes questions de recherche, ou dans le cadre de collaborations. Ce chapitre présente les langues australiennes que j’ai contribué à décrire (dalabon et autres langues gunwinyguan, et kriol, créole d’anglais parlé en Australie), puis retrace mes publications concernant les classes possessives, les constructions bénéfactives, l’ergativité optionnelle en dalabon ; les marqueurs réflexifs, la modalité, et l’élision du sujet en kriol. Le Chapitre 3 décrit mes travaux concernant la typologie sémantique dans le domaine des émotions. Ceux-ci concernent à la fois des ressources descriptives (i.e. ce qui relève du lexique) et des ressources expressives, qui indexent un état du locuteur, comme les interjections ou la prosodie par exemple. Ces études, menées pour certaines à l’échelle du continent australien (e.g. métaphores émotionnelles et parties du corps, noms émotionnels, interjections) ou sur des échelles plus larges (« réflexème », i.e. constructions réflexives lexicalisant un sens émotionnel, morphologie évaluative), permettent notamment d’observer des associations préférentielles entre certains outils linguistiques et certaines catégories émotionnelles. Par exemple, les interjections (australiennes) expriment souvent la surprise, mais jamais la honte ; la morphologie exprime très fréquemment l’attendrissement et l’affection ; les réflexèmes décrivent souvent la colère ; etc. D’une manière générale, les ressources expressives sont plus spécialisées et différenciées que les ressources descriptives, qui ont tendance à converger vers un panel d’émotions relativement comparables (avec des nuances). Ce panel comprend notamment des catégories génériques, comme se sentir bien ou mal ; ainsi que des émotions dites ‘primaires’, comme la colère par exemple. Ces observations soulèvent naturellement la question des motivations de ces régularités. Dans le Chapitre 4, j’aborde la question de l’influence possible de l’architecture grammaticale, propre à chaque langue, sur les outils disponibles dans cette langue pour parler des émotions – et ainsi, on peut l’imaginer, sur la manière dont les locuteurs communiquent leurs émotions. Si cette question évoque l’épineuse hypothèse whorfienne de la relativité linguistique, elle est en réalité plus abordable, car elle se situe intégralement à l’intérieur du langage, et ne convoque pas les représentations ou pratiques des locuteurs. J’avais déjà envisagé cette problématique dans ma thèse sur l’encodage des émotions en dalabon, où j’avais identifié une possible influence de la grammaire sur les métaphores émotionnelles. Les noms émotionnels du dalabon se comportent grammaticalement comme des noms inaliénables de parties du corps, et cela semble empêcher la formation de certaines métaphores de réification ou de personnification par exemple. Pour tester plus avant cette corrélation entre grammaire et métaphores, j’ai comparé, avec l’aide de collègues, les métaphores émotionnelles d’un plus grand nombre de langues sur le continent australien. Cette comparaison a mis à jour d’autres cas d’influence de la grammaire sur les représentations figuratives des émotions. La comparaison du dalabon et du warlpiri (pama-nyungan) suggère que la levée, en warlpiri, des contraintes grammaticales qui s’appliques à certains noms en dalabon annule aussi les contraintes sur les types de métaphores. En gija (jarragan), le système des coverbes produit des métaphores totalement différentes de celles construites à l’aide de noms d’émotions ou de parties du corps (Kofod & Crane 2020). Il semble donc que les métaphores émotionnelles ne sont pas strictement l’écho de conceptions culturellement partagées – comme il a été soutenu – mais peuvent également être influencées par les caractéristiques grammaticales des langues. Cette influence de la grammaire sur les métaphores émotionnelles est également corroborée par l’analyse d’une situation de changement linguistique, rapportée au Chapitre 5. Cette étude compare l’encodage linguistique des émotions en dalabon et en kriol, le créole à base d’anglais qui remplace le dalabon. Les deux langues sont radicalement différentes formellement – le dalabon étant une langue polysynthétique, et le kriol une langue isolante. Sur le plan des représentations figuratives des émotions, les deux langues diffèrent : les métaphores émotionnelles mobilisant des noms de parties du corps incorporés à des verbes sont omniprésentes en dalabon, mais beaucoup plus marginales en kriol. Cela confirme d’autant plus la sensibilité des représentations figuratives à la grammaire que pour d’autres ressources émotionnelles, le contraste entre les deux langues est moindre. Par exemple, elles présentent de nombreuses similarités sur le plan lexical. Même là où certaines ressources « manquent », pour ainsi dire, en kriol (comme dans le cas des diminutifs morphologiques), les locuteurs ont créé des outils linguistiques pour couvrir les mêmes catégories sémantiques qu’en dalabon. Ainsi, mis à part les représentations figuratives, le passage d’une langue à l’autre n’a pas d’effets flagrants sur la manière dont la langue permet aux locuteurs de communiquer au sujet des émotions. Le Chapitre 6 explique comment mes travaux répondent à des préoccupations des membres des communautés linguistiques avec lesquelles je travaille, ou du grand public en général. C’est le cas notamment de mon travail sur le changement linguistique, qui éclaire des idéologies linguistiques très répandues, et souvent dommageables aux communautés qui ont adopté une nouvelle langue. En effet, la comparaison entre le kriol et le dalabon montre que le changement linguistique n’entraine pas nécessairement une « perte » culturelle, et ne menace donc pas l’identité des locuteurs de ces langues. Le Chapitre 6 montre également comment les communautés de locuteurs bénéficient de la documentation linguistique que j’ai eu l’occasion de produire avec eux, et comment j’ai valorisé mes résultats par le biais d’articles ou de pages web accessibles aux non spécialistes. Le Chapitre 7 présente mes projets de recherches pour les années à venir. Ces projets s’appuient sur mes résultats de recherche les plus notables, à savoir : • En typologie sémantique (Chapitre 3) : o Les ressources émotionnelles expressives sont plus spécialisées et différenciées que les ressources descriptives, qui ont tendance à converger vers un panel d’émotions relativement comparables. o Ce panel comprend notamment des catégories génériques, comme se sentir bien ou mal ; ainsi que des émotions dites « primaires », comme la colère par exemple. • Influences de la grammaire sur l’encodage linguistique des émotions (Chapitres 4 & 5) : o Les représentations figuratives émotionnelles présentes dans une langue peuvent dépendre en partie des propriétés grammaticales de cette langue. o En revanche, le changement linguistique à l’échelle d’une communauté ne modifie pas fondamentalement la manière dont les locuteurs peuvent communiquer autour des émotions à l’aide des outils disponibles.
Fichier principal
Vignette du fichier
Ponsonnet HDR thesis_20220801.pdf (2.18 Mo) Télécharger le fichier
Origin : Files produced by the author(s)

Dates and versions

tel-03919642 , version 1 (07-02-2023)


  • HAL Id : tel-03919642 , version 1


Maïa Ponsonnet. Towards a semantic typology of feelings. The linguistic encoding of emotions across languages, and what it tells us. Linguistics. Université Lyon 2 Lumière, 2022. ⟨tel-03919642⟩
159 View
128 Download


Gmail Facebook X LinkedIn More