Organisation sémantique et mécanismes de généralisation des concepts abstraits

Semantic organisation and generalisation mechanisms of abstract concepts

• Auteur / Autrice : Dounia Lakhzoum
• Direction : Marie Izaute, Ludovic Ferrand
• Type : Thèse de doctorat
• Discipline(s) : Psychologie
• Date : Soutenance le 16/12/2021
• Etablissement(s) : Université Clermont Auvergne (2021-...)
• Ecole(s) doctorale(s) : École doctorale des lettres, sciences humaines et sociales (Clermont-Ferrand)
• Partenaire(s) de recherche : Equipe de recherche : Laboratoire de psychologie sociale et cognitive (Clermont-Ferrand ; 1984-...)
• Jury : Examinateurs / Examinatrices : Diane Pecher, René‏‎ Zeelenberg
• Jury : Rapporteur / Rapporteuse : Simona Maria Brambati, Marc Brysbaert

Mots clés

Mots clés contrôlés

Vocabulaire

Français (langue)

Sémantique -- Psychologie

Apprentissage des concepts

Représentation (linguistique)

Sémantique structurale

Mots clés libres

[pas de mot clé en français]

Abstract concepts processing

Semantic similarity

Verbal association

Semantic network analysis

Picture-word priming

Abstraction

Generalisation mechanisms

Résumé

[pas de résumé en français]

The study of semantic representation of abstract concepts entails answering some fundamental questions such as: (1) How is the meaning of the word represented? (2) How does it relate to other words whether similar or associated? (3) How does meaning relate to conceptual structure? (4) How is the meaning abstracted and generalised to all instances of the concept? In order to answer such question, researchers need to have access to normed [s]timuli. The literature on abstract concepts still lacks from the avaibility of such stimuli. The first experimental chapter (Chapter 2) introduces a database of semantically similar pairs of French words with varying levels of abstractness. This database is to the best of our knowledge and at the time of writing this manuscript, the first to introduce semantically similar pairs of abstract words in the French language. Chapter 3 introduces another database for word associations between concepts using the same words as in the previous database. Correlation analyses revealed that cue words presented to the participants elicited response words of a similar level of concreteness meaning that concepts are organised in the mental lexicon according to a gradient of concreteness. Analyses from associative strength have shown that concrete concepts elicit stronger associations compared to abstract concepts. The large amount of data generated by the word association task allowed for the implementation of mathematical graph analyses. Consequently, Chapter 4 introduces the first semantic network built from French association data and the first to compare topological parameters for concrete and abstract concepts. Results have shown that the French mental lexicon is structured according to a small-world pattern characterised by a sparse density, a short average path length between nodes and a high clustering coefficient. Comparison analyses between the networks for concrete and abstract networks respectively have shown that concrete concepts are organised in denser communities compared to abstract concepts. In addition, the concrete word nodes are more influential in the network and can spread information better due to their position and patterns of connectivity. In Chapter 5, we explored the richness of abstract concepts in a picture-word priming paradigm and compared the role of situational and intangible features in their processing. Results have shown that even when compared with situational picture primes, extraction mechanisms still occurred in the case of abstract pictures. We interpreted these results as further evidence of the richness of abstract concepts which can be composed from features other than lexical or situational, namely also abstract and intangible. These intangible features could be construed as evidence of abstracted representations from varied exemplars of concepts becoming generalised statistical traces in long-term memory. In Chapter 6, we investigated the effect of similarity and contextual diversity on the ability to generalise the meaning of abstract concepts to all instances. Results have shown that exposure to contextually diverse exemplars enhanced performance in the testing phase compared to similarity-based exposure. We conclude this thesis by proposing a theoretical model based on our findings with the purpose of bridging the gap between levels of investigation of abstract concepts processing. Taken together, our findings describe a reversed pattern of processing with lower levels ruled by similarity-based processes while higher levels of processing are rules by diversity-based mechanisms.