The purpose of this study was to determine whether or not individual scores derived by the Illinois Test of Psycholinguistic Abilities can be used as predictors of performance on visual motor tasks of the Wechsler Intelligence Scale for Children and the Frostig Developmental Test of Visual Perception for a child who has learning problems.

Automatic emotion recognition has long been a focus of Affective Computing. It has become increasingly apparent that awareness of human emotions in Human-Computer Interaction (HCI) is crucial for advancing related technologies, such as dialogue systems. However, performance of current automatic emotion recognition is disappointing compared to human performance. Current research on emotion recognition in spoken dialogue focuses on identifying better feature representations and recognition models from a data-driven point of view. The goal of this thesis is to explore how incorporating prior knowledge of human emotion recognition in the automatic model can improve state-of-the-art performance of automatic emotion recognition in spoken dialogue. Specifically, we study this by proposing knowledge-inspired features representing occurrences of disfluency and non-verbal vocalisation in speech, and by building a multimodal recognition model that combines acoustic and lexical features in a knowledge-inspired hierarchical structure. In our study, emotions are represented with the Arousal, Expectancy, Power, and Valence emotion dimensions. We build unimodal and multimodal emotion recognition models to study the proposed features and modelling approach, and perform emotion recognition on both spontaneous and acted dialogue. Psycholinguistic studies have suggested that DISfluency and Non-verbal Vocalisation (DIS-NV) in dialogue is related to emotions. However, these affective cues in spoken dialogue are overlooked by current automatic emotion recognition research. Thus, we propose features for recognizing emotions in spoken dialogue which describe five types of DIS-NV in utterances, namely filled pause, filler, stutter, laughter, and audible breath. Our experiments show that this small set of features is predictive of emotions. Our DIS-NV features achieve better performance than benchmark acoustic and lexical features for recognizing all emotion dimensions in spontaneous dialogue. Consistent with Psycholinguistic studies, the DIS-NV features are especially predictive of the Expectancy dimension of emotion, which relates to speaker uncertainty. Our study illustrates the relationship between DIS-NVs and emotions in dialogue, which contributes to Psycholinguistic understanding of them as well. Note that our DIS-NV features are based on manual annotations, yet our long-term goal is to apply our emotion recognition model to HCI systems. Thus, we conduct preliminary experiments on automatic detection of DIS-NVs, and on using automatically detected DIS-NV features for emotion recognition. Our results show that DIS-NVs can be automatically detected from speech with stable accuracy, and auto-detected DIS-NV features remain predictive of emotions in spontaneous dialogue. This suggests that our emotion recognition model can be applied to a fully automatic system in the future, and holds the potential to improve the quality of emotional interaction in current HCI systems. To study the robustness of the DIS-NV features, we conduct cross-corpora experiments on both spontaneous and acted dialogue. We identify how dialogue type influences the performance of DIS-NV features and emotion recognition models. DIS-NVs contain additional information beyond acoustic characteristics or lexical contents. Thus, we study the gain of modality fusion for emotion recognition with the DIS-NV features. Previous work combines different feature sets by fusing modalities at the same level using two types of fusion strategies: Feature-Level (FL) fusion, which concatenates feature sets before recognition; and Decision-Level (DL) fusion, which makes the final decision based on outputs of all unimodal models. However, features from different modalities may describe data at different time scales or levels of abstraction. Moreover, Cognitive Science research indicates that when perceiving emotions, humans make use of information from different modalities at different cognitive levels and time steps. Therefore, we propose a HierarchicaL (HL) fusion strategy for multimodal emotion recognition, which incorporates features that describe data at a longer time interval or which are more abstract at higher levels of its knowledge-inspired hierarchy. Compared to FL and DL fusion, HL fusion incorporates both inter- and intra-modality differences. Our experiments show that HL fusion consistently outperforms FL and DL fusion on multimodal emotion recognition in both spontaneous and acted dialogue. The HL model combining our DIS-NV features with benchmark acoustic and lexical features improves current performance of multimodal emotion recognition in spoken dialogue. To study how other emotion-related tasks of spoken dialogue can benefit from the proposed approaches, we apply the DIS-NV features and the HL fusion strategy to recognize movie-induced emotions. Our experiments show that although designed for recognizing emotions in spoken dialogue, DIS-NV features and HL fusion remain effective for recognizing movie-induced emotions. This suggests that other emotion-related tasks can also benefit from the proposed features and model structure.

No description available.

[pt] Esta dissertação se insere na área de Psicolinguística e está em interface com os Estudos Cognitivos da Tradução. Examina o processo de segmentação textual de legendas no âmbito da investigação sobre a expertise tradutória em Legendagem, uma modalidade de Tradução Audiovisual subexplorada nesse aspecto. Considerou-se, em particular, a segmentação textual entre linhas de uma mesma legenda, uma habilidade com elevado grau de especificidade e indispensável a tradutores dessa área ao produzir uma redação que otimize o processamento nessas condições. Partindo de estudos sobre Expertise em Tradução (Shreve, 2006) e de pesquisas sobre processamento de sentenças, com foco particular na chamada Teoria do Garden-Path (Frazier e Fodor, 1978), foram concebidos dois experimentos envolvendo tradutores aprendizes e experientes, com vistas a avaliar o desempenho dos dois grupos no processo de segmentação textual (experimento 1) e revisão de segmentações (experimento 2), nos dois casos com legendas em Português. Os paradigmas com alta validade ecológica, cujas tarefas experimentais eram compostas de fases reais do trabalho de profissionais da área, foram desempenhados em um software específico de legendagem. Os resultados do experimento 1 indicaram um desempenho superior do grupo de tradutores experientes, o que foi confirmado no experimento 2, favorecendo as hipóteses dos autores de que i) a experiência é fator relevante no processo de segmentação de legendas; ii) tradutores experientes são sensíveis a segmentações que podem comprometer a leitura (sentenças do tipo garden path). As contribuições desta pesquisa são: a caracterização preliminar da expertise em tradução para legendagem e o aprofundamento da discussão sobre a TGP. / [en] This work is an effort in the Psycholinguistics area and it is interfaced with that of Cognitive Translation Studies. It examines the process of textual segmentation of subtitles in the research on translation expertise in subtitling, an under investigated Audiovisual Translation mode in this regard. We consider, especially, the textual segmentation between lines of the same subtitle, a highly specific and indispensable skill that allow subtitlers to render their work in a way that optimizes the processing under these conditions. Based on studies of Expertise in Translation (Shreve, 2006) and language processing with focus on the Garden-Path Theory (Frazier e Fodor, 1978), we designed two experiments involving novice and experienced translators to evaluate the performance of both groups on the textual segmentation process (experiment 1) and segmentations revision (experiment 2); in both cases subtitles were in Portuguese. Experimental paradigms with high ecological validity, where the tasks were comprised of actual phases of professional work, were performed in a specific subtitling software. The results of experiment 1 showed superior performance of the professional group, which was confirmed in experiment 2, favoring the hypothesis of the authors that i) the experience is a relevant factor in the subtitle segmentation process; ii) experienced translators are sensitive to segmentations that can compromise reading (like garden path sentences). The preliminary characterization of expertise in subtitling and the new data added to the discussion on TGP are the main contributions this work achieves.

Il y a (au moins) deux questions fondamentales que l’on est amené à se poser lorsqu’on étudie le langage: comment acquiert-on le langage? —le problème d’apprentissage —et pourquoi les langues du monde partagent certaines propriétés mais pas d’autres? —le problème typologique. Dans cette thèse, j’entreprends de relier ces deux domaines en me focalisant sur le lexique, l’ensemble des mots de notre langue et leur sens associés, en posant les questions suivantes: pourquoi le lexique est-il tel qu’il est? Et est-ce que les propriétés du lexique peuvent être (en partie) expliquées par la façon dont les enfants apprennent leur langue? Un des aspects les plus frappants du lexique est que les mots que nous utilisons sont ambigus et peuvent être confondus facilement avec d’autres. En effet, les mots peuvent avoir plusieurs sens (par exemple, les homophones) et sont représentés par un ensemble limité de sons qui augmentent la possibilité qu’ils soient confondus (par exemple, les paires minimales). L’existence de ces mots semble présenter un problème pour les enfants qui apprennent leur langue car il a été montré qu’ils ont des difficultés à apprendre des mots dont les formes sonores sont proches et qu’ils résistent à l’apprentissage des mots ayant plusieurs sens. En combinant une approche computationnelle et expérimentale, je montre, quantitativement, que les mots du lexique sont, en effet, plus similaires que ce qui serait attendu par chance, et expérimentalement, que les enfants n’ont aucun problème à apprendre ces mots à la condition qu’ils apparaissent dans des contextes suffisamment distincts. Enfin, je propose que l’étude des mots ambigus permet de révéler des éléments importants du mécanisme d’apprentissage du langage qui sont actuellement absents des théories actuelles. Cet ensemble d’études suggère que les mots ambigus et les mots similaires, bien que présents dans le langage, n’apparaissent pas arbitrairement dans le langage et que leur organisation reflète (en partie) la façon dont les enfants apprennent leur langue. / Modern cognitive science of language concerns itself with (at least) two fundamental questions: how do humans learn language? —the learning problem —and why do the world’s languages exhibit some properties and not others? —the typology problem. In this dissertation, I attempt to link these two questions by looking at the lexicon, the set of word-forms and their associated meanings, and ask why do lexicons look the way they are? And can the properties exhibited by the lexicon be (in part) explained by the way children learn their language? One striking observation is that the set of words in a given language is highly ambiguous and confusable. Words may have multiple senses (e.g., homonymy, polysemy) and are represented by an arrangement of a finite set of sounds that potentially increase their confusability (e.g., minimal pairs). Lexicons bearing such properties present a problem for children learning their language who seem to have difficulty learning similar sounding words and resist learning words having multiple meanings. Using lexical models and experimental methods in toddlers and adults, I present quantitative evidence that lexicons are, indeed, more confusable than what would be expected by chance alone. I then present empirical evidence suggesting that toddlers have the tools to bypass these problems given that ambiguous or confusable words are constrained to appear in distinct context. Finally, I submit that the study of ambiguous words reveal factors that were currently missing from current accounts of word learning. Taken together this research suggests that ambiguous and confusable words, while present in the language, may be restricted in their distribution in the lexicon and that these restrictions reflect (in part) how children learn languages.

[pt] Esta tese, que se constitui na interface entre a Psicolinguística e os Estudos da Tradução, tem como objetivo investigar questões associadas a custo de processamento na tradução de DPs em inglês, com núcleo nominal precedido de dois ou três modificadores substantivos, adjetivos ou mistos, para o português. Além disso, busca-se refinar, a partir da integração entre as duas áreas, uma caracterização do processo tradutório com base em evidências da tradução dos DPs e de custo associado ao mesmo, de modo a contemplar tanto aspectos mais globais quanto microestruturais. Para tanto, faz-se uma revisão da literatura sobre os estudos processuais de tradução, com o intuito de verificar os tipos de pesquisa realizados na área, os objetos de investigação e os principais modelos do processo tradutório. São abordadas, também, a estrutura do DP e as características da pré-modificação em inglês, com vistas a avaliar em que medida as diferenças estruturais entre as duas línguas acarretam dificuldades para a tradução desse tipo de sintagma e, consequentemente, custo de processamento. Dentre os fatores que podem acarretar custo, examinam-se, em especial, os atrelados à estrutura do DP, como a diferença de posição entre núcleos e modificadores em inglês e português – o que exigiria a aplicação de uma operação de conversão –, número e tipo de modificadores e a potencial ambiguidade estrutural de alguns tipos de DPs, podendo implicar reanálise. São verificados, também, custos associados à tarefa tradutória em um nível mais global, considerando-se a interrupção do fluxo tradutório, as modificações realizadas nos DPs e as fases em que elas ocorrem. Para verificar custo de processamento associado aos DPs, foram elaboradas três atividades experimentais, que fazem uso de técnicas utilizadas nos Estudos da Tradução (Translog e Camtasia) e também em Psicolinguística (PsyScope). A atividade aplicada a tradutores experientes pretende investigar o processo de tradução de DPs com múltipla modificação tendo em vista as pausas relacionadas à tradução da estrutura, os tipos de modificação realizados e as fases do processo tradutório em que ocorrem (redação ou revisão). Ademais, procura-se avaliar se o comportamento global dos participantes se repete no nível local, relativo à tradução dos DPs. As atividades aplicadas aos tradutores novatos buscam investigar a noção da diferença entre núcleos e modificadores em estruturas de DP com múltiplos modificadores e, também, a consideração de elementos de desambiguização, como posição do adjetivo e gênero dos modificadores, ao se avaliar a adequabilidade de uma tradução. Os resultados sugerem que o número de modificadores e tipo de modificador estão relacionados a custo de processamento da estrutura em inglês. Os resultados também mostram que os tradutores experientes e novatos conseguem identificar corretamente o núcleo da estrutura e utilizam estratégias de minimização de custo, como realizar modificações ao longo do fluxo do processo tradutório, optar por estruturas não ambíguas e que não causem algum tipo de estranhamento na língua-alvo. Percebe-se, portanto, uma atenção à organização da estrutura em português e uma preocupação em produzir e optar por construções menos custosas para o tradutor e o potencial público leitor. / [en] This thesis, which lies at the interface between Psycholinguisticsand Translation Studies, is aimed at investigating issues related to processing cost involved in the translation of complex DPs with multiple premodification (by two or three noun and adjective modifiers) in English. In addition, based on an interdisciplinary approach, this thesis is also aimed at characterizing the translation process based on evidences regarding DPs and cost related to it. Therefore, in order to verify the types of research conducted in the field of Translation Process Research, as well as the main translation process models, the literature on Translation Process Researchis reviewed. DP structures and the characteristics of premodification in English are also mentioned, so that it can be possible to evaluate the degree to which structural differences between the English and Portuguese languages cause difficulties concerning the translation of this type of phrase and, consequently, processing cost attributed to it. Among the factors that may lead to processing cost, we especially analyze cost related to the DP structure, such as the difference of head and modifier positions in English and Portuguese (making it necessary to switch the order of the items in the structure), the number and types of modifiers and the potential syntactic ambiguity present in some types of DPs, which could possibly lead to phrasereanalysis. We also verify cost related to the global translation process, considering the interruption of the translation process flow, the editing procedures applied to DPs and the phases of the translation process in which they take place. In order to analyze cost related to the DPs, three experiments were designed by employing the techniques used in Translation Process Research (Translog and Camtasia) and in Psycholinguistics research(PSyscope).The activity conducted with experienced translators is aimed at investigating the process involved in the translation of DPs with multiple modifiers regarding the pauses related to the translation of the structure, the types of editing procedures adopted and the phases of the translation process in which they occur (drafting or revision phase). The activities conducted with novice translators are aimed at evaluating if they are aware of the differences regarding heads and modifiers in DP structures with multiple modifiers in both languages and if, when evaluating the plausibility of a translation, they take into account morphosyntactic aspects related to the adjective position and gender of the modifiers in order to avoid possible ambiguity. The results suggest that the number and type of modifiers are related to processing cost involved in the DP structure in English. The results also show that the experienced and novice translators are able to correctly identify the head of the structure and make use of strategies in order to reduce cost, such as adopting editing procedures during the flow of translation process, choosing non-ambiguous structures in the Portuguese renderings or avoiding any option that could call the attention forbeing awkward. It is, therefore, possible to observe that the participants pay attention to the organization of the structure in Portuguese and they are also worried about rendering or choosing structures that are less cognitively demanding for the translator and the reader.

Le problème de représentation automatique de la signification logique des énoncés ambigus en langage naturel a suscité l'intérêt des chercheurs dans le domaine de la sémantique computationnelle et de la logique. L'ambiguïté dans le langage naturel peut se manifester au niveau lexical / syntaxique / sémantique de la construction de sens, ou elle peut être causée par d'autres facteurs tels que la grammaticalité et le manque de contexte dans lequel la phrase est effectivement prononcée. L'approche traditionnelle Montagovienne ainsi que ses extensions modernes ont tenté de capturer ce phénomène en fournissant quelques modèles qui permettent la génération automatique de formules logiques. Cependant, il existe un axe de recherche qui n'est pas encore profondément étudié: classer les interprétations d'énoncés ambigus en fonction des préférences réelles des utilisateurs de la langue. Ce manque suggère une nouvelle direction d'étude qui est partiellement explorée dans ce mémoire en modélisant des préférences de sens en alignement avec certaines des théories de performance préférentielles humaines bien étudiées disponibles dans la littérature linguistique et psycholinguistique.Afin d'atteindre cet objectif, nous suggérons d'utiliser / d'étendre les Grammaires catégorielles pour notre analyse syntaxique et les Réseaux catégoriels de preuve comme notre analyse syntaxique. Nous utilisons également le Lexique Génératif Montagovien pour dériver une formule logique multi-triée comme notre représentation de signification sémantique. Cela ouvrirait la voie à nos contributions à cinq volets, à savoir, (i) le classement de la portée du quantificateur multiple au moyen de l'opérateur epsilon de Hilbert sous-spécifié et des réseaux de preuve catégoriels; (ii) modéliser la gradation sémantique dans les phrases qui ont des coercitions implicites dans leurs significations. Nous utilisons un cadre appelé Montagovian Generative Lexicon. Notre tâche est d'introduire une procédure pour incorporer des types et des coercitions en utilisant des données lexicales produites par externalisation ouverte qui sont recueillies par un jeu sérieux appelé JeuxDeMots; (iii) l'introduction de nouvelles métriques sensibles au référent basées sur la localité pour mesurer la complexité linguistique au moyen de réseaux de preuve catégoriels; (iv) l'introduction d'algorithmes pour l'achèvement des phrases avec différentes mesures linguistiquement motivées pour sélectionner les meilleurs candidats; (v)l'intégration de différentes métriques de calcul pour les préférences de classement afin de faire d'elles un modèle unique. / The problem of automatic logical meaning representation for ambiguous natural language utterances has been the subject of interest among the researchers in the domain of computational and logical semantics. Ambiguity in natural language may be caused in lexical/syntactical/semantical level of the meaning construction or it may be caused by other factors such as ungrammaticality and lack of the context in which the sentence is actually uttered. The traditional Montagovian framework and the family of its modern extensions have tried to capture this phenomenon by providing some models that enable the automatic generation of logical formulas as the meaning representation. However, there is a line of research which is not profoundly investigated yet: to rank the interpretations of ambiguous utterances based on the real preferences of the language users. This gap suggests a new direction for study which is partially carried out in this dissertation by modeling meaning preferences in alignment with some of the well-studied human preferential performance theories available in the linguistics and psycholinguistics literature.In order to fulfill this goal, we suggest to use/extend Categorial Grammars for our syntactical analysis and Categorial Proof Nets as our syntactic parse. We also use Montagovian Generative Lexicon for deriving multi-sorted logical formula as our semantical meaning representation. This would pave the way for our five-folded contributions, namely, (i) ranking the multiple-quantifier scoping by means of underspecified Hilbert's epsilon operator and categorial proof nets; (ii) modeling the semantic gradience in sentences that have implicit coercions in their meanings. We use a framework called Montagovian Generative Lexicon. Our task is introducing a procedure for incorporating types and coercions using crowd-sourced lexical data that is gathered by a serious game called JeuxDeMots; (iii) introducing a new locality-based referent-sensitive metrics for measuring linguistic complexity by means of Categorial Proof Nets; (iv) introducing algorithms for sentence completions with different linguistically motivated metrics to select the best candidates; (v) and finally integration of different computational metrics for ranking preferences in order to make them a unique model.

[pt] O processamento da concordância de número sujeito-verbo na produção de sentenças por falantes do dialeto culto do português brasileiro é investigado. A dissertação focaliza os chamados erros de atração e seu principal objetivo é identificar os fatores que interferem no processamento da concordância e prover uma explicação psicolingüística que seja compatível com pressupostos do Programa Minimalista da Lingüística Gerativa. Mais especificamente, busca-se examinar: i) as condições que favorecem os erros e as propriedades sintáticas que levam um núcleo interveniente a ser tomado como o controlador da concordância; ii) a interferência de informação morfofonológica de número dos elementos que integram os modificadores do DP sujeito; iii) a interferência de informação semântica de número no estabelecimento da concordância. Ainda como objetivo específico busca-se distinguir em termos estruturais os DPs responsáveis pelos erros de atração daqueles que licenciam uma forma singular ou plural do verbo - as chamadas construções partitivas. Aplica-se uma tarefa psicolingüística envolvendo julgamento de gramaticalidade a fim de investigar diferenças de processamento da concordância entre as construções partitivas e os DPs complexos. A relevância dos tópicos investigados se deve ao fato de estes permitirem uma discussão mais ampla acerca da autonomia do formulador sintático. Parte-se de vasta revisão da literatura, na qual se reportam interferências sintáticas, semânticas e morfofonológicas no processamento da concordância em diferentes línguas. Explicações apresentadas por modelos de produção interativos e não-interativos são discutidas. Inclui-se ainda uma caracterização da concepção minimalista de língua, com o tratamento da concordância como processo de valoração de traços formais, e um modelo de produção de natureza serial, não-interativo, que incorpora um parser- monitorador funcionando paralelamente à formulação dos enunciados - modelo PMP (produção monitorada por parser). Em seguida, reportam-se 5 experimentos com falantes de português. Os resultados indicaram efeito de marcação e de distância linear entre o núcleo do sujeito e o verbo, com mais erros para núcleo do sujeito não-marcado (singular) e linearmente distante do verbo, e efeito de posição estrutural do núcleo interveniente, com maior incidência de erros para os núcleos hierarquicamente próximos do nó mais alto do DP sujeito e núcleos inseridos em PPs argumentos. Um efeito semântico de distributividade associado a efeito de marcação também foi obtido. Quanto a fatores morfofonológicos, a informação de número no determinante (e não no nome) mostrou-se crucial para a identificação do número do DP sujeito. É proposta uma versão ampliada e revista do modelo PMP que unifica explicações para os erros de concordância em termos de uma escala de acessibilidade da representação do DP sujeito pela memória de trabalho e que leva em consideração as expectativas do parser como possível fator de interferência em erros de atração. Essa interferência ocorreria após o parsing do primeiro DP e afetaria a codificação morfofonológica do verbo. Em suma, a tese aqui veiculada é a de que os erros de concordância não ocorrem na computação sintática e que o formulador sintático atua de forma autônoma. / [en] The processing of subject-verb number agreement in sentence production by speakers of the standard dialect of the Brazilian Portuguese is investigated. The dissertation focuses on attraction errors and its main aim is to identify the factors that interfere in agreement processing and to provide a psycholinguistic account, which is compatible with assumptions of the Minimalist Program of Generative Linguistics. This work examines, in particular: i) the conditions that favor attraction errors and the syntactic properties which make an intervenient head to be considered as the agreement controller; ii) the role of morphophonological information on number provided by the DP subject modifiers; iii) the interference of conceptual number in agreement. Additionally, the work intends to distinguish, in structural terms, the DPs responsible for attraction errors from those DPs that allow a singular and a plural form of the verb - the so-called partitive constructions. A psycholinguistic procedure of grammaticality judgment is conducted in order to verify agreement processing differences between partitive constructions and complex DPS. The relevance of these topics is due to the fact that they enable a more comprehensive discussion on the autonomy of the syntactic formulator in language production. An extensive review of the existing literature was carried out and results are reported, concerning the interference from syntactic, semantic and morphophonological factors on agreement processing in different languages. Explanations provided by interactive and non-interactive models are discussed in this work. The minimalist conception of language is presented according to which agreement is described as a feature valuation process and a serial non-interactive production model is characterized, which incorporates a monitoring-parser that works in parallel with speech formulation - PMP model (parser monitored production). A total of 5 experiments with Portuguese speakers are reported. The results show an effect of markedness and linear distance between the subject and the verb, with more errors caused by non-marked (singular) subject heads that are linearly distant from the verb, and an effect of the structural position of the intervenient head, with a large number of errors for intervenient heads that are near to the upper phrasal marker of the DP and for heads which are inserted in PP arguments. A semantic effect of distributivity associated with an effect of markedness was also obtained. As far as morphphonological factors, number information of the determiner (and not of the noun) has shown to be critical to subject number identification. A revised and improved version of the PMP model is proposed that unifies possible explanations for agreement in terms of on an accessibility scale of the DP subject representation in the working memory and that takes into account parser predictions as a possible factor of interference in attraction errors. This interference would occur after the parsing of the DP and would affect the morphophonological encoding of the verb. In sum, the main argument of the thesis is that agreement errors do not occur in the syntactic computation and that the syntactic formulator works autonomously.

大多數台灣英語學習者在進入大學前已經習得相當數量的英文字彙，即便如此，他們仍然會誤用常見的搭配詞 (例如: [do/make+noun])。本論文藉用兩種語料庫 (分別為台灣英語學習者語料庫及英國國家語料庫)來分析、觀察[do+noun]和[make+noun]的異同以及英語母語人士及台灣英語學習者使用[do/make+noun]狀況。結果顯示台灣英語學習者和英語母語人士最大的不同在[make+noun]：就語意層面而言，最常被英語母語人士使用的[make+noun]為 ‘to perform, to carry out’ (例如: make a speech, make a fine judge, etc.) 而台灣英語學習者偏好 ‘to create’ (例如: make a sushi, make a robot, etc.)；就名詞特性而言，母語人士偏向使用抽象的名詞 (例如: comment, progress, etc.) 而學習者習慣使用具體的名詞 (例如: robot, sushi, etc.)。除了語料庫語料分析，本論文還透過心理語言學實驗測驗 (即圖片引述實驗－受試者描述他們不熟悉的動作) 觀察母語人士和學習者使用常見的搭配詞－[do/make+noun]－的差異。台灣英語學習者使用為數不少廣義的[do+noun] (例如：do exercise 或 do sports) 而英語母語人士傾向使用帶有具體意義的動詞 (例如:sit-up) 或搭配詞 (例如: do sit-up)。幾乎沒有母語人士使用[make+noun]而大多數學習者使用的是[make+noun]－當make做為causative的用法。根據此實驗分析，本論文提出一個模型來探討英語母語人士和英語學習者對[do/make+noun]的使用異同。 / Learners of English in Taiwan are estimated to reach a certain command of vocabulary size before they enter colleges. However, they still differ from native speakers in producing the commonly-used patterns, such as [do/make+noun]. In order to observe the similarities and differences of [do+noun] and [make+noun], as well as their uses by EFL learners, this paper inspects their senses using two types of corpus data, namely a Taiwan-based learner corpus and the British National Corpus. The results show that learners differ from native speakers mainly in their use of [make+noun]. For example, the most frequent sense used by native speakers is ‘to perform, to carry out,’ as in make a speech, make a fine judge, etc., whereas that used by Taiwanese learners is ‘to create’ as in make a sushi, make a robot, etc. With respect to the characteristic of the noun following make, native speakers tend to choose abstract nouns, such as comment, progress, etc., whereas learners prefer concrete nouns, such as robot, sushi, etc. A psycholinguistic experiment is also included in order to see whether learners use language with general meanings, such as [do/make+noun], more in describing situations unfamiliar to them. Results show that [do+noun] patterns with a more general meaning (e.g., do exercise or do sports) are more often used by the learners in our experiment while native speakers prefer language with a more precise meaning (e.g., sit-up or do sit-up). Few [make+noun] constructions are found in native speakers’ language whereas learners produce numerous [make+noun] constructions, mostly the causative uses of make.

Les unités phraséologiques représentent actuellement un objet d’étude commun à de nombreuses disciplines en science du langage. C’est dans ce cadre que s’inscrivent les travaux exposés dans la présente thèse. Adoptant une optique pluridisciplinaire alliant linguistique et psycholinguistique et impliquant l’approche comparative interlangue, nous étudions un phénomène linguistique complexe, le figement, en français et arabe dialectal tunisien. Nous proposons, en premier lieu, une description des expressions figées dans les deux langues mises en contraste, en repérant la typologie et le fonctionnement morphosyntaxique, lexical et sémantique de ces séquences. Notre attention se porte essentiellement sur les deux notions importantes dans l’étude du figement: la scalarité et l’iconicité. La méthodologie et les résultats d’une étude psycholinguistique et expérimentale réalisée auprès d’adultes francophones natifs et non natifs (des arabophones tunisiens apprenant le français L2) sont, en second lieu, exposés. L’intérêt est de déterminer si les facteurs linguistiques, degré de figement et iconicité, facilitent le traitement du langage figé au cours d’une tâche de mémorisation impliquant l’encodage et la reconnaissance des séquences figées du français. Une troisième expérience consiste en un test de familiarité qui examine l’effet des facteurs linguistiques, la scalarité et l’iconicité, ainsi que personnels, l’âge et le sexe, sur la connaissance des expressions figées du dialectal tunisien par des locuteurs arabophones natifs. Ces recherches psycholinguistiques, portant sur une population d’adultes, sont suivies d’une étude exploratoire sur l’émergence des expressions figées ou semi-figées chez les enfants. Ce travail suit la double problématique de l’émergence et de la comparaison interlangue. Nous comparons les productions de deux enfants: un enfant arabe tunisien et un enfant français. / The phraseological units currently represent a common object of study in many disciplines in language science. It is in this context that we present our work in this thesis. Adopting a multidisciplinary approach combining linguistic and psycholinguistic involving comparative cross-language approach, our goal is to study a complex linguistic phenomenon, the fixation, in French and Tunisian Arabic dialect. We propose, first, to contrast the fixed expressions in both languages by identifying the type and the morphosyntactic, lexical and semantic operation of these sequences. Our interest will focus on two important concepts in the study of the fixed sequences: scalarity and iconicity. The methodology and results of a psycholinguistic and experimental study among native and non-native francophone adults (Tunisian Arabic speakers learning French as second language) will be then exposed. Our aim is to determine whether language factors, scalarity and iconicity, facilitate the processing of fixed expressions during a memorization task involving the encoding and recognition of French frozen sequences. A third experiment is a test of familiarity that examines the effect of linguistic factors, scalarity and iconicity as well as personal factors such as; age and sex, on the knowledge of Tunisian dialect frozen sequences by Arabic native speakers. This psycholinguistic work carried on an adult population is followed by an exploratory study on the emergence of fixed and semi-fixed expressions in children communication. This work follows the dual problem of emergence and cross-language comparison. We compare the productions of two children: a Tunisian Arab child and a French child.