Cette thèse travaille sur l’élucidation du sens poétique à l’aide d’outils linguistiques ; elle propose une définition cognitive, phénoménologique et énonciative de la distinction abstrait/concret à partir de statistiques établies sur l’œuvre de John Donne (1572-1631), Meditations upon Emergent Occasions et The Complete English Poems. Le concret résulterait d’une forme de lyrisme indirect, c’est-à-dire non sémiotisé, implicite, auquel le poète aurait recours lorsqu’il serait impliqué affectivement dans une situation de discours. L’expressivité du sujet parlant repose sur une décatégorisation momentanée de catégories disponibles en langue, décatégorisation qui lui permet de dénoncer (implicitement) l’insuffisance des cadres abstraits prévus pour tous pour désigner la singularité de son expérience sentimentale. Les noms en –ness constituent une métaphore grammaticale car ils résultent d’une décatégorisation grammaticale, tandis que les métaphores traditionnelles mettent en jeu une décatégorisation lexicale. La métaphore porte l’empreinte affective du sujet parlant, qui se réapproprie ainsi le langage. La poésie métaphysique de John Donne, oscillant entre métaphores et comparaisons, entre le concret et l’abstrait, apparaît alors comme l’expression d’un travail du sentiment, sentiment d’autant plus travaillé qu’il est religieux, et se construit en l’absence de l’être aimé. / This dissertation studies poetic meaning using linguistic tools. It offers a cognitive, phenomenological and enunciative definition of the distinction between the abstract and the concrete, based on statistics carried out on work of the metaphysical poet, John Donne (1572-1631): Meditations upon Emergent Occasions and The Complete English Poems. I argue that the “concrete” is the result of indirect – implied, unsemiotized – lyricism, a form of lyricism used by the poet when s/he is emotionally implicated in a speech situation. The speaker’s expressivity relies on a temporal decategorization enabling him to (implicitly) claim that generalized (abstract) terms are insufficient to articulate the specificity of his own sentimental experience. Words in –ness – grammatical metaphors – result from a grammatical decategorization, while traditional metaphors derive from a lexical decategorization. Metaphors reflect the affect of the incarnate speaker, who thus repossesses language. Lastly, I understand John Donne’s poetry – hinging on metaphors and comparisons, concrete and abstract elements – as expressing a working of feelings, which is the strongest when the feeling is religious and needs to make up for the absence of the beloved.

過往研究指出當個體受到社會排斥威脅後，可能產生兩種相反的行為，一者是具攻擊性的行為，而另一者則是希冀與其他人連結之親和傾向。本研究提出社會排斥之兩階段模式，認為當個體受到排斥威脅時，會同時有兩種不同的行為反應，一個是『避免傷害』，當此反應被激發時，被排斥個體會展現出『戰或逃』的行為傾向；而另一個反應則為『尋求歸屬』，在此反應被激發時，被排斥個體則會展現出『親近與示好』的行為傾向。而個體會出現何種反應，端視其之後互動的對象本身的屬性而定。若後續互動對象與先前排斥者間具有高相似性，此拒絕相似線索會引發個體展現出較高的戰或逃行為。相對的，若是後續互動對象所具有的特性與其自身之特性具有高相似性，亦即具有接納相似線索時，個體則會出現較多的示好行為。而這種隨著後續屬性，個體會出現不同反應歷程的假設在已完成的四個實驗中獲得初步的驗證。 實驗一（83名實驗參與者）採取2 (社會排斥：有、無) × 5 (類別相似線索：高拒絕線索vs.中拒絕線索 vs.高接納線索 vs.中接納線索 vs. 中性線索)之混合設計，結果發現受社會排斥威脅者會有較高意願加入帶有高可能接受線索之團體、較不願意加入帶有高拒絕線索之團體。實驗二（47名實驗參與者）採取閾下觸發來操弄社會排斥威脅，顯示社會排斥的確會讓人加速處理那些與拒絕者與可能接受者相關之訊息。而實驗三（74名實驗參與者）重複驗證了實驗一之發現，並且發現受社會排斥威脅之參與者在反應時間上會較快決定拒絕帶有高拒絕線索之團體，也會較快決定加入帶有高接納線索的團體（兩者皆對比於對中性線索團體之決策反應時間）。 實驗四（75名實驗參與者）則採取2 (社會排斥：有、無) × 2 (認知負荷：高、低) × 3 (類別相似線索：高拒絕線索vs.高接納線索vs. 中性線索)之混合設計，並發現當被社會排斥者處於高認知負荷狀態時，他們雖然還是能使用拒絕線索來選擇團體，但卻不會使用接受線索，顯示拒絕相似性線索為較為優先之線索。 實驗五A（67名實驗參與者）則採取2 (社會排斥：有、無) × 2 (性格回饋：有、無) × 4 (類別相似線索：拒絕線索對象vs.接納線索對象vs. 中性線索對象vs.原拒絕者)之混合設計，並發現當被給予性格回饋之後，受社會排斥威脅者會對於帶有接納線索的對象有較高的評價。而實驗五B（31名實驗參與者）採取2 (社會排斥：有、無) × 4 (類別相似線索：拒絕線索對象vs.接納線索對象vs. 中性線索對象vs.原拒絕者)之混合設計，發現受到社會排斥威脅者會對於帶有拒絕線索的對象及原拒絕者都會有較低的評價。 而實驗六與七則認為當個體對於相似性線索的認知被去、再類別化作業改變時，個體的反應也應隨之改變。實驗六採取（51名實驗參與者）則採取3 (社會排斥組別：社會排斥一般作業組、社會排斥再類別化組、控制組) × 4 (類別相似線索：拒絕線索對象vs.接納線索對象vs. 中性線索對象vs.原拒絕者)之混合設計，並發現當受到社會排斥者完成再類別化作業後，比起完成一般作業的參與者，他們對於原先帶有拒絕線索對象的評價會提升，而他們亦會提升對於原拒絕者的評價，顯示再類別化作業可以讓受威脅者感知到互動對象跟自己可以共屬另一個有意義之團體、自己與互動對象是相連結的，因而提升對這些對象之評價。 實驗七採取（46名實驗參與者）則採取3 (社會排斥組別：社會排斥一般作業組、社會排斥去類別化組、控制組) × 4 (類別相似線索：拒絕線索對象vs.接納線索對象vs. 中性線索對象vs.原拒絕者)之混合設計，並發現去類別化作業可以削弱類別相似性線索，因此使得受排斥者降低對原先帶有接納線索之對象的評價，亦會提升原先帶有拒絕線索對象之評價。換言之，由於去類別化作用會降低被威脅個體與互動對象的連結感，亦即他們對於互動對象與自己的相似性知覺會降低，個體因而降低對此類對象之評價；另一方面，去類別化作用也會降低帶有拒絕線索對象以及原拒絕者的相似性知覺，因而提升對於此類對象之評價。 綜合以上所述，本研究以一系列之研究來檢驗：個體受到社會排斥後，後續互動對象身上之類別相似性線索如何影響受威脅者之反應傾向。實驗一至五發現，當個體被社會排斥之後，他們對於那些帶有拒絕線索的個體會展現出戰或逃的行為傾向，並對帶有接納線索的個體展現親近與示好之行為，而拒絕相似性線索較為優先。而實驗六及七分別使用『再類別化』及『去類別化』兩種作業來改變類別相似性線索，來改變受社會排斥威脅者對於接納線索的知覺使其後續行為隨之改變。 / Previous research shows that social exclusion may cause either fight-or-flight behavior toward innocent people or a substantial increase in affiliation behavior. A two-stage model is proposed to explain people’s reactions after experiencing social exclusion. For these threatened individuals, groups that resemble their rejecter would provoke a defensive attitude and hostile behaviors. These individuals also show hospitality to those groups that resemble themselves to gain inclusion. The former process is more primary than the other. Experiment 1 (83 participants) and 3 (74 participants) used a scenario story to manipulate social exclusion. Groups with different categorical cues categorical similarity cues were provided. The results suggested that people after experiencing social exclusion tended to reject groups that resemble to previous rejecter. By contrast, groups that fit the characteristics of the target person were much preferred and received more resources. The second experiment (47 participants) used subliming priming to manipulate social exclusion. Main dependent measure was the decision time of lexical decision task. As predicted, compared to participants in control condition (all priming words in random order), those in social exclusion condition (being primed with rejection-softball and inclusion-enjoy-learning) processed rejection associated words (related to the reject group) and acceptance associated words (related to the perspective group) significantly faster than no association words. Experiment 4 (75 participants) used a 2 (social exclusion vs. control condition) x 2 (cognitive load: heavy vs. light) x 3 (categorical similarity cues: high rejective cue, high perspective cue, irrelevant cue) mixed-design. The results indicated that the excluded participants who had a heavy cognitive load could only avoid groups that resemble the previous rejecter and showed no preference for groups that might be suitable for them. To defend is the primary process. From experiment 5A to experiment 7, cyberostracism was employed to manipulate social exclusion. Experiment 5A (67 participants) used a 2 (social exclusion: social exclusion condition vs. control condition) x 2 (personality feedback: feedback vs. no feedback) x 4 (categorical similarity cues: rejective cue, perspective cue, irrelevant cue, rejecter) mixed-design. It was found that only those who with perspective cue could receive better evaluation from excluded participants. Experiment 5B suggested that excluded participants would evaluate people with rejective cues worse than those who in control group. The experiment 6 and 7 used decategorization and recategorization task to change the perceptions of categorical similarity cues. Experiment 6 (51 participants) showed that recategorization task could weaken the effect of the rejective cue and thus excluded participants would give better evaluation to people with rejective cue than participants in social exclusion only condition. Experiment 7 (46 participants) suggested that decategorization task might weaken the effect of perspective cue and therefore the evaluation tended to decrease for people with perspective cue. Eight experiments provided convergent evidence to this study to suggest that categorical similarity cues of possible affiliated people could elicit different reactions from excluded people.