One of the most important hallmarks of human language is the use of symbolic signs in that words just “symbolically” represent the objects they refer to. That is, neither the written word ‘tree’ nor the spoken sequence of sounds /triː/ has any inherent similarity or analogy to a real tree. This is a core assumption of classic linguistics, which states that words are paired with objects, mental images, or concepts in an arbitrary fashion. Thus, the sound of a word per se has no inherent semantic content, nor does it play any contributing role in shaping the meaning of words. The goal of this dissertation is to contemplate, to rethink, and to examine the aforementioned statement, which has dominated language research throughout the last century. Indeed, there are a vast number of counterexamples showing how meaningful single phonemes, or their combination—as in nonwords—can be. Consider the standalone role of sound in poetry, or the use of single syllables or phonemes in various sacred rituals, or the prevalence of onomatopoetic words, i.e., words that sound similar to what they mean (e.g., click, zigzag), across all of the languages in the world, or the tendency of language users toward using harsh-sounding words as swear words, or cross-linguistic phenomena such as a preference to match the nonword BOUBA with a curvy round shape and KIKI with a spiky angular shape: These all constitute excellent examples that could potentially falsify the radical assumption of the sound of words being per se meaningless. The present dissertation now wants to shed new empirical light on this old debate, which dates back to Greek antiquity, yet faces a number of unanswered questions regarding the cognitive and neuropsychological mechanisms underlying the potential effect of sound on the processes of meaning making. More specifically, my focus is on the existence of sound-meaning relationships in the affective domain, termed affective iconicity, and on the investigation of different aspects of this phenomenon in both everyday language and poetry. By taking an interdisciplinary approach, the present work combines a variety of methods and techniques, such as behavioral and neuroimaging experiments, phonological and acoustic analysis of a large-scale lexicon, computational modelling, and corpus analysis, in order to provide comprehensive answers for this multi-faceted phenomenon. The theoretical part of the dissertation explores and integrates linguistic perspectives on the iconic mappings, explaining how a linguistic sign can acquire meaning based on similarities between its form and the object it refers to. A surprising neglect of the role of emotion in empirical models of language in general, and in previous investigations of iconicity in particular, is discussed. Specific hypotheses are formulated via the predictions made by the recently proposed Neurocognitive Poetics Model (NCPM), and through reviewing the previous works on the topic. Accordingly, three main questions are formulated that the present dissertation aims to address: i) Does the sound of words evoke affective responses observable at the behavioral and neural level? ii) Does the sound of words influence the processes of meaning making in the affective domain? iii) Does the sound of words in a poem contribute to its global affective meaning as perceived by readers? Six empirical studies attempt to address these questions which are subdivided into six more precise research questions. Results of the empirical part provide a comprehensive picture of the interplay between sound and meaning at different levels of processing (i.e., rating, semantic decision, and passive listening) for different presentation modalities (i.e., visual, and auditory) and for different textual levels (i.e., single word, and entire text). In short, results of Study 1 and Study 2 indicated a high similarity between the affective potential of the sound of words and other types of affective sounds (e.g., nonverbal emotional vocalization and affective prosody) at both the level of psychological perception (Study 1) and the level of neural correlates and substrates (Study 2). Furthermore, when giving their affective judgments (valence and arousal) about the meaning of words, participants, as shown in Study 1, were implicitly influenced by the sound of words even when words were presented visually and read silently. These results were extended in Study 3 in which iconic words, as operationalized by congruence between affective sound and affective meaning, were evaluated more quickly and more accurately than their non-iconic counterparts, suggesting that a similarity between the form and meaning of a word may help language users to more readily access its meaning through direct form-meaning mappings. Study 4 investigated the neural mechanisms underlying the facilitative effect observed in Study 3. Results showed an enhanced fMRI signal in the left amygdala, known for its role in multimodal emotion integration, for both a comparison between iconic and non-iconic words, as well as functional connectivity between two seed regions representing the sound (superior temporal gyrus) and meaning (inferior frontal gyrus) of words modulated by iconic condition. Lastly, results of Study 5 and Study 6 emphasize the role of foregrounded phonological units in the affective and aesthetic processes of literary reading. This clearly supports the initial hypotheses that iconicity is a feasible indicator of the affective qualities of a literary text as evoked by particular phonemic structures. The presented method for measuring the basic affective tone of the poems investigated could account for a considerable part of the variance in the ratings of their general affective meaning. In summary, this dissertation provides strong psychological and neuroimaging evidence for a device that has long been deployed in poetry and the arts, i.e., evoking affective (and aesthetic) responses by the use of certain words with specific sound patterns. The results were used to upgrade the standard models of visual word processing by conceiving corresponding modules responsible for the evaluation of affective sound and its interaction with the evaluation of the affective meaning of words. Lastly, at the more complex level of the whole text, the findings of this dissertation confirm the central assumption of the NCPM regarding the role of foregrounded elements in enhancing affective perception, although the Panksepp-Jakopson hypothesis might need to be extended to human-specific brain regions which originally evolved for other, more simple, tasks. Also, the literary model of reading may need to be updated by adding feedback loops from resulting reading behavior (e.g., fluent reading) to the perceived emotions (e.g., lust and play) based on the findings concerning the facilitative role of iconicity in language processing.