The mechanisms by which the mind encodes meaning into words and reconstructs it from them has been the subject of philosophical speculations at least since Plato and Aristotle in the 4th century B.C. Our current understanding of how the brain is involved in these processes, however, only started in the 19 ^th century, with precise descriptions of language impairments following brain injury. More recently, the development of new, non-invasive techniques for studying brain function has provided a wealth of information about the neural and cognitive underpinnings of the language capacity. The present work contributes to this line of inquiry, focusing on two main questions: (a) how is meaning represented in the brain; and (b) how does the brain create sentence-level meaning from individual words. The dissertation reviews the relevant literature on these topics and points out limitations with our current understanding of them. It also reports four experiments that aim at providing answers to these problems and indicating new possible frameworks for the investigation of the relationship between meaning and the brain. Chapters 1 and 2 summarize the currently prevailing views on the relationship between language, action representation, and the brain. Chapter 3 reports an fMRI study that investigated the organization of action representations in the human premotor cortex, as well as the possible role of these representations in processing language meaning. We found that observation of different categories of actions involving the hand led to the activation of distinct portions of the premotor cortex, in agreement with electrical stimulation studies of the premotor cortex in the monkey. These results indicate that specific actions—not only body parts—play a role in neuroimaging studies of embodied semantics. Finally, chapter 4 describes a series of behavioral and ERP studies on the processes involved in compositional meaning and their time courses, whose findings challenge sequential models of sentence comprehension, supporting a parallel processing view.