A lot of what humans communicate about concerns cognitive contents of various kinds produced by others or themselves: speech, thought, writing, emotional states, attitudes, hopes, and the like. Languages have developed specialized ways to structure the representation of such contents, especially in various dedicated forms of speech and thought representation. Represented content can also include embodied behavior, such as gesture, whether in cospeech gesture or in sign language. What is represented need not actually have been previously produced: represented contents can be future, hypothetical or nonexistent, and forms of so-called fictive interaction can be used in which the model of face-to-face interaction is used to talk about a variety of other meaning types. Speech and thought representation presupposes the existence of two speech events—a current and a represented one—and each comes with a speaker, defined linguistically in terms of their central deictic coordinates, I—here—now. The interplay of deictic features and different forms of structural integration can define specific types of construction, showing different degrees of access to the embedded mental space of the represented speaker's speech or thought, such as direct, indirect, and free indirect speech or thought. Social media forms of direct speech or thought merit separate investigation, as do subjective uses of reporting clauses such as I think, which form a distinct construction type, using a subset of the grammar of speech and thought representation for different purposes.