"[The long-standing debate about the nature of visual mental imagery]…hinges on the distinction between the format versus the content of images: The format is the type of code, whereas the content is the information conveyed. The same content can be conveyed using many different formats; for example, the information in this sentence (i.e., the content) could be conveyed using the dots and dashes of Morse code, using the words and grammar of French, speaking English aloud, and so on (each of which is a different format). The debate focuses on the format of the representations that give rise to the experience of imagery, with one side arguing that at least some of these representations preserve key aspects of pictures (e.g., Kosslyn, 1980) and the other arguing that these representations are in no sense pictorial but rather rely on abstract symbols of the sort used in language (e.g., Pylyshyn, 1973, 2002, 2003).

This debate arose at the very beginning of scientific psychology (for a review of the history of the debate, see Kosslyn, 1994; Tye, 1991). For example, John B. Watson (1913), the founder of behaviorism, rejected the idea that picturelike mental images exist; instead, he claimed that subtle movements of the larynx accompany imagery and that these movements are all that is important. In contrast, others, such as William James (1890), argued that mental images are in some sense pictorial. More recently, the debate was recast by Pylyshyn (1973) in terms of “propositional” representations, which are abstract, language-like symbolic representations that capture the meaning of an assertion. According to Pylyshyn (1973, 1981, 2002, 2003), Anderson and Bower (1973), and others, all cognitive processes rely exclusively on propositional representations very much like those used in language. In contrast, others have argued that—remarkably—introspection reveals properties of the representations used in information processing (e.g., Kosslyn, 1980, 1994; Kosslyn & Pomerantz, 1977). According to this view, representations that ultimately give rise to the experience of imagery are actual images—insofar as each part of the representation corresponds to part of the represented pattern, and the distances among the representations of the parts correspond to the distances among the parts themselves. Such representations are not descriptions but rather are geometric patterns that depict the appearance of objects. The issue is not whether propositional representations are used in imagery—in some processes they probably do play a role (e.g., stored descriptions may well be used to compose images of objects into scenes; see Kosslyn, 1980, 1994). Rather, the issue is whether only propositional representations are used in imagery or whether depictive representations also play a role.

Following many years of behavioral research, this debate reached an impasse: The results from these studies could be explained by theories that posit only propositional representations or theories that also posit depictive representations (see, e.g., Anderson, 1978; Kosslyn, 1980). Modern neuroimaging techniques now provide an opportunity to press this issue forward."

When is early visual cortex activated during visual mental imagery?, Psychological Bulletin. 2003 Sep Vol 129(5) 723-746, p.723-724.