Anderson, J. R. (1978). Arguments concerning representations for mental imagery. Psychological Review, 85, 249-277.

The argument for a propositional over a pictorial representation for visual imagery has largely taken the form of an attack on the logical coherence of pictorial representations. These attacks have not been valid; one can develop a coherent dual-code model involving pictorial and verbal (nonpropositional) representations. On the other hand, empirical demonstrations that are claimed to support pictorial representations fail to discriminate such representations from propositional ones. It is argued that the failure of the antipictorial and the propictorial arguments stems from a fundamental indeterminancy in deciding issues of representations. It is shown that wide classes of different representations, and in particular propositional versus dual-code models, can be made to yield identical behavioral predictions. Criteria such as parsimony and efficiency in addition to prediction of behavior may yield further constraints on representation; and, in particular, it may be possible to establish whether there are two codes, one for visual information and one for verbal, or whether there is a single abstract code. However, the conclusion of this article is that barring decisive physiological data, it will not be possible to establish whether an internal representation is pictorial or propositional.