This paper addresses the surprisingly difficult task of defining mental imagery. It argues against recent influential definitions, particularly those by Bence Nanay (2023) and Margherita Arcangeli (2020), and proposes an alternative definition that avoids their limitations.

Drawing on Arcangeli's approach, I begin by distinguishing the varied uses of “mental imagery” in philosophical and scientific discourse. Psychological research refers to mental imagery in multiple ways, including afterimages, eidetic imagery, memory imagery, and imagination imagery (Richardson, 1969; Finke, 1989). By contrast, philosophers have used mental imagery to refer to phenomenal experience, neural representations, or intentional objects (Block, 1981; Arcangeli, 2020). These diverse uses reveal conceptual ambiguity, requiring a clearer theoretical framework.

I identify five key meanings in the various uses of mental imagery: (a) a phenomenal experience, (b) a format for mental representations, (c) a memory process, (d) an imaginative process, and (e) the persistence of a visual representation after a stimulus disappears. However, I argue that those meanings can be reduced to three core referents relevant to craft a definition: (a') phenomenal experience, (b') a representational format, and (c') a mental/neural process. I then challenge the view that phenomenal experience is necessary or sufficient for defining mental imagery, though it often accompanies it. I thus propose the following definition:

Mental imagery is (i) a top-down, quasi-perceptual process (ii) producing analog/depictive mental representations.

This definition captures both the processual and representational aspects of mental imagery, distinguishing it from perception while maintaining coherence with cognitive science. By refining the concept, this paper contributes to greater clarity in the theoretical and empirical literature on mental imagery and imagination.