Introduction. The paper claims that each thought experiment, despite of its particular subject, reveals the general principles of the work of the imagination – these are elements of the transcendental scheme, which I. Kant controversially defines as pure representation, that is both intellectual and sensory. In this sense, imagination is the meta-subject of a thought experiment. Theoretical analysis. With the help of thought experiments operating marginal subjects, the necessary synthesis schemes related to fundamental concepts are clarified. The connection between a thought experiment and imagination is obvious and is considered trivial – as R. Sorensen notes, in a thought experiment «imagination substitutes for perception» and the reliability of the result depends on how well imagination «can fill a role originally intended for perception». At the same time, the ambiguity in the interpretation of imagination, characteristic for the intellectual culture of modern times, is directly reflected in the assessment of the epistemological status of a thought experiment closely associated with it. Imagination is understood, on the one hand, as something opposite to reason – the source of chimeras and delusions; on the other hand, imagination is a fundamental condition for cognition, since outside the synthesis, from which any subject of knowledge begins, cognitive activity is generally impossible, and “to know” means “to be able to visualise”, i.e. to imagine. The structure of this ability precisely reveals a thought experiment in which non-realistic premises are used to demonstrate the work of transcendental schematism, i.e. the necessary method of synthesis - just as a map reflects not only the terrain (which might not exist), but also the cartographer’s way of making his job (which will necessarily be depicted). Conclusion. In addition to the subject, which is the obvious goal of a specific thought experiment, it reveals a subject that is invariant for all thought experiments – the principles of the imagination.