The Homogeneity of the Imagination

Luke Roelofs is a postdoctoral researcher at Ruhr-University Bochum, working on social cognition, empathy, and the metaphysics of consciousness. His book ‘Combining Minds’, dealing with the combination problem for panpsychism and related topics, is under contract with OUP. 

Luke Roelofs is a postdoctoral researcher at Ruhr-University Bochum, working on social cognition, empathy, and the metaphysics of consciousness. His book ‘Combining Minds’, dealing with the combination problem for panpsychism and related topics, is under contract with OUP. 

A post by Luke Roelofs.

In ‘The Heterogeneity of the Imagination’, Amy Kind argues that no single mental faculty can do all the work imagination has been assigned by philosophers. I can’t address every point Kind makes, but I’ll sketch a case for the homogeneity of the ‘imagination’ appealed to in the four contexts she focuses on.

1. The Case for Heterogeneity

Kind’s case for heterogeneity targets the ‘simulationist’ account of imagination, on which ‘imagination’ is the capacity to create simulations of other mental states. To visualise something, we make a simulation of a visual perception; to suppose something, we make a simulation of a belief; and so on.

Kind argues that the way simulationism conceives ‘imagination’ (tailored primarily to explaining mindreading) directly conflicts with what’s needed to explain imagination’s role in modal epistemology, engagement with fiction, and pretence. Consider three more specific problems she mentions:     

Problem 1: Modal Epistemology and the Supposition-Imagination Distinction

First, simulationism treats supposition as a type of imagination, namely the simulation of belief. But in modal epistemology it is crucial that supposing something and imagining it are different activities, since (roughly) we must be able to suppose impossible things (e.g. to prove their impossibility by reductio) but not to imagine them (since the possibility of imagining something is prima facie evidence of its possibility). We must distinguish round squares (unimaginable but easily supposed) from flying pigs (imaginable, though not actual). Otherwise “we sever the evidentiary connection between imagination and possibility.” (151)

Problem 2: Engagement with Fiction and Imaginative Affect

Second, simulationism assimilates hypothetical suppositions and engaged-with fictions. But there seems to be a difference in the affective responses they produce: “when I imagine Frank Jackson’s Mary… I do not experience any anger or indignation about her imprisonment.” (153) If we can imagine tragic situations without affect, then imagination by itself is not enough to explain engagement with fiction.

Problem 3: Pretence and the Link to Action

Third, simulationism understands simulations as ‘offline’, standardly glossed as “disconnected from action-generating systems.” (Currie & Ravenscroft 2002: 67) But, Kind points out, this makes imagination ill-suited to explain pretend-play, which does generate action. In Kind’s example, where “Christopher is pretending to be Obi-Wan Kenobi and his brother Sean is pretending to be Darth Vader” (142), a key explanandum is that they swing their sticks (light sabers) at each other. How does something ‘disconnected from action-generating systems’ explain that?

2. A First-Pass Response

Let’s start with problem 3. Although ‘offline’ is often defined as ‘cut off from generating action’, a broader definition is sometimes given: “detached from its usual function” (Nichols et al. 1996: 42). This definition handles pretence better: Christopher’s i-belief that, say, Sean is Darth Vader, does serve to guide action, but is clearly not performing its usual function, which would produce effects like overpowering fear, profound confusion, questioning why Darth Vader is so short, etc.

But this risks making ‘offline’ hopelessly nebulous. Certainly we can’t say that any state not playing its normal role is an simulation: then every instance of irrationality, thoughtlessness, or eccentricity would be imagination.

Here is a first pass at a better definition:

An ‘offline’ state is one such that which parts of its normal role it plays is open to voluntary control.

An offline state, we might say, has been ‘detached’ from its usual function, but can be ‘re-attached’ as and when the subject wishes. Christopher ‘re-attaches’ some simulations to action-generation, while someone else might have the same simulations but not do so, keeping them a purely internal fantasy. And (to touch on problem 2) someone might imagine the events of a narrative while deliberately letting them generate emotions, while someone else imagines the same events but doesn’t, keeping them ‘at arm’s length’ emotionally.

‘Offline’ is meant to be a sort of ‘second-order functional role’. It presupposes that states already have sets of characteristic causes and effects, and then identifies a status they can have which modulates that primary role (rather as ‘being attended’ often allows a state to produce its characteristic effects more strongly).

Note that the phrase ‘open to voluntary control’ does not mean ‘always voluntary’: simulations may occur and have their effects spontaneously. Nor does it mean ‘perfectly subject to voluntary control’: simulations may have effects in excess of our intentions. But the same things are true of moving our limbs, and we still call such movements ‘voluntary’.

3. A Second Pass Response

Suppose the first-pass response can explain why offline states can produce action, or not, and produce affect, or not. But problem 1 is different. Recall that Kind accuses simulationism of collapsing the distinction between supposing that P and imagining that P, thereby undermining the latter’s ability to justify the belief that P is possible.

It might seem like simulationists could treat forming modal beliefs as an output, like action or affect, which simulations can be freely detached from (in supposition) or re-attached to (in a more robust sort of imagination). But this won’t work. If we could simply simulate a belief and then choose whether or not to generate a genuine modal belief, we could form any modal beliefs we wanted.

Consider instead another output that simulations may be allowed to produce, or not: other simulations. Just as focusing on a given mental state will tend to bring to mind any number of others, so by focusing on a simulation we can often generate from it any number of other, connected simulations. Upon forming the i-belief that a flying pig exists, I can allow that i-belief to generate, for example, visual images of winged pigs, speculations about how they might evolve, questions about their relationship to humans, images of people trying to catch them, feelings of awe at their majestic escapes, and so on. Or, I can simply hold the i-belief in mind without letting it prompt such an enriching.

This suggests the following second pass:

An ‘offline’ state is one such that which parts of its normal role it plays, and whether other mental states it produces are themselves kept offline, is open to voluntary control.

My suggestion is that the distinction between imagining and supposing is really a difference of degree between richer and poorer sorts of simulation.‘Imagining that P’ generally means ‘simulating the belief that P and letting that i-belief generate other connected simulations’, and ‘supposing that P’ generally means ‘simulating the belief that P and keeping it at that’. The distinction makes an epistemological difference because letting the i-belief prompt many other simulations gives us a decent chance of detecting contradictory results.

Kind is right that it’s hard for simulationism to explain how simulations can sometimes be cut off from action, affect, and belief, but sometimes motivate action, provoke affect, or justify beliefs. I’ve suggested a defence based on a more flexible understanding of ‘offline’ than is normally provided: an offline state is not completely and finally separated from its normal effects, but rather provisionally detached from them such that which of them it does or does not produce is open to voluntary control.


Currie, G., and Ravenscroft, I. (2002). Recreative Minds. Oxford: Oxford University Press.

Kind, A. (2013). ‘The Heterogeneity of the Imagination.’ Erkenntnis 78 (1):141-159.

Nichols, S., Stich, S., Leslie, A., and Klein, D. (1996). ‘Varieties of Off-Line Simulation’. In Theories of Theories of Mind, eds. P. Carruthers, P. Smith, Cambridge University Press: 39-74.