Joshua Myers is a postdoctoral fellow at York University. His research lies at the intersection of philosophy of mind and epistemology, with a focus on imagination and imagistic representation. He received his PhD in philosophy from New York University.
This week at The Junkyard, we’re hosting a symposium on Rob Hopkins’ recent book: The Profile of Imagining (Oxford University Press, 2024). On Monday, we began with an introduction from Rob. Commentaries and replies are running Tuesday through Thursday.
* * *
Commentary from Joshua Myers
In this clear, thoughtful, and sharply written book, Hopkins presents a comprehensive account of the nature, content, and phenomenology of sensory imagining that is inspired by Wittgenstein, Sartre, and Ryle.
One of the core claims of this account is that, unlike what we perceive, what we imagine cannot overflow our awareness and is therefore not amenable to observation. Consider what it is like to perceive an apple. The apple seems to have a nature that is open to investigation. For example, you might be surprised to turn the apple around and find a bruise on the other side. The objects of perception overflow our awareness of them and, as a result, we can extend our awareness of them by observation. By contrast, when you imagine an apple, there is no aspect of the imagined apple that you are not aware of. If you imagine rotating the apple to reveal a bruise on the other side, then either you already imagined the bruise before rotating it, or you didn’t, in which case there was no pre-existing bruise to discover. Either way, you will not be surprised by the bruise. As a result, you cannot extend your awareness of imagined objects by observation.
A large portion of the book is dedicated to defending this contrast between perception and imagination. I will focus on just one thread which comes out in Chapter 7 while investigating how imagination can yield knowledge about the actual world.
Hopkins pits two models of learning by imagination against each other (well, really three, but the third gets set aside to be dealt with in later chapters). The first is the Observation Model, according to which we gain knowledge from imagination by observing what we imagine, and then drawing a conclusion about the world. The second is the Inference Model, according to which we gain knowledge from imagination by moving between imaginings in systematic ways. By unfolding our imaginings such that later imaginings draw out the consequences of earlier imaginings, we can then draw a conclusion about the world.
The Observation Model threatens to undermine Hopkins’ central claim that imagined objects cannot overflow our awareness. So, Hopkins is concerned to defend the Inference Model against the allure of the Observation Model. I am broadly sympathetic to the Inference Model (shameless plug: in Myers 2021 I argue, in a similar vein, that learning from imagination is a form of reasoning). However, in this short commentary I want to raise two questions that I hope will help to further develop and clarify what the Inference Model is committed to.
Consider two ways of extending our awareness of the world. One is to change our perceptual relations to objects to uncover features that were hidden from our initial perspective. This is what you do when you turn the apple around to discover the bruise. Another is to employ our recognitional capacities to recognize and categorize objects. This is what you do when you recognize that the apple is a Granny Smith. In both cases, you extend your awareness of a feature that the apple had all along. So, there is a sense in which both count as observation. But there is another sense in which they differ. The former is purely perceptual and involves sensing the feature, while the latter it partly conceptual and involves recognizing it. Let us call this latter model the Recognition Model.
I suspect that the Recognition Model can handle many of the cases that Hopkins uses to motivate the Inference Model. Here is one of them: suppose you imagine a capital ‘E’ with a capital ‘P’ adjoined to its right, and thereby come to learn that the resulting figure would contain a rectangle. According to the Inference Model, this involves first imagining the arrangement of lines, and then transitioning to imagining the arrangement of lines as a rectangle. The latter imagining draws out the consequences of the former. But are these really distinct imaginings? It isn’t obvious. The Recognition Model applies much more naturally: you imagine the arrangement of lines, and then use your capacities for recognizing geometrical shapes to recognize a rectangle in what you imagine. This last step does not involve a distinct imagining. Indeed, it does not involve anything distinctively imaginative at all. Instead, it involves relying on your domain-general recognitional capacities for rectangles. These very same recognitional capacities would be activated were you to perceive the arrangement of lines, rather than merely imagine them. If recognitional capacities can be a source of knowledge in perception, then why not in imagination?
My first question for Hopkins is whether he sees the Recognitional Model as a genuine competitor to the Inference Model, and if so, how to adjudicate between the two.
Hopkins fleshes out the Inference Model by appealing to his profiling view of imaginative content. According to the profiling view, imaginings represent non-perspectival properties by placing sensory representations (conjurings, in Hopkins’ terminology) of perspectival properties into sensory profiles, which in turn involve commitments to represent other perspectival properties in other circumstances. For example, imagining a cube involves committing to imagining certain visible figures when viewed from certain perspectives. These commitments determine how imaginings unfold over time, and explain how later imaginings can unfold the consequences of earlier imaginings in a way analogous to inference.
The profiling account naturally lends itself to explaining how we can learn from imaginings that involve shifting perspective, such as mental rotation exercises. However, it seems to struggle to accommodate learning from imaginings that do not involve shifts in perspective. For example, suppose you are wondering whether a basketball that is currently moving through the air will land in the basket or not. You might imagine the basketball continuing to move along its trajectory in order to find out. It is natural to think of this case as involving a kind of inference. You first imagine the basketball at its initial point and then unfold the imagining to determine where the basketball will end up. But how this imagining unfolds is not determined by the sensory profiles of the basketball alone. It does not just depend on one’s expectations of how the basketball will look from various perspectives but also expectations about how objects tend to move through space over time given the various forces acting on them.
Consider an even more sophisticated example. You are playing a game of chess. You imagine making a certain move, and then imaginatively simulate the likely consequences of that move. You imagine it to have favorable consequences, and on the basis of this imagining you conclude that the move is a good one. Once again, how this imagining unfolds is not determined by the sensory profiles of the various chess pieces. Instead, it depends on your knowledge of the rules of chess. This knowledge is much more cognitively and conceptually sophisticated than the low-level sensory commitments implicated in the profiling view.
My second question for Hopkins is whether the profiling view can be extended to account for these cases, or whether we need to bring in a much richer set of resources that go beyond sensory profiles, such as intuitive physical expectations and conceptual knowledge, to account for their epistemology.
Reply to Myers
In his thought-provoking and constructive commentary, Josh Myers asks two questions about my attempt to reconcile the claim that everything imagined is imagined consciously (No Unconscious Imagining) with the fact that we can sometimes learn from our imaginings.
The first question is whether the model of learning from imagining that Myers proposes, the Recognitional Model, is a genuine competitor to the model I favour, the Inference Model; and, if so, how we should settle which is best.
Now, if our interest is in defending No Unconscious Imagining, Myers’ Recognitional Model does not clearly pose a threat. The model claims that one way to extend awareness of what we imagine is by bringing it under recognitional concepts. After all, this happens in perception: Myers’ example is coming to see the apple before me as a Granny Smith. Here, bringing that concept to bear does indeed extend my awareness of what I’m seeing, in some sense of ‘awareness’. (I’m happy to accept, as Myers clearly assumes, that this awareness is more than merely intellectual - in the case of seeing the apple, that it is perceptual.) But does this shift make me aware of a feature of which I was previously wholly unaware? That is less clear. Myers expressly contrasts the case with one in which I move, bringing previously hidden facets of the object into view. In that case features entirely without my ken now enter it. That’s less obviously so when I merely apply a new concept. If we turn to imagining, then, does applying the Model there entail that the subject has already imagined features of which she is only later aware in any way at all, or only that she imagines features she is initially not aware of in one way, though she does enjoy some other form of awareness of them? Only the former threatens No Unconscious Imagining.
However that may be, Myers’ proposal is certainly interesting. It parallels an account of our perception of (external) pictures, on which they engage our recognitional capacities (Schier 1986, Lopes 1996). I think we should be wary of the idea that imagined contents can be conceptualized in the way that things we perceive, whether ordinary objects or the marks on a picture's surface, can be. Assuming as much is precisely to succumb to the assimilationist tendency, to treat imagining as some inner analogue of perceiving, that the book sets out to resist. Of course, it's one thing to favour a different general approach, another to argue for its superiority with respect to the issue in hand. Let me, then, raise at least one issue confronting proponents of the Recognitional Model.
We should distinguish between merely applying a concept to some content one grasps and applying such a concept correctly. Myers’ interest must be in the latter, since only then is there any prospect that in conceptualizing what she has imagined the subject discovers something about it of which she was not previously aware. But what here could make it correct to apply one concept, rather than another? If one is seeing an apple, the answer lies in the nature of the perceived object: is it a Granny Smith, or not? If one is looking at a picture of an apple, the answer might lie in the artist’s intentions. But if one is merely imagining an apple, a thing initially imagined not under that concept, what could secure that one is right to conceptualize it as a Granny Smith, rather than, say, merely as something that looks like one?
In some cases, there will be an answer to this question. Those are precisely the cases that, in the chapter Myers is engaging with, are at issue between my Inference Model and the alternatives. So perhaps the question just posed will seem out of place. It does, however, serve a purpose. The unproblematic cases will be ones in which the features the subject (according to the Recognition Model) comes to recognize are logical or geometrical consequences of features of which she is already aware. So, the scope for appeal to recognition in the realm of imagining will be far more limited than in the case of perception (be it of ordinary things or of pictures of them). The assimilationist character of Myers’ proposal makes this fact easy to overlook. It wrongly encourages us to think that, if we can recognize apples on seeing them, we can recognize them when imagining them. And it is precisely those temptations that the book sets out to eradicate, by rejecting the assimilationist assumptions that drive them in the first place.
Myers’ second question concerns the range of cases my view might hope to cover. As he points out, we may learn from imagining, not only about such matters as whether two shapes are congruent, but also about, e.g., whether a moving basketball will make it through the hoop. The Inference Model of learning I develop leans heavily on the notion of a sensory profile, and, Myers worries, that notion does not apply to the trajectory of the ball. Isn’t my view rather limited in scope?
I agree that we cannot make sense of this case by appeal to the notion of a sensory profile. That is a matter of how the perspectival appearances of things bearing a certain feature shift with our relations to those things. However, as I note in the book (117-118), the profiling approach can be extended. If sensory profiles are patterns in perspectival appearances that characterize non-perspectival properties such as surface colour and 3D shape, other profiles will be patterns in those non-perspectival features. The profile of the property of being a horse, for instance, will be in part a pattern in the various 3D shapes an object can assume. (Profiles don't define the properties they characterize: they correspond 1:1 with such properties, insofar as the latter can be present to us in perception: pp. 81, 84.) Similarly, various dynamic properties, such as being on a certain trajectory, will be patterns in the positions the object might occupy in a given 3D space. In order to make sense of our being able to imagine more than shape, colour, sound and the like at all (i.e. regardless of any questions about what such imaginings may teach us, and how), I have to bet that this extension of the idea of a profile makes sense. Being committed to that, I'm happy to take on a further bet. This is that we can use the idea of those higher-order profiles to make sense, in a simple expansion of the terms the Inference Model offers, of such cases as using imagining to learn where the basketball will fall. Of course, to tell you what I’m betting on is not to argue that those bets will come off. I’m merely sketching a strategy for addressing the challenge Myers's second question poses. But I hope this at least gives an idea of how I would attempt to meet it.
References
Lopes, Dominic. 1996 Understanding Pictures Oxford University Press: Oxford.
Schier, Flint. 1986 Deeper into Pictures Cambridge University Press: Cambridge.