Dr Tom McClelland – The Architecture of Consciousness Project – University of Manchester
There are some properties you can see and some you cannot. When you look at the picture below, for instance, what do you see? I see colours such as the yellowness of the banana, I see shapes such as the banana’s curve, I see spatial relations such as the banana’s proximity to the man’s head and I see textures such as the smoothness of the man’s necktie. There are other properties I don’t see. I don’t see the banana’s property of being a source of potassium or its property of costing 28p. And I don’t see the man’s property of being a member of the Labour Party or his property of being an elder brother. On the basis of what I see I might judge that the things I’m looking at have these properties, but that’s not the same as actually seeing those properties. After all, properties like ‘being a source of potassium’ just aren’t the kind of thing that one could see.
The examples I’ve mentioned shouldn’t be too contentious, but there are many kinds of property that do cause controversy. For instance, can you see what kind of object something is, such as seeing the smaller object as a banana and the larger object as a man? Can you see causal properties such as the banana being supported by the hand, or affordances such as the banana being edible? Can you see aesthetic properties such as the banana’s beauty, or moral properties such as the man’s virtue? Can you see the identity of objects, like seeing the man as David Miliband?
There is a great deal of debate in philosophy about these contentious cases, and the disputants fall into two camps. The first camp are conservatives, and they say that our visual experiences are limited to the basic kinds of property I first listed: colours, shapes, spatial relations and textures (e.g. Prinz 2012; Brogaard 2010). These conservatives shouldn’t be confused with political Conservatives, but like political Conservatives they are big on austerity – they take an austere view of visual experience that excludes all the contentious properties. The second camp are liberals, and this camp adopts a much more inclusive view of perception (e.g. Siegel 2012; Bayne 2009). They hold that at least some of the contentious properties can be visually experienced. Again, this kind of liberal shouldn’t be confused with political Liberals, but like political Liberals they are endlessly arguing among themselves about just how liberal they should be — the property of being a man is surely permitted as a visible property, but might permitting the property of being virtuous be a step too far?
Now, which camp are you in? The questions I’ve been asking are about what it’s like for you to have the visual experience you have when you look at the photo above. Conservatives would offer an austere description of your experience involving only the limited range of properties that they countenance. If you think that such a description fully captures what your visual experience is like, then you’re a conservative (don’t worry — that doesn’t come with any political commitments). If, on the other hand, you think there’s more to your visual experience than is captured by the austere description, then you’re some kind of liberal, and will have to reflect carefully on just how wide the range of properties you can see is.
I’m a liberal, but I’m thinking carefully about just how liberal we should be. Specifically, I’m interested in whether we can see a special category of property called ‘scene categories’. When we open our eyes we don’t just see objects – we also see the wider environments in which those objects are embedded. The philosophy of perception tends to focus on our perception of objects — there is endless discussion of whether we can see an object as a pine tree, for instance, but no real discussion of whether we can see a scene as a forest (e.g. Siegel 2012). I think this is an oversight and that we should ask ourselves whether we can perceive scene categories such as being a forest, being a beach, being a field, being a street, or being a carpark.
Consider the image above. Besides seeing the various shapes, colours, spatial relations and textures in this image do you also see the scene as a forest? Is the scene’s property of being a forest part of your visual experience? Conservatives would say that it is not, and would deny that any such scene category can be perceived. They would accept, of course, that we recognise the scene as a forest — they would just deny that this recognition is perceptual. On their view, we see certain patterns of colour and shape and then judge that the scene is a forest. However, I think that a combination of empirical and philosophical considerations cast doubt on this conservative view. There are good reasons to adopt a liberal view that acknowledges we can see scenes as forests or as beaches in much the same way as we can see objects as green or as tall. Conservatives will need some convincing that we visually experience scene categories, and you might need some convincing too. My case for this has two steps: the first step concerns the ‘visual’ bit of ‘visual experience’ and the second step concerns the ‘experience’.
If conservatives deny that we perceive scene categories, they have to say that we recognise scene categories through some kind of post-perceptual cognitive process, such as making a judgement on the basis of what we see. The empirical data counts against such a view in at least four ways. First, judgement is relatively slow, but our recognition of scene categories is incredibly fast. Thorpe et al. (1996), for instance, found that when subjects were shown images in a scene categorisation task, their brains showed Event Related Potentials (ERPs) as early as 150 milliseconds after being shown the image. Second, it is generally thought that only attended areas of the visual field are available to judgement, but our recognition of scene categories often seems to be inattentive (see Li et al, 2002). Third, the speed at which we make discriminative judgements about a stimulus can generally be improved if we’re familiar with the stimulus, or if we form appropriate expectations about the stimulus. However, an early study by Biederman et al (1983) suggests that familiarity and expectation do not speed up our categorisation of scenes, indicating that scene categorisation is an automatic perceptual process. Fourth, perceptual processes display a phenomenon known as ‘perceptual aftereffects’ (which you can find more about here). Post-perceptual processes do not display this effect, but a study by Greene & Oliva (2010) indicates that scene categorisation is susceptible to aftereffects.
Interpreting this data is not always straightforward, but it certainly looks like scene categories can be recognised perceptually, not just through post-perceptual judgements. But I’m not home free yet. It’s one thing to perceptually process a property but quite another to perceptually experience it. Since I claim that we perceptually experience scene properties, I have more work to do. This is where some philosophical considerations need to be introduced to supplement the empirical data. Liberals use something called ‘contrast cases’ to show that our visual experience is more rich than conservatives think. Contrast cases are pairs of visual experiences that differ from each other in ways that conservatives are unable to account for. Such cases drive the following argument against conservatives:
- The two experiences are alike with respect to all conservative-permitted properties i.e. they represent all the same colours, shapes, spatial relations and textures.
- The two experiences are nevertheless different i.e. what it’s like to undergo the first visual experience is different to what it’s like to undergo the second.
- Therefore the two experiences must differ with respect to properties not permitted by conservatives.
Here is a classic example used by liberals:
To begin, this image looks to most people like a meaningless jumble of black and white patches. But if you look closely you can recognise it as a picture of a cow (the face is on the left and is looking towards you). This revelation changes what your visual experience is like, but the conservative can’t explain this change because there is no difference in the colours, shapes (etc.) that you see. Surely what changes is that you start to see the image as a cow? Conservatives deny that we see this kind of property, but this contrast case suggests they are wrong. Perhaps a similar example can be found in which we come to visually experience a scene category. Consider the following image:
Again, you might start by seeing meaningless patches of black and white but then come to recognise that this scene is a waterfall. To make sense of this change, it seems we must say that we visually experience the property of being a waterfall. Here’s another kind of example often used by liberals:
You might first recognise this image as a rabbit then recognise it as a duck. Your visual experience represents the same conservative-permitted properties in both cases, so the change must involve some more contentious property, such as visually experiencing the image first as a rabbit then as a duck. Again, we might be able to find a counter-part to this example involving scene categories. Consider the following image:
These sand dunes look a lot like waves, and you might be able to switch between visually experiencing this scene as a desert then visually experiencing it as a sea. If so, this would again be a case in which we see scene categories.
Although these brief arguments are far from conclusive, they offer a taste of the larger case I hope to make in favour of the visibility of scene categories. Ultimately though, there’s only one way to decide where you stand on these issues, and that is to ask yourself what you can see!
REFERENCES
Bayne, T. (2009). Perception and the Reach of Phenomenal Content. Philosophical Quarterly, 59(236), 385–404.
Biederman, I., Teitelbaum, R. C., & Mezzanotte, R. (1983). Scene Perception: A Failure to Find a Benefit From Prior Expectancy or Familiarity. Journal of Experimental Psychology, 9(3), 411–429.
Brogaard, B. (2013). Do we perceive natural kind properties? Philosophical Studies, 162 (1), 35–42.
Greene, M. R., & Oliva, A. (2010). High-Level Aftereffects to Global Scene Properties. Journal of Experimental Psychology, 36(6), 1430–1442.
Li, F. F., VanRullen, R., Koch, C., & Perona, P. (2002). Scene categorization in the near absence of attention. Proceedings of the National Academy of Sciences of the United States, 99(14), 9596–9601.
Prinz, J. (2012). The Conscious Brain: How Attention Engenders Experience. Oxford: OUP.
Siegel, S. (2012). The Content of Visual Experience. Oxford: OUP.
Thorpe, S., Fize, D., & Marlot, C. (1996). Speed of Processing in the Human Visual System. Nature, 381, 520–523.
Tom,
thank you for the interesting post, I hope it’s not too late to leave some observations — sorry if they are going to be quite long, but I’m fascinated by the issue as well.
First, a somehow provocative question. Your main claim is that “we don’t just see objects — we also see the wider environment in which those objects are embedded”. Does anyone really disagree with this claim? It’s true that philosophers tend to focus on examples such as the perception of an apple in order to ask questions about what properties we can visually experience, isolating individual objects in a way that may seem arbitrary. Yet I am unsure whether they would deny that the broader scene is visible too. After all, a beach or a street is made of objects.
Perhaps you want to claim that — over and above individuals with their properties — we can perceive groups, collections and other aggregates as holistic wholes instead as of mere sums of individual objects. On the face of it, this seems a plausible idea. Being a forest, being a crowd, being a flock may be good examples. Does this make sense to you?
Second, let’s grant that the empirical evidence suggests that recognition of scene categories is not something like an inference based on perception which leads to a judgement. Thus, perceptual recognition is not the same as a judgement. Still, someone may argue that the fact that we can recognise scene categories without necessarily forming explicit judgements does not yet show that we can perceive those categories. Recognising is one thing, perceiving is another. The former — according to at least some theorists — is not required for the latter.
Third, aspect-downing pictures. Without any need to think that we can perceive properties such as being a cow, you may explain the visual phenomenal switch you undergo when an object appears to you in the picture by appeal to properties such as organisational or Gestalt properties. Paying attention to the colours and shapes on the 2D surface leads you to group the visible elements in such a way that a 3D object emerges. Sure, the object is a cow; and since you know what cows look like, you recognise a cow being depicted in the picture. But is it necessary to recognise it as a cow or to see it as a cow in order to have a phenomenally different experience from the experience of seeing a flat surface covered in colour?
Think of simpler examples of ambiguous figures, such as this (discussed in Peacocke’s Sense and Content,1983). You can see the dots either as columns or as rows (depending on whether you give more weight to grouping by colour or by proximity). Such a grouping, though it may involve attention, does not depend on conceptual or recognitional capacities. Perhaps a similar explanation also works for the aspect-downing cow picture. True enough, the cow and waterfall pictures are indeed pictures, “in” which we experience 3D objects that are not really there. So there may well be a further level which distinguish picture perception from the perception of simple visual groupings. Still, recognising the seen-in object as a cow or a waterfall does not seem necessary to undergo a phenomenal switch.
Incidentally, I’d like to highlight how your scene-perception proposal may have the advantage of avoiding whatever arguments one may advance against the proposal that we can perceive natural kind properties (such as being a cow). In order to do so, however, you would need to insist that e.g. seeing a pine forest as a forest — as a whole, as a unified scene… — does not imply seeing it as a pine forest, i.e. it does not imply seeing its trees as being pine trees instead of being another kind of tree.
Finally, I would be interested in knowing more about your the desert/sea example. One natural description of the case is this: the picture is indeterminate about the object or scene it depicts; notably, it is indeterminate with respect to the object’s or scene’s colours; thus, we may equally interpret it either as a picture of a sea or as a picture of a desert. Yet this interpretation does not amount to a perceptual difference. For instance, it is not the case that in the “sea” experience you see the curved dark lines as depicting convex 3D shapes whereas in the “desert” experience you see them as depicting concave 3D shape (in contrast with an ambiguous case like this). Perhaps you may start imagining that the scene you see in the picture is blue in one case and sandy in the other, but visual imagery is not perception. Surely we visually experience a 3D object or scene over and above colours and shapes, but why it should be such a specific property as being a sea as apposed to being a desert. Can you explain why you think it’s a case of perception of those scene categories?
I understand you weren’t able to go into much detail in a blog post, but if you are able to respond in the comments I would be interested in better understanding your view. Thank you!
Hi Giulia. Many thanks for these comments. They’ve given me a lot to think about. I’ll post a separate reply for each comment…
I don’t think anyone would disagree with my claim that we perceive scenes and not just individual objects. My worry isn’t that philosophers deny the existence of scene perception but that they tend to *ignore* it. Some philosophers might assume that perceiving a scene is just a matter of perceiving some complex of objects that constitute the scene, but there are at least two reasons to doubt this: 1) some scenes aren’t constituted by objects (or at least not by perceptible objects) – ocean scenes aren’t made of objects and although a beach might technically be made of grains of sand we don’t have to perceive these individually to perceive the beach; 2) the psychological data seems to suggest that much scene-processing occurs in parallel to object-processing rather than being built upon object processing. This suggests that much of what’s involved in recognising a scene is psychologically independent of object recognition. You mention that I might want to say that we perceive groups (etc.) as more than the sum of their parts. As it happens, I do want to say this – there’s some interesting data supporting such a view. However, for the reasons above I don’t think that this captures what’s involved in scene perception. Recognising something as a forest involves more than representing an aggregate of trees.
Your second comment highlights the fact that a crude dichotomy of perception and judgement is probably too simplistic to capture the range of psychological states we enjoy. If it was a straightforward choice between saying a) that we can only judge scene categories or b) that we can perceive scene categories, I hope to have tipped the balance in favour of ‘b’. But if we introduce kinds of recognition between perception and judgement I’m open to the possibility that scene recognition falls into one of those other kinds. The problem is that I don’t think we yet have a workable account of what those other kinds of recognition are.
Your third comment highlights a couple of crucial questions in the philosophical literature surrounding the content of perceptual experience: 1) might changes to attention capture what’s going on in contrast cases?; 2) can perceptual experience represent ‘natural kind’ properties that are typed by their underlying nature rather than by their outward appearance?
My response to the first question is that I don’t think changes in attention exhaust what happens when we recognise an image as (for instance) a waterfall. My reasons for this are a little complex, but the two main points are; a) that scene recognition often seems to occur inattentively, suggesting we can perceptually experience scene categories without having to attend to that scene, and; b) that scene recognition (unlike object recognition) isn’t achieved by attending to local fine-grained features of a scene but rather by processing the overall global properties of the visual field.
My response to the second question is that, as you suggest, there are good reasons to doubt that perceptual experience can represent natural kinds. This indicates that the only scene categories that are perceptible are those that aren’t typed by natural kinds, thus being a pine forest would be inadmissible but being a forest is probably admissible. Of course, another option is to deny that these worries about natural kinds stand up to scrutiny, and I wouldn’t want to rule out that possibility too hastily.
My argument assumes that *something* in your perceptual experience changes when you shift from seeing the image as a sea scene then seeing it as a desert scene. If you don’t have this kind of switching experience, my argument doesn’t get off the ground. If you do have this switching experience, my argument is that perceptual representation of scene categories is the best explanation of the switch. The switch can’t be a matter of coming to see the image as 3D because both experiences represent the scene to have the same contours. The switch can’t be a matter of attending to local low-level properties associated with being the sea, then attending to local low-level properties associated with being a desert, because the picture doesn’t involve any such diagnostic properties. Nor can the switch be a matter of imaging the scene as blue in the first experience then imagining it as yellow in the second experience because you can (I think!) achieve the relevant switch without imaging the scene as having any colour. Having ruled out all of those, it seems we have to start looking at high-level properties instead of low-level properties, and here I think that scene categories like being the sea and being a desert are the most obvious candidates.
I concede that there’s a limit to how much weight I can put on the contentious sea/desert case. That’s why I’ve been trying to accumulate a larger collection of examples that support my argument. I post some of these below…
…actually, no I don’t because I can’t post pictures. But trust me, they’re REALLY CONVINCING! 😉
Hi Tom,
thank you for the helpful replies. In point of fact, I also tried to post links to pictures (Peacocke’s example of dots) and a picture of shaded circles that you can see either as concave or as convex semispheres.
I’ll have a closer look at your comments and try to respond asap.
Tom,
many thanks for your replies. Here are some point-by point follow-ups.
1) Thank you for clarifying your claim. In order to have a better grasp of the distictiveness of scene perception, may you tell us more about the empirical data you mention? Can you give an example of experiment showing how scene-processing does not depend (or is not wholly determined by) object processing?
This leads me to a much wider question: how is processing related to perceptual experience and its content? I certainly don’t expect you to offer us a full theory of such a relation, but you may have a sketchy idea. Questions like this would probably always appear at some point when you are using (certain kinds of) empirical evidence in support of a philosophical thesis.
2) Fair enough. So my question becomes: do you think perception always implies recognition, so that there is no such a thing as perceptual discrimination without perceptual recognition? Relatedly, what does recognising an object — or recognising the object as having a certain property — require?
As you may imagine, I’m more of a supporter of the independence of perceptual capacities from recognitional capacities. Yet this is entirely irrelevant, as I just want to see where you are coming from.
1) One the best studies that seems to show scene recognition is not driven primarily by object recognition is Greene, M. R., & Oliva, A. (2009). Recognition of natural scenes from global properties: seeing the forest without representing the trees. Cognitive Psychology(58), 137–176. (they also cite a number of other studies that point to the same conclusion).
I don’t have anything especially insightful to say about the relationship between processing and perceptual experience. I think it’s a *necessary condition* of a property being part of the content of perceptual experience that it is processed perceptually. The real challenge for philosophers and psychologists is to establish the *sufficient conditions* but I can avoid this worry because I never say “this property is processed in this way, therefore it’s part of the content of perceptual experience.”
2) I think I’m using ‘recognition’ as a label for processing a property that’s neutral on whether the property in question is represented in perception, in belief or in some other way. Perhaps that term is misleading. I think that to perceive a property is to represent it in some way. You might call such representation recognition, or might preserve the label ‘recognition’ for a particular kind of representation. I’m not sure.
3) I do agree with you that attending to low-level properties such as shapes, colours, spatial positions etc. cannot explain the phenomenal change we undergo with aspect-downing pictures and (twice) with ambiguous pictures. It cannot explain the change we experience with the simple dots figure either. Coming to perceive more detailed shapes or more determinate colours thanks to focusing our attention is clearly not related with those phenomenal switches in any interesting way.
Your b) point — we achieve scene recognition by processing global properties of the visual field — is consistent with the view that differences in perceptual attention can account for the phenomenal switches at least with the simpler, non-representational figures. By attending to the shapes and colours, we do not merely perceived them as more determinate, but we group them/organise them in such a way that a new property is perceived. This should be some kind of Gestalt property, which is a good example of global property since it is neither reducible to nor supervenient on lower-level perceivable properties (in her book, Siegel briefly mentions a similar proposal as an alternative explanation of an alleged instance of natural kind perception).
However, if your a) point is correct — attention is simply not required for scene recognition / perception — then the proposal wouldn’t be a viable option.
3) I like the idea that perception of global properties is analogous to perception of gestalt properties. Of course, I’d then need to work out what the relationship is between perceptually experiencing those global gestalt properties and perceptually experiencing scene categories.
As you say, point ‘a’ would help my argument a lot. Unfortunately, the evidence in favour of this is not clear-cut, so I can’t rely on it too much.
4) As for the sea/desert picture, I’m thorn between denying that there is a phenomenal difference — in contrast with visually ambiguous figures, this picture does support two interpretations but does not support two different visual experiences — and looking for another explanation of the overall contrast between the “sea” and the “desert” experience. I undoubtedly agree that neither attending to local low-level features (see reply 3 above) nor going from seeing a 2D surface to seeing to a 3D scene can account for the phenomenal difference.
By the way, I really want to SEE the other pictures — I’m sure you appreciate how this is different from trusting your report
Of course, I wouldn’t expect you to trust my report 🙂 I’ll send you the most recent draft of my paper then you can see all the examples and a bibliography of some of the most interesting studies I’ve looked at. Thanks again for all your comments — it’s been very helpful to work my ideas through further.
: — ) Can you direct me to studies where the examples are discussed?