09 Jan How Are Perceptual Beliefs Justified?
In this essay, I will answer the above question with a version of direct realism: Perceptual beliefs are justified because they arise out of a direct awareness of physical objects. Those beliefs are justified as long as there is no defeater.
There are objections to this view. I will consider four which I consider most weighty and try to show why they do not refute my version of direct realism.
I. A Direct Realist Theory That Yields Justified Perceptual Beliefs
I.1 A broad picture of perception
Before going into the minute work of explicating my theory, I think it helpful to present a broad and general picture of what perception is composed of. I follow Robert Audi (Audi 2002, p. 16-17) on his “four elements in perception”:
- the perceiver
- the object
- the perceiver-object-relation (a causal connection)
- the sensory (perceptual) experience
Virtually all theories of perception agree on point 1. Element 2 is more contentious. Phenomenalism and Idealism deny the existence of external objects, whereas realist accounts such as mine confirm it. Point 3 seems to be partly denied by intentionalism (Crane 2005, p. 31-38), although even here some causal connection between object and perceiver must occur at some point or another. At any rate, my theory clearly assumes the existence of a causal connection. Lastly, the existence of sensory experiences (4.) is something that we all know from introspection, hence this point is not a matter of debate. What is controversial is what we have the sensory experience of. Indirect realists will claim that is “sense-data” – non-physical objects caused by the physical objects – that we have direct perceptual experiences of; hence, the experience of physical objects is only indirect. A direct realist such as me believes that our perceptual experiences are directly of the external objects. Notice that perceptual experience is a part of perception and therefore does not equal it. I will come back to that distinction later. For now, let it suffice that perceptual experience is what we cannot help but have when we let our sense organs do their job. Perceptual experiences “just happen to us”.
I.2 How Phenomenally Conservatist Direct Realism accounts for the justification of perceptual beliefs
I call my theory “Phenomenally Conservatist Direct Realism” (in short PCDR).
It is a version of naïve realism because it assumes the physical world to exist independently of our minds. The qualifier “naïve” expresses the view that physical objects retain all their properties when unperceived, not just secondary qualities (for more on that see section II.4).
It is direct because the objects of awareness (the objects we perceive) are directly the external (physical) objects and not some intermediary objects representing the external objects (as posited in indirect realism). Directness here includes a causal chain from the external objects to the perceptual experience in the perceiver. By a causal chain I mean the established scientific sequence that begins with physical stimuli such as light or heat which turn on biochemical reactions in cells of our sense organs, which in turn trigger afferent neurons to fire and ultimately end up with neurons firing in the (visual, tactile…) cortex. Let me add here that in my view the causal chain does not end with brain neurons changing their states; rather, these brain states give rise to the perceptual experiences in our minds, which are part of our immaterial substance (for the record, I’m a substance dualist).
Finally, my theory is phenomenally conservatist because it builds on the principle of Phenomenal Conservatism (PC) as rendered in Huemer 2001, p. 89:
(PC) If it seems to S as if P, then S thereby has at least prima facie justification for believing that P.
Now, how do those elements work together in justifying perceptual beliefs? Let’s take an example.
I look out my window and see a snowy mountain. As PCDR is a realist theory, it assumes that there really is a mountain, and that the mountain out there causes my perceptual experience. According to PCDR, I could also say that I am aware of the snowy mountain – this is one consequence of its being direct realist (an indirect realist would say he is aware of some sense-data caused by the mountain, not the mountain itself). Awareness according to my usage of the term describes a mental state in which I have a representational content in my mind that represents the world as it is. To count as perception, this representational content must meet two criteria:
- Content-satisfaction: It must correspond at least roughly in character to the things in the world.
- Non-accidentality: It must have been formed non-accidentally, i.e. by a causal relationship between the things out there (here the snowy mountain) and me (the perceiver).
To 1): Surely what I see when looking outside the window corresponds to what is really there. The qualifier “roughly” is important because if I suffered from, say, short-sightedness, I would see the mountain blurred. This would still count as perception, even if it does not render the object perfectly. However, if I had the perceptual experience of a roaring polar bear, this would be no perception of the mountain, but a hallucination or an illusion, because the corresponding representational content (of a roaring polar bear) is too dissimilar to the external object.
To 2): This obviously seems to be the case. How else could the perceptual experience of the mountain come about? Even if someone had put a deceivingly realistic wallpaper of the mountain in front of my window, it would still be this wallpaper that causes my perceptual experience (of course, my belief that I see a mountain would be false, but that’s another story). Well, there is in principle one setting in which I could have perceptual experiences of my environment in a way that does not fulfill the above requirement of causality: Suppose a scientist with a very elaborate apparatus stimulated my brain so that it produces the brain states causing the perceptual experiences of things around me. This would not be the required causal chain. But as long as I don’t have reasons to believe that my brain is manipulated in such a way, I can trust that my perceptual experiences have come about in the right causal way.
Now we’ve established an account of what it means to have a perceptual experience. This is presumably what Robert Audi calls “simple perception” (Audi 2002, p. 17). As depicted earlier, a perceptual experience is a necessary part of perception, but not the whole story. It seems that a perceptual experience per se does not contain a belief, though it usually leads to the formation of a belief and is foundational for it. When we perceive (as opposed to a mere perceptual experience), we perceive something (which physically exists, on a realist account) and hence form a belief about that something. We call those beliefs perceptual beliefs.
Perceptual beliefs are, in my theory, always grounded in and caused by perceptual experiences. Now this characteristic is very important, for it makes perceptual beliefs foundational. Foundational beliefs are beliefs not based on other beliefs. To be sure, this does not mean that they are not grounded in anything (in our case they are grounded in the perceptual experiences). Nor does it mean that we cannot have reasons for holding foundational beliefs or support them by reasons; it just means that reasons aren’t required for foundational beliefs.
Just as any other belief, a perceptual belief can be true or false. Here we come full circle to the notion of perception: In my view, one should speak of perception only if the perceptual beliefs formed by the perceptual experience are true and justified, in other words if they constitute knowledge. Consider the famous “bent” stick example: It looks as if the stick is bent under water, but this isn’t true. Hence, if I formed the belief that the stick is bent (as opposed to believing that it looks bent), this would be no knowledge and therefore no perception, but some form of illusion.
Here, a critical question arises: How do I know that I am not mistaken frequently or even all the time about my perceptual beliefs? Recall the mountain: It is possible that I see just a huge, true-to-life wallpaper. I could go double-check by making a hiking tour and touch the mountain. But, first, touching (or smelling, for that matter) the mountain is just another perceptual way of learning about the object, one that might also be misled. Second, none of us so frantically scrutinizes his perceptual beliefs unless in very special circumstances; very reasonably, we take most of our perceptual beliefs to be true and live our lives quite successfully by them. The key to a healthy, trusting, but not credulous way to perceptually grasping the world is the aforementioned principle of Phenomenal Conservatism (PC):
(PC) If it seems to S as if P, then S thereby has at least prima facie justification for believing that P.
If it seems to me that I see a mountain, then I have at least prima facie justification for believing that I see a mountain. That means, I am justified as long as no defeater comes in, e.g. evidence that a huge mountain wallpaper has been erected in front of my window. Notice that the kind of justification here is internalist, i.e. from the subject’s point of view, as opposed to externalist justification, a justification from an “omniscient” third-person perspective. According to the latter, I am of course not justified in believing that I see a mountain when in fact a wallpaper is before my eyes (not even if I don’t know that I am deceived). But in my view, externalist justification is useless for a philosophy that seeks to give answers relevant for everyday life. Its standard is too high. Internalist justification, on the other hand, is exactly what befits a healthy epistemic stance: not too skeptical, but not too apodictic either.
II. Objections to PCDR and responses to them
Notice that the following objections go against direct realism and try to support indirect realism. They don’t deny that perceptual beliefs can be justified, only that they are justified in the way direct realism says.
II.1 The Argument from Perspective
Michael Huemer coined the term “Argument from Perspective” (Huemer 2001, p. 119) to delineate a Humean argument that is different from the Argument from Illusion (see section II.2):
The table which we see seems to diminish as we remove farther from it; but the real table, which exists independent of us, suffers no alteration; it was, therefore, nothing but its image which was present to the mind. (Hume 1975, XII.1, p. 152)
Hume is not talking about an illusion here, but about the change of perspective that occurs when we move away from a table. The more we move away, the smaller the table seems to us; but as its actual size remains the same, Hume concluded that what we see is not the table itself, but just a mental image of it.
In response, Michael Huemer (Huemer 2001, p. 120) points out that we should distinguish between the linear size and the angular size of an object. The linear size is what we normally call size, say 1 m x 1,20 m x 0,8 m in the table case. This is what remains constant no matter where I stand in relation to the table. What changes is the angular size, represented by the angle at point A:
This concludes matters for Hume’s argument: there is no need to posit sense-data, just different (objective!) size-properties. In fact, the linear-angular distinction not only deflects Hume’s argument, but also yields positive evidence for direct realism. The changing angular size is precisely what we expect if our perception represents things as they are. Notice that the changing angle is a matter of geometrical necessity and therefore an a priori truth. I think we couldn’t wish for a more sound support for direct realism.
II.2 The Argument from Illusion
Michael Huemer presents the argument in the following, revised form (Huemer 2001, p. 126):
- When you look at the stick, you are directly aware of something that appears bent.
- No (relevant) physical object is bent at this time.
- If you are directly aware of a thing, then the thing is the way it
- Therefore, the thing you are directly aware of is something nonphysical.
Of course, the argument aims at establishing indirect realism as the best explanation for our perceptual experiences. Huemer points out that premise 3 is not true without qualification (Huemer 2001, p. 126). Indeed, according to the content-satisfaction principle, the perceptual experience must correspond roughly to the physical things, not exactly. I take the bent-looking stick to meet this criterion, but that’s surely debatable. However, there is another way of showing that the bent-looking stick does not necessitate the conclusion that the object of our direct awareness is a non-physical sense-datum. As in the table case, there is a property of the stick distinct from its geometrical form; call it “optical form”. It depends on the media the light passes before reaching our eyes. In the present case, light travels from part of the stick first through water and then through glass and air. Depending on the refraction at the interfaces between two media, the object or part of it looks displaced. This is as much explainable physically as the angular size is explainable mathematically. In summary: Whichever version you prefer – requiring the perceptual experiences to only roughly meet the character of the objects or positing a property that accounts for the “distorted” look – there seems to be no need to postulate sense-data.
II.3 The Argument from Hallucination
Having a hallucination means in my terms having a perceptual experience of something that isn’t there. Indirect realists have always held that the best way for accounting for hallucinations is positing sense-data or something similar. The argument goes roughly as follows:
1) A perfectly faithful (visual) hallucination of a mountain is subjectively indistinguishable from an ordinary perceptual experience of that mountain.
2) What one is directly aware of is the same sort of (non-physical) thing in a perfect hallucination of a mountain as in an ordinary experience of seeing a mountain.
3) What is directly seen in a hallucination of a mountain is not a mountain.
4) What is directly seen in an ordinary perceptual experience of a mountain is not a mountain.
This argument has several problems.
First, premise 1 is debatable. By definition, what I hallucinate is not really there. Can one have the same kind of acquaintance with a hallucinated object as with a really existing one? I doubt one can. It might be responded that I have no other way but my experiences to check and verify, and saying they are subjectively indistinguishable means only that the subject cannot distinguish them, not that, for example, God could not distinguish them. But I am not content with that. Perhaps a person that first hallucinates the mountain and then, shortly after, has a genuine perception of the same mountain, could tell the difference. This is, however, not the place to discuss that further, so I’ll proceed to the next premise.
Premise 2 has the problem that it does not follow from 1. Even if hallucinations were subjectively indistinguishable from ordinary perceptual experiences, it does not follow that the underlying mental process is the same. More still, to speak of “awareness of things” is to bring indirect realism through the backdoor. On my PCDR account, one is aware of things in the external world, not of perceptual experiences. In fact, it follows from PCDR that hallucinations are not awareness at all, which fits well with my assumption that hallucinations and perceptions can in fact be told apart. But even if we choose the more agreeable term “mental process” instead of “mental things/objects”, it is a non sequitur to conclude from (arguendo granted) subjective indistinguishability that hallucinations and perceptions include the same mental process. Two different mental processes might lead to the same subjectively indistinguishable experience. Perhaps one day science can show that the brain processes are identical in hallucinating and perceiving persons; this would then be good evidence in for the “mental processes identity” thesis. But for now we need not accept premise 2.
Premise 3 is widely agreed upon and evident. But the conclusion is blocked because the argument fails at the transition from premise 1 to premise 2.
I’ve said much to rebut indirect realism, but I wish to give a brief PCDR account of hallucinations too. Direct realism accepts that perception hinges on a causal chain of physical and biological processes, some of which can misfire to lead to false positives. We know from LSD consumers that they have strong hallucinations, obviously caused by the drug’s effects in the brain. If similar effects can take place in a non-drug-consumer, we have a very plausible way of explaining hallucinations without committing oneself to sense-data. The PCDR theory even accounts for the justification that arises out of hallucinations: As long as the hallucinating person has no counterevidence, he or she is justified that what he or she hallucinated is true.
II.4 The Argument from the Illusoriness of Secondary Qualities
This argument is based on a distinction between primary and secondary qualities of physical objects. Primary qualities are taken to be those that objects have independent of perceiver, e.g. shape, size or mass. Secondary qualities are those that depend on a perceiver, such as smell or color. The argument comes in two forms. One is directed against naïve realism (supporting scientific realism) and the other one against direct realism (in support of indirect realism). As my theory counts as naïve and direct realism, I shall consider both forms.
II.4.1 Against Naïve Realism
The argument against naïve realism relies on Ockham’s razor as an “umpire” in metaphysical questions:
[W]e hould not admit the existence of a sort of property if we can avoid it. […] For we can say that the sorts of explanation that contemporary physics offers give us no need to suppose that the secondary qualities are independent properties of physical objects. (Dancy 1991, p. 148)
In other words, the explanations of physics that operate with particles, waves, forces etc. should not be complemented by additional, non-physical qualities such as smell or color; the smell should be taken as molecules binding to specific receptors and the color as certain wavelengths of the electromagnetic spectrum, period. According to this view, the secondary qualities exist only in the perceiver, not in the objects. The consequence is that naïve realism becomes untenable and hence that people should be aware that the secondary qualities they ascribe to objects exist in their minds only.
Two points can be replied to rebut scientific realism. First, the distinction between primary and secondary qualities might be unsound. For how do we know that objects do not retain their secondary qualities when unperceived? The problem is that we cannot – or can hardly – conceive of a world in which only primary qualities exist but secondary ones do not. Would it be a world similar to the Matrix, with tables retaining their shape but displaying the wavelengths of the light reflected by them with some kind of abstract information? What could that look like? Whatever is visible apparently must have some kind of color (or else it would be invisible). Furthermore, even if colors and smell could be displayed in a purely abstract way, why not shape and size as well? In this light, the primary-secondary-distinction seems arbitrary.
The second point is that both primary and secondary qualities seem to present themselves to us with equal force in perception. To account for that, and if secondary qualities do not belong to the objects, it seems that we must adopt some form of indirect realism. For surely, something looks colored, and if it cannot be the physical object, it seems it must be some non-physical object like a sense-datum. But if one wishes to stick to direct realism, for which there are many reasons even beyond those I give in this essay, this option is not viable.
In summary, I take it that the arguments for scientific realism are too weak to overthrow naïve realism.
II.4.2. Against Direct Realism
There are three main lines of the argument from the illusoriness of secondary qualities against direct realism (cf. Huemer 2001, p. 135-146).
1. The problem of color variation under different lighting conditions.
Colors look different under different lighting conditions. If an object has one objective distinct color, there seems to be no way for us to determine that color. Why then not drop the notion of objective color at all and understand color as a property of sense-data rather than physical objects?
There is an easy and common sense answer to that from a direct realist point of view:
[W]hat are the conditions under which we perceive the true colors of objects? The obvious answer is: normal lighting conditions. That means reasonably (but not blindingly) bright, white light. So the pink look of objects under red light is just an illusion, as is the bluish look of objects in dim light. (Huemer 2001, p. 143)
True, “normal lighting conditions” does not specify an exact mixture of wavelengths. Indeed, one and the same object will look slightly different if the wavelengths vary slightly, but this is still in accordance with the content-satisfaction principle of PCDR.
2. The problem of variations of color experiences among humans and animals
It is a fact that people sometimes disagree about a color, even if lighting conditions are constant. It is also a fact that other species perceive colors differently than we do, fail to perceive colors we perceive or perceive colors we do not perceive (like UV or IR radiation).
The indirect realist chimes in here, claiming that the best explanation for these variations is that colors aren’t properties of things themselves, but of sense-data. But this need not be the answer. I suggest that direct realism has an at least equally good answer: namely that different color experiences are the result of different qualia of the perceptual experiences, not of different sense-data. A quale (pl. qualia) is, roughly speaking, “the way an experience feels like”. A person might take a shirt to be red (and have a red* quale), while another believes it to be orange (because she has an orange* quale). From this does not follow that the shirt has no (true) color; the differences in qualia might be due to differences in the retina or the neuronal transmission. In animals, we know quite well that different architectures of their visual system accounts for their different quales.
3. The problem of colorlessness of submicroscopic particles
There is widespread agreement that atoms and subatomic particles are colorless. From this, another objection to the direct realist view arises (cf. Huemer 2001, p. 145):
- If an object is composed entirely of parts that are colorless, then that object is colorless.
- All middle-sized physical objects are composed entirely of subatomic particles (protons, electrons, neutr ons, and so on).
- Subatomic particles are colorless.
- Therefore, all middle-sized physical objects are colorless.
Using (4) as a premise it is further argued that:
- Colorless objects are invisible.
- We perceive physical objects.
- Therefore, the objects we directly see must be some kind of sense-data caused by the physical objects.
Premises 2 and 3 are uncontroversial. Premise 5 can be debated (cf. Huemer 2001, p. 139-140), but I won’t do this here. Instead, I aim my guns at premise 1, on which all else hinges. Premise 1 completely ignores the possibility of emergent qualities. An emergent quality is one that a thing has but all its constituent parts lack. This seems precisely to be the way with colored objects. Now of course premise 1 is widely applicable in cases of middle-sized to microscopic objects. If I create a glass plate by melting myriads of tiny, transparent glass shivers, the plate will be transparent as well (barring chemical modifications during the melting process). Conversely, when I tear a sheet of red paper into pieces, the pieces are all red. But from those experiences it does not follow that the principle applies equally to sub-microscopic or even subatomic particles. We have no direct acquaintance with them, so it is hard to assess the putative changes they undergo when composed to a greater whole.
I suggest to construe color as an emergent quality. This fits well with the following color theory, which I find very plausible:
Colors are dispositions that physical objects have to reflect light in certain ways.
Those dispositions, of course, depend on the kinds of atoms, their chemical bonding and arrangement etc. These thoughts can and ought to be elaborated more, but for now they should suffice to show that a direct realist account of color perception can be held reasonably.
 Of course I might be hallucinating the mountain, but that possibility does nothing to refute the realist assumption that there are mountains. In fact, the very notion of “hallucination” presupposes the existence of a mind-independent world. How else could I make a conceptual distinction between hallucinated and existing objects? By the way, this question is one of the severe problems Idealism and Phenomenalism suffer from. As they deny the existence of physical objects altogether, the term “hallucination” becomes conceptually void for those theories.
 Which I borrow mainly from (Huemer 2001, p. 60-65)
 Of course, in such cases of apparent hallucination we cannot speak of „seeing“ or „perception“.
 This is presumably even the case when we have no concept of the perceived object. Accordingly, Robert Audi differentiates between two belief-forming perceptual modes, objectual perception and propositional perception (Audi 2002, p. 17-19). He defines objectual perception as perception to be and propositional perception as perception that. For example, when a little child encounters a dryer for the first time, she will not believe that she sees a dryer, but probably she will believe this thing to be something that makes noise.
 Of course, it might be the case that by chance I hallucinate exactly what is before my eyes anyway. What differs from perception here is the lack of an appropriate causal connection.
 By saying that, I do not wish to deny that we can become aware of our perceptual experiences if we wish to. But that is not what happens normally in perception.
 see (Huemer 2001, Ch. VII-VIII)
Audi, Robert. 2002. Epistemology: A Contemporary Introduction. 2 edition. New York: Routledge.
Crane, Tim. 2005. “What Is the Problem of Perception?” Synthesis Philosophica, no. 20:237–64.
Dancy, Jonathan. 1991. Introduction to Contemporary Epistemology. 1 edition. Oxford, UK ; New York, NY, USA: Wiley-Blackwell.
Huemer, Michael. 2001. Skepticism and the Veil of Perception. Lanham, Md: Rowman & Littlefield Publishers.
Hume, David. 1975. “An Enquiry Concerning Human Understanding.” In Enquiries Concern- Ing Human Understanding and Concerning the Principles o FMorals, edited by L.A. Selby-Bigge. Oxford: Clarendon.