Available via license: CC BY-NC
Content may be subject to copyright.
DOI: 10.1111/phis.12176
ORIGINAL ARTICLE
Do looks constitute our perceptual evidence?
Harmen Ghijsen
Radboud University
Correspondence
Harmen Ghijsen, Radboud University.
Email: h.ghijsen@ftr.ru.nl
Funding information
Nederlandse Organisatie voor Weten-
schappelijk Onderzoek
Abstract
Many philosophers take experience to be an essential
aspect of perceptual justification. I argue against a spe-
cific variety of such an experientialist view, namely,
the Looks View of perceptual justification, according
to which our visual beliefs are mediately justified by
beliefs about the way things look. I describe three
types of cases that put pressure on the idea that per-
ceptual justification is always related to looks-related
reasons: unsophisticated cognizers, multimodal iden-
tification, and amodal completion. I then provide a
tentative diagnosis of what goes wrong in the Looks
View: it ascribes a specific epistemic role to beliefs
about looks that is actually fulfilled by subpersonal
perceptual processes.
KEYWORDS
amodal completion, dogmatism, looks view, multimodal identifi-
cation, perceptual justification, perceptual evidence, subpersonal
processes
1 INTRODUCTION
Many views about perceptual justification grant perceptual experience an essential role in provid-
ing (part of) the evidence required for having justified perceptual beliefs. According to these views,
the belief that there is now a laptop in front of me is, at least in part, justified in virtue of my percep-
tual experience of that laptop. Such “experientialist” views of perceptual justification have been
defended in different forms: either perceptual experience itself constitutes the perceptual evidence
This is an open access article under the terms of the Creative Commons Attribution-NonCommercial License, which permits use, distribu-
tion and reproduction in any medium, provided the original work is properly cited and is not used for commercial purposes.
© 2020 The Authors. Philosophical Issues published by Wiley Periodicals LLC
132 wileyonlinelibrary.com/journal/phis Philosophical Issues. 2020;30:132–147.
GHIJSEN 133
required for perceptual justification (Berghofer, 2020; Chudnoff, 2018; Conee & Feldman, 2004;
Huemer, 2001; Moretti, 2015); or perceptual experience gives rise to distinct perceptual seemings,
which then constitute the required perceptual evidence (Brogaard, 2013,2018; Pace, 2017;Reiland,
2015;Tucker,2010); or perceptual experience immediately justifies propositions about perceptual
looks, which then constitute our perceptual evidence (McGrath, 2017,2018). All of these views take
perceptual experience to be an important and sometimes even crucial nexus in the transitioning
from perceptual input to perceptual belief, and, because of this, all of these views grant percep-
tual experience an important role in providing perceptual justification. I think it’s very natural to
arrive at this idea of perceptual justification on the basis of reflection on how we ordinarily seem
to form beliefs on the basis of perception. I also think this idea might be quite radically mistaken.
In what follows I will focus specifically on the recent “Looks View” of perceptual justification
(McGrath, 2017,2018), which has the benefit of nicely accommodating a lot of our discursive prac-
tices involving perceptual reasons. In section 2, I set out the Looks View in more detail, and briefly
discuss its benefits over another main experientialist view of perceptual justification, namely, dog-
matism. In section 3, I then attempt to make plausible that there are cases in which we can have
perceptual justification even if we don’t have sufficiently strong looks-related reasons. In section
4, I conclude by providing a tentative diagnosis of what goes wrong in the Looks View: it ascribes
a specific epistemic role to beliefs about looks that is actually fulfilled by subpersonal percep-
tual processes.
2 THE LOOKS VIEW OF PERCEPTUAL JUSTIFICATION
The “Looks View” of perceptual justification has been defended in a series of papers by Matthew
McGrath (Comesana & McGrath, 2016; McGrath, 2017,2018). According to the view, everyday
perceptual beliefs, such as the belief that there is an apple in front of me, are mediately justified
on the basis of beliefs about the way these things look. To be precise, McGrath limits his view
to what he calls simple visual beliefs, visual beliefs that “(1) [...] are manifestations of a stable
disposition to categorize the perceived object as F upon having certain sorts of visual experience,
[...] and (2) [...] enjoy the phenomenology of “just seeing that the thing is F”” (McGrath, 2018,
p. 111). McGrath’s claim is thus that simple visual beliefs, like the belief that the thing in front
of me is an apple, are mediately justified by means of beliefs about how things, like the apple,
look. But note that McGrath hopes to extend the account to perceptual beliefs and appearances
in general, and not just visual beliefs and looks (2018,p.112n5).
A first thing to get more clear about, is the type of “look” that is relevant to the view under
consideration. The way an object looks is not supposed to be a property of the viewer’s experi-
ence, but a publicly available property of the object. However, looks also don’t just consist in a
combination of view-point independent properties like shape, size and color, as one might agree
that an object is not black but still contend that it looks black. Instead of either being a property
of experience or a view-point independent property, McGrath (2018, p. 130) suggests that looks
might be identified with view-point relative cues to view-point independent properties like shape,
size and color. Thus, if I believe that the object in front of me is an apple, this belief depends on
the belief that apples have a specific sort of look, a look which consists out of a constellation of
view-point relative cues.1
Now that we have at least some grip on the type of look under consideration, we should ask what
having beliefs about those looks comes down to. What does it take to, say, believe that apples have
a specific sort of look? One thing that’s certainly not necessary is being able to produce a list of
134 GHIJSEN
features that could distinguish apples from non-apples. If this were a necessary requirement for
having beliefs about looks, then many people would fail to have beliefs about the looks of objects
which they are nevertheless certainly capable of recognizing. Instead, McGrath (2017,pp.16–7)
suggests that a helpful way of thinking about beliefs about looks can be found in, for instance,
Papineau’s (2006) work on perceptual concepts. Having a perceptual concept involves having a
sensory template that was set up on one’s original encounter with the referent, a template that
can then be reactivated through a further encounter with the referent, or even autonomously
through imagination. The best way to describe the relevant type of beliefs about looks would then
be as believing that apples look, or are, W, where Wrefers to the (visuo-)perceptual concept of
an apple.
All of this provides us with some necessary background for understanding the view that visual
beliefs are mediately justified by beliefs about looks. But what are the reasons for accepting the
view? First, patterns of defeat and epistemic dependence fit nicely with the idea that simple visual
beliefs are supported by beliefs about looks (McGrath, 2018, pp. 120–3). Suppose I’m looking at a
clarinet and believe it to be an oboe until someone informs me that I have been mixing up the way
clarinets and oboes look (perhaps because of reading an otherwise reliable book that mislabeled
the pictures). My earlier justification for the visual belief that the thing in front of me is an oboe
will now be defeated, because my justification for the belief that the thing looks like an oboe is now
defeated. And there’s an even more interesting case: suppose I’m justified in suspending judgment
on whether the thing in front of me looks like an oboe. Again, in that case it seems I cannot have a
justified visual belief that the thing in front of me is an oboe. Just imagine someone saying: “Well,
I’m not sure whether it looks like an oboe but, nevertheless, it’s an oboe”. We would immediately
assume that this person had some other way, distinct from visual recognition, to ascertain that the
thing is an oboe. So simple visual beliefs appear to epistemically depend on beliefs about looks in
just the way the Looks View would predict.
Second, our dialectical, explanatory and recognitional practices are also nicely in line with the
idea that simple visual beliefs are supported by beliefs about looks (McGrath, 2018, pp. 124–6). In
cases of (visual) disagreement, people can appeal to looks to convince their opponents that the
thing in front of them really is, say, a hoverfly rather than a wasp; in cases of mistaken beliefs,
one can explain one’s beliefs by appealing to the fact that the thing really looked like a wasp; and
in cases where one is unsure what one is looking at, one can explicitly consider various looks-
propositions to base one’s best bet on. All of this seems to count in favor of the Looks View of
perceptual justification.
Despite these positive features, the Looks View goes strongly against what might be called the
dominant theory of perceptual justification, i.e., dogmatism, which holds that at least some simple
visual beliefs are immediately justified on the basis of visual perceptual experience. One important
difficulty for dogmatism consists in spelling out which simple visual beliefs enjoy such imme-
diate justification: e.g., whether such beliefs need to have the same contents as the experience
(McGrath, 2018; Silins, 2011,2013), or whether you need to be seemingly aware of a truth-maker
for the content that is presented as true by the experience (Berghofer, 2020; Chudnoff, 2018). Here
I will simply skip these difficulties and assume that there is a way in which a dogmatist can delin-
eate a certain class of simple visual beliefs for which it is also intuitive to hold that they can be
immediately justified by perceptual experience.2
The dogmatist view has a lot of plausibility. It fits the phenomenology of just seeing that
something is, say, an apple, without the need for any further reflection. It also makes sure that
perceptual justification is available for unsophisticated perceivers, as it doesn’t posit any further
complex beliefs on the basis of which the simple visual belief is justified. And it appears to have
GHIJSEN 135
anti-sceptical punch: if a perceptual experience can immediately justify beliefs of the sort that
something is an apple, then we have a large commonsense foundation (beyond the foundation
of mere appearances) on which to ground further beliefs. Nevertheless, according to McGrath,
there is good reason to think that dogmatism is false.
This reason comes from further reflection on the clarinet/oboe mix-up case. In this case (where
a subject believes a clarinet to be an oboe based upon a justified false belief about the looks
of oboes) the subject seems to have a justified belief that the thing in front of him is an oboe.
The Looks View explains this result by pointing to the justified belief about the looks of oboes
on which the visual belief is based. However, it’s unclear in what way the dogmatist could also
explain this result in terms of immediate justification. How could the perceptual belief that the
thing is an oboe be immediately justified by perceptual experience if the thing actually is a clar-
inet and clearly looks like a clarinet as well?3Note that one line of response, that the subject
is only immediately justified in believing that the thing has a certain combination of low-level
view-point independent properties (size, shape, color, etc.), immediately diminishes the attrac-
tiveness of dogmatism as a theory that can explain perceptual justification for ordinary perceptual
beliefs.
The mix-up case thus presents a strong case for allowing the possibility of mediately jus-
tified (looks-based) perceptual beliefs that are not also immediately justified. All that is now
further needed for a strong argument against dogmatism, is the claim that the mix-up case
is relevantly similar to ordinary cases of perception. Even in cases where there’s no mix-up,
subjects will still have learned about the looks and labels of the objects they recognize, simi-
lar to the way the subject has (mistakenly) learned about those things in the mix-up case. But
this means that we can explain ordinary cases of perceptual justification by means of looks-
related reasons as well. There would then be no further need for an immediate justification
of those perceptual beliefs; this type of immediate justification would simply be epistemically
superfluous.
In addition to this indirect line of argument against dogmatism, McGrath (2017,2018,p.128)
also appeals to the following direct argument against the importance of immediate justification.4
Suppose one has a justified belief that this is an apple, at least partly based on the knowledge that
apples look W. What happens if we take away this knowledge, and even take away the subject’s
prima facie justification for the belief that apples look W? In that case, it seems the subject will
no longer be justified in believing that this is an apple. But if the subject really had immediate
justification for believing that this is an apple, then taking away the looks-related reason (that
apples look W) should, presumably, still leave this immediate justification intact. Given that this
does not appear to be the case, we should conclude that the subject did not have any immediate
justification to begin with.
3 PERCEPTUAL JUSTIFICATION WITHOUT LOOKS
I take the previous two lines of argument to raise serious worries for dogmatism about perceptual
justification. However, I also think that a line of argument similar to the first can be used to argue
against the Looks View as well. To that end, I will first attempt to provide three cases in which
subjects lack justification for (sufficiently strong) looks-related propositions, but still have justi-
fied perceptual beliefs (in parallel to McGrath’s case of mediate justification without immediate
justification). In section 4, I will then attempt to diagnose where the Looks View has gone wrong,
despite its intuitive second line of argument.
136 GHIJSEN
3.1 Unsophisticated cognizers
One case that is often used against accounts of mediate perceptual justification is that of unso-
phisticated cognizers, like non-human animals and small children. If perceptual beliefs are always
mediately justified by, for instance, beliefs about looks, then that might make perceptual justifica-
tion too difficult to achieve for unsophisticated cognizers. Such cognizers appear capable of form-
ing perceptual beliefs, but it is not at all clear that they are also capable of forming the required
beliefs about looks.
Now, one thing that counts in favor of the Looks View is that it construes beliefs about looks
as involving, for instance, a perceptual concept. Believing that apples have a specific look thus
comes down to thinking that apples look, or are, W, where Wrefers to a perceptual concept. This
already makes it far more plausible that even unsophisticated cognizers are capable of having the
required beliefs about looks.
Nevertheless, there is still reason to be careful in ascribing such beliefs to unsophisticated cog-
nizers. A belief about looks shouldn’t just be identified with the disposition to recognize (or cat-
egorize) an object by means of vision. The belief about looks is in fact supposed to be the thing
that explains this disposition; according to the Looks View, it is because we believe that apples
look Wthat we classify this W-looking thing as an apple. But it is unclear whether a belief about
looks is necessarily prior to a disposition to recognize; why couldn’t a cognizer first be able to
recognize and only later be able to think (in a minimal sense) about the way in which (s)he was
able to recognize?
To illustrate this point, think about empirical studies investigating source knowledge in chil-
dren.5In a study by Gopnik and Graf (1988), 3-year old children performed significantly worse
than 5-year olds when they had to identify the source of their beliefs about the contents of a
drawer in a forced choice between having seen the contents, being told about the contents, or hav-
ing inferred the contents—although they did perform above chance. In another study (O’Neill &
Chong, 2001), experimenters were just perplexed about young children’s responses to questions
about the sources of their knowledge: “[C]hildren not only responded with irrelevant answers
(which could be expected if they simply did not know the answer), but also at times by demon-
strating or naming an action that could not possibly have led to the knowledge in question. To
watch a child sniff a swimming pool and tell you that that is how they found out it contained cold
water is quite striking!” (pp. 812–3).
Now, of course these results do not prove that small children don’t have beliefs about the looks
of objects. The set-up is such that small children are required to verbally cite or demonstrate that
they used a specific sensory modality to come to know, say, the color of a ball. This seems to be
far more demanding than just believing that the ball looks W. It’s possible that children have
such justified beliefs about appearances, yet are still unable to connect those appearances to the
relevant sources.
Nevertheless, the results are also compatible with the idea that children simply do not have
a good grasp of the different types of appearances (visual, auditory, etc.) an object might have.
If that idea is correct, then children might not be capable of having a belief involving a specific
visuo-perceptual concept of an object. Insofar as there is a belief about appearances at play then, it
would be one that uses a perceptual concept which encompasses all of the modalities: this thing
is (e.g.) apple-ish.6
How bad would this result be for the Looks View of perceptual justification? Let me mention
two possible lines of response. First, although the result would go against the letter of the Looks
GHIJSEN 137
View, perhaps it is still in line with its spirit. After all, even if there is no specific looks-related
reason in this construal of the case at hand, there is still an appearance-related reason which can
justify the relevant perceptual belief. A proponent of the Looks View could thus maintain that
theories endorsing immediate perceptual justification should still be rejected, at least for the types
of perceptual beliefs under discussion (where an object is categorized as F). But note that the type
of appearance-related reasons we are now accepting as the relevant bases for the Looks View are
quite general in nature. As such, even the chicken-sexers or blindsighters of philosophical lore
might surprisingly have what it takes to form mediately justified perceptual beliefs.7
Second, even if small children do not have the suggested type of looks-related reasons for their
perceptual beliefs, we need not suppose that the same goes for adult perceivers. Indeed, McGrath
(2017, p. 28) explicitly considers it an open possibility that adults are in need of more demanding
justifications for their beliefs than children are. However, even if this were the case for all-things-
considered justification, it would still be surprising if adults somehow lost the prima facie justi-
fication available to children. What’s more, if we can find an account of perceptual justification
that fits both children and adult perceivers, this could give it an edge over the Looks View.
But let’s not exaggerate the point against the Looks View so far. It’s possible that children, or
unsophisticated cognizers in general, don’t have the type of beliefs the view proposes as necessary
for perceptual justification, but the adduced experimental evidence certainly does not prove this.
The experimental evidence merely illustrates the possibility of perceptual justification without
looks-related beliefs. The next sections will thus look at different cases that also put pressure on
the idea that perceptual justification requires beliefs about looks.
3.2 Multimodal identification
In the previous section I focused on children’s possible inability to form specific looks-related
beliefs about objects rather than more general appearance-related beliefs. This section, in con-
trast, will focus on cases where the relevant appearance-related beliefs might be insufficient to
epistemically support the relevant perceptual beliefs, given that the categorization of the object
as F actually depends on multimodal evidence that need not be accessible to the subject. To get
the relevant point across, I will sometimes use a slightly extended version of the Looks View by
applying it to different modalities (i.e., justification of auditory beliefs by means of how things
sound, justification of olfactory beliefs by means of how things smell, etc.).
It should come as no surprise that the way we ordinarily perceive the world is not unimodal
but multimodal. The existence of cross-modal illusions shows that we use and integrate infor-
mation from all of our senses to acquire our beliefs about the world, rather than relying solely
on just one of the modalities. For instance, in the ventriloquism effect, the perceived location of
a certain sound source is influenced by what can be seen as the source, such as perceiving the
voice of the ventriloquist as stemming from the puppet’s mouth (Alais & Burr, 2004). Or think
about the parchment-skin illusion, in which people perceive their skin as being drier (like parch-
ment) because the sound of rubbing their hands together is modified in a specific way (Jousmäki
&Hari,1998). Cross-modal integration can even go as far as creating a misperception of one flash
of light as multiple flashes of light when multiple tones are heard, or multiple taps on the skin are
provided (Shams, Kamitani, & Shimojo, 2000; Violentyev, Shimojo, & Shams, 2005).8Although
the experimental set-ups in such illusions deliberately create misperceptions, they do so with the
aim of finding out how multimodal integration usually leads to (more) accurate perception of
the world.
138 GHIJSEN
The fact that we have good evidence for multimodal integration puts pressure on the idea that
we can sensibly delineate purely visual beliefs as manifestations of a stable disposition to catego-
rize an object as F upon having certain sorts of visual experience. In many cases, we’ll be relying
on more than just one modality for our identifications. This will make the relevant disposition
to categorize either dependent on more than just visual experience, or it will problematize the
whole notion of purely visual experience. Even so, one might expect to salvage the Looks View by
focusing on perceptual beliefs in general as manifestations of dispositions to categorize objects as
F upon having certain sorts of multimodal experiences.9
However, multimodal integration also points to a different problem with the Looks View. In
many cases of classification we won’t be aware of relying on other modalities next to the one we
deem most important: that is why we are so surprised about cross-modal illusions. Who would
have thought that the amount of flashes we perceive can be influenced by the amount of sounds
that we hear? The fact that we are often unaware of the influence of other modalities means that
it’s not at all clear that we have all of the relevant beliefs about appearances that the Looks View
would require for perceptual justification.
Take the McGurk-effect as an illustration (McGurk & MacDonald, 1976). When subjects hear
a voice articulating a particular syllable (e.g., [ba]) while simultaneously looking at a video of a
face articulating another syllable (e.g., [ga]), they often perceive the sound of a different syllable
being spoken (in this case [da]). The perceived sound thus comes about as the result of integrating
the incongruent auditory and visual stimulus.10 But, if you’re like me, then you will probably lack
beliefs about the way (lip movements for) particular syllables like [ba], [ga] and [da] look. There’s
no doubt I could come to form such beliefs by visualizing the movements of my mouth while
articulating [ba], [ga] and [da], but, prior to such a visualisation effort, I wouldn’t describe myself
as believing that those syllables looked a particular way W. If that is correct, then I lack some of
the appearance-related reasons that should be relevant in justifying my belief that the syllable [da]
was spoken.
One important response to this particular example (as well as to other cases of cross-modal
illusion) holds that there is actually a different appearance-related reason doing the justificatory
work. After all, the case is described as a cross-modal illusion precisely because subjects experience
a[da] sound rather than a [ba] sound. This means that the relevant appearance-related reason can
be constituted by the beliefs that (1) “[da] sounds W” and (2) “this sounds W”. The perceptual belief
that the syllable [da] was spoken could then be justified based on these justified beliefs about the
way things sounded, in line with an extended version of the Looks View of perceptual justification.
This response works for this particular type of example. Note, though, that it surprisingly dis-
misses part of the cognitive story as epistemically unimportant: the way the syllables look is no
longer given any epistemic role, though they do have an impact on the experience that provides
the relevant appearance-related reason. Even disregarding this point, it also seems that there can
be different cases in which the same type of response doesn’t work. What’s needed is a case of
perceptual categorization in which (1) two or more lines of evidence from different modalities
combine to sufficiently support a perceptual belief together (and not individually), and in which
(2) the subject does not have the relevant beliefs about appearances for at least one of those lines
of evidence.
We can imagine such a case for speech perception as well. Suppose that you’ve just perceived
someone uttering a certain syllable, but that what you heard was actually ambiguous between
[ba] and [da]. Nevertheless, you might still form a justified perceptual belief that the uttered syl-
lable was [da] because of the simultaneously perceived lip movements. This could be so even if
you were not aware of the fact that you were using those lip movements as evidence for your
GHIJSEN 139
perceptual belief. In a case like this, the sound-related reason appears insufficient to fully support
the perceptual belief even though there is no accessible looks-related reason (assuming that you,
like me, do not have beliefs about the looks of syllables at your disposal).
Two important lines of response remain open for the proponent of the Looks View. First, the
proponent might maintain that we do have beliefs about the looks of spoken syllables. The fact
that we are not aware of using them merely shows that these beliefs are unconscious. But that
is compatible with them still being used as reasons, albeit reasons of which we are not aware.
The question now is: do we really want to commit ourselves to the thesis that subjects have such
modality-specific beliefs for all of the relevant cases of multimodal identification—even if they’re
not able to cite those reasons upon being asked (for instance, what does parchment skin sound
like)? We should not commit to this idea too easily just because subjects are able to use the rel-
evant information in arriving at their perceptual beliefs. After all, part of the reason of how we
perceive 3D shapes is related to the visual system’s assumption that lighting stems from above,
but that doesn’t mean that we should commit to our having an unconscious belief that lighting
stems from above (Lyons, 2016). Not all information used by our subpersonal perceptual systems
is appropriately seen as part of the contents of our beliefs.
Second, the proponent of the Looks View might also deny that the envisaged scenario is actu-
ally possible. If we accept that cases of cross-modal illusions are cases in which modality-specific
experiences are altered by multimodal integration, then we might expect that this is bound to
happen in the above scenario as well. That is, if the visual evidence of the lip movements is really
taken into consideration by the visual system, then this will have an effect on how the sound is
experienced. It is thus impossible that the experienced sound would remain ambiguous between
[ba] and [da] if the visual evidence already clearly points to [da].
Note that this response surprisingly commits the Looks View to a specific empirical hypothesis
about multimodal integration and perceptual experience. But once we elaborate on this hypothe-
sis, it immediately becomes less plausible. Recall that “looks” and “sounds” on the extended Looks
View are meant as view-point relative cues to view-point independent properties. The hypothesis
under consideration thus posits that the subpersonal information provided to the visual and audi-
tory system, which in combination support the belief that [da] was uttered, gives rise to a specific
auditory experience of view-point relative cues that are now on their own sufficient to conclude
that [da] was uttered. This seems like an illicit boosting of a specific line of evidence, and it appears
unclear why our perceptual system would go through that trouble.
Perhaps the most sense could be made of the suggestion if we accept that the combination
of subpersonal auditory and visual information gives rise to a genuinely multimodal experience
in which the [da] utterance is represented without a specific unimodal representation of that
utterance. The relevant appearance-belief would then be something like “this syllable is [da]-ish”,
without this belief specifically representing looks, sounds, or other modality-specific view-point
dependent properties.11
This final suggestion is similar to the one in the previous subsection, where it was used to
explain small children’s possible inability to form beliefs about modality-specific appearances. It
also has the same drawbacks: if we actually use such general appearance-beliefs as evidence to
categorize objects as F, then such evidence might even be available to subjects who appear to be
devoid of experience, such as blindsighters and chicken-sexers.
Moreover, we’ve arrived at a point where the Looks View might have lost a lot of its intuitive
motivation: from a view that said our perceptual beliefs were justified by beliefs about how things
look, sound, etc., we’ve now arrived at a view that claims our perceptual beliefs are justified by
beliefs about what type of object (multimodally) appears to be in front of us. Not only is it unclear
140 GHIJSEN
FIGURE 1 An example of amodal completion
what those types of beliefs are exactly, they are also very far removed from some of the original
motivations for the Looks View, namely, that it captures a lot of our discursive practices related
to perceptual reasons. We don’t often appeal to what a thing generally (as opposed to modality-
specifically) appears to be to help recognize it as, say, a hoverfly rather than a wasp. Nor do we
attempt to convince others of the fact that an object is a hoverfly rather than a wasp by appealing
to its general appearance, or use general appearances to explain our perceptual beliefs (“it really
had that waspish appearance!”). Whichever way we turn, cases of multimodal identification put
pressure on the (extended) Looks View of perceptual justification.
3.3 Amodal completion
A final type of case that puts pressure on the thesis that visual beliefs are always mediately justified
by beliefs about looks, is that of amodal completion.12 In amodal completion, we represent parts
of an object even though we receive no sensory stimulation from those parts. For example, when
looking at a circle that is partly hidden behind a large rectangle, we still represent the circle as a full
circle—even though the shape could just as well stop at the intersection with the larger rectangle
(see figure 1). But amodal completion also occurs for the other senses, as, for instance, one can hear
a tune continuing ‘under’ the sound of a large bang. For our purposes, what’s especially relevant is
that amodal completion is ubiquitous in ordinary perception (just as multimodal integration is),
and that we’re often not aware of the fact that it happens (just as with multimodal integration).
Consider, now, a case of amodal completion in more detail. Suppose I’m looking at a circle partly
hidden behind a rectangle and come to believe, unreflectively, that it’s a circle (just imagine it to
be two pieces of paper lying on top of another). I take it that this perceptual belief can be justified,
even if there is a possibility, not excluded by my evidence, that the circle is in fact a weird sort of
GHIJSEN 141
different shape (part circular, part rectangular for instance). Note that, if one doesn’t accept this
idea, then a lot of perceptual beliefs will come out as unjustified or at best inferentially justified
by things other than looks-related reasons, given the ubiquity of amodal completion in ordinary
cases of perception.13 So, supposing that the perceptual belief that it’s a circle is justified, what is
the looks-related reason for this belief?
Recall once more that the relevant belief about looks is one that involves a certain perceptual
concept related to view-point relative cues. Now suppose one has the relevant belief that circles
look W(just imagine a circle, and think about the way it looks). It’s hard to see that the circle
in Figure 1will actually manifest the look Wthat is specific to circles as there only seems to be
three-quarters of Wpresent. To the extent that people will appeal to a looks-related belief here,
these beliefs might even be false and unjustified. Just imagine asking a person why he thought
that the object behind the rectangle was a circle. This person might very well answer: “Because it
looked like a circle”. But if this answer is supposed to express the belief that the thing looked W,
then it seems that it is both false and unjustified.
On a different reading, what people might be expressing is the belief that circles behind rect-
angles look this particular way. We now have a different look, say V, that is relevant for the
case at hand. But here one can start to question to what extent Vcan actually provide the epis-
temic support needed to justify the conclusion that things that look Vare circles. That is, even
if we grant that subjects might have a justified belief that circles (under certain circumstances)
look V, and a justified belief that the thing in front of them looks V, then it is still not clear
that these beliefs provide the necessary support to justify the perceptual belief that this thing
is a circle. After all, other things besides circles could look Vas well (tilted pacman-shapes, for
instance).
So far this example might look a bit artificial. Sure, random geometrical figures could turn out
to be lots of ways based on a partially occluded look, but similar considerations are not applicable
to the natural objects in our environment. For instance, a dog behind a picket fence will be partly
occluded, but surely the visible parts of the dog (its genuine looks) are still sufficient to justify the
perceptual belief that there is a dog behind the fence—even if one accepts that the justification is
a bit weaker in comparison to seeing the full dog. Although I agree with this claim, the proponent
of the Looks View still owes us a story of how this justification comes about exactly. Dogs might
be occluded in any number of ways, and it is unlikely that I believe for each of these ways that
its accompanying look is indicative of dogs (prior to seeing or imagining a dog being occluded
that way). Perhaps we could posit a belief that the actual occluded look is “close enough” to a
stereotypical dog-look, but then we have to explain where the justification for this belief comes
from. Surely the content of this comparitive belief is not a part of the content of the experience,
unlike the content of the beliefs about looks.
And the problems for the Looks View don’t end there. Even if we have beliefs about the stereo-
typical looks of objects, those beliefs don’t always play a role in amodal completion. Figure 2
(from Hazenberg & van Lier, 2016) displays a case where the object in the middle has the look
of an (elongated) elephant, even though we will all agree that this is incompatible with the way
we believe stereotypical elephants look (indeed, that is why we won’t believe that there is one
occluded elongated elephant there despite its look). This example thus indicates that beliefs about
stereotypes need not be involved in the outputs of amodal completion. This will count against
positing such beliefs about stereotypical looks as bases for perceptual beliefs in cases of amodal
completion.14
142 GHIJSEN
FIGURE 2 Amodal completion in opposition to beliefs about stereotypical looks
4SUBJECTS’ BELIEFS VS. PERCEPTUAL ASSUMPTIONS
The previous sections have attempted to make plausible that there are ubiquitous cases of having
justified perceptual beliefs without there being (sufficiently strong) looks-related or, in general,
appearance-related reasons. Although there will also be cases in which looks-related reasons are
present to be used (think of identifying objects from photographs), this is not sufficient to keep the
Looks View up and running. Without an account of many ubiquitous cases of perception, such
as those involving multimodal identification and amodal completion, the view would be severely
hampered at best. So how is it possible that the Looks View appears to be so successful for many
ordinary cases of perception, but appears to fail in others? In this last section I will provide a
tentative diagnosis of this situation.
There’s something intuitive about the idea that we need to have beliefs about looks to visually
recognize objects as apples, oboes, etc. After all, if I did not justifiably believe, for instance, that
oboes looked this way, then it seems I could not have justifiably concluded that the object in front
of me was an oboe. It is only because I know what oboes look like that I am able to recognize an
oboe as an oboe when I see one. As mentioned at the end of section 2, McGrath (2017)uses(an
elaborated version of) this line of reasoning to support the importance of looks in our perceptual
knowledge. Add to this the fact that we ourselves often appeal to looks, sounds, smells, etc. to
justify our perceptual beliefs to ourselves and others, and it might seem that the Looks View has
to be correct.
The above perceptual cases of unsophisticated cognizers, multimodal identification and
amodal completion are all attempts at describing cases in which it is relatively clear that the sub-
ject under consideration does not plausibly have the relevant belief about looks, even though she
is still able to recognize the relevant object as F. In at least some sense, though, even in these cases
subjects do know what an F looks (or sounds, smells, etc.) like: they do have an ability to recognize
Fs on the basis of how they appear, presumably because their perceptual system has learned to
latch on to those specific appearances. This recognitional ability, however, should not be mistaken
for a (justified) belief about looks, as the cases hopefully make plausible that such beliefs could
very well be absent.15
GHIJSEN 143
From these plausible cases of perceptual beliefs without beliefs about looks we can extrapolate
to other cases. Even if subjects do have beliefs about looks at their disposal, that does not mean that
those beliefs are actually operative in forming perceptual beliefs with the “just see” phenomenol-
ogy. In fact, given that the perceptual processing will be the same as in the cases without beliefs
about looks, it’s likely that these beliefs do not play a role as a basis for simple perceptual beliefs.
The role they could play might have more to do with taking the output of the perceptual system
at face value or not: if one believes that a certain look is not indicative of Fs, then one can reject
how things seem. Cases of known illusions could fall under this heading, as could the earlier case
in which one learns that one has mixed up the way oboes and clarinets look.
Still, there might be a lingering doubt. When one first learns to recognize objects as Fs, one
often has to learn to connect the look of the object to a certain verbal category. Although one can
learn to visually recognize a certain type based on repeated exposure to tokens of that type, one
can only recognize, say, an oboe as an oboe if one has learned that that is what this type of object
is called.16 Doesn’t this show that many perceptual beliefs are mediately justified in precisely the
way that the Looks View proposes?
It doesn’t, precisely for reasons outlined above. We don’t have beliefs that specify all of the dif-
ferent ways an oboe might look under different circumstances, including circumstances in which
it is partly occluded. Although we might start off with a belief specifying what this object in front
of us is called (or a belief that specifies what to look for, such as the belief that copperheads have
copper red heads), at some point our perceptual system takes over. With repeated exposure to
examples, our perceptual system automatically outputs a certain category label in response to
certain looks, an output which can even change when one acquires more expertise (Ransom,
2020;Stokes,2020): for instance, bird experts will automatically recognize and classify a kingbird
as a kingbird rather than as (just) a bird. So although beliefs about looks might diachronically
influence our perceptual processes, it would be a mistake to think that they are also epistemically
relevant as bases for our perceptual beliefs.17
What has gone wrong, I want to suggest, is that the Looks View has ascribed a specific epistemic
role to subjects’ beliefs that is actually fulfilled by (the assumptions of) their perceptual processes.
Although beliefs about looks do have specific epistemic functions to fulfill—related to our more
sophisticated practices of recognition, explanation and justification—in cases of “just seeing” that
an object is F, they are not actually operative in producing the relevant perceptual beliefs. Instead,
what’s operative are, at best, assumptions of the perceptual system, relating to which features are
indicative of being F. The fact that we could also have beliefs with similar contents is what creates
an illusion of our perceptual beliefs being justified by beliefs about looks.
What does this tell us about our own practices involving perceptual reasons? First, we certainly
can focus on how things look to decide whether a thing is, say, a wasp or a hoverfly in a case where
we’re unsure about the answer. In such a case, we genuinely are using a belief about looks to come
to a belief about what is in front of us. If the belief about looks is justified, then the belief about
what’s in front of us will be mediately justified. But notice how different this case is from a case
in which you “just see” that there is a hoverfly in front of you.
Second, we can convince other people that something is a hoverfly instead of a wasp by pointing
their attention to its looks. This could either be a case in which we convince those people that only
hoverflies look W(and wasps don’t), or a case in which we help people (by directing their attention
in a certain way) to just see that the thing is a hoverfly. The former case would be one in which the
belief about looks is a genuine basis for the belief about the hoverfly, the latter case is one where
the belief about looks merely influences one’s patterns of attention without being a (direct) basis
for the belief about the hoverfly.
144 GHIJSEN
Third, we can explain and justify our perceptual beliefs by an appeal to the looks of things. Given
that we are often unaware of how our perceptual beliefs exactly came about (e.g., the influence
of other modalities and amodal completion), these justifications will often be only part of the
story at best. One could even suspect that many of these explanations will confabulate looks-
related reasons that were not relevant to the formation of the perceptual beliefs. After all, we know
that people are, in general, prone to confabulate reasons. Ask them why they chose this pair of
stockings over three (unknowingly) identical pairs of stockings, and they will provide you with
reasons (Nisbett & Wilson, 1977); ask them why they thought the person on this picture was more
attractive (even though they just judged the person on the other picture to be more attractive), and
they will provide you with reasons (Johansson, Hall, Sikström, & Olsson, 2005). It should only be
expected that when people are asked why they believed that, say, that person was their friend
Marc, they will also provide you with reasons that need not have actually influenced their belief-
formation (e.g., he really looked the same). This means that we should not take these dialectical
justifications as an accurate reflection of what actually justifies perceptual beliefs.
5CONCLUSION
I have argued against the Looks View of perceptual justification on the basis of several cases in
which it’s plausible that one has perceptual justification without having sufficiently strong justi-
fied beliefs about looks. At the very least, these cases will push the proponent of the Looks View
to refine its relevant notion of beliefs about looks to one that can accommodate cases of unsophis-
ticated cognizers, multimodal identification and amodal completion. However, it seems the more
this type of belief is adapted, the less the view will be recognizable as the intuitive Looks View
with which we started.
I have further suggested that we can extrapolate from the cases of unsophisticated cognizers,
multimodal identification and amodal completion to all cases of simple perception (the “just see”
cases). Even if beliefs about looks are present, we shouldn’t usually expect them to be actually
operative in forming perceptual beliefs. This means that we should not use our discursive practices
involving perceptual reasons as an accurate picture of how our perceptual beliefs are justified. If
that is correct, then experientialist frameworks that do take such practices seriously are definitely
in trouble.
ACKNOWLEDGMENTS
I’d like to thank the members of the research group headed by Bence Nanay at Antwerp University
for their extremely helpful comments to an earlier draft of this paper. I would also like to thank
Matt McGrath for some excellent comments on an earlier version. This work is part of the research
programme “Should we trust our own eyes? Epistemic challenges from cognitive penetration and
implicit bias” with project number 275-20-056, which is financed by the Netherlands Organisation
for Scientific Research (NWO).
ENDNOTES
1Perhaps it would be better to say specific sortsof looks, given that there will be different constellations of view-
point relative cues that will all be indicative of apples. This point will come up in discussion in section 3.3.
2But see Ghijsen (2014,2015) for discussion.
3McGrath (2018, pp. 126–8) discusses and rejects several responses on behalf of the dogmatist which I will not
rehearse here. Also see Vaassen (2016) for more about this perceptual learning problem for dogmatism.
GHIJSEN 145
4Note that McGrath (2017) provides a very detailed discussion of this direct argument that I cannot do full justice
to here. I’ll briefly revisit the argument in section 4.
5I focus on children, but see Shieber (2017) for an argument that non-human animals are capable of visual recog-
nition without being capable of having beliefs about looks.
6Note that this line of thought doesn’t rely on the controversial idea that children are unable to distinguish
between appearance and reality (see McGrath (2017, p. 29) for discussion). The thought here is that children
might only have general appearance-beliefs instead of the specific looks-beliefs that the Looks View posits.
7See Ghijsen (2016) for more discussion of these cases.
8For a small overview of such cross-modal illusions and their philosophical repercussions, see, e.g., O’Callaghan
(2012), Bayne and Spence (2015).
9In fact, if theories of predictive processing are on the right track, then all sorts of expectations besides appear-
ances from several modalities might influence one’s categorization of objects (Clark, 2016; Ghijsen, 2018;Hohwy,
2013). Think about recognizing oranges in the supermarket: why are you justified in believing that those are
oranges rather than, say, large tangerines? Part of the answer might lie in background beliefs or expectations
about where the oranges are usually located in the supermarket, but those beliefs do not appear to influence
the way the oranges look in terms of view-point relative cues. If this goes for many perceptual beliefs, then all
of those will need to be mediately justified by many more beliefs than just beliefs about appearances—if we are
even willing to count the relevant expectations as full-fledged beliefs.
10 The resulting percept is often described as a “fusion” of the auditory and visual stimulus, but see e.g. Tiippana
(2014) for discussion of this interpretation.
11 At this point, the Looks View appears to become dangerously similar to dogmatism involving perceptual seem-
ings (e.g., Brogaard, 2013,2018; Pace, 2017; Reiland, 2015 Tucker, 2010). Such versions of dogmatism claim
that sensory experiences give rise to (possibly amodal) perceptual seemings, which then justify the relevant
perceptual beliefs.
12 See, e.g., Nanay (2018) for more on amodal completion and its philosophical consequences.
13 Note that Chudnoff (2018) accepts this cost, but still believes there to be sufficiently many interesting cases of
immediately justified beliefs that can be explained by means of a specifically constrained dogmatist view.
14 What about positing a belief with conditional content; e.g., if an object has large ears, tusks and a trunk, then
it’s an elephant? Although we might posit such a belief in response to this example, the problem here is (again)
that’s it’s unlikely that we’ll really believe this for all features that our perceptual system takes to be indicative
of being a certain object.
15 See McGrath (2017) for an argument that not just any recognitional capacity will work to counter his “only-
because” argument, and see Shieber (2017) for discussion.
16 See Lyons (2005), Pace (2017) for more on the distinction between phenomenally classifying and describ-
ing kinds.
17 How, then, should we think about perceptual justification? I can see two options one might like to explore.
First, one might accept that perceptual systems quite literally use assumptions about appearances to produce
their outputs. In that case, one might still think that perceptual beliefs are mediately justified. It’s just that
the mediate justifier is no longer taken to be a subject’s belief about looks, but rather an assumptionabout looks
made by the subject’s visual system. In this case, we would be broadening the scope of the kind of entities that can
be justified or unjustified, similar to what (Jenkin, 2020) proposes for representations of core cognition. Second,
one might ground the justification of perceptual beliefs with reference to the reliability of the processes of the
perceptual system (Lyons, 2009). Assuming that multimodal integration and amodal completion are reliable
processes, such a view could easily account for justification in all of the earlier cases. One difficulty this view
faces has to do with explaining what goes wrong when one has used an unjustified belief about looks to calibrate
one’s perceptual processes without making those processes any less reliable. See Vaassen (2016) for a discussion
of this problem.
REFERENCES
Alais, D., & Burr, D. (2004). The ventriloquist effect results from near-optimal bimodal integration. Current biology,
14(3), 257–262. https://doi.org/10.1016/j.cub.2004.01.029
146 GHIJSEN
Bayne, T., & Spence, C. (2015). Multisensory perception. In M. Matthen (Ed.), The Oxford handbook of Philosophy of
Perception (pp. 1–21). Oxford: Oxford University Press. https://doi.org/10.1093/oxfordhb/9780199600472.013.024.
Berghofer, P. (2020). Towards a phenomenological conception of experiential justification. Synthese,197,155–183.
https://doi.org/10.1007/s11229-018-1744-5.
Brogaard, B. (2013). Phenomenal seemings and sensible dogmatism. In C. Tucker (Ed.), Seemings and Justification:
New Essays on Dogmatism and Phenomenal Conservatism (pp. 270–289). New York: Oxford University Press.
Brogaard, B. (2018). Phenomenal dogmatism, seeming evidentialism and inferential justification. In K. McCain
(Ed.), Believing in Accordance with the Evidence (pp. 53–67). Springer Verlag. https://doi.org/10.1007/978-3- 319-
95993-1_5.
Chudnoff, E. (2018). Epistemic elitism and other minds. Philosophy and Phenomenological Research,96(2), 276–298.
https://doi.org/10.1111/phpr.12308.
Clark, A. (2016). Surfing Uncertainty: Prediction, Action, and the Embodied Mind. Oxford University Press.
Comesana, J., & McGrath, M. (2016). Perceptual reasons. Philosophical Studies,173, 991–1006. https://doi.org/10.
1007/s11098-015-0542-x.
Conee, E., & Feldman, R. (2004). Evidentialism: Essays in Epistemology. New York: Clarendon Press.
Ghijsen, H. (2014). Phenomenalist dogmatist experientialism and the distinctiveness problem. Synthese,191(7),
1549–1566. https://doi.org/10.1007/s11229-013-0348-3.
Ghijsen, H. (2015). Grounding perceptual dogmatism: What are perceptual seemings? Southern Journal of Philos-
ophy,53(2), 196–215.
Ghijsen, H. (2016). The Puzzle of Perceptual Justification.Springer.
Ghijsen, H. (2018). Predictive processing and foundationalism about perception. Synthese,(Online first).https://
doi.org/10.1007/s11229-018-1715-x.
Gopnik, A., & Graf, P. (1988). Knowing how you know: Young children’s ability to identify and remember the
sources of their beliefs. Child development, 1366–1371. https://doi.org/10.2307/1130499.
Hazenberg, S. J., & van Lier, R. (2016). Disentangling effects of structure and knowledge in perceiving partly
occluded shapes: An erp study. Vision Research,126, 109–119. https://doi.org/10.1016/j.visres.2015.10.004.
Hohwy, J. (2013). The Predictive Mind. Oxford University Press UK.
Huemer, M. (2001). Skepticism and the Veil of Perception. Lanham, MD: Rowman & Littlefield Publishers.
Jenkin, Z. (2020). The epistemic role of core cognition. The Philosophical Review,129(2), 251–298. https://doi.org/
10.1215/00318108-8012850.
Johansson, P., Hall, L., Sikström, S., & Olsson, A. (2005). Failure to detect mismatches between intention and
outcome in a simple decision task. Science,310(5745), 116–119. https://doi.org/10.1126/science.1111709.
Jousmäki, V., & Hari, R. (1998). Parchment-skin illusion: Sound-biased touch. Current biology,8(6), R190–R191.
https://doi.org/10.1016/s0960-9822(98)70120-4.
Lyons, J. (2005). Clades, capgras, and perceptual kinds. Philosophical Topics,33(1), 185–206.
Lyons, J. (2009). Perception and Basic Beliefs: Zombies, modules and the problem of the external world.USA:Oxford
University Press.
Lyons, J. (2016). Unconscious evidence. Philosophical Issues,26(1), 243–262. https://doi.org/10.1111/phis.12073.
McGrath, M. (2017). Knowing what things look like. Philosophical Review,126(1), 1–41. https://doi.org/10.1215/
00318108-3683602.
McGrath, M. (2018). Looks and perceptual justification. Philosophy and Phenomenological Research,96(3), 110–33.
https://doi.org/10.1111/phpr.12289.
McGurk, H., & MacDonald, J. (1976). Hearing lips and seeing voices. Nature,264(5588), 746–748. https://doi.org/
10.1038/264746a0.
Moretti, L. (2015). Phenomenal conservatism. Analysis,75(2), 296–309. https://doi.org/10.1093/analys/anu153.
Nanay, B. (2018). The importance of amodal completion in everyday perception. Iperception,9.https://doi.org/10.
1177/2041669518788887.
Nisbett, R. E., & Wilson, T. D. (1977). Telling more than we can know: Verbal reports on mental processes. Psycho-
logical review,84(3), 231. https://doi.org/10.1037/0033-295X.84.3.231.
O’Callaghan, C. (2012). Perception and multimodality. In E. Margolis, R. Samuels, & S. P. Stich (Eds.), The Oxford
handbook of philosophy of cognitive science (pp. 92–117). New York: Oxford University Press. https://doi.org/10.
1093/oxfordhb/9780195309799.013.0005.
GHIJSEN 147
O’Neill, D. K., & Chong, S. (2001). Preschool children’s difficulty understanding the types of information obtained
through the five senses. Child Development,72(3), 803–815. https://doi.org/10.1111/1467-8624.00316.
Pace, M. (2017). Experiences, seemings, and perceptual justification. Australasian Journal of Philosophy,95(2), 226–
241.
Papineau, D. (2006). Phenomenal and perceptual concepts. In T. Alter & S. Walter (Eds.), Phenomenal Concepts and
Phenomenal Knowledge: New Essays on Consciousness and Physicalism (pp. 111–144). Oxford University Press.
https://doi.org/10.1080/00048402.2016.1202991.
Ransom, M. (2020). Expert knowledge by perception. Philosophy,95(3), 309–335. https://doi.org/10.1017/
S0031819120000157.
Reiland, I. (2015). Experience, seemings, and evidence. Pacific Philosophical Quarterly,96(4), 510–534. https://doi.
org/10.1111/papq.12113.
Shams, L., Kamitani, Y., & Shimojo, S. (2000). What you see is what you hear. Nature,408(6814), 788–788. https:
//doi.org/10.1038/35048669.
Shieber, J. (2017). Looks and the immediacy of visual objectual knowledge. Analysis,77(4), 741–750. https://doi.org/
10.1093/analys/anx126.
Silins, N. (2011). Seeing through the ’veil of perception’. Mind,120(478), 329–367. https://doi.org/10.1093/mind/
fzr030.
Silins, N. (2013). The significance of high-level content. Philosophical Studies,162(1), 13–33. https://doi.org/10.1007/
s11098-012-9991-7.
Stokes, D. (2020). On perceptual expertise. Mind & Language,Online first.https://doi.org/10.1111/mila.12270.
Tiippana, K. (2014). What is the McGurk effect? Frontiers in Psychology,5,725.https://doi.org/10.3389/fpsyg.2014.
00725.
Tucker, C. (2010). Why open-minded people should endorse dogmatism. Philosophical Perspectives,24(1), 529–545.
https://www.jstor.org/stable/41329456.
Vaassen, B. M. K. (2016). Basic beliefs and the perceptual learning problem: A substantial challenge for moderate
foundationalism. Episteme,13(1), 133–149. https://doi.org/10.1017/epi.2015.58.
Violentyev, A., Shimojo, S., & Shams, L. (2005). Touch-induced visual illusion. Neuroreport,16(10), 1107–1110. https:
//doi.org/10.1097/00001756-200507130-00015.
How to cite this article: Ghijsen H. Do looks constitute our perceptual evidence?.
Philosophical Issues.2020;30:132–147.https://doi.org/10.1111/phis.12176
