Talking about Looks
- 1.5k Downloads
In natural language, looks-talk is used in a variety of ways. I investigate three uses of ‘looks’ that have traditionally been distinguished – epistemic, comparative, and phenomenal ‘looks’ – and endorse and develop considerations in support of the view that these amount to polysemy. Focusing on the phenomenal use of ‘looks’, I then investigate connections between its semantics, the content of visual experience, and the metaphysics of looks. I argue that phenomenal ‘looks’ is not a propositional attitude operator: We do not use it to ascribe propositional attitudes to subjects, but to directly ascribe looks to objects, where looks are relational properties. However, I go on to argue that, given the way we use phenomenal ‘looks’, these relational properties are ultimately best understood as phenomenal relational properties, i.e. in terms of relations involving experiences. Along the way, I endorse Byrne’s argument against Jackson’s claim that phenomenal ‘looks F’ only takes predicates for colour, shape, and distance, and raise the issue of compositionality for the resulting view according to which phenomenal ‘looks F’ is context-dependent in a way that allows it to take a vast range of predicates. I conclude by arguing that these considerations concerning the natural language use of ‘looks’, and in particular its phenomenal use, are water on the mills of phenomenal intentionalism, a position in the philosophy of perception according to which experiences are propositional attitudes with phenomenal looks-contents.
One morning in late summer, Alma looks out of the window of the cottage she and Martha have rented for their vacation. Alma looks at the old farm house across the meadow and the mountains beyond. “It looks as if the neighbours moved away a long time ago, doesn’t it”, she says, but Martha is in the kitchen making coffee and listening to the radio. “Let’s go hiking today”, she calls. The sun is shining but there are dark clouds gathering behind the mountains. “Absolutely,” Alma answers, “but it looks as if we will need our rain gear.” She walks into the kitchen. “What did you think of Bob?”, Martha asks. Bob is an old friend of hers they ran into at the country store the day before. “He looks like Mads Mikkelsen”, Alma says. “I liked the look of his sweater. Old-fashioned. A bit like the old abandoned farm over there.” She gestures in the direction of the house across the meadow. “It looked so red in that weird light.” Martha laughs, but then her face darkens. “It looks as if the stock market is crashing again”, she tells Alma while gesturing at the radio. They sit down at the kitchen table to listen. “We should switch to the other station,” Alma says, “they usually are more up to date.” They switch and learn that the situation has already much improved.
In this little story, there is a lot of talk about looks. Like everybody else, Alma and Martha are using looks-language in a variety of ways; they talk about the looks of objects and how they compare, but they also use ‘looks’ in their descriptions of the states of affairs that might or will obtain in the world around them. Such looks-language is the main topic of this paper. How do looks-locutions work, and what do they mean? Quite a bit has been said about this in the philosophical literature since the 1950s. Most, if not all, of this work has been in the context, and service, of the theory of perception. Claims about looks-language have been made to support a wide variety of accounts of (visual) perceptual experience, from sense data theory via the adverbial theory and intentionalism all the way to anti-intentionalist relational accounts.1
This paper will be no exception here – ultimately, I am interested in looks-language to see whether natural language does, or does not, furnish a ready-made sense of ‘looks’ that can be used to interpret one of the two central claims of a position I have been developing: “phenomenal intentionalism”. Phenomenal intentionalism combines the claim that perceptual experience is a subspecies of belief with the claim that visual experience takes what I have called looks- or Lp-contents, contents of the form x looks F (to S at t), or it looks as if p (to S at t) (cf. Glüer 2009, 2014). It seems to me that previous investigations of looks-language, instructive and insightful as they are, have not always been looking at sufficiently varied examples from natural language. Despite all that has been said about the uses of ‘looks’, there still is a need, it seems to me, for careful exploration of, and theorizing about, looks-language. My aim in this paper is quite modest: To make some progress on these fronts.
I shall proceed as follows. In Section 2, I shall follow Chisholm and Jackson in distinguishing three main uses of ‘looks’: epistemic, comparative, and phenomenal uses. I shall explore these and their interrelations in some detail, and investigate whether their existence amounts to something like ambiguity or polysemy. In the sections to follow, I shall then focus exclusively on the third, or phenomenal, use of ‘looks’. In Section 3, I shall be concerned with its significance, if any, for the theory of perceptual experience. In particular, I shall argue that, prima facie at least, the existence of a third use is compatible with a wider range of positions on perception than one might expect. The only position directly threatened by the very existence of a third use appears to be Martin-style parsimony about looks. But I shall then go on to present a range of additional observations about the way phenomenal ‘looks’ is actually used in natural English. These, or so I shall argue, ultimately do speak both against construing it as a propositional attitude operator as well as against construing it as ascribing mind-independent, or non-phenomenal, relational properties to objects. These observations point towards the conclusion that the best semantics for phenomenal ‘looks’ treats looks as phenomenal relational properties. In the final section, I shall mainly investigate the range of predicates that can be modified by phenomenal ‘looks’, and what such combinations mean. I shall argue with Byrne and against Jackson that phenomenal ‘looks’ can take a vast range of predicates, and raise the issue of compositionality. I shall conclude that nothing in this paper prevents me from modelling looks-contents for experience on natural language phenomenal ‘looks’.
2 The Uses of ‘Looks’
It looks as if the neighbours moved away.
It looks as if we will need our rain gear today.
It looks as if the stock market is crashing.
The evidence indicated by epistemic uses of ‘looks’ need not be perceptual, however. Even though Martha’s evidence for thinking that the stock market is crashing comes through an auditory channel, its epistemic source is not perception, but testimony. Such non-perceptual epistemic uses of ‘looks’ most clearly show a feature of this use stressed by Brogaard (cf. Brogaard 2013, 4; 2015, 239): defeated epistemic looks “disappear”. Once Alma and Martha have switched channels and received a better evaluation of the situation on the stock market, the reasons they initially had for thinking that it was crashing have been defeated and the stock market no longer looks to be crashing. Let’s call this feature “defeated disappearance”.
Epistemic ‘looks’ reports are evidence-bearing. It cannot epistemically look as if p if one’s total evidence indicates that not-p. (...) So, when epistemic ‘looks’ reports are accurate, they give us information about the speaker’s total subjective evidence (Brogaard 2013, 4).
It looks as if the neighbours moved away.
It looks as if the neighbours moved away, but I happen to know that they are hiding in the attic.4
However, there is defeated disappearance even in “attic cases”. There is a strong contrast between (4) and (1) here. While (4) seems fine, asserting (1) after learning that the neighbours are hiding in the attic is just as unacceptable as asserting (3) after getting the better report about the stock market. This becomes very clear, I think, if you imagine that right after Martha tells Alma about the neighbours in the attic, Emma drops by and Alma tells her that it looks as if the neighbours moved away. It would be only natural for Martha to be somewhat upset about that. She might burst out with: “What? I just told you that they are hiding in the attic.” On the weak interpretation, that would not only be impolite, but quite unmotivated.
(4 ′) It looks as if p but I happen to know that not-p.
It looks as if the stock market is crashing, but I happen to know that it isn’t.
I think this is clearly unacceptable. All in all, it thus seems to me that the strong interpretation of epistemic ‘looks’ does somewhat better than the weak interpretation. But endorsing it leaves us with the question of how to make sense of the acceptability of sentences like (4).
A clue can be gotten, I think, from thinking about upset Martha again. She told Alma that the neighbours are hiding in the attic, but Alma nevertheless tells Emma that it looks as if they moved away. Alma can, it seems to me, defend herself by saying something like: “I didn’t say that they moved away, only that it looks that way”. But note that this works only if she stresses ‘looks’. And that suggests that what Alma exploits in her defense is in fact an ambiguity in ‘looks’ – it suggests that the use of ‘looks’ that makes (4) acceptable is not epistemic, after all. Rather, what we have here is a use of ‘looks’ describing the visual appearance of things. If that is correct, it would explain why sentences of the form (4 ′) are acceptable in perceptual cases, but not in cases involving clearly non-perceptual epistemic uses.
Bob looks like Mads Mikkelsen.
Bob’s sweater looks like the old farm house.
Bob’s sweater looks old-fashioned.
It looks as if we will need our rain gear today.
It looks as if it’s going to rain.
(6 ′) ∃y (WLy & has (Bob, y) & has (Mads, y)),
That looks like a tomato.
(10 ′) ∃y (has(o, y) & (WL(y) & ID(C(F, WL, c), y))),
where ‘WL’ is a way of looking, o is the demonstrated object, ‘C(F, WL, c)’ the contextually salient way of looking characteristic of F s, and ‘ID’ is the identity function.
Construing comparative uses of ‘looks’ along lines such as these, we take them to be quantifying over looks. And we explain the comparative use of ‘looks’ in terms of looks, too; i.e. by using ‘looks’. This raises the questions of what the things quantified over in comparative looks-sentences are, and how to interpret ‘looks’ when used to explain what these sentences mean. These uses, it seems, cannot themselves be comparative uses (cf. Maund (1986) and Byrne (2009), 441). There seems to be a third use.
Chisholm and Jackson certainly thought there was a third use. Chisholm called it the “non-comparative use”, while Jackson called it the “phenomenal use”. These days, the terms “non-comparative” and “phenomenal” are often used interchangeably. But the range of expressions Chisholm thought could be used non-comparatively is significantly wider than those that according to Jackson are phenomenal uses. Even though I agree with Chisholm on the range of the third use of ‘looks’, I shall for the most part stick with the label “phenomenal use”, but initially without thereby buying into any particular interpretation of the third use. All I shall initially assume about this use is this: what we do when using ‘looks’ this way is “directly” ascribing looks to things (or scenes) – where “directly” is opposed to “by comparison”. What these looks are is precisely the question I want to initially leave open.
Bob’s sweater looks red.
It looks triangular.
The top line looks longer than the bottom line.
That looks centaurian.
The farmhouse looks old.
The farmhouse looks red and old.
x looks F.
It looks like/as if x is F.
Tom looks red,
It looks as if the stock market is crashing.
It looks as if Tom is red.
*It looks as if (the stock market is crashing and Tom is red).
*The chair looks red and like a sofa.
x looks red.
There is a contextually salient, characteristic way of looking that x shares with red things.
∃y (has(o, y) & (WL(y) & ID(C(G, WL, c), y))),
o looks F and o looks G,
o looks F and G
That one looks red.
*That one looks red and grey.
That one looks red but it doesn’t look red.
That one looks red – just like the other red things around here – but it doesn’t look red.
3 ‘Looks’, Looks, and Visual Experiences
A natural idea about the phenomenal sense of ‘looks’ is the following: Phenomenal ‘looks’ can be used for reporting visual experience. In telling us how things phenomenally look, the idea is, speakers (typically) report on how they visually experience them or what the world around them is like – according to their visual experience. This chimes well with the fact that we typically talk about looks in situations where these are misleading, or where we suspect them to be misleading. Thus, John in Sellars’s classical story first learns to use ‘looks F’-locutions in a scenario where he has to pick a tie by colour but under dubious lighting conditions (cf. Sellars 1963, 142ff). According to Sellars, what John learns when he learns to use the sentence ‘this necktie looks green to me’ is “a way of reporting an experience” (Sellars 1963, 146), more precisely, a way of reporting an experience that “as an experience” is identical to a seeing that something is green (ibid., 145). It thus seems natural to think that phenomenal looks-reports not only report visual experience, but also that they are sensitive to precisely the phenomenal characteristics of these experiences.
In present day philosophy of perception, one of the big divides is between what I shall call “intentionalism” and “anti-intentionalism”. Intentionalists and anti-intentionalists agree that perceptual experiences are (conscious) mental states with distinctive (sensory) phenomenal character. They tend to disagree on the nature of phenomenal character, but the crucial difference concerns the question of content.
Intentionalism is any position that construes perceptual experience as a mental state with representational content. As I shall use the terms here – in a wide and very uncontentious sense – this amounts to construing perceptual experience as a propositional attitude.15 According to standard intentionalism, experiences ascribe sensible properties to ordinary material objects. An experience as of something F thus has a content of the form x is F. There is a lot of discussion about further details, in particular about whether these contents are singular or general, but I shall for the most part abstract from that here. As I said right at the beginning, my own form of intentionalism – phenomenal intentionalism – does not construe experience as having such standard contents. According to phenomenal intentionalism, visual experience in particular has “looks-contents”, i.e. contents of the form x looks F. Intentionalism is often combined with an explanatory ambition regarding phenomenal character: What is usually called “representationalism” is the claim that the representational content of perceptual experience determines, and thus explains, its phenomenal character.16
Anti-intentionalism on the other hand denies that experience has representational content. It mostly comes in the form of relationalism. According to relationalism, veridical perceptual experiences are relations between subjects and (mind-independent) objects. Here, the object of a veridical experience is a constituent of the experience. Anti-intentionalists have a tendency to deny the existence at least of any interesting phenomenal use of ‘looks’, while intentionalists mostly are happy to simply assume its existence.17 This is slightly curious – as it is at least not immediately obvious what exactly motivates these tendencies.
With hindsight, [Jackson] could have taken it to index the content of perception instead: if o looks p (the subscript indicating the phenomenal use) F to S then S exes, of o, that it is F (Byrne 2009, 442).
An experience report of the form o looks p F (to S) is true iff S exes that o is F.
And so does Brogaard:
The force of the indeterminacy objection relies on ‘looks’ being understood comparatively. If ‘looks’ is understood noncomparatively, (...) then the way things look fixes the content of experience (Schellenberg 2011, 9).
Intentionalists employing something like the standard indexing principle in response to the indeterminacy challenge construe the phenomenal properties that looks-reports are sensitive to as properties the experience has in virtue of its content. So construed, a looks-report does not commit the reporter to the truth of the reported experience’s content – to o’s being F – but only to S’s having an experience with the content that o is F. Employing the standard indexing principle thus, in effect, amounts to treating phenomenal ‘looks’ as a propositional attitude operator: ‘o looks F (to S)’ is treated as equivalent to ‘It looks (to S) as if o is F’ and both are construed as true iff S exes that o is F, where ‘exes that’ is a propositional attitude operator just like ‘believes that’. Let’s call proposals along these lines “attitude operator semantics” for phenomenal ‘looks’.
If phenomenal ‘looks’ actually works this way in natural language that would be very good evidence at least for the claim that pre-theoretically, we conceive of perceptual experience as (what a philosopher would call) a propositional attitude. In other words: If phenomenal ‘looks’ in natural language indeed is a propositional attitude operator, that is (at least some) evidence for intentionalism. Of course, we can have good reasons to give up on (parts of) our pre-theoretical conceptions of pretty much anything – but if something is so deeply engrained as to be encoded in the syntax and semantics of natural language, giving it up is a rather high cost. The anti-intentionalist would be able to avoid paying this price if they instead could convince us that there is no phenomenal use of ‘looks’.
But while attitude operator semantics certainly is not lacking prima facie plausibility, it is by no means clear that it is the best way to go. Other options become more visible once we think in terms of the metaphysics of looks. According to attitude operator semantics, looks presumably are experiences. Or maybe they could be construed as properties things only have as experienced: An object has a look iff it is the object of someone’s experience. This isn’t implausible, but neither is it implausible to think that things have looks even if no-one is looking (someone who makes heavy weather of this is Martin (2010)). Of course, some feel that such properties are best understood in terms of dispositions of objects to look certain ways if someone were to look at them. But others feel that looks are more “objective” than that, i.e. that they are mind-independent properties of objects. Let’s call all accounts of looks according to which looks are mind-independent properties of objects “non-phenomenal accounts”.
According to Noë, for instance, looks are what he calls “perspectival properties”, or “P-properties” (cf. Noë 2004, 84). These are relational properties of objects, but they are not relations to experiences. “P-properties are, in effect, relations between objects and their environment”, Noë explains (Noë 2004, 83). P-shapes, for instance, are determined by an object’s shape and its relation to the location of a (possible) perceiver. More precisely, P-shapes are occlusion shapes: “[t]he P-shape is the shape of the patch needed to occlude the object on a plane perpendicular to the line of sight” (Noë 2004, 83). If P-properties are looks, it is only natural to expect there to be a use of ‘looks’ that, in effect, ascribes P-properties to objects. Thus, attitude operator semantics would seem to have a rival: “P-property semantics”.20 We could thus have a “relational semantics” for the third, or phenomenal, use of ‘looks’ without adopting a phenomenal account of looks – where a phenomenal account of looks construes looks as relational properties involving experiences.
Favoring relational semantics over attitude operator semantics for ‘looks’ does not prevent us from hanging on to the idea that phenomenal ‘looks’ reports experiences (and their content). Noë, for one, thinks that P-properties are represented in the content of experience (along with intrinsic properties of objects). And according to phenomenal intentionalism, all experience contents are looks-contents, so phenomenal ‘looks’ can indeed be used to report experience.21
Sam looks straight.
Sam looks bent.
As far as I can see, it is only the adoption of the most extreme non-phenomenal parsimony regarding looks that is hard to reconcile with the existence of a third use of ‘looks’, a use of ‘looks’ interpreted as directly ascribing looks to objects. Martin suggests identifying looks with intrinsic properties of objects, more precisely, with their basic visible properties (such as shape and color) (cf. Martin 2010, 161; 207ff). Directly ascribing a looks-property such as looking bent to Sam then simply amounts to ascribing bentness to him. A “parsimonious semantics” for phenomenal ‘looks’, that is, cannot explain the intuitive truth-value changes for sentences like (32) and (33). In order to have a chance at making a parsimonous account of looks compatible with the intuitive context-dependence of the truth values of looks-sentences, we need to make it plausible that virtually none of them are direct ascriptions of looks to objects. Rather, they need to be construed as comparative.22
With the exception of extreme parsimony about looks, then, there are, it seems, various accounts of looks available that could inspire prima facie quite plausible interpretations of phenomenal ‘looks’, and among them are interpretations that do not construe ‘looks’ as an attitude operator. Is it, then, only the most extreme relationalist that really has anything to fear from the existence of a third, or phenomenal, use of ‘looks’? I ultimately don’t think so. But to make any further progress on how best to interpret phenomenal ‘looks’, we need to try a new tack. For all that has been written about the uses of ‘looks’, we still know too little about how ‘looks’ is actually used in natural language. In what follows, I shall draw attention to certain patterns in the usage of phenomenal ‘looks’ that are hard to reconcile not only with attitude operator semantics but also with non-phenomenal semantics. Again, the observations I shall present will hopefully push the debate forward, but hardly clinch it.
Tom looks red (to S).
*It looks (to S) that Tom is red.
S sees that Tom is red.
Tom smells/feels/tastes/sounds ripe (to S).
*It smells/feels/tastes/sounds (to S) that Tom is ripe.
*S smells/feels/tastes that Tom is ripe.
Tom seems/appears ripe (to S).
It seems/appears (to S) that Tom is ripe.
It looks as if Superman is flying by.
It looks as if Clark Kent is flying by.
In phenomenal looks contexts, co-phenomenal expressions can be substituted salva veritate.
o looks green.
o looks triangular.
o looks triangular.
o looks trilateral.
o looks red.
o looks red 52.
It looks as if Tom is on the table in front of me.
It looks as if Tim is on the table in front of me.
The observations made so far already provide a good case against attitude operator semantics. Phenomenal ‘looks’ allows for salva veritate substitutions that no propositional attitude operator should allow. At this point, it therefore looks as if the phenomenal use of ‘looks’ is better interpreted as ascribing relational properties to things rather than as ascribing propositional attitudes to subjects. But which? Should we go for a phenomenal or for a non-phenomenal semantics for the third use of ‘looks’?
Sam looks bent.
Dubbit looks rabbity.
Dubbit looks ducky.
Dubbit looks rabbity to S at t.
Dubbit looks ducky to S at t.
What’s more, an object doesn’t need to be an “ambiguous figure” – something like a duck-rabbit or a Necker-cube – in order to undergo changes in looks that have nothing to do with changes in its P-properties (or any other non-phenomenal relational properties). Painters learn to make pretty much any old object undergo a certain kind of overall Gestalt switch: They learn to make objects have what is sometimes called “painter’s looks”. An object’s painter’s look is a “flattened Gestalt”: Having it’s painter’s look, an object looks like a paper cut-out, i.e. as if it was a very thin, flat object perpendicular to your line of sight. Learning to see objects this way involves learning to switch Gestalt also on all the depth clues they typically provide. Thus, what typically looks like a shadow will look like an difference in objective colour when an object has its painter’s look. These changes cannot be understood as changes in an object’s non-phenomenal relational properties. No differences in looks that result from nothing but mere Gestalt switches can, and pretty much any object can (be made to) switch Gestalt.29 Phenomenal ‘looks’ clearly is sensitive to changes in look that result from mere Gestalt switches, not just to changes that result from differences in an object’s relations to its environment. Ultimately, a phenomenal semantics for phenomenal ‘looks’ therefore has a bit of an edge over a non-phenomenal semantics.
To sum up: In this section I have argued that phenomenal ‘looks’ is not a propositional attitude operator. When we use phenomenal ‘looks’ in natural language, we do not ascribe experiential attitudes to subjects; rather, what we seem to be doing is ascribe relational properties to the objects of experiential attitudes.30 These properties in turn are best construed as phenomenal relational properties, i.e. in terms of relations between objects and the experiences they are the objects of.31 Therefore, it seems to me, the third use of ‘looks’ is aptly named the phenomenal use, after all. In the next and final section, I shall explore its (syntax and) semantics just a bit further.
4 More on Phenomenal ‘Looks’
x looks red.
It looks like/as if x is F.
o looks F.
o looks to be F.
It looks as if o is F.
It looks as if something is F.
Something looks to be F.
(56 ′) L 1 (∃x (F x ))
∃x ((L 2(F))(x)) ≡ d f ∃x L 1(F x)
The farmhouse looks red and old.
The farmhouse looks old.
Mora looks old.
Mora looks bald, pink, and wrinkled.
If phenomenal ‘looks F’ indeed ascribes the salient F-look to an object, then, it seems to me, Byrne has to be right in thinking that ‘red’ in ‘looks red’ does not have the semantic value it normally has. It is, after all, not the look that is red.
Similarly, if someone looks p Scandinavian, and so looks to have the stereotypical Scandinavian bodily features (straight blond hair, small nose, pale skin, etc.), he can be as he looks p without being Scandinavian. (...) [T]hat animal, which looks p centaurian, can be as it looks p without being acentaur. ‘Looks p F’ is therefore idiomatic in the interesting way ‘red hair’ is. ‘Red hair’ does refer to hair of adistinctive colour similar to red (and so is an example of polysemy), but that orangeish shade is not the semantic value of ‘red’. (‘Looks p Scandinavian’ and ‘red hair’ are thus quite different from paradigmatic idioms like ‘blue blood’ and ‘green thumb’) (Byrne (2009), 444, subscripts changed).
However, I also think that it would be quite unfortunate if we had to settle for the conclusion that phenomenal ‘looks F’ is idiomatic – even in the interesting way ‘red hair’ is. Already on the basis of the examples we have looked at, it seems safe to predict that phenomenal ‘looks’ can modify a very large range of predicates. Moreover, the semantic effects of such combination seem to be quite systematic. All of this will be hard to explain if we go idiomatic on phenomenal ‘looks F’. We should also not forget the peculiar substitution behaviour we observed earlier. This presumably would receive no explanation at all on an idiom account.
What a large range of meaningful combinations with systematic effects would seem to call for is, of course, a compositional semantics predicting those effects. But a (classically) compositional semantics is precisely what we cannot have if ‘F’ does not have its usual semantic value when combined with phenomenal ‘looks’. This is not as much of a dilemma as it might look to be, however. For even if we cannot have a semantics that is classically compositional, we might be able to have one that is general compositional (cf. Pagin and Westerståhl (2010a, 2010b)). And that is just as good: It gives precisely the kind of explanation of the semantic effects of composition that we would like to see.
If two predicates ‘F’ and ‘G’ are co-phenomenal (in a context c), then phenomenal ‘looks F’ and ‘looks G’ are synonymous (in c).
What we need next is a systematic semantic explanation of the synonymy. The basic idea would be the following: phenomenal ‘looks’ is a “semantic evaluation switcher”.36 A phenomenal looks-modified predicate F does not denote F-ness but rather phenomenal F-ness. When in the scope of phenomenal ‘looks’, that is, a predicate F is evaluated by means of a function assigning sets of phenomenal F s (at worlds and contexts) as its extensions – instead of sets of F s. Such a switcher semantics is not classically compositional, but it is general compositional. Obviously, this is the merest sketch of the idea, which I hope to develop in much more detail in a paper of its own.
This paper, I would like to round off by revisiting the propositional attitude operator question. For even if we find the substitution behaviour of phenomenal ‘looks’ peculiar, we might wonder whether phenomenal looks contexts are really so very different from, for instance, belief contexts. After all, there is much a speaker needs to know besides their meanings to understand that two expressions can be substituted salva veritate in a phenomenal looks context. This is especially important if phenomenal ‘looks’ indeed meaningfully combines with lots of predicates over and above the basic sensible ones. One might therefore wonder whether there really is any deep difference here. After all, belief contexts allow for all sorts of peculiar substitutions, too – provided the speaker has sufficient extra-linguistic knowledge. And isn’t that just knowledge of the very kind required for substituting into phenomenal contexts, too?37
Bob believes that o is F.
What does all of this mean for the theory of perceptual experience? To the extent that phenomenal ‘looks’s being a propositional attitude operator in natural language would provide support for standard intentionalism, this is bad news for standard intentionalism. Its not being a propositional attitude operator, however, does not favor anti-intentionalism over intentionalism in general: If anything, it is water on the mills of phenomenal intentionalism.
Just a few examples for illustration: Jackson (1977) uses claims about looks-language to support a sense datum theory, Chisholm (1957) to argue for adverbialism, Alston (2002) for the theory of appearing, Sellars (1963), Byrne (2009), and Brogaard (2013, 2015) for intentionalism, and Martin (2010) claims that there are no semantic arguments against identifying looks with basic sensible properties such as redness and roundness, thereby providing support for a form of anti-intentionalist relationalism.
Epistemic ‘looks’ thus looks quite a bit like a kind of epistemic modal. I tend to think that it, in fact, is one. For a slightly different take, cf. Brogaard (2015).
Cf. Jackson (1977), 31. Jackson here is discussing the idea that epistemic ‘looks’ should be understood in terms of “tentative, guarded, etc. assent”, but as (rational) assent is based on total evidence, his observations are equally relevant to the question of whether epistemic ‘looks’ is best interpreted weakly or strongly.
The example again is one of Jackson’s, cf. Jackson (1977), 31.
I have used this example numerous times, and nobody has ever found it anomalous – except for Peter Pagin.
Thus, Byrne argues that both ‘looks red’ and ‘looks old’ can be used in all three ways. It seems to me that he is right about this. We’ll get back to this in Section 4.
- 8.Some, but very few, people report not being able to read (20) any way other than epistemically. If you are one of those, try
*The stock market looks to be crashing and Tom to be red.
instead. I think that ‘the stock market looks to be crashing’ can be read epistemically, and indeed has to be in the extended treacherous tomato scenario. I also think that in this scenario (i) is not acceptable.
On the assumption that comparative uses quantify over phenomenal looks, we would predict that comparative looks do not disappear when defeated, either. As far as I can tell, that is the case. On the further assumption that (20) can be read comparatively, we can run an analogous argument for the claim that the distinction between epistemic and comparative uses of ‘looks’ amounts to polysemy.
Rather, the ambiguity here is – somehow – created by the combination with ‘like’ (cf. Brogaard 2013, 20f).
The same holds for perceptual epistemic uses of ‘looks’.
In Section 4, I will argue that even phenomenal ‘looks F’ comes with a (hidden) context parameter and means something like the salient red-look. But, or so I shall also argue, at least for basic sensible predicates such as ‘red’, the resulting range of looks they can be made to denote when looks-modified will be very limited. It certainly will not include shades of grey.
- 14.Breckenridge (2007) suggests an event-semantics for ‘looks’ according to which looks are events and ‘looks F’ is a definite description of a property of such events: the way F s look. As this description is analyzed as a quantifier construction, this might predict that conjunction reduction for sentences of the form (27) always is unproblematic. As explained below, I think it isn’t. Moreover, and also pace Breckenridge, I think that construing them as definite descriptions makes the wrong predictions regarding the modal profile of phenomenal ‘looks F’. Thus, I think that (i) is true, while Breckenridge predicts that it has a false reading:
Red things might not have looked red.
In this wide sense, propositions are those things that essentially have, or determine, truth-conditions (including so-called “satisfaction” or “accuracy conditions”), and any mental state that has a proposition as its content is a propositional attitude.
Inverse representationalism holds that in experiences, phenomenal character determines representational content, and strong representationalism is the claim that, for perceptual experience, there is an equivalence relation between representational content and phenomenal character. There are also intentionalists who hold that only some, but not all, aspects of phenomenal character are to be explained by means of representational content (thus, for instance, Schellenberg (2008)).
17Travis (2004) implicitly denies the existence of a third use of ‘looks’, while Martin (2010) accepts that there is such a use, but construes it as ascribing certain properties to looks, properties such as splendidness. This appears to be motivated by parallel constructions involving sound or smell, but both the extent and the force of the parallel remain unclear to me – why exactly are we to conclude that, for instance, ‘looks red’ would have to be read as attributing redness to a look? Intentionalists often simply accept and sometimes utilize the existence of a third use of ‘looks’. Arguments and discussion are provided by Maund (1986) and more recently by Byrne (2009) and Brogaard (2013, 2014, 2015).
According to Byrne, experiences have singular contents, but why that would need to be displayed by means of a de re attitude report in this context, I do not know. Nor does it matter for present purposes.
Siegel (2010, 2011, 59ff,) accepts that there might not be ause of ‘looks’ in natural language that exclusively tracks what’s presented in visual experience, but argues that such publicly available looks wouldn’t be what’s relevant for the determination of experience content in any case. Moreover, she argues, there is no good reason to think that we couldn’t devise “a special, regimented sense of looks that did track what’s presented in visual phenomenology” (Siegel 2011, 63). For such aregimented sense, standard indexing would be true – by regimentation.
Yet another rival semantics for phenomenal ‘looks’ employing mind-independent relational properties can be derived from Brewer’s account of looks (cf. Brewer 2011, 118ff). Brewer construes looks in terms of an object’s visually relevant similarities with paradigms, relative to a point of view and other circumstances of perception.
And even if you don’t think experiences have contents, you could construe phenomenal looks reports as reporting on them: Gauker suggests interpreting such reports as telling us what assertions an experience would dispose an agent to make were she to be guided only by that experience (Gauker forthcoming, 28).
- 22.It should be noted, however, that this also requires further complicating the logical form of comparative looks-sentences. To have a chance at explaining the truth value data, these sentences need to be construed as doubly context-sensitive. As long as we think of (33) as saying that there is a salient, characteristic look that both Sam and bent things share, the sentence will be as false in all contexts as the sentence ‘Sam is bent’. Instead of
Martin thus (in effect) suggests the following logical form for comparative looks-sentences:
(10 ′) ∃y (has(o, y) & (WL(y) & ID(C(F, WL, c), y))),
(10 ″) ∃y (has(o, y) & (WL(y) & SIM(C(F, WL, c 1), y, c 2)))
where ‘SIM(x, y, c 2)’ is the kind of similarity between x and y relevant in context c. (33) then is true if the look that Sam has is relevantly similar to whatever is the salient and characteristic look of the bent things in c. Martin explains: “The stick is similar to bent things simply with respect to how it strikes me, or the subjective bearing it has on me” (Martin 2010, 215). More precisely, looked at when half-way immersed in water, Sam is such that a speaker uttering (33) “enjoys a visual experience relative to which they would find the stick similar to a bent thing” (ibid., 221). It is here that the context dependence of SIM gets to do its work: the context somehow has shifted the similarity measure to something involving visual experiences of Sam.
The need for double context dependence is due solely to the parsimonious account of looks. Even if we were to grant that all the relevant sentences are indeed comparative, no relational account needs to go doubly context dependent. Relational accounts of looks thus have the advantage of explaining the truth value data by means of a simpler semantics (cf. Glüer (2013) for a more detailed case in support of this claim). Nor is it the case that comparative sentences standardly are construed as doubly context dependent (or that there is any obvious reason for doing so). We therefore have good reason to reject Martin’s claim that “there is no reason from the semantics alone to attribute to objects more than the basic visible properties that we were committed to positing anyway” (Martin 2010, 222).
As an anonymous referee notes, ‘want’ might be taken to be another propositional attitude operator somewhat unhappy with that-clauses.
If you like, you can construe the example as one showing hyperintensionality, but that requires assuming a controversial semantics for proper names according to which co-referential proper names are co-intensional. I don’t subscribe to such a semantics (cf. Glüer and Pagin 2006), but nothing here will hang on the question of whether looks-contexts are intensional or hyperintensional.
Why not? The short answer is that what guarantees substitutivity salva veritate in attitude contexts is precisely either co-intensionality or co-hyperintensionality. Both entail co-extensionality. And the substitution behaviour of attitude contexts should be like that; if an operator is not sensitive to precisely differences in semantic content or meaning, it is not a propositional attitude operator, it seems to me. Phenomenal ‘looks’ is not: As we shall see, there are expressions that are not even co-extensional but substitutable in looks-contexts. For a bit more detail on this, see Glüer (2014), fn 38.
- 26.If you have difficulties reading ‘It looks as if p’ phenomenally, consider the following pair instead:
Tom looks to be on the table in front of me.
Tim looks to be on the table in front of me.
- 27.If you think ‘rabbity’ and ‘ducky’ are not part of English, consider the following pair instead:
Dubbit looks to be a rabbit.
Dubbit looks to be a duck.
In many expressions of the form ‘x looks to be F’ the infinitive can be deleted (cf. Brogaard 2015, 244). But this does not work for ‘looks to be a G’ where ‘G’ is a count noun.
The Gestalt an object takes is somehow determined by our visual system, of course. But whatever is responsible for Gestalt switching is not something we are aware of. What determines the intuitive truth values of looks-sentences, however, are changes we are eminently aware of.
Noë describes the acquiring the ability to see an object’s painter’s look as acquiring the ability of recovering an earlier stage of visual processing, i.e. recovering something like a two-dimensional image of the seen object (cf. for instance Noë (2004), 175ff), but I think that gets the phenomenology slightly wrong. An object’s painter’s look is as much a fully gestalted look as any other; having its painter’s look, the object looks like a very thin, flat material object (for more on this, see Glüer (2016a)).
An anonymous referee wonders whether phenomenal ‘looks’ might also be construed as itself ambiguous: Couldn’t it be the case that it sometimes works as a propositional attitude operator and at other times is used to ascribe relational properties to the objects of those attitudes? This hadn’t occurred to me, but I guess the question to ask in return is whether there really are any use data regarding phenomenal ‘looks’ forcing us to posit further ambiguity – data we cannot account for in terms of relational property ascription. So far, that does not seem to be the case to me. Note, too, that if the relational properties ascribed by phenomenal ‘looks’ are best construed as phenomenal relational properties, an ascription of such a property to an object o implies the existence of the experience o is the object of, and can thus be used to “report” on that experience (as noted above, Section 3).
Obviously, much more would need to be said about these properties. But this is not the place to do so. For a bit more on how I think phenomenal properties are best construed, see Glüer (2016b).
If this is right, one would also expect sentences containing ‘looks to be F’ to strike at least native speakers of English as entirely natural. While that might hold for native speakers from England, it does not quite generally seem to be the case, however.
Byrne himself calls the third use the “non-comparative use”.
What Byrne is interested in here is showing that phenomenal ‘looks’ does not index the content of experience. Here’s why: Mora is a baby. Like all naked mole rats, Mora was born looking old. But when you look at her, your experience might be perfectly veridical. Assuming that perfectly veridical experiences do not have false contents, Byrne concludes that the experience you can report by means of (58) does not ascribe being old to Mora. What it does ascribe is being bald, pink, and wrinkled.
A helpful analogy here would be personal pronouns. ‘Looks F’ expressions seem to pick up the property they ascribe from the context in as neat and systematic a way as personal pronouns pick up their referents.
This objection has independently been raised by Elisabeth Coppock, Josep Macià, and Elia Zardini.
I would like to thank Peter Pagin, Gunnar Björnsson, Brit Brogaard, Elisabeth Coppock, Manuel García-Carpintero, Chris Gauker, Josep Macià, Josep Prades, Andreas Stokke, Pär Sundström, Elia Zardini, audiences in Barcelona, Dubrovnik, Oslo, Uppsala, and Umeå, as well as two anonymous referees for very helpful comments and conversation. This project has received funding from the European Union’s Horizon 2020 research and innovation programme under the Marie Skłodowska-Curie grant agreement No 675415 as well as from the Swedish Research Council (VR 2013-737).
- Alston, W.P. 2002. Sellars and the myth of the given. In Philosophy and Phenomenological Research LXV: 69–86.Google Scholar
- Austin, J.L. 1962. Sense and Sensibilia. Oxford: Oxford University Press.Google Scholar
- Brogaard, B. 2013. Do ‘looks’ reports reflect the contents of perception? Unp. MS.Google Scholar
- Brogaard, B. 2015. Perceptual reports. In The Oxford Handbook of Philosophy of Perception, ed. Matthen M. , 237–253. Oxford, Oxford University Press.Google Scholar
- Chisholm, R. 1957. Perceiving: A Philosophical Study: Cornell University Press, Ithaca, NY.Google Scholar
- Gauker, C. forthcoming. The argument from ‘looks’-talk. In In the Light of Experience: Essays on Reasons and Perception, eds. Gersel J. , Jensen R.T., Thaning M.S., and Overgaard S. The Mind Association Occasional Series, OUP.Google Scholar
- Glüer, K. 2013. Martin on the semantics of ‘looks’. Thought 1: 292–300.Google Scholar
- Glüer, K. 2014. Looks, reasons, and experiences. In Does Perception have Content? ed. B. Brogaard, 76–102. Oxford University Press.Google Scholar
- Glüer, K. 2016a. Constancy in variation. An argument for centering the content of experience? In De Se Thought and Communication, eds. Torre S. and García-Carpintero M., 56–85. Oxford, Oxford University Press.Google Scholar
- Glüer, K. 2016b. Defeating looks. In Synthese (online first). doi: 10.1007/s11229-016-1186-x.
- Glüer, K. , and P. Pagin. 2012a. General terms and relational modality. Nous 46: 159–199.Google Scholar
- Glüer, K. , and P. Pagin. 2012b. Reply to Forbes. Analysis 72: 293–303.Google Scholar
- Jackson, F. 1977. Perception. Cambridge: Cambridge University Press.Google Scholar
- Noë, A. 2004. Action in Perception. Cambridge, MA: MIT Press.Google Scholar
- Pagin, P., and D. Westerståhl. 2010a. Compositionality I: definitions and variants. Philosophy Compass 5: 250–264.Google Scholar
- Pagin, P., and D. Westerståhl. 2010b. Compositionality II: arguments and problems. Philosophy Compass 5: 265–282.Google Scholar
- Pagin, P., and D. Westerståhl. 2010c. Pure quotation and general compositionality. Linguistics and Philosophy 33: 381–415.Google Scholar
- Sellars, W. 1963. Empiricism and the philosophy of mind. In Science, Perception and Reality. London: Routledge and Kegan Paul, 127–196.Google Scholar
- Thau, M. 2002. Consciousness and Cognition. New York: Oxford University Press.Google Scholar