The possible worlds theory of visual experience

ABSTRACT

When we watch movies, or are tricked by a trompe-l'oeil painting, we seem to be visually representing possible worlds; often non-actual possible worlds. This suggests that we really can visually represent possible worlds. The suggested claim is refined and developed here into a theory of visual experience that holds that all visual experiences, both veridical and non-veridical, represent possible worlds, many of which are non-actual.

KEYWORDS:

• Perception
• singular content
• particularity
• hallucination
• representation
• possible worlds

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 ‘Items’ is a neutral term, and can refer to objects, property types and tokens, relations, and events.

2 As we have described them here, theories that identify the represented-items with property-complexes – i.e. theories that say that all visual experiences are fundamentally property-representations – are what Schellenberg (2011, 7) calls pure property-cluster theories. Yet other theorists might insist that objects are represented by their properties in another way – by experiences having wholly descriptive propositions as contents. This view is sometimes called generalism about visual experience (McGinn 1982; Pautz 2007; Hill 2019). All criticism of the former view will apply to the latter. Generalism is discussed further in footnote 12.

3 Due to both space and thematic constraints, this essay only takes up the major non-cognitive and non-Fregean representational theories of visual experience. By ‘cognitive theories’ we have in mind Montague's (2016), according to which visual experiences are partially constituted by bare demonstrative thoughts, what Montague calls ‘object-positing’. On Schellenberg's (2018) Fregean particularism, the representational content of experience is constituted by ordered sets of de re modes of presentation, which are gappy in hallucinatory cases. While these theories have their own merits and costs – and indeed are concerned with some of the very phenomenological facts at issue here – they will be set aside.

4 See McDowell (1994) and Schellenberg (2018).

5 Masrour (2020) is a possible exception to the general endorsement of C2, but he might be better understood as denying that t-hallucination is possible, instead re-constructing purported cases as extreme forms of illusion. The difference between the PW theory and Masrour's will become clear as we proceed. See also Rezaei (ms).

6 Naïve realists-disjunctivists will deny C3 (Martin 2004).

7 For PC1, see Tye (2014a, 2014b, 2014c). For PC2, see Tye (2014b): 85. For PC3, see Dretske (1995, 102) and Tye (2014a, 41, 51).

8 The ‘P’ in ‘P-Particularity’ stands for ‘phenomenal’. Because P-Particularity is a phenomenological claim – a claim about how things seem to the subject – P-Particularity should not be confused with what Schellenberg (2010) calls relational particularity, which is a claim about the constituents of experience (i.e. whether experience is constituted by particulars), not necessarily a claim about how things seem. Those who claim that experience is wholly general (e.g. Mehta 2014) and thus deny particularity in this relational, metaphysical sense can still attempt to endorse P-Particularity. In this way, whether P-Particularity is best explained in term of metaphysical particularity is a substantive open question. When we use ‘p-particularity’ (i.e. not in small caps), we are referring to the property, not the thesis that visual experiences have this property.

9 In other words, if P-Particularity concerned only generic phenomenal-particularity, wherein one seems to be aware of a particular (as opposed to a universal), the first description would do fine (French and Gomes 2019, 43; Gottlieb and Rezaei 2020). But we are concerned with specific phenomenal particularity. We are assuming that phenomenal particularity is not a post-perceptual phenomenon.

10 PC theorists assume that properties are universals (e.g. Dretske 1999, 162–163). This falls out of PW3. We discuss the relevance of this point further in Section 4.2.2.

11 Hence Tye's (2019a) alternative suggestion – that an HE can be of an F without there being a particular F being represented – misfires. See fn. 9. Many thanks to Ali Rezaei for discussion.

12 Generalism (see fn. 2) faces the same problem. Here Jane's HE represents that: $\mathrm{\exists }\mathrm{x}(\mathrm{x}\mathrm{is}\mathrm{red}\wedge \mathrm{x}\mathrm{is}\mathrm{round})$An experience with an existentially quantified proposition as its content may represent an object, but only whatever object happens to satisfy the description laid down by its content. And adding to the description (e.g. Searle 1983) will not help the generalist any more than adding properties and relations to the property-complex helped the PC theorist. Nor does, as Hill (2019) suggests, making the descriptions especially fine-grained. See Gottlieb and Rezaei (2020) for discussion.

13 P-Particularity does not say you seem to be aware of a physical particular apple; properties like being physical do not obviously manifest themselves phenomenologically, and being mental or being non-physical don't obviously either.

14 See, e.g. McGinn (1982, 39).

15 Subjects, unless otherwise noted, are assumed to have good vision; and scenes contain objects that can reflect or transmit light, but (for the sake of simplicity) do not generate it.

16 From Jane sees x it does not follow that Jane sees every part of x.

17 For a somewhat similar approach to this issue in a different context, see French and Gomes (2019)

18 Lewis' concrete modal realism is a non-Meinongian version of possibilism insofar as it rejects any distinction between being and existence. For Lewis, ‘actual’ is indexical, and the property actual is relational, with no ontological gravity. The PW theorist who endorses object-dependent singularity might have options beside Lewis' possibilism. For example, she might opt for Williamson's (2013) necessitism. Whether this is less palatable than Lewis's concrete modal realism will not be broached here. For something approximating a Meinongian approach to t-hallucination, see Smith (2002). Lycan (1987) also appeals to non-actual ‘merely intentional’ objects in t-hallucination, but he does not spell out how this works. For a discussion of t-hallucination and counterpart theory, see Lewis (1999/1983).

19 PC theorists are forthright that, for them, properties are universals. (See fn. 10). So trope views are set aside here. Regardless, trope views have the same difficulties with t-hallucinations as Aristotelian views.

20 For discussion, see Churchland (2005). See also Ffytche and Howard (1999) and Ffytche (2008), as cited in Schellenberg (2011, fn. 11). These ‘chimerical colors’ are uninstantiated at the actual world; on the PW framework, they are instantiated in scenes at non-actual worlds. This is consistent with the empirical data. Further, if one is a (figurative) color projectivist, on the PW framework all colors will be instantiated only in scenes at non-actual worlds. See Section 4.4.2.

21 This observation is in the spirit of Schwitzgebel's (2014) ‘crazyism’ about the metaphysics of mind.

22 Note too that we would not hereby be positing concrete non-actual worlds and their inhabitants simply because of visual experience and the p-particularity of t-hallucination. The friend of the PW theory could avail themselves of any other reasons to be a modal realist, independent of visual experience. Thus, the PW theorist is not on this score in the same boat as the sense datum theorist. There is no reason to posit sense data outside of t-hallucination.

23 Causal approaches to the how-question fall under the more general ‘tracking’ theoretic approach to the how-question. On this family of views, a token experience E's representing F is, roughly, a matter of E carrying information about, or in some way correlating with instances of F. See, e.g. Millikan (1984), Fodor (1987) and Tye (2000). The marriage of representationalism with tracking accounts of representation yields ‘tracking representationalism’.

24 This is not an abandonment of a causal-tracking approach. Suppose one t-hallucinates a red circle. On the PC theory, this t-hallucination represents uninstantiated redness and roundness, so cannot be caused by uninstantiated redness and roundness. Nonetheless, this token t-hallucination is of a type that normally tracks (and is caused by) redness and roundness. That's all that is needed to say that the token t-hallucinations represent uninstantiated properties, or so the story goes. See Tye (2019b).

25 See, for instance, Crane (2013) and Hansen and Rey (2016).

26 Intentional objects, whatever they are, are equivalent to represented-items on the assumption that (i) intentional objects are represented, and (ii) intentional objects figure into phenomenal character. Note that intentional objects are not contents, at least on one good sense of ‘content’. Two mental states can have the same intentional object but represent that object in different ways. For example, I can think that Jane is generous, but I can also think that Jane is greedy; these beliefs have the same intentional object (Jane) but differ in content (cf. Crane 2013, 97). And if Jane is generous, my first thought is true whereas my second is false. In this way, content – again, at least in one sense – bears on a state's evaluability, where this is understood in terms of truth, accuracy, or satisfaction conditions. But content itself is not what I am thinking about; I am thinking about Jane, not a proposition, or whatever. Some mental states with intentional objects (like fear) might be classified as ‘non-propositional attitudes’ insofar they are about things (they have intentional objects) but not in virtue of relating subjects to propositions concerning those things. The present point covers both propositional and non-propositional attitudes; while both have intentional objects, they differ over whether the aboutness-relation is propositionally mediated. Propositions themselves can be intentional objects. As Grzankowski (2013) points out, liking the proposition that p is not a propositional attitude, but a (quite odd) non-propositional attitude with the proposition that p as its intentional object. The PW theory is not committed to saying that visual experience is a propositional attitude.

27 ‘Intentionally directed’ might be a more comfortable expression to describe perceptual experiences than saying they are ‘about’ things (as we do with, say, thoughts), but this linguistic artifact does not mark any deep difference concerning their nature (Crane 2013, 90). While there are some non-perceptual states that are fairly clear candidates for lacking intentional objects altogether (e.g. so-called ‘undirected anxiety’), within the perceptual domain things are less obvious. Two potential perceptual examples: the experience one has upon opening your eyes in a completely dark room and, perhaps less so, ganzfelds (see Block (ms) for the former). Unsurprisingly, these are also potential examples of visual experiences without p-particularity.

28 If singular-content-representations turn outs to be object-dependent, then objects will be amongst the represented-items, but – given PW2, PW3, C4* and the pressures from generality – this is because their existence will be a pre-condition on singular representation. If singular-content-representations turn out to be object-independent, then objects will be amongst the represented-items in the same way, but these objects may only be merely intentional objects in t-hallucination. For an account of how the latter might work using mental files, see Crane (2013).

29 The accuracy condition embedded in PW-core is used here to explain Jane's inaccurate representation of the museum's wall. PW-core includes the accuracy condition because an attempt to explain an inaccurate representation of a scene in the actual world in terms of an inaccurate representation of a scene a non-actual possible world would be circular. This point is embellished below.

30 This point has been made in discussions of pictorial representation. For example, Newall (2009) notes how trompe l'oeil paintings typically do not involve seeing-in or twofoldness; when they work right, we simply experience the painting's subject matter. Moreover, Newall argues – in line with (6) – that one's experience in such a case is non-veridical.

31 Seeing is inessential to having a visual experience; in t-hallucination, one sees nothing. But veridical, illusory and partially hallucinatory visual experiences involve seeing.

32 If ‘and its background’ is eliminated from SEE, S would not see black objects. Subjects see the edges of colored shapes (a kind of object), and thus their shapes, by seeing color contrasts. So to see a colored shape one needs to see the colored shape against a background.

33 Light can be absorbed, reflected and transmitted by non-radiating objects. In SEE ‘in the right way’ means that the reflected light from Fx reaches S, and is the stimulus that causes S's visual system to represent x as an x instantiating F. If one takes a picture of Fx, or a video of Fx, and S sees the picture or video, this does not count as a case where S sees Fx, because it is not the light leaving Fx that stimulate S's visual system. However, S can see Fx in a mirror (the light from Fx has been reflected) or through a piece of glass (the light from Fx has gone through a transparent medium). Objects that generate light will not be discussed (see fn. 15).

34 Since ‘scene’ is defined in terms of what a person in a possible world can see in that world, PW-core tacitly assumes that the objects in scenes reflect light in much the same way that objects do in the actual world.

35 To learn how strong the illusion is in your case go to: https://michaelbach.de/ot/sze-muelue/index.html

36 Many thanks to Ali Rezaei and an anonymous referee for Inquiry for helpful comments and discussion.