THE UNITY OF PERCEPTUAL CONTENT

Representationalists hold that perceptual experience is a conscious representational state with content, something which is accurate or inaccurate in certain conditions. The most common version of Rep-resentationalism takes perceptual content to be singular in the object-place and otherwise consisting of attribution of properties (Singularism/Attributionism). Schellenberg has recently developed a version on which perceptual content is singular even in the property-place in containing a de re mode of presentation of a property-instance (Particularism). In this paper, I show that Particularism faces a version of the problem of the Unity of Perceptual Content. Namely, its supporters haven’t told us how objects can be bound together with property-instances into a content such that it represents them and has accuracy-conditions. Furthermore, I argue that Particularists face an in-principle obstacle in solving it. In contrast, Attributionists can solve the problem and that establishes their view as the only game in town.


I. INTRODUCTION
You see a red tomato on a table.It is natural to describe your mental state as one in which you experience a tomato and its redness.But how should we think of your perceptual experience?
Naïve Realists hold that perceptual experience is a 'genuine' conscious relation to an object and its property-instances (among many others: Campbell 2002 ;Martin 2002bMartin , 2004 ; ;Fish 2009 ;Brewer 2011 ;Johnston 2011Johnston , 2014 ; ;Genone 2014 ).A 'genuine' relation, rather than an 'intentional' one, is one in which two things can stand in only if they both exist.To emphasize the genuine relationality of perceptual experience Naïve Realists call the experiential relation acquaintance , awareness , or presentation and take it to be primitive.On this view, in your tomato experience you're acquainted with the tomato and its redness, perhaps in a particular way.1This is frequently thought to amount to the best articulation of our pretheoretic view of how perceptual experience seems to us.However, since such a genuine relation is absent in the case of hallucinatory experiences, this view has work to do in explaining how perceptual experiences and hallucinations can be internally indistinguishable and what goes on in the latter cases.
In stark contrast, Representationalists hold that perceptual experience is a conscious representational state with content, something which is accurate or inaccurate in certain conditions (among many others: Burge 1991, 2010, 2022, Byrne 2009 ;Davies 1992 ;McGinn 1982 ;Nanay 2012 ;Pautz 2009 ;Schellenberg 2018 ;Speaks 2009 ).Since perceptual experiences and hallucinations can share their contents or aspects of their content, this view is supposed to smoothly account for how perceptual experiences and hallucinations can be indistinguishable and what goes on in the latter cases.
Different versions of Representationalism vary along numerous dimensions of which the most important for us is the distinction between general vs. singular content. 2 On the former version of the view, perceptual content is thought to be entirely general in containing only properties, for example, it could be an existentially quantified one: there is an x: x is red.(McGinn 1982 ;Davies 1992 ;Hill 2021 ;Pautz 2009 ).On this version your tomato experience represents there being something that is red.Call this view Generalism .On recently much more popular versions, perceptual content is thought to be singular in the object-place in containing an object or a de re mode of presentation of an object ( = MOP), but general in the property-place in containing a property which is attributed to the object: < o, redness > , < MOP(o), MOP (redness) > (McDowell 1986 ;Burge 1991 ;Speaks 2009 ).On these versions, your tomato experience represents the tomato as being red.This view is frequently called Singularism , but here we'll mostly call it Attributionism to emphasize the latter aspect.
Naïve Realists and others have offered multiple objections to specifically Generalist views of which I want to focus on the following two: first, it is false to the phenomenology of experience which at least seems to be of particular objects and their property-instances; and, secondly, it is not able to capture how experience grounds demonstrative thought (Campbell 2002 ;Martin 2002a ;Soteriou 2000 ). 3 These objections have led to developments of versions of representationalism like the move from Generalism to Singularism/Attributionism.An especially striking recent development, explicitly motivated by its ability to capture certain further Naïve Realist insights, is a version of the view on which perceptual content is singular even in the property-place in containing a property instance or a de re MOP of one: < o, o's redness > , < MOP (o), MOP (o's redness) > (Schellenberg 2011b(Schellenberg , 2018 ; ;Nanay 2012 ).On this view your tomato experience somehow represents the tomato and its redness .Call this view Particularism .
Naïve Realism and Particularism are similar in taking your tomato experience to relate you, one way or another, to the tomato and its redness.This helps them to account for phenomenology of particularity and explain how experience grounds demonstrative thought.It would be genuine progress if these were the two views left standing, that is, if Particularism were the best version of Representationalism. Then the debate between Naïve Realism and Representationalism would be only over their fundamental difference.Namely, the fact that even though both claim that in perceptual experience you're related to objects and property-instances, Naïve Realists further insist that the relation at least partly determines the experience's phenomenal nature or character while Particularists deny this.On the standard construal, this is taken to further entail that perceptual experiences and indistinguishable hallucinations can't share their phenomenal nature or character ( = Disjunctivism ). 4 In contrast, the Particularist view allows for veridical experiences and hallucinations to share their character and avoids Disjunctivism.Thus, if Particularism were the best version of Representationalism then the debate between Naïve Realism and Representationalism would reduce to the debate over whether the relation determines phenomenal character, and given standard assumptions, Disjunctivism.
Particularism is therefore an intriguing and important view.Unfortunately, it faces a heretofore unnoticed problem.Namely, its supporters haven't told us how objects can be bound together with property-instances into a content such that it represents them and has accuracy-conditions.Since this problem is comparable to the problem of the Unity of the Proposition , we can call it a version of the problem of the Unity of Perceptual Content .
My first aim in this paper is to present the problem and argue that Particularists can't solve it.In contrast, Attributionists can say that objects are bound together with properties into a content because the latter are attributed to the former (Burge 2010(Burge , 2022 ; ;Rescorla 2020 ).This establishes Attributionism as the only Representationalist game in town.I will further argue that Attributionism, properly understood, can capture some purported Naïve Realist insights that the Particularists are after, but not all of them.As a result, the debate between Naïve Realism and Representationalism doesn't reduce to the debate over Disjunctivism.
I will proceed as follows.I fill first set things up in more detail by explaining how we should think of property-instances for the purposes of philosophy of perception (Section II ).I will then explain three arguments for thinking that we experience property-instances and make clear the perceived advantages of Naïve Realism that Particularists want to secure by taking us to represent them (Section III ).Next, I will present Schellenberg's view, present the version of the problem of Unity of Perceptual Content it faces, and argue that it can't solve it (Sections IV and V ).Finally, I'll explain how Attributionism can solve the problem with the help of perceptual attribution of properties and clarify which Naïve Realist advantages it can secure and which it can't (Section VI ).

II HOW TO THINK OF PROPERTY-INSTANCES
Naïve realists tell us that we experience not only objects, but their propertyinstances: …when we see an object -such as the Pacific Ocean or the Taj Mahalit is not simply blueness or pinkness we are aware of, but specific instances of blueness and pinkness; the blueness of the Pacific Ocean and the pinkness of the Taj Mahal.(Fish 2009 : 13) …it is not the fact of the leather's being red that falls within the ambit of experience, but a quite different thing, namely the redness of the leather, which is a particular exemplification of redness.(Johnston 2014 : 128)

II.1 Particularists agree. But how should we think of property-instances?
Let's set aside wholesale scepticism about property-instances since the views of perception we're interested in this paper assume them. 5For our purposes, it is essential to understand the difference between property-instances, metaphysicians' repeatable properties or universals, and what philosophers of language (hereafter semanticists), call properties, things with satisfaction-conditions. Property-instances are plausibly designated by expressions like 'o's redness', 'the tomato's redness' (Kriegel 2005 : 85;Moltmann 2013 : Ch. 2).Metaphysicians' properties are designated by expressions like 'redness'.Semanticists' properties are designated by expressions like 'being red', 'the property of being red', and ' λx: x is red'.
On one conception of property-instances they are abstract particulars .They're abstract not in the sense of existing outside of space and time, but in the sense of contrasting with concrete.One rough-and-ready way of explicating the concrete-abstract and universal-particular distinctions that will do the trick for us here is spatiotemporal (Kriegel 2004 : 9;2005 : 87).Concreta exclude other things from being at their location, abstracta, if located at all, don't.Particulars can exist only at one location at a time, universals, if located at all, can be multiply located.Objects hegemonize their locations and are therefore concreta. 6Property-instances and properties don't and are thus abstracta.Objects and property-instances can only exist at one location at a time and are therefore particulars.7 Properties don't and are thus universals.
What we can and should leave open for the purposes of philosophy of perception is the further ontological status of objects and property-instances conceived of as abstract particulars vis à vis each other.On the historically most prominent view, objects are independent of their property-instances while property-instances are dependent on objects.In such traditions propertyinstances are frequently called modes or moments (Heil 2012 ;Lowe 2005 ;Mulligan, Smith & Simons 1984 ).However, in the 20th century many who talked about property-instances took objects to be nothing but bundles of them and thus reversed the dependence.In such a tradition property-instances are usually called tropes (Campbell 1981(Campbell , 1990 ; ;Ehring 2011 ;Williams 1953 ).But there is no reason for our purposes to take a stand on this issue.Note also that according to both traditions property-instances form classes or kinds which can then be taken to be the repeatable properties like redness.
On a very different conception of property-instances they can be thought of as immanent universals considered in abstraction of their bearer and their other instances .Armstrong's influential views of properties placed repeatable properties, universals in the centre (Armstrong 1989 ).His Aristotelian immanent universals are supposed to be wholly located wherever they're instantiated.But a notion of property-instance can be constructed even on this view: it's the universal that is wholly located, considered in abstraction from its bearer and its other instances.On this view it is this thing that we refer to with 'o's redness'.
Insofar as we need the notion of a property-instance in philosophy of perception we can and should remain entirely neutral on how they're metaphysically to be thought of.
Both property-instances ('o's redness') and metaphysicians' properties ('redness') should be sharply distinguished from semanticists' properties ('being red').Metaphysicians repeatable properties are, depending on how one thinks of them, kinds of modes/tropes or universals.They are things that have instances not things that have satisfaction-conditions or things that are true of objects.Metaphysicians typically, though not universally, deny that every wellformed predicate expresses a property.For example, most think that there is no universal that corresponds to 'being not red'.In contrast, semanticists' properties are semantic values of predicates and as such they're not things that have instances, but things that have satisfaction-conditions or things that are true of objects.They have predication, as if, already built in (Lowe 2008 : 179-81).These are universally taken to be abundant in that every well-formed predicate expresses one.For example, the semanticists' property of being not red doesn't have any instances, but is satisfied, true of an object o iff o is not red.It should be obvious, though it's troublingly commonplace not to notice, that metaphysicians' properties and semanticists' properties are completely different sorts of entities (for comments pressing this point see Johnston 2011Johnston : 194-5, 2014 : 138; : 138;Lowe 2008 : 179-81).This will become important a bit later.

III EXPERIENCE OF PROPERTY-INSTANCES
Why think that we experience property-instances?What perceived advantages of Naïve Realism do Particularists want to secure by claiming that we represent them?In this section, I will review three different arguments for the claim that we experience property-instances: a causal one, a phenomenological one, and one from grounding thought.The last two arguments will make clear the perceived advantages of Naïve Realism that Particularists want to secure.

III.1 The causal argument
The causal argument has been offered by metaphysicians arguing for the importance of property-instances, but it has received considerably less discussion from philosophers of perception (Lowe 2005 : 24-5, for brief discussion see Nanay 2012 : 6-7).It is very simple.Perception is a causal relation.The relata of causal relations are property-instances.Thus we are perceptually related to property-instances.Something like this argument is suggested by Schellenberg in the following passage: Perceptual relations are a kind of causal relation.So when we perceive, say, the shape of the cup in front of us, that shape must be causally efficacious-Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 otherwise we could not perceive it.Thus, given plausible assumptions about causation, the shape of the cup must be a concrete spatio-temporal particular rather than a universal.After all, universals are neither spatio-temporally located nor causally efficacious.I will assume an Aristotelian view on which properties are understood in terms of their instances.Hence, I will assume that we perceive property-instances (Schellenberg 2018 : 15).
Even if one grants that the relata of causal relations are property-instances, there is an easy way to resist this argument.Consider your experience of a tomato.One might simply argue that even though in this case it's the tomato's redness that is causally efficacious, that it is what causes your experience, it is not what you experience.You don't experience the tomato's redness, but just experience the tomato as being red. 8o make progress it must be established that it is plausible that we experience o's redness.This is where the further arguments come in.

III.2 The phenomenological argument
Most early discussions of particularity in perception targeted the view of perceptual content as general, existential content and were focused on objects (Burge 1991 ;Soteriou 2000 ;Martin 2002a ).Let's discuss these arguments first since they provide a guide to the arguments targeting property-instances.
There are two different sorts of phenomenological arguments related to phenomenal particularity of objects: a cheap one and a contentious one.First, experiences clearly at least seem to be present us with particular objects (Schellenberg 2018 : 17; French and Gomes 2019 : 42-3, for scepticism, see Hill 2021 : 13-15).We can call this datum Phenomenology of Object Particularity .This is to be interpreted as a cheap, uncontroversial claim about all experience, even hallucination.It is unclear whether Generalists can account for this since their account of content, something which they take to determine phenomenology, doesn't mention any particulars.Whether this is true is not important here.What matters is that this is one reason why many representationalists take perceptual contents to be singular and of the forms: Of course, Naïve Realists don't think that this helps.They offer also a more contentious phenomenological argument.Experiences not only seem to be as of particular objects, but such objects must be part of the phenomenal nature or character of the experience.Otherwise, we would be cut off from the objects in the world (for discussion, though not support, see Martin 2002b : 397-9).Thus, if the above conception of singular content is developed in a non-disjunctivist manner such that phenomenal character is determined by some object-related aspect of the content that can also present in hallucination, then this cuts us off from objects even in the veridical case.However, this is not a problem if the singular content view is developed in a disjunctivist way (McDowell 1986 ).
But Naïve Realists are not satisfied with views like McDowell's on which experience has disjunctivist singular content since it leaves out our relation to property-instances.This is why many recent discussions of phenomenology of particularity focus on properties instead (Fish 2009 : 22-3;Genone 2014 : 347-50).
There are, again, cheap and contentious phenomenological arguments.First, and supposedly more neutrally, it is claimed that the properties experiences present us with at least seem to be property-instances (Fish 2009 : 22-3;Genone 2014 : 347-50).One argument that could be given to support this relates to change.Suppose you see a red tomato turn into black because it's subjected to a blowtorch.It might be argued that in this case you see the tomato's redness get destroyed and see it acquiring a new color, blackness (Lowe 2008 : 185-6).But this presupposes that you see the tomato's redness.We can call this datum Phenomenology of Property Particularity . 9Again, this is to be interpreted as an uncontroversial claim about all experience, even hallucination .There has not been much discussion whether Attributionism can account for this.We will come back to this in the last section.What matters is that this is one reason why particularists take perceptual content to involve property-instances (Nanay 2012 ;Schellenberg 2018 ).
Of course, Naïve Realists won't think that this helps in this case either.They offer again a more contentious phenomenological argument.Experiences not only seem to be as of property-instances, but such property-instances must be part of the phenomenal nature or character of the experience.Otherwise, we would be cut off from the property-instances in the world (Genone 2014 : 348-9).Thus, if the above conception of particular content is developed in a non-disjunctivist manner such that phenomenology is partly determined by some aspect of the content that is also present in hallucination, then this 9 Note that this is intended by the above Naive Realists as a datum about our experience of sensory qualities like redness.Almang has argued that we experience property-instances on the basis of experience of property constancy .It's uncontroversial that we can experience an object as unchanging over time despite changes in our experience of it.Almang points out that we can also experience an object's property as unchanging despite changes in experiences of its presented quality.It's one thing to see an object change in color like in the above tomato case.It's another thing to see an object as remaining the same in color despite changes in the presented sensory quality, for example, when you move closer to tomato T1 and your experience of it becomes consecutively more determinate: from dark to red to tomato red (Almang 2016 : 46).Almang argues that the only way to make sense of this is if we take the unchanging color to be a property-instance or a trope.But this is quite a different claim than the datum at issue above on which T1's darkness, T1's redness, etc. themselves are property-instances.So Almang's point can't be used to support Phenomenal Property Particularity as it is intended.An interesting further line of inquiry is whether intuitions supporting Phenomenal Property Particularity can be defeated by taking only the constant properties to be property-instances.Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 cuts us off from property-instances even in the veridical case.Nothing short of Disjunctivism will help.But at this point Particularists will balk and claim that by capturing the Phenomenology of Property Particularity they've secured all the advantages of Naïve Realism they want to, while avoiding Disjunctivism.

III.3 The thought-grounding argument
Most early discussions of particularity in relation to grounding demonstrative thought again targeted the view of perceptual content as general, existential content and were focused on objects.
Experience seems to ground our abilities for demonstrative thought about objects (Campbell 2002 : ch. 6, for criticism see Burge 2022 : 401-5).Take the following case.You see a red tomato T1 and then close your eyes at which time the tomato is swapped for a qualitatively indistinguishable tomato T2.Your experience before the swap enables you to think about T1, the experience after enables you to think about T2.Generalists arguably can't accommodate this (Martin 2002a : 182, for discussion see Schellenberg 2018 : 19-22).It is common to think that the only way experiences can play this role is when they're individuated in terms of the particular objects they're experiences of.We can call this thesis Relational Object Particularity .Naïve Realists accommodate it by taking experiences to consist in a genuine relation to objects such that it affects their phenomenal nature, meaning that experiences of distinct tomatoes have differing phenomenal natures , even if they share their qualitative character (Martin 2002a : 194-5). 10Representationalists can accommodate it by taking experiences to have singular content in the object-place while allowing that this doesn't affect phenomenal character, meaning that experiences of distinct tomatoes share their phenomenal character. 1110 This relies on Martin's distinction between phenomenal nature and phenomenal character, or, what I think is the better term, qualitative character .Naïve Realists think that an experience's conscious character is determined by acquaintance with particular objects and their propertyinstances.Thus, particular objects and property-instances figure in an experience's conscious character ( = phenomenal nature).However, take two experiences of qualitatively indistinguishable objects.Such experiences will have differing phenomenal natures, yet there's clearly an aspect of their conscious character that they share.This aspect can be called qualitative character.For discussion, see French and Gomes ( 2019 : 47-53).Also relevant is Beck's distinction between broad and narrow conceptions of phenomenology (Beck 2019 ).
11 As above, Naive Realists are not satisfied with this and give further arguments to the effect that representationalists can't explain how experience grounds capacities for thought since the formation of perceptual representations seems to presuppose such capacities (Campbell 2010 : 202).Since at least some concepts are perhaps acquired via experience, this argument has perhaps bite against those who think perceptual content is conceptual.However, it is not particularly forceful against those who take it to be non-conceptual.In fact, many representationalists have themselves argued that the ability of experience to ground conceptual thought entails that experience has non-conceptual content (Peacocke 2001 ;Roskies 2008 ).And many representationalists think that the non-conceptual capacities involved in the formation of perceptual representations are not acquired via experience, but innate (Burge 2010(Burge , 2022 ; ;Schellenberg 2018 ).Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 Much as with objects, experience might be thought to ground our capacity for demonstrative thought of property-instances.Taking the above case, your experience before the swap enables you to think of T1's redness, the experience after enables you to think of T2's redness.And again, it seems that the only way that experiences can play this role is when they are individuated in terms of the actual property-instances they're experiences of.We can call this thesis Relational Property Particularity .
Generalists and Attributionists arguably can't accommodate this since they take content to involve only properties.Naïve Realists can accommodate it by taking experiences to consist in a genuine relation to property-instances such that it affects their phenomenal nature, meaning that experiences of distinct tomatoes have differing phenomenal natures, even if they share their qualitative character.Particularists try to accommodate it by moving to a conception of perceptual content that involves property-instances and takes us to be related to such instances in the veridical case (Nanay 2012 : 3-5; Schellenberg 2018 : ch. 1).In the next section, we'll take a closer look to Schellenberg's view and how she proposes to do this.

IV FREGEAN PARTICULARISM
Schellenberg's Particularism is developed in a Fregean fashion as follows.In experiencing one is employing general capacities that function to discriminate and single out objects and property instances under MOP-s where singling out is a non-conceptual analogue of reference (Schellenberg 2018 : 25).The capacities themselves are semantically general, that is, repeatable, and their use yields the following content schema (where ' O ' signifies an object-place and ' F ' a property-place): In the case of your two tomato experiences the use of the general capacities results in singling out the relevant tomatoes and their property-instances and the resulting token contents are thus supposed to be the following: These two contents are supposed to be accurate just in case T1 has the relevant property-instance and T2 has the relevant property-instance.
In contrast, when you're under an illusion and experience T3 as red while it is yellow, the MOP RED is gappy in the sense that one is not related to a Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 particular property-instance.And when you're hallucinating a red tomato, both MOP-s are gappy: The first content is accurate just in case T3 has some property-instance of the relevant kind.The second content is accurate just in case some object of the relevant kind has some property-instance of the relevant kind.However, since in both cases at least some of the relevant particulars one is supposed to be related to are missing, such contents are necessarily inaccurate (Schellenberg 2018(Schellenberg : 94-5, 2019 : 745-6) : 745-6).
With this account of the application of general capacities and the resulting particular contents at hand, she puts it to work to secure the perceived advantages of Naïve Realism while avoiding Disjunctivism.First, such contents explain why experiences at least seem to be as of particular objects and property-instances and can therefore capture both the Phenomenology of Object -and Property-Particularity .Secondly, in the case of veridical experiences, you are singling out both the object and the relevant property-instance and this can therefore satisfy both Relational Objectand Property-Particularity .However, since it is the application of the general capacities that grounds the experience's phenomenal character, it can be the same in all four cases.
It is important to note that to capture the Phenomenology of Particularity the application of the relevant general capacities for singling out is supposed to suffice.Even hallucinations exhibit phenomenology of particularity.However, to capture Relational Particularity the applications need to single out an actual object and/or property-instance.Illusions don't ground capacity for thought of property-instances and hallucinations don't ground capacity for thought of anything.So, the actual relations to the objects and property-instances matter.In claiming that such relations exist, Particularism gets maximally close to Naïve Realism, just differing on the question whether such relations influence phenomenology.
The main question about Particularism is whether it can really make sense of the above sort of perceptual content.In the next section, I will argue that it can't.

V UNITY OF PERCEPTUAL CONTENT
How can objects and property-instances be bound together into a content such that it represents them and has accuracy-conditions?This is the version of the problem of the Unity of Perceptual Content that Particularism faces.We can Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 get a sense of the problem when we first look at the analogous problem about propositional representation.
Those who trade in Russellian or Fregean structured propositions usually model them as follows: < o, F-ness > or < MOP(o), MOP(F-ness) > .But these structured n-tuples are not themselves propositions, but mere models of the actual things which must have their truth-conditions essentially.The question is how to make sense of the actual things as some combination of the elements such that the combination has truth-conditions.This is one version of the famous problem of the Unity of the Proposition (King 2007 : ch.1; Soames 2010 : ch.2; Hanks 2015 : ch.2).The Platonist option, as always, is to just postulate the relevant entities with their relevant properties and call it a day (Merricks 2015 ).However, the recently popular Naturalist, Act-Based views explain this by appealing to the act of predication and making propositional acts more basic than propositions (Soames 2010 ;Hanks 2015 ;Reiland 2022 ).Such views are best understood in a two-step manner.In the first step, it is claimed that when a subject predicates a property like redness of an object o, they perform an act with propositional content: for example, a judgement that o is red which is true iff o is red.This is where unity is secured.In the second step, it is then claimed that a proposition itself can be identified with some derivative entity, for example a type of such an act.Now take Schellenberg's view on which perceptual contents are supposed to be the following: As before, such structured n-tuples cannot themselves be the contents but are mere models of the actual things which must have their accuracyconditions essentially.Much like in the case of propositional representation, the question is how to make sense of the actual things as some combination of the elements such that the combination has accuracy-conditions.
It would be natural for Schellenberg to try to appeal to an act or operation analogous to predication .In other words, once the object and property-instance have been singled out, it would be natural to appeal to some act that binds or glues them together.But it is very important to be clear that not any form of binding will do.On her view they need to be bound together not just into any entity, but a perceptual content , something that has accuracy-conditions.After all, Schellenberg holds a broadly Fregean view of content on which they're things which themselves are true or false, accurate or inaccurate and thus have truth-or accuracy-conditions. 12 Contents are one thing, their truth-and accuracy-conditions, the conditions under which they're true or accurate are another.Their truth-and accuracy-makers are yet another thing (compare Schellenberg 2018 : 92-3).Thus, on her view, the object and property-instance need to be bound together into a content , something with accuracy-conditions.
To illustrate the importance of being clear about what binding needs to do, consider one view of what she could take it to consist of.In a related discussion of how Schellenberg can make sense of illusory cases, McGrath considers the view that such binding could be generated by the application of a capacity to single out instances of co-location (McGrath 2019 : 737).Consider again the case where you're under an illusion and experience T3 as red while it is yellow: McGrath first asks what binds the two parts together and considers the view that this is done by the application of a further capacity to single out instances of co-location.His problem with this is that in the illusory case there is no such instance of co-location.But there is a more fundamental problem with the appeal to co-location.Even in the veridical case where one would single out an object, single out a property instance, and single out an instance of their co-location we do not have anything that amounts to a perceptual content, something that has accuracy-conditions.After all, singling out an instance of co-location is not something that can be accurate or inaccurate. 13he basic worry about Schellenberg's view is that it faces an in-principle obstacle in making sense of binding in a way that generates accuracyconditions.Namely, that it just features two particulars and two particulars are of the wrong logical or metaphysical categories to be able to form a content.Much like you can't make a subject-predicate sentence out of two singular terms nor a proposition out of two objects, you can't make a perceptual content out of two particulars.And it is not hard to see why.The only known representational and have truth-conditions.This notion is not only adopted by direct followers of Frege, but all of those who think that propositions are representational, even if they reject lots of other aspects of Frege's picture.For example, King, Soames, and Hanks all think that propositions qua contents are things that themselves represent or have truth-conditions and proceed to give theories of how that can be (King 2007 ;Soames 2010 ;Hanks 2015 ).Similarly, Burge thinks that perceptual contents are things that represent or have accuracy-conditions (Burge 2010(Burge , 2020)).It contrasts starkly with the Russellian notion of content on which it is an entity at the level of reference , the world.On such views contents are not things that are representational and have truth-conditions, but they are the truth-conditions of thoughts, the things they are about .
way of thinking of the sort of binding that generates accuracy-conditions is as amounting to categorizing or sorting things into classes or kinds based on what they have in common (Burge 2022 : 44-5;Hanks 2015 : 64-5).And to categorize or sort things into classes we need as a principle of categorization or sorting something general , something that different things can share or be characterized by.This is why the above Act-Based views take us to predicate properties.But Schellenberg's view features only property-instances.And we can't categorize or sort things into classes or kinds based on property-instances which are by their nature particulars and not something different things can share or be characterized by.It doesn't matter that they are abstract particulars.They are still just particulars and not something general.
To sum up, not only do we have no clue what it would be to bind an object o together with a property-instance like o's redness in a way that the resulting entity is accurate or inaccurate, it seems that this can't be done.This is a particularly egregious instance of a missing case of glue. 14 In the next section, I will argue that Attributionism can easily solve the version of the same problem they face because they take us to attribute general properties, and discuss Schellenberg's objections against it, as well as which Naïve Realist advantages it can and can't secure.

VI FREGEAN ATTRIBUTIONISM
Let's go back to Attributionism which has been developed by Burge and defended by Rescorla (Burge 2010(Burge , 2022 ; ;Rescorla 2020 ).We will develop it here in a fashion that is maximally similar to Schellenberg's Fregean Particularism to avoid noise. 15In fact, Schellenberg's earlier work can perhaps be interpreted in this vein (Schellenberg 2010 : 35).On this view we claim that 14 McGrath and Rescorla have come closest to recognizing this problem for Particularism, but neither spells it out nor emphasize how fundamental it is for making sense of Schellenberg's view (McGrath 2019 : 737;Rescorla 2020 : 725).One might wonder whether appealing to featureplacing views of early vision might help here.On one model of how this works that derives from Treisman's work and has been extensively developed by Austen Clark, there is a part of early vision that refers to sensory features, a part that places them at locations at the sensory field, and then some further part that binds them together via their represented locations into a proto-object (Clark 2004 ;Gelade and Treisman 1980 , for criticism see Burge 2022 : ch.10).Don't feature-placings consist of binding sensory features, which can perhaps be thought of as property-instances, together with sensory locations, another particular?Yes, but again, featureplacings aren't, on their own, representational!For this you need an independent further story on which the sensory feature stands in some sort of representational relation to an actual feature in the external world and the sensory location stands in a representational relation to an actual location in the external world, etc.Perhaps a story like this could be given and perhaps this is the direction Schellenberg should go with her view, but it is very different from what she's given us thus far. 15Burge takes the form of perceptual content comparable to the complex demonstrative 'This F' in that the attributive 'F' guides reference and doesn't feature in pure predication like in Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 in experiencing one is employing general capacities for singling out objects and attributing properties to them where attribution is a perceptual analogue of predication.The capacities themselves are semantically general, that is, repeatable, and their operation yields the following content schema: In the case of your two tomato experiences the use of the general capacities results in singling out the relevant tomatoes and attributing redness to them and the resulting token contents are thus the following: As a result, you represent the tomato as being red.These two contents are accurate just in case the relevant object is red.
In contrast, when you're under an illusion and experience T3 as red while it is yellow you still attribute redness to it.As a result, you represent the tomato as being red.And when you're hallucinating a red tomato, there is seeming reference which in fact fails, but you still attribute redness to the seeming referent: The first 16 content is accurate just in case T3 is red.The second content is accurate just in case some object of the relevant kind is red.However, since in this case the relevant particular you're supposed to be related to is missing, this content is necessarily inaccurate.
In contrast to Particularism, this Attributionist view can solve the version of the problem of Unity it faces.Much like Act-Based views which tell you that the case of 'This is F' (Burge 2010(Burge : 541-2, 2022 : 156-9) : 156-9).This is an important difference, but we will disregard this for the time being (though see fn. 17).
16 This reflects a recent debate between Rescorla and Schellenberg on the contents and accuracy-conditions of illusions.Rescorla says that intuitively the accuracy-conditions of a veridical experience of a red tomato and an illusion of a yellow tomato as red are the same (Rescorla 2020 : 722-4).This is so on the above Attributionist view, but not on the Particularist view on which the content of the illusion is gappy in the property-place.Schellenberg responds by insisting that the accuracy-conditions are not the same since the veridical experience is accurate only if you're related to the actual property-instance while this is not the case for the illusion (Schellenberg 2019 : 746).Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 propositional representation results from predicating a property of an object, this view tells us that perceptual representation results from attributing a property to an object.The solution is exactly the same in both cases.If it works in one case, it works in the other.
For some reason Schellenberg moved away from her earlier Attributionistlike view on which contents were built out of objects and properties to her current view on which they're supposed to be built out of objects and propertyinstances (Schellenberg 2010: 35 vs. Schellenberg 2018 : 67-9).Why the move away from perceptual attribution of properties?As we saw above, causal considerations mentioned by Schellenberg fail to have any bite since one can grant that in your tomato experience the tomato's redness is causally efficacious in causing your experience while still thinking that what you experience is not the tomato's redness, but just the tomato as being red.
A more serious worry is that Attributionism can't account for the Phenomenology of Property-Particularity .To recall, the datum is that experiences at least seem to present us with property-instances.The worry is that thinking of content and phenomenology in terms of attribution of properties, of generalities, leaves property-instances out of the story.The basic challenge is that if predication of properties doesn't, in any way, feature property-instances then why would perceptual attribution do so?
But this is too hasty.Predication and its relation to metaphysicians' properties are generally poorly understood.Most of the models we have of predication take the things involved to be semanticists' properties.For example, take the Frege-Tarski model of predication on which it consists of something like feeding an object to a function from objects to truth-values (for discussion, see Rescorla 2009 : 176-7).A function from objects to truth-values has, as if, predication already built in.It is essentially predicative (Lowe 2008 : 179-81).But once we start thinking of predication or attribution as a separate act or operation we or our sub-personal systems perform, we should move to thinking of the things we attribute, in at least the basic cases, as metaphysicians' properties, universals instead (compare Burge 2022 : 159-60).Then we can think of predication of redness of o as consisting of taking it to have an instance of redness.Similarly, we can think of perceptually attributing redness to the tomato as consisting of taking it to have an instance of redness.
Thus, take the above models of content: Such models are best understood as saying that the subject does something like referring to an object T1 or T2 under some MOP, indicating redness under RED and then attributing the property to the object.When are these experiences accurate?The first one is accurate when T1 is red.But if to attribute Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 redness to o is to take o to have an instance of it then this just means that it is accurate when T1 has an instance of redness.The accuracy-condition of the experience is that T1 has an instance of redness and the accuracy-maker is the actual red T1.The second one is accurate when T2 is red.And this just means it is accurate when it has an instance of redness.The accuracy-condition is that T2 has an instance of redness and the accuracy-maker is the actual red T2.Thus, even when the content is formed by attributing universals, it's the instances that figure in the accuracy-conditions.This seems sufficient to account for Phenomenology of Property-Particularity .On Attributionist views it is attribution that grounds phenomenology, that is, whether you're actually related to the property-instance doesn't matter.However, if we think of attribution of a universal to o in terms of taking o to have an instance of it, then we can see how attribution can make it the case that you're at least seemingly presented with a property-instance.
The most serious worry is that the above form of Attributionism can't account for Relational Property-Particularity .To recall, the supposed datum is that experiences ground our capacity for demonstrative thought of propertyinstances.For experiences to do this, they have to be individuated in terms of the actual property-instances they're experiences of.Schellenberg tried to accommodate it by taking us to be related to such instances in the veridical case.However, on the above form of Attributionism we do not get such relations to actual instances.To make this vivid, on the Attributionist view a veridical experience of a red tomato and the illusion of a yellow tomato share their contents and accuracy-conditions.In both cases, one attributes redness to the tomato and thereby takes it to have an instance of redness.But this means that even in the veridical case neither the content nor the accuracyconditions are individuated in terms of the tomato's actual instance of redness.Thus, the above form of Attributionism remains a further step removed from Naïve Realism than Particularism.But to see whether this is a cost, more work needs to be done on how important it is to take us to be experientially related to property-instances as well as whether more sophisticated forms of Attributionism might be able to accommodate such relations. 17 Are there other reasons why one might object to Attributionism from the Particularist point of view?Schellenberg claims that it is discrimination that is basic and constitutive of perception, and not attribution: The reason I focus throughout on discriminatory, selective capacities is that these low-level capacities are constitutive of perception.It is not clear what it 17 Of course, if Attributionists think that you can refer to property-instances then they can allow relations to them in the object place, as in < MOP O (o's redness), MOP VIBRANT (vibrancy) > .This by itself doesn't help with capturing Relational Property Particularity which is about relations to property-instances in the attribute-place.However, some Attributionists take a more sophisticated view of perceptual content where it always involves referential relations to both objects and property-instances as well as attributions of properties to both (Burge 2022 : 165-6).I leave discussion of such views to future work.Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 would be to perceive a particular without at the very least discriminating that particular.I do not disagree with Rescorla that there might be capacities employed in perception that are not discriminatory, selective capacities, such as for example attributional capacities.But I see no argument that those capacities are necessary for perception (Schellenberg 2019 : 747).
However, one can grant that discrimination of particulars is basic and plays a fundamental role while arguing that attribution of properties is necessary as well.The Unity problem provides the missing argument for the claim that attributional capacities are necessary for perceptual content.It's not that one can't analyze experience by appealing only to discrimination.This is what Naïve Realists do.But the Unity problem shows that this is not open to Representationalists who posit perceptual content with accuracy-conditions. 18 VII.CONCLUSION Particularism is an intriguing and important view since it gets as close as possible to Naïve Realism, which is arguably the best articulation of our pretheoretic view, while being able to avoid Disjunctivism, which many view with suspicion.As I suggested in the beginning, it would be genuine progress if Particularism was the best version of Representationalism. Then the debate between Naïve Realism and Representationalism would reduce to the debate over Disjunctivism.
I have argued, however, that Particularism fails in the face of a version of the problem of the Unity of Perceptual Content .Attributionism with its view that perceptual content results from the attribution of properties is the only Representationalist game in town.However, as developed here, it is a further step removed from Naïve Realism in not featuring any relations to actual propertyinstances.Hence, it is not true that the debate between Naïve Realism and Representationalism reduces to the debate over Disjunctivism and progress 18 Schellenberg mentions two other reasons to avoid Attributionism, both of which seem to me to be problematic.First, she claims that Attributionism can't make sense of cases where we perceive only property-instances.She mentions olfactory, gustatory, and tactile experiences without giving any examples (Schellenberg 2018 : 68).However, there is a danger of changing the subject here.It's easy to see how in having sensations , whether visual sensations (phosphenes or 'seeing stars'), auditory ones like ringing in the ears or tactile ones like itches, we're not related to objects.But sensations are not perceptual experiences.In their case it makes no sense to talk about hallucination.It's unclear whether something counts as genuine perceptual experience when it doesn't exhibit objectification and constancies (for discussion, see Reiland 2015 : 513-6).Furthermore, Attributionism can make sense of perception of only property-instances by taking us to refer to them (Burge 2022 : 163-4).Secondly, she claims that appealing to attribution runs the danger of overintellectualizing perception in giving it proto-sentential structure, whereas perception can be iconic (Schellenberg 2018 : 69).However, it's not clear that there is much to this charge.First, Attributionists can construe of perceptual capacities as low-level, innate, nonconceptual ones (Burge 2010(Burge , 2022 ) ).Secondly, appealing to attribution is entirely compatible with non-sentential structure and iconicity (Burge 2010(Burge , 2018(Burge , 2022 ) ). Downloaded from https://academic.oup.com/pq/advance-article/doi/10.1093/pq/pqad105/7379039 by Vienna University Library user on 16 November 2023 on it can be also made by thinking about how important it is to take us to be experientially related to property-instances.

FUNDING
This work was funded by Austrian Science Fund (Grant number: M 3373-G).