- Title: The Quality of Thought - Author: David Pitt - Tags: #summary #llmgenerated #toread #book # Introduction ## I.1 Analytic Phenomenology - [[The author]] advocates for an approach called "Analytic Phenomenology" which brings the ethos and methodology of [[analytic philosophy]] to the study of experience from the first-person perspective. > Analytic Phenomenology sets aside the problematic relation between experience and the brain, in order to study experience independently of the vexed metaphysical issues that have long plagued analytic philosophers of mind. This Phenomenology is analytic because it brings the ethos and methodology of [[analytic philosophy]] to bear on the study of experience per se. It is Phenomenology because it is the study of experience from the first-person perspective. (Introduction, p. 2) - Analytic Phenomenology resists [[the idea]] that experience must be ultimately explicable in naturalistic terms. It focuses on studying experience on its own terms using disciplined first-person methodology. ## I.2 A Plea for Agnosticism - [[The author]] argues we should be agnostic about whether consciousness is a physical phenomenon or not. There are reasons on both sides: > We have at once good reason to think that [[conscious experience]] could not be physical (i.e., its manifest subjective and qualitative nature), and good reason to think that it must be (its intimate relation to physical phenomena). How should we respond to this antinomy? It seems to me that the counsel of wisdom is agnosticism. (Introduction, p. 12) - In developing the phenomenal intentionality of thought thesis, [[the author]] recommends being noncommittal on the ultimate metaphysical nature of consciousness and [[its relation]] to the physical. ## I.3 The Plan of the Book - The book defends and develops the thesis that conceptual/propositional content is experiential. This is a "pretty radical" view in relation to traditional [[analytic philosophy]] of mind and language. > The thesis that conceptual/propositional content is experiential is, in relation to traditional [[analytic philosophy]] of mind and language, pretty radical (some might say heretical). I do not think it has been widely appreciated, even among those who endorse it, just how radical the thesis is, how far-ranging its consequences are, and how uncompromising its development must be. Many "standard" doctrines in philosophy of mind and language must be thoroughly revised, or discarded entirely, if it is true. (Introduction, p. 18) - The chapters address challenges for this thesis, including indexical concepts, singular thought, unconscious thought, and provide alternative experiential accounts. # Chapter 1: Phenomenal Intentionality ## 1.1 Causal-Informational-Teleological Theories of [[Conceptual Content]] - Causal-informational theories of [[conceptual content]], inspired by Dretske's work, face serious internal difficulties in spite of their promise. These include: 1) The proliferation problem - concepts seem to represent anything that causes them 2) The disjunction problem - concepts get assigned indeterminate or wildly disjunctive contents 3) The stopping problem - there are too many candidate links in the causal chain - Attempts to solve these problems by appealing to learning, teleological function, or asymmetric dependence are unsatisfying. ## 1.2 Phenomenal Intentionality of Thought - The problems for causal-informational theories arise from excluding the subjective aspects of intentionality. The phenomenal-[[intentionalist approach]] avoids them. > There have all along been dissenters from causal-informational theories, who have insisted that they fail because they ignore the essential role of experience in constituting intentionality. (Chapter 1, p. 24) - On this internalist view, conceptual intentional content is entirely fixed by and identical to internally instantiated experiential properties of the proprietarily conceptual kind. ## 1.3 Phenomenal Intentionality of Perception - [[The author]] discusses [[the role]] of phenomenal character in determining the intentional content of perceptual states. Two contrasting views are presented: 1) The Pure Causal View - phenomenal character plays no role; a state is a perception of an object/property iff it is caused by that object/property 2) The Matching View (Montague) - there is a limit to how badly the phenomenal character can misrepresent the cause before the state ceases to be a perception of it - The truth lies between these extremes. Causation is necessary for genuine perception, but resemblance also plays an essential role. ## 1.4 Content Internalism ### 1.4.1 Propositional Hallucination - Reflection on perceptual hallucination motivates belief in internal, mind-dependent representations. The mere possibility a representation could occur without what it seems to represent suffices to show this. > If the possibility that it can appear to us that we are perceptually representing an object o when there is no such thing should lead us to believe in internal perceptual representations, then the possibility that it can appear to us that we are conceptually representing the proposition that p though there is no such thing should lead us to believe in internal conceptual representations. (Chapter 1, p. 35) - By parity of reasoning, the contents of hallucinatory conceptual representations must be internally determined. And veridical and hallucinatory contents do not differ. ### 1.4.2 Propositional Illusion - Conceptual illusion, unlike perceptual illusion, is impossible. There is no mind-independent connection to a proposition that could ground misrepresentation of it. > For what would such a thing be? It would be a case in which what one seemed to be thinking (where, again, this is not assumed to be some kind of conscious appearance) was not in fact what one was thinking. It would be a case in which what one was representing—a proposition, some kind of mind-independent thing—did not have the properties one represented it as having: the object of one's thought is, say, the proposition that cats are finicky, whereas one's mental representation is as of the proposition that cats are feline. (Chapter 1, p. 36) - Since conceptual illusion is impossible, conceptual contents must be determined by mind-internal properties of the thinker's representations. # Chapter 2: The Experience of Thinking ## 2.1 An Epistemological Argument for Propositional Phenomenology - The main argument is that conscious occurrent thought contents can be consciously identified and individuated introspectively and non-inferentially. But this requires each type of conscious thought to have a proprietary, distinctive, individuative phenomenology. > (K1) It is possible to identify and individuate one's occurrent conscious thoughts introspectively and non-inferentially. But (K2) It would not be possible to identify one's conscious thoughts in this way unless each type of conscious thought had a proprietary, distinctive, individuative phenomenology. Therefore, (P) Each type of conscious thought (each state of consciously thinking that p, for all thinkable contents p) has a proprietary, distinctive, individuative phenomenology. (Chapter 2, p. 39) - This implies propositional phenomenology is identical to propositional content. Phenomenology constitutes rather than represents content. ## 2.2 Knowing What It Is Like and Knowledge by Acquaintance - Non-inferential introspective knowledge of conscious thought content is knowledge by acquaintance, grounded in acquaintance with the tokening of a propositional-phenomenal content. - Acquaintance-knowledge, the fundamental mode of knowledge of phenomenal properties, is not propositional. It consists simply in experiencing conscious tokens of phenomenal properties. > Acquaintance-knowledge is not the same as, and cannot be explained in terms of, knowledge by acquaintance. Perhaps acquaintance-knowledge is what Russell (1911; 1912) meant by 'knowledge by acquaintance'. But there is an important distinction to be made between propositional knowledge based on acquaintance and the acquaintance it is based on. (Chapter 2, p. 43) ### 2.2.1 Mary - When Mary leaves the black-and-white room and sees red for the first time, she gains acquaintance-knowledge of what red looks like simply in virtue of experiencing it. This is a kind of knowledge in itself. #### 2.2.1.1 Retaining Acquaintance-Knowledge - When Mary is not actively experiencing or imagining red, she does not have acquaintance-knowledge of it. Having a capacity to imagine or recognize a quality is not the same as knowing what it is like. > Suppose Mary, still in the room, has never tried to imagine red, but could in fact do so if she tried. If she tries, she will come to know what red looks like. But before she imagines it, she does not know what it looks like, because she has never been acquainted with it. Hence, having the capacity to imagine red is not knowledge of what it looks like. (Chapter 2, p. 45) - Acquaintance-knowledge exists only in the conscious moment of experiencing a phenomenal property. ### 2.2.2 Tye on Knowing What It Is Like - Tye argues acquaintance cannot be identified with knowing what it's like because of the logic of knowledge-wh statements. The author responds: > Tye's argument is invalid because it has the same form as the following obviously invalid arguments: 1a. Samantha knows the color red. 1b. The color red is what my favorite color is. Therefore, 1c. Samantha knows what my favorite color is. (Chapter 2, p. 46) - These arguments equivocate on two senses of 'know' - acquaintance in the premises vs knowing-that in the conclusion. Fixed to use a univocal acquaintance sense, the arguments are valid. ## 2.3 Objections to the Epistemological Argument The author considers three alternative explanations for non-inferential conscious knowledge of occurrent conscious thought content: ### 2.3.1 Reliable Mechanisms - Reliable belief-forming mechanisms cannot explain immediate conscious knowledge of occurrent thought content. Either the higher-order belief is unconscious, in which case the subject's knowledge is not consciously immediate, or it is conscious, in which case its content must be immediately consciously available, requiring phenomenal properties. ### 2.3.2 Extrospectionism - The view that self-knowledge of attitudes is gained by considering outward phenomena rather than inner states (Moran, Byrne) cannot explain direct access to conscious contents. - It implicitly relies on introspective knowledge - to apply an outward-looking rule like "If p, believe you believe p", you must know whether you recognize that p. ### 2.3.3 Levine - Levine distinguishes implicit self-knowledge (thinking with understanding) from explicit self-knowledge (higher-order beliefs) and argues neither requires cognitive phenomenology. - But implicit knowledge cannot be conscious without phenomenology, and explicit knowledge faces a regress - if content is extrinsic to the higher-order representation, self-knowledge would require an infinite series of higher-order states. > You cannot have implicit conscious knowledge of what you are thinking in virtue of tokening an unconscious mental representation. Levine's account does not seem to allow for there to be an epistemic difference between conscious and unconscious thinking. (Chapter 2, p. 58) - Levine's claim that asserting direct introspective knowledge of conscious thought content begs the question fails, as this knowledge still requires explanation. ## 2.4 A Metaphysical Argument for Propositional Phenomenology - For all non-conceptual mental states, sameness and difference within consciousness are phenomenally constituted. By parity, the same should hold for conscious thoughts. - Two principles support this: 1) The Principle of Phenomenal Difference - there can be no difference in consciousness without difference in phenomenology 2) The Principle of Phenomenal Individuation - conscious states are individuated by their phenomenology > In general, then, there can be no differences, or changes, in consciousness without differences or changes in phenomenology. We can call this the Principle of Phenomenal Difference. It is a corollary to the Phenomenality Principle: if there can be no consciousness without phenomenology, there can be no difference in consciousness without difference in phenomenology. (Chapter 2, p. 61) - If conscious thoughts are states of a distinctive kind, not identifiable with other conscious states, they must have a proprietary phenomenology. Conscious phenomenal differences between thought-types must suffice to individuate them. ## 2.5 Phenomenal Contrast Arguments - Arguments for sui generis cognitive phenomenology based on contrasting experiences, e.g. hearing a language one understands vs. one doesn't, are dialectically vulnerable. - Critics can always attribute the phenomenal contrast to differences in sensory or emotional phenomenology rather than a proprietary cognitive kind. Disputes can devolve into conflicting introspective reports. > Though I think that phenomenal contrast argument cases do in fact show that there is conceptual phenomenology, I also think that in such a controversial context a stronger argument is preferable. (Since Burge's case for anti-individualism rests so fundamentally on the commonsense intuitions he evokes, meeting it on its own terms requires equally natural commonsense intuitions and principles whose application in the situations he describes shows that his conclusions are unjustified.) (Chapter 2, p. 65) - The epistemological and metaphysical arguments aim to provide a more robust case for cognitive phenomenology. ## 2.6 Against Conceptual Phenomenology ### 2.6.1 The Humean Objection (Cognitive Acuity) - Conceptual phenomenology may elude introspection because it is less vivid than other kinds of phenomenology and drowned out by concurrent sensory experience, especially inner speech. - Analogously, human conceptual experience may be primitive compared to other possible minds, like flatworm visual experience compared to ours. We are not great at precise, sustained abstract thought. > Perhaps one day we will have stopped discovering new things, and everything will have been shown to be reducible to, or to superdupervene (Horgan 1993) on, or be grounded in, basic physics as presently understood (or conservatively extended). But we should not assume that this must happen, and we should not let such an assumption determine in advance what can count as really real, or what is a genuinely scientific hypothesis, or what kind of methodology is legitimate. (Introduction, p. 4) - Overcoming the "natural attitude" and attending to experience itself takes effort. Searches for conceptual phenomenology are hampered by its elusiveness, brevity, and the fact that we use thought to look for thought. ### 2.6.2 Carruthers and Veillet - Carruthers and Veillet argue conceptual content cannot be phenomenal because, unlike sensory phenomenology, it is effable rather than ineffable. - The author replies this conflates two senses of content: what an experience/concept is of (its referent) vs how it represents. Sensory and conceptual phenomenology are equally ineffable. - Carruthers and Veillet also argue conscious thought is not phenomenal because, unlike sensory experience, it does not generate an explanatory gap. Inferentially isolated phenomenal concepts are required for gap-supporting scenarios. - But the author contends modificed gap-inducing scenarios are coherently conceivable for thought, showing conceptual properties can produce explanatory gaps. ### 2.6.3 Tye and Wright - Tye and Wright object that, pace Geach, thoughts cannot be part of the stream of consciousness because they do not unfold over time. So they cannot bear phenomenology. - But thoughts, while brief, do take some time - instantaneous thoughts could not be conscious. Thoughts can form temporal streams through repetition or succession. > Tye and Wright seem to think that the fact that all of the constituents of a thought must exist at any moment the thought exists means that thoughts can have no temporal extension: thoughts endure, they say, but do not perdure (2011: 342); "they are wholly present at each moment that they exist" (2011: 342). But being wholly present in this sense has nothing whatever to do with the distinction between endurance and perdurance. (Chapter 2, p. 76) - A deeper issue is their conflation of thinking with judging and believing. The critique may apply to attitudes but not to conscious entertaining of contents. ## 2.7 Internal Worries ### 2.7.1 The Matching Content Challenge - - Bayne and McClelland argue that if intentional content supervenes on phenomenal character, states with the same content should have the same phenomenology. But perceptual and conceptual states seem to represent the same properties while differing in phenomenology. - In response, the author denies that conceptual and perceptual representations can have exactly matching content. Conceptual contents cannot include sensory phenomenology due to the Phenomenal Immiscibility of distinct experiential modalities. > One can no more think colors or smells than one can hear pains or itches. These kinds of experiencing are constituted by the instantiation of radically distinct kinds of phenomenal properties, each with its distinctive experiential modality. There can be no cross-modal experiences. (Chapter 4, p. 140) - Conceptual and perceptual states can co-refer to the same properties while differing in mode of presentation and hence intentional content. ### 2.7.2 Phenomenal Compositionality - If phenomenal character can change depending on context, as in color contrast effects, it seems to threaten compositionality - the content of a complex concept or thought would not be a function of the contents of its parts. - One response is to hold contents fixed but relativize truth conditions - the content of 'The king is in his counting house' is the same whether tokened in the US or France, but has different truth values. - However, the author favors denying that conceptual phenomenology is context-dependent at all, based on introspection. There is no evidence for contrast effects or changing appearances of thought contents. > It is simply left open whether or not a thought like the king is the king or this is this is true, in spite of the fact that both tokens of its constituent concepts have the same contents. Likewise, it is left open whether a thought like the king is not the king or this is not this is false. In both cases sameness of conceptual content does not entail sameness of conceptual referent; hence, analyticity does not entail truth, and antonymy does not entail falsity. (Chapter 5, p. 136) ### 2.7.3 Complexity and Phenomenal Presence - Mendelovici argues that occurrent thoughts often fail to have their complete contents phenomenally present, instead having gisty or schematic phenomenal content. - In response, the author maintains that in cases where one lacks the time or ability to entertain a thought's full content, one entertains a different thought content related to but not identical with the full propositional content, while using language to express or refer to the complete content. > What is thought is that a bearer of 'John' is a philosopher. This is analogous to thinking he is tall with no one in particular in mind. (Chapter 5, p. 168) - Limitations of time and cognitive resources may restrict how much content we can consciously entertain at once, but this is compatible with the thesis that consciously entertained contents are in general phenomenally constituted. # Chapter 3: Externalism ## 3.1 Putnam - Putnam's Twin-Earth thought experiment was meant to challenge the traditional idea that linguistic meaning is derived from psychological content. - Since the twins' words 'water' have different extensions despite their identical psychological states, meanings cannot be determined by psychological content alone but depend also on environment. - An internalist can resist by holding that intension determines extension only relative to context. The twins' word 'water' means the same, but can differ in extension across contexts. > An obvious way for an internalist to resist Putnam's conclusion is to maintain that it is not the principle that meaning is psychologically grounded that should be given up, but, rather, the principle that intension determines extension—at least as it is stated. (Chapter 3, p. 89) - Putnam's arguments do not establish semantic externalism because he equivocates between the claim that natural kind terms are indexical and the claim that difference in extension entails difference in meaning for them. ## 3.2 Burge - Burge's thought experiments aim to show that propositional attitude contents can be socially determined, by eliciting intuitions about literal belief ascription together with a ceteris paribus principle that natural attributions are typically true. > Burge is surely right that in the situations he describes it is natural to attribute beliefs to his subjects using the very words they utter, even though their understanding of those words is at odds with their socially determined meanings. (Chapter 3, p. 91) - However, several common practices suggest these intuitions should not always be taken as guides to literal truth: ### 3.2.1 The Structure of the Argument and the Role of the Naturalness Intuition - The thought experiments' real point is that even when a subject's errors raise suspicion of incomplete understanding, homophonic attribution still seems appropriate. This reveals the domineering role of the literal interpretation principle. ### 3.2.2 Objections - But the naturalness of an attribution does not guarantee its literal truth. Reflection on several types of cases suggests the ceteris paribus clause is often false. #### 3.2.2.1 The Principle of Charity - Attributing contradictory or conceptually incoherent beliefs based on a subject's sincere assertions violates principles of charity and rationality. - Burge fails to show that cases like 'arthritis' do not involve conceptual error of the sort that defeats homophonic attribution, since it seems part of the concept that arthritis is necessarily a disease of the joints. #### 3.2.2.2 Malaprops, Misnomers and Other Miscellaneous Mistakes - Homophonic attribution is natural even in clear cases of linguistic incompetence and communicative error where literal interpretation is obviously inappropriate. > That is, my view is not committed to the regress Levine charges it with (2011: 108–9). One has implicit knowledge of the second-order thoughts whose occurrence constitutes explicit knowledge of the contents of first-order thoughts. (Chapter 2, p. 59) - The mere naturalness of an ascription cannot override strong independent reasons for non-literal interpretation. So it is not a reliable guide to mental content. ### 3.2.3 Anti-Individualism without the Principle and the Intuition? - Burge suggests his argument need not rely on non-dissonant cases and can succeed even granting some errors force reinterpretation, as long as others don't. - But non-dissonant cases cannot establish anti-individualism on their own, since in those cases the match between thought and utterance content can be explained by the subject's choosing the right words for their thoughts. ### 3.2.4 Perceptual Anti-Individualism - Burge's arguments for perceptual anti-individualism neglect to give phenomenal character any role in determining representational content. - Stable causal relations between perceptual states and environmental features are not sufficient for the states to represent the features - the states must attribute properties in a way that (at least largely) matches the instantiated properties. > An experience that in no way resembles an elephant—an experience whose phenomenal character conveys no information about the elephant's distinctive features—cannot, according to Montague, be a perception of the elephant that caused it. Causation is not sufficient for perception. (Chapter 1, p. 30) - Phenomenal character itself, not just external relations, plays a crucial role in determining perceptual content. ### 3.2.5 Externalism without Thought Experiments? - Yli-Vakkuri argues externalism follows simply from three widely accepted principles: 1) Duplicate subjects' corresponding beliefs have the same narrow contents 2) The truth value of a belief matches that of its content 3) Duplicate subjects' corresponding beliefs can differ in truth value - But (3) does not follow from (1) and (2) without a further premise that content determines truth value absolutely, not relative to context. This begs the question against the internalist. ## 3.3 Phenomenal Externalism, the Paraphenomenal Hypothesis - Representationalist theories identifying the phenomenal qualities of experiences with properties of their external objects face a challenge in cases of illusion and hallucination. - Experiences seem to have the same phenomenal character whether veridical or not. But if qualities are external, how can they be present when the objects are absent? - Appeals to uninstantiated universals or merely possible objects to make up the "missing" qualities fail, because such abstracta lack the sensible, appearential properties of concrete particulars. > Intuitively, it may seem unproblematic to speak of non-veridical experience in this way. If you hallucinate a baboon wearing a pink party hat in the living room, it seems perfectly natural to say that your experience represents an object that might have been, but is not, in the living room, and a color that might have been, but is not, locally instantiated. But interpreting this to mean that your experience represents an object that is located in the (or a) living room in some other possible world, or an uninstantiated color, is not consistent with the reductive representationalist's claim that the qualitative features of experience are features of the objects of experience, and not experience itself. (Chapter 3, p. 114) - Disjunctivism fares no better, since denying phenomenology to hallucinations to avoid the problem is highly implausible. Experiences, as appearances, cannot fail to have appearential properties. # Chapter 4: Indexical Thought - The thesis that thought contents are phenomenally individuated is prima facie in tension with standard externalist semantics for indexicals, on which their contents are individuated by contextual referents. - However, analogous examples involving non-indexical concepts show sameness of reference across tokenings is neither necessary nor sufficient for sameness of content: > Suppose I think centenarians are rare in 2000 and then again in 2024. The extension of centenarian in 2000 is not the same as its extension in 2024. But have I thought different thoughts? Has my concept centenarian changed its content? I have a strong intuition that I do not, and it has not. (Chapter 4, p. 121) - Changes in truth value of thoughts across worlds and times also do not force referential individuation of senses. Contents can remain constant while extensions vary with context. ## 4.1 What Is Thought - While allowing distinct "narrow" and "wide" contents may accommodate some externalist intuitions, it is unmotivated on the phenomenal view, where cognitive phenomenology is intrinsic and determines both content and referent. > Obviously, if meanings are abstract objects, then they cannot be located anywhere, including in people's heads. And it is hard to see how a speaker's knowing the meaning of a term could not involve being in some mental state that would count as grasping it. (One cannot know a proposition one has never even entertained.) (Chapter 3, p. 88) - Fregean senses are better construed as abstract but mental experiential types tokened in consciousness, not external referents or truth conditions. ## 4.2 Modality - Indexical contents seem to require actual referents to be constituents in order to deliver the right truth values in counterfactual scenarios. But this is not inevitable. - It makes sense to ask if what a thinker actually thought is true or false of different individuals and in different contexts considered as actual. Specification of counterfactual context, not just referent, matters. > Now suppose that we transplant Penelope's brain, hooked up to yours in the same way as in the previous paragraph, into your skull. Given that Penelope's brain is now housed in your skull, I think there is reason to say that you now have two brains—just as we might say that you have two hearts if someone else's heart were transplanted into your chest alongside your original one. (Chapter 6, p. 194) - While it may be a strong default to hold actual referents fixed, this is not semantically required by the meaning of the indexical concept deployed. Context is always needed. ## 4.3 Understanding - Indexical contents are very thin - knowing the referent is not required for grasping the meaning, even if it would enrich understanding. One can entertain a thought like I really don't like it here without knowing where here refers to. - Frege-cases also show grasp of meaning does not require knowing the referent - one can understand 'Hesperus is Phosphorus' without knowing it expresses a trivial truth. > You will have a much better idea of the nature of the state of affairs that this sentence refers to if you know who he is and where here is. But it does not follow either that you do not understand the utterance, or that if you repeated it yourself (perhaps replacing 'here' with 'there') you would not understand what you said, or, indeed, that you have not succeeded in saying anything. (Chapter 4, p. 131) - The information indexical contents encode is minimal. But that does not mean no content is grasped when referents are not known. ## 4.4 Attitude Reports - In indirect discourse, indexicals in the content clause are typically adjusted to match the reporter's context, suggesting reported contents differ from original. - However, Castañeda and others have argued this is a pragmatic feature of reporting practices, not indicative of a semantic shift in content between the original thought and the report. > The intuitions on both sides are respectable, and that the correct account of the role phenomenology plays in the fixing of perceptual content lies between the extremes of the Pure Causal View and the Matching View. (Chapter 1, p. 29) - Our interest in what others think about often concerns the referents of their thoughts, not necessarily exactly how they represent those referents. ## 4.5 Agreement and Disagreement - If indexical contents are not referentially individuated, it may seem we cannot account for agreement and disagreement - different thinkers' I-thoughts would have the same content despite being about different individuals. - But a thought's truth-conditions can vary between thinkers even if the content (mode of presentation) is shared, because the referents differ. That is sufficient for agreement and disagreement. > Agreement does not entail that thinkers grasp the same content, and disagreement does not entail that thinkers grasp different contents: agreement requires thinking thoughts that attribute the same property to the same individual, and disagreement requires thinking thoughts that attribute incompatible properties to the same individual. And this applies to indexical thoughts as well. (Chapter 4, p. 134) - What matters is co-reference and predication of the same properties, not the specific indexical vehicles used to refer or how they are connected to referents. ## 4.6 Tautology, Contradiction, Analyticity and Antonymy - The necessity of "I am I" thoughts and the impossibility of "I am not I" thoughts seem to support referential individuation, as purely descriptive indexical contents could not deliver these logical properties. - However, these status depend on a thought's tokens co-referring, not on contents containing referents. Co-reference of a tautologous form does not imply the thought content itself is necessary: > So how should the intuition of necessity be accommodated? I submit the following. Let us say that a singular term is logically singular if it cannot, as a matter of the interpretive rules of a language, be assigned to more than one individual... Since names in natural language are not logically singular, using our language with our meanings to describe non-actual possible worlds does not entail that a given name will refer to one and only one thing in any of them. (Chapter 4, p. 135) - Whether indexical tokens co-refer is a further question beyond the content grasped - no semantic guarantee of co-reference exists for non-logically-singular terms. ## 4.7 Phenomenal Demonstratives - Many have argued phenomenal concepts involve samples or demonstrations of the phenomenal properties they refer to, e.g. a concept of red might involve a red visual image. - But phenomenal properties cannot literally be constituents of concepts, as they are fundamentally different kinds of things. Concepts must be thinkable - smells, colors, sounds, etc. are not. - A deeper reason is the Phenomenal Immiscibility of distinct experience modalities - phenomenal properties of one kind cannot mix with or be part of phenomenal states of a different kind. > I thus have the same reasons for rejecting name-experience-containing metalinguistic concepts as I do for rejecting sensory-experience-containing phenomenal concepts. Whatever the proposed hybrid entities might be, they cannot be concepts. (Chapter 5.5, p. 172) - No sensory-involving concepts are acquired when Mary leaves her black-and-white room. She gains acquaintance-knowledge, but not new thinkable contents. # Chapter 5: Thinking with Names - The standard view that proper names have only their referents as meanings, not descriptive senses, is in tension with an experiential view of thought contents. - Frege's puzzle cases of co-referring names with different cognitive significance and informativeness seem to support senses and present problems for Millian views. > According to the description theory criticized by Kripke, a description that gives the meaning of a name must be sufficient for determining the name's referent: it must be uniquely satisfied by the name's bearer. In order to do this, it must be substantive: it must encode enough information about the referent of the name to pick it out, and nothing else. (Chapter 5, p. 144) - However, purely referential semantic contents cannot constitute what is grasped in thinking, on the phenomenal intent ## 5.1 Metalinguistic Descriptivism - The author defends a metalinguistic description theory of proper names, on which the meaning of a name N is given by the description 'a bearer of "N"', not just the referent. - This avoids Fregean puzzles, since co-referring names can differ in metalinguistic descriptive content. Empty names can also be meaningful on this view. > On the version of the metalinguistic description theory I develop here, the meaning of a name N (and the content of the nominal concept it expresses) is given by the indefinite description 'a bearer of "N".' (Chapter 5, p. 145) - Using indefinite rather than definite descriptions better captures the actual meaning of names - most names have multiple bearers and are used to refer to a single individual at a time. ### 5.1.1 Kripke's Objections The author argues the metalinguistic view is not vulnerable to Kripke's main objections to descriptivism: #### 5.1.1.1 Modality - Kripke argued names cannot mean descriptions because 'N is D' would then be analytic and necessary, whereas no such statements are necessary. But this assumes a flawed notion of analyticity. - The author follows Katz in holding analyticity is a structural relation between concepts, not sufficient for necessary truth. Analytic statements can be contingent or even false if terms fail to refer. > Kant has been interpreted as offering two distinct characterizations of analyticity. On the first, a judgment is analytic if its predicate concept is (covertly) contained in its subject concept...On the second, a judgment is analytic if it is necessarily true, because its denial is a contradiction. These characterizations are obviously not equivalent. The former entails the latter, but not vice versa. (Chapter 5, p. 151) - 'Aristotle is a bearer of "Aristotle"' is not necessarily true even if analytic - its terms might not co-refer in a possible world where Aristotle is not so-named. #### 5.1.1.2 Kneale - Kripke charged that just as we cannot conclude 'horses' means things called 'horses' from the triviality of 'horses are called "horses"', we cannot conclude the meaning of 'N' is metalinguistic from the triviality of 'N is called "N"'. - But triviality is not the only reason for adopting metalinguistic descriptivism - there are independent semantic and metaphysical grounds. Truth-security of 'N is called "N"' does not by itself entail N's meaning is metalinguistic. #### 5.1.1.3 Non-Circularity - Kripke charged that metalinguistic descriptivism fails a non-circularity condition, since saying the referent of 'N' is the thing 'N' refers to is vacuous and cannot determine reference. > Kripke's criticism therefore misses its mark. Moreover, the metalinguistic theorist is not constrained to endorse the circular answer "He refers to the man to whom he refers." One may answer the question "To whom does he refer by 'Socrates'?" by saying "He refers to a man who is a bearer of 'Socrates'." (Chapter 5, p. 154) - But the view is not a theory of reference determination - it does not aim to identify a name's referent from its meaning alone. Meanings are only partial constraints on reference. #### 5.1.1.4 Ignorance and Error - If 'N' meant 'the F', an 'N' user discovering the referent is not F should withdraw the name. But Kripke claimed this does not happen for proper names as it would for descriptions. - However, in true cases of reference shift via an error in naming or reference-fixing, we would withdraw the name if the mistake were discovered, as the descriptive theory predicts. ### 5.1.2 Back to Frege? - The notion of analyticity without necessary truth might seem to allow a return to Fregean descriptivism - e.g. taking 'Aristotle' to mean 'the teacher of Alexander' without making 'Aristotle taught Alexander' necessary. - But the metalinguistic view denies this, since analytic truths involving proper names are still truth-secured - necessarily true if they refer at all. 'The teacher of Alexander' fails this test. > What it says is that if a description d gives the meaning of a name N, then the sentence 'N is d' is truth-secured. But it is not the case that, necessarily, if 'Aristotle' refers, then (21) ['Aristotle was the teacher of Alexander'] is true. Hence, 'the teacher of Alexander the Great' cannot give the meaning of 'Aristotle'. (Chapter 5, p. 156) - No substantive definite description can give the meaning of a proper name on the metalinguistic view, only the minimal 'bearer of "N"' descriptions. ## 5.2 Direct Reference, Rigidity and Necessity - The author rejects the ideas that proper names are rigid designators, that identity statements flanking names are necessary if true, and that names are directly referential. - Multiple bearerhood and the non-descriptionality of names block rigid designation. Identities are not semantically guaranteed to hold across all uses/worlds even when true. > If names have metalinguistic descriptive content, however, we should no more think that, for example, 'Ringo Starr is Richard Starkey' is necessarily true because 'Ringo Starr' and 'Richard Starkey' refer to the same individual than we should think that 'the first woman in space was the pilot of Vostok 6' is necessarily true because 'the first woman in space' and 'the pilot of Vostok 6' both refer to Valentina Tereshkova. (Chapter 5, p. 159) - Directly referential names would be Kripkean rigid designators. But names are semantically associated with reference-fixing descriptions, not just referents. ## 5.3 Kripke's Puzzles The metalinguistic view solves Kripke's famous puzzles about names and belief ascription: - Pierre has contradictory beliefs because 'Londres' and 'London' differ in metalinguistic content (bearer of 'Londres' vs bearer of 'London'), allowing rational inconsistency. - Peter can believe Paderewski had musical talent and lacked musical talent because the indefinite descriptions do not require a single referent. Apparent singular terms can be used generally. > Finally, the 'Paderewski' case does not lead to paradox because the propositions a bearer of 'Paderewski' had musical talent and a bearer of 'Paderewski' had no musical talent are not contradictory (cf. 'A book I read was difficult' and 'A book I read was not difficult'). (Chapter 5, p. 167) - The puzzles are predicated only on undefended Millian assumptions about names' semantic contents that the metalinguistic theory rejects. ## 5.4 Determining Nominal Reference - The metalinguistic contents associated with names do not suffice to fix a unique referent, due to multiple bearers. Users must rely on extra-linguistic contents to pick out a specific individual. - Names function analogously to indexicals in this respect - character does not determine content, so speaker intentions and contextual salience must fill the gap. > What is thought is that a bearer of 'John' is a philosopher. This is analogous to thinking he is tall with no one in particular in mind. The thought John is a philosopher is true just in case a bearer of 'John' is a philosopher. But if the user of a name is to refer to some particular bearer of 'John', they must have some descriptive access to it, even if this is indirect. (Chapter 5, p. 168) - Reference is constrained but not fully determined by names' linguistic meaning. Perception, memory, and descriptive identification all play a role. ## 5.5 The Paratactic Theory of Nominal Concepts - A Fregean objection: names in metalinguistic descriptions ('bearer of "N"') do not refer to senses, so such descriptions cannot give senses or be part of thoughts. - The phenomenal intentionality theory faces an analogous problem - if nominal concepts are individuated by cognitive phenomenology, they cannot contain non-phenomenal name-experiences. > There are no concepts the grasp of which requires experience of the phenomenal properties they are of, and no special conceptual knowledge of what it is like—no phenomenal thought. (More precisely, there can be no non-conceptual-phenomenal thought. On the view defended in this book concepts and thoughts are themselves phenomenally constituted, so of course there are phenomenal concepts and thoughts in this sense.) (Chapter 4, p. 138) - The author's solution: names are not constituents of metalinguistic concepts, but are displayed alongside them, serving as referents of tokens of the concept a bearer of this name. Nominal concepts refer to names, not their referents. > The concept associated with a name N is the concept a bearer of this name, and the concept comes to be associated with the name by being demonstratively applied to a displayed token of it: a bearer of this name: N. (This why I call this account the "paratactic" theory of nominal concepts.) (Chapter 5, p. 171) - So nominal concepts have metalinguistic senses (involving names) without names being phenomenal parts of the concepts. # Chapter 6: Unconscious Thought - The phenomenal intentionality of thought thesis prima facie implies that if there are unconscious intentional states, there must be unconscious phenomenology. But this is controversial. - If there is no unconscious phenomenology, a phenomenal intentionalist must either accept eliminativism about unconscious intentional states or give up the phenomenal intentionality of thought thesis. > Still, introspection is, I would argue, where we begin the study of the mind. Indeed, we only know there is such a thing from the first-person perspective. Analytic Phenomenology insists that empirical study of the brain and central nervous system is neither the starting point nor the sine qua non of the study of experience. (Introduction, p. 7) - Before considering those options, the author examines compromise positions that grant unconscious states a derivative or second-class intentionality. ## 6.1 Searle - Searle maintains intentionality requires "aspectual shape", an experiential notion. So unconscious intentional states must at least potentially be conscious. - But the author charges Searle with inconsistency, since he also says unconscious states are entirely neurophysiological. Neural properties cannot constitute aspectual shape. > Aspectual shape properties, therefore, would appear to be the likely candidates for those properties in virtue of which an unconscious intentional state is intentional. For the last two claims of the preceding paragraph together seem to entail that a state need not be conscious in order to have aspectual shape. Yet, Searle also says that "The ontology of mental states, at the time they are unconscious, can only consist in the existence of purely neurophysiological phenomena" (1991: 53). (Chapter 6, p. 174–5) - Searle equivocates between two notions of unconscious aspectual shape - one distinct from consciousness and one a mere potential for consciousness. Neither delivers a stable view. ## 6.2 Strawson - Strawson claims true intentionality is only found in experiential beings and episodes. But he allows unconscious thoughts in conscious creatures to "borrow" intentionality somehow. - The mere fact that an unconscious state occurs in a conscious creature cannot explain its status, since it seems intrinsically identical to unconscious states in non-conscious beings. - Strawson later suggests the intentionality is inherited from a prior conscious thought through a causal process. But this cannot constitute real intentionality without imparting intrinsic phenomenal properties. ## 6.3 Horgan and Graham - Horgan and Graham consider three views: 1) Unconscious states are not genuinely intentional, having only as-if content 2) Unconscious states can only be intentional if potentially conscious (Searle's view) 3) Unconscious states inherit content through integration with conscious states - The author objects that as-if content is not real content. Nor can merely being caused by or disposed to cause conscious intentional states constitute genuine intentionality. ## 6.4 Kriegel - Kriegel identifies original intentionality with experiential intentionality. But he allows unconscious states to count as intentional if an ideal interpreter would attribute content to them to explain behavior. - The author objects that interpretability is not genuine intentionality - a state's intrinsic properties are not changed by being interpreted as contentful. > Interpretation does not change the intrinsic nature of the interpreted state. So, assuming, as Kriegel does, that intentionality is phenomenally constituted, and that phenomenality and consciousness are intrinsic properties, in spite of the fact that it is treated as if it had intentional content, it does not really have it. (Chapter 6, p. 180) - No compromise view fully reconciles unconscious intentionality with phenomenal intentionality. As-if content is not the real thing. ## 6.5 Smithies - Smithies distinguishes "access consciousness" from "phenomenal consciousness." Unconscious thoughts can be access conscious and so assessable for rationality. - He maintains unconscious thoughts "inherit" content from the conscious thoughts they dispose the thinker to have. The contents are "accessible" because they are possible contents of consciousness. - But the author sees this as another failed compromise. Mere accessibility or counterfactual relations to consciousness cannot constitute phenomenal, so non-experiential, properties. > If genuine cognitive content is phenomenally constituted, but there is no unconscious phenomenology, then no unconscious state can literally have genuine cognitive content. Relational phenomenal individuation does not suffice for phenomenal constitution, and so cannot confer phenomenally constituted content. (Chapter 6, p. 181) - All attempts fail to get intentionality on the cheap. Non-phenomenal properties cannot do the work of the phenomenal, whether attributed dispositionally or interpretively. ## 6.6 Unconscious Phenomenology? With compromises rejected, the author considers two non-eliminativist approaches: ### 6.6.1 Intrinsic Unconscious Phenomenology - Unconscious states could have phenomenal properties without being conscious, if phenomenality and consciousness are distinct properties not necessarily co-instantiated. - Some assume phenomenality entails consciousness conceptually. But phenomenality could be seen as a determinable of which consciousness is a determinate, or consciousness itself as a species of phenomenality. #### 6.6.1.1 Blindsight - Blindsight patients have no conscious experience in a region of their visual field, yet can make visual discriminations there, suggesting unconscious qualitative states. - Normal perceivers need color phenomenology to distinguish red from black. If blindsighters can make chromatic discriminations, that suggests unconscious qualia are involved. > If blindsighters could make distinctions on the basis of color that the achromat cannot, I think we would have a stronger (though not, of course, conclusive) reason for thinking that the phenomenon of blindsight, or at least some cases of it, do show that there can be unconscious phenomenology. (Chapter 6, p. 185) - But an opponent can always appeal to non-phenomenal information to explain these discriminations, so the argument is suggestive but not decisive. #### 6.6.1.2 Phenomenal Sorites - Gradual phenomenal continua (e.g. color samples each indistinguishable from its immediate neighbors yet globally distinguishable) seem to require phenomenal differences that are not consciously detectible. - If sample X is indistinguishable from sample Y, and Y from Z, but X is distinguishable from Z, then Y must not have the same color phenomenology as both X and Z. One of those appearances is illusory. > Now suppose we have a series of internal samples—experiences e, f and g—for which this phenomenon arises...When speaking of e, f and g, I will use 'color' to refer to their subjectively experienced intrinsic chromatic properties...If the color of f were the same as the color of e, then f ought to be colorwise distinguishable from g, since e is. But, by hypothesis, f is not distinguishable from g. Hence, the phenomenal characters of e and f must differ. But their difference is consciously undetectable: as far as consciousness goes, the colors of e and f are the same. But their colors are not the same; hence, there are (subjective) colors that in certain circumstances elude conscious perception. (Chapter 6, pp. 187–8) - Subjective indistinguishability in spite of subjective difference requires an appearance/reality distinction for qualia, and so unconscious phenomenology. ### 6.6.2 Unconscious Consciousness - Another view is that all phenomenal states are conscious, but some are conscious in themselves without being conscious for the subject - creature consciousness without state consciousness. - This could happen if one part of a split brain generates experiences that affect behavior without being accessible to the subject's main center of consciousness. The inaccessible experiences would be conscious, just not consciously experienced by the subject. > If consciousness does not require a self, a view of the sort sketched here would not have to populate our heads with any others, but merely with ownerless intrinsically conscious states. If this is possible, then what we call our unconscious states, because we are not conscious of them, may nonetheless be intrinsically conscious, and, hence, phenomenal; and they may be states of the individual in whose brain they arise in the functional sense of figuring in the etiology of the individual's behavior, and, hence, in rationalizing explanations of it. (Chapter 6, p. 195) - So phenomenal intentionality could be preserved without unconscious qualia, by granting phenomenality to in-themselves-conscious states underlying some behavior, even when not conscious for the behaving subject. ## 6.7 Unconscious Contentlessness - The other non-eliminativist option is to deny that the unconscious states posited in psychological explanations are genuinely intentional or contentful at all. - Unconscious brain states may be contentless but systematically related to contentful conscious states, as unconscious encoding states are related to the conscious experiences of the images they encode. > Consider a computer program that yields, e.g., encodings of consequents of conditionals given encodings of the conditionals and their antecedents. Pace Turing, we are not constrained to treat the processes themselves as logical inferences, even if they produce logically entailed outputs. If they are non-conscious, and there is no unconscious content (because there is no unconscious experience), then such processes are not inferences, since inferences are transitions between contentful things (states, sentences, propositions,…). (Chapter 6, p. 197) - Computational processes can mirror rational inference without being themselves inferential or intentional. ### 6.7.1 The Causal Role of Content - How can content be explanatory if content-bearing states result from content-free processes? By being a structuring cause - conscious intentional states program the computational processes that generate them. - Evolution and learning imbue unconscious processes with a structure that enables them to respect logical relations between contents of the conscious states they subserve, without them needing to bear contents themselves. - But this cannot explain all fine-grained intentionality - the conscious mind must be able to program the unconscious to make and manipulate distinctions beyond the latter's native powers. > Evolutionarily speaking, conscious creatures capable of finer-grained distinctions in content could evolve unconscious states encoding for such fine-grained conscious states, as well as unconscious processes that respect the content relations of the conscious states. But the pressure comes from above. The fine-grained distinctions available to consciousness are not available to unconscious Nature nor, therefore, to the unconscious brain. The unconscious brain must be molded by the conscious mind to subserve its fine-grained representational system. (Chapter 6, p. 200) - Consciousness can thus shape the unconscious to respect its intentional structure without imbuing it with genuine content of its own. ## 6.8 Concluding Remarks - Three responses to the paradox of unconscious intentionality for the phenomenal intentionalist remain viable: 1) Ascribing intrinsic unconscious phenomenology to content-bearing unconscious states 2) Ascribing in-itself-but-not-for-me consciousness to some unconscious content-bearing states 3) Denying genuine intentional content to unconscious states while granting them content-respecting causal roles - While the third option is perhaps the most promising, elements of all three may have a role to play in a complete picture of the mind. > I have suggested three ways to respond to the unconscious thought challenge to the phenomenal intentionality of thought thesis. The first argues for unconscious states with intrinsic phenomenal properties; the second denies that there can be such states, and postulates intrinsically conscious states not conscious for their possessor. The third denies that there are unconscious thoughts, but shows how unconscious states lacking content could encode conscious content and be processed in ways that respect fine-grained content relations. (Chapter 6, p. 201) - The question remains difficult and the ultimate truth likely complex. But the phenomenal intentionality thesis is not defeated by the prima facie problem of unconscious intentionality. # Chapter 7: Conceptual Reference - Phenomenal intentionality is fundamentally an "intensionalist" rather than extensionalist theory of content. It explains representational content in experiential terms rather than referential or truth-conditional ones. - But critics may charge that this fails to account for a key aspect of intentionality - aboutness, the relation between thoughts and the worldly objects and properties they concern. > The focus of this book has been what we can call "intensional intentionality"—i.e., the kind of meaningfulness or contentfulness that a concept that lacks an extension (e.g., phlogiston, the largest natural number) can have. (Chapter 7, p. 203) - A complete theory of intentionality should explain both content and reference. Otherwise phenomenal intentionality may seem to tackle only "narrow" not "wide" content. ## 7.1 Extensions of Complex Concepts - Perceptual experience enables demonstrative reference and singular thoughts by presenting particular objects to serve as referents of perceptual and conceptual representations. - For non-present referents, descriptions and imaginative representations can substitute for perceptual acquaintance in providing a referential context for thought. > The things we perceive are the objects of our perceptual experience (what we perceive), analogously to the way in which the things we conceive of are the objects of our thought (what we think about). For something to be a perceptual object for us is for it to be singled out and presented as distinct from yet present to the subject—as out there, in the mind-independent world, yet at the same time spatially related to us in specific ways, and capable of being acted upon and acting upon us. (Chapter 7, p. 205) - The extension of a complex concept is a function of the extensions of its parts and their mode of combination. So determining extensional reference reduces to determining simple conceptual reference. ## 7.2 Extensions of Simple Concepts - Causal theories identify content properties with the objective properties whose instances cause a concept's tokenings. So they yield co-extensive concepts and contents automatically. > Causal-informational approaches have the virtue of establishing general conceptual content and extension simultaneously: the properties instances of which reliably cause tokenings of conceptual representations (in the proper circumstances) are their contents, and the extensions of those properties are the extensions of the concepts. (Chapter 7, p. 205) - But on the phenomenal theory, concepts cannot denote the same properties that constitute their contents, on pain of equating instantiating a concept with instantiating its referent. - Instead, concepts must denote properties by describing them, as Fregean senses do. The key question is what this description relation amounts to. > On the phenomenalist view, concepts are contents—conceptual type-phenomenologies; and concept tokens are tokens of those phenomenal types (contents), not content-expressing syntactically individuated representational "vehicles." There are no mental analogs of Frege's concept-words. However, like Fregean senses, conceptual experience types denote properties, and thereby inherit their extensions. (Chapter 7, p. 208) - Conceptual phenomenology is intrinsically and primitively intentional - it is its very nature to describe and point beyond itself, just as it is in the nature of sensations to feel a certain way. > And the same is true for conceptual phenomenology. It is intrinsically, primitively, inexplicably, descriptively intentional. It is of its nature to be about things—to point beyond itself; to describe. To ask how it is that conceptual phenomenology can describe is like asking how it is that pain can hurt. These are basic facts. (Chapter 7, p. 209) - No further explanation of the intentional directedness of thought is needed or possible. It is bedrock in the conception of the mental. ## 7.3 Harmony and Disharmony of Thought and World - The phenomenal model does not guarantee an automatic alignment between the properties we conceptually represent and the actual properties that typically cause our conceptual representations. > If we happen to get it right without further ado, it is a matter of luck. (Chapter 7, p. 210) - Our thought, while intentionally directed at reality, often misconceives it. Bringing concepts into alignment with the world's actual structure takes effort and is always uncertain. - But this epistemic predicament is not a special problem for the phenomenal theory - it is our general lot as finite inquirers. No account of intentionality can or should eliminate it. > And surely any chip in the series would have the color it looks to have at least in some circumstances—e.g., if it were the only thing being perceived. (Chapter 6, p. 187) - In conclusion, the phenomenal intentionality theory can account for the referential directedness of thought not by reducing it to causal relations, but by taking it as a basic, irreducible feature of conceptual experience as such. The mind's intentional grip on the world it represents is fallible but real.