Mary and Propositional Knowledge

Abstract

I introduce what I call the Propositional Stance - the consensual claim that Frank Jackson’s Mary acquires new propositional knowledge - and discuss two prominent arguments that have been put forth in its favor. The first argument concerns the semantics of what-it-is-like sentences. The second argument concerns the epistemology of phenomenal belief. I argue that both arguments are unconvincing or lead to undesirable consequences, and then motivate non-propositional knowledge approaches to the puzzle of The Knowledge Argument.

Keywords:
Knowledge argument; Varieties of knowledge; Knowing what it is like; Phenomenal knowledge

Introduction

Frank Jackson’s Mary goes out, sees colors for the first time and gains new knowledge, or so it seems. But what does she come to know? Jackson originally held that she comes to know new non-physical information (1982) or non-physical facts (1986), thus providing what is known as the Knowledge Argument against physicalism (sometimes abbreviated as ‘KA’ in what follows). But not only is this conclusion highly undesirable for current philosophical and scientific customs, but it is also not at all an obvious consequence of Jackson’s setting.

Amid discussions of Nagel’s (1974) bat problem, Laurence Nemirow (1980, p. 475) proposed that one understands a given experience (e.g. what it is like to be a bat, or what it is like to see red) not by knowing some set of facts about such experience, but by having the ability to recreate it in imagination. Nemirow bases that claim on the more general thesis that “some modes of understanding consist, not in the grasping of facts, but in the acquisition of abilities” (op. cit., same page), a thesis famously attributed to Gilbert Ryle (1949). Later on, David Lewis (1983, 1988) picked up Nemirow’s ideas and applied them to the context of Jackson’s Knowledge Argument, arguing that Mary’s new knowledge consists solely in new abilities or new know-how, and that she does not come to know new facts. Similarly, Earl Conee (1994) argued later that Mary’s new knowledge is to be explained solely through the notion of acquaintance (usually credited in philosophical discussions to Bertrand Russell), a notion that captures yet another mode of understanding that is independent of knowing facts.

The specifics of the above-mentioned proposals won’t be central for the present work. Here I’m interested in what they share as a general approach to the Knowledge Argument. What these proposals have in common is that they appeal to non-propositional knowledge to explain Mary’s situation. If that general approach is correct, it accommodates Mary’s new knowledge while also successfully blocking Jackson’s conclusion. This is so because Jackson’s argument, as it will become clear in what follows, depends on Mary acquiring new propositional knowledge.

Despite the advantage just mentioned, non-propositional knowledge approaches remain fairly marginal within the KA debate. As put by Nida-Rümelin & O’Conaill (2024, section 4.6), “many philosophers find it hard to deny that Mary gains new factual knowledge after release”.¹ Examples to illustrate that are not hard to find. David Chalmers (1996, p. 129) claims that “no doubt Mary gains some abilities when she first experiences red, (...) but it certainly seems that she learns something else: some facts about the nature of experience”. Van Gullick (2004, pp. 375-6) writes that the view “according to which Mary comes to some new propositional or factual knowledge about the nature of red experience and what it is like seems too compelling”. Michael Tye (2009, p. 132) argues that Mary “discovers (learns) what it is like to experience red. This (...) involves a mixture of factual and objectual knowledge”.

Those who posit new propositional knowledge when analyzing Mary’s epistemic situation adopt the position which I shall call the Propositional Stance. Although commonly taken, the position lacks substantial dedicated defense. A prominent exception is William Lycan’s A limited defense of phenomenal information (1995), in which Lycan’s main goal is exactly to argue in favor of (what I call) the Propositional Stance. Although the present work can be seem as an attempt to advance Lycan’s more general project of finding the most compelling lines of argument for and against the Propositional Stance, his main goal in the mentioned work was to argue for the Propositional Stance, while my main goal here is to argue against it.²

The discussion is structured as follows. The first three sections will prepare the ground for assessing the central arguments. In section 1, I offer a definition of what I call the Propositional Stance and of a close and more general idea, the Hypothesis of Phenomenal Information. In section 2, I finesse the thought experiment by presenting Nida-Rümelin’s variant, and then a variant of my own, which should help us grasping more firmly the relevant claim of the Propositional Stance. In section 3, I offer a clarification and some constraints on the notion of propositional knowledge. In section 4, I discuss two arguments in favor of the Propositional Stance. The first argument will be called the Linguistics Argument, which is construed around the semantics of knowledge-wh. The second argument will be called the New Concept Argument, which is an argument that appeals to propositional knowledge that uses phenomenal concepts as its components. I reason against both arguments. If my arguments are sound, proposals that posit new propositional knowledge while analyzing Mary’s epistemic situation lose considerable strength in the context of the Knowledge Argument debate.

1. The Propositional Stance

Before advancing, we can start with a working definition of the Propositional Stance and some associated ideas. The central thesis can be captured as thus:

Propositional Stance (PS): the position according to which Mary acquires new propositional knowledge when she experiences colors for the first time.

At first, PS is only a local claim about the epistemology of the Knowledge Argument. It is very plausible to suppose, however, that PS depends on a more general thesis about the relation between (types of) experiences and propositional knowledge, a thesis which is very close to what Lewis (1988) dubbed as the Hypothesis of Phenomenal Information. Adopting the label, we can capture the idea in the following terms:

Hypothesis of Phenomenal Information (HPI): There is propositional knowledge that can only be acquired through experiential means.

HPI is not difficult to unpack: it says that there are some pieces of propositional knowledge that, to be acquired, depend on being accessed through experiences, i.e. through the first-person conscious perspective.³ HPI provides a good way to ground the Propositional Stance: Mary can only acquire certain propositional knowledge after experiencing colors, and not before that, because that knowledge depends on experiential means that were not available to her from inside the room.⁴ This leads us to another important definition:

Experience Condition: The condition that some pieces of propositional knowledge must meet in order to be acquired.

If HPI is correct and some pieces of propositional knowledge can only be acquired through experiential means, then those pieces of propositional knowledge obey some condition, which we may call the Experience Condition. The Experience Condition explains well the position of the PS proponent: only when Mary experiences colors for the first time can she acquire some pieces of propositional knowledge, because before that she did not meet the proper experiential condition.

With those definitions under our sleeve, we are ready to progress.

2. Finessing The Thought Experiment

To properly understand what it is at stake for proponents of the Propositional Stance, we can give some finesse to the thought experiment. We can start following a distinction due to Nida-Rümelin (1995, 1998). It is common to express Mary’s new knowledge in terms of the locution ‘what-it-is-like’. For this, we can say that Mary obtains new what-it-is-like-knowledge. But, according to Nida-Rümelin, this way of putting things can lead to failure to distinguish two different epistemic steps taken by Mary at once. To properly separate those two steps, Nida-Rümelin asks us to imagine Marianna, a character that, at t ₁, is in the exact same initial epistemic scenario as Mary. Instead of, like Mary, going out from t ₁ directly to the open world, Marianna first, at t ₂, goes to a room full of manufactured objects, such as chairs, walls, tables, etc., that display arbitrary colors (e.g. some chairs are blue, but others are red, and so forth for the others). Nida-Rümelin claims that, in such situation, Marianna would be unable to relate those new experiences of colors at t ₂ to the correct color names and what she knew about them at t ₁. She may wonder which of those colors she sees is red or blue, but she wouldn’t be able to know without further information. Then, at t ₃, Marianna goes to the open world and sees the sky, red apples, grass and other objects that have characteristic colors, and, given her extensive knowledge at t ₁, she becomes able to tell which colors are which. At t ₃ she may say, while pointing to a ripe tomato, “so this is red!”.

Nida-Rümelin’s distinction is helpful in one regard, but it can also be quite misleading in another. The distinction is helpful insofar as t ₂ isolates the new relevant experiences from purported new true thoughts. Some approaches, especially those who appeal to non-propositional knowledge, analyze exactly t ₂ as the epistemic step where Mary learns what it is like to see colors, what makes t ₂ a useful tool to have. The distinction can also be very misleading, though, for Nida-Rümelin’s room cheats on the original setup of KA as it deprives Marianna of physical information. Marianna should have descriptive/physical knowledge of, for example, the chemical composition and the reflectance properties of the objects in the room.⁵ Another way to reach the same end would be to know the coordinates of each object and the color of each one, something that Mary/Marianna could know at t ₁. Depriving Marianna from such knowledge at t ₂ can only function as an arbitrary tool of analysis. Maintaining the plausibility of a scenario where those characters know all physical facts and still can’t sort blatant physical properties of objects can do only harm for our comprehension of the main issue posed by the argument.

To avoid the issues mentioned and to focus on what is really central for KA, I propose the following variant of the thought experiment, which I’ll call Chromophobe Mary. From her room, Mary develops chromophobia. She has been studying colors for so long and with such passion that she grows scared of actually experiencing them. Since she knows that she is being released from her room soon, something must be done. Being a super scientist from a completely developed science, she proceeds without much effort to develop lenses that make her vision monochromatic, but that emulate the fine-grained level of informational processing of standard human eyes. The lenses then display that information to Mary through text in real time. When freed, she can finally interact with all sorts of things that were not around in her room. She sees a dog for the first time and says this is a dog. She can point to the sky and say this is the sky. She can also, with the help of her lenses, point to any object and truthfully say this is red, this is turquoise, this is green₃₉ and so forth, even if she still sees it all in black and white. After many years, she is habituated to the outside world and is finally prepared to take off her lenses. She takes a deep breath, stares at a red wall and, with a final help with her lenses, says this is red. She then takes off the lenses and finally experiences colors. What does she learn?

The Chromophobe Mary variant contributes to Jackson’s thought experiment by giving to Mary all contextual knowledge (or indexical knowledge, if you may) available before her relevant experiences. The acquisition of such contextual knowledge is arguably not central for evaluation of what Jackson’s Mary learns. This finesses out some confusions and help us focus on the crucial element for KA: the new experiences that Chromophobe Mary will have and the immediate connection of the phenomenal qualities of those experiences to what she knew before, plus any knowledge that might come from that. It is on the epistemic step of Chromophobe Mary taking off her lenses, and not on Nida-Rümelin’s Marianna going from t ₂ to t ₃, that the proponent of the Propositional Stance needs to find some new propositional knowledge.

3. Propositional Knowledge

Before moving on to actual arguments around the Propositional Stance, we can use a final clarification. PS relies on the notion of propositional knowledge, so we need a good grasp of that notion. One way to make the relevant distinctions is to talk about objects of knowledge. Knowledge is a relation between a subject (the knower) and its object of knowledge (what is known). In the case of propositional knowledge - usually construed in the philosophical discourse as a member of the broader class of propositional attitudes -, the object of knowledge is a proposition (cf. Perry 1994; Zagzebski 1999). Grasping propositions is usually thought as depending on sophisticated intellectual activity, such as abstract reasoning and language usage. Propositions are also commonly linguistically marked as the designata of that clauses. So ‘S knows that P’ is the most basic example of a relation of propositional knowledge. An immediate effect of this characterization is that propositional knowledge relies on truth. Propositions are usually accepted to be true or false (i.e. to have truth-value), but in the case of propositional knowledge the subject needs to be related to a true proposition: if S knows that P, it has to be the case that P. Conversely, if P is not the case (if P is false), then it seems that S cannot legitimately know that P.

If the above is right, propositional knowledge is an intellectual activity that involves grasping true propositions.⁶

For our current purposes, the main upshot of the discussion above is that the objects of propositional knowledge are true propositions. This contrasts with the objects of other purported types of knowledge. Practical knowledge, for instance, is sometimes said to relate the subject to an ability, while objectual/acquaintance knowledge is said to relate the subject to a thing or an experience.⁷

It is, then, a constraint of the above discussion that the proponent of PS is committed to the view that Mary acquires new propositional knowledge in virtue of entertaining an intellectual relation that constitutes knowledge and that has a true proposition as a component.

4. Two Arguments for The Propositional Stance

We are now in a better position to move to our main topic. In what follows, I will spell out what I consider to be the two most prominent lines of reasoning in favor of the Propositional Stance, and then will argue against them. To be thoroughly evaluated, those arguments deserve dedicated work and more detail than I will give them below, but here I’m more interested in assessing the Propositional Stance as a whole. As said at the outset, this attempt shares much with Lycan’s 1995 work. There Lycan discusses nine arguments in support of the Hypothesis of Phenomenal Information. Some of the arguments presented by Lycan are a bit rushed, and some of them merge with one another as different lines of reasoning to state the same point. Here I’m trying rather to provide a smaller number of arguments that are strong enough to survive initial scrutiny. The two presented arguments draw from various literatures, but, in a broader characterization, the first argument can be seem as an argument from linguistics, and the second one from philosophy of mind.

4.1. First Argument: The Linguistics Argument

The first argument we’ll look into builds on the claim that what Mary learns is better expressed through the locution ‘what it is like’ (in what follows, sometimes abbreviated as ‘WIL’). The argument is motivated by the observation that the semantics of what-it-is-like sentences fit the general semantic treatment of other knowledge-wh ascriptions,⁸ which ascribes knowledge of an answer to an embedded question. Knowing answers to questions can satisfy without much difficulty some (and maybe all) the criteria of propositional knowledge discussed in section 3. So we have the makings of an argument for the Propositional Stance. The rationale of the argument can go as this:

• P1) The relevant knowledge that Mary obtains is what-it-is-like knowledge.

• P2) What-it-is-like knowledge ascriptions are to be semantically treated through the semantics of knowledge-wh.

• P3) The semantics of knowledge-wh poses propositional knowledge to the subjects of knowledge-wh ascriptions.

• (C) The semantics of knowledge-wh poses propositional knowledge for Mary’s new knowledge.

Let’s call the above argument the Linguistics Argument. This argument goes back at least to Lycan (1995, pp. 244-5), who argues that on the “indirect-question (...) model, ‘S knows what it’s like to see blue’ means roughly ‘S knows that it is like Q to see blue”. The Linguistics Argument recently received great attention in the literature and figured in discussions of Stanley & Williamson (2001), Tye (2009; 2011), Stoljar (2016; 2017),

Habgood-Coote (2018), Tye & Grzankowski (2019), Cath (2019), Lynch (2020) and others. For illustration: Tye (2011, p. 304) writes that “if the standard semantics is correct, knowing what it is like to see red is knowing an acceptable answer to the question ‘What is it like to see red?’” Similarly, Stoljar (2017, p. 5) says that, according to similar rationale to the Linguistics Argument, the sentence ‘Mary knows what it is like to see red’ is “true just in case Mary knows some fact that answers the question ‘what is it like to see something red?’”.

Before advancing, a quick caveat. It is important to notice that it does not follow from (C) that Mary’s new knowledge necessarily involves propositional knowledge. (C), if correct, strengthens the Propositional Stance, that is for sure, but linguistics should offer only but an evidence for epistemological discussions, not decide those discussions altogether. This point should go without saying, but some part of the recent literature (Stanley & Williamson, op. cit., being an influential locus) precisely extract strong epistemological conclusions out of linguistic discussions.

Whatever the weight someone gives to considerations from linguistics to epistemology, however, the Linguistics Argument is not a safe tool for the proponent of PS, for the argument is dubious in its own domain. As pointed out by Stoljar (2016, 2017), there is a well known distinction in the linguistics literature between interrogative and free-relative readings of wh-sentences. While the interrogative reading relates the subject to a proposition embedded in the wh-clause, the free-relative reading relates the subject to an object. Consider the following sentences:

• (1) I wonder what the square of fifty-nine is.

• (2) I eat what is cooked.

(1) has a clear interrogative reading: what the subject wonders about is the proposition that is a correct answer to the indirect question ‘what is the square of fifty nine?’. On the other hand, (2) has a clear free-relative reading: what the subject eats is the object that is cooked, not a proposition. A natural explanation for the phenomenon at hand is that ‘wonder’ allows indirect questions as complements, whereas ‘eat’ does not.⁹ Now consider the following sentence with the verb ‘know’, crucial to epistemology and to the Knowledge Argument debate:

• (3) Alice knows where the conference is.

Stoljar (2017) claims that (3) admits both interrogative and free-relative readings. In its interrogative reading, (3) is true just in case Alice knows a proposition that answers the embedded question ‘where is the conference?’. If the conference is in Rio, then (3) will be true in virtue of Alice knowing the proposition that the conference is in Rio.¹⁰ In its free-relative reading, however, (3) is true only if Alice knows the particular place denoted by the noun phrase ‘(the place) where the conference is’, which we stipulated to be Rio. Here what matters is the epistemic relation to the object, a place, and not to a proposition.

The point can be made clearer if we use an example that has a more natural free-relative reading than an interrogative one. Suppose that I am thinking of my partner, and that my dog knows her. Now consider the following sentence:

• (4) My dog knows who I am thinking of.

The more natural reading of (4) is that what the dog knows is the person denoted by the noun phrase, which constitutes a free-relative reading. That (3) and (4) have free-relative readings is supported by other languages, such as Portuguese, French, German, Spanish and and Latin, languages that have two different verbs deeply associated with knowledge. For example, Portuguese has ‘saber’ and ‘conhecer’. French has ‘savoir’ and ‘connaître’. German has ‘wissen’ and ‘kennen’.¹¹ In those languages, free-relative readings of (3) and (4) would be translated by using the second verb of the pairs (conhecer, connaître, and kennen), whereas interrogative readings would be translated by using the first verb (saber, savoir and wissen). This suggests that free-relative readings are possible for knowledge verbs, but that English lacks a lexical subtlety to make it linguistically clearer when those readings are due.

If the above discussion is correct, and ‘know-wh’ sentences admit those two readings, it is bad news for proponents of the Linguistics Argument. It could very well be the case that, when we ascribe new WIL-knowledge to Mary, we ought to treat the linguistic phenomena through a free-relative reading, which plausibly doesn’t pose propositional knowledge. This amounts to a falsification of the third premise of the argument, which claims that the semantics of knowledge-wh poses propositional knowledge to the subjects of knowledge-wh ascriptions.

I see the following courses of action to the friend of the Linguistics Argument from here: a) To claim that ‘know-wh’ sentences do not admit free-relative readings; b) To accept the two different readings, but to claim that free-relative readings of ‘know-wh’ sentences involve propositional knowledge nonetheless; c) To accept the two different readings, but to claim that some interrogative WIL-knowledge can only be possessed in virtue of meeting some experiential condition, and that the phenomena at hand is better accounted by appealing to such demanding interrogative knowledge.

Alternatives a and b go against current convention on linguistics and philosophy,¹² so I won’t analyze them here. Alternative c, that some interrogative WIL-knowledge can only be possessed in virtue of meeting some experiential condition, is more promising.¹³ But take Chromophobe Mary before taking off her lenses. It seems that she has all possible answers to what-it-is-like questions. What could be an acceptable interrogative knowledge that she lacks? I see three ways of cashing out the intuition behind c. The first is through some sort of pragmatics of knowledge ascription. Take two individuals, Elliott and Smith, who are specialists on the experience of going to war. Both have written numerous books on the matter and, as far as descriptive knowledge can go, they can be considered to possess exactly the same amount of knowledge about going to war. But there is one difference. Smith has actually gone to a war, whereas Elliott has not. Because of that, the majority of people would most likely prefer to hear lectures on the issue given by Smith, even if those lectures are exactly the same as those given by Elliott.

One could take the case above to suggest that Smith has some interrogative WIL-knowledge that Elliott lacks, since the answers (lectures, books, etc) of the former are more credited than those by the latter.¹⁴ This approach is not implausible, but I won’t stress it here, due to it not being commonly defended in the KA literature. But in any case, it could very well be the case that the best explanation here (and it is indeed the one that I find to be more appealing) is that we better value listening to Smith on the matter precisely because he has free-relative WIL-knowledge, even if Elliott has the exact same interrogative knowledge. In other words, it could be the case that the pragmatics of the situation is to be accounted for in terms of free-relative knowledge-wh, rather than in terms of interrogative interpretations.

Another option would be to appeal to gradability or mastery of interrogative WIL-knowledge. Here Mary, from the room, already has all the pieces of interrogative knowledge, but what matters is that she brings the relevant pieces to specific grades or past a specific threshold.¹⁵ A major problem with that suggestion is that it is not clear that talk of mastery supports the Propositional Stance, which is a claim about new propositional knowledge.¹⁶ A further concern is that here too we can account for the phenomena by appealing to new free-relative knowledge, rather than by appealing to improvement of old interrogative knowledge: when Mary leaves her room she deepens her knowledge of what it is like to see red because she does that in a free-relative manner, even if from inside the room she has full interrogative mastery of what it is like to see red. It is not feasible to properly develop such a view here and I can only hint it, but the view is one that is not only a possible and competing way to account for the phenomena, but it is also one that enjoys elegance of explanation. While it is hardly intelligible how Chromophobe Mary improves her old answers to embedded WIL-questions, free-relative WIL-knowledge neatly fits on her epistemic progress.¹⁷

A third option to cash out c would be to say that some interrogative WIL-knowledge gets its experiential condition borrowed from the experiential condition of some concepts, which may be called “phenomenal concepts”. Since phenomenal concepts depend on being acquired through experiential means, some answers do too, for they are based on those concepts. This idea has a long tradition in the philosophy of mind, and for that it deserves more careful analysis, which is the goal of the next subsection. The reader will be able to assess the arguments by their own shortly, but, in advance, I will reject the proposal on the grounds that such demanding criterion of concept acquisition has the consequence of positing concepts that are private, and hence inexpressible in public language. If my arguments there are sound, leaning on that same reasoning to cash out alternative c would lead to equally private answers. This outcome renders the proposal unable to account for a vast array (if not the entirety) of the question-answer dynamics - given its anchoring on public exchanges between different subjects -, which is unsatisfactory for the linguistic purposes at hand. In any case, it seems that appealing to special “concepts” that are somehow lurking beneath sentences is already a departure from proper linguistic analysis and a venture into sheer philosophical grounds. Given those considerations, my assessment is that the Linguistics Argument is ill supported by linguistic analysis, which undermines its sole purpose.

4.2. Second Argument: The New Concept Argument

The second argument is perhaps the most direct defense of the Propositional Stance. Its rationale builds on the close connection between experience and concepts, then on the role of concepts as constituents of propositional attitudes.¹⁸ The argument runs roughly like this:

• P1) Mary acquires new concepts that can be used to formulate new propositional attitudes

• P2) Using her new concepts, Mary formulates new propositional attitudes that satisfy propositional knowledge conditions.

• P3) If one formulates new propositional attitudes that satisfy propositional knowledge conditions, one acquires new propositional knowledge.

• (C) Mary acquires new propositional knowledge.

Call the above the New Concept Argument. In direct or indirect form, it is historically the most representative argument of PS proponents.¹⁹ If any of the premises is false, the conclusion is not established. There is room to argue against all three premises. For example, against P3 one could argue that it rests on ambiguity and that mere new propositional attitudes that satisfy propositional knowledge conditions don’t suffice for new propositional knowledge, on the basis that the former can be understood as new fine-grained cognitively significant thoughts, whereas the latter has stronger and more coarse-grained requirements.²⁰ However, I will here focus on pressing P1 and P2.

P1 expresses the idea that when Mary leaves her room and has new experiences of color, she also acquires new thought content. This is so because she then supposedly becomes able to formulate new thoughts about those very experiences, i.e. thoughts that use the phenomenal aspect of the newly experienced colors as material. This idea of phenomenal thought content is most frequently cashed out in terms of phenomenal concepts. So, according to this approach, after her release Mary can attend to the phenomenal aspect of red and think “this is beautiful!”, where this captures some type of quality of her experience and gives her grounds for coining a new concept. Since the concept picks out the phenomenal aspect of an experience that was not available for her prior to the release, it seems to many that such concept was also unavailable before. If this is right, then, in virtue of acquiring a new concept, Mary becomes able to formulate all sorts of new thoughts.

Is this account correct? According to Ball (2009) and Tye (2009), it is not. Drawing on remarks based on Burge’s (1979, 1982, 1986) and Putnam’s (1970, 1975) work on social externalism, both Ball and Tye contest the demanding experiential condition for phenomenal concept acquisition required for the picture presented above to work. The point can be illustrated through a quick look at Burge’s (1979) popular story of the arthritis man. In that story, a man who has a range of true and false “attitudes commonly attributed with content clauses containing ‘arthritis’ in oblique occurrence” (op. cit., p. 77) goes to a doctor and claims that he has arthritis in his thigh. The doctor corrects the patient and informs that arthritis is an inflammation of the joints and that it thus cannot occur in the thigh. The patient is surprised by the information, but he defers such knowledge to the doctor and then proceeds to revise his beliefs that employ the concept ARTHRITIS.²¹ In virtue of the exchange with the doctor, the patient changes his belief that ARTHRITIS CAN OCCUR IN THIGHS to the belief that ARTHRITIS CANNOT OCCUR IN THIGHS.

The moral of the arthritis man story is that concepts are rather easy to possess, for even one who poorly comprehends a concept can, in virtue of the linguistic community, still possess that concept and employ it in propositional attitudes. Concept possession does not require full mastery of the given concept, and one can acquire the concept simply by linguistic deference. Ball and Tye then argue that such moral defeats the accounts that posit rigid criteria for the possession of phenomenal concepts.²² Their arguments depend on the claim that social externalism is true of phenomenal concepts.²³ Both argue on similar grounds to establish that. They give various arguments, but for matters of scope I will focus on that which I consider to be the strongest and most prolific one. By adapting Burge’s remarks on sameness of content between the doctor and the patient, Ball and Tye argue that there is na immediate advantage of applying social externalism to Mary’s case. Consider Mary, from inside her room, uttering:

• (5) I will never know what it is like to see red. Then she finally goes out and utters:

• (6) I was wrong, now I finally know what it is like to see red!

And, alternatively to the exchange (5)-(6):

• (7) I used to wonder what red would feel like, but now I know.

There seems to be a continuity of propositional content from (5) to (6), and one seems intuitively to be the contradiction of the other. The phenomenal concept story has trouble to explain this. That approach is committed to the thesis that (6) employs a phenomenal concept, whereas (5) does not, for at the moment of the (5) utterance Mary lacks possession of the phenomenal concept of red altogether. It is also committed to the thesis that, in (7), either a) both ‘wonder’ and ‘know’ do not take a phenomenal concept as a complement cause, or b) that only ‘know’ takes it, because a third option, where ‘wonder’ takes a phenomenal concept at its complement, is immediately ruled out on the phenomenal concept picture. Both a and b are unfortunate. Alternative a renders Mary’s purported newly acquired phenomenal concept absent from her rationale. Alternative b, in turn, renders the adversative conjunction ‘but’, plausibly displayed in Mary’s utterance and thought, completely idle to her rationale.²⁴

Veillet (2012) champions the attempt to save the New Concept Argument from objections like the ones above. She presents two arguments. The first aims to account for (5)-(6), while the second aims to account for (7). The first one is introduced through her story of “Annie the Astronomer”:

Imagine that Annie is at first quite naïve, at least inasmuch as she possesses the concept HESPERUS but not the concept PHOSPHORUS. Nonetheless, Naïve Annie can be introduced to the word ‘Phosphorus’ in just the way that Inexperienced Mary is introduced to the word ‘R’ while still in her room. Naïve Annie can be told only that the word is used in her community to express a concept that picks out the very same planet her concept HESPERUS picks out. (VEILLET, 2012, p. 113)²⁵

The goal of this story is to show that Mary, in being introduced to a word that is meant to be the vehicle of a phenomenal concept (such as ‘R’ for the phenomenal concept R), would be in an analogous position to Annie’s. Annie has the concept HESPERUS, but not the concept PHOSPHORUS. Someone then introduces to her the word ‘Phosphorus’ and says only that it is associated with a concept that has the same reference as the concept HESPERUS. Veillet argues that in such a case, when Annie utters sentences that contain the word ‘Phosphorus’, she would be actually expressing her concept HESPERUS and that she would be still to acquire the concept PHOSPHORUS. Veillet then claims that Mary is in the same spot: when she utters sentences with her new word ‘R’, she would be actually expressing an old nonphenomenal concept that she learned to associate with the new word, as she would be still to acquire the phenomenal concept R. Such reasoning is then employed by Veillet (op. cit., p. 114) to substantiate the claim that (5) and (6) actually employ different concepts and that, despite linguistic appearance, they do not contradict each other at all.

I find this case unconvincing, even if we grant to Veillet that mere usage of new words is not sufficient to prompt new concepts. The only way for Mary or Annie to not prompt a new concept is if the new words are introduced to their vocabularies in a very poorly informative way, as it indeed is the case in Veillet’s story. In effect, Veillet’s Annie is basically told something like: “here’s a new word for your old concept!”. She is not told, for example, that ‘Phosphorus’ stands for a concept that people used to believe to have a very different referent from the one of the concept HESPERUS, or that ‘Phosphorus’ stands for a concept that used to be associated with the morning, while HESPERUS used to be associated with the evening.

Now such concealment of information is not really applicable to Mary and KA, for here it is a starter that Mary has all information that is possible to get prior to her experience of colors. When introducing ‘R’ to Mary, we could say to her that this new word stands only to the concept R, and to no other concept. We could say that R performs different cognitive roles from other concepts associated with the color red, we could say that R captures a very specific what-it-is-like-ness, we could say that many philosophers find it hard to believe that she could possibly think about R, and so forth. That information would be enough to prompt a new concept that is different enough from the nonphenomeal ones. Because of that, I conclude that Veillet’s first argument is unsuccessful and that (5)-(6), as the linguistic appearance suggest, indeed do employ the same concept and contradict each other.²⁶

This is enough to counter Veillet’s whole defense, but it is worth looking into her rationale towards (7). Veillet thinks that (7) would be troublesome for the phenomenal concept story even if her account of (5)-(6) succeeded. And it indeed is, since, as I claimed, the two open, phenomenal concept friendly, alternatives for addressing (7) are not very appealing. In her story, after unequivocally acquiring the concept PHOSPHORUS, Annie could not truthfully say:

• (8) I used to wonder whether Phosphorus is bright, but now I know.

Annie could not (without a defect in her reasoning) truthfully utter (8) with the goal of expressing the proposition I USED TO WONDER WHETHER PHOSPHORUS IS BRIGHT, BUT NOW I KNOW THAT PHOSPHORUS IT IS BRIGHT, for she did not have the concept PHOSPHORUS previously. Veillet acknowledges this issue, but, in a surprising move, she claims that such problem is not transposed to (7), for there is an important disanalogy between (7) and (8). She reasons as follows:

In uttering (7), Mary is not singling out any one particular fact about red that she used to wonder about and that she now knows: she is not saying that she used to wonder whether some thing was true of red; she is not saying that she now knows that fact to be true. Instead, she is saying that she used to wonder, much more generally, what R feels like, and what she now knows is, generally speaking, what R feels like. However, wondering about or knowing what R feels like is interestingly unlike wondering whether or knowing that, e.g., seeing red feels like R. (VEILLET, 2012, pp. 116-7)

The attentive reader will notice that Veillet’s claims only strengthen the overall argument of the present paper. The preceding quote comes very close to claiming that sentences such as (7) are to be read in a free-relative way, while (8) is to be read interrogatively. Indeed, Veillet proceeds to make analogies between (7) and sentences employing ‘who’, such as (4), suggesting that what she has in mind is a relation to an object, rather than to a proposition. I do think that a free-relative reading is by a great margin the best way to account for (7), as it is explicit in my discussion of the Linguistic Argument. However, this is far from being a great outcome to the proponent of the New Concept Argument. For if a free-relative reading is the best way to account for sentences such as (7), then it is also the best way to account for any sentence through which Mary cashes out her new knowledge, such as (6). But in free-relative readings of those sentences no new special concepts are required to explain the epistemic progress. So this line of reasoning cannot really be put in service to the picture of phenomenal concepts. I take those arguments to show that Veillet’s defense of phenomenal concepts against social externalism is unsuccessful.

From here, the proponent of the New Concept Argument may be tempted to adopt a familiar strategy: resorting to demonstrative concepts. Mary can only acquire the phenomenal concept when she becomes able to think that thisthis is making an ostention to her experience (or its features).²⁷ Since her experiences are private, any ostension to them is bound to be private as well. The consequence of this suggestion is that phenomenal concepts are private and not expressible in public language.²⁸ is red. But what is the demonstrative making ostension to here? Certainly not to red objects, such as the previous square, since, if you recall, Chromophobe Mary has access to all of those before taking her lenses off. I take it that such suggestion can only work if Mary’s

My conclusion from the above arguments is that claiming that phenomenal concepts are not expressible in public language is the most promising (and perhaps the only) alternative to proponents of the New Concept Argument (at least to those who are not willing to reject the Burge-Putnam cases across the board). Indeed, all the arguments presented in this subsection, from both sides of the dialectics, depend on the assumption that phenomenal concepts are expressible in public language. As Ball (2009, p. 250) points out, this is an assumption that phenomenal concept theorists typically make. Ball proceeds to argue that the claim that phenomenal concepts are not expressible in language isn’t plausible. However, his arguments are rather underdeveloped and unconvincing.²⁹ So, at this point, I take it to be a valid option for the proponent of the New Concept Argument.

I am now ready to proceed by stating the dilemma that we reached:

First Dilemma - Either phenomenal concepts are inexpressible in public language or they cannot perform the role requested by the New Concept Argument.

Embracing the public inexpressibility of phenomenal concepts might not be, at the start, too big of a loss for the phenomenal concept theorist. After all, they are usually more than ready to accept special features of those concepts. However, there is a price to pay. By postulating essentially private content, the position violates the principle of publicity of thought, one famously related to Frege (1892). The publicity of thought is an important, if not crucial, mechanism to explain successful interpersonal communication through a framework of propositional attitudes, so the incompatibility here comes as a methodological disadvantage for the position under consideration.³⁰ Another problem comes from the incompatibility of private content with theories that construe language and meaning as an enterprise that is necessarily public, such as semantic pragmatism.³¹ Here the disadvantage is a theoretical incompatibility with influential accounts of meaning and language.

This is not all. Embracing essentially private concepts bears further implications for the New Concept Argument. The second premise of the argument states that, by using her newly acquired concepts, Mary formulates new propositional attitudes that satisfy propositional knowledge conditions. Remember that, as discussed in section 3, propositional knowledge requires true propositions. As emphasized above, propositions that employ publically inexpressible concepts are themselves publically inexpressible. The outcome is that, if P2 is correct, there are true propositions that are publically inexpressible. Hence there are publically inexpressible truths.³² From this, we can state a second dilemma:

Second Dilemma - Either phenomenal concepts cannot perform the role requested by the New Concept Argument or there are essentially private, publically inexpressible truths.

I take the above dilemma, which is rarely acknowledged,³³ to be a sour outcome for proponents of the New Concept Argument, one that makes the argument highly unattractive. On contemporary standards, the notion of truth is deeply associated, philosophically and otherwise, with public validation or verification, what makes it hard to provide a robust and respectful proposal that posits private truths. As Wittgenstein has famously remarked on Philosophical Investigations (entry §258), “correctness” cannot be applied to private thoughts. I can see three main dialectic paths for the advocate of private content here: a) to claim that while there is private content, such content lacks truth-value; b) to claim that private content has truth-value, but that its truth-value is parasitic on the truth-value of public content; c) to claim that private content has intrinsic truth-value.

Alternative a cannot be used to defend the Propositional Stance. PS, and the Knowledge Argument debate, are about new true thoughts, so thoughts without truth-value won’t do it. Alternative b is perhaps the most common one taken by physicalists who are proponents of the New Content Argument. The main problem with this suggestion is that its intelligibility is hostage of a theory that draws a parallel between content construed privately, on one hand, and content construed publically, on the other. I know of no theory that accomplishes this, and I think that the prospects are rather dim. If we can account for the phenomena while avoiding that heavy explanatory burden, we should. Alternative c is, in my opinion, the most stable one as a defense of PS.³⁴ Its evident problem, besides the already rather mysterian commitment to private truths, is that it is dualist in letter. Since most contemporary philosophers have an anti-dualism attitude, I’ll here limit myself to consider to be self-evident that endorsing c is undesirable.

So here is how things stand. Although I didn’t construe the dilemmas as some “knockdown argument”, the outcome of positing essentially private truths puts considerable pressure on the New Concept Argument, an argument that, under its surface, shows to be highly committing. I hope to have shown that considerations weigh against it. The price for endorsing the argument may be too high, and, to avoid problems such as the ones discussed above, we may be better off exploring alternative routes to account for Mary’s epistemic progress, such as focusing on non-propositional knowledge and free-relative readings of what-it-is-like sentences.³⁵

This leads me to the assessment that the Propositional Stance is not convincingly supported by the New Concept Argument. Endorsement of the Propositional Stance through that argument should therefore be withdrawn or at the very least be met with greater skepticism than it has been done.

Conclusion

We surveyed two arguments in favor of the claim that Mary acquires new propositional knowledge. I argued that they are unconvincing. My arguments for rejecting the Linguistics Argument appealed to free-relative readings of what-it-is-like sentences, which plausibly provide the ground for ascription of non-propositional rather than of propositional knowledge. My arguments against the New Concept Argument appealed to the application of social externalist considerations to conclude that phenomenal concepts are not attractive tools to the proponent of the Propositional Stance, since they commit the proponent of the argument to essentially private truths.

My overall conclusion is obviously negative, given that my main goal was to demotivate the Propositional Stance, hence to show what Jackson’s Mary does not learn. There is still the hard task of carefully explaining what Mary does actually learn (if she indeed learns something), but I cannot pursue that goal here and will leave it for future research. There are many promising routes to that end that do not depend on propositional knowledge, but they have been fairly under-explored in the literature. I call for a direction of efforts towards that goal.

Acknowledgements

I would like to thank André Abath, Marco Aurélio Alves, Joel Temporal, Victor Moura and audience to a talk at the XX ANPOF conference for helpful comments on the paper.

Research Data Availability:

All research data are available within the main text of the article.

References

• ABATH, André J. (2022). Knowing What Things Are: An Inquiry-Based Approach. Synthese Library, no. 466.
• ALTER, Torin (2013). Social Externalism and the Knowledge Argument. Mind 122(486).
• _______. (2019). Concept Mastery, Social Externalism, and Mary’s New Knowledge. In: The Knowledge Argument. Ed. Sam Coleman. Cambridge University Press, 2019, pp. 141-60.
• ALTER, T. & WALTER, S. (2006) Introduction. In: Alter, Torin Andrew & Walter, Sven (eds.). Phenomenal Concepts and Phenomenal Knowledge: New Essays on Consciousness and Physicalism. New York, US: Oxford University Press, pp. 1-13.
• BAKER, C. 1995. English Syntax (2nd edition) MIT Press, Cambridge MA.
• BALL, Derek (2009). There are no phenomenal concepts. Mind 118(472): 935-962.
• _______. 2013. ‘Consciousness and Conceptual Mastery: Reply to Alter.’ Mind 122: 497-508.
• BURGE, Tyler (1977). Belief De Re. Journal of Philosophy 74(6): 338-362.
• _______. (1979). ‘Individualism and the Mental’. Midwest Studies in Philosophy, 4, pp. 73-121.
• _______. (1982). ‘Other Bodies’. In Woodfield 1982, pp. 97-120.
• _______. (1986). ‘Intellectual Norms and the Foundations of Mind’. Journal of Philosophy, 83, pp. 697-720.
• BYRNE, Alex (2002). Something about Mary. Grazer Philosophische Studien 63(1): 27-52.
• CATH, Yuri. Knowing What It Is Like and Testimony. Australasian Journal of Philosophy, 97: 1, pp. 105-120, 2019.
• _______. (2024). Knowing What It Is Like and The Three “Rs”. In Íngrid Vendrell-Ferran & Christiana Werner (eds.), Imagination and Experience: Philosophical Explorations. New York, NY: Routledge.
• CHALMERS, David J. (1996) The Conscious Mind: In Search of a Theory of Conscious Experience. Santa Cruz: Department of Philosophy, University of California, 1995.
• _______. (2002). The content and epistemology of phenomenal belief. In Aleksandar Jokic & Quentin Smith (eds.), Consciousness: New Philosophical Perspectives. New York: Oxford University Press. pp. 220-72.
• CONEE, Earl. Phenomenal Knowledge. Australasian Journal of Philosophy, 72: 2, pp. 136-150, 1994.
• DAVIDSON, Donald (1992). Three varieties of knowledge. In A. Phillips Griffiths, A. J. Ayer: Memorial Essays. New York: Cambridge University Press. pp. 153-166.
• DENNETT, Daniel C. (1991). Consciousness Explained. Penguin Books.
• FARKAS, Katalin (2019). Objectual Knowledge. In Jonathan Knowles & Thomas Raleigh (eds.), Acquaintance: New Essays. Oxford, United Kingdom: Oxford University Press. pp. 260-276.
• FREGE, Gottlob (1892). On Sense and Reference. In: Translations from the Philosophical Writings of Gottlob Frege. ed. Peter Geach and Max Black. Oxford: Blackwell, 1952, 56-78.
• GETTIER, Edmund (1963). Is Justified True Belief Knowledge? Analysis 23(6): 121-123.
• GRZANKOWSKI, Alex & TYE, Michael. What Acquaintance Teaches. In: Acquaintance: New Essays. Ed. Jonathan Knowles & Thomas Raleigh. New York: Oxford University Press, 2019.
• HABGOOD-COOTE, Joshua (2018). Knowledge-how: Interrogatives and Free Relatives. Episteme 15(2): 183-201.
• HELLIE, Benj (2004). Inexpressible truths and the allure of the knowledge argument. In Peter Ludlow, Yujin Nagasawa & Daniel Stoljar, There’s Something About Mary: Essays on Phenomenal Consciousness and Frank Jackson’s Knowledge Argument. MIT Press. pp. 333.
• JACKSON, Frank. Epiphenomenal Qualia. The Philosophical Quarterly, Vol. 32, No. 127, pp. 127-136, Abril 1982.
• _______. 1986. What Mary Didn’t Know. Journal of Philosophy 83: 291-295.
• _______. 2022. A Change of Heart. In: Philosophers on Consciousness: Talking About The Mind. Bloomsbury Academic, 2022, pp. 39-47.
• LEE, A. Y. 2023. Knowing What It’s Like. Philosophical Perspectives 37(1): 187-209
• LEWIS, David K. (1983). Postscript to “mad pain and Martian pain”. Philosophical Papers 12: 122-133.
• _______. (1988) What Experience Teaches, in Mind and Cognition: An Anthology (Second Edition), ed. W.G. Lycan, Oxford: Blackwell: 447-460.
• LYCAN, William G. (1995). A limited defense of phenomenal information. In Thomas Metzinger (ed.), Conscious Experience. Paderborn: Ferdinand Schoningh. pp. 243-58.
• LYNCH, Kevin (2019). Knowing What an Experience Is Like and the Reductive Theory of Knowledge‐wh. Analytic Philosophy 61(3): 252-275.
• MARGOLIS, Eric & LAURENCE, Stephen (2007). The ontology of concepts: Abstract objects or mental representations? Noûs 41(4): 561-593.
• MENDOLA, Joseph (2008). Anti-externalism. New York: Oxford University Press.
• NAGEL, Thomas. What Is It Like to Be a Bat? The Philosophical Review, Vol. 83, No. 4, pp. 435-450, Outubro 1974.
• NEMIROW, Laurence (1980). Review of Nagel’s mortal questions. Philosophical Review 89(3): 473-7.
• NIDA-RÜMELIN, Martine (1995). What Mary couldn’t know: Belief about phenomenal states. In Thomas Metzinger (ed.), Conscious Experience. Paderborn: Ferdinand Schoningh. pp. 219-41.
• _______. (1998) “On Belief About Experiences: An Epistemological Distinction Applied to the Knowledge Argument”, Philosophy and Phenomenological Research, 58: 51-73.
• NIDA-RÜMELIN, Martine and DONNCHADH O. CONAILL, “Qualia: The Knowledge Argument”, The Stanford Encyclopedia of Philosophy (Spring 2024 Edition), Edward N. Zalta & Uri Nodelman (eds.), URL: <https://plato.stanford.edu/archives/spr2024/entries/qualia-knowledge/>.
  » https://plato.stanford.edu/archives/spr2024/entries/qualia-knowledge/
• ONOFRI, Andrea (2018). The Publicity of Thought. Philosophical Quarterly 68(272).
• PAPINEAU, David (2011). Phenomenal Concepts and the Private Language Argument. American Philosophical Quarterly 48(2): 175.
• PARENT, T. 2014. Knowing-Wh and Embedded Questions, Philosophy Compass 9/2: 81-95.
• PAVESE, Carlotta, “Knowledge How”, The Stanford Encyclopedia of Philosophy (Fall 2022 Edition), Edward N. Zalta & Uri Nodelman (eds.), URL: <https://plato.stanford.edu/archives/fall2022/entries/knowledge-how/>.
  » https://plato.stanford.edu/archives/fall2022/entries/knowledge-how/
• PERRY, J. (1994) Intentionality. In S. Guttenplan (ed.), The Blackwell Companion to the Philosophy of Mind (Oxford: Blackwell), 386-95.
• _______. (2001), Knowledge, Possibility, and Consciousness, Cambridge, MA: MIT Press.
• PIETROSKI, Paul M. (2000). On Explaining That. Journal of Philosophy 97(12): 655.
• QUINE, Willard V. (1975). The nature of natural knowledge. In Samuel D. Guttenplan, Mind and language. Oxford [Eng.]: Clarendon Press. pp. 1975-67.
• PUTNAM, Hilary (1970). Is semantics possible? Metaphilosophy 1(3): 187-201.
• _______. (1975). The meaning of ‘meaning’. Minnesota Studies in the Philosophy of Science 7:131-193.
• RYLE, Gilbert. The Concept of Mind. Chicago: University of Chicago Press, 1949.
• STANLEY, Jason & WILLIAMNSON, Timothy. Knowing How. Journal of Philosophy, 98(8), pp. 411-444, 2001.
• STANLEY, Jason (2011). Know How. Oxford, GB: Oxford University Press.
• STOLJAR, Daniel (2005). Physicalism and phenomenal concepts. Mind and Language 20(2): 296-302.
• _______. (2016) ‘The Semantics of “What It’s like” and the Nature of Consciousness’, Mind 125(500), 1161-98.
• _______. (2017). The Knowledge Argument and Two Interpretations of ‘Knowing What it’s Like’. In Dale Jacquette (ed.), The Bloomsbury Companion to the Philosophy of Consciousness. New York: Bloomsbury Academic.
• TOMASELLO, Michael (2003). Constructing a Language: a usage-based theory of language acquisition. Cambridge, MA: Harvard University Press.
• TYE, Michael (2009). Consciousness Revisited: Materialism without phenomenal concepts. Cambridge, MA: MIT Press.
• _______. (2011). Knowing What It Is Like. In John Bengson & Marc A. Moffett (eds.), Knowing How: Essays on Knowledge, Mind, and Action. Oxford, England: Oxford University Press USA. pp. 300-13.
• VAN GULICK, Robert (2004). So many ways of saying no to Mary. In Peter Ludlow, Yujin Nagasawa & Daniel Stoljar (eds.), There’s Something About Mary: Essays on Phenomenal Consciousness and Frank Jackson’s Knowledge Argument. MIT Press.
• VEILLET, Bénédicte (2012). In Defense of Phenomenal Concepts. Philosophical Papers 41(1): 97-127.
• WITTGENSTEIN, Ludwig (1953). Philosophical Investigations. New York, NY, USA: Wiley-Blackwell. Edited by G. E. M. Anscombe.
• ZAGZEBSKI, Linda (1999). “What Is Knowledge?”. In John Greco & Ernest Sosa (eds.), The Blackwell Guide to Epistemology. Malden, Mass: Wiley-Blackwell. pp. 92-116.

• 1
  I’ll use factual knowledge and propositional knowledge as equivalent notions. The notion will be discussed in greater detail in section 3.
• 2
  What I call the Propositional Stance can be seen as a standard conjecture within the positions which are commonly classified as dualism and type-B materialism (see Chalmers (2002, pp. 221-2) or Alter & Walter (2007, p. 5) for this exact characterization). In a sense, my goal here of arguing against the Propositional Stance can be seem as supporting type-A materialism. But bear in mind that such terminology and typology are not very precise for our current topic. The negative part of blocking new propositional knowledge in KA goes hand in hand with type-A materialism, but it is compatible with that initial movement to explain the positive side of what Mary learns through non-propositional acquaintance knowledge. It is unclear if a proposal like that should ultimately be classified as materialist.
• 3
  As pointed out by some scholars (e.g. by Stoljar 2005, Ball 2009 and Byrne 2002) the core idea here can be seem as very similar to the old empiricist claim, defended by Hume and Locke, that some “ideas” of colors and sounds cannot be entertained by the congenitally blind and deaf.
• 4
  HPI can be ambiguous between a weak and a strong reading. The weak reading says that there is propositional knowledge that initially can only be acquired through experiential means. The strong reading says that there is propositional knowledge that can only be acquired through experiential means, and never by other means. It is clear that the Propositional Stance depends on the strong reading. Jackson’s scenario satisfies with ease the condition imposed through the weak reading of HPI, for Mary does have access to all descriptive knowledge, including a rich phenomenology of colors. For that, my position in this work is only incompatible with the strong reading of HPI, and my arguments should be viewed as compatible with the weak one.
• 5
  Ball (2009, p. 242) makes the same point.
• 6
  Such characterization is as general as it can be, and it is deliberately so, because properly understanding both poles of the propositional knowledge relation proves to be philosophically difficult. On the side of the knower, there is a need to specify the criteria on what put the subject in good condition to know. Since the Greeks it has been thought that the state of knowledge of truths is to be analyzed through the notion of belief. It is also commonly accepted that the subject needs to have good justification for her belief. On top of that, since Edmund Gettier’s (1963), work there is a pursuit for a further requirement or for an alternative model to the justified true belief one. On the side of what is known, there is a need to specify the exact nature of propositions, on which there is great dissension in the literature. The relation of propositional knowledge can be affected by clarifications on those topics, but fortunately our current discussion can move forward with the general characterization that was presented.
• 7
  A strong “intellectualist” about knowledge may disagree that there is knowledge of things other than true propositions, i.e. they may disagree that there are types of knowledge other than propositional knowledge. This epistemic reductionism of types of knowledge is more commonly done to practical knowledge (see Pavese 2022 for an overview). I am not aware of a recent well-developed global epistemic reductionism that may bear consequences upon the KA debate, so I won’t consider the possibility here.
• 8
  The standard analysis treats phrases containing ‘know’+paradigmatic ‘wh’ words (who, what, where, why, when, whether and, sometimes included, how) as involving knowledge of a proposition that answers the question embedded in the wh-clause. According to this analysis, the sentence ‘John knows who Billy Wilder is’, for example, ascribes to John knowledge of a proposition that answers the embedded question ‘who is Billy Wilder?’. See Parent (2014) and Abath (2022, ch. 3) for detailed overview.
• 9
  C.f. Baker, 1995, p. 205.
• 10
  To make things simpler here, we can, without much loss, avoid discussions of mention-some/mention-all problems.
• 11
  Many scholars point out the same distinction in order to press a variety of points. See, for example, Stanley 2011 (pp. 36-7) and Farkas 2019 (p. 264).
• 12
  The conventions being, respectively, that free-relative readings are linguistically legitimate and that propositional knowledge is not necessary for objectual knowledge.
• 13
  This route is explicitly endorsed by Lee (2023) and Cath (2024).
• 14
  For illustrative purposes, take discussion by Abath (2022, p. 34), who writes that on an interrogative inquiry-based approach to knowledge-wh “we should not take subjects as knowing-wh in terms of de re knowledge, as “knowledge of a thing as a thing to which the wh-predicate applies”. Rather, knowledge-wh is the result of us successfully settling the questions we are inquiring into. Knowledge-wh is thus knowledge of propositions that properly answer to questions (that are true and settle them) in a given context”. One could from here argue that only Smith has the relevant propositional knowledge-wh because only he can settle the relevant questions.
• 15
  See Cath (2019) for an illuminating articulation of this proposal.
• 16
  The emphasis on new propositional knowledge is not a mere verbal demand. The Knowledge Argument is an argument about a substantial epistemic gap between pre and post-release Mary’s propositional knowledge. Proponents of PS typically accept that gap, and do so by positing new propositional pieces. It is unclear if improvement of old pieces is suited to do the same job (cf. Ball 2013, p. 506; Alter 2019, pp. 151-3).
• 17
  Cath (2019, pp. 10-1) draws on suggestion by Tye (2009; 2011) to present such improvement of interrogative WIL-knowledge by talking of phenomenal ways of interrogatively knowing-wh, where those phenomenal ways are based on acquaintance relations. However, laying out the specifics of that idea can be very challenging. See Grzankowski & Tye (2019), who try to pursue the idea further but end up with a list a problems and no satisfactory way out.
• 18
  I am here taking concepts to be entities that can occur in thoughts and other propositional attitudes and that are individuated in a fine-grained way. I am taking conception to be the “content” that one has under a given concept. I am also taking subject S to (necessarily and sufficiently) possess concept φ if S is able to exercise φ in their thoughts and other propositional attitudes.
• 19
  For direct defenses of it, see e.g. Nida-Rümelin (1995, p.p. 233-7) and Perry (2001 pp. 101-2).
• 20
  See Tye (2011, p. 307) for the distinction between propositional knowledge as cognitive significance and as elimination of open metaphysical possibilities. In the context of the KA debate, David Lewis famously (e.g. in 1988) required the latter as a necessary condition for new propositional knowledge.
• 21
  One who claims, contra Burge and Putnam, that the patient, in virtue of having an inadequate conception, cannot have the concept ARTHRITIS is committed to the claim that the patient is employing a different concept altogether, such as T-ARTHRITIS. This view faces known difficulties to explain the disagreement between doctor-patient and the latter’s correction of belief. Meanwhile, social externalism has an easier time to handle those cases. Since proponents of the New Concept Argument typically accept the Burge-Putnam cases (cf. Veillet 2012, p. 98), I will take them for granted here. But see Mendola (2008) for a general attack against externalism.
• 22
  Given those charges, Alter (2013; 2019) recasts the argument as a claim about concept mastery, a proposal that I can’t properly survey in the present paper. But at any rate, at the end of subsection 4.1 we discussed how talk of mastery is dubious as support for the Propositional Stance. Similar worries appear here, and Alter (2019, pp. 151-3) himself is aware of the limitation. A further worry is a disanalogy between purported phenomenal mastery and typical Burge-Putnam mastery cases. One implication of the social-externalist position is that “there are no concepts off limits” and therefore “no propositional thoughts of limits to Mary” (Grzankowski & Tye, 2019, p. 75). This particularity of Mary’s epistemic situation makes it very different from usual examples of conception impoverishment. On Burge’s story, the doctor has concept mastery, whereas the patient does not. Arguably the difference in conception of ARTHRITIS between the two is largely conceptual and propositional. There are many concepts that the patient lacks, and many propositional truths about arthritis which he is ignorant about. If the patient were to acquire mastery of ARTHRITIS, this would come in virtue of acquisition of many new propositional pieces. In the case of Mary there are no such pieces unavailable. Mary knows them all. Here it is tempting to claim that there is not even propositional mastery.
• 23
  Social externalism is a thesis that defends that meaning and contents of thought are not solely determined by an individual apart from its community and environment. It can be illustrated by Putnam’s (1975, p. 144) famous remark that “meanings just ain’t in the head”. It seems natural to assume that, once we accept that thesis, we ought to apply the same conditions to phenomenal concepts, i.e. to assume that phenomenal concepts are no exception to the rule. If for some reason we should not apply the rule, then the burden of giving an account to sustain the exception is on the side of the phenomenal concept strategist. I’m still to find a convincing account for that end. But in any case, since the phenomenal concept theorist is more than willing to claim that phenomenal concepts are special and different from all other concepts, due argument to establish that they are not follows.
• 24
  One could be tempted to argue here that this would not be an issue, since defects of reasoning such as the one that alternative b suggests happen all the time. However, Mary is stipulated to be a perfect reasoner in the debate. Introducing imperfect reasoning just to validate alternative b would be an obvious ad hoc move that should not be taken seriously.
• 25
  I am adapting Veillet’s technical terminology to match ours.
• 26
  A reviewer has prompted me to clarify a potential ambiguity at this point. They argue that while Annie is introduced to a new word in natural language, Mary is not. The point is rather, they claim, that Mary expresses different concepts through the same word. However, Veillet is considering a scenario where, by stipulation, a new word is actually introduced to Mary’s vocabulary (see the quote on p. 16). An illustrative way to appreciate the point is to imagine a scenario where there are twin Marys. One of them leaves the room, the other does not. The one who leaves creates new color concepts and comes up with neologisms to exclusively capture those new concepts. Rompadompa, she says while attending to redness, captures this. Then “outtie” Mary interacts with “innie” Mary through the computer and brings up the new word. Veillet’s point is that merely introducing the word wouldn’t prompt the relevant concept for innie Mary.
• 27
  Maybe it will be argued here that the demonstrative is picking a token that exemplifies a type of brain state. However, Tye’s (2009, pp. 42-3) “cerebroscope” case renders this suggestion implausible.
• 28
  See Byrne (2002) for a similar conclusion. See also Papineau (2011, p. 179) for discussion of the privacy of phenomenal concepts when contrasted with Wittgenstein’s Private Language Argument. A reviewer has suggested that an advocate of phenomenal concepts could argue here that a phenomenal concept is just partially incommunicable. My response is that no one should deny that Mary isn’t in an ideal position in regard to concepts of colors. When she sees red for the first time, she will no doubt enrich her concept of phenomenal redness, similarly to how Burge’s patient would enrich his concept of arthritis. For example, it is a compelling interpretation that now Mary would then be able to store samples of the experiences of seeing colors in the “conception” of her concepts. The claim here is only that Mary does not depend on that richness to possess the concept.
• 29
  Ball presents two arguments. The first consists in claiming that Mary expresses her new knowledge by saying ‘that is what it is like to see red’ and that such utterance would indicate, on a phenomenal concept account, expressibility of a phenomenal concept. The second argument consists in claiming that it is possible to introduce a new dedicated word to precisely be a vehicle for the phenomenal concept. These arguments are not convincing, for both can only be made if it is assumed from the outset that phenomenal concepts are expressible. On a charitable reading, what Ball seems to be arguing is that, since we have good candidates to perform the role of vehicles of those concepts, there is good reason to think that they indeed express the concepts. This conclusion is, however, too quick and calls for more work.
• 30
  See Onofri (2018) for an overview of discussion about the publicity of thought. See also Margolis & Laurence (2007), who argue that Frege’s anti-psychologism arguments in favor of the publicity of thought rest on a confusion about the token/type distinction. Even if they are right, the same rationale has limited applications here, since, as I have argued, the outcome of positing phenomenal concepts is that there are thought tokens which cannot be typified as to be made public.
• 31
  The classic example is Wittgenstein (1953), but see Tomasello (2003) for a recent and empirically informed articulation.
• 32
  I am taking truth to be individuated in a fine-grained way and to stand one for one to true propositions, which is fairly common. So, in this usage, the true propositions HESPERUS REFERS TO VENUS and PHOSPHORUS REFERS TO VENUS are two different truths.
• 33
  But see Hellie (2004), who faces the consequence and, in favor of a picture of phenomenal concepts, openly accepts that there are inexpressible truths.
• 34
  See Chalmers (2002) for a detailed proposal of this sort.
• 35
  While entertaining the idea that my position is correct, a reviewer rightfully prompted me to clarify how rejecting the New Concept Argument would prevent difficulties concerning the privacy and inneffability of some aspects of experience. My response is that while talk of non-propositional knowledge does not eliminate private aspects of experience (I’m not sure that this is possible), that attitude has the advantage of accounting for the phenomena while preserving the publicity of concepts, thoughts and, better yet, truth. In my opinion, this alone renders the position more desirable.
• Article info
  CDD: 121.
• Funding:
  This work was supported by CAPES: 88887.685831/2022-00.

Publication Dates

• Publication in this collection
  07 Nov 2025
• Date of issue
  2025

History

• Received
  09 Dec 2024
• Reviewed
  30 June 2025
• Accepted
  05 Aug 2025