The Philosophical Significance of Secondary Uses of Language in Wittgenstein’s Late r Philosophy

This paper aims to provide an account of Wittgenstein’s employment of the distinction between primary and secondary use of words. Against views that circumscribe its relevance to aesthetics and ethics, the paper demonstrates that there are many instances of secondary uses in Wittgenstein’s work that are not reducible to those limited applications. Additionally, as secondary uses are often interpreted as having an expressive function, the paper argues that we cannot reduce secondariness to a single unifying principle, because the distinction is philosophical , as it works as a powerful device to tackle different, often unrelated, philosophical issues.

It is unclear where the philosophical significance of such a distinction lies, as well as its general function within Wittgenstein's conception of language. In the literature, at least three different interpretative approaches can be pointed out regarding secondariness. The first sees secondary uses as mostly irrelevant, almost uninteresting by-products of our life with language. As Oswald Hanfling points out, Wittgenstein's examples are "idiosyncratic" and "abnormal" (1990: 122), they belong to the "margins of language " (2002: 152). If secondary uses are marginal, exceptional -not "essential" to the concept of meaning and language, as Hans Johann Glock argues (1996: 40) -we should not be much bothered to understand their nature.
On the other hand, many scholars have employed secondariness to understand our aesthetical discourse (Tilghman 1984, Hanfling 1990, 2002, Budd 2006. Ben Tilghman, for instance, complains that the examples 'for me the vowel e is yellow' and 'Thursday is lean' are idiosyncratic and infelicitous, as they risk discrediting the distinction between secondary and primary uses and its broader -and fruitful -applications to art and aesthetics (1984: 160). In a nutshell, these authors argue that many aesthetic descriptions in art -for instance when we say that a painting is 'dynamic,' a musical theme 'triumphant' or 'plaintive,' a dance step 'woody,' and so forth -are based on secondary uses of ordinary words. The appeal to secondary use to explain the logic of our attribution of aesthetic qualities has the undeniable advantage of dismissing or bypassing any theory that explains said attribution through the problematic notion of resemblance (see Hanfling 1990: 117-119 andBudd 2006: 135 -141). Along the same line, Cora Diamond employs secondary uses to account for our ethical discourse. In particular, the distinction between absolute and relative good that Wittgenstein makes in the Lecture on Ethics, is interpreted as an application of the later distinction between primary and secondary uses of the word 'good' (Diamond 1991).
Unlike the first, this second approach conceives secondary uses as a crucial feature of our life with language, which can be fruitfully employed to clarify certain issues in aesthetics or ethics. Their philosophical significance, even though circumscribed only to a limited area of language, may it be art or ethics, is fully recognized.
Finally, Michel ter Hark's work on Wittgenstein's philosophy of psychology outlines a third theoretical option for understanding the logic of secondary use. According to ter Hark, secondary uses of words such as 'meaning' and 'experience,' as well as 'Thursday,' 'yellow,' and 'lean,' have an expressive function (2011: 515): we employ words secondarily to express ourselves, to convey something about us and our experience. Accordingly, secondariness should be bound to our psychological discourse about ourselves and our inner experiences.
In this paper, I aim to offer an overview of Wittgenstein's employment of the distinction between primary and secondary uses of words, and develop a fourth interpretative option, which refuses to deny any significance to secondariness (as in the case of the first approach), without circumscribing it only to a limited area of language (as in the case of the second and the third). I will show that ter Hark's interpretation, even though substantially correct, is theoretically and philologically partial, insofar as it risks excluding those employments of words that Wittgenstein himself often mentions in the Nachlass, that are hardly expressive, and yet can be said to be secondary. It is the case of telling absurdist tales, attributing emotions to inanimate objects, giving instructions about how to play music, and describing aesthetic qualities. It follows that secondariness is far from being at "the margins of language." Second, in the light of this discussion, I will argue that the distinction is not reducible to an overarching principle, or function, able to capture every instance of secondariness, as its function is primarily philosophical, that is, it is meant to be a logical tool for distinguishing different uses of those words that can lead to confusions if conflated. Accordingly, the distinction is strictly problem-relative, as it is usually mentioned or introduced to tackle those puzzles that are caused by such confusions.
The paper is structured as follows. First, I will give a quick description of those criteria Wittgenstein points out to define secondariness in the second part of the Investigations. I will focus on the difference between secondary uses and metaphors, meant to highlight an important feature of secondariness: its unavoidable and yet spontaneous character. Second, I will illustrate ter Hark's interpretation and show thateven though it covers many cases of secondariness in the Nachlass -it is still unsatisfactory to confine secondariness to expressive language. Finally, I will proceed to show how this proliferation of examples is due to the philosophical nature of the distinction.

1.Unavoidable and yet Spontaneous: Secondariness and Metaphors
In the second part of the Investigations, Wittgenstein elaborates on the distinction between primary and secondary use by laying out three main criteria to identify secondariness.
The first involves explanation: we cannot point to words used secondarily as an explanation of their use. If we must teach the meaning of 'yellow,' we certainly point to a sample of the colour, drawings might be involved. Surely, the vowel e is not mentioned as an example of a yellow object (PPF: § 275). The fact that words used secondarily do not change the way we explain them, is also the reason we are not inclined to talk about different meanings here: the words 'fat,' 'lean,' and 'yellow,' are all explained in their usual way. The second involves use (PPF: § 276): one cannot use a word secondarily without knowing its primary meaning. We are here dealing with, to say with ter Hark, a form of "logical dependency 1 of a use upon another one" (2011: 515).
These criteria are perfectly extendible to metaphors and figurative language too. The third criterion, then, is meant to rule these out of the picture. The main difference between metaphors and secondary uses is that metaphors are mostly optional figures of speech that can be explained by paraphrase and thus avoided if they bring about misunderstandings. To use Cora Diamond's example, if I say that 'man is the cancer of our planet,' I can rephrase my thought way less emphatically by stating that humankind is an invasive species that is destroying natural ecosystems (1991: 227). This kind of explanation is precisely what is excluded in the case of secondary uses: we cannot appeal to another piece of language to paraphrase what we mean. Secondary uses are in this sense not optional; they are unavoidable, insofar as we are bound to use those specific forms of expression to convey the meaning we want to convey. "I could not express 1 Sometimes, in order to emphasise this kind of logical dependency, secondary uses are said to be parasitic (Hanfling 1990: 131, Baker Hacker 1990. However, I will avoid this qualification to define secondariness. Even though correct, it hides an implicit evaluation of secondary uses as a philosophically irrelevant phenomenon, something I explicitly challenge in this paper.
what I want to say in any other way than by means of the concept of yellow," Wittgenstein points out (PPF: § 278). At the same time, the unavoidable character of secondariness is not to be read as a psychological compulsion. That would mean that we could contemplate alternative forms of expression and not choose them because we feel urged otherwise. On the contrary, there is no alternative we can envisage when we use words secondarily. There is no other way to convey what I mean by saying that e is yellow than saying that e is yellow and the words I apply here feel like a natural and spontaneous extension of their use, which flows naturally from the very meanings of the words involved. 2 Incidentally, it is important to stress that Wittgenstein's point is entirely negative. He is not sketching in this regard a theory of metaphors, that is, he is not committed to the claim that every metaphor must be paraphrasable and must be optional. If so, it would be hard to defend such a theory. For instance, dead metaphors -the cannon is said to have a mouth, the table legs -are metaphors which became part of the common heritage of our language, and they are far from being optional; we just use them as the most appropriate forms of expression in certain circumstances.
The very notion of paraphrase is not that uniform either. In the case of Diamond's example, it is easy to appeal to non-figurative applications of words to convey the same thought. Not so for the controversial Shakespearian themed example of 'Juliet is the Sun,' where the paraphrase too is clearly figurative. We can explain that Juliet's role in Romeo's life is a centre of gravity that illuminates and makes his life bright, but these would be, again, figurative uses of the words 'centre of gravity' or 'illuminate.' Thus, there are metaphors that are paraphrasable through other figurative words, others that are not, whereas others are hardly optional, as in the case of dead metaphors. This indeterminacy, however, is not an argument against Wittgenstein's point, which remains narrow: the comparison with metaphors aims only to stress the unavoidable character of secondary uses, and together the fact that they are bound to the primary meanings of 2 I am here using the word 'spontaneous' with the sole intention to stress the difference with psychological compulsion. It is not a Wittgensteinian way to define secondariness. Spontaneity is mentioned in a (rather obscure) remark in the second part of the Investigations, in a way that is, however, compatible with my use: "What is new (spontaneous, 'specific') is always a language game" (PPF: § 335). Notably, spontaneity is connected here with novelty, a distinctive trait of secondary uses (LW I: § 61). their words. As such, to convey what I want to convey by saying that e is yellow I surely can use synonyms if needed, but I would be nevertheless bound to the meanings of the words I choose. Not so, if I want to describe the role of Juliet in my life as a lover or the physical prowess of my friend Jason when I say that he is a bull.
To sum up, secondary uses are in a way figurative, without being optional or explainable by paraphrase. They have an unavoidable character, as the meaning we want to convey is bound to the forms of expression we use, without however being compulsory. They feel as a natural extension of the use of words, in this sense they are an immediate or spontaneous use of language, without however constituting the primary meanings of words (by saying that Thursday is lean I am not changing the meaning of weekdays to be found in a dictionary). As in the case of metaphors, by using words secondarily the speaker means something through what their words conventionally mean.
Differently from metaphors, however, what we mean when words are used secondarily is directly and immediately inscribed in the words and concepts we use, so that any explanation that is not a mere repetition of the same words (or synonyms) is in principle opted out.

2.Secondariness and Expression
Even though the second part of the Investigations lays out the criteria for secondariness, it is dramatically short of examples. In order to get a clearer picture of what Wittgenstein had in mind, beyond the puzzling cases of coloured vowels and lean Thursdays, we should look in the Nachlass. There, as already noticed by Michel ter Hark, Wittgenstein looks committed to the idea that we use words secondarily to express ourselves, to convey a certain experience, as the following remark implicitly suggests: But why do you use just this expression for your experience? -such a poor fit! -That To give some context to the remark, Wittgenstein is here investigating the use of those requests, or orders, asking to utter an ambiguous word, and together mean only one of its meanings. Wittgenstein's example involves the German word Bank, which can mean either 'bench' or 'bank' (RPP II: § 571). These orders make sense only if we experience the meaning of our words, something a meaning blind is frequently said to be unable to do (Z: § 184, LW II: § 711).
Hence, secondary uses are mentioned in a context where Wittgenstein is exploring the role of experience of meaning in our use of language. The sentence 'the vowel e is yellow' is connected to a certain experience, an experience we express precisely by using the words secondarily. Primitive expressions of pain, such as cries and yells, are here introduced as a simple term of comparison of language used expressively. Importantly, Wittgenstein suggests that, if we overlook the function our sentence is supposed to have, we might take it "the wrong way," that is, as if those words have a different function (a descriptive one, for instance), and confusions might arise.
We might wonder what kind of experience we do really express by talking about coloured vowels. It is hard to see how we would vent to our own feelings by simply talking about the colours of sounds. It has been suggested that the experience in question is the one of psychological synaesthesia, a particular psychological condition that makes us experience involuntary cross-modal sensations, such as seeing a sound as colourful, or hearing a colour with a certain tone (Kindi 2009). It is likely that synaesthetes would use such expressions to compare their experiences, or to exemplify their own experience to somebody for whom letters do not have colours. However, the twin example of the lean Thursdays puts cross-modal sensations out of the picture. Even in this case an experience is expressed, but no synaesthesia is involved.
Accordingly, we can admit that psychological synaesthesia is expressible through secondary use, but the concept of experience Wittgenstein had in mind was probably broader. As ter Hark already suggested (2011: 516), the experience in question is more likely to be what Wittgenstein calls, in the context of his discussion of the experience of meaning, the atmosphere of our words. 3 A murky notion in Wittgenstein's Nachlass, the atmosphere can be defined as "the corona of faintly indicated uses" that familiar words carry in use (PPF: § 35). In the Last Writings, it is analogously defined as "a picture of a word's use" (LW II: 39). To sum up Wittgenstein's scattered employment of the term, we can say that a word acquires an atmosphere -a sort of psychological trace the word carries in use -once we get accustomed to applying the word in the multifaceted contexts of our life and culture. The atmosphere can also be defined as a felt experience of the meaning of our words, which is rooted in the broader context of our ordinary life and language use. What is more, such an experience can only be referred to "by repeating the original expression" through which we convey it in the first place (ter Hark 2011: 516).
By saying that e is yellow, then, I am giving expression to the peculiar symbolism the vowel has incorporated in my own life, by associating it with a certain colour (the same goes for the weekdays example). Such a symbolism might be naturally caused by the vowel's sound, the role of the colour yellow in our life, certain associations to our cultural habits, or whatever. As the following remark points out, the experience can be expressed only through these words, just as we can express pain through natural pain reactions: Would it be more correct to say that yellow 'corresponds' to e than e is yellow'? Isn't the point of the game precisely that we express ourselves by saying e is yellow? Indeed, if someone were inclined to say that e 'corresponds' to yellow and not that it is yellow, wouldn't he be almost as different from the other as someone for whom vowels and colours are not connected? (LW I: § 59) We get struck by the strangeness of the expression 'e is yellow. ' We "take it the wrong way," and think that it should be reformulated, as it cannot be that we are really attributing a colour to a sound (sounds cannot be colourful, after all!). So, we might suggest that what we really mean is only that the vowel 'corresponds' to the colour.
However, here lies the double mistake. First, the verb 'to correspond' is likewise used secondarily (the quotation marks are there to stress this aspect). We would not really be getting further, as we would only swap one controversial word with another one, equally controversial. Second, it is the point of the game that we use the verb 'to be' here. If we replace the words that we use to express what we mean, with 'correspond,' or with the expression 'is like' because we think that we deal with a simile here, we lose it, we do not play that game, we simply do not convey what we want to express. We would not be expressing the atmosphere of the words. If somebody is really inclined to talk about correspondence and not about being here, their case would be similar to that of a person who simply does not get the kind of connection we envisage between vowels and sounds at all, because they have a completely different experience of those very words.
The expressive function of secondary expressions is less controversial if we think of another, less idiosyncratic, example: 4 The feeling of the unreality of one's surroundings […] Everything seems somehow not real; but not as if one saw things unclear or blurred; everything looks quite as usual. And how do I know that another has felt what I have? Because he uses the same words as I find appropriate. But why do I choose precisely the word "unreality" to express it? Surely not because of its sound. (A word of very like sound but different meaning would not do.) I choose it because of its meaning. But I surely did not learn to use the word to mean: a feeling. No; but I learned to use it with a particular meaning and now I use it spontaneously like this. One might saythough it may mislead-: When I have learnt the word in its ordinary meaning, then I choose that meaning as a simile for my feeling. But of course what is in question here is not a simile, not a comparison of the feeling with something else. (RPP I: § 125) We can talk about 'a sense of unreality' when everything feels unreal. This expression is usually employed to report a particular feeling of detachment 5 and alienation, and it 4 There are other examples that stress the connection between secondariness and expressiveness from the Nachlass. LW I: § 69-73, for instance, examine the forms of expression that we employ to describe the character of proper names as secondary uses. For instance, in the sentence 'the name "Schubert" fits Schubert's work' the verb 'to fit', is used secondarily. Again, Wittgenstein points out that we do not really describe Schubert or its music by using these expressions. We rather formulate "a pathological statement about the speaker" (LW I: § 73), as we reveal something about our own cultural world (our musical taste, beliefs, and opinions concerning Schubert's music).
characterizes certain psychological disorders. Notably, this expression is used secondarily, as 'unreality' is not learnt to mean a feeling, and we would not mention the sense of unreality to explain its meaning. The use is in a sense figurative, but not metaphorical; it does not constitute a simile either, as no comparison is really being drawn. We learn the word in its primary meaning, and we use it in a new way, we apply it in a new context in which the point is to express a particular feeling. The word can be seen as "the bearer of another technique" (RPP I: § 126), that is borrowed or co-opted from another language game. This extension of use is spontaneous -no one teaches it to us -and is based on the primary meaning of the words in question.
Wittgenstein says that this feeling can be conveyed to others, as whoever talks about a feeling of unreality uses the same words secondarily as other members of the same community would. There is no other criterion for understanding apart from our tendency to use the same forms of expression in the same circumstances. As ter Hark suggestively frames it, the person will know what we are talking about because we are "in tune with the same expression" (2011: 516). Strikingly, what is in common between people using the same words secondarily is not only a cluster of rules (the conventions through which we learn the primary meanings of the words involved). What is common is that we use the same forms of expression to convey the same experience. We use the same expressions because we are accidentally attuned with each other in the same community of experience. The same can be said about the coloured vowels and the fat Thursdays.
Finally, it is worth mentioning that Wittgenstein talks about experience and secondariness also in the case of telling our dreams. We have already encountered dreams in RPP II: § 574, where the expression 'in my dream I knew that…' was paired up with the sentence 'e is yellow'. The most explicit passage about dreams is the following, where, once again, drams are mentioned in connection to experience (in this case, the feeling of unreality): Now I am not using the word (meaning) for something else; rather, I am using it in a different situation [just as I am not using 'know' to refer to two different things when I say 'In my dream I knew.' CF. also: feeling of unreality]. (LW I: § 57) When we report the content of a dream, it can happen that we say that in the dream we knew something. This 'knowledge' is different from the knowable information we share in the every-day, insofar as it entertains a different relationship with the dreamer: when we say that, in a dream, we know something, we are not really committed to it, we do not really believe that we knew anything. Furthermore, and more importantly, what we know in a dream usually does not require the same level of epistemic warranty: we simply know things, even though we have no reasons or grounds to believe so. We do not learn to use the verb 'to know' this way, yet we use it: we co-opt the word for other means. The same thing can be said of other analogous forms of self-attribution, such as They are compared to dreams in the following way: Is a dream a hallucination? -The memory of a dream is like the memory of a hallucination, or rather: like the memory of a real experience. This means that sometimes you would like to say: "I just saw this and that", as if you really had just seen it. (LW I: § 965) The memory of a hallucination is similar to the memory of real experience and to that of dreams. We can say, for instance, that we saw something we later found out not to be real, yet we use the verb 'to see' even if we did not really see anything. The verb is here used secondarily, as much as the verb 'to know' when reporting a dream: a natural extension of a word used in a new context, to play a different game.
The cases of dreams and hallucinations are more complex and ambivalent than the other cases of expressive uses of language. For sure, we are neither expressing ourselves when talking about our dreams, nor are we talking about how we feel, as in the case of the feeling of unreality. It is true, however, that while telling a dream we give substance to an experience we had, rather than describing an actual thing in the world, or a certain knowledge. Whoever does not dream would not understand our practice of recalling of a dream, very much like the case of a speaker who does not understand what we are getting at by saying that e is yellow. In a community where no one has ever hallucinated, it is likely that we could not be understood when describing a mirage either.

3.Beyond Expression: Description and Absurdism
Now, if we focus only on the examples of vowels and weekdays, or the one about the sense of unreality, as ter Hark does in his paper, we might be tempted to conclude that secondary uses of words have only an expressive function. 6 This, however, would be a mistake. Let us think about the case of teaching music and art or giving instructions on how to play a song. The famous jazz musician Wayne Shorter was once reported to have said to one of his band members, Danilo Perez, to 'put more water in the chords'. 7 This expression is secondary, as it is logically dependent on the primary meaning of its words, we cannot use it as an example to elucidate the meaning of 'water' and it is not a metaphor either, because it is not paraphrasable. The sentence presupposes a certain understanding on how to play music and, if we want, even a certain experience of music 6 This is because ter Hark's main goal is to clarify Wittgenstein's employment of the concept of experience: "Other concepts […] e.g., 'secondary use', 'illusion', 'inclination', and 'primitive reaction', turn out to be part of one and the same conceptual survey of meaning-experience" (2011: 502). We find a similar, even stronger, claim in Gilead Bar-Elli, as he claims that "the phenomenon of using words in their secondary sense depends on the experience of meaning" (2006: 2043). Differently from ter Hark, however, Bar-Elli envisages in Wittgenstein a theory of meaning as experience and interprets secondary senses as an epiphenomenon of the experience that every word supposedly bears in use. As I will show in the following paragraph, this general approach to secondary uses is partial, as it implicitly denies that words can be employed secondarily without an experience being expressed or involved. 7 As attested in Shorter's biography (Mercer 2007: 302). and sound. 8 However, it is neither a sentence through which Shorter expressed anything about himself, as in the case of the feeling of unreality, nor it is a sentence to express the atmosphere of our words; he was rather giving an instruction about how to play a song through an implicit description of the sound he wanted to achieve. Descriptions, Wittgenstein points out, are "instruments for particular uses" (PI: § 291). There is no univocal model of what a description is, it varies from case to case and depends on what we want to do with it. As such, I can describe a sound as watery if I want to convey a certain information, even though I feel as if I cannot use any other form of expression to do it. In this case, I would use language secondarily to provide a description.
Other valuable examples of descriptive secondariness come from poetic language. It is the case, for instance, of poetic synaesthesia. Expressions like 'soft silence,' 'black scream,' and 'silver voice,' are conjured because of their evocative power and poetical effectiveness, yet they can be used to describe actual features in the world. It might be suitable to call a scream 'black,' for instance, when it is loaded with dread, bereavement, and despair. 9 In this case, using the synaesthesia can be considered to be the most natural, appropriate, even accurate way to provide a description. Aesthetic qualities, in general, usually require secondary employments of words to be described: paintings that are dynamic, melodies that are plaintive, wines that are round if tasted, and so forth. 10 Besides, in the Investigations, there is another important instance of secondariness to consider, that can hardly be explained as expressive or descriptive. Amidst the private language argument, while addressing the privacy of pain and its relationship with pain behaviour, Wittgenstein addresses the objection that we can talk about pots and their feelings in a fairy tale, even though there is no pain behaviour imaginable accompanying pain in this case. This is how Wittgenstein addresses this issue: We do indeed say of an inanimate thing that it is in pain: when playing with dolls, for example. But this use of the concept of pain is a secondary one. Imagine a case in which people said only of inanimate things that they are in pain; pitied only dolls! (When children play trains, their game is connected with their acquaintance with trains. It would nevertheless be possible for the children of a tribe unacquainted with trains to learn this game from others, and to play it without knowing that it was imitating anything. One could say that the game did not make the same kind of sense to them as to us). (PI: § 282) Here, the notion of secondary use is employed to neutralize the objection that, insofar as we can attribute pain to inanimate objects, pain is somehow logically independent from pain behaviour. This conclusion can be easily blocked if we carefully distinguish the different uses of the word 'pain' in this case. It is true that we can attribute pain to dolls -this is what children frequently do while playing with their toys -but this use of the word is secondary, and nothing about the nature of pain follows from it, just as nothing about the nature of weekdays can be assessed if we say that Thursday is fat or thin.
It is easy to see why we deal with a case of secondariness. First, we do not point at dolls to explain what pain is (at most, we can point at a certain expression in a doll face as an example of a pain expression). Second, this use is logically dependent on another one, as we need to begin by learning the word 'pain' in its connections with pain behaviour. Wittgenstein underlines this kind of logical dependency, so typical of secondariness, when he asks us to imagine a group of people that pity only dolls. We can certainly imagine something like that; only, the word would have a different sense for them, as it would not share the same connections with our life as our word does (analogously, if the word 'yellow' were used only to talk about vowels, its meaning would be quite different). The case is comparable to that of children playing with toy trains: in the case of a tribe of children playing with trains but ignoring their connection with real trains, the game would not have the same sense -the same role -in their life: it would not present the same connections with its context, the things children would say about it would be different, the way the game is played would diverge in significant ways, and so forth. 11 Third, this kind of use is not metaphorical, as there is no other way to say that a doll is in pain in the context of a children's game.
Notably, it is clear that children neither describe anything while playing, nor express their own feelings. They just play a game of pretence, where dolls cry and suffer because they are taken to be real people and represent living bodies. Even though it consists of a spontaneous and natural extension of a word's use in a new context, this case of secondariness is not forced on us in the same way the coloured vowels and the other expressive uses are. The attribution of pain is in fact markedly stipulative: children expand the use of words by establishing a new instance of use in the context of a certain game. Thus, here secondariness acquires a further dimension that is not reducible to the other examples already described, a feature that is however fully captured by the three criteria for secondary uses. The general picture of secondariness becomes even more complex. Now, as the case of the attribution of pain to dolls is introduced as a term of comparison to understand why we can say that pots talk and feel, we can make a step beyond Wittgenstein's text, and suggest that secondariness is also involved when we tell fantastic and absurdist tales. Pots do not talk and feel, we cannot even imagine what the behaviour of a pot in pain would be like. Where does a pot have a mouth, how could it express its pain? Yet when we tell a story -within the specific language game of telling a fiction -a sentence like 'the pot shrieked in pain' is perfectly fine: it does what it is required to do in the context of telling a tale. Words are spontaneously projected into a new context of use and combined to represent absurd and unimaginable situations.
11 Oswald Hanfling argues that the case of toy trains is an example of a weaker form of secondariness, as we can play toy trains without being acquainted with real trains and thus knowing the primary use of the word (1990: 127). However, this is misleading. The fact that the whole sense of the game would change means nothing but the fact that we would play a different game.
The example that Wittgenstein makes about talking and moaning pots might be silly, yet it should not be overlooked, as it gives us the chance to understand the role of secondariness when it comes to understanding certain employments of words in literature. Let us think about Franz Kafka's Metamorphosis, an eerie and absurdist novella about a man, Gregor Samsa, inexplicably transformed into a monstrous cockroach. This example is not accidental, as it struck the attention of Oxonian philosophers in their disputes about attributions of identity, as reported in Isaiah Berlin's memories: 12 The principal example of the latter [the problem of identity] that we chose was the hero of Kafka's story Metamorphosis, a commercial traveler called Gregor Samsa, who wakes one morning to find that he has been transformed into a monstrous cockroach, although he retains clear memories of his life as an ordinary human being. Are we to speak of him as a man with a body of a cockroach, or as a cockroach with the memories and consciousness of a man? 'Neither' Austin declared. 'In such cases we should not know what to say. This is when we say "words fail us" and mean this. We should need new words. The old ones just would not fit. ' (1973: 11) Kafka depicts an impossible scenario -a man wakes up and finds himself to be transformed into a cockroach -and in order to do that he employs old words in a new context of use: the word 'cockroach' is employed to describe a human being. This description is neither figurative nor derogatory; it is secondary.
In the story, Samsa is a cockroach. We cannot make sense of this strong identity. We may in fact wonder: what does it mean for a human being to be an insect? How can a cockroach be angry, suffering, thoughtful, as Samsa as a human being can be? Yet, in the context of storytelling, all these questions are pointless, just as it is pointless to question whether dolls are in pain or not when a child plays with them. Most importantly, we cannot weaken the identity -by saying that Samsa is rather a man in the body of a cockroach, for instance (a formula that is also committed to a certain dualistic preconception of the mind-body problem: this is what Berlin and Austin were interested to investigate) -without losing the sense of Kafka's tale, as much as we lose what we want to express with 'e is yellow' if we rather say that e is like yellow or corresponds to yellow. The strong identity is just a more appropriate form of expression. Kafka's Metamorphosis can then be seen as an exercise in secondary language use: within a certain narrative, words can be employed in a new context, their use is extended to convey a certain meaning, a meaning that is bound to the specific combination of words we employ to express it. It is not, as Austin suggested, that we lack words to explain Samsa's condition, that words fail us, and thus we should invent a new vocabulary to make sense of it. The point is exactly the opposite: to express Samsa's condition we can use only those words as Kafka uses them.

4.A Philosophical Distinction
Considering the wide application of the notion of secondary uses in Wittgenstein's work, and its further fruitful applications beyond his texts, it is hard to see how we can reduce the plurality of examples we have discussed so far to a single unifying principle, such as a univocal function that secondariness is supposed to fulfil. We can use words secondarily to express ourselves, to convey a certain experience, or feeling, and to give instructions on how to play an instrument. Sometimes, secondary uses have a descriptive function, sometimes they are stipulative, as in the case of pain attribution to dolls, or in fiction.
The reasons why we can provide so many different instances of secondary uses in Wittgenstein's work, is that the distinction is meant to be primarily philosophical, that is, it is designed to be a helpful tool to distinguish different shades of word use, and thus clarify those areas of language that confuse us and lead us astray. As the primary aim of Wittgenstein's philosophy is clarification through the dissolution of philosophical problems, then, the distinction is markedly problem-relative, that is, it acquires a certain meaning in relation to the problem it is designed to solve. Accordingly, a single unified account of secondariness is in principle impossible.
Let us examine how Wittgenstein appeals to secondary uses to clarify language and tackle a variety of different problems. First, it should be noticed that Wittgenstein mentions the coloured vowels for the first time in the context of an elaborated discussion on commonality (BB: 138-139). The example is introduced to tackle a certain preconception, according to which every application of a word needs to have something in common with all the others in order to be used legitimately . 13 We can define 'blue' to be the colour that all its specific shades have in common, but this does not necessarily imply that blue is a thing that can be pointed at and recognized before learning and applying the word. One conclusion, reported in the Brown Book, is that, when we talk about blue as the thing common to all its shades, we are merely saying that we use the word in in all those cases, and nothing more: we are certainly not committed to assuming a common thing that we can point at which all the shades share (BB: 135).
In other words, there is no commonality that can be first acknowledged and then appealed to as a reason for our use of the word. There is nothing beyond use that works as a justification for it: we simply apply the same word to all these shades of colour.
Notably, ordinary descriptive sentences like 'the shades A and B are both blue' and expressions like 'e is yellow' are akin, as in both cases we employ a colour word without having a reason to do so, that is, without having a clear perception of a commonality that could justify the new use in the new context. 14 In this case, then, the sentence 'the vowel e is yellow' is useful for proving a general point about the reasons we provide for using words as we do, and dissolve a certain picture of commonality as a required condition for use.
In the case of PPF: § 274-278 -those remarks where the distinction is clearly laid out -a careful analysis of their context helps us understand that Wittgenstein's aim was to clarify our ordinary employment of the word 'meaning' to talk about the way we experience the meaning of words, as when we say that a word is 'loaded' with meaning.
Especially while reading a poem loud, Wittgenstein points out, words acquire a special meaning, a different ring (PPF: § 264) that we feel. Now, this use of the word 'meaning' is apparently problematic, as it does not refer to the use of the words in question, rather to an experience or a feeling, and we know that use for Wittgenstein defines the meaning of meaning (PI: § 42). Is, then, the word "meaning" polysemous, or ambiguous? Should 13 The gist of this discussion is also present in the Investigations, where Wittgenstein lists different examples of our use of the expression "to see what is in common" (PI: § 72), in the same sections addressing family resemblance. 14 Those interpretations that aim to account for the attribution of aesthetic qualities to works of art through secondary uses, emphasise this aspect of secondariness, as it allows us to avoid any appeal to commonality or resemblance to justify why we use words as we do when describing works of art (see, in particular, Tilghman 1984 andBudd 2006). The employment of secondariness to tackle certain specific aesthetic problems is thus consistent with the general idea that the distinction between primary and secondary uses is a useful philosophical tool to solve specific confusions.
we dismiss the definition of meaning as use because sometimes the word seems to be attributed to a feeling of a sort? Wittgenstein implicitly rejects all these questions, in the following way: But the question then remains why, in connection with this game of experiencing a word, we also speak of 'the meaning' and of 'meaning it'. -It is a characteristic feature of this language-game that in this situation we use the expression 'We pronounced the word with this meaning' and take this expression over from that other language-game. (PPF: § 273) Wittgenstein is here rephrasing the criteria for secondariness through the vocabulary of language-games. Much as in the case of secondariness, we use the word 'meaning' in virtue of its meaning to give expression to an experience; this new use is logically dependent from its primary meaning, as we must borrow the expression from a language-game to another, an expressive one. The word meaning is neither polysemous nor ambiguous, as secondary uses do not constitute a new meaning, and we can maintain the definition of meaning as use, as it is because of it that we can use the word 'meaning' secondarily. As a further proof, the remarks that immediately follow PPF: § 273 are those that introduce the distinction between primary and secondary uses: further examples are provided to strengthen the idea that in language these cases of use are more frequent than we might expect.
We might wonder, however, what is the problem Wittgenstein aims to tackle by appealing to secondariness here. The answer is the following: through secondariness, we can give a perspicuous description of a certain employment of the word 'meaning' -the one through which we refer to the experience of words -that could lead us to relapse into some form of mentalism, that is, the theory of meaning according to which meaning is an inner or private experience of sorts that accompanies the words in use. 15 If we do not carefully distinguish the primary use of the word 'meaning' from its secondary use, we might be tempted to take the secondary use of the word as an actual proof that meaning is something mental accompanying our words. It is not so: we talk about 15 Mentalism is also the philosophical target of other two instances of secondary uses of words we did not mention, the case where we say that we 'calculate in the head' (PPF: § 279, LW I: § 801, 802, 804) and read silently (LW I: § 803). In both cases, to stress that words are used secondarily is helpful to avoid the temptation to claim that calculating in the head or reading silently actually refer to an inner, hidden, process in our mind. meaning as an experience in those cases where we are giving expression to the feelings a certain word evokes in us, for instance in a poem, and we can do so by employing words secondarily.
Given that Wittgenstein's remarks are scattered and sketchy, we cannot easily point out what kind of philosophical problem the case of dreams is related to. Arguably, however, it can be seen as an important case study to tackle the premise of Cartesianism, that is, the assumption that what we know in a dream and what we know in real life are indistinguishable, so that philosophy acquires the task of grounding our knowledge and proving that we do not live in a dream of sorts. On the contrary, if we stress that our talk of knowledge, while reporting a dream, is secondary, then we are less tempted to assume that Cartesian doubt is legitimate. It is legitimate, only if we conflate two distinguishable uses of the same verb, 'to know.' Analogously, we can see the philosophical importance of distinguishing between primary and secondary uses when it comes to understanding the logic of our discourse about hallucinations: if we stress that in this case words are used secondarily, we are less tempted to consider the report of an hallucination as the same as the description of a state of affairs or perceptual reports. If so, then we can target scepticism regarding perception at its core.
The case of telling fantastic stories was originally introduced in the private language argument to tackle the idea that we can detach pain from pain behaviour. Indeed, we can attribute pain to dolls and pots while telling stories. Only, this is a secondary use that does not reveal anything significant about the primary uses of our pain vocabulary (just as Kafka's cockroach does not reshape our zoological taxonomy). This vocabulary is learnt in the broader context of our life, in close contact with the pain expressions and behaviour of the other members of our linguistic community. Without it, there would be no pain vocabulary as we know it. Thus, we can point out that pain attribution to dolls is secondary, and so we neutralize an observation that is meant to back up the idea that our concept of pain refers to an inner thing that is logically independent from our behaviour.
Beyond Wittgenstein's examples and actual remarks, the case of giving instructions in how to play music might be helpful to clarify our concepts of musical understanding and musical explanation. To ask somebody 'to put more water in the chords' is a form of expression that Shorter used to feel as the only appropriate one to convey a certain idea on how to play. The instruction works as a tool to lead somebody to understand how the piece should be played, much as we do when we lead others to grasp the meaning of a poem (PI: § 533). As in the case of the experience of meaning, to distinguish different cases of what we call 'explanation' and 'understanding' in language might be useful in order to avoid any temptation to conflate different cases of understanding, and thus relapse into any mentalist preconception, according to which understanding always requires a mental act, or an experience. It can do so in the case of understanding music, but only through a secondary usage of language.