domingo, 9 de abril de 2017

# WITTGENSTEIN ON MEANING

Corrected last draft. The text will be published in the book Philosophical Semantics by CSP in 2017.





- III -
WITTGENSTEINIAN SEMANTICS

Philosophers constantly see the method of science before their eyes and are irresistibly tempted to ask and answer questions the way science does. This tendency is the real source of metaphysics, and leads the philosopher into complete darkness.
Wittgenstein

The philosopher does not belong to any community of ideas – this is what makes him a philosopher.
Wittgenstein

My aim in this chapter is not so much to interpret Wittgenstein, as to reconstruct and sometimes develop his insights on meaning in a way that makes them more coherent and hopefully more convincing than they may seem at first sight. What I search for here is what in his own terminology could be called a surveillable representation (übersichtliche Darstellung) of the grammar of the concept-word ‘meaning,’ particularly in regard to representative language. Before beginning, I would like to offer my views on something we could call the ‘semantic-cognitive link.’

1. Semantic-cognitive link
In this book, I support the most common viewpoint concerning the referential mechanism, according to which referential expressions can only refer because of some intermediary link able to associate them with their reference. This view originated in classical antiquity. In my understanding, this link has a semantic-cognitive nature, in the sense that it can always be considered from two contrasting perspectives: semantic and cognitive.[1] From a cognitive or psychological perspective, the link is usually called an idea, representation, intention, conception, thought, belief and cognition (Aristotle and Locke are models of semanticists who have adopted this perspective). From a semantic perspective, the link is more often called sense, meaning, use, application, intention, connotation, concept, informative content, belief-content, proposition, criterion, criterial or verificational rule (the Stoics, Frege and Husserl are models of semanticists of this last persuasion). This can be clarified with a diagram:

LANGUAGE
(a) COGNITIVE LINK:
idea, representation, cognition, intention, conception, thought...
(b) SEMANTIC LINK:
sense, meaning, content of thought, intention, use, application, semantic rule, criterial rule, criteria, proposition…
WORLD

At this point, an old question arises: What is the appropriate link? Which set of terms should be excluded? Should we exclude psychological terms, so as not to contaminate semantics with empirical contingency? Or should we abandon a possible commitment to questionable abstract semantic entities, exchanging them for the more feasible concreteness of the psychological, the only really able to justify mental causality? Should we read an ambiguous work like the Critique of Pure Reason from a semantic or from a psychological perspective?
   Many philosophers have dealt with this problem by assuming that each alternative excludes the other. In my view, this is the real mistake. I see this assumption as a false dilemma generating useless philosophical confusion. The psychological and semantic perspectives should be seen not as mutually exclusive alternatives, but as complementary and at the bottom inseparable ones.
   The source of the illusion that these two perspectives are irreconcilable lies in the fact that the abstract character of the semantic perspective seems to be committed to some form of realism (Platonist or in a sense Aristotelian) about universals. In contrast, the cognitivist perspective seems committed to some kind of nominalism or particularism attached to the contingency of the psychological subject. Since these ontological commitments are incompatible, the two alternatives also seem incompatible.
   However, when we perceive that these ontological commitments are avoidable, it becomes easy to conclude that the intermediary link between words and things can be dealt with in these two apparently contradictory ways without a real conflict. We must only realize that when we consider the intermediate link from a semantic perspective, we are not committed to the appeal to really abstract entities. What we are doing is leaving out of consideration the inescapable fact that meaning can only exist insofar as it is spatio-temporally embodied in some specific psychological subject.
   To clarify the complementarity that I am suggesting: We can consider the intermediate link as both: (i) a cognitive link, consisting of semantic elements that must be spatio-temporally realized as ephemeral intentional acts realized in specific psychological individuals; (ii) a semantic link. This is referred to as something considered in abstraction from its spatio-temporal instantiations as intentional acts going on in some specific psychological subject in a specific time and space, but not in abstraction from any spatio-temporal instantiation as an intentional act going on in any particular psychological subject. So considered, the semantic link can be distributed among an undetermined number of unconsidered cognitive subjects, which does not make it de-psychologized or disembodied.
   In other words: the proposed abstraction cannot be made in a sense where the semantic link is considered as somehow transcending the realm of specific psychological and physical subjects, since it always needs some form of cognitive spatio-temporal intentional embodiment to be object of consideration. In fact, the word ‘abstraction’ means here simply ‘leaving out of consideration the natural association between a meaning and this or that specific psychophysical individual who instantiates the meaning,’ and focusing on the signs that can convey this meaning, insofar as they can be understood by some other psychological interpreter. This is the only way to make a semantic-cognitive link semantically independent of ephemeral cognitive instantiations.
   A very simple example illustrates my point. When I recognize a patch of vermilion of cinnabar (a precisely characterized shade of color), it is because the patch I see matches the memory image of vermilion that I have stored in my long-term memory from earlier experience. Now, when I speak of a general concept of vermilion of cinnabar, I can be speaking not only of this image, that may be made conscious in my mind, but also of any other qualitatively identical[2] image of this color that may be made conscious in any other mind.[3]
   In other words, against the idea that our semantic link is a type that is a unique abstract Platonic entity, what I am proposing is that we conceive the semantic link in the sense of an arbitrarily chosen model, ideally, as any token that stands for any other token that is qualitatively identical to it.[4] In short, we can define a semantic link as:

A semantic link X = an occurrence of X arbitrarily chosen to serve as a model for any other occurrence of some X that is qualitatively identical to the model.

Since all these possible occurrences need to be psychological (and certainly also physical), we don’t need to transcend the domain of the psycho-physical in order to reach the semantic domain. Moreover, we do not need to have an instantiation of the semantic type in any specially chosen psychological particular. What we really need is that at least one psychological particular, no matter which, should embody the semantic cognition. But this condition, as we will see later, can easily be accommodated in our commonsense ontological framework of particularized properties.
   This compromise solution is strengthened if we note that even some sub-items of (a) and (b) show an approximate correspondence to each other. Thus:

(i)                    the psychological word ‘idea’ has meaning proximity to the semantic words ‘sense’ or ‘meaning,’ as well as to ‘concept’;
(ii)                  the psychological word ‘representation’ has some meaning proximity to the semantic phrase ‘criterial rule’;
(iii)            the psychological word ‘mental image’ has meaning proximity to the semantic phrase ‘criterial configuration’;
(iv)            the psychological word ‘belief’ has meaning proximity to the semantic phrase ‘belief-content.’
(v)                  the psychological phrase ‘occurrence of thought’ has meaning proximity with the semantic terms ‘proposition’ and ‘content of thought.’

2. Why reference cannot be meaning
When we consider the semantic link, the words that more easily come to mind are ‘sense’ and ‘meaning’ (generally used as synonyms) understood as semantic or informational content. But what is sense or meaning? Perhaps the simplest answer is what may be called semantic referentialism, a doctrine that in its crudest form holds that the meaning of a linguistic expression is its own reference. This conception either denies the existence of a semantic link between word and object or minimizes its importance. Wittgenstein described this way of understanding meaning at the beginning of his Philosophical Investigations, where he commented on the so-called ‘Augustinian conception of language’:

These words, it seems to me, give us a particular picture of the essence of human language. It is this: individual words in language name objects – sentences are combinations of such names. – In this picture of language we find the roots of the following idea: Every word has a meaning. This meaning is correlated with the word. It is the object for which the word stands. (Wittgenstein 1984c, part I, sec. 1)

Wittgenstein’s aim in this passage was to object to semantic referentialism, a theory championed by him in his first and only published book, the Tractatus Logico-Philosophicus. According to his version of semantic referentialism, when completely analyzed, language proves to be composed of atomic propositions constituted by atomic names whose meanings would be the simple and indestructible objects necessarily referred to by them.[5]
   Semantic referentialism is not devoid of intuitive appeal. After all, it is usual to explain the meaning of a concrete substantive by pointing to objects that exemplify what it means. In our childhood, we learned what the word ‘chair’ means because adults showed us examples of this artifact. And we learn the name of a particular person when this person introduces herself to us with her name. Moreover, we learn what a word means or does not mean respectively through positive and negative examples of its application. All this makes credible the idea that meaning may be the object actually referred to. This view has at least a nearly palpable simplicity: ‘here is the name, there is its meaning.’[6]
   However, there are strong well-known arguments against this naive view of meaning. The most obvious is that you cannot predicate of a meaning what you predicate of an object: if a pickpocket steals your wallet, you do not say that the meaning of your wallet was stolen, and if you say that Abraham Lincoln was assassinated you do not intend to say that the meaning of his name was assassinated.
   Another argument is that many natural terms have the same reference, while their senses or meanings are obviously different: the singular terms ‘Socrates’ and ‘the husband of Xantippe’ point to the same person, although they clearly have different meanings. And the opposite seems to be the case with general terms: the predicate ‘...is fast’ in the statement ‘Bucephalus is fast’ allegedly refers to a particularized property of Alexander’s horse Bucephalus; and the same predicate ‘…is fast’ in the sentence ‘Silver is fast’ allegedly refers to a particularized property of another horse, Silver. Although the speed of Bucephalus is numerically different from the speed of Silver, in both sentences the word ‘fast’ preserves precisely the same meaning.
   The most decisive argument against the referentialist view of meaning, however, is the most obvious: it concerns the fact that even when a referential expression has no reference, it does not lose its meaning. The singular term ‘Eldorado’ and the general term ‘phlogiston’ do not have any reference, but by no means do they lack a meaning.
   For a long time, semantic referentialism has been criticized by ordinary language philosophers as based on a primitive and misleading understanding of mechanisms of reference. As Searle noted, it is ‘a good illustration of the original sin of all metaphysics, the attempt to read real or alleged features of language into the real world’ (1969: 164). He saw in its adoption a major source of pseudo-problems, although some proponents of the metaphysics of reference have more recently tried to reassert it (e.g. Salmon 1993).

3. Failure of Russell’s atomistic referentialism
Well aware of difficulties like those presented above, Bertrand Russell tried to defend semantic referentialism in a minimalist fashion, taking into account only alleged atomic elements of language and the world. It is instructive to consider his attempt. For Russell, the meaning of at least some terms – called by him logically proper names – would have their objects of reference serving as their proper meanings. This could be the case, perhaps, with the word ‘red’. After all, as he noted, a blind man is unable to learn the meaning of red, since he is unable to see the color (Russell 1994: 194-5: 201-2).[7]
    However, it is untenable that the meaning of any word can be given by its reference tout court. Changing his example a little, suppose that someone demonstratively applies the word ‘vermilion’ to an occurrence of vermilion of cinnabar, which is a shade of red that the human eye practically cannot further subdivide. (In this sense, it is a simpler candidate for ‘simple’ than Russell’s red color, since it does not need to include gradations and has specified limits.) Could such an occurrence be the meaning? There is an obvious reason to think that an occurrence of vermilion could not be its meaning: the absence of any identity criteria. When we consider the occurrence of vermilion – be it physically thought of as an externally given spatio-temporal property, or phenomenally thought of as an appearance, a sense-datumthe occurrence will always be different for each new experience. But if the meaning of ‘vermilion’ is nothing but a detected occurrence irrespectively of its relation to the others, then each new occurrence of vermilion should be a new and distinct meaning – an intolerable conclusion!
   Russell must have understood this problem, for he found a way to defend his view against such objections. But, as we will see, it was at the cost of becoming entangled in even worse difficulties. He suggested that the object-meaning of a logically proper name would be something immediately accessible – such as sense data picked out by pronouns like ‘this’ or ‘that’ – only as long as we keep these sense data present in consciousness. This means that the meaning also lasts only as long as our personal experience of a word’s object of application! (Russell 1994: 201, 203). But this is a desperate answer, as it is clear that it leads to solipsism.[8] For what criteria of correction could we apply to fix this meaning, in order to know in what cases the word ‘vermilion’ can be reapplied to another occurrence of the sense data that would at least qualitatively be the same sense data? Moreover, how could we insert this fugitive meaning of a proper name in our common language – a language composed of words whose meanings are permanently shared by their speakers?
   Indeed, in our language, to know the meaning of a word like ‘vermilion’ presupposes at least the ability to recognize an occurrence of vermilion as being precisely similar to other occurrences of vermilion. But this recognition is not included in the idea that the meaning of the word is nothing but the occurrence of its reference. The concept of a word’s meaning requires essentially that we should be able to unify its different applications to the same referent, which is not possible here.
   It is true that if the meaning of a word like ‘vermilion’ were the vermilion-type – understood as an abstract entity common to all occurrences (tokens) – we would be able to solve the difficulty pointed out above. But this solution might commit us to accepting some form of Platonic-realism, raising justified suspicion of an unintelligible reification of the type in a topos atopos. Alternatively, one would need to consider the vermilion-type as being a certain class of occurrences of sense data that are precisely similar to each other. This lowers the danger of realism, but does not eliminate it, since classes are often seen as abstract entities, and if they are not, then they need some limiting intention. In addition, classes may be larger or smaller depending on how many members they have, while the meaning of the word ‘vermilion’ has no proper magnitude, neither increasing nor decreasing.
   The most feasible alternative seems to be that we consider the meaning of ‘vermilion’ as some occurrence of vermilion that we are using as a model. This could be either a sense datum or some particularized property in the outside world, able if necessary to be arbitrarily replaced by another like it or any other occurrence that is precisely similar to this model. So, if I recognize what is currently being offered as an occurrence of vermilion, it may be because I realize that this occurrence is qualitatively similar to others that were previously given to me as being those of vermilion. This resorts to a model whose copy I have stored in my memory, giving me an awareness of it as a color qualitatively identical to colors I have previously experienced. Thus, recalling the various experienced occurrences of vermilion {V1, V2... Vn} and the model-copy Vm that I have stored in my memory, I can say that {V1 = Vm, V2 = Vm... Vn = Vm} and, therefore, that {V1 = Vm = V2}, etc., without resorting to any Platonic entity or to any multiplication of identities of identities or even to the concept of an intensionally defined set – problems often thought to burden particularistic strategies for handling universals.
   What this view amounts to is that what we call the meaning of the word ‘vermilion’ must be identified with a referential connection. This is namely with a rule that relates cognitive experiences of occurrences of a color to occurrences of color that we in some way use as models, in order to produce an awareness of what is experienced as being a precisely similar vermilion color in each case. Moreover, this internal semantic cognition is produced in association with ‘vermilion,’ a word for an entity. In this way, both a reference and its word turn out to be in principle interpersonally accessible, once the precise similarity between occurrences allows for interpersonal accessibility and for the implicit agreement necessary to create a linguistic convention, even if in itself the semantic cognition, as a matter of fact, is not interpersonally accessible.[9] We should also point out that the semantic rule that uses memorized models to identify any new instance of vermilion is independent of this or that particular occurrence of vermilion, for it only relates to instantiations of possible occurrences that can satisfy it. This view is the one I believe to be workable.
   But this view has a price: we see on reflection that by adopting it we have already left behind us the referentialist conception of meaning. Even to establish a meaning as simple as that expressed by the word ‘vermilion of cinnabar’ we must appeal to something that is more than a rough object of reference and is independent of it, namely, a semantic rule.
   Even if Russell’s semantic referentialism is unsustainable, there is a lesson to be learned from discussing it. Our last suggestion salvages an important idea derived from his semantic referentialism, namely, the idea that the existence of an object of reference is necessary for the names of objects taken as simple in the context of a linguistic praxis. We understand the phrase ‘simple object’ as inevitably having a non-absolute sense (Wittgenstein 1984c, I, sec. 45-48) and restrict it to a non-decomposable entity in the framework of some linguistic practice. This seems to be the case with the sense datum of vermillion of cinnabar as a trope (a spatio-temporally particularized property that may be given to experience). Anyway, the positive conclusion is that for such ‘simple names’ to acquire meaning they need to have reference.
   This is why, in an important sense, a blind man cannot learn the meaning of the word ‘red.’ Since the color red is in a sense simple[10] and knowledge of it is based on acquaintance, and since the blind cannot have this sensory experience, a blind man cannot learn and apply the conventional criterial rule for the shared referential meaning of the word ‘red.’ At least in the case of this subrogate of a logically proper name restricted to a certain linguistic practice, the existence of some object of reference is indispensable. But this obviously does not lead to the idea that a word’s reference is its meaning. What it means is only that in some cases a given object of reference is indispensable for the formation of the semantic rule whereby a word acquires its referential function.

4. Meaning as a function of use
We shall now move on to a second candidate for the semantic link: use or application. This candidate was privileged by Wittgenstein, who suggested that the meaning of a linguistic expression is its use (Gebrauch) or application (Verwendung). As he wrote in a famous passage of Philosophical Investigations:

You can, for a large class of cases of use of the word ‘meaning’ – if not for all cases of its use –, explain it like this: the meaning of a word is its use in a language. (Wittgenstein 1984c, part I, sec. 43).

 This suggestion applies to both words and sentences. It clearly applies to what has been called directive meaning: the illocutionary forces of expressions, which establish kinds of interaction between speaker and hearer in speech acts and can be made explicit, particularly by so-called performative verbs (like ‘I order’, ‘I promise’, ‘I quit’…). Together with expressive meaning, aiming to express internal psychological states, these first two kinds of meaning aren’t very important for us here. We are more interested in the kind of meaning able to link our linguistic expressions with the world, something we may call referential meaning. Our concern here, as was clear from the start, is the content of declarative sentences, which is the kind of referential meaning we call cognitive, epistemic, informative, descriptive or factual, able to link language with the world and to be endowed with truth-value (Aristotle’s apophantic speech). Such epistemic, informative or descriptive contents should be of major philosophical importance, because by being able to relate language with the world, they should have epistemological and ontological implications.
   However, the identification of meaning with use doesn’t apply so easily to the referential, cognitive or informative meanings of our expressions. Consider, for example, a declarative sentence like ‘The tide is high.’ It is easy to imagine an illocutionary use for this sentence, such as: warning or informing. But by doing this we would revert to meaning as force. According to Searle’s theory of speech acts, all utterances must have the form F(p), where (explicitly or not) F expresses an illocutionary force, and p expresses a propositional content. Here we are not interested in F, even if F expresses an assertive illocutionary force; we are instead interested in the use of p as p. But it is not very natural to speak of the use of a statement separately from its assertive force. The only sure way of approaching pure referential and cognitive meaning with an appeal to use consists in producing an acceptable extension of the concept of use. This would suggest that what is at issue in the case of cognitive meaning is the use involved in the act of communication by means of which a speaker intends to share with a hearer his awareness of a real or possible fact. Thus, when a speaker says ‘The tide is high,’ in addition to using this sentence with the illocutionary force of affirmation, for example, the use may be the spelling in which a propositional content is expressed. This is normally added to assertive force and made to communicate both, with the intention to reproduce a corresponding judgment (the same propositional content plus judicative force) in hearers’ minds.
   To make clear what is at stake, we can isolate cognitive meaning from assertive force, as when we employ the Fregean device of expressing a sentence’s content only as being regarded, depriving it of any assertive force. We can do this with the sentence ‘The dog has run away’ in the subordinate sentence of ‘It is possible that the dog has run away.’ The spelling of the complementary sentence ‘…that the dog has run away’ expressing epistemic content, even if non-asserted – could also be seen as a use.
   But what about the hearer’s understanding of a statement? The hearer is surely not using phonetic shapes in his understanding of its meaning. In order to maintain the view that even in this case meaning can be viewed as use, we need to resort here to a second and bolder extension of the word ‘use.’ It seems possible to say that we use referential expressions simply by thinking them. When a hearer thinks the tide is high, it is possible to say that he actually uses this sentence in an epistemic mode by thinking it, for if Paul understands the sentence ‘The tide is high’ or Anne believes that ‘the dog has run away,’ with or without words they are repeating this judgment or its content to themselves. In normal communication, this use that a hearer gives to heard words by understanding them should be identical to what a speaker has in mind when using words to convey cognitive meaning. Hence, not only the epistemic sense as the speaker’s thought, but also the hearer’s thought, could be viewed as internalized epistemic uses, with or without the addition of assertive force, which in its internalized form is called a judicative force (what Frege called Urteilskraft). Finally, if Plato was right that discursive thought is ‘a silent dialogue of the soul with itself,’ we can generalize this process of internalization and consider any cognitive act as associated with language as use, even without being associated with communicative action.[11] We can call this the cognitive use of an expression, of which judicative and assertive forces are dispensable complementary elements.
   It is easy to question the relevance of the two proposed extensions of the meaning of the word ‘use’ that I am employing in order to save the understanding of meaning as a function of use: though they are not wrong or confusing, they can be considered too cumbersome to be required. However, as will become clear, the reason why Wittgenstein identified meaning with use was a pragmatic advantage, namely, that of locating meaning in its most proper place from the start: normal linguistic praxis – the concrete speech-act situation – even in the usual mental practice of thinking with words. This enables us to individuate the meaning of an expression where it exercises its proper function, so that in this way we achieve the highest level of interpersonal corrigibility, not excluding or distorting anything.
   This is what Wittgenstein’s identification of meaning with use is all about: It allows us to individuate meanings precisely as they are, while in doing philosophy we are too often prone to exempt and distort meanings in order to produce highly illusory speculative insights. In this sense the maxim that meaning is a function of use can help us in practicing what Wittgenstein called philosophy as therapy, which aims to untie the knots of thought tied by philosophers, insofar as it brings our words back from their metaphysical holidays to their daily chores (Wittgenstein 1984c, part I, sec 116).

5. Meaning as a kind of rule
A more basic difficulty arises when we understand that the identification of meaning with use cannot be one of meaning and episodic use tout court, namely, a mere spatio-temporal occurrence (token) of a linguistic expression. This isn’t possible, because each occurrence differs from others in its spatio-temporal location. If this were the case, each new occurrence would have a new meaning, which would spark a semantic catastrophe by making the number of meanings of any linguistic expression unlimited.
   There is, however, a more intuitive alternative. We can understand the words ‘use’ (Gebrauch) or ‘application’ (Verwendung) as a way of use (Gebrauchsweise) or a way of application (Verwendungsweise), since the same word can be used many times in the same way. But what is the way of use? Well, it doesn’t seem to be anything other than something-with-the-form-of-a-rule (etwas Regelartiges) that commands episodic uses. Wittgenstein himself came to that conclusion in an important, though less well known passage of his last work, On Certainty:

The meaning of a word is its mode of application (Art der Verwendung) ... Hence, there is a correspondence between the concepts of ‘meaning’ and ‘rule.’ (Wittgenstein 1984a, sec. 61-62)

In fact, to use a word meaningfully is to use it in accordance with its mode or way of use or application, it is to use it correctly, and to use an expression correctly, in the right way, is to use it in accordance with those rules that give it its meaning. By analogy, we can say that we use a screwdriver according to its way of use when we use it correctly, according to a rule, turning it clockwise in order to tighten a screw. Consider the following examples of ways of use taken from the Linguee online dictionary:

Way of Use: Apply several times to the skin and rub in for several minutes with a circular motion, until completely absorbed.
Way of Use: To color and cover up grey hair, we recommend 20 ml. 6% of a cream oxidizing agent in the proportion of 1 + 1.
Way of Use: Never dispense any pharmaceutical product without a prescription detailing way of use, site, withdrawal periods and other relevant information signed by a veterinarian.

Of course, here ‘way of use’ means rules or sequences or combinations of rules for correctly using things. Now we see clearly that meaning can only be identified with use in the sense of something-of-the-kind-of-a-rule determining episodic uses. And what holds in general for a word’s use also holds here for epistemic or referential use.
   In fact, the identification between meaningfulness and rule is more primitive. Consider the following two signs: ‘00O’ and ‘Oà.’ The second seems to us ‘more meaningful,’ since we have the habit to link it with a rule pointing in a particular direction. Rules are the intrinsic source of meaningfulness.

6. Meaning as combinations of rules
However, why does Wittgenstein prefer to say that meaning is determined by rules? Why can’t the meaning of our linguistic expressions be identified with rules simpliciter? In my view, at least part of the answer was also approached by him with his analogy between language and calculation. (Wittgenstein 1984f: 168; 1982: 96-97). This understanding is reinforced by the many otherwise unjustified considerations in his Brown Book of how complex sequences of rules could be followed in relatively simple language-games, understood as systems of rules. In use, linguistic expressions normally involve calculations, which should be understood as nothing more than combinations of conventional rules. And the cognitive meanings that these expressions have can consist essentially in combinations of more or less implicit, automatized semantic conventions, knowledge of which speakers tacitly share.
   Arithmetic can serve as an illustration. If the meaning of a mathematical proposition is constituted by its proof, considering that the proof is a combination of rules, this meaning is also a combination of rules. Some people can do the multiplication ‘120 x 30 = 3,600,’ for instance, by combining three rules, first multiplying 100 by 30, then multiplying 30 by 20, and finally adding the results 3,000 and 600 to get the result 3,600. The meaning, understood as the cognitive content of multiplying ‘120 x 30 = 3,600,’ would be given by this and other methods of calculation. Together they should amount to essentially the same general signification, insofar as they proceed in different but complementary ways, i.e. beginning at the same starting point and reaching the same result, in some cases by direct application of a single rule.
   We see that what we called something-of-the-kind-of-a-rule can be understood as possible combinations of rules that starting from some initial conditions bring us to some result. The cognitive meaning of a linguistic expression must also be the same as (i) a specific semantic-cognitive rule or (ii) one or more combinations of semantic-cognitive rules that determine a correct episodic use of the rules – which could be called a rule-complex (Regelkomplex). And the cognitive meaning of a linguistic expression is a rule or possible rule-complex that when applied or satisfied brings about a cognition of some state of affairs. – In this book, I normally use ‘rule’ in a broad sense that includes combinations of rules or rule-complexes. This is ultimately a justified extension of the term ‘rule,’ since combinations of rules that produce the same results have the same functions as rules. Although irreducible to shared conventions, such combinations can still be seen as conventionally grounded, since they are constituted by elementary rules, namely, ones usually established by convention. Thus, when someone says, as Wittgenstein sometimes does, that meaning is determined by rules, what can be reasonably understand by this is that cognitive meaning may be the application of some rule-complex enabling us to reach a cognitive result, and nothing more.
   Since we are interested in the problem of reference, the meaning that will be considered will be a content – cognitive, informative, epistemic or referential – that is, something reducible to semantic-cognitive rules responsible for our linguistic awareness of what can be objectively given, which are also criterial rules. So, we are dealing with cognitive criterial rules responsible for the epistemic or referential significance of declarative sentences. Criteria are, in Wittgenstein’s own terms, ‘what confers to our words their ordinary meanings,’ (Wittgenstein 1958: 57). As I understand him, semantic-cognitive rules are based on criteria or criterial configurations, which are conditions generated by rules, hence parts of them, and hence part of their meaning-giving function. On the other hand, criteria (having process-product ambiguity) must be also those cognitively independent conditions that once given satisfy such criterial conditions, making us realize that something is the case. Using Wittgenstein’s own example, if someone says ‘It’s raining’ and this statement is true, this involves applying a criterial rule, a rule which requires that certain conditions must be given – say, drops of water falling from the sky – so that a cognition follows – say, awareness that it’s raining (2001: 28). And this resulting awareness, the cognition, could be understood, as already suggested at the end of the first chapter, as the availability to the system of what results from what we may consider as (effectively or only possibly) satisfied criterial conditions.
   However, if an analysis of the appeal to use leads us to cognitive reference-rules, why appeal to use? Why not just start with an investigation of these rules and their combinations? The answer was already given. Language is primarily an instrument of action, and meaning, cognitive or not, is there to facilitate action. Moreover, since semantic rules are not abstract objects, they only exist when they are applied. Attention to correct use helps us to individuate meaning and to find the real cognitive-criterial rules or combinations of rules that must of necessity be applied in order to confer meaning.
   We can further elucidate what is at stake by appealing to a metaphor: when a post office delivers a letter, it has general indications as to the addressee’s state, city, and locality. These general indications can be compared with the grammatical meaning of a sentence. Although necessary, they are insufficient, as too many other sentences have the same grammatical meaning, because too many other addressees live in the same state and city. To reduce this vagueness, mail carriers also need the name of the street, the building or house number and if applicable the addressee’s apartment number. Without these singularizing details, it is impossible to deliver mail to its proper destination. The same holds for cognitive meaning. What is decisive is the way of applying our expressions in the given context – not only the concrete, but also the whole discursive context, including the context we find in philosophical texts. What an appeal to use does is to lead us to semantic details necessary to find what really matters. In other words: the more general traits of an expression’s way of use are less relevant in the sense that they are common to many other expressions and for this reason are not able to individuate meanings. What matters at most are the more specific traits of meaning: ways of use. These are traits that expressions can only gain in the real contexts where they are applied. Consequently, these can only be completely explored in linguistic praxis. For this reason, it is so important to explicitly consider occasions of use. These can be responsible for subtle semantic variations that an expression can have in different concrete discursive contexts. As we will see, consideration of such subtle semantic variations is of particular importance for correcting or criticizing misuse, since this enables us to correct misconceptions arising from philosophical attempts to use words beyond the limits of meaningful language. Particularly serious cases of overstepping the limits are ones involving the metaphysics of reference, which in most cases result in errors by systematically confusing semantic elements originated from linguistic praxis.

7. Meaning and language-games
There is more to be said about meaning as a function of use. The first thing to note is that a linguistic expression only makes sense when used within a system of rules, within a language game. Here again we may appeal to a metaphor. We can compare a linguistic expression with a chess piece, and its use with a move in playing chess. When you move a chess piece, the meaning of the move is not given only by the rule that governs the piece’s move. What the move really means in the important sense of the word will depend on the game situation. It will be given by the contextually determined tactic, by the calculation of possible combinations of rules in anticipation of possible moves by the opponent and responses that could be made. This is a calculation made in playing chess and could be different in a different game, even if the pieces were the same.
   Something not very dissimilar occurs with linguistic use. The linguistic rules governing what Wittgenstein called ‘superficial grammar’ could be compared with the rules for moving chess pieces. But grammatical rules – even those of logical grammar – may not be what really matters. Often what is essential are rules or rule-complexes belonging to what Wittgenstein called ‘deep grammar’ (1984c I, sec. 668). These may have more resemblance to semantic rules like those we exemplified before (for the words ‘Aristotle’ and ‘chair’) that if combined would justify moves depending on chess players’ tactical calculations, particularly when we consider what takes place in a dialogic speech.
   To give an example, one knows that the sentence ‘Caesar visited Calpurnia’ is grammatically correct, and one may even know that its logical form is aRb. But this will be of no help if one does not know who Caesar and Calpurnia were, what relationship they had, and cannot even imagine when or why he could have visited her. Superficial grammar (or syntax) gives expression to a grammatical sense that is often the same for semantically different sentences. The rules and combinations of rules that constitute what is meant by a linguistic expression are more flexible, changing in accordance with the concrete and linguistic contexts.
   Furthermore, in the same way that the rules responsible for a strategic move in chess depend on a context provided by the system of rules that constitutes the game of chess, the rules determining the application of linguistic expressions are able to produce meaningful utterances only when combined within a system of rules. This system is called by Wittgenstein a language-game (Sprachspiel) or a linguistic practice (sprachliche Praxis). Language-games can be characterized as linguistic systems that typically include syntactic, semantic and pragmatic rules. Examples of language-games given by Wittgenstein are:

Giving orders and acting according to them, describing an object by its appearance or measures, informing… speculating about an event, making and testing hypotheses… making up a story, reading… solving a riddle, telling a joke, describing a landscape, acting, asking, thanking, cursing, greeting, praying, etc. (Wittgenstein 1984c, sec, 23)

   But he also uses the same idea in a wider sense, pointing to more extended domains of language like:

The language of colors, the language of proper names, or even the important  ‘knowing games’ from On Certainty, like the game of doubt and the languages of history, physics, chemistry and arithmetic. (Costa 1990: 50)

   That is: it seems that almost any chunk of our language can be seen as a language-game. Language-games include themselves, one within another, like the case of Cantor’s theory of infinite numbers within the theory of numbers and the language of mathematics, and they can partially overlap one another, as when someone describes a scenario and using ambiguities simultaneously tells a joke, insofar as we remain able to distinguish them (Wittgenstein 1984c, sec. 46-48). Fundamental is that these practices remain distinguishable at the interpersonal level.
   The concept of language-game or linguistic practice contains the concept of the speech act, systematically studied by J. L. Austin and John Searle, but it is much wider. This is why Wittgenstein was not mistaken when he wrote that there are countless language-games (1984c I, sec. 23).
   By making the meanings of expressions the result of rule combinations belonging to rule-systems typified by language-games, Wittgenstein was endorsing what was later called semantic molecularism: What we call the meaning of an expression does not depend on the expression in isolation (semantic atomism), nor on its insertion in language as a whole (semantic holism). It depends more essentially on the context of the linguistic practice in which it is located (molecular subsystem of language).
   Finally, it is a mistake to believe that meaning is a matter of all or nothing. It is more plausible to think that when used according to the rules of a language-game some part of a word’s meaning extends to the group of games to which this game belongs, gradually merging with them.
   In support of the idea that we use and give meaning to our expressions in language-games, in the Brown Book Wittgenstein described natural language as a great nebula of language-games, and later, in his Philosophical Investigations, he compared it to a great old city:

The language of the adult presents itself to our eyes as a massive nebula, ordinary language, surrounded by more or less defined language-games, which are technical languages. (Wittgenstein 1984e: 122)

Our language can be seen as an ancient city: a maze of little streets and squares, of old and new houses, and of houses with additions from various periods; and this surrounded by a multitude of new boroughs, with straight, regular streets and uniform houses. (Wittgenstein 1984c, sec. 18) 

The nebula, the city, begins with what was built in its original center: the practices of ordinary language, expressing our ordinary commonsense wisdom. To this, there come new insights, like those arising with the emergence of new scientific fields. As with games, the great old city can be subdivided in many distinct ways, one part including another, or one overlapping another.
   There is a noteworthy relation of dependence here: learning and teaching these new scientific practices, even the possibility of their understanding and creation, depends on prior acquisition of more basic practices governing ordinary life. This coheres with our principle of the primacy of modest common sense: Rejection of its proper assumptions by means of science is a questionable approach, and it would be logically incoherent to reject them as a whole based on science.
   A question that now arises is: in such circumstances, what criteria would we have for identifying meaning variations, or, less ambitiously, what criteria would we have for identifying the language-game in which an expression is used or even misused? Considering that language can be subdivided in multiple and varied ways, it seems that we can apply different criteria to the same move, insofar as we are able to interpersonally identify and share the criteria we are applying... But in this case, what guides us in choosing a criterion? Is this identification really possible?
   My tentative suggestion is that the identification of a language-game according to the criteria for the use of a word, which also establishes its meaning, involves identifying state-contexts (any of the many actual states of a game) created by two factors:

(i)                the relevant factual and linguistic circumstances in which the word is used, together with
(ii)              the speaker’s intention in using the word, insofar as this intention is made interpersonally clear, even if only in a tacit way.

I suggest that in the normal case these two factors allow the identification of the language-game in which a speaker is using a linguistic expression if the speaker succeeds in giving a clear idea of the circumstances and the intention he has in using an expression. In this way he is identifying a state-context in a certain language-game, the relevant system of linguistic rules for determining how he is using the expression. These are the intended rules constitutive of what is meant. And if a hearer correctly identifies the speaker’s state-context (circumstances plus intention), he identifies the language-game the speaker has in mind and will be able to understand correctly what the speaker means. (For example: if I tell my students Aristotle said that friendship is only possible among equals, the context shows everyone that I am playing a game of naming in which I intend to lecture on the famous Greek philosopher and not about some homonym.)

8. Meaning and forms of life
There is a last important concept in the understanding of Wittgenstein’s explanation of meaning. The linguistic practices that form the nebula find their ultimate raison d’être as constituents of what Wittgenstein called a form of life (Lebensform). As he wrote in his few passages on this concept:

…the word ‘language-game’ is used here to emphasize the fact that speaking a language is part of an activity, or of a form of life. (Wittgenstein 1984c, I, sec. 23)
Right or false is what human beings say; and in the language they agree on. This is no agreement in opinions, but in form of life. (Wittgenstein 1984c, I, sec. 241)
What is taken for granted, the given, we could say, are forms of life (Wittgenstein 1984c, II: 572)

Wittgenstein arrived at this idea probably influenced by an article written by the great anthropologist Bronislaw Malinowski, who suggested that in order to learn the language of a primitive people one needs to share life with them in their society (Malinowski 1989).[12] One example used by Malinowski to illustrate this point can be revealing here: when fishermen in the Trobriand Islands use the phrase ‘paddling in a place,’ they mean they are navigating close to an island village. The waters around the islands are so deep that it is not possible to use a pole to propel the canoe, so they need to paddle their boats to reach the village. Only by knowing speakers’ life circumstances can we find the information needed to understand what their expressions mean.
   The relevance of much that Wittgenstein wrote consists in his having seen the importance and comprehensiveness of some ideas. For him the phrase ‘form of life’ means the way of life in a society. More precisely: the complex of regularities that govern the lives of people in a social environment considered in its totality.
   We can compare the idea of a form of life with what is involved in two technical terms introduced by J. R. Searle. These are (a) the network of meanings involved in the determination of an intention, and (b) the background of abilities, skills, dispositions and ways of doing things which is linked with the corresponding network (Searle 1983, Ch. 5). Though including what Searle means by network and background, the concept of form of life is more comprehensive, since even the landscape in which natives lived should be comprehended by the concept and may have an influence on the meaning.
   More auspicious is a comparison between the concept of form of life and Husserl’s concept of life-world (Lebenswelt), which for the latter author can be the whole of our shared communal world of human activity (Husserl 1954, Vol. VI: 105 f.), grounding in this way all possible knowledge. The life-world is able to undergo changes. For Husserl, the life-world, which can be subdivided into different home-worlds (Heimwelten), forms the holistic framework within which all knowledge is acquired, serving therefore as the ultimate foundation of all human cultural endeavors, gradually extending into scientific ones. Furthermore, although there are different life-worlds, they must have basic common aspects (like spatio-temporality, materiality, life, birth, death, instincts, hunger, thirst, etc.).
   Wittgenstein would probably share this view at least in its non-theoretical aspects. It is helpful to see that there must be something common in the most basic levels of our different forms of life. For much of this, commonality should be a condition that allows us to be transplanted into a different form of life and be able to learn other cultures’ languages, assuming that we all share much of a common human nature (a generally uncontroversial assumption).

9. Tying the threads together
We can now summarize. Language appears in Wittgenstein’s philosophy as an immensely complex system of syntactic, semantic and pragmatic rules: a system we can subdivide in many ways into subsystems called languages, linguistic practices and language-games, which are in turn rooted in a wider ground: the life-form made up of regularities that determine the lives of people in a society. Linguistic practices constituting our ordinary language originate spontaneously from our form of life and depend upon it. Here again, we see that creating and learning the specialized language-games of science is only possible because of the assumption of more central practices of ordinary language ultimately entrenched in life-forms. This is also why a computer will never be able to give meaning to the signs with which it operates: a silicon-based machine is a by-product of a life-form and not a biological agent naturally growing within it.
   We can synthesize our considerations so far in the following formula:

A meaning of an expression X = any episodic use of X made in accordance with the rules of an appropriate linguistic practice (the language-game) rooted in a certain life-form.[13]

This is a characterization of meaning as something that belongs to the praxis of language as it is understood and to our extensions of the concept of use as what is meant in mental acts. This assimilation of cognitive meaning to action by means of an extended notion of use as a rule-in-its-application is what makes it unnecessary to hypostasize semantic rules as abstract objects. Meaning is what we think of or speak about as being meaningful; and what we think or speak is meaningful insofar as it is correctly used, namely, used in accordance with the meaning-rules of linguistic practices rooted in our life-form.
   This is what I believe we can achieve, based on Wittgenstein’s views, an uncomfortably vague but sufficiently plausible and minimally distorted surveillable representation of the grammar of the concept of meaning. This kind of representation is important insofar as it plays a role as a semantic foundation for philosophy as therapy. This is also why a surveillable representation of the grammar of meaning is central to Wittgenstein’s later thought: it is the sustaining core of his philosophy, as much as the doctrine of ideas was the sustaining core of Plato’s philosophy.

10. Criteria and symptoms revisited
Another important distinction that we owe to Wittgenstein, already introduced in the first chapter of this book (sec. 9), is the distinction between criteria and symptoms. Semantic-cognitive rules are criterial rules. Criterial rules are ones based on conditions called criteria. As we have also noted, words like ‘criteria,’ ‘symptoms’ and ‘conditions’ involve process-product ambiguity. Often they mean the internal conditions belonging to the semantic-cognitive rule (criterial rule) that we are able to consider independently of its effective application in the world and hence belong to meaning. But criteria can also mean the cognitively independent external conditions that, once really given, make possible the application of a semantic-cognitive rule. When criteria are understood in this last way, they can be seen as particularized properties or tropes or configurations of them, as we will see. Moreover, as we will also see, as criteria, tropes are understood to necessitate the application of the semantic-cognitive rule and, when this rule is the verifiability rule of a statement, they are what necessitate its truth – what we call its truth-maker.
   There is, as we have also noted, a fundamental difference between criteria and symptoms. Criteria are conventionally grounded conditions that, once accepted as really given, warrant for us the application of a semantic-cognitive rule. Symptoms, on the other hand, are conventionally grounded conditions that, once accepted as really given, make the application of a semantic-cognitive rule only more or less probable. A criterion should establish the sufficient conditions for the application of an expression, though not essentially, since criteria can be multiple and varied, as our investigation of proper names has shown. Because of this sufficiency, Wittgenstein also called them definitional criteria; they are primary criteria, while symptoms are also called secondary criteria (cf. Wittgenstein 2001: 28). In this book I will use the expression ‘secondary criteria’ to qualify those stronger forms of symptoms able to considerably increase the probability of their bearers’ existence.
   One example makes Wittgenstein’s distinction clear: a criterion for the application of the concept-word ‘malaria’ is actually finding a bacterium – plasmodium falciparum – in a patient’s blood. If we accept that we have found this, by definition we are warranted in saying that in some way the patient has malaria. But if all we find is that the person has cyclically high fever, we have a symptom of malaria, something that makes it only probable that the patient has contracted the disease.
   Insofar as criteria are understood as internal constitutive conditions of the semantic-cognitive rules for the referential use of a conceptual expression, they must belong to its meaning, since these rules (whether effectively applied or only cognitively regarded) are constitutive of meaning. When Wittgenstein wrote that criteria ‘give words their common meanings’ (Wittgenstein 1975: 57), he was referring to criterial rules.
   Finally, criteria play the role of criteria only in the context of the language-games to which they belong. This is the main reason why Wittgenstein says that there can be a grammatical oscillation between criteria and symptoms: with the alternation of linguistic practice, criteria can become symptoms and vice versa (Wittgenstein 1983c, sec. 79, 354). That is: the same condition that works as a criterion in one practice can serve only as a symptom in another practice and vice versa. And similar changes can also occur as a result of the evolution of language, which may change and improve our conventions, often turning criteria into symptoms by replacing them with new conditions.
   The distinction between criteria and symptoms is also important for the critique of language. Philosophers are all too often inclined to treat symptoms as though they were criteria. To give a very simple example: consider peoples’ facial features and physique. These are the physical characteristics by means of which we are able to immediately identify people we know. At first sight, it seems that they are the real criteria for identifying persons – and within some superficial language-games they may work in this way. But in a deep way it is obvious that they aren’t. If a person, as happens in fairy tales, were transformed into a donkey, but continued to behave no differently than before, talking to us and in full possession of his memories, personality, knowledge and abilities, we would be forced to admit that he remained the same person, even though in a different body. This and other similar thought-experiments show that people’s bodily appearances are not primary criteria at all, but only useful symptoms that make their identification very probable in an easy and immediate way. To find the ultimate criteria of personal identity is still today a controversial philosophical problem.[14]

11. Transgressions of the internal limits of language
In the Tractatus, Wittgenstein was interested in ascertaining what David Pears called the external limits of language and its transgressions (1970, Ch. 5). This is relatively easy to spot: a logical contradiction is an external transgression. However, he came to see that most philosophical confusions are caused by the more subtle transgressions of the internal limits of language. These transgressions happen because many of our expressions can be used in different linguistic practices, undergoing in this way more or less subtle changes in meaning. As Wittgenstein wrote, ‘The place of a word in grammar is its meaning’ (1984d, sec. 23), a place that cannot be fixed beforehand, since it may change. Now, when an expression is used simultaneously in different practices, where it should receive a different meaning or meaning-nuance, it turns out to be easier to confuse what we mean with it.
   According to Wittgenstein’s philosophy, we can find two forms of confusion or misleading uses of expressions, which we may call equivocity and hypostasis.[15]
   These two forms of transgression have a striking similarity to the psychoanalytic distinction between the two mechanisms of the primary process (primäre Vorgang), called by Sigmund Freud displacement (Verschiebung) and condensation (Verdichtung). Hence, it is interesting to explain this process here very briefly. According to Freud, our thought can involve two distinct processes: the secondary process (sekundäre Vorgang) and the primary process (primäre Vorgang). The secondary process is the typically conscious process of rational thought, particularly scientific thought. In this process, affective or emotional charges (Besetzungen) are firmly associated with their respective representations (Vorstellungen). The primary process, on the other hand, is found in dreams, neurotic symptoms, humor, artistic creations, religion, and… philosophy! In these cases, emotional charges are not rigidly associated with their respective representations (or thoughts) and can be transferred to different representations, insofar as the latter can easily be associated with the former representations. The primary process is what produces the conscious manifestation of unconscious or pre-conscious thoughts, in the latter case understood as non-repressed and consequently always able to become conscious.
   The two fundamental mechanisms of the primary process, displacement and condensation, are more clearly illuminated in Freud’s Interpretation of Dreams (Freud 1900, Ch. 7).
   Displacement occurs when the emotional charge of a repressed representation is transferred to another representation, which is able to elude censorship and become conscious, thereby releasing its endo-psychic tension into consciousness. We have displacement when representation R1, repressed and therefore unable to become conscious, has its charge transferred to representation R2, able to evade censorship and become conscious. A Freudian example of displacement is the story of a Jewish woman who couldn’t marry the man she loved because he was a Christian. However, she dreamed that she gave him her comb. This is her conscious representation in the dream; but in her unconscious, the repressed representation is the idea of giving herself to him in love. The emotional charge passes from the repressed representation to the innocuous one, which is able to outwit censorship, becoming conscious as a dream. This makes it possible for the charge to be released in the dreamer’s consciousness, diminishing the endo-psychic tension.
   The mechanism of condensation is somewhat different. Here a representation (or group of interrelated representations) transfers its affective charges to a partial representation belonging to it, which becomes liberated in consciousness. We can represent this by saying that the charges belonging to the representations {R1, R2… Rn} are usually condensed in one of them, say, R2, which enters into consciousness, in this way allowing the release of emotional charges into consciousness. One example of condensation would be a case in which the same woman dreams that the man she loves forgot his scarf at her home... The scarf is part of the whole representation of the man, and the emotional charges associated with the whole are condensed in this partial representation and released into consciousness. It is worth remembering that according to Freud, displacement requires full unconsciousness by being a product of repression, while condensation requires only pre-consciousness, since it isn’t necessarily a product of repression.
   Now, an investigation of the two mechanisms by which the internal limits of language are transgressed brings into sharper focus the sometimes noted relation between philosophy as therapy and psychoanalysis (e.g. Wisdom 1969), for it shows that philosophical activity is affected not only by a lack of semantic awareness, but also by affirmative unconscious motivations.
   Let us see now how the primary process works in cases of confusion arising from linguistic transgressions of normal uses of expressions. By using an expression equivocally, a philosopher shifts the use of this expression, applying it in a state-context of a linguistic practice B, though following the semantic rules that this expression should have in linguistic practice A. This equivocity amounts to displacement, since the emotional charges associated with the first use are transferred to a new representation. On the other hand – in what we call hypostasis – the philosopher tries to apply an expression that can be used according to the rules of two or more linguistic practices, say, A, B, C, etc. simultaneously in a neutral state-context. It is as if there were a single linguistic practice able to bring together these different uses, adding their emotional charges, when in fact this practice neither exists nor has any justification to be created.
   Philosophical examples of these mechanisms can be complicated and difficult to describe, since philosophers, being masters of deception (and self-deception), construct their spiderwebs of far more abstract and complex material than ordinary mortals can imagine. Hence, I will consider only two very simple examples.
   For the case of displacement, consider the following skeptical paradox attributed to the Megarian philosopher Stilpo, denying the possibility of predication. For Stilpo, if I say that Socrates is wise, this is a contradiction, because I am denying that Socrates is Socrates. That is: I can say of something that it is what it is, but if I want to say something more than this, I fall into a contradiction, for I am denying that it is what it is… The upshot is that all that we can do is to express the identity of a thing with itself or remain silent.
   We can explain Stilpo’s fallacy as due to a failure to distinguish the ‘is’ of predication (copula) from the ‘is’ of identity. We can distinguish a linguistic practice of type A with its identifying state-contexts – in which the verb ‘to be’ means ‘is the same as’ (e.g. ‘Socrates is Socrates.’) – from linguistic practices of type B with their own identifying state-contexts – in which the verb ‘to be’ is used in the predicative sense (e.g. ‘Socrates is wise.’). However, Stilpo recognizes the verb ‘to be’ as having only one correct use: that which is found in practices of type A. As a result, each time he observes people using the verb ‘to be’ in state-contexts of practice B, he understands their use as following the rule of use that the verb has in practice A – meaning ‘is the same as.’ In this way, he equivocally and systematically displaces the real use from practice A to practice B. Since he sees that in state-contexts of A he cannot apply the ‘is’ of identity typical of practice B, he falsely concludes that true predication is impossible.
   I will now offer an easy example of hypostasis in philosophy. Consider this suggestion made to me by a philosopher, according to whom the verb ‘to be’ must have a truly primordial sense, which is not only that of copula, but also of identity and of existence together! To justify this, he considered the sentence: ‘To be is to be’ (Sein ist Sein). This sentence says not only that ‘to be’ has the property of being, but also that ‘to be’ is the same as ‘to be,’ and finally that ‘to be’ has the property of existing (of being).
   Against this folie metaphysique, a critique of language will tell us that it is much more plausible to think that what the philosopher seeks with the ‘is’ in the sentence ‘To be is to be,’ although grammatically correct, is semantically only an incoherent mixture of different senses of the verb ‘to be.’ These were created for different practical purposes but attain no justification when mixed, except the satisfaction of an ad hoc philosophical claim. It is an hypostasis: a condensation mixing three different modes of use or meanings of the same word in a neutral state-context. These three modes of use belong to three actually distinct practices, say, the predicative practice A, the identifying practice B, and the practice of attributing existence C. In the best case, this is multiple ambiguity; but since the philosopher is claiming to have discovered a way to achieve the primordial sense in an arbitrary way, the diagnostic is of mere incoherence and illusion.
   I offer these explanations because in criticizing the metaphysics of reference, we very often denounce equivocity and hypostasis. Wittgenstein suggested that philosophical maladies have their origins in a ‘craving for generality’; in efforts to achieve generalization without enough fundament, very often reductively influenced by the greater success of natural science (Wittgenstein 1975: 18). We can now suggest that here as well the frequent case of equivocity may work as a compensatory byproduct of repressing some kind of undesirable awareness.
   An additional point is that striving for generalization is inherent in the philosophical endeavor (at least as revisionary metaphysics) even if it may be ultimately doomed to fail. This is, I think, the reason why Wittgenstein’s concession that the philosophical unavoidably bumps up against the walls of language has the mark of profundity (Wittgenstein 1984c, sec. 111). The reason for this is that these confusions, when effective, have the potential to point to relevant issues after forcing us to search for the right way out of the illusions they produce in us. As I intend to show, much of the metaphysics of reference is grounded upon the forms of confusion described above, particularly equivocation (displacement), and can be the object of a critique of language.

12. The form of semantic-cognitive rules
In an approximate way, we can now expose the general form of a cognitive or criterial semantic rule (basically the identification rule of singular terms, the attributive rule of general terms, and the verification rule of declarative sentences). This rule is constituted, on one hand, by a relation that can be summarized in the sign ‘~>,’ which means either an inductive inference (for empirical knowledge) or a deductive inference (for logico-conceptual knowledge). By ‘C’ we mean the criteria to be satisfied and, furthermore, by the result ‘A’ we mean meaning-awareness. Here is the basic schema:

C ~> A

This schema of a semantic-cognitive rule is a simplification, for the criteria can be multiple, varied and staggered in complex procedures. The satisfaction (always in the context of some practice) of a (definitional) criterion under the state-context of a certain practice gives place to the occurrence of the (usually non-reflexive) meaning-awareness A, which in the case of a statement could be expressed by a declarative sentence. The cognitive content or meaning or sense is the whole procedure of rule-following, including still unverified criteria, insofar as they also belong to the rule.
   Complementing what was said, there is a second cognitive element associated with the semantic-cognitive rule, which is the awareness of the consequences of the satisfied content – of the rule or rule-complex. That is, what I believe can be explained by theories of consciousness such as those briefly summarized at the end of chapter II. It is, for instance, what has been called the ‘availability of content to reasoning and action’ (Block), the ‘transmission of content for the mind’s global workspace’ (Baars), ‘brain celebrity’ (Dennett), etc. It is the consciousness of a meaningful reference.
   Calling the meaning or cognitive content [{C1 ˅ C2 ˅… ˅ Cn} ~> A], where each criterion is seen as sufficient for the meaning awareness A, and calling E its cognitive effect (like transmission of content to the global workspace), we can summarize the more common form of a semantic-cognitive rule added to its conscious effects as follows:

[{C1 ˅ C2 ˅… ˅ Cn} ~> A] > E
                                      
In order to better understand the rule, suppose that C1 and C2 are assumed criteria for the meaning awareness of what can be expressed by the statement ‘Caesar visited Calpurnia.’ The conclusion that Caesar visited Calpurnia is a meaning awareness that can be silently thought or spelled out.[16] Now, we can consider three situations:

(a) When we take this A in isolation from any criterion for identifying Caesar or Calpurnia; in this case all that we get is mere grammatical meaning.
(b) When we add informative content to A, associating it at least potentially with some conventional procedure from which it results, for instance, C1 ~> A. Then we have semantic-cognitive meaning; the rule is regarded as potentially applicable. This regarding is an imaginary rehearsal of the true application. Here the cognitive meaning, say, the verifiability rule, can only be considered. But this does not mean that propositions or thoughts are statically regarded as abstract entities. What is meant is that they are applied in our imagination, even if in a limited way, only felt as potentially applicable, with the result that we make ourselves aware of the semantic-cognitive content as a possible occurrence of a rule-in-its-applicability. This already makes us to a certain extent aware of the foreseeable effects E. Since we are using it as an instrument in a search for possible consequences of its satisfaction, we need not be reflexively aware of the relation A ~> E.
(c) Finally, when a criterion, say, C1, is seen as actually satisfied; then we have the application of the semantic-cognitive rule, which can be symbolized as C1 & [{C1 ˅ C2 ˅… ˅Cn} ~> A] ~> E. This fulfilled A inevitably produces a referential awareness, which could be nothing other than A’s availability for reasoning and action, its transmission to the mind’s global workspace, brain celebrity, etc. given by theories of consciousness, since it is what results from consciousness of a real state of affairs. Here the semantic-cognitive rule is effectively applied. In this case, if as a result we spell a sentence internally, adding to it a judicative value and we associate this cognitive application of the rule with its spelling, we have an assertion: a statement spelling out a sentence whose content is accepted as true. Notice that what is judged or asserted is the whole content: the verifiability rule along with the satisfaction of its criteria.

It is interesting to note the proximity between our conclusion and inferentialist approaches to meaning. If we say that a content, a semantic-cognitive rule, is available for reasoning and action, we also mean that the content (which is in itself inferential) would be inferentially open to those related contents. This is what I believe can be understood as the cognitive effect of the satisfaction of the semantic-cognitive rule. However, I will not risk mixing this inferential openness proper of the cognitive awareness of content with the real meaning, because this openness is a consequence of the instantiation of referential or cognitive meaning won through the application of its semantic-cognitive inferential rules.
   The relevance of these sketched formulations isn’t clear now, but I think that it will gradually justify itself in the course of this book.

13. What is wrong with the private language argument?
I do not believe that there is only one possible interpretation of the so-called private language argument (Wittgenstein 1984c, I, sec. 244-271), a term that isn’t even present in Wittgenstein’s text. There are a variety of more or less interesting interpretative alternatives. To interpret Wittgenstein is like trying to assemble a jigsaw puzzle, knowing from the start that some pieces will inevitably be left over. This isn’t a problem for me, since my aim here is not interpretative.[17] What I want is to reconstruct Wittgenstein’s ‘argument’ in a way that makes its consequences as philosophically strong as is reasonably possible. This philosophically strong formulation will be important, because if it is right it means the destruction of all our human subjectivity, not only as it is currently understood, but as it is understood in all of traditional philosophy (e.g. in the cogito or regarding sense-data). A private language argument with trivial conclusions would be of scant interest.
   We can begin with the contrasting case: public physicalist language. How do we learn to identify and distinguish different types of physical objects? For example: how does a child learn to identify references of the word ‘ball’? This doesn’t happen by means of verbal definitions, but ostensibly: adults point to examples and say things like, ‘This is a ball’ or ‘That isn’t a ball’... and the child eventually learns what types of objects are round balls. But this learning is only confirmed when a new ball is presented and the child shows adults that it is able to re-identify the object as belonging to the ball type. In this case, based on agreement among other speakers of the language regarding correct re-identification, it is possible for everyone (adults and their children) to know that the child has learned the rule for identifying ball-type objects. That is, the only way to know that we have learned a rule is ultimately if our way of application is confirmed by interpersonal checking.
   Consider now what happens when we try to identify internal mental entities of a phenomenal nature. In this case, we cannot do any checking of interpersonal re-identifications. Suppose that a person is expected to learn to identify an internal state, for example, a feeling of pain. Other people cannot teach her to do this, because they cannot know if and when she feels pain or how it feels to her. But let’s suppose that independently of any public language a person decides to point inwardly to some feeling and identifies his feeling through a sign that she himself has invented. Suppose this sign is ‘P’ (for ‘pain’). Imagine now that the next time she feels pain, she says to herself ‘P,’ intending to point to the same internal mental state. In this case she won’t be able to know if she is really pointing to the same phenomenal state that she initially pointed to, because there are no other speakers who can check the correctness of his rule application, i.e., who are able to confirm or refute his identification. As Wittgenstein realized:

‘I impress it on myself’ can only mean: this process brings it about that I will remember the connection correctly in the future. But in this case, I have no criterion of correctness. One would like to say: whatever seems right to me is right. And that only means that here we cannot talk about ‘right.’ (1984 sec. 258)

Where interpersonal correctness criteria cannot be found, we cannot distinguish between following a rule and the mere impression of following a rule. However, this distinction is indispensable, because without it we have no way to construct something that we may effectively call ‘a rule.’
   Since language is a system of rules, the generalization of this result leads us to the radical conclusion that there cannot be a language whose objects of reference are internal phenomenal states. The only construable language is one based on behavioral expressions of internal states, transforming expressions like ‘Ouch!’ into ‘I feel pain.’ Nevertheless, he concedes the existence of these mental states, rejecting behaviorism. This is in my view an incoherent move, since if behaviorism weren’t the case, mental states would be beyond the reach of linguistic rules and therefore cognitively unspeakable and in the end senseless… once he also confirms that something about which nothing can be said has as much value as nothing.
   The problem, as Ernst Tugendhat once told me, is that the private language argument is too counter-intuitive to be correct. The point, however, is to discover where the weakness of the argument lies. In order to find this, we need to make two things clear. The first is that we will probably only stop regarding a rule as a rule if we conclude that it is logically impossible to correct it. A rule does not cease to be a rule just because for some contingent reason it was not in fact interpersonally checked. After all, it is an indisputable fact that many of the rules we follow, for one reason or another, have never been interpersonally checked. I can invent for myself the rule of never eating creamed spinach, and nobody needs to be informed of this rule. There are rules that for merely circumstantial reasons cannot be checked, such as those made by a shipwrecked sailor who is never rescued and consequently lives and eventually dies alone on a remote uninhabited island.
   The objection that could be made to this interpretation is that Wittgenstein’s argument demands that any rule, in order to be a rule, must be publicly checked for correctness, and not just be able to be publicly corrected (correctable). Even if this interpretation were true, it would be uninteresting. For it expresses only an absurdly implausible and methodologically anti-Wittgensteinian idea: it would jeopardize our commonsense certainty that we are able to follow rules that have not yet been checked by others. In fact, if we wish to overstate skepticism, we could also argue that no rule can be applied in situations where it cannot be subjected to simultaneous interpersonal correction – after all, there is no guarantee that in the absence of this control the rule will be correctly interpreted and applied... However, this gratuitous skepticism is too implausible to tempt us.
   With this in mind, let us now interpret Wittgenstein’s argument as assuming that the rules of a phenomenal language must be logically incorrigible. Let’s suppose that every morning when waking up I unintentionally follow the rule to remind myself of the first sentence of Dante’s Divine Comedy, but that I always immediately forget doing this. Here we are already close to nonsense, and we would reach total nonsense if it could be proved to be logically impossible to know if this happens.
   We conclude that it is the assumed logical incorrigibility of phenomenal language that convinces us of the plausibility of the private language argument: it seems very plausible to assume that a rule that logically cannot be corrected cannot be considered a rule. If the rules of our (supposedly) private phenomenal language are logically incorrigible, it seems that they cannot, ultimately, be distinguished from mere impressions of rules.
   This reconstruction of Wittgenstein’s argument is not only the most interesting and reasonable. It also uncovers what I believe to be an important implicit assumption made by him. Once he noted, for instance, that even though person A’s nervous system could be connected to that of person B, so that A could feel a wasp stinging B’s hand, only the location of pain would be shared, but not the pain itself. This is because pain felt by A would be A’s pain, while pain felt by B would still be B’s pain (Wittgenstein 1975: 54). In his most famous article, surely read by Wittgenstein, Frege noted that if another person could enter our minds to observe a visual representation, the representation she experienced would be her own and not ours (Frege 1892: 30). Such considerations lead us to a dogma generally assumed by earlier Twentieth Century analytical philosophers: the thesis that phenomenal states are logically non-sharable.[18] If this thesis is correct, then interpersonal corrigibility of phenomenal language would be logically impossible, which would support the private language argument.
   At this point, all we need to destroy the private language argument’s main foundation is to show that the logical non-shareability of phenomenal states is a false principle. That is, we need to show that although the rules of a phenomenal language have never been interpersonally corrected, they are – contrary to what Wittgenstein and many philosophers assumed – logically corrigible from an interpersonal perspective, this being the hidden flaw that tacitly supports the private language argument.
   It’s hard to imagine a thought-experiment able to prove that phenomenal states are logically shareable. We can begin by making an analogy with computers. Suppose A and B are updated versions of the primitive kind of automata called by Grey Walter machina speculatrix, which fed on light and spent all their time in search of it. Suppose automaton A meets automaton B, and that A is able to read the information content that B has accumulated in its searching. Although automaton A can copy these data first, and only afterwards read them in its own system, so that such ‘contents of experience’ become an unshared part of itself, there is no contradiction in thinking that A can read these ‘contents’ directly in B, as if they were its own, thereby sharing them with automaton B! This would in fact be the simplest and most direct method. Why should we think that in a similar situation we humans would need to be different from machines?
   Perhaps it is even possible to imagine that someday there will be two human beings, A* and B*, who somehow share some functioning of parts of their brains. Suppose that their limbic system is in some way tuned essentially the same, while the neocortical regions of A* and B* remain distinct. Now, it seems conceivable that a mental state of pain that occurs in relevant parts of the same limbic system could be shared by subjects A* and B*, even though their conscious interpretation of pain, made in their distinct neocortical regions, are quantitatively different. If we understand pain as essentially a process occurring in a limbic system, then A* and B* really could share the same pain, demonstrating possible interpersonal checking of the same internal phenomenal state.
   The thought-experiments considered above suggest that it is logically possible to distinguish:

(a)   the subjective interpretation of a phenomenal mental state X
from
   (b) the phenomenal mental state X in itself.

In fact, this separation seems possible. We know cases of hypnosis where people are led to feel pain even though a source of this pain is absent, or the case of a patient at the dentist who, because he is afraid of treatment, believes he feels pain when he really only feels the sensation of friction…
   Now, if we accept that it is logically possible to separate (a) and (b), then the interpersonal sharing of mental phenomenal states turns out to be logically possible as well, which at least in principle makes possible interpersonal checks of identification rules for mental states. In this case, the private language argument fails because the logical unshareability of phenomenal states is a false principle. In this case, the rules of phenomenal language acquire an epistemic status that does not differ from that of the rule I made for myself of never eating creamed spinach; in principle, both rules could be checked. Consequently, we are entitled to assume that what we now believe to be the rules of our phenomenal language may in fact be the actual rules, since they are at least logically susceptible to interpersonal correction.
   Furthermore, we are also entitled to say that the rules for the identification of phenomenal states are highly probable, since this probability is very well confirmed in an indirect way by a multitude of systematically related associations between interpersonally accessible physical phenomena and reports of internal phenomenal occurrences. For example: if wrinkling the forehead is often associated with repeating the statement ‘I feel pain’ when one believes one has a feeling x, wrinkling the forehead indirectly reinforces the probability that when applied the words will really refer to the same feeling – even if only minimally.[19] Our case is not different from the case of concluding, based on a large amount of convincing circumstantial (indirect) evidence, that a person was in fact murdered by a psychopath. Even if no one actually saw the murder take place, a great quantity of circumstantial evidence could be rightly seen by a jury as inductively mutually reinforcing, and taken together as highly convincing. (cf. Costa 2011, Ch. 5)

14. Concluding remarks
Returning to our initial question about the nature of the intermediate link, we can now see more clearly why and how the intermediate link between words and things can be read in two different complementary modes. These are the psychological mode, which considers some particular cognitive bearer of the link, and the semantic mode, in which particular bearers of a link and their psychological particularities are left aside.[20] That is: cognitive meanings are semantic-cognitive rules that can be considered in their possible or effective application and that when regarded, thinking of their conditions of satisfaction, can be called semantic-cognitive criterial rules. As will be seen in chapter V, the cognitive meaning of a statement should be nothing but a verifiability rule that really applies when some criterial configuration required by it is adequately satisfied, making the statement true.[21]
   Nonetheless, it is important to maintain a clear distinction between the semantic and the psychological aspects of the intermediate link, as philosophers like Frege and Husserl insisted, even if only in a misleading way. The semantic aspect is conventionally grounded and grammatically necessary; the psychological aspect is spatio-temporally given, and in its physical particularity contingent. But contrary to what these philosophers have supposed, nothing semantic can really exist outside of cognitive instantiations. Semantic entities are nothing more than conventional structures that exist only when embodied in mental acts, in applications, even if considered in abstraction from their contingent bearers. To assume that semantic entities can exist without any psychological basis is to hypostasize their nature.[22]




[1] This semantic versus cognitive dichotomy can be traced at least as far back as Aristotle. The latter viewed the intermediary link as an affectation of the soul (ton en têi psychêi pathêmáton) or thought (noêmata) – a psychological perspective – while the Stoics, who appealed to ‘what is said’ (lectón) or ‘what is meant’ (semainómenon), associated the intermediary link in some way with language – a semanticist view. (Manetti 1993: 93 ff.)
[2] Qualitative identity is the identity between different things; it is opposed to numeric identity, which is the identity of a thing with itself.
[3] Of course, one could also do the same thing without drawing on color memory: suppose that people carry with them templates of vermilion and when necessary compare the patches of color they see with these templates. This shows the indispensability of the existence of some empirically given model and the complementary role of memory.
[4] It is true that this last ‘any’ allows us to infer that there is a class, called the class of all tokens that are qualitatively identical, but that this class does not belong to the definition and does not need to be known by anyone.
[5] As Wittgenstein wrote in the Tractatus: ‘The name means its object. The object is its meaning.’ (1093c sec. 3.203)
[6] The view was ironized by Gilbert Ryle as the ‘Fido-Fido’ theory of meaning (1957).
[7] As Russell recognizes, logical atomism was first suggested by Wittgenstein, who defended it in a full-fledged way in his Tractatus Logico-Philosophicus.
[8] This kind of difficulty already arises in the last public discussion of Russell’s speech in ‘The Philosophy of Logical Atomism’, 1994: 203. (For criticisms see Tugendhat 1976: 382 and Kripke 2013: 15-16.)
[9] See my discussion of Wittgenstein’s private language argument later in this chapter.
[10] One could object that since there are many different shades of red (one of them being vermilion), red cannot be simple. But with Wittgenstein we can answer that what we call ‘simple’ depends on whatever system we have adopted: we can use an old language-game with only three basic colors: red, yellow and blue. Here red will be considered simple; and in this case, distinct shades of red will not be taken into account, even if they are perceptually distinguishable. Instead of being qualitatively identical to the pattern, a new red patch must only be sufficiently identical, insofar as it is possible to determine its limits using the other two colors.
[11] Language not only has a communicational function, but also an organizational function, in the sense that we also use it to think, to organize our ideas and our plans of action (Vygotsky). At first sight, the identification of meaning with use doesn’t seem to do justice to its organizational function. But it doesn’t have to do so. It makes sense to say that when I think that the Leaning Tower of Pizza could come crashing down, I am using this name referentially in my mind, in thought, that is, in an internal dialogue with myself.
[12] Although Wittgenstein expressly disliked K. Ogden and I. A. Richards’ book ‘The Meaning of Meaning’, he must have appreciated the short supplement of the book written by Malinowski where these ideas are presented.
[13] See (Costa 1995, Ch. 1). The assumption that grounds my reconstruction is that Wittgenstein was not making repeated attempts to explain the nature of meaning which always ended in some kind of failure erratically followed by new attempts, as some interpreters seem to believe. What he did was to develop different approximations, often analogical suggestions, each addressing the same issue from a new perspective, such suggestions being largely complementary. In this way, it is possible to find continuity in Wittgenstein’s semantic conceptions, which began with the Notebooks 1914-1916 and ended with On Certainty.
[14] I outlined what I believe to be the most workable solution in Costa 2011, Ch. 5.
[15] These two forms were already noted by Anthony Kenny (1973).
[16] C. S. Peirce’s view, according to which all thought is in signs, seems to be wrong when we consider that we are surely able to think without using words. But it is plausible that in having these thoughts we are unconsciously using signs that are, if not linguistic, at least imagistic or emotive.
[17] Although I still believe my interpretation is the most faithful to a central line of Wittgenstein’s thought.
[18] See, for instance, A. J. Ayer 1972: 196.
[19]  Against this, Wittgenstein imagines a situation in which when he believes he has a particular sensation P the manometer always shows that his blood pressure has increased. This assures a correlation between his thinking he has the feeling P and an increase in his blood pressure; but it does not guarantee that the rise in his blood pressure will be correlated with the same sensation on various different occasions (1984c, sec. 270). Indeed, it does not guarantee that the last correlation will be the same, but we feel that in small measure it increases the probability that P is being correctly correlated with the same feeling. As we normally have a very great interweaving of such correlations, in the end we reach a very high probability of associating the belief that we are having e.g. ‘pain’ with the pain that we really feel.
[20] While semantic theories like that of Davidson fall short of the mark, the Gricean psychological theory of meaning overshoots the mark. H. P. Grice suggests that to display what he calls a non-natural meaning (our semantic-cognitive meaning) of p the speaker must have (i) the intention that the hearer should come to believe that p, (ii) that the hearer should recognize the intention (i) of the speaker, and (iii) that by means of the recognition (ii) the hearer will come to believe that p. However, what Grice thereby analyzes is not the non-natural meaning in itself, but only the standard procedure by which the non-natural meaning is communicated. (cf. Grice 1991, Ch. 5, 6, 14, 18).
[21] Note that there are non-referential cognitive rules: we can have rules that relate (a) the empirical data to cognitions, (b) cognitions to other cognitions, and (c) cognitions to actions. But concerning the issue of reference, what matters is the first kind of rule, which is responsible for referential meaning.
[22] As I see it, there is a great variety of ways to make this hypostasis. One of them is to identify sense/meaning with Platonic entities (Frege, Husserl); another (already criticized in the Appendix to Chapter II) is to identify linguistic meaning with something external like essences of things (Putnam); another is to identify meaning with minimum units of reference (Russell); and yet another is the attempt to identify meaning with psychological communicative intentions (Grice).