Nonsense and the Privacy of Sensation
Juan José Acero
Among the many threads that make up the so-called Private Language Argument (PLA, in what follows) comes to the conclusion that the philosopher who gives credence to the possibility of a private language does not have an insight into the proper function of words like `pain', `I', `know' and a few others. This is what I will call the Epistemic Privacy Way.Foot note 1 The specific argument that follows this route lays the blame on such a philosopher for thinking that there are linguistic vehicles -- I will restrict my attention to certain sentences -- that are apt to convey incorrigible and immediate knowledge of our own minds. It is alleged that `I know I am in pain', `But I must know if I am in pain!' or `Only you can know what you feel', among many others, belong to the kind of sentences that can be used in this manner. In contrast to such a view, it has been adduced that to recruit those sentences for the purpose of serving the demands of an infallibilist view of inner perception is, as Wittgenstein showed many years ago, doomed to failure. There is no such a use, and philosophers that are not aware of it end up speaking nonsense, i.e. depriving those sentences of the function speakers require from them. The Epistemic Privacy Way thus leaves room enough for language analysis to play a role in critical philosophy and highlights the fact that not everything in PLA depends upon some or other subtlety, whether metaphysical or epistemological. As is well known, Wittgenstein considered that the philosopher's task consists of bringing words back from their metaphysical to their everyday use (PhI, § 116), and the Epistemic Privacy Way nicely fits in with this metaphilosophical advice.Foot note 2
On the other hand, if we aim at explaining why the private language advocate does not succeed in bending language -- that is, the part resorted to -- to his or her will, a detailed account is due that pinpoints where the philosophical mistake lies. Which amounts to spelling out a view of nonsense that squares with what PLA ends up. In other words, it is one thing to embrace the Epistemic Privacy Way and to agree that the believer in the possibility of a private language has not succeeded in endowing his or her claims with sense; it is another thing to found such a criticism on a systematic view of nonsense. This is what this paper aims at. Accordingly, in § 1 the view of nonsense expounded by Cora Diamond in her book The Realistic Spirit (Diamond 1990) will be spelled out. In § 2 I will examine this more in-depth in order to emphasize its affinity to a certain condition, the Meaning Restricted Exportability Requirement (MRE), which contains the essentials of Diamond's analysis of nonsense. I will also argue, in contrast to Diamond, that MRE is compatible with the Meaning Compositionality Principle, according to which the meaning of a complex expression is a function of the meaning of their parts and of the expression's constituent structure. In § 3 two aspects of MRE will be unfolded, its syntagmatic and its paradigmatic dimensions, in order to pave the way to acknowledging two kinds of relationships concerning the exportability of sentences and their constituents. In § 4 I will pause to take up an objection aimed at applying Diamond's view of nonsense to cases very similar to the ones that matter in ALP. Answering the objection hinges upon distinguishing meaning from thought, a distinction that echoes back to Wittgenstein's distinction between depth and surface grammar and at the same time sheds light on the sort of temptation to which philosophers are prone. Finally, in § 5, I reconstruct the steps to be followed by anyone adhering to the Epistemic Privacy Way and advance the main thesis in this paper, namely, that the possibility of a private language is an idea fuelled by the failure to comply with the paradigmatic exportability constraints to which certain constituents of sentences and thoughts are subjected.
1 Diamond on nonsense
In «What Nonsense Might Be» Cora Diamond offers a weighted answer to the question of what nonsense consists in. Nonsense has, in principle, two sources. Sentences lack sense either because of a categorical error or because of the presence of a constituent that bears no meaning or sense; and thoughts lack content because of either a categorical error or the presence of a constituent devoid of it. According to the first attempt at an explanation, nonsense is the outcome of a categorical incompatibility between the meanings of two constituents. Both sentences and thoughts are complex, articulated entities, and the composition of their constituents' meanings and contents must be made in accordance with rules that fix which combinations are permitted and which others are forbidden. Any failure to be in line with those rules results in a lack of sense. This is what Diamond calls the natural view of nonsense. According to the second source of nonsense, not having been successful in providing a sentential or thought constituent with a meaning or content, i.e. not having managed to give it a use in the context of the sentence or the thought where the constituents occurs, is what accounts for the lack of sense. This view of nonsense Diamond attributes to both Frege and Wittgenstein, and that is the view she herself adopts. Each view has its own banner. The first one waves in sentences like (C) (for Carnap); the second, in sentences like (M) (for Moore):
(C) Caesar is a prime number
(D) Scott kept a runcible at Abbotsford
While (C) illustrates the first kind of deficiency, the second one is discerned in (D). Of course, this is not the final analysis, because the advocates both of the natural and of the Frege-Wittgenstein views mean to give a general diagnosis of where nonsense originates, which forces them to justify why the other's banner could be waved in his or her own parade.
Diamond is on Frege and Wittgenstein's side. This view's very idea is that, regardless of the appearances, the lack of sense of sentences like (C) is due to the same principle that explains the senselessness of (D) and similar cases. The natural view, Diamond points out, confronts the difficult prospect of construing cases like (D) as examples of categorical mistakes -- not an easy task since `runcible' is not an English word and a fortiori has no meaning or content that can contribute to any sentence it could occur in. The Frege-Wittgenstein view, on the other hand, is adequately equipped to construe (C) in harmony with what (D) is supposed to show. Diamond's argument for this depends on Frege's Context Principle, that is, the claim that a word has meaning only as a constituent of a sentence (alternatively, that a concept only contributes a content to a thought only as a constituent or a part of such a thoughtFoot note 3).
According to this, (M) is nonsense because no rule has been stated that assigns `runcible' a meaning, and this amounts to saying that no rule has been laid down that makes `runcible' either a meaningful constituent or provides it with a syntactic function vis-á-vis the remaining constituents which make up what seems to be a sentence. Therefore, its occurrence in (M), alongside with `Scott kept a ___ at Abbotsford', does not therefore result in a meaningful sentence. No thought could find expression in this combination of words. Now, what is remarkable, according to Diamond, is that the same diagnosis can be made of (C). Though (1) is perfectly in order,
(1) Caesar crossed the Rubicon
it does not follow from that fact that any of its constituents, i.e. `Caesar', has to bear the same meaning anywhere else, independently of its being a constituent of that sentence. The step that mediates between the premise that in (1) `Caesar' names the history character, the Roman Julius Caesar, and the conclusion that `Caesar' also names that Roman dictator in (C) is wrongly taken. When no rule has been enforced that adjusts the role played by `Caesar' vis-à-vis `___ is a prime number', no meaning can be assigned to such a name, i.e. no sense can be made of it, in such a linguistic context. (C) lacks sense, not because any clash of semantic categories, not because arithmetical properties cannot be predicated of human beings -- as the natural view of nonsense would suggest, but because the sort of imbalance between `Caesar' and `___ is a prime number' that exists in (C) is the same you find in (M) between `runcible' and `Scott kept a ___ at Abbotsford'. As it happens, there is no need for postulating two sources of nonsense, one for (C) and another one for (M).
It is worth stressing that Diamond's diagnosis smoothly squares with an idea of meaning and content Wittgenstein made popular, namely, that the meaning of a word is its use in language, and that the content of a concept is its use in thought, i.e. in judgement, inference, deliberation and so on. Words are tools and should be understood as such. Despite the fact that the analogy is widely acknowledged nowadays, one of its consequences has to be highlighted one again. Only as long as they are properly connected to further words they can contribute a meaning to a sentence, in the same way as the `t' key on my computer keyboard can add a `t' token to a text on pressing it, if it is properly connected to rest of the system. Its role is played in so far as it is part of the sentence gear assembly. If we take the analogy seriously, then the Fregean Context Principle becomes a natural restriction to abide by in searching for word meaning and concept content. Words are like cogwheels or levers in a machine, and for them to have a meaning to contribute to a sentence requires them to be geared with the rest of sentence constituents.
Therefore, Diamond's defence of the Frege-Wittgenstein view of nonsense exploits the virtues of the Context Principle to untie the knot of nonsense. It is now evident that her relying on the Context Principle to explain how nonsense emerges is in line with a use view of what meaning and content understanding consists in.
The next step is that of extending the Frege-Wittgenstein view to examples of nonsense much less apparent that either (C) or (M). They combine words in a manner that to decree their non-sentencehood would be to fly in the face of common sense -- though, needless to say, common sense would be now informed by linguistic theory. They are sentences, but for a number of reasons it can be argued -- in fact, it has been argued -- that they might not express thoughts. That is, there is no guarantee that any of their utterances is the vehicle of a thought. Put briefly, their part-time nonsensical nature does not immediately strike us. Consider the following examples for discussion:
(2) It is 5 o'clock p.m. on the Sun
(3) Sapin is above New Zealand
(4) I know I am in pain
(5) He is in pain or he isn't
(6) Milu believes that Tintin is in danger
To begin with, I am almost certain that in today's world we can think that it is 5 o'clock on the Sun. However, the same assurance cannot be credited anyone who believed in the same thought eighty or one hundred years ago, when the electronic or atomic clock technologies had not been developed and Earth rotation was measured according to the Sun's apparent movement. (Actually, this is not the whole truth, as the knowledge of how to calculate the time on the Sun by means of pendulum's movement dates from ancient times. But, again, the position of the Sun over the horizon played a role in making the usual estimates.) The reason for such an impossibility is that in order for us to think that it is 5 o'clock on the Sun, it is necessary to think that the Sun keeps with itself the kind of relation that holds when it is 5 o'clock p.m.; that is, the kind of relation that would hold if the Sun could be seen from the Sun's very surface as occupying the 5 o'clock position! Even if someone were to assure us that he or she is able to think or to be thinking such a thought, we should insist that (1) expresses an apparent thought. To be in position to think that thought, one has to have acquired a rather sophisticated intellectual training. Somewhat less demanding are the conditions that control the expression of thoughts by utterances of (3). In this case, the difficulty is made apparent by the relational predicate `is above'. No ordinary English speaker hesitates to tell whether the picture is above the chimney or the clouds are above the mountains, if the circumstances are the usual ones. On the other hand, to overstate the use of `above', moving away from the norm designated by those central cases, to the point of demanding a yes-or-no answer to whether Spain is above New Zealand is to drive us over the edge of nonsense. If anyone utters (3) while thinking that New Zealand is on the antipodes of Spain, then to resort to the predicate `above' can be interpreted as the choice of an austere and common mean instead of a more demanding one. Maybe the speaker is about to make a joke. However, `to be on the antipodes of' is a complex relational predicate whose use is subject to rather strict conditions. If all one takes into account when uttering (3) is a quick glance at a map of the world globe that prompts in your mind an image of Spain's being over New Zealand, then nothing warrants that a real thought has been framed. I would say that, if the only means available to that speaker is an image together with the usual linguistic resources, such a speaker could well fail to articulate a thought. Clearly, this use of `above' significantly differs from that of `the picture is above the chimney'. Should it be true that the picture is above the chimney, then it follows that it has to be false that the chimney is above the picture. But if someone feels free to claim that Spain is above New Zealand, then he or she could add that New Zealand is above Spain as well. The fact that we have here different uses of `above' and, correspondingly, different concepts of spatial relations, but only one word to cover them, gives rise to an illusion, namely, that it will suffice to entertain a clear image to think a thought. As for (2) and (3) I would add that the standard linguistic and conceptual resources available to people who lack a specialised training, and that are unable to see through them, contain blind spots, and this explains why they occasionally produce apparent thoughts. An unreflective experience of the Sun has created a blind spot on the conceptual system with which we estimate time; and a superficial understanding of the nature of closed, curved surfaces has given rise to another blind spot, this time on the conceptual system we use to map spatial relations.
Wittgenstein, who is the source of the problem (2) and (3) pose,Foot note 4 believed that the reason why they could express apparent thoughts makes clear why (4) and (5) could lack sense as well. In these cases, nevertheless, the question whether they make sense or not gets mixed up with the question of whose side you are on in some debates in philosophy of logic and mathematics -- is the Law of Excluded Middle valid? -- as well as of philosophy of mind -- are incorrigible your judgements on your mental states and processes? This relation makes it reasonable for us to take precautions against focusing on such sentences if we seek to settle the question whether they are senseless or not. Accordingly, I will follow the strategy of putting them aside for the time being, aiming at a general analysis along elsewhere and then trying to see what might follow for the cases in point. But this in no way means that I would welcome to set up a principled distinction between, say, (6), and (2) or (3). I am not. Hergé and Goscigny repeatedly picture situations in which Milu thinks that Tintin is in danger. If we did think that dogs cannot frame a thought with the conceptual and behavioural complexity required to have the content that a certain person is in danger -- no such content is possible without concepts, and concepts either are or evince abilities dogs do not have, it would follow not only that (6) is false, but also that it is a nonsensical sentence.
So, what can be said about (2) and (3) from the Frege-Wittgenstgein view of nonsense? In case the final answer should be that none of them makes sense, the corresponding justification would have to mirror the kind of argument Diamond provides for (C) and (M) that we have put forward above. This task must confront the difficulty of bringing a dislocated item in (2) and (3) to view, as happens with `runcible' in (M). This does not seem to be a minor point, as such an item seems to be absent in (2) and (3). However, Diamond's penetrating suggestion is not that nonsense originates in the presence of so-called runcible words and items. The point about these cases is -- exploiting an analogy with jewellery craftmanship -- whether each and every word is mounted on the sentence the way they should, not whether they are already acknowledged in language. (Of course, if it is not, they cannot be fitted with further words or constituents in a sentence frame.)Foot note 5 Thus, turning to (2), the question that has to be answered is whether we have managed to provide `the Sun' with a use vis-à-vis the rest of constituents. As for (3) the question is: Does `is above' have a meaning vis-à-vis the rest of the sentence constituents? Both questions must be answered in the negative. The design of a linguistic framework to state time estimates no matter where one might be situated on the Earth must entertain the possibility of combining a name of such a place (`Mexico, D.F.', `Granada', `Helsinki') with the predicable `It is 5 o'clock on ___' while implicitly barring `the Sun' from being taken as a legitimate choice. (The conscious inclusion of `the Sun' in the list means that another point of reference to estimate solar time has been adopted.) Why not interpret this condition as a proof that `the Sun' bears no meaning when inserted in the predicable blank, i.e. when it occurs in the sentence (2)? An analogous argument could be run that would justify why by uttering (3) someone did not succeed in expressing a thought. (3)'s lacking sense in some occasions of their use is due to the fact that, far from being true that (3) results from importing the relational constituent `is above' from, let's say, `The picture is above the chimney' to the predicable `Spain ___New Zealand', `is above' cannot be mounted on the latter. In the same way, (6) could be involved in nonsensical chatter, because the proper name `Milu', unlike `Captain Haddock' cannot be set off for the predicable `___ thinks that Tintin is in danger'. Following Frege and Wittgenstein, Diamond ought to opt for this way of analyzing (2), (3) and (6). Though I share this analysis, in the next section I will set out where my disagreement with her lies. Every word and every constituent is a runcible somewhere (sometime).
2 Restricted Exportability
One reason why I believe that Diamond's view on nonsense should be seriously taken into consideration is its flexibility. Though she does not seem to be interested in the kind of nonsense (2) -- (6) illustrate, her analysis of nonsense can be extended to them in a natural way. Moreover, her analysis squares with the existence of blind spots in our conceptual system and promises to shed light on their nature. On the other hand, Diamond claims, finding inspiration in the work of both Frege and Wittgenstein, that nonsense is what one should expect to obtain if the Functionality (or Compositionality) Principle is applied unrestrictedly. This principle holds that the meaning of a complex expression, i.e. a sentence, is a function of the meaning of their constituents and of the syntactical mode of combining them, i.e. of their syntagmatic relations. She interprets this condition as requiring that the meaning of a complex expression is a complex extralinguistic entity made up of the meaning of the expression's constituents and their mode of syntactic combination. Therefore, nonsense is what results if the meanings of the constituents do not fit into each other in the right way, something that takes place when rules of semantic combination do not operate on the required arguments. The Compositionality Principle, argues Diamond, is tied to the natural view of nonsense:
`México D.F.', `Granada' and `Tombuctú' are names that can be combined with the predicable `It is 5 o'clock p.m. on ___'. `The Sun' cannot. And an analogous remark is valid of `is above' and `Spain ___ New Zealand'. The fact that these constituents make a semantic contribution to any sentence they might occur in does not support the conclusion that they can do so in any other sentences as long as syntax and morphology requirements are complied with. As I see it, the most suitable way of making this pointFoot note 6 has it that words and concepts acquire an use in more or less specific or paradigmatic circumstances, and that such a use develops into new ones as words get anchored to new situations, which to a certain extent may be similar to the former ones. I take it that we first learn to say and think that it is 11 a.m. 5 in Granada, 5 p.m. in Helsinki and so on; to learn and think that the picture is above the chimney, that Los Angeles is above San Andres fault, etc. However, as soon as we move beyond the circumstances in which the target expressions receive a use, the risk of giving rise to an apparent thought becomes real. And this is what in fact transpires in both (2) and (3). No word or expression carries with itself the meaning it has in a sentence or phrase to any other linguistic context.
It is worth dwelling on this point, because Diamond seems to go into a dead end in making (M) the measure of nonsense, i.e. in claiming that nonsense results from having failed or forgotten to mount an expression on a sentence with the rest of its constituents. The Compositionality Principle certifies that the meaning of a sentence is a function of the meaning of its constituents and of their syntagmatic relations. Once these meanings and relations are fixed, the meaning of the sentence is univocally determined. By invoking the principle we gain the right to assert that (M) lacks sense because one of its constituents either is senseless or its syntagmatic relations to other constituents in the sentence have not been settled. If this explanation has to guide us, it would follow either that (2) and (3) are perfectly in order -- an uncomfortable conclusion -- or that they lack sense for the same reason that (C) does. Thus, the Compositionality Principle traces the analysis back to the natural view while denying the Frege-Wittgenstein view. Because of it, she is doomed to rejecting a compositional approach to meaning and content, and particularly the condition, built into the compositional approach, that semantic combinatory rules apply across the board, that is, no matter what constituents are acknowledged in the lexicon. Put alternatively, Diamond opposes the following
Meaning Free Exportability Requirement (MFE)
Diamond openly rejects that words should be mapped onto grammatical categories, and thus that their meanings should fit into modes of compositions, independently of the contexts they might occur in. Neither word's meanings nor concept's content are freely exportable from contexts to contexts. Sentences (2) and (3) deserve close attention, since they reveal that MFE is all of a piece with a way of understanding language and thought that generates unexpected troubles. It is the so-called mechanical conception of meaning and thinking.Foot note 7 Diamond rejects the Compositionality Principle because she conceives of it as a mechanism, though described in abstract terms, which compounds meanings and thoughts out of simpler meanings and concepts. Those outcomes are the values of a complex function which take as its arguments the meanings of constituents and whatever syntactical categories they belong to. According to such a conception the Compositionality Principle does its best when taken as psychic mechanism, maybe endowed with a computational nature. Seen to this light, the principle plays an essential role within a representational conception of mind and language. Which thought we may think, which content it may have, will at last depend on what representations the thought is made up of, what syntactical patterns these representations fit into, and what meanings or contents they have. All these pieces are put together by the Compositionality Principle, which accordingly plays a centrally strategic function in depicting human mind as a sort of mechanism. In Baker and Hacker's opinion, it is an essential part of a calculus view of language as much as of a generative theory of meaning understanding. Regardless of whether sentence meanings are made up of the meaning of its constituents or simply are functionally determined by them,Foot note 8 subject to their mode of syntactical composition, either to understand a sentence or to grasp a thought consists of calculating a value out of the value of its parts and their syntactical relationships.Foot note 9
To my mind, Diamond's rejects the Compositionality Principle because she makes an unnecessarily restricted reading of it. This becomes plain when she explains how she believes our knowledge of language rules, which governs not only the meaning of its expressions but their combinatorial possibilities as well, should not be understood. After making it clear that meanings and syntactical categories are not unconditionally exportable, she points out that language rules do not say what role expressions play in sentences, including their semantic contribution:
Diamond's alternative to a conception of rules that views them as content containers is a conception that understands rules as means of analyzing the expressions they govern. The analysis carries out two tasks: rules split up sentences -- complex expressions -- into pieces or parts, i.e. names, predicates, relation terms, and so on, and fix their meanings as constituents of sentences. What really matters, she says, is that «neither kind of rule will apply unconditionally to a given sentence» (Diamond 1990: 109&f.). When she comes down to the last details, she insists on the idea that language rules capture what language users do with words and structures, instead of preceding those uses in the language constitution order. Thus, when tackling the sentence `Venus is more massive than Mercury', she writes:
Nevertheless, I see nothing here that an advocate of the Compositionality Principle may be not ready to embrace. Such a theoretician would refuse to accept the view that, for example, the constituent `got burnt' makes the same contribution both to (7a) and (7b):
(7a) The forest got burnt (= Some burnt the forest)
(7b) The forest got burnt (= The forest was consumed by the fire)
In order to mark and expose differences of meaning or thought that could matter, the believer in compositionality will look for syntactic categories, appropriate assignments of items to them and the corresponding semantic combinatorial rules that provide (7a) and (7b) with different meanings, i.e. truth-conditions. Therefore, the analysis put forward by the theoretician will keep an eye trained on (7a) and (7b) and the other one trained on the meanings they respectively express, aiming at the conditionality requirement, to which the application of language rules are subject, as Diamond insists. If follows that the Compositionality Principle can be read in a manner that it does not become part and parcel of the mechanical conception of thinking and language understanding. On the contrary, the Compositionality Principle leads to the acknowledgement of systematic relations, of a functional nature, between the meanings of complex expressions and the meanings of their constituents. It simply claims that the meaning of a sentence, the meaning of their constituents and its structure are systematically related.Foot note 10 No more demanding requirements should be allowed to bear upon it. Moreover, when so interpreted, the conflict between the Compositionality Principle and the Context Principle, which some philosophers have pointed out, smoothly dissolves. On the one hand, the Compositionality Principle is addressed to the conditions that rule the links between the meaning of any complex expression, particularly any sentence, and the meanings of its constituents. On the other hand, in its most natural interpretation the Context Principle claims that words and expressions in general can contribute a meaning as long as they are sentence constituents. Thus, no unbearable tension results from these balanced readings of both principles.Foot note 11
Once the Compositionality Principle is given an explanatory role to play independently of the mechanical conception of thinking and meaning, Diamond's objections to the natural view of nonsense become arguments against MFE. And there is nothing in meaning compositionality that stands in the way of replacing that constraint by a weaker one. The following one is a natural choice:
Meaning Restricted Exportability Constraint (MRE)
Of course, `appropriate to its constituent' is the key clause in MRE. It alludes to nothing that contemporary grammarians ignore nowadays. Far from being true, it is a significant part of their business. By assigning words and expressions to syntactic categories in a systematic way, they attempt not only to capture, but also to anticipate, which contexts are appropriate to which items. Those assignments provide one sort of standard to catch those adjustments among expressions that have to be preserved and those that must be ruled out. Another standard lies in the subtleties involved in syntactic and semantic composition, because they become a powerful instrument to fix those contextual restrictions we deem necessary. Thus, the challenge that (7a) and (7b) pose to us can be met as soon as we make it sure that compositionality and contextuality requirements neatly square with each other. This calls for two arrangements. Firstly, the verb `to burn' is assigned to two different syntactic categories, one reserved to medial verbs and the other one to non-medial verbs. This step is taken in accordance with the spirit of the Context Principle. Secondly, the rules that govern meaning composition and in particular the contribution of the verb `to burn' must be redesigned to answer the fact that `The forest got burnt' is ambiguous.Foot note 12
3 Two dimensions of meaning exportability
To deal with the kind of nonsense that crops up along the Epistemic Privacy Way, it is necessary to have a closer look to MRE and figure out what makes a context an appropriate or inappropriate one to import either a word or a sentence constituent. I will claim that meaning exportability can be assessed along two dimensions, in order to be faithful to the fact that two kinds of valuations have be taken into account to understand a word's disposition to fit into a sentential context gear assembly. I will borrow a distinction, which has been repeatedly tried out in structural linguistics, to characterize those dimensions and give them a name and speak of syntagmatic and paradigmatic relations (Lyons 1968: 70&ff.). The justification for borrowing such a pair of terms lies in its utility to the analysis of (4) to be carried out below (in § 5).
Compare (8) to (9):
(8) I don't know whether there is a red book on the shelf.
(9) I do not know whether I have ever been in China
Wittgenstein rejects that (9) will make any sense in normal circumstances: «Normally Europeans do know whether they have been in China or not» (OC, § 333). While the sentential context (› 9) is not appropriate for the expression `I do not whether' to be exported to it, context (ç 8) is:
(ç 8) ___ there is a red book on the shelf
(ç 9) ___ I have ever been in China
`I do not know whether' is exportable to (ç 8), but it is not exportable to (ç 9). I will say, accordingly, that by not loosing sight of its exportation behaviour, we gain an insight into how different the syntagmatic relation it maintains to (ç 8) is if compared to those it holds with (ç 9). Those differences help explaining why (8) may be used to express a thought while (9) cannot. The exportability of `I do not know whether' to (ç 8) makes it possible to express a thought; the failure to exporting `I do not know whether' from (8) to (ç 9) explain why (9) is nonsense, i.e. why it usually cannot be used to express a real thought. Why certain words or expressions are not liable to be exported to specific contexts, while others are, is an important and difficult question. The Frege-Wittgenstein view of nonsense has it that exportability is restricted to those cases in which the exportable item has not been provided with a use vis-à-vis the target context; in other words, when language and thought use admit of mounting the moving item on the target frame. On this dimension what counts is whether these linguistic and conceptual gears successfully work or not.Foot note 13
In caring about paradigmatic relations we pay attention to quite a different sort of concern. We do not consider a moving item Ej relatively to a target context C, but how does Ej behave relatively to a number of further exportable items E1,... Ej-1, Ej+1,... En with respect to C. Any difference between two members of this class with respect to C -- one of them can be exportable to C, while another one cannot -- is a difference in their paradigmatic relations. Alternatively, we could say that they have different paradigmatic properties. Wittgenstein made the most of paradigmatic relations in his sustained battles against the bewitchment of our intelligence by means of language:
Here he is suggesting that the use of `I do not know whether' is appropriate when, i.e. in those contexts in which, it is also appropriate to use `I have reasons to doubt that'. It follows from this constraint that both `I do not know whether' and use `I have reasons to doubt that' are not exportable to either (ç 10) and (ç 11):
(ç 10) ___ I came into the world without parents
(ç 11) ___ cats do not grow on trees
If we are not sensible to these nuances, we run the risk of mistaking apparent thoughts for real ones, that is, mistaking (10) and (11) for meaningful sentences:Foot note 14
(10) I know I came into the world without parents
(11) I know cats grow on trees
And conversely: that any context to which `I do not know whether' may be exported is also a context to which `I have reasons to doubt' may be exported. In the structural linguist's jargon, whether `I know' and `I have reasons to doubt' belong to the same paradigm.
The fact that words and expressions share paradigms is something philosophically worthy of attention, because it offers clues as to the range of application of our concepts, as well as to the range of our words. It is not possible to export `I do not know whether' either to (ç 9), (ç 10) or (ç 11), because ignorance and doubt are somewhat correlative: one is a natural move to make in language use just where the other one is a possible choice. Thus, if there is room for doubting whether such-and-such is the case or not, it must be possible to ignore that such-and-such. Doubt can take root only where the possibility of ignorance is built into. Conversely, nobody can ignore that such-and-such unless the possibility of doubting whether such-and-such is built into the very situation in which our ignoring takes shape. It follows that to determine the scope and limits of meaning and concept exportability, we are forced not only to detect what compatibilities and incompatibilities happen to take place between exportable resources and target contexts, but to polish off the ability to identify when those resources belong to a certain paradigm and when they do not. We do not grasp the function of a concept unless we somewhat map its place in the thought system. Put it in less abstract terms, we must be acquainted with its syntagmatic and paradigmatic properties to be able to pinpoint what difference it means to the possible meaning and thinking.
It is evident, then, that what lies behind MRE's opaque clause `appropriate to that constituent' is the class of syntagmatic and paradigmatic relations any constituent enters into. It is also clear what is involved in looking at sentences like (2) and (3) through the emblematic (M). Those sentences make no sense because some of their constituents, i.e. `the Sun' and `is above', have not been provided with a use relatively to the rest of their respective constituents. Their syntagmatic and paradigmatic relations fix their functions' scope and limits. The fact that `the Sun' holds such a specific paradigmatic relation to names like `Mexico, D.F.', `Granada', `Helsinki', and so on speaks for itself of how delicate its place in the time estimation conceptual system is. And the fact that `is above' has got the very particular syntagmatic properties it exhibits speaks for itself of how much subtlety there is in the system we have designed to locate places on the globe.
4 Meaning and thought
I have argued why MRE states what is essential to the Frege-Wittgenstein-Diamond view of nonsense. The time has arrived to face up to a difficulty which has been looming large since the moment we addressed the question whether or not it possible for (M) to model (2) and (3)'s oddity. Up to now, I have tried to explain why (2) makes no sense, i.e. why an utterance of it may express no more than an apparent thought. However, it might be adduced not only that the possibility of there being a thought with the content that it is 5 o'clock p.m. on the Sun is an illusory one, but also that the sentence `It is 5 o'clock p.m. on the Sun' is in order according to both syntactic and semantic criteria. There might be no such a thought, but the sentence meaning is out there, ready to be translated to other languages (`Son las 5 de la tarde en el Sol' in Spanish, `On kello auringossa viideltä iltapäivällä' in Finnish). Translation preserves meanings, at least to a certain point, something that it would not be possible to achieve, if there were no such meanings to capture in different languages, at least to a certain point. Translation, when conceived of in a certain way and performed well, aspires to leave intact some sort of content or semantic value. However, this practice demands intellectual abilities that are, partially or wholly, independent of those required to frame a thought out of a concept repertoire. This remark can be made extensive to (3) -- (6). In conclusion, there is something to arguing that (2) does have a meaning, although it expresses no thought. Since the terms `meaning' and `sense' have been chosen to mark the difference between (1), on the other hand, and (C), (M) and (2) -- (6), on the other, are we not in the grip of an more or less explicit contradiction?
Throughout the discussion in §§ 1 -- 3 no distinction was made between a sentence's capacity to have a meaning and its capacity to express a thought. Moreover, a serious commitment was advanced -- see note 4 above -- namely, that whatever would be claimed on words and sentences when dealing with nonsensical sentences would also have a straight counterpart within the domain of concepts and thoughts, so that it would be superfluous to break up the symmetry between the realms of language and thought. To claim that (2) could lack any sense in one or another of theirs utterances was another way of taking it as expressing no thought in those very occasions. The meaning that a sentence has in one occasion is the thought the sentence expresses in it. By adopting this equation we have managed to maintain the complexity of our discussion within reasonable clear bounds. Or so I hoped. In spite of it, the threat of contradiction is strong enough to modify our previous commitment without giving up the thesis that MRE is the cornerstone of nonsense analysis. The way-out consists of distinguishing thoughts from other kinds of semantic values. It will help alleviating the tension our first commitment gave rise to if we reserve the term `meaning' to predicate semantic properties and relations of words and sentences. (« -- What does `Son las 5 de la tarde en el Sol' mean? -- It means that it is 5 o'clock p.m. on the Sun.»)
Before hastening to welcome this reaction, it could be argued that the objection is based on a premise that concedes too much. It could be alleged that translation only makes sense if it is of a bit of speech or of a text that contains or expresses thoughts; in other words, that thought is what makes translation a serious business. When we discuss what people and texts say, the objects of everybody's assertions are the same as the objects of everybody's thinkings. It cannot be asserted that p unless it can also be thought that p, and conversely. Since there could be no thinking that it is 5 o'clock p.m. on the Sun, there could be no saying it too. Any language that created the room for the expression of such a thought would be an illogical language, as would one which allowed to say that the execution of Charles I is in the crater of Vesuvius:
I take Diamond to be claiming not only that there is no saying that the execution of Charles I is in the crater of Vesuvius or that it is 5 o'clock p.m. on the Sun, but that those things cannot be thought as well. Therefore, (2) and (3) are no more strings of words than they are vehicles of thought. Taken at first value, they look like sentences, but they are not really. If we remember that Diamond thinks that (2) and (3) are like (M), we can infer from her words that the argument which adduced that there is something in (2) that makes it suitable for translation does not touch her at all. Not being sentences, why worry about it? (It would be tantamount to playing the game of translating into Spanish Chomsky's famous example `Colourless green ideas furiously sleep', at most an extreme and degenerate case of translation.)
I part from this reply, because there is a sense of `saying' according to which someone can say that it is 5 o'clock p.m. though it is overlooked that there could be no thought behind his or her saying it. It seems to me that it is perfectly possible for anyone to be blind to nuances built into the estimation of time conventions and think it perfectly normal to ask what time is it on the Sun. If he or she does it and the speech occasion is not a scientifically sophisticated one, it becomes appropriate to object that his or her words lack any sense. It is natural to conclude, then, that assertions and thoughts do not have the same objects, though these objects may often coincide. This departure from the argument line developed until now does not forces us to give up the conclusion that nonsense results from a failure to mount an item on an appropriate conceptual frame. Grammar constraints do not have to be sensitive to each and every condition which features in the conceptual systems we employ. In spite of the fact that our sentences are usually adequate guides to our thought' contents and that, because of it, in setting up the limits of meaning we are more or less mapping the limits of our thought, it does not follow that sentence meanings must strictly keep track of thought's junctures. (2) and (3) are worth taking into consideration since they show that such a kind of gap is not merely a theoretical possibility. Explaining (2)'s oddity is laying out in what respect language only partially mirrors thought.
It is not necessary to enter now into sa ystematic reflection that justifies the need for distinguishing thought from sentence meaning. From the point of view adopted here, the border between those two domains is passed over on replacing MFE by MRE, i.e. on restricting meaning exportabiliby from certain sentential contexts to further sentential contexts. (These are also the limits beyond which the validity of the Compositionality Principle is in danger.) The reason on which MRE is founded -- one that has repeatedly emerged above and that will reappear in the next section, where we will be concerned with some Wittgenstein's arguments against the very possibility of a private language -- can be briefly stated in the following way: the limits to constituency exportation in cases like (2) and (3) are the limits to which the use of the resultant sentences are subject to. Unless there are use conditions for those sentences, they are vehicles of thought. Since (2) is useless in the circumstances described above, in expresses no thought in them, and nobody could think it is 5 o'clock p.m. in the Sun there. On the other hand, this sentence means the same as `Son las 5 de la tarde en el Sol'.Foot note 15
Wittgenstein could have welcomed a distinction close to the one I have put forward in separating meaning from thought. Those cases of nonsense we have been focusing our attention on typically reproduce an scheme in which words have a use as long as they go together in accordance with patterns that language reproduces again and again. That aspect of word use that becomes easily recognizable through sight and ear is something Wittgenstein refers to when he talks of word and sentence surface grammar:
However, there are aspects of language use that cannot be perceived either through sight or ear, aspects we must be acquainted with if words and sentences have to make sense to us, namely, those circumstances in which it is correct to use words and sentences, the consequences that follow from using them here and there, and so on. Mere sentence form is not enough to turn a string of words into a useful tool; put briefly, to have learnt the right technique of using it. Therefore, having access to a sentence syntactic structure, even to its logical form, and to the meaning of its constituents falls short of knowing its depth grammar.
Sentences (2) and (3), as well as `that the execution of Charles I is in the crater of Vesuvius' are word formations devoid of the necessary technique of use. We do not know what they are good for, what to do with them.Foot note 16 Their being somewhat meaningful does not meet their being able to express a thought.
If the gap between meaning and thought must be acknowledged, then the theory of meaning and the theory of thought aim at different targets.Foot note 17,Foot note 18 The discussion carried out in §§ 3 -- 4 belongs to the theory of thought and can now be rephrased in a manner -- backed by MRE -- that makes it clear what endeavour it is part and parcel of. Theory of meaning, and grammar generally, we might say as a short summary, lays down on phrase and sentence constituents less demanding and more subtle constraints than those the theory of thought enforces on any concept to be exportable to a predicable.
5 Epistemic privacy and conceptual exportability
The casual comments made on sentences (4) and (5) have got round what there is in them that matters. I have restricted myself to saying that they serve to reiterate those conclusions obtained from the analysis of (2) and (3). However, both (4) and (5) belong to the class of sentences the debate on the Private Language Argument (PLA) has centred around. In particular, (4) has been at the heart of attention. My purpose is to make plausible the thesis that one of Wittgenstein's argumentative lines in his PLA is to argue that (4) expresses no thought, if understood in the way the advocate of epistemic privacy intends. Thus, the attack on the Epistemic Privacy Way Wittgenstein mounts claims that thinking that only I can know whether I am in pain is an illusion due to overlooking those specific paradigmatic restrictions the expression `I know' is subject to. As in the case of (2) and (3), MRE lies at the bottom of the illusion. It follows from it that to understand the meaning of an expression as its use is the central finding we become aware of at the end of the Epistemic Privacy Way. I will elaborate on this by deploying exegetical evidence. Anyone who embraces Wittgenstein's criticism will see why sentences (4) and (5) are on the list in § 1, as vehicles of apparent thoughts. Since I have just distinguished meaning from thought, there is no alternative left but to accept that (4) and (5) are meaningful sentences, thus feeding the philosophical illusion that sensations are private objects on a parade with no more than one watcher. I hope to have shown in detail how this illusion's roots can be pulled up.
Wittgenstgein's PLA is a group of arguments that from different stances discuss and display the baselessness of maintaining that private languages are possible. By the private language of a subject S it is meant a language whose words denote S's private experiences, experiences nobody except S can be acquainted with in principle. As against this conceptual possibility -- firmly entrenched in the philosophical modern and contemporary tradition -- Wittgenstein essayed a number of arguments. It has been pointed out that the main argumentative line in PLA elaborates on the remark that to name anything, i.e. a sensation of pain, requires to competently follow complex and demanding techniques of name using; and that though use techniques share a family air, there being no essence of the naming relation, to name an experience is not a sui generis kind of naming. It seems to follow from this that the possibility of providing a private something, i.e. a sensation, with a name cannot be discounted. This amounts to acknowledging that in `naming a private object', the adjective `private' has to mean `incidentally hidden to other people', not `logical or conceptually impossible to be known by anyone else'.Foot note 19 Among the rest of PLA's argumentative lines there is one, the Epistemic Privacy Way, that concludes that sentences like (4) and (5) do not express the kind of thought the philosophers Wittgenstein confronts would like to speak aloud. My contention is that Wittgenstein's run along Epistemic Privacy Way can be reconstructed by making use of the kind of analysis I have put forward in §§ 2-3 and the subsequent interpretation of the distinction between surface and depth grammar. The way PLA relates to the Frege-Wittgenstein-Diamond view of nonsense is as I have just stated.Foot note 20
This is how Wittgenstein, in PhI, § 246, states the argument I am about to reconstruct:
The observations made in this remarkable passage close off a possible escape for the believer in the possibility of a private language might wish to follow. As is manifest, Wittgenstein faces the doctrine that our sensations are private, i.e. sui generis entities we have an experience of, everyone their own, in foro interno, entities whose existence is certain for the one who perceives them and to whom he or she can refer through acts of inner reference, but which no one else can know what they are like. I have already pointed out that PLA's main line objects to such a possibility that private sensations -- in the sense just fixed -- are not suitable for bringing the techniques of using name expressions to bear. The partisan of the Epistemic Privacy Way might now reply by insisting on the fact that each and every of us maintains an epistemic link, an immediate and incorrigible one, to our own experiences. It is this link that warrants the truth of the thought expressed by our uttering (4). While only you can know whether you are in pain, the rest of us can guess at it, conjecture it or whatever, but not being certain of it. Well, in the above passage Wittgenstein rejects two things about how (4) could be used. On the one hand, he rejects the idea that in using `to know' as people commonly do, any utterance of (4) expresses a false thought, since other people know whether I am in pain or not. On the other hand, the one that truly matters here, he insists that if a different use of `to know' is intended, nonsense is what one gets. Not only (4) do not thus express any thought, but (12) not (13), unlike (15) and (16), are in the same bag:
(4) I know that I am in pain (alternatively, Only I can know whether I am in pain)
(12) I doubt whether I am in pain
(13) I learn that I am in pain
(14) I am in pain
(15) I know that she is in pain
(16) I doubt whether she is in pain
These judgements found on reasons, provided in the last part of PhI, § 246(a) as well as in PhI, § 246(b,c), that can be resumed in the point that where no room is left for distinguishing between to know that p and to have reasons either to have reasons to think that p, to doubt whether p or to learn that p, to resort to `to know' gives rise to nonsense. Moreover, it is not just the contrast between `to know', on the one hand, and `to have reasons to think', `to doubt whether' and `to learn' that matters. To the very same contrast class belong `to believe (in some of its uses)', `to suspect' and `to find out', among others.
Of course, this is not all. Wittgenstein openly admits that (4) has, or might have, another use, namely, the same (14) could be put to. Nonsense that results from uttering (4) in intending to claim that my sensations belong to the realm of my own acquaintance originates in ignoring that there is no movement in the language game that acknowledges the existence of a cognitive relationship between our pain experiences and ourselves. The basic error, which Wittgenstein focuses on in PhI, §§ 247 -- 255, is due to mistaking the use of sentence (4): we believe we are in the presence of a cognitive thought, one that plays a central role in articulating the metaphysics of subjectivity, while the truth is that (4) does nothing but expressing a grammatical rule. The metaphysical emphasis laid on `I know' fades away.Foot note 21 Put it in more systematic terms, that (4) is an apparent thought is the conclusion of the following argument:
[P1]For every proposition p: I express/conceive a thought on uttering/saying silently to myself that I know that p, if I express/conceive a thought on uttering/saying silently to myself that I can learn p
[P2]I do not express/conceive any thought on uttering/saying silently to myself that I can learn that I am in pain
[C] I do not express/conceive a thought on uttering/saying silently to myself that I know that I am in pain
Premise [P1] captures the fact that `I know' belongs to a paradigm or contrast class which `I believe', `I have got reasons to think', `I surmise' and `I can find out' also belong to. The paradigm member Wittgenstein explicitly mentions in PhI, § 246(c) is `I can learn'. Of course, the remark that `I know' and `I can learn' belong to the same paradigm far from bending the usual meaning or use of `I know' it is fully faithful to it.Foot note 22 The existence of paradigmatic relations between different expressions or conceptual resources -- relations referred to in [P1] -- is an aspect of thought that Wittgenstein thoroughly exploits in arguing that having a meaning and expressing a thought are not equivalent properties.
On the other hand, premise [P2] completes the argument by pointing out the impossibility of importing `I can learn' to the following sentential context:
(ç 13) ___ that I am in pain
In other words, (ç 13) is not an appropriate context for `I can learn'. It would then follow that `I know' cannot be exported to (ç 13) as well. This is why (4) is nonsense in one of its uses. This amounts to arriving at the end of PLA's Epistemic Privacy Way. However, there remains the need for answering a couple of questions to finish reconstructing the argument. The first question is why Wittgenstein specifically focuses on that particular paradigm item, i.e. `I can learn', compares it with `I know' and concedes it such an emblematic role. The answer: Other items, like `I have found out' and `I surmise', would have suited equally well. What really matters is that, drawn by philosophical prejudices, it may cost us dear to appreciate the particular relation between `I know' and (ç 13). The second question is more difficult to answer: Why cannot `I can learn' be exported to (ç 13). Without trying to exhaustively answer this question -- since many philosophers are familiar with at least a significant part of this realm, three different argumentative lines can be adduced to explain why MRE rules out this possibility:
a) There is no way of mounting `I know' on (ç 13), thus succeeding to make up the vehicle of a thought, because (14) does not provide the traditional philosopher with the needed propositional object such a thought should have. This is due to the fact that (14) does not work as a piece of representational or intentional system, but as a piece of an expressive system, i.e. (14) is used to moan. It is an avowal, like `It hurts'. «[W]ords» -- suggests Wittgenstein as a possibility -- «are connected with the primitive, the natural, expressions of the sensation and used in their place» (PhI, § 244(a)). Now, what focusing on the use of `I know' brings out is that these words cannot have a role to play both in the expressive and in the intentional dimensions. Unlike other expressions that belong to the same paradigm, i.e. `I believe', `I know' is devoid of such a potential. (See OC, § 180.)
b) Leaving now aside what has just been put forward in a), let's seriously consider whether `I know' has an intentional or representational function -- not only an expressive one -- which lends itself to being used to express a thought. A language game in which it would be proper to use `I can learn', one that also allows `I know' to be applied, requires from the language user the possibility of taking on initiatives to improve his or her epistemic backing with respect to what the language user might get to know. To improve one's epistemic guarantees about p will often consist of either stocking up with reasons for p or correcting and completing those reasons for p already gathered. It seems clear, then, that the possibility of exporting `I can learn' to context (ç 13) has to be dismissed, since the result of that move is the improvement of my epistemic backing by finding out reasons that support the conclusion that I am in pain -- an apparent thought to all effects. Yet no such a possibility is in the offing. In what direction should one look for ways of improving one's epistemic stance? The most natural option seems to be the observation of our own behaviour, provided that keeping watch on other people's behaviour is what we resort to in trying to know whether they feel pain. Since this is an absurd choice, Wittgenstein concludes that «I cannot be said to learn of them» (PhI, § 246(c)). This, of course, does not bring the answer to an end, because the possibility remains open for any of us of knowing whether we feel pain without having to examine our own behaviour, namely, by checking inside oneself. However, Wittgenstein also blocks this way. What one should now adduce is that (14), either uttered in the open or silently said to oneself, expresses a thought whose truth none of us would know how to question. Does not pain afford the sort of foundation on which our certainty sustains? Do not we know then that we are in pain? No -- and this is the final answer. The exportability of `I can learn' to context (ç 13) is not underwritten by any successful research work. At the apparent truth of (14) «[w]e do not arrive [...] as a result of investigation» (OC, § 138). The conjuring trick lies in acknowledging that any reason or foundation for its truth would be less compelling than the very thought's truth whose support we could be looking for. `I can learn' was chosen because `I know' is a right movement to make in a language game in which the possibility of articulating, correcting and accumulating reasons must be built into from the very beginning. If the shaping, correcting and gathering of reasons do not provide us with the desired support, then the only conclusion left is that we are not in the knowledge business. Thus, the use of `I know' is not a possible movement in such a language game. `I know' announces that reason supplies are on offer:
c) One more option before closing. The one who utters (14), `I am in pain', either speaking loudly or talking silently for him or herself, really wishes to say that he or she is aware of a feeling of pain, but awareness is now conceived as the phenomenal modality of consciousness. In such a case, (14) would be used to expresses the same thought as `I feel I am in pain'. Nevertheless, no real, non-apparent thought would have thus been expressed. In this occasion I won't take any time over the ensuing argument (see Wittgenstein 1980: § 913).
6 Summing up
In this essay I have put forward and analyzed the view on nonsense Cora Diamond has carried out starting from the work of Frege and Wittgenstein. From Diamond's views I have derived a condition, the Meaning Restricted Exportability Constraint (MRE), and subsequently applied it to a kind of nonsense -- the blind spot thought variety -- which is found in one of the arguing lines that constitute Wittgenstein's so-called Private Language Argument. According to the reconstruction of such a line, the Epistemic Privacy Way, Wittgenstein would have made MRE a strategic function by pointing out that the advocate of epistemic privacy overlooks what constraints govern the paradigmatic exportability relations of expressions like `I know' and `I can learn', among others.Foot note 23
[Foot Note 1]
Here I follow Hacker (1990: 54&ff.)
[Foot Note 2]
Hacker explicitly traces out this connection when he writes that Wittgenstein «insisted that it is wrong, even nonsense, to say `I know that I have a pain' [...]. Such an utterance can be grammatical assertion, or merely an emphatic affirmation, but not, as philosophers typically take it to be, an epistemic claim» (Hacker 1990: 56). However, the most penetrating attempt I know of to explain why `I know I am in pain' lacks sense is carried out in Kripke (1982: 117&ff.) I do like Kripke's analysis because of the connection he traces out between that sentence and `It is 5 o'clock p.m. on the Sun' -- an example I will be largely concerned with below.
[Foot Note 3]
In what follows, and in order to make the text more readable, I will abide by the following convention, namely, that any conclusion about sentences and their constituents I may arrive at will also apply, in what should be a plain parallelism, to thoughts and their constituents, i.e. concepts thoughts are made of. I will also follow the converse norm. I choose to follow these conventions because the kind of nonsense I want to discuss in this paper, and my reconstruction of PLA in particular, does not force anyone to adopt the more restrictive, albeit more realistic as well, starting point of accepting that thoughts are what sentences express relatively to extralinguistic contexts. In general, the sort of examples tI will bring up below does not play so significant a role that should oblige us to take into account extralinguistic roles separately. Put differently, the only kind of context for either a word or a concept that will matter to assess their contributions to sentence meanings and thought contents are further words and concepts. Nevertheless, in some of Wittgenstein's texts quoted below (in § 3), especially the one extracted from OC, § 333, extralinguistic or extraconceptual context makes a substantial contribution. If justice had to be made to this example, the exposition that follows would contain an excess of detail and would be much too involved. I expect, however, that by taking this shortcut I am not leaving out anything worth of consideration.
[Foot Note 4]
Wittgenstein (1953/2001: §§ 350& f.) is also the source of (2) and (3).
[Foot Note 5]
I prefer this analysis to arguing with Hertzberg (2001: § 1) that both (2) and (3) express mental contents that contain gaps to be filled, because by putting things in this way I feed the impression that these sentences are assimilated to (C) -- what would be a mistake.
[Foot Note 6]
Due to Wittgenstein. See his PhI, § 347&ff.
[Foot Note 7]
Baker and Hacker (1980: 267&ff.), Heil (1981), McDonough (1989) and Wright (1989) contain criticisms of this conception. In another place Diamond asserts that «[i]t is just a phrase that we have put together by analogy with other phrases that we use» (Diamond 2000: 69). This suggests that she might welcome what Robbins has called the Minimal Compositionality Constraint: «The content of indefinitely many complex concepts is exhaustively determined by the concepts of their constituents and the rules governing the combination of those constituents» (Robbins 2002: 319). This constraint would allow excluding (2) and (3) as meaningful sentences. I gather that Robbins consciously articulates his constraint by substituting `every' for `indefinitely many'.
[Foot Note 8]
Baker and Kacker (1980: 263) and a few others do not make this distinction, which it is forced on us anyway. Meaning compositionality does not only requires `composition' to be literally understood, but it makes it necessary that there be a functional relation between the meaning of the complex and the meaning of its constituents. Obviously, a composition relation is functional, but not the other way around. Unless a difference between functionality and compositionality is acknowledged, we should follow Hale (1997) in distinguishing Weak Compositionality from Strong Compositionality.
[Foot Note 9]
See Baker and Hacker (1980: 263&ff.).
[Foot Note 10]
Some have argued that this relation is supervenience. See Szabó (2000).
[Foot Note 11]
To quote only one author, Dummett (1981a; 1981b) is the classic reference to put on record the fact that philosophers have not overlooked the existence of some sort of incompatibility between meaning compositionality and contextuality. It must be added that the conciliatory reading of the Context Principle I have chosen above, no matter how much reasonable it may sound, is not unanimously embraced nowadays. See note 12 to catch a glimpse of Diamond's troubles to deal with both principles simultaneously.
[Foot Note 12]
Therefore, MRE is in line with the structured nature of thought, a view that Diamond does not wish to deny, because it contains the key piece to explaining language learning and thought productivity. It also guarantees the existence of systematic relations among thoughts, provided that the Compositionality Principle makes it clear why two thoughts differ that result from combining the same conceptual items in different ways; and why thoughts that do not combine the very same items are distinct. It is no coincidence that Diamond rejects meaning functionality while openly welcoming the idea that the meaning of a sentence depends on the meaning of their constituents. It is no coincidence as well that she forcefully opposes MRE while, immediately after doing so, adding that this does not entail the idea that words have to be assigned to syntactic categories (cf. Diamond 1990: 104). In fact, to my mind, the main problem that Diamond's paper poses for any reader is how to harmonize all these claims. For example, she seems to find in the Fregean doctrine that «we cannot properly ask for the meaning of a word except in the context of a sentence» (Diamond 1990: 109) the counterweight she is looking for to decide what is prior over what, whether the sentence over the word or the other way around, the word over the sentence. All this leads the reader to conclude that Diamond values the Context Principle higher than the Compositionality Principle. However, it is difficult to know whether this choice is full of content or not. In effect, compositionality is a blunt scalpel if Frege's Context Principle is read, as Diamond does following Quine, as requiring that «it is through the sense of the whole that the parts get their meaning» (Diamond 1990: 109). Let's take seriously the advice that sentences have priority over the words they are made up of and, accordingly, to go on to segment sentences assigning words to syntactic categories and meanings to the bits so obtained. Once this task has been accomplished, it follows that the meaning of the sentence is a function of the meaning of its parts and of its syntactic mode of combination. True enough; but the compositionality requirement is trivially satisfied if the sense of the whole comes first. It cannot be put to service to explain how language is learnt and thought is understood, unless we allow the domain to which the Context Principle is applied to be a proper subset of the whole set of sentences that define language. Sentence meanings and thought contents must be out there, many of them anyway, before we start cutting them into pieces. Therefore, Diamond denies compositionality, because she takes it to be incompatible with sentence priority, while on the other hand she needs it as a condition to carry on those explanatory tasks. That her position is unstable is also shown by her brief comments on language understanding. A sentence is grasped by applying rules that analyze it into words and map meanings onto them, but to follow these rules does not consist of using the competence of recognizing its constituents, their meanings, and putting them together. To understand a sentence is to make sense of it; in other words, «to make the sentence his, but using the rules» (Diamond 1990: 111). This remark is hardly helpful, to say the least, because Diamond has not solved the tension described above.
[Foot Note 13]
Things are not so straightforward, as a more thorough consideration of (9) leads us to recognize. Wittgenstein points out that (9) makes no sense for Europeans, i.e. Western people: (9) does not express a thought if uttered in a conversation among them. However, as soon as we widen the language context by including in it the place where people are born and grown, then syntagmatic relations become quite distinct, and the possibility of exporting a word or an expression to a context now makes a difference for the speaker's CV. (See note 3 above again.)
[Foot Note 14]
Wittgenstein immediately adds that what human beings think is reasonable changes not only among persons but from one culture to another, and within one and the same culture, and from one time to another. See OC, § 336. (ç 11) is dealt with in OC, § 282.
[Foot Note 15]
In good faith it should not be alleged that exportability relationships are squarely understood by taking them to be computational relations. In a sense there is no denying that they are relations between sentential contexts, and that these contexts are complex symbols of a computational system. However, this is only part of the truth, because exportability relationships supervene on what use sentences involved have, i.e. in what circumstances they can be used.
[Foot Note 16]
This is the apt way Diamond describes the situation. See her Diamond (2000: 69).
[Foot Note 17]
Deflationist theories of truth, among which I include Disquotationalist theories of truth, deal with the sort of linguistic abilities Wittgenstgein would comprise under the label of `surface grammar'. In a typical statement of Deflationism, as that of Field (1994), semantic theory aims at giving the truth-conditions of any sentence belonging to the language the semantic theory is a theory of. What confers Deflationism such an extremely superficial character, in the Wittgensteinian sense, to meaning analysis is that, according to such a semantic approach, meaning theory aims at theorems of the form «`p' is true if, and only if, π», where «`p' is true» and «π» are cognitively synonyms. Since it can be argued that, let's say, (2) does not have substantial truth-conditions -- because it expresses no thought, Deflationism is forced to distinguish surface truth-conditions from deep (or substantial) truth-conditions. Field's Deflationism could only admit of surface truth-conditions.
[Foot Note 18]
Nowadays philosophy of language hardly leaves room for distinguishing the theory of meaning from the theory of thought -- a situation to which it is not alien the fact that the question of nonsense has disappeared from both philosophy and grammatical theory. In my opinion, such a distinction echoes another one, this time between nonsense that has a sense from nonsense without any sense, to which Carnap gave his approval, but which Wittgenstein rejected (cf. Whitherspoon 2000). However, to those who favour the Wittgensteinian attitude we can reproach their not being sensitive enough towards the fact that sentences (2) -- (6) as well as many others that matter to philosophy (like `colours are mind sensations' or `this piece of marble has extension') are meaningful, in some sense of the word `meaning'. A possible exception to this situation is Glock, who is aware that the demands of formal logic openly differ from what philosophical grammar requires (see Glock 1996: 222). Whitherspoon (2000: § 5) contains the most elaborated defence I know of the significance that sentences like those discussed in this paper have, a defence in which the notions of quasi-interpretation, quasi-understanding and quasi-meaning, among others, play a central role.
[Foot Note 19]
Concerning the criss-crossing of arguments in PLA, I follow Hacker (1990: 22&ff.). On what PLA's central argumentative line is, I follow Stern (1994).
[Foot Note 20]
As far as this relation is concerned, I only know two precedents. The most direct one is Groenendijk and Stockhoff (2004), where the Epistemic Privacy Way is approached to in outline from the perspective of the Compositionality Principle. On the other hand, in the interesting discussion on Moore's Paradox in Schulte (1993: ch. 9) I appreciate the presence of the Meaning Free Exportability Requirement (see § 2 above) under the guise of a Uniformity Principle supposedly governing the use of the verb `to believe'. «The advocate of uniformity», Schulte writes, «makes the mistake of expecting all expressions to behave in the same uniform manner and, if he encounters unexpected forms of behaviour, to avoid facing up to undesirable characteristic of our use of linguistic expressions by side-stepping and appealing to misleading similarities. He is tempted by surface analogies to assume that there are corresponding analogies of function and meaning. And he tends to overlook that surface analogies may hide completely different ways of functioning» (Schulte 1993: 150). As for the case I am concerned with in this paper, the privacy of sensation is but an outcome of yielding to one of those misleading similarities. My aim here is to bring out the details of the confusion.
[Foot Note 21]
Wittgenstein writes in On Certainty: «It is as if `I know' did not tolerate a metaphysical emphasis» (OC, § 482).
[Foot Note 22]
«I would like to reserve the expression `I know' for the cases in which it is used in normal linguistic exchange» (OC, § 260).
[Foot Note 23]
The results here reported were obtained while developint two research projects of the Spanish Ministerio de Educación y Ciencia: BFF2000-1073-C04-01 and HUM2004-118. Without the help of Fran Camós much of the stress I put on some pasagges in which exportability restrictions and conditions of use sentences are taken up would be wrongly placed. My debt to Jaroslav Peregrin and José Manuel Morillas is acknowledged as well. However, sound judgement counsels to warn that neither of them are responsible for any detailed point made in these pages.