Abstract: Donald Davidson says there is no such thing as a language. His reasons for this claim stem from the improvisatory nature of linguistic communication, as manifested in our understanding of malapropisms. If there are no languages, there can be no translations either. However, I argue that justice can be done to the improvisatory aspect of linguistic communication in a way that is consistent with the existence of languages and translations. Davidson, like many philosophers and linguists, assumes that languages, if they exist, must be sets of some kind. I bring to bear an account of artifacts, including abstract artifacts, that I have developed elsewhere in order to argue that languages are abstract artifacts made out of, but not identical to, sets of certain kinds. This enables me to describe the differences in the kind of improvisations needed to understand things like malapropisms and those needed to translate from another language.
What is a translation? Here is a definition which, although it might require some refinement, seems to capture the core of the concept:
TR) x is a translation if and only if i) x is an utterance in a language L, and there is an utterance y(≠x) in a language L’(≠L), such that ii) x and y have (more or less) the same meaning, and iii) x is produced with the intention that ii) is the case.
If something along these lines is correct, then the existence of translations, as condition i) makes clear, depends on the existence of languages. A theorist of translation, therefore, is likely to be bemused by Donald Davidson’s startling pronouncement that:
there is no such thing as a language, not if a language is anything like what many philosophers and linguists have supposed. There is therefore no such thing to be learned, mastered, or born with. We must give up the idea of a clearly defined shared structure which language-users acquire and then apply to cases.1
Given TR, Davidson’s position implies that not only is there no such thing as a language, but there is no such thing as a translation. Davidson’s conclusion rests explicitly on concerns stemming from the nature of linguistic communication and implicitly on certain ontological presuppositions about what a language would be, if it existed. In this paper, I shall attempt to bring to light those ontological presuppositions. I shall argue that by taking a different approach to the ontology of languages, we can accommodate Davidson’s linguistic concerns without giving up on the common sense idea that there are both languages and translations.
First, then, let us see what Davidson’s reasons are for his claim that there is no such thing as a language. The suppositions of “many philosophers and linguists” to which he adverts are these:
1) a language is systematic – the meanings of complex expressions depend on the meanings of their parts and the ways those parts are put together;
2) a language is shared – the typical case of linguistic communication is one in which all parties share a language;
3) a language is conventional – the assignments of meanings to parts and the semantic effects of composition are determined by conventions adopted by speakers of the language.2
Now, suppose Mrs. Malaprop says to Captain Absolute:
if I reprehend anything in this world it is the use of my oracular tongue and a nice derangement of epitaphs.3
Absolute may initially assume that he should understand Mrs. Malaprop in the light of conventions that associate the word “reprehend” with the meaning reprehend, “oracular” with oracular, “derangement” with derangement, and “epitaph” with epitaph.4 These conventions are (part of) what Davidson calls Absolute’s “prior theory” (since it is a theory he holds prior to this encounter with Mrs. Malaprop). In this case, it is clear that communication is not facilitated by Absolute’s relying on his prior theory. That theory would generate an interpretation of Mrs. Malaprop’s utterance as meaning that if she reprehends anything in this world it is the use of her oracular tongue and a nice derangement of epitaphs. But no-one could mean that! And even if someone could, the context of Mrs. Malaprop’s utterance (she is defending herself against an accusation that she cannot speak English properly) shows that she certainly doesn’t mean that. So Absolute must modify his theory on the go, yielding what Davidson calls a “passing theory.” On this passing theory, “reprehend” is associated with apprehend, “oracular” with vernacular, “derangement” with arrangement, and “epitaph” with epithet. On the basis of the passing theory, Absolute interprets Mrs. Malaprop as saying that, if she apprehends anything in this world it is the use of her vernacular tongue and a nice arrangement of epithets. In other words, she certainly knows how to speak her own language and can string adjectives together very nicely, thank you!
Both the prior and the passing theories are systematic: they pair words and meanings and contain semantically significant syntactic rules that determine how meaning is compounded. Thus they both satisfy the first of the three conditions on languages given above. But neither, Davidson argues, can correctly be taken as being or representing a language. His prior theory fails to satisfy the second condition because it is not shared by Mrs. Malaprop. (It associates “reprehend” with reprehend whereas Mrs. Malaprop associates “reprehend” with apprehend.) And his passing theory fails to satisfy the third condition because its word-meaning pairings are not determined by convention. They are posited ad hoc by Absolute.
How does Captain Absolute reach his passing theory, if not by following existing conventions? Davidson says:
there are no rules for arriving at passing theories, no rules in any strict sense, as opposed to rough maxims and methodological generalities. A passing theory really is like a theory at least in this, that it is derived by wit, luck, and wisdom from a private vocabulary and grammar, knowledge of the ways people get their point across, and rules of thumb for figuring out what deviations from the dictionary are most likely.5
This is a wonderful passage. It portrays linguistic communication as a free-spirited exercise in bricolage rather than as the dour province of schoolteachers and language primers. Absolute must rely on “wit, luck, and wisdom,” and all the rest of it, to make the leap from the assignment, in his prior theory, of the meaning oracular to the word “oracular,” to the assignment, in his passing theory, of the meaning vernacular to that same word. One of Davidson’s main contentions in “A Nice Derangement of Epitaphs” is that both Absolute’s predicament and his solution to it are ubiquitous and constitute a paradigm of linguistic communication. We are all bricoleurs when it comes to language, constantly adjusting our prior theories in the light of wit, luck, and wisdom, employing whatever comes to hand to make sense of people. For Davidson, then, linguistic communication does not give a place to anything that corresponds to what we take a language (as characterized by the three conditions above) to be.
If there are no languages, what should we say about TR? Are there no translations either? Or can we amend our account of what a translation is so as not to rely on the concept of a language?6 Suppose we simply eliminate the references to languages:
TR*) x is a translation if and only if i) x is an utterance and there is an utterance y(≠x) such that ii) x and y express (more or less) the same thing, and iii) x is produced with the intention that ii) is the case.
The problem is that this is much too permissive. It would count as translations many things that we do not think should count as such. For example, three people are talking, all in English. A says something but C simply doesn’t hear it. B repeats what A said, with or without some minor variations. B’s utterance will count as a translation of A’s, according to TR*. Intuitively, the salient feature of instances of TR* that are genuine translations is that they involve not just the reproduction of meaning but its reproduction across some boundary or barrier, a boundary or barrier that is captured, in the original TR, by the requirement that the two utterances be in different languages.
We have, then, a conflict. On the one hand, there is the common-sense idea that there are such things as translations and that an account of what a translation is, like TR*, that does not invoke languages will erase the difference between genuine translations and a host of other cases involving reproduction of meaning. On the other hand, Davidson has given a characterization of linguistic communication that stresses its improvisatory nature but that leads him to reject the existence of languages, and hence translations. Can we do justice to Davidson’s very appealing vision while hanging on to the existence of languages and translations?
Davidson says, as we have seen, that “there is no such thing as a language, not if a language is anything like what many philosophers and linguists have supposed.” We looked at three conditions that characterize languages, but none of these characterizations speaks to what a language actually is. What do philosophers and linguists suppose a language is? I cannot here survey all the views that have been expressed on the topic.7 But one answer that many philosophers and linguists have given is that a language is a set of some kind. There is difference of opinion over what the members of such a set are, but there are two tendencies that are common. One takes a language to be something like a (possibly infinite) set of sentences, or sentences plus interpretations; the other takes a language to be a (probably finite) set of elements – rules, expressions, words – out of which sentences can be constructed. Chomsky provides this definition:
I will consider a languageto be a set (finite or infinite) of sentences, each finite in length and constructed out of a finite set of elements.8
Both tendencies are evinced here – with some resulting unclarity that I shall not dwell on. In what follows, I shall take the set-theoretic approach in its second tendency, as identifying a language with a set of word-meaning pairings and semantically significant syntactic rules. Although Davidson does not explicitly bring in talk of sets, it should be clear that when he rejects the existence of languages as they are commonly thought of by philosophers and linguists, he is assuming that if there were languages, they would be what this second tendency of the set-theoretic approach takes them to be.
There are two main problems with thinking of a language as a set. The first has to do with temporal and modal flexibility. Languages change over time. They gain and lose words, the meanings of words change, and the syntactic rules change. And not only do they change in various ways, but they might change in others, even though they do not. Actual and merely possible changes make languages temporally and modally flexible. Sets, by contrast, cannot change their members over time, nor is it possible for them to have different members. There are, of course, set-theoretic expedients at hand to try and deal with this mismatch between languages and sets. We could identify a language not with the set of elements that characterize the language at a time, but rather with the set of its elements throughout time (so that English would include all its past, present, and future word-meaning pairings, for example), or, to account for a language’s modal flexibility as well, all its elements throughout time and its merely possible elements. Or we could take a language to be a function from times to sets, so that a language would associate with each time the set of elements characterizing it at that time. Or again, to accommodate the modal flexibility, a function from pairs of possible worlds and times to sets. None of these expedients, in my view, is adequate. The problem is that they all take something that is, in its essence, dynamic and identify it with something static, albeit spread through time and modal space to provide a simulacrum of a language’s essential dynamism.9
The second problem with identifying languages with sets, of any sort, is that languages are artifacts but sets are not the kinds of things that can be made. The artifactual nature of languages is particularly evident in the case of invented languages: Esperanto, Volapük, Klingon, Láadan and literally hundreds, if not thousands, more.10 Since it is the artifactual nature of languages, in my view, that is key to understanding what a language is, let me say something about artifacts in general before returning to languages and translation.11 Start with concrete artifacts. When someone makes a chair, she works on some matter – wood, for example – with various intentions. So there is a labor component of the making and an intentional component. The nature of the labor is adventitious; it is whatever is needed to bring the matter into line with the intentions. The intentions will be both instrumental, such as the intention to nail this piece of wood to that piece, but also, crucially, must include what I call a creative intention, an intention to make an object of a certain kind. It is the intention to create that results, where the labor is successful, in the coming into being of a new object, an artifact of the kind specified in the creator’s intentions. The process of making might be thought of as the imposition of a concept onto the matter; the result is an entity that is essentially a product of intentional making, an intentional object, so to speak. The object has matter but is not identical to that matter. Chairs, and other common artifacts, can change their matter over time. It is part of the very concept chair, the concept that is imposed on the matter in the act of the chair’s being made, that chairs are the sorts of things that can undergo changes, not just with respect to location, color, and so on, but also with respect to the matter they have at any given time. They can lose matter as they are subject to ordinary wear and tear, or more drastically, if a whole piece of wood falls off; and they can gain matter, as broken parts are replaced or refurbished. From an epistemic point of view, it is possible to encounter the matter of the chair without realizing that there is a distinct object of which it is the matter. This might happen if one lacked the concept chair or, for some reason, was unable to apply the concept in a particular context. If, for example, a chair was sufficiently unusual in its shape, and was encountered in some unhelpful context (perhaps in the middle of a huge pile of scrap wood), one might encounter the wood that is the matter of the chair without apprehending the chair itself. One would encounter the wood, in such a case, without any way of understanding how that wood could change or be replaced consistently with the continued existence of a distinct thing of which the wood was the matter.
Essentially the same account applies in the case of abstract artifacts, such as musical works, poems, or fictional characters but there are two special points in such cases.12 First, the artifacts are abstract because their matter is abstract. Secondly, as a consequence of the abstractness of the matter, the nature of the labor on it by which an abstract artifact is created is generally somewhat different from that involved in the creation of material artifacts such as chairs. By way of illustration, consider the case of musical works. Platonists about musical works identify musical works with sound structures. I assume that a structure is itself some kind of set-theoretic construct. So the identification of musical works with sound structures is analogous to the identification of languages with sets. As in the language case, the identification of musical works with sound structures has several unwelcome consequences. First, it implies that musical works are not brought into existence by an act of the composer, and hence that they are not artifacts at all. Secondly, it implies that musical works cannot change over time and could not have been any different from how they are. But although we should not identify musical works with sound structures, that does not mean that sound structures have no ontological relation to musical works. Rather, we should see a sound structure as the (abstract) matter out of which a musical work is made. How does one ‘work on’ something abstract like a sound structure to make a musical work out of it? One way to think of the work would be as indicating the sound structure, perhaps by writing a score that represents it, as opposed to other sound structures.13 If a musical work, then, were treated along the same lines as a chair, allowing for the differences raised by the abstractness of the matter, we would be able to hold that it, too, might change its matter over time. For example, a composer might revise a musical work, perhaps changing a sequence of crotchets into quavers plus rests, so that its matter at one time is a sound structure distinct from the sound structure that was its matter at an earlier time.
Now, returning to languages, we can ask what happens when someone invents a language, as Suzette Elgin invented Láadan?14 She makes an abstract artifact. The matter of the language is just what the set-theoretic approach has taken a language itself to be: a set of elements such as word-meaning pairings, syntactic rules, etc. The labor the language inventor performs with respect to this matter is selecting, or indicating, it, which she might do by writing down representations of the members of the set (i.e. by writing a lexicon and grammar). The set of elements is not created by the language maker.15 The set is not an entity that can be created, and it cannot have, or come to have, different members; the abstract artifact which is the language, by contrast, is created and can have, or come to have, different matter. In just the same way, a given sound structure is not created and cannot have, or come to have, different sounds as elements of it; but a given musical work is created and can have, or come to have, different sound structures as its matter. What is responsible for the existence of an artifact that has matter is the imposition, by the maker, of the concept of that kind of artifact onto the matter, through the process of making. In the case of a language, that means that the maker of a language brings the language into existence by selecting or indicating the word-meaning pairings and syntactic rules with the intention of making a language.
What goes for invented languages can be extended to ordinary natural languages. There is, of course, no single point at which the concept language is imposed on some set of pairings and rules. But over time, the speakers of a language maintain the imposition of the concept on sets of rules and pairings, one replacing another as the matter of the language as people come to use new forms and drop older ones. A natural language is a collectively built and maintained artifact, something like an abstract city.
Let us now return to Davidson’s claim that there are no such things as languages and the challenge it poses for the existence of translations. Davidson’s skepticism arises because he sees a conflict between the conventional nature of languages, as traditionally understood, and the improvisatory nature of actual linguistic communication. Our prior theories are likely to be inadequate on a regular basis and so we will be thrown back on the resources of our “wit, luck, and wisdom” as much when we speak with people who speak ‘the same language’ as when we deal with those who speak ‘a different language.’ When an upcountry Norwegian farmer speaks with a hipster from Oslo, she may find herself far more reliant on “wit, luck, and wisdom” to understand her interlocutor than when she speaks with an upcountry Swedish farmer from just across the border, despite the fact that, as we would put it, in the first case the two are speaking the same language and in the second they are not. We will find a role for languages, then, if, consistently with Davidson’s insights about the roles of “wit, luck, and wisdom,” we can distinguish between different ways in which departures from our prior theory are generated, a difference that tracks what we would pre-theoretically think is a departure necessitated by difference of language and what we would pre-theoretically think is a departure from existing convention that is yet within the same language. And this, I think, a conception of languages as sets cannot do, but a conception of them as abstract artifacts, along the lines suggested above, can.
Davidson’s problem arises because he assumes, in line with the set theoretic understanding of what a language is, that if there were such things as languages, they would have to be identical to sets of conventional word-meaning pairings, etc.. Hence, where someone cannot rely on just those conventions, they can no longer count as drawing on the resources represented by knowledge of the language associated with those conventions. I suggest that the set of conventions represented in the prior theory does not fully fix the identity of a language itself (just as a given sound structure does not fully fix the identity of the musical work with which it is associated, nor some wood that of the chair). Recall that I explained above that one does not properly encounter the chair merely by encountering the wood which is its matter at a time; one must also apprehend the concept under which the matter has been worked on, in order to know such things as where the boundaries are of the chair (how much of the wood one is seeing, as a given chair lies atop a pile of scrap wood, belongs to the chair), what changes can occur without destroying the chair, in what ways could the wood have now been different and yet still be the matter of the same chair, and so on. Davidson, in effect, is looking at the prior theory and thinking that if there were a language associated with what that theory represents, the language and what the theory represent would have to be identical. But what he is ‘seeing’ is merely the matter of the language. In order to apprehend the language, we must see that matter as organized by, or falling under, the concept of a language. This concept is what will allow us to distinguish between leaps in understanding that fall within a single language, and those that do not.
I shall illustrate this, briefly, with two examples. But before I give them, let me say how much I am in agreement with what Davidson says in the whole of the “wit, luck, and wisdom” passage quoted above. My disagreement with him comes only in my thinking that what he says there actually gives us the resources to go much further in understanding the different ways of departing from a prior theory to construct an interpretation of someone than he himself does.
As a first example, consider Mrs. Malaprop again. When someone constructs a passing theory to interpret her utterance, she must indeed employ wit, luck, and wisdom, and rules of thumb about what deviations from the dictionary are most likely, etc. But the prior theory does have a distinctive role to play, if we remember that what is given there is the matter of something that involves the imposition of the concept of a language. Roughly, since a language involves elements with phonic and graphic properties (spoken and written words), the concept of a language makes salient to the interpreter who apprehends the prior theory under the concept of a language, and not simply as a set of elements, the possibilities of deviations from the dictionary that are based on certain kinds of phonic and graphic similarities. Indeed, it is not simply because the pairing of, say, the word “reprehend” with the meaning reprehend is part of the matter of a language that Absolute is able to get to the passing pairing of “reprehend” with apprehend, but because the first pairing is part of the matter of a particular language, English, with certain regularities of word formation, certain principles of etymology, and so on. The word “apprehend” has not just a phonic, but a structural relation to the word “reprehend.”16 Knowing what a language is, and hence apprehending the various word-meaning pairings that are parts of the prior theory as parts of something with such regularities and history, as elements of a language, is key to the ability to understand Mrs. Malaprop. True, as Davidson says, there are no rules at work; but the concept of a language, and the apprehension of a language that that enables, play more than a merely adventitious role in allowing transition from the prior theory to the passing theory. Thus, if Mrs. Malaprop were to make her utterance, and someone were to ask Absolute what she had said, and Absolute were to say “if I apprehend anything in this world it is the use of my vernacular tongue and a nice arrangement of epithets,” he would not count as offering a translation because his report is based on the resources of English, even though Mrs. Malaprop had used non-English pairings of words and meanings, and he English pairings.
My second illustration concerns change over time. Here is a passage from the King James version of the Bible:
Why died I not from the womb? why did I not give up the ghost when I came out of the belly? Why did the knees prevent me? or why the breasts that I should suck? (Job 3, 11-2)17
Given my prior theory, the author of this text seems to be saying something absurd. If Mrs. Malaprop had uttered this nice derangement of epitaphs (and it really does read like an epitaph!), we would laugh at her misuse of the word “prevent.” For after all, Job’s mother’s knees and breasts did not prevent his birth but, to the contrary, facilitated it, the former by opening at the appropriate time, the latter by filling with milk “that he should suck.” So we must construct a more adequate passing theory. With a little knowledge of Latin (or other Romance languages) and some a priori reasoning, we can infer that there must be an archaic usage of “prevent” to mean anticipate (pre-venire: to come before). In other words, Job is lamenting that his mother’s knees opened, and her breasts filled with milk, in anticipation of, or preparation for, his being born. In apprehending English, the matter of which no longer includes a pairing of “prevent” and anticipate, we apprehend not just its current matter – which includes a pairing of “prevent” and prevent – but an object with a certain history, a history which can be at least partially reconstructed. Part of that history is that the matter of English formerly included a pairing of “prevent” and anticipate. Again, Davidson is right that there are no rules that will lead us to this knowledge. It is, as he says, “wit, luck, and wisdom” that get us there. But the wit and wisdom involved in a case like this manifest themselves in a special kind of knowledge, knowledge of how languages can develop historically. Knowing something of this is integral to apprehending English, a language, though it is not integral to apprehending the matter of English (a set of elements) at a given time. Just so, in apprehending a chair, we encounter something that we know must have had a maker, though this knowledge is not integral to apprehending what is merely the matter of the chair at a given time.
In the cases of Mrs. Malaprop and the excerpt from the book of Job, we are dealing with cases in which our attempts to construct a passing theory exploit not just the elements of our prior theories, but the knowledge that those elements are the matter of a particular language. Hence reaching a passing theory does not involve crossing the barrier of languages and does not count as translation. If we were attempting to construct a passing theory for someone speaking an entirely unfamiliar language, such knowledge would not be of use to us. In that case, we could indeed still draw on “wit, luck, and wisdom” to understand another, and the results, if successful, would be a translation; but the shape of our attempts to construct a passing theory in the context of translation would be quite different from the shape of those attempts that occur ‘within’ a language.18
1 Donald Davidson, “A Nice Derangement of Epitaphs,” in Ernest Lepore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson (Oxford: Blackwell, 1986), 446.
2 Davidson, “A Nice Derangement,” 436, offers these three conditions as governing what he calls first meaning (his version of literal meaning). I think it does no violence to his view to take them as governing languages.
3 Richard Sheridan, The Rivals, Act 3, Scene 3. Text quoted from Sheridan, The School for Scandal and Other Plays, edited by Michael Cordner (Oxford: Oxford University Press, 1998), 46.
4 We need not worry here about what exactly meanings are. And the idea that a language, or knowledge of a language, includes a lexicon with correspondences between words and entities of any kind is not beyond controversy. I don’t think what I say here depends on a substantive view of these matters, so I conduct the discussion in these very simple terms for convenience.
5 Davidson, “A Nice Derangement,” 446.
6 I say (incorrectly, though colloquially) “the concept of a language,” rather than “the concept language” to emphasize that I am concerned with languages and not language. (What the relation is between language and languages – or science and sciences, religion and religions… – seems to me a difficult and intriguing question but one I cannot explore here.)
7 In particular, I cannot do justice to the rich discussion in Robert Stainton’s “Philosophy of Linguistics,” Oxford Handbooks Online (2014): 1-17. Stainton discerns four answers to the issue he frames as the “the metaphysics of natural languages and their parts:” the physicalist, the mentalist, the platonist, and the social. Stainton, however, is more focused on what the “parts” of languages are than the nature of languages as such. For example, the physicalist view is that a language is a set of physical objects (inscriptions or utterances of words, etc.). These physical entities are the “parts” of languages, but Stainton does not confront the issues that arise from the identification of a language with a set of such (or any other) entities.
9 Obviously, a lot more needs to be said to motivate my skepticism about these approaches. For more on the problems of understanding dynamic entities in terms of rigid simulacra, see my Making Objects and Events: A Hylomorphic Ontology (in progress), especially 1.2.3, but also 2.1 and 2.3.
10 See Arika Okrent’s fascinating book, In the Land of Invented Languages: Adventures in Linguistic Creativity, Madness, and Genius (New York: Spiegel and Grau, 2010).
11 The following two paragraphs present a dogmatic and highly truncated version of my views on artifacts. These views are presented at much greater length, and I hope less dogmatically, in Making Objects and Events, especially chapters 3 and 4. See also my “Constitution and Qua Objects in the Ontology of Music,” British Journal of Aesthetics 49 (2009): 203-17 and “Ready-Mades: Ontology and Aesthetics,” British Journal of Aesthetics 53 (2013): 407-23.
12 For a general treatment of abstract artifacts, see Amie Thomasson, Fiction and Metaphysics (Cambridge: Cambridge University Press, 1999). See also the two papers of mine cited in the previous footnote.
13 This is the account given by Jerrold Levinson, in “What a Musical Work Is,” Journal of Philosophy 77 (1980): 5-28, of musical works as indicated sound structures. See my “Constitution and Qua Objects in the Ontology of Music” for a discussion of Levinson in these terms.
14 See Elgin, Native Tongue (New York: DAW Books, 1984).
15 I have not said anything about the ontology of these elements themselves. What is a word-meaning pairing or a syntactic rule? Perhaps they themselves are artifacts. And if one makes some entities of that kind, perhaps we may want to say that one counts, by courtesy, as making the sets of which those entities are members. I ignore such complications in the main text.
16 Of course, as Davidson says (“A Nice Derangement,” 434), not all malapropisms work by exploiting this particular feature of words. Sometimes a speaker might not use other real words at all. I am merely illustrating one of the ways in which the concept of a language enables us to understand someone who departs from the conventions that supply the matter of a language at a given time.
17 Thanks to Robert-Louis Abrahamson for supplying this splendid example to suit my needs.
18 I would like to thank my fellow participants in the conference on The Philosophy of Translation and the Translation of Philosophy at the University of Zürich, February 2014, for their helpful questions and comments on an earlier version of this paper. In particular, thanks to Joachim Adler and Štefan Riegelnik for organizing the conference and inviting me to it.