Meaning as convention: a puzzling effect of Aristotelian theory of names

1.    Why Aristotle's theory of names seems puzzling?

In “On interpretation” Aristotle writes that name (onoma) is a sound which has meaning by convention (syntheke) (Aristotle 1963, 16a20). He says both that no sound is a name by nature, but becomes one when is used as symbol (Aristotle 1963, 16a26)1, and that spoken words (“what are in sounds”) are symbols of im- pressions of the souls (psyche pathematon).2 According to this view meanings of names are impressions, i.e. subjective entities3, which are to be assigned conventionally.

This paradigm of linguistic meaning is still very popular: conventionalism is one of the most elaborated explanations of how linguistic entities acquire mean-ings,4 and meanings are usually considered as at least partly subjective. Quite plausible is the hypothesis that many names of natural languages mean real things, events, properties etc. But when ordinary people come to know which real entities they mean by using certain names they get acquainted with them either by way of direct or indirect (mediated by some image) observation of some of their features or through understanding of some descriptions of what they are. In both cases “in the head” of an agent a subjective representation of a referent of a name is created; and many theories of names assume that in cases when meanings of linguistic signs, names in particular, are their references they have these meanings due to certain mental representations (senses, intensions etc.).5

But if meanings are subjective and their assignments are purely internal acts taking place in individual minds, then there is no overt behavior by producing which a name's user could demonstrate her conformity to a convention about this name's meaning. The conformity should consist in the combination of an external behavior with some internal mental activity. Then patters of a proper mental activity must be somehow established by a linguistic convention governing the use of a name with certain meaning. But how something purely subjective may be collectively used, and how can it be publicly demonstrated as a pattern? This is what seems puzzling relative to any theory of names within the Aristotelian paradigm. Can any notion of convention make the claim that meanings of names are both conventional and subjective consistent? This problem I will address in this article.

It may be immediately objected, though, that the problem is superficial: whatever people do, at least intentionally, they do by using some mental processes; and in this broad sense any convention between people should be a convention about something mental. If conventions are social entities, as they are usually supposed to be, they need some interaction between people in order to be established. An idea of interaction presupposes that its participants act so as to produce some expected effects in other participants, and may be responded to by their actions. And actions of people are normally considered as complex entities involving an external behavior along with some mental (intentional) component: thus we would reject to recognize a behavior looking like greeting as an act of greeting if we had serious doubts that the behavior was motivated by a greetingintention. Then if there is a problem it concerns not only conventions about linguistic meanings.

But there is a difference in how this problem relates to different types of conventions. If the convention is about producing some public effect, such as, e.g., shaking hands as a sign of greeting, it presupposes that some observable results may be signs of conventionality of the behavior which made them occur. Of course we cannot be absolutely certain that what looked like shaking hands was actually shaking hands unless we could observe internal, intentional or motivational sides of the behavior of the participants of an interaction. But we can at least observe in such situation something which would allow to say saying with a good plausibility6 that shaking hands took place, and therefore the behavior was conventional. Therefore by creating or making salient some patterns of public behavior normal social conventions may show people what to do in order to conform to them. Meanwhile if the convention is about meaning as such the results which could be signs of conformity to this convention should be completely internal - some mental activity or state. Unobservable by others, these results cannot be publicly demonstrated and persons cannot learn from each other how to achieve them.

2.    Meaning by agreement

Besides, even if people established a convention without making any agreement we may require that they at least could reestablish it by a relevant agreement. Suppose that two people talk about dogs, and suddenly one of them says seriously that there is a high rate of suicides among dogs. This would sound confusing as a report of what the agent believed if she was known as sufficiently ra- tional and competent. Then the confused participant of the talk may ask: what did her co-talker mean by “dogs”? They both used the word as if they mean the same thing by it, i.e. conventionally; but now this needs to be justified. In order to do this they have to reestablish the convention by way of an explicit agreement about the meaning of the word “dogs”. Having come to such agreement they may then explicate the differences in their beliefs about dogs. If the participants of some collective behavior are incapable to preserve it due to an explicit agreement between them this may be a good reason to doubt that they freely chose their parts of the behavior in question among some relevant alternatives.

Unlike an agreement that something is true or has place, an agreement to do certain thing presupposes that its participants become collectively disposed to act in the agreed way. And in order to become thus disposed the participants need to know or have a proper idea of what the agreement is about: the object of an agreement must be introduced to them. But, however established, in order to persist and develop in time social conventions need to involve the same procedures of introduction of the object of convention to new members who have to learn how to act conventionally. When a person is learning how to assign meaning to the expression conventionally a pattern of conventional meaning should be introduced to her by someone else.

The question then is: can something subjective be literally introduced to a person by someone else? And as a result, can a group of people literally agree to assign some mental impression to a certain form of sound as its meaning?

If an object to be introduced is mental impression or subjective assignment of meaning it cannot be introduced to other people by direct demonstration. It may be introduced by description though. But since the object of a description would still be subjective this introduction could not make other people literally learn, and therefore agree, to assign someone's subjective impression to a name or reproduce someone's private mental activity in their minds. The reason is that people cannot literally learn or agree to do something which as they know they cannot do.

However by "impression" not only some purely mental experience, but also something which is common to different mental experiences may be understood. Mental impression is supposed to represent thing, which caused it (directly or through a chain of interactions), due to being in relation of resemblance with it.9 Therefore, different impressions, even belonging to different persons, may be both caused by the same thing and resembling it. Then they may have common representative content which may be also called an impression of a thing. Thus Aristotle claims that impressions which different souls get by experiencing the same thing are similar (Aristotle 1963, 16a5). Indeed, however different individual experiences of a thing in different situations may be, impressions of the same thing formed by different people may result from extended series of experiences in which the most essential properties of the thing were demonstrated to them, and combined with information provided by descriptions according to some common rules of rationality. Such complex impressions may have similar representative contents, and that which is similar in them may be introduced to other people by descriptions.10

We presume that our ordinary ways of language-learning somehow provide us with similar or very close understandings of meanings of many words and expressions which we collectively use. But if meanings are conventional, conventions must rather cause these learning effects than be caused by them. However if a conventional meaning is to be introduced by a description, since descriptions always use names, meanings of some names must be somehow given prior to conventions. Then either some such meanings must not be conventional or at least some conventions about meanings must be provided by direct demonstration of patterns of individual meaning-assignment which is impossible. Therefore, even if some conventions may be established through introduction of impressions as meanings of names this may not be a universal method of establishing and preserving conventional subjective meanings.

But anyway, it is doubtful that meanings can be effectively introduced by descriptions, i.e. so as to make the receivers of the introduction be disposed to assign those meanings. From introductions like (1) we can know that in communication members of the convention should use "hound" instead of "dog" in certain contexts, i.e. where “dog” is normally used to refer to dogs. This convention is apparently about the use of a name; but is it about its meaning? An agent may use “hound” instead of “dog” and nevertheless assign different meanings to these two words so far as she understands what “hound” means outside the convention (or at least that it has outside it some standard meaning which is different from that of “dog”). We usually see the difference between using a word so as to cause certain public effects (e.g., pronouncing the word along with pointing at certain thing) and using it with certain meaning. In order to infer a meaning of “hound” from such introduction as (1) its recipients must have no antecedent meaning for this word. But even in this case (1) would only tell what to do with the word, not how to understand it.11

The claim like (2) “Let's mean dogs by ‘hound’” only seems introducing meaning more explicitly; for “to mean” usually means the same as “to call” under one of the readings described above. If it means the same as “to (try) to refer to certain (kind of) thing by using the word thus defined”, it does not define what thing the word should refer to, it only names it by “dogs”.

Descriptions like (2) may be further extended so as to list all (allegedly) essential properties of the referent of the word, as in: (3) “‘Aristotle’ means ‘the great ancient philosopher who was born in Stagira, taught Alexander the Great, wrote Metaphysics…’.” Since different people ordinarily infer their understandings of referents of names they commonly use from different descriptions, conventions based on close similarities of objective content of subjective impressions provided by descriptions are rare and cannot be the norm of communication.12

Besides, (3) also may be read as introducing a relation of mutual substitutability between the name and the description, hence as a definition of the name's use rather than meaning. But even if read as an introduction of a reference it says only that the name should refer to the same thing as that which the description of the properties following “means” refers to. But the referents of the name and the description may not coincide; thus, if Aristotle was not born in Stagira the description introducing the referent of “Aristotle” in (3) would not have referent, whereas the name still could. Often descriptions purporting to define meanings of proper names may be satisfied by more than one object in the world. And anyway, they always may be said to introduce not referents of certain names but whatever happen to be bearers of the properties which they list.13

3.    Coordination and common knowledge

The most popular conception of the nature of social conventions stipulates that the conventional behavior is based on specific reasons, i.e. beliefs that other members of convention do the same (their parts of a conventional job).14 The most famous 20th century version of this view was proposed by D. Lewis. According to him conventions are means of coordination. It is common to suppose that the need for coordination generally emerges when two or more individuals have some common purpose (such that each of them tries to achieve it in the current situation), neither of them can achieve it in the given situation individually, and its achievement is possible in this situation only through some combination of actions, one for each participant (Millikan 2005: 9). There may be more than one such combination accessible in the situation; some of them may be relative to a chosen system of coordinates: e.g. some may be simpler to realize or maximizing the utility for every participant in comparison with the rivals. But they all may be recognized as conventions.

Conventional behavior according to Lewis is rooted in rational expectations concerning the behavior of other participants of an interaction, and their rational expectations concerning the behavior and rational expectations of other participants. Convention then is a many-leveled construction constituted by common beliefs of its members (beliefs with the same propositional content which all members of a group share) which Lewis calls common knowledge . He proposes that "it is common knowledge in a population P that ___ if and only if some state of affairs A holds such that: 1) everyone in P has reason to believe that A holds, 2) A indicates to everyone in P that everyone in P has reason to believe that A holds, 3) A indicates to everyone in P that___" (Lewis 1969, 56). And "a regularity R in the behavior of members of a population P when they are agents in a recurrent situation S is a convention if and only if it is true that, and it is common knowledge in P that, in any instance of S among members of P , 1) everyone conforms to R , 2) everyone expects everyone else to conform to R, 3) everyone prefers to conform R on condition that the others do, since S is a coordination problem and uniform conformity to R is a coordination equilibrium in S " (Lewis 1969, 58).¹⁵ Thus the conventional behavior in this model is based not only on the readiness of each member of convention to conform to it but also on the belief of each of them that all other members have the same motivation. In general "the expectation of conformity to the convention gives everyone a good reason why he himself should conform" (Lewis 1983, 167).

Lewis defines a language as a function that assigns truth-conditions to sentences. There are infinitely many possible languages in theory, but in practice a population P uses a certain language L iff in P prevails a convention, sustained by a general interest in communication, according to which a speaker tries to avoid uttering sentences not true in L , and believes that sentences uttered by other speakers are true in L (Lewis 1983, 167).¹⁶ Speakers who intend to communicate

Alexey Chernyak / ΣΧΟΛΗ Vol. 11. 1 (2017) 87 with others then have reasons to conform to the convention which prevails in the population.

Lewis says only that meanings of sentences of a language are conventional under the proposed conditions; his theory is not about meanings of names. Meanwhile, it is fairly obvious that often, though perhaps not always, we understand meanings of sentences by understanding meanings of their linguistic parts, including names, and applying to them some compositional rules in order to infer the meaning of the whole from the meanings of its parts. But Lewis's theory may be adapted to explain meanings of names as well. Roughly it may be said that there is a convention about the meaning of a name n in a population P if only it is common knowledge in P that n means x (in certain context c ), everyone intends to use n to mean x (in c ), and avoids using it otherwise (in c ), everyone expects everyone else to use n this way (in c ) , and everyone prefers to use n (in c) this way on condition that the others do the same.

It is easy to see that in order to achieve such level of coordination the members of convention need to have access to each other's minds: each has to be capable to know or infer with high probability that the others mean (intends) the same as she does by using certain name in order to keep using the name conventionally. If conventions about names' meanings were only conventions about their uses the coordination in question would be easier to achieve. But if meanings are subjective conventions about them should provide coordination of subjective meaning-assignments or intentions. And no empirical reason a person could have concerning such matter seems to be good enough.17 Similarities of observable pieces of personal behavior which constitute certain tokens of a name's use are consistent with occurrences of these tokens with different intentions or subjective meaning-assignments.18

4.    Natural conventions

Unlike being grounded in certain rationality being a matter of conformity seems to be necessary condition of convention. The conformity need not be conscious though; it may consist in mere reproduction of conventional patterns. It is also not necessary for a convention to be conformed to by each of its members in each relevant situation; but some agents in some relevant situations should act so as to be seen as conforming to the convention.²⁰ Thus according to R. Millikan the use of a natural language is based on what she calls natural conventions which consist in reproducible patterns of behavior and some principles of their reproduction. She writes “Specific linguistic forms survive and are reproduced together with cooperative hearer responses because often enough these patterns of production and response benefit both speakers and hearers. Like conformity to other biological norms, conformity to these patterns need not be universal or even average…. Conformity is needed only in a critical mass of cases, enough to insure that the cooperative use constituting the norm—the convention—continues to be copied, hence continues to characterize some interactions of some speaker– hearer pairs” (Millikan 2005, vi). Reproduction of a pattern, from her point of view, does not necessarily require a conscious act of copying; it may be unconscious as in the case of a choice of a proper social distance in conversation (Millikan 2005, 6).

The conformity may be generated in different ways. It is likely that a conventional action may be chosen because it has proved to be effective in the past or because it is simpler or more convenient to be reproduced in certain circumstances in comparison with its rivals, or yet because it is a matter of habit or cus- tom to act this way in certain circumstances within some community or group. Also, a person may choose a conventional action just because she knows no other way to do what she intends in her current situation. Millikan believes, though, that a pattern of behavior is conventionally reproduced by a group partially due to the weight or significance of its precedents for the members of a group (Millikan 2005, 2, 7). According to her, although behavioral patterns may be reproduced because they are functional, there are conventionally reproduced patterns which either are less useful than their rivals or at least have functional alternatives. But truly conventional behavior wouldn't have place if there was no pattern of this behavior in the past which this behavior reproduces. On the other hand if past patterns were different the actual convention would be substituted by another one (Millikan 2005, 7).

Indeed some samples of presumed conventionality, like wearing green at St. Patrick's Day or using chopsticks, seem to be reproduced in spite of the fact that the same people wouldn't be worse off if they used other patterns.21 If we agree to count such behavior as conventional we can explain, in particular, why even some utterances which sound meaningless or confusing to the audience may be conventional.

Millikan recognizes that reproduced “speaker–hearer patterns generally involve inner as well as outer acts so that direct copying is not a means of reproduction for the whole” (Millikan 2005, 6).22 She proposes that these patterns are reproduced by way of a counterpart reproduction when a hearer, reacting to an utterance, chooses the interpretation which would fit it, or a speaker chooses the utterance so as to fit (cause) certain hearer's reaction (Millikan 2005, 4, 6).

But as a linguistic or phonetic form an utterance (the production of sounds or inscriptions) may mean whatever or nothing; when we say that the interpretation fits an utterance we usually mean that it fits the utterance's meaning or the speaker's (writer's) communicative intention. In order to fit an utterance in this sense the interpretation should be chosen between possible alternatives according to the rules of a language and relative to the meaning of the utterance in the conversation or text. In many cases this meaning is derivative from or at least correlated with what the speaker (or writer) intended the utterance to mean.23 In such case an interpretation may not be literally chosen to fit the corresponding utterance unless it may be chosen to fit the meaning which this utterance had to have from the point of view of its speaker. But in order to be fitted by the interpretation the speaker's meaning or intention should be given to the interpreter in communication; and this cannot be accomplished in the current state of the reality. Nevertheless, a reproduction of a pattern may take place in spite of the fact that some of its parts may not be literally chosen to fit some other parts. When someone in order to greet someone else chooses the phrase which is commonly used by her group as a normal mean to this end – “Greetings” – and the one who is thus greeted understands the utterance as a sign of greeting, just as many other persons do in the same situation, this seems to be a perfect example of a reproduction of a communicative pattern. If we agree to count thus reproduced patterns as conventional, then we seem to have many examples of linguistic conventions governing not only uses of linguistic (and phonetic) forms, but also their meanings.

But how this applies to names? The pattern in the proposed model must become a norm of communication in order to count as conventional. This means that it must become normal for different users of a name N in different situations to refer to the same thing x by N, and to assign to N the reference to x in interpretation which is the reaction to someone's use of N in a relevant context. But if meanings of names are subjective nothing would prevent different agents of speaker-hearer pairs constituted by the uses of N as a name for x from identification of different things as x according to their personal impressions of x which are different due to different ways by which they were acquainted with x , and modes under which x was presented to their minds. The reproduced patterns then will probably not consist in coincidences of meanings assigned to the name by speakers and hearers correspondingly. This idea of convention is consistent with there being different and even inconsistent meanings for the same expression in the contexts of its normal use and interpretation.

Suppose the case of an agreement which Robinson Crusoe and Friday from the famous D. Defoe's story establish relative to their mutual use of the word “Friday”. First, Robinson points at Friday and says “Friday”; then Friday reacts by pointing at himself which he accompanies by saying “Friday”. Robinson understands this as a sign that Friday grasped the meaning which Robinson wants to assign to the word in their mutual use of it; he repeats his previous action and utterance, and Friday understands this as a sign that he properly understood the

Robinson's original intention. Suppose that in their further interactions both Robinson and Friday will normally try to use and understand “Friday” as referring to Friday. It seems pretty obvious that in many cases these intention and interpretation will be perfectly coordinated: the intention will be fitted by the interpretation satisfying it, and vice a versa. But what sort of satisfaction do we have here? Will it provide coordination of meanings?

Still Robinson's and Friday's ideas of the reference of the name may, and most likely would (if the case was real) be based on different personal conceptions of Friday derived from different pieces of past experience in which Friday was represented in different ways to Robinson and to himself. They may and quite plausibly would identify different fragments of the same world as Friday,24 and hence as the referent of the name in question.

The established norm of the word's use and interpretation may serve to the mutual benefit of both members of the small community under consideration; but this may be partly because the differences between personal concepts of Friday do not affect interactions between Friday and Robinson. Nevertheless interactions based on such pattern would remain counterfactually sensitive to the differences in meanings mutually assigned to the term within this pattern by the agents of its reproduction, if affected by these differences the mutual benefit from the reproduction of the pattern would decrease relative to its current level. On the one the hand this shows that even if subjective meanings assigned to the name by the agents of its conventional use and interpretation were different, the communicative pattern, construed according to the analyzed model, might be the same. Therefore these meanings don't look being chosen due to the weight of the pattern. On the other hand the counterfactual sensitivity of communication to differences in subjective meanings shows that these sorts of meanings may not be completely substituted by ways of a name's use or whatever capable to be understood as both constituting reproducible communicative patterns and providing coordination of meanings as a result of such reproduction.

I suppose that most patterns constituted by normal uses and interpretations of at least proper names of natural languages are sensitive to differences in subjective meanings in this respect even if actually beneficial for their agents in spite of these differences. As I see it, this feature of collectively reproduced patterns of uses and interpretations of names shows that if they are conventions, they govern conventional uses of names and some publicly observable reactions to them, but not their meanings and interpretations.

Besides, if subjective meanings assigned to the same name tend to be different the patterns in question seem to be reproducible only within groups constituted by the same persons; no such pattern may survive the substitution of the members of a group by other persons with the same group features but different mental contents. The proposed notion of convention does not explain how we can learn from other persons how to use a name with its conventional meaning or interpret it conventionally. In order to be publicly learnable conventional meanings still need to be intersubjectively introducible, and not only collectively reproducible.

Conclusion

Anyway, it is still possible that each member of a group intends to coordinate with its other members in communication and assigns to the name N a meaning which is by chance or due to some natural regularity similar to that which all or the majority of the other members of a group assign to N in the same contexts. Although no participant of such collective use of a name could check out whether she is right assuming that the others use the name with the same meaning, she may be luckily right in this respect. This collective outcome may be then considered as a coordination of subjective meanings. And as we saw above there is the notion of convention which allows considering such collective activities as conventions.

After all, it may be even claimed that conventions need not be necessarily social. Thus, for example, R. Carnap regarded conventional any stipulation resulted from an individual free choice among alternatives (Carnap 2002). This simple view reduces conventionality to mere arbitrariness of choice. It may be objected that arbitrariness is not an essential feature of conventions though: if, e.g., arithmetic is conventional it seems to be an instance of convention established in the absence of a functionally adequate alternative, i.e. anything which the same peo- ple could use as a mean of solving the problems which arithmetic actually solves.25 But even if there are conventions constituted by mere arbitrary choices among alternatives or reproductions of communicative patterns in which similar subjective meanings are assigned to the same names in the same contexts in spite of the lack of coordination between the agents of this communication necessary for such similarities, they cannot explain the majority of meanings assigned to the names in everyday communication. For in too many cases subjective meanings may be different. Therefore, the problem does not vanish: it is still unclear, how subjective meanings assigned to names of natural languages may be literally conventional.26