Monday, 13 August 2012

Thoughts on Language

 I apologise for not having updated with anything recently.  I have been extraordinarily busy.  I have also been unable to follow the comments on blogs I read, and so I've unfortunately left some hit-and-run comments on some of them in the spare time I've had.  Anyway, here is a post on some thoughts I've been having about meaning and language.*

It seems inevitable that language should change – and to some extent it is. But language is based on the same decision-making process as any other form of human action, and as such it is dependant on a wider range of variables than we commonly appreciate. By considering variables beyond the desire to communicate efficiently, it is possible to imagine circumstances under which language would not change at all. Consider, for instance, the idea of a group of robots who have taken control of a particular human society. These robots are, for whatever reason, linguistic prescriptivists, and they believe that all humans in this society should speak flawless Received Pronunciation with no deviations or contractions. They punish anyone who says 'didn't' instead of 'did not' with electric shocks, and likewise anyone who fails to distinguish between the vowels in 'trap' and 'bath', whether they were brought up speaking with this distinction or not. They always punish the offender instantaneously, and they are never dissuaded from doing so. Nor do they tire of it, and repeated offenses are simply met with repeated shocks.

It is easy to imagine that the humans in this fortunately hypothetical society would speak as if their language were in stasis, and that if the robo-dominance persisted for a thousand years then the Received Pronunciation spoken in the community would likewise remain unchanged throughout this period. The advantages gained through using contractions and easily-computable forms of speech would be outweighed by the desire to avoid violence at the zappers of the robot overlords. In this case, the desire to avoid physical pain is something that can potentially affect the likelihood of selecting particular word and sentence forms in communicating. The form of human language is dependant on variables such as these, and for a complete understanding of language, you can't see it only as a device that seeks to maximise the efficiency of communication, even though that seems to be the dominant determining variable.

What we see in human natural languages is that they change in different ways and at different rates. Icelandic is the archetypal conservative language; while it would be an exaggeration to say that modern Icelandic speakers can read the medieval sagas with ease, it is certainly true to say that the differences between medieval and modern Icelandic in pronunciation, syntax, and vocabulary are relatively slight. Old and Middle English were, by comparison, spoken entirely differently from modern English as a result of a wave of sound changes that swept the language in the middle of the second millennium CE, and a great deal of new vocabulary has been generated through interactions with an enormous number of widely-separated groups on all continents, from shark (from Yucatec Mayan xok) to kangaroo and bungalow. The English and Icelandic languages are quite closely related, but they have shown entirely different rates of change. Language change may be inevitable in some sense, but certain factors can make it more advantageous for speakers of one language to use certain forms, whether deliberate conservatism and lack of an overseas empire in the case of Icelandic, or growing familiarity with non-Germanic terms for certain objects and the absorption of large numbers of non-native speakers with their own linguistic predilections in the case of English.

Vocabulary is something historical linguists use to track language families and infer historical relationships, but this is not entirely stable either, and if a group of people has used a word descended from an ancient precedent then it is because using this word is more advantageous than using another one. The word for shark before 1585 appears to have been “sea-dog”, a term we now associate with the pirates and privateers of that era. The word shark came into the language through familiarity with the Mayan word in the communities most commonly in contact with such creatures (English sailors in the Caribbean), and the word sea-dog was presumably becoming ambiguous in English due to its associations with both pirates and Selachimorpha. Using shark instead of sea-dog to refer to the large fish would have been a natural step for those wishing to avoid ambiguity in their speech, and the word shark has thus become the standard term in English for the creatures.

There is also an aspect of language commonly called 'social'; the use of certain words, phrases, and phonemes as shibboleths, or words representing allegiance to particular social groups. Sailors and pirates – those cool, dashing heroes of Elizabethan England – would have used the word shark in place of sea-dog as a result of their familiarity with the Caribbean coast of Mexico and its inhabitants. Shark might have served as a sailor's shibboleth, imitated by those who wanted to associate themselves with the dashing pirates (this is a just-so story, not reality, but the mechanism is certainly present). Such 'social' and contextual factors clearly influence the choices people make in using language, whether in deciding to adopt slang, archaisms, or new loanwords.

Perhaps the most important variable in determining which sounds we emit or letters we inscribe in communicating with others is the belief that the listener or reader will understand what it is that you intend to convey.  This operates through common knowledge, or collective intentionality. Common knowledge amounts to a beliefs about the beliefs of others about your beliefs, and so on ad infinitum.  It isn't just your beliefs about the meaning of a word that determine your using it; you must also believe that the hearer will believe something about your belief about the meaning of the word, and so on.  This is a recursive process, and is an excellent example of human language's dependence on recursion. As Richard Elwes puts it in the very handy Maths 1001:
In ordinary language a piece of information X is common knowledge to a group of people if each of them knows it.  Within game theory, this is first-order knowledge.  Common knowledge has a much stronger meaning.  It is required not only that everyone knows X, but that everyone knows that everyone knows X, and furthermore that everyone knows that everyone knows that everyone knows X, and so on.  That is to say, common knowledge is knowledge of every order.  The typical way for a piece of information to become common knowledge is for it to be announced publicly.

When I use the word shark, I am implicitly expecting that you will understand what I mean by the word. The process is actually dependant on the ability to process several orders of recursion. In using the word shark, I believe that you will know what object I am referring to, and you believe that when I use the word shark I am referring to the same object you are thinking of when you use the word shark (or, similarly, that if you don't believe that I'm using the word to refer to the same object, at the very least you have a belief about my belief about the meaning of the word). Likewise, I believe that you believe this, and that you believe that I believe that you believe this, and potentially so on ad infinitum.

This process is notable when speaking with non-native speakers of your language; not only will you speak more slowly, but you will use vocabulary that you believe you are justified in using on the basis of a belief that the person you are speaking to will know what you mean by it. You select words on the basis of your belief that the other person will know, or be able to infer, what you intend by them, and that they will mean what you mean by them, and that they will believe that what you believe they mean by the words will be the same as what...

Of course, it isn't possible for brains to truly process infinity (my mind is utterly boggled after only six or seven orders of recursion, as I expect yours is), and so the recursive aspect of common knowledge must have a limit beyond which we find it impossible to easily compute the information.** And it is equally obvious that this is not always a conscious process. It would require too great a cognitive load to always think consciously about the contents of another's mind when communicating basic pleasantries, and so some scripting of conversation and the use conventional terms and syntax – the stuff we abstract as 'language' – is a natural and obvious thing to do. Instead of having to process this cycle of beliefs about beliefs every time we use any word, our brains unconsciously abstract meanings from the terms employed, as a labour-saving device. 'Shark means a big toothy fish' is one such abstraction, saving you from necessarily processing my use of the word in such a burdensome way. If it seems obvious that I am not using the word shark to refer to a toothy fish, then you can adjust accordingly by using all of the information available to discover just what it is that I do mean.

 We learn language not simply by assigning abstract meanings to arbitrary collections of syllables, but by unconsciously thinking about what others mean by the words they use and by extracting rules from observed consistencies in use in order to minimise cognitive load when communicating. Thought of in this way, it is easy to see how context affects meaning; instead of taking the conventional meanings of words at face value, we factor in other information from our experiences in discovering the meaning behind what the speaker has said. The use of conventional vocabulary and syntax is something people do because it makes life and thought easier for them, not because language is necessarily constrained in this way (although this is a different issue to the possibility of there being an innate mentalese language all humans use to construct their utterances). The consistency of this conventional use and the fact that it is dependant on what individual speakers mean is what allows us to say that 'shark means a big toothy fish' while also expecting that someone could potentially use the syllable shark to express something completely different if given reason to do so.  This is analogous to the way in which we can understand other social facts and other rules of behaviour, whether implicit or explicit ones.  They are abstractions based on some relatively consistent examples rather than entire descriptions of behaviour that constrain individual action.

With regard only to language, what all this means is that when people speak or write in conventional language or 'proper' forms, they are doing so fundamentally because it is expedient, or because it is appropriate for accomplishing the goals the speaker has. It would be impossible to outline all of the variables involved and all of the goals a person might have in using language, just as it would be impossible to outline all of the beliefs and desires people can possibly have, but there are enough commonalities in the reasons people have for using language to allow us to extract some generalisable principles of language change or some usually-applicable statements about what people mean when they utter certain sounds.

This has repercussions for our understanding of genetic relationships between languages. When we are able to establish genetic relationships between languages (ie, between English and Icelandic or Indonesian and Hawaiian) due to the discovery of cognate vocabulary and similar phonemes, it isn't on the basis of a necessary property of language that it rely on conventional forms and inevitable sound changes. Instead, consistent use of the same phonemes and conventional forms is a method for maximising the efficiency of individual speech acts. Languages do not necessarily change, either; they merely do so because the changes are, for whatever reason, expedient, and they change in consistent ways due to consistencies in the ways in which humans maximise the efficiency of what they do rather than because of inevitable, inviolable rules of language change.

Historical linguistics works because people are lazy, or efficient.  Whichever you prefer.

* I'm trying to find a way to condense this position into a much shorter piece, but it is quite a tricky thing to do.

** If you don't believe that there's an upper limit to the ability to process recursion, try to keep track of this: George believes that Margery was lying when she said that Melanie thinks that Kwame holds to be true that Egbert proposes that Napoleon was a rapist.
In any case, the ability to process recursion is probably the most important human cognitive ability, possibly responsible for everything from language to consciousness.  We're still terrible at processing it when it gets too much, though.

No comments:

Post a Comment

You can post anonymously if you really want to, but I would appreciate it if you could provide some means of identifying who you are, if only for the purpose of knowing who has written what.