blog.kennypearce.net

Implicature and the Interpretation of Foreign Language Texts

I've just read Grice's "Logic and Conversation" (ch. 2 in Studies in the Way of Words) for (I'll admit) the first time. Something that struck me while reading it, which Grice does not explicitly recognize, is that his model helps to explain a phenomenon that causes a lot of trouble when one tries to interpret texts (or speech) in a language in which one is not fully fluent.

Grice's basic model works like this: sometimes a speaker says something which, taken in its perfectly straightforward sense, seems quite odd. The oddness (at least in the cases in which Grice is most interested) comes from the fact that the utterance somehow violates the ordinary rules of conversation, rules that say things like "do not say what you believe to be false" or "be relevant" (p. 27). In such a case, it may be that if some other proposition, besides the meaning of the speaker's sentence, were also true, then the maxims would not be violated. To use one of Grice's examples, "A asks B how C is getting on in his job, and B replies, Oh quite well, I think; he likes his colleagues, and he hasn't been to prison yet" (p. 24). In this example, B implicates that C was (for some reason) likely to go to prison if he didn't "get on" well at his job. This is implicated because if it wasn't true B would have said something irrelevant to the conversation.

Now, the thing that stuck out to me came from a pair of examples Grice gives later in the paper, in talking about what he calls 'generalized' implicature:

Anyone who uses a sentence of the form X is meeting a woman this evening would normally implicate that the person to be met was someone other than X's wife, mother, sister, or perhaps even close platonic friend. Similarly, if I were to say X went into a house yesterday and found a tortoise inside the front door, my hearer would normally be surprised if some time later I revealed that the house was X's own. (p. 37)

The reason these examples struck me is that they clearly don't hold up cross-linguistically. In the first case, in several languages (including at least ancient Greek and modern German) the word for 'woman' also means 'wife'. Now, 'a wife' might or might not, in such a language, imply a wife other than his own. However, secondly, the use of articles and possessives is highly variable across languages. Furthermore, many languages lack one or both of the articles English has, and some languages have other articles.

This suggests to me that Grice's position is in need of some modification. First, a sentence might not strictly speaking be at the right level of informativeness, but not be considered the least bit odd simply because it is so commonly said. For instance, in English, when we say that someone is eating, it is extremely common to specify what meal she is eating. As a result, even though the question of which meal she is eating may be (a) irrelevant and (b) deducible from the time of the utterance anyway (and so violate the maxim of Quantity by being unnecessarily prolix; p. 27), the inclusion of this detail will not implicate anything. On the other hand, if a fluent speaker uses an uncommon turn of phrase or choice of words, even if it is at the right level of informativeness, is relevant, etc., the hearer will be left trying to determine why the speaker chose those words, and this may generate an implicature. The easiest examples of this are with word choice. For instance, in contemporary American English, 'to go for [or on] a walk' is much more common than 'to go for [or on] a stroll.' However, 'walk' and 'stroll' are near synonyms. If I say 'he went for a stroll', my choice of this unusual word will cause my hearer's attention to be drawn to the subtle differences between walking and strolling, and so will have the implication that the event in question was especially leisurely or enjoyable.

The 'oddness' of a sentence depends on what one is expected to say in such circumstances. The implicature is generated by the search for an explanation of the discrepancy between expectation and result. As a result, in order to know what implication is carried by an utterance one must know what is ordinarily said (including what words and syntactic structures are ordinarily used) by speakers of the language in that context.

This brings us to the cross-linguistic problem I mentioned at the beginning of this post. Consider this example: the Greek word logos carries with it a lot of philosophical and theological baggage. As a result, interpreters of (e.g.) the New Testament who don't know much Greek but are armed with Strong's Concordance may be inclined to read much or all of this baggage into every occurrence of logos. However, as anyone whose spent much time reading ancient Greek knows, logos is an extremely common word with a lot of perfectly mundane uses. For instance, it is the normal word for any kind of public speech. So if a NT writer calls a particular public speech a logos, this doesn't imply anything at all. It is simply the normal mode of expression.

The moral of the story: in order to know whether it's appropriate to go digging for implications, one has to know whether there is something unusual about the utterance, interpreted literally and straightforwardly. In order to know this, one has to know how speakers of that language ordinarily talk in that situation, and this requires a lot of familiarity with the language in question.

Posted by Kenny at July 6, 2010 2:54 PM