feel do think

The individual cognitive turn in linguistics suggests the idea of language as a code and a person using language as a computer – processing information. In this conception, people might operate like intelligent machines, and I think this view may relate to three other big events in post-war science…

1. DNA was a pretty new discovery in the late 1950s, when transformational grammar was appearing (Syntactic Structures – 1957) I think the dual structure of language has some parallels with nature’s way of storing information about proteins – even if there are considerably more phonetic forms (43-ish for English) than bases in DNA and RNA (5 for all life-forms). See Crick’s central dogma (1958)

2. Artificial intelligence was a booming research field in the 1950′s  - with support from the US defence budget at Carnegie Mellon and MIT. People confidently expected the “problem” of AI to be solved rather quickly – or at least they told the sponsors that. It’s worth noting that the best tests of AI are essentially linguistic ones. Wikipedia notes “in 1956 optical character recognition (OCR) was considered AI”. The Turing test is a linguistic test relying on intuition about the “personality” one is communicating with.

3. Neuroscience had few tools at its disposal. (PET scanners first developed in 1974) In the 1950s the idea of “neuroplasticity” was exotic. Many neurologists thought it was impossible for the brain to make any new nerves cells after birth (according to Wikipedia this view only changed in the 1970s). This suggests that the Chomskian “acquisition device” would have to be a physical structure present at birth too… it would be hard to imagine as social. Allowing for neuroplasticity, we don’t need any more to make such a radical choice between nature and nurture when it comes to language – it seems likely to me that the brain structures connected with language grow in response to particular stimuli.

I suppose I don’t really agree with Minsky that consciousness is just a kind of parallel  thinking. I suspect that electronic computers will never “learn to mean”. They are not “embodied” and lack emotions – (feel do think….is the idea of this site). Furthermore it is very hard to see how we could give a machine the biographical experience of growing up in particular home in a particular culture.

There are some useful pointers in this podcast LSE lecture “The Most Human Human“. Brian Christian notes the two main strategies used by programmers trying to get their software to pass for human, either:

a) the interactive book. This is likely to be successful in simulating obsessive speakers who always bring the subject back round to their favourite hobby-horse – thus staying in predictable territory.

b) the corpus picker. Using a huge data base of how people usually respond to particular questions. Here the difficulty is that the replies are probably “stateless”, that is to say, they don’t represent a consistent personal biography and perspective.

In both cases what is lacking is a self- however this aspect of humanity has probably receded during the 20th century. There is a sense in which the “information society” is making humans more robotic.  Christian gives an interesting technical example – the extra delay introduced to cellphone conversations changes the dynamic of turn-taking compared with land-line connections. There is an extra half second added between when you stop speaking and the moment the listener hears you stop speaking. (lag on a land line is 0.1 seconds, a cell phone is 0.6) There will be a 1.2 second period which lasts from the moment when a person who has a made a pause (“I can be done”….giving an opportunity for you to say something else) and then continued…actually finds out that they have ceded their turn to the other speaker.  ”we’ve lost the timing ballet” of turn-taking, says Christian.

Consulting the Routledge encyclopedia of Language Teaching and Learning enabled me to produce this potted history of applied linguistics:

pre WWII – Pure linguistics in the English-speaking world meant American Structural linguistics – this formed around the study of Indian languages which had no written form. Linguists focussed first on their phonology and then grammar. Applied Linguistics appears together with war-time foreign language training for the military. Fairly similar assumptions drive the syllabus of language learning in Russia to the present day – where beginners start with a survey of the sounds of English, and quite likely the phonetic alphabet as well as the Latin one. In this period, applied linguistics meant: “an academically respectable way of talking about language teaching theory”.

70s-80s The period when pure linguistics was focussing on more abstract syntactic concerns (suggested by Chomsky’s approach), coincides with the appearance of the Communicative Approach in language teaching. Applied linguistics was a “refuge discipline” for linguists who saw language in a more social and interactive way…thus the encyclopedia describes it as a period of tension because pure linguistics did not lend itself easily to applications. However, it was a period in which applied linguistics proved a productive source of new approaches and methods in teaching.

Recent years: The broad school of applied linguistics splits into various directions producing socio-, computational , and other “hyphenated linguistics”. Meanwhile EFL teaching has moved from the “communicative approach” being the dominant idea to “task based learning” although the theoretical justifications for it may not be very different. We have arrived, in teaching, at a sort of post-method situation – where teachers may refer to a whole collection of different ideas to discuss their work. Logged-in OU classmates may refer to my week 02 post on contradictions as they appear in my professional life.

There is an extraordinary range of potentialities in different living things. When it comes to language I think there is a radical gap between humans and everything else in nature… Just as there is a radical gap between spiders and everything else in making webs, and between the elephant’s trunk and every other animal’s nostrils. This argument is lucidly deployed by Steven Pinker in “The Language Instinct” (1994). It shows that the symbolic / linguistic realm is a space that only our species inhabits, using arguments that are perfectly consistent with evolution and a scientific view of nature as a whole.

Pinker imagines that if elephants could work as biologists, they would very likely spend a lot of their time discussing the trunk and its “unique place” in nature. (p364 in my paperback). They might pick on the hyrax which shares 90% of the same DNA as the elephant and try to train them to make more use of their nostrils…

picture of a rock hyrax, distantly related to elephants and sea cows but weighing only about 4kg… with thanks to orchidgalore on flickr.

That does not mean of course, that we humans are wrong to pay a great deal of attention to language. But it does perhaps go some way to explaining…(now that biology is becoming top science and a basis for explanations elsewhere) why we should predict that most people will say that monkeys are closer to talking than bees are. Logged in classmates can see a more specific post on this question in my OU blog.