For this lecture I am afraid you will be reading one of my old papers, Smith (2002). The paper is based around the matrix/network model we have been working with, and introduces iterated learning. Iterated learning happens when learners learn by observing the behaviour of other individuals who learnt in the same way. Language is transmitted by iterated learning: you learn a language by observing the linguistic behaviour of people who learnt their language by observing the linguistic behaviour of people who learnt their language by etc etc etc. The majority of the rest of this course will be spent working with models of iterated learning. In particular we’ll be thinking about what kinds of linguistic systems emerge out of iterated learning, what determines the kinds of systems that iterated learning produces, and what this means for our understanding of language and the human capacity for language.
You have already seen nearly all of the components of the model that I used in this paper, and in the next lab class (after Iterated Innovative Learning Week) you will get to mess about with iterated learning and try to reproduce my results. A couple of things to bear in mind when reading the paper:
- I introduce the model in a rather formal way – I guess I had a mathematical notation fetish in 2002 – but the thing to remember is that you already know how this network model works. You know how meanings and signals are represented, you know how winner-take-all production works, and you know how learning works, with a range of learning rules.
- [I AM ADDING THIS ONE FOLLOWING A QUERY IN THE POST-READING QUIZ] ^ in the paper (it appears in the definition of a rule that is +learner) means “and”.
- This was only the second paper I ever wrote, when I was a very green PhD student (so if you don’t like it, cut me some slack). The reason we are working with the Open Access version and not the journal’s nice typeset version involves a sad sad story. When you publish a paper they send you ‘proofs’ before they print it in the journal – you are supposed to check the proofs and catch all the ridiculous errors that the typesetters have introduced (spelling mistakes galore, missing out or randomly inserting the word ‘not’ in crucial sentences, reversed bits of maths, etc). Since I was only a youngster I didn’t know what the proofs were for and thought they were some sort of weird souvenir, so I just chucked them in a drawer without looking at them. When the paper finally came out I spotted 3 spelling mistakes that the typesetters had introduced in the abstract alone, and wept. So this Open Access version is the only unmangled, accurate one available.
As usual, once you have done the reading, do the post-reading quiz then look at my comments.
References
Smith, K. (2002). The cultural evolution of communication in a population of neural networks. Connection Science, 14, 65-84.