Iterated Learning
In the lecture, we’re going to look at some issues with using learning biases to explain language structure (like regularity). The solution to these problems will be to move beyond the individual learner and think of how languages evolve as they are passed from learner to learner over time. This kind of cultural evolution is called iterated learning.
I will briefly cover two experiments that use iterated learning to explain regularity in language. I’d like you to look at the papers behind these experiments after the lecture. You should have a look at the first of these before you do the lab, though.
Reali & Griffiths (2009) covers a model of frequency learning – this is the same model we have been working with in Labs 3 and 4, so don’t be intimidated by the slightly mathsy framing in sections 2 and 3 of their paper, you know how this model works – and 3 experiments. In Experiment 1 they verify they get something that looks roughly like probability matching in their word-learning task (as e.g. Hudson Kam & Newport, 2005, do in a much more complex language task), in Experiment 2 they run an iterated learning experiment, and then in Experiment 3 they do the same thing with a non-linguistic task where people seem to have radically different priors.
Smith & Wonnacott (2010) is a short report on a frequency-learning iterated learning experiment – it’s relatively similar to the Reali & Griffiths (2009) experiment, but the task is framed slightly differently so we can look at the evolution of conditioned variation (remember, one of the interesting features of natural language is that variation occurs but is conditioned on the context). Depending on what topics we settle on for the last few weeks, we may look at extending our current model to handle the evolution of this kind of conditioned variation.
Reali, F., & Griffiths, T. L. (2009). The evolution of frequency distributions: Relating regularization to inductive biases through iterated learning. Cognition, 111, 317–328.
Smith, K., & Wonnacott, E. (2010). Eliminating unpredictable variation through iterated learning. Cognition, 116, 444–449.
This page was written by Kenny Smith and modified by Simon Kirby. All aspects of this work are licensed under a Creative Commons Attribution 4.0 International License.
Project maintained by centre-for-language-evolution Hosted on GitHub Pages — Theme by mattgraham