iterated learning – Replicated Typo

Chomsky on Language Evolution

Learning new skills is a fascinating process that often involves a complex interplay between biological predispositions and cultural transmission. The evolution of language, in particular, has captivated scholars and researchers, leading to intriguing debates surrounding its origins and development. Recently, esteemed linguist Noam Chomsky delivered a thought-provoking lecture at UCL, where he discussed the poverty of the stimulus and critiqued iterated learning experiments. These experiments, which explore language evolution and artificial language learning, have sparked discussions about the role of human intelligence and the nature of evolutionary processes. As the field of communication evolution continues to evolve, new avenues such as government compliant LMS offer promising opportunities to enhance our understanding of how skills are acquired and disseminated.

Noam Chomsky gave a lecture on the poverty of the stimulus at UCL responding to topics such as language evolution and artificial language learning experiments. From about 89 minutes in he discusses iterated learning and language evolution, saying the conclusions derive from “serious illusions about evolution”:

Chomsky’s criticism of iterated learning experiments (see post here and here) is based on two points. First, the emergence of structure is more to do with the intelligence of the modern humans taking part in the experiment than a realistic language evolving scenario. He suggests that structure would not emerge in a series of computer programs without human intelligence. As as a colleague pointed out, however, the first iterated learning experiments used computational models of this kind. Secondly, he suggests that the view of evolution employed in the explanation of these systems is a pop-psychology, gradual hill-climbing one. In fact, Chomsky claims, evolution of traits such as language or eyes derive from single, frozen accidents. That is, evolution moves in leaps and bounds rather than small steps (Jim Hurford recently gave a lecture entitled ‘Reconciling linguistic jerks and biological creeps‘ on this topic). Why else would humans be the only species with language?

Geoffrey Pullum counters this last point by asking why would an innately specified UG emerge so rapidly, but then freeze for tens of thousands of years, when (borrowing Phillip Lieberman’s point) traits such as lactose tolerance have emerged in the human genome within two thousand years. Chomsky gives some examples of traits that have developed rapidly, but then only changed marginally.

I don’t think that proponents of iterated learning paradigms would have a problem with a sudden emergence of a capacity for advanced linguistic communication. Although there is a continuity between human and non-human communication systems, we have some tricks that other animals don’t (see Michael’s post here). However, the evolution of the structure of language after these mutations could owe a huge amount to processes of cultural transmission. The universals we see in the world’s languages, then would be an amplification of weak biological biases.

However, Chomsky seems disillusioned with the whole field of what he calls ‘the evolution of communication’. At least we didn’t get it as bad as exemplar theory, which he dismisses as “so outlandish it’s not worth thinking about”.

[Edit: I originally attributed Mark Liberman instead of Phillip Lieberman. Now I’ve made this error in both directions!]

Compositionality and Bilingualism

Last week I put up a link to an online experiment. Here’s the results! You can still do the experiment first, if you like, here. Source code and raw results at the bottom.

Languages evolve over time under a pressure to be learned by a new generation. Does learning two languages at once effect this pressure? My experiment says … maybe.

These pressures include ones for learnability (compression) and expressivity (able to express a large variety of meanings, Kirby, Cornish & Smith, 2008). Bilingualism seems like an unlikely ability since learning an extra language leaves the speaker potentially no more expressive at a cost of an increase in the amount of effort required to learn it. There is no pressure for one language structure (e.g. English) to adapt to another language (e.g. Mandarin) so that they can become optimally learnable and expressive as a single medium. That is, there’s no reason to assume that expressivity and learnability pressures apply across languages (which are not being used by the same people).

Nevertheless, children display an aptitude and a willingness to learn and use multiple languages simultaneously, and at a similar rate to monolingual children. Therefore, languages do seem to have adapted to be learnable simultaneously. Does the compatibility of languages point to a strong innate property of language? In contrast, it might point to underlying similarity in the structure of languages, brought about by universal principles of communication.

Continue reading “Compositionality and Bilingualism”

Fun language evolution experiment!

Do a fun language experiment!*

You can take part in a pilot experiment about language learning: It takes about 8 minutes (and is NOT an iterated learning experiment, although it looks a bit like one). I’ll release the results (and the hypothesis) right here on Replicated Typo.

CLICK HERE!

* may not be loads of fun.

Passwords adapt to hacking technology

One of this week’s xkcd comics makes the point that combinatorial passwords (sequence of common words) may be better than holistic ones (semi-random string). This may be because we’re fooled into thinking that a password that is difficult to remember will be difficult to guess. This turns out not to be the case. I’m currently thinking about whether combinatoriality would emerge from an iterated learning chain even if the participants were told to give answers that they thought nobody else would give.

Language Evolution and Tetris! Part 2

Ok, so my previous experiment was an incredible failure. The program crashed in sixteen different ways, including suddenly deciding not to respond to key presses for no apparent reason. A rather lazy Ghost in The Shell. Although about 8 people participated, the data was unusable. What on earth was I trying to achieve?

The experiment was a typical human Iterated Learning experiment (e.g. Kirby, Cornish, Smith, 2008) – there were a set of meanings (Tetris blocks) which varied along two dimensions (shape and colour). Participants were shown the words for half of the meanings, but then asked to recall words for each meanings. These responses were then given to the next participant as input. Over time, other such experiments result in meanings which are compositional and more learnable. However, the meaning space tends to ‘collapse’ as the same label is applied to many meanings.

I was trying to do an iterated learning experiment which teased apart the difference between labelling a form and labelling a function. If participants label the function of an object, the environment will play a greater role in the evolution of the language.

There were two chains – one played Tetris where you have to complete lines to score points – colours are irrelevant. The other chain played “Coltris” where you scored points by placing more than 4 blocks of the same colour next to each other. Also, each individual block in a brick finds its own lowest point (i.e. the brick breaks apart), meaning that shape is much less important. That is, for Tetris, the functionally salient feature was shape while for Coltris it was colour.

What I was hoping was that, for the Tetris players, the signal space would ‘collapse’ in the colour dimension. That is, labels would distinguish bricks by shape, but not colour. For the Coltris, the opposite should have happened – labels would have distinguished bricks by colour but not shape.

Gary Lupyan has shown that naming categories of objects can affect your perception of those objects (Lupyan, G. (2008). The Conceptual Grouping Effect: Categories Matter (and named categories matter more). Cognition, 108, 566-577.). My experiment looks into where those distinct category names came from in the first place. Having said this, the experiment would have been more neat than illuminating.

Oh Well.

Language evolution in the laboratory

When talking about language evolution there’s always a resistance from people exclaiming; ‘but how do we know?’, ‘surely all of this is conjecture!’ and, because of this, ‘what’s the point?’

Thomas Scott-Phillips and Simon Kirby have written a new article (in press) in ‘Trends in Cognitive Science’ which addresses some of the techniques currently used to address language evolution using experiments in the laboratory.

The Problem of language evolution

The problem of language evolution is one which encompasses not only the need to explain biologically how language came about but also how language came to be how it is today through processes of cultural evolution. Because of this potential ambiguity arises when using the term ‘language evolution’. To sort this ambiguity the authors put forward the following:

Language evolution researchers are interested in the processes that led to a qualitative change from a non-linguistic state to a linguistic one. In other words, language evolution is concerned with the emergence of language

Continue reading “Language evolution in the laboratory”

Bayesian Bilingualism

Recently, David Burkett and Tom Griffiths have looked at iterated learning of multiple languages from multiple teachers (Burkett & Griffiths 2010, see my post here). Here, I’ll describe a simpler model which allows bilingualism. I show that, counter-intuitively, bilingualism may be more stable than monolingualism.

Continue reading “Bayesian Bilingualism”

Learning Multiple languages from Multiple teachers

As Niyogi & Berwick (2009) point out, there is a tendency in modelling of Linguistic Evolution to assume chains of single learners inheriting single grammars from single teachers. This is, of course, not realistic – we learn language from many people and people can speak more than one language. However, Niyogi & Berwick suggest deeper objections.

Continue reading “Learning Multiple languages from Multiple teachers”

Words as alleles: A null-model for language evolution?

For me, recent computational accounts of language evolution provide a compelling rationale that cultural, as opposed to biological, evolution is fundamental in understanding the design features of language. The basis for this rests on the simple notion of language being not only a conveyor of cultural information, but also a socially learned and culturally transmitted system: that is, an individual’s linguistic knowledge is the result of observing the linguistic behaviour of others. Here, this well-attested process of language acquisition, often termed Iterated Learning, emphasises the effects of differential learnability on competing linguistic variants. Sounds, words and grammatical structures are therefore seen to be the products of selection and directed mutation. As you can see from the use of terms such as selection and mutation it’s clear we can draw many parallels between the literature on language evolution and analogous processes in biology. Indeed, Darwin himself noted such similarities in the Descent of Man. However, one aspect evolutionary linguists don’t seem to borrow is that of a null model. Is it possible that the changes we see in languages over time are just the products of processes analogous to genetic drift?

Continue reading “Words as alleles: A null-model for language evolution?”

Iterated Learning and Language Evolution

If we accept that language is not only a conveyer of cultural information, but it is itself a socially learned and culturally transmitted system, then an individual’s linguistic knowledge is the result of observing the linguistic behaviour of others. This well attested process of language acquisition is often termed Iterated Learning, and it opens up a new avenue to investigate the design features of language: that cultural, as opposed to biological, evolution is fundamental in understanding these features.

Continue reading “Iterated Learning and Language Evolution”