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

In the figure above biological changes are shown with solid arrows and cultural changes are shown with dashed arrows.

One of the main points in this review are that the processes of biological and cultural change should not be considered in isolation. That is because although biological processes can equip individuals with particular cognitive adaptations, which constrain abilities for linguistic learning,  how languages emerge in dynamic structured populations also needs to be understood. It is this last point of understanding the cultural evolution of language which the experimental techniques explored here are aiming for. This is in the hope that the insights collected using these methods will be able to combine with our understanding from the biology side of things, to get a more complete picture of language evolution.

New experimental approaches

This article explores a new approaches which allow for experimental control whilst extracting data from real human participants. Many of these experiments are based either directly or indirectly on computer models. These models include those of iterated learning as implemented by Kirby et al. (2008) and also studies exploring non-communicative behaviour taking on a communicative role as implemented by Scott-Phillips,T. C. (2010).

I have listed some of the methods used below under headings:

Signal creation

The following methods seek to identify cognitive mechanisms which detect communicative intent and how this influenced the form of signals.

Embodied Communication Game (ECG)

2 participants have to communicate using nothing but movement around a 2x2 grid. This creates a problem in communication in that a strategy must be implemented in order to reveal that the movements are communicative. Because of this participants must lay down some common expectations between eachother so that deviations from these expectations become salient and therefore communicative.

Tacit Communication Game (TCG)

This also uses 2 participants but this time with a 3x3 grid. One of them is assigned as the 'sender' and the other the 'receiver'. The receiver is told to interpret the senders actions within the grid in communicative terms. These expectations take the same role as the common ground of expected behaviour in the ECG experiments.

The emergence of communication systems

Graphical Communication Tasks

One participant draws an intended meaning and the other must guess the referent. This can result, through a process of iterated learning between the pair, in a learned symbolic communication system emerging from an originally iconic one. Systematicity can also emerge from these tasks in that a feature that is common to more than one item is represented in the same way for each different item.

Cultural Evolution

Iterated learning tasks

Iterated learning tasks can be used to illustrate the effects of vertical cultural transmission. This can be done in the lab.

Participants were asked to learn a language that consisted of a series of strings of syllables paired with pictures (i.e. meanings). The set of meanings was structured (each item is one of three shapes that takes one of three colours and travels in one of three ways), but the initial strings were not. Participants were tested on their knowledge of this language and their answers were then used as the training data for the next participant.

This can then demonstrate the effects on the language after this process is repeated over several generations. The features within the language which are effected can include systematicity, expressivness and learnability.

Population dynamics

Garrod et al. (2010) found, using graphical communication tasks, that using methods of vertical cultural transmission drawings retained a much more iconic appearance than when the interaction was only between two people.

These findings, among others summarised in the article, outline the need for the exploration of how different population structures affect the dynamics of evolving communication systems, and how they interact with one another.

These population structures in cultural transmission are displayed below:

In this diagram solid arrows indicate information that is being transmitted culturally. The dotted arrows in the middle figure indicate how the population changes over time.


In conclusion this review selects some of the most prominent laboratory experiments used in the study of language evolution today.This is not a definitive list and there are still many more methods not explored here which Scott-Phillips and Kirby list briefly in their concluding remarks.

The work reviewed which uses real human participants brings new validity to the computational models that have gone before. It also illustrates and explores how individuals interacting can give rise to linguistic phenomena on a level of population and the various different ways in which information can be passed between members of a community.

As well as answering many questions this research has also created problems for further research in that it has outlined that a simple equation between a cognitive model of the human capacity for language and the linguistic phenomena we wish to explain is not possible.

This reasearch is also limited in that it can only use modern humans which is not desirable when trying to come up with a convincing story of how early humans developed language. The point of these experiments is seemingly not to replicate the evolutionary history of language, but to investigate the nature of the various phenomenon explored and, through doing this, to gain insights on how experimental methods can uncover the origins of language.


Garrod,S. et al. (2010) Can iterated learning explain the emergence of  graphical symbols? Interact. Stud. 11, 33–50

Kirby, S. et al. (2008) Cumulative cultural evolution in the laboratory: an experimental approach to the origins of structure in human language. Proc. Natl.Acad.Sci.U.S.A. 105, 10681–10686

Scott-Phillips,T. C. (2010) The evolution of communication: humans may be exceptional. Interact. Stud. 11, 78–99

Scott-Phillips, T., & Kirby, S. (2010). Language evolution in the laboratory Trends in Cognitive Sciences DOI: 10.1016/j.tics.2010.06.006

  • Great post! The Kirby et al. (2008) experiments are really cool IMO and it was really nice to see them appear in the BBC Horizon - Why do we talk? documentary. You're certainly right in pointing out that one the biggest problems of these experiments is that they can only use genetically modern humans. This is all the more problematic given that in the view of Kirby, Scott-Philipps and many others there would have been no genetic change from "protolanguage" to "language." And, as you pointed out, this conviction simply cannot be tested with modern humans who may bring other linguistic or non-linguistic capacities to the tasks of iterated learning and communication games than pre-modern ones.

  • I certainly agree with Michael here, but I do think we can use things like musical or rhythmic "languages" as a proxy for syllable-based ones. I've done a similar experiment to the Kirby et al (2008) experiment, teaching participants an alien language made up of musical intervals, and then asking them to reproduce it. I didn't use iterated learning, as I was looking at literacy and practice, rather than cultural transmission, but what the experiment did tell me is how well people respond to musical stimuli. While it might seem difficult to expect people to learn a musical language (and the analysis was a nightmare), people could do it, and did it well!

  • Pingback: Iterated learning using YouTube videos | Replicated Typo()