Meaningful mistakes in language behaviour [...]

3 • Back to the actual verbal behaviour

See the full set of slides from this talk.

In the previous section I have briefly reviewed the experimental iterated learning method for studying language evolution dynamics in laboratory conditions. On an example of a simple model of lectal structure – an evolect – we could see how its repeated transmission through generations of learners would inevitably lead to the emergence of compositional syntax.

We could also recognise that there is a kind of deterministic force at work: the linguistic signals, no matter how much unstructured they are at the beginning, will always gravitate towards the already given (in this case artificially created) categorial distribution of ‘meanings’ available in some meaning space.

This finding, repeatedly confirmed through similar modelling work, have revealed that the combinatorial dynamics of the developing lectal structure follows a very general pattern which seems to have the force of a scientific law. To date, however, most studies on this subject were conducted either by using (computational, mathematical) models or by using experimental settings such as the one described above. Hence, the obvious question arises:

Can we find empirical evidence for this phenomenon in (a) natural language?

In fact (as far as I know), there are as yet no studies that try to validate the results of modelling work on evolects with corresponding linguistic data. I would like to present a case study of a single verb paradigm, in the structure of which we can observe exactly the same glossogenetic dynamics as the one implied by the model under consideration.

First of all, comparing the meaning space considered in the presented case study with the set of meanings most often implemented in Iterated Learning simulations, a critical difference needs to be emphasised. As we saw in the previous section, in the simulation designed by Kirby et al. (2008) the linguistic transmission was modelled by means of memorising and recalling names for specific objects. For example, ‘renana’ stood for a red square moving forward. In many similar models evolects are also designed in this manner.

However, while we obviously rely on nominal reference in learning a language and using it for communication, we do so only to a certain extent. Even though the lexical strategies for naming objects make up a substantial contribution to the expressive power of language, it is not the nominal reference but the predicate describing an action that (almost universally) lies at the heart of every verbal expression.

Perhaps, as the results of extensive and ongoing research into mirror neurons (see e.g. Arbib, 2012) would seem to indicate, we should give some thought to the role that the capacity for imitation of others’ bodily actions might have played in the origins of early protolects. To my mind, these findings are sufficiently suggestive to entertain the hypothesis that the development of the very first germs of lectal structure had been initiated by the emergence of conventional signals for performing some socially desirable activities.

JPEG - 109.7 kb

Let us imagine an early Homo sapiens running forward and trying to communicate this very action to its conspecifics every time, when an urgent need for sudden flight arises. Which expression would we use in such situation? In a modern language like English?

As many of you would concur, the most obvious candidate for such an expression would involve some morphological form of the verb „go“.1 Symptomatically, a significant part of the conjugation paradigm of this verb (and many others), in its present form towards which it developed glossogenetically, mirrors exactly the distribution of categories in the meaning space of the evolect designed by Kirby et al. (2008).

JPEG - 170 kb

Furthermore, some of the verb forms belonging to this inflectional paradigm continue to change on the glossogenetic time scale, on trajectories similar to the ones observed in the simulation model. The morphological structures in the spelling pattern are becoming more and more regular, even though the whole paradigm has already reached a high level of compositionality. We can observe this adaptive development in the paradigm structure of the English verb „to go“. In the course of its grammaticalisation only one inflectional distinction (for gender) remained in the 3rd person singular. In fact, most English verbs in their present tense paradigm have followed a similar trajectory of historical change.

In the next section I will discuss this phenomenon more thoroughly, using an illustrative example of a corresponding verb paradigm in Polish. One last remark is due on the relation between this study and the iterated learning simulation reviewed here. As I have tried to show, the modelling work has revealed that the process of cumulative cultural transmission can have fundamental implications for the origins and further development of lectal structure. We must keep in mind, however, that the evolect itself consisted of artificially designed, randomised signals for naming an abstract object, whereas the conclusions of this research were formulated with reference to the structure of a whole language.

I have chosen to explore the inflectional paradigm of a verb – a really existing, commonly used word expressing an action (or a condition of a person). The main reason for that was that verb paradigms in many modern languages display well-structured, clearly delimited meaning spaces which are very close to the one designed for the evolect study of Kirby et al. (2008). The crucial difference is that the categorial distinctions in a verb paradigm have not been artificially set but are still being implicitly imposed by the constraints of communicative interactions.

Although the evolutionary dynamics under study cannot be controlled in the same way, it is exactly the similarly limited space of combinatorial possibilities which allows us to compare the naturally arisen changes in the structure of a verb paradigm with the evolect structure developed in laboratory conditions.

In other words, in the search for empirical evidence which (at least to some extent) could validate the predictions from the modelling work, I am going to look at some significant changes in data taken from the actual verbal behaviour of modern humans, the dynamics of which can be digitally captured, statistically analysed and graphically processed by adopting the very same methods and measures.


by Arkadiusz Jasiński
published on: 26.09.2015

Creative Commons License

This work is licensed under a CC BY-SA 4.0
Creative Commons · Attribution-ShareAlike 4.0 International License




  • Arbib, M. A. (2012). How the Brain Got Language: The Mirror System Hypothesis. Oxford/New York: Oxford University Press.
  • Kirby, S., Cornish, H. & Smith, K. (2008). Cumulative cultural evolution in the laboratory: An experimental approach to the origins of structure in human language. Proceedings of the National Academy of Sciences 105(31), 10681–10686.



1 By the way, „go“ is a very good candidate for The Vocal Iconicity Challenge and it actually should have been included in the list of verbs which denote „Palaeolithic-relevant“ meanings. For a similar competition, follow this link.