Meaningful mistakes in language behaviour [...]
4 • An exploratory single-case study
See the full set of slides from this talk.
The previous sections set the stage for an introduction to a new concept for studying the evolution of language by reference to the changes occurring in actual linguistic data on a much shorter time-scale. This empirical approach should be considered as complementary to the studies conducted by using (computational, mathematical) simulation models or by using experimental settings such as the one reviewed above.
The modelling approach provides a framework in which computational methods can be applied to measure the dynamics of structural changes occurring in an evolect – a limited representation of linguistic structure. In one line of this research, evolects were designed for testing the hypothesis about a convergence of the early protolects towards a systematically ordered repertoires of symbolic conventions. According to this hypothesis the evolutionary dynamics which gave rise to the origins of a shared language (resources) in a relative small community cannot be explained solely with reference to a biological „capacity for language” (e.g. Hauser et al., 2014).
Rather, the earliest language resources might have emerged under selection pressures which reflected weak cognitive biases in favour of processing and memorising structural analogies perceived in the signalling behaviour of others and ones own performances. Amplified by repeated episodes of transmission, the early germs of the emerged lectal structure could have given rise to the systematic, rule-like organisation of a shared, highly conventionalised communicative behaviour.
In the following I will present two exploratory case studies (one qualitative and one longitudinal study) aiming to address these questions empirically and to identify similar processes of structural adaptation occurring in vivo, in the verbal behaviour of modern language users.
I will consider only a part of the verb paradigm, typical for many modern languages, where as far as the grammatical number is concerned only singular and plural forms will be taken into account, leaving behind the dual form (which of course still exists in some dialects), and not even mentioning more complex paradigms.
In order to better capture the idea behind this specific example, let us take a closer look at the glossogenetic changes in the verb paradigm of English. In the Old English of Anglo-Saxon times there was an explicit distinction between singular and plural form of personal pronouns in the 2nd person in all core tenses. In modern English this distinction has disappeared and only the gender distinction remained in the 3rd person singular of most verbs in the present tense.
Now, if we put this in the perspective set by the Iterated Learning Model, it will help us to embrace the conceptual leap needed to capture the correlation between the conclusions that came from this line of research and the actual dynamics of what is happening on the micro-scale.
This coalescent tree (Fig. 1 in Cornish et al., 2009:194) shows the subsequent stages of the “evolving” evolect. Zooming in, let’s focus on one of the lineages of this developing structure, specifically looking right in the middle of it.
Thanks to the simulation we could observe the mechanism underlying the adaptation dynamics of the evolect. As we can see in the trajectories of changes occurring during this simulation, it is actually driven by transmission errors! They become combinatorially meaningful, in the sense that they rule out the less adaptive variants and replace them with new ones, contaminated with already existing parts of the evolving paradigm.
So, on the one hand these reproductive blends contaminate the previous stages of the evolving structure but on the other hand they make the whole evolect more stable, shaping its form towards a combinatorial solution, which will be more resistant to random mutations. The critical conceptual leap is that we can picture that fragment of the coalescent tree as a tiny moment in the historical time of the glossogenetic changes of a verb paradigm.
Let’s look at an interesting example of a similar process occurring at the present time in the past form of the Polish verb “to go” (which happens to be also a suppletive form, cf. English “went”). Polish is a highly inflected language, but it slowly loses some grammatical features. Many European languages undergo a similar process of gravitating towards more analytic, positional syntax (e.g. German, Dutch). Anyway, Polish is morphologically far more complicated than English at the moment.
In a verb paradigm like this we have to do with a space of ecologically significant meanings, which (in the course of the glossogeny of this particular lect) made themselves available to be occupied by signals – signals which would explicitly articulate the most salient subcategory distinctions. Following the uniformitarian assumption (Christy, 1983), in a hypothetical evolutionary scenario, some kind of such subcategorisation might have naturally arisen out of the ecological circumstances in which the early language users found themselves far back in time, at the very dawn of humankind.
Leaving that speculation aside for the present, let’s turn our attention to the dynamics of the recent glossogenetic changes occurring within this rather restricted meaning space set by the grammatical categories of a verb paradigm. This case study, although conducted on such a minimal example, is transparent enough to observe the effects of the already mentioned selection pressures, which shape the lectal structure in a self-organising manner.
The presented case of the grammatical paradigm of the Polish verb “to go” in the past tense is emblematic for a large number of finite verbal forms in many languages. As already said, it is by far not a universal distribution, however we can safely assume that in most languages the available verb forms can be analysed as belonging to more or less complex morphological paradigms which constrain the variation possibilities of the linguistic behavioural patterns occupying somehow restricted meaning space.
What makes the presented case particularly interesting, are the transmission errors occurring in the actual verbal behaviour of Polish speakers, which are made due to the pressure for learnability. In this part of the grammatical paradigm the masculine singular forms
SZEDŁEM (1st pers. sg.),
SZEDŁEŚ (2nd pers. sg.) change under the influence of the feminine forms into more learnable
SZŁEŚ, whereas in plural we observe an opposite alignment of feminine forms under the influence of the masculine ones.
If we look at that from the evolect-perspective, this reproductive blends ensure better compressibility of the whole paradigm. We do not talk here about regionalisms or dialects but about a standard variety of Polish, the social status of which has been very high, at least in recent history. And the verb itself is also a frequent one because it expresses a very basic and useful conceptualisation. Although at present both variants are used by Polish people, often even by the same person depending on social circumstances (see below), the forms
SZŁEŚ are more naturally acquired by children. This is the best indicator of the ongoing alignment of the lectal structure constrained by a meaning space, due to the learnability pressure imposed by its cultural transmission.
But there is another phenomenon here, which makes the case even more interesting. We observe a constant fluctuation between the variants mentioned above, especially the 1st and 2nd person masculine singular forms of this verb, i.e.
SZEDŁEM, SZEDŁEŚ) vs.
The second variant tends to be easier to learn and to use because of the irresistible analogy to the respective feminine forms.
And this has some significant socio-cultural implications. If you do not want your kids – boys, because gender is crucial in this case – to be corrected by teachers at school, you will have to suppress their natural inclination to say
SZŁEM (as well as po
SZŁEM and all the other variants with productive prefixes) and redirect them to use
In that respect this case is also a perfect example of the adaptive influence of the third selection pressure on linguistic structure, namely the pressure for social conformity. The socially conform (in the normative sense) variant
SZEDŁEM is the one that is far harder to master.
The incredible power of the pressure for social conformity manifests itself in the socially integrative behaviour of Facebook users: this Facebook page is entitled: “Language mistakes you’re making everyday of which you’re not even aware of”.
Although the easily learnable variant
SZŁEM is by now far more frequent than the normatively correct one, it is very difficult to provide statistical evidence for that propagation. This is due to the vast amount of meta-narratives and discussions concerning the (in)correctness of these forms, which skew the actual frequency of their usage.
That is why I have conducted another study on an idiomatic construction, the changes in the occurrence of which I managed to measure.
by Arkadiusz Jasiński
published on: 26.09.2015
This work is licensed under a CC BY-SA 4.0
Creative Commons · Attribution-ShareAlike 4.0 International License.
- Christy, C. (1983). Uniformitarianism in Linguistics. Philadelphia/Amsterdam: John Benjamins. ↩
- Cornish, H., Tamariz, M. & Kirby, S. (2009). Complex adaptive systems and the origins of adaptive structure: what experiments can tell us. Language Learning 59(Supl.1), 187–205. ↩
- Hauser, M.D., Yang, C., Berwick, R.C., Tattersall, I., Ryan, M., Watumull, J., Chomsky, N. & Lewontin, R. (2014). The mystery of language evolution. Frontiers in Psychology 5(401), DOI: 10.3389/fpsyg.2014.00401. ↩