Computational and evolutionary aspects of language
Nowak et al. (2002): Computational and evolutionary aspects of language
This paper starts with a highly interesting abstract, which makes the reader curious about what is to come:
Language is our legacy. It is the main evolutionary contribution of humans, and perhaps the most interesting trait that has emerged in the past 500 million years. Understanding how darwinian evolution gives rise to human language requires the integration of formal language theory, learning theory and evolutionary dynamics. Formal language theory provides a mathematical description of language and grammar.
According to the authors, the genetic code was a "generative system", and until "very recently", it was the only one, which changed when human language emerged.
It enables us to transfer unlimited non-genetic information among individuals, and it gives rise to cultural evolution.
What is the aim of this paper? The authors list several ones:
Currently there are many efforts to bring linguistic inquiry into contact with several areas of biology including evolution, genetics, neurobiology and animal behaviour. The aim of this Review is to formulate a synthesis of formal language theory, learning theory and evolutionary dynamics in a manner that is useful for people from various disciplines. We will address the following questions: What is language? What is grammar? What is learning? How does a child learn language? What is the difference between learning language and learning other generative systems? In what sense is there a logical necessity for genetically determined components of human language, such as ‘universal grammar’? Finally, we will discuss how formal language theory and learning theory can be extended to study language as a biological phenomenon, as a product of evolution.
The authors start off by writing about formal language theory, stating that there is a "fundamental aspect of human language that makes it amenable to formal analysis: linguistic structures consist of smaller units that are grouped together according to certain rules" and that "[i]ndividual languages have specific rules". What follows is the classical formal language theory as taught at universities (e.g. in courses on "Theoretical Computer Science"), including the Chomsky hierarchy. You may also read an introductory article which I once wrote about this topic .
Afterwards, a chapter on learning theory follows, in which the authors state:
Learning is inductive inference. The learner is presented with data and has to infer the rules that generate these data. [...] Neural networks are an important tool for modelling the neural mechanisms of language acquisition. The results of learning theory also apply to neural networks: no neural network can learn an unrestricted set of languages.
Then, the topic is evolutionary language theory. Among other things, the authors write in this chapter:
The central question of the origin of human language is which genetic modifications led to changes in brain structures that were decisive for human language. Given the enormous complexity of this trait, we should expect several incremental steps guided by natural selection. In this process, evolution will have reused cognitive features that evolved long ago and for other purposes. Understanding language evolution requires a theoretical framework explaining how darwinian dynamics lead to fundamental properties of human language such as arbitrary signs, lexicons, syntax and grammar.
These basic statements are followed by a couple of mathematical formulae which are supposed to illustrate evolutionary dynamics.
The article also contains an excursion on statistical learning theory.
In the conclusions chapter, the authors end the article with the following words:
The study of language as a biological phenomenon will bring together people from many disciplines including linguistics, cognitive science, psychology, genetics, animal behaviour, evolutionary biology, neurobiology and computer science. Fortunately we have language to talk to each other.