Skip to Main content Skip to Navigation

Statistical physics of language evolution : the grammaticalization phenomenon

Abstract : This work aims to study grammaticalization, the process by which the functional items of a language come to be replaced with time by content words or constructions, usually providing a more substantial meaning. Grammaticalization is therefore a particular type of semantic replacement. However, language emerges as a social consensus, so that it would seem that semantic change is at odds with the proper working of communication. Despite of this, the phenomenon is attested in all languages, at all times, and pervades all linguistic categories, as the very existence of grammaticalization shows. Why it would be so is somehow puzzling. In this thesis, we shall argue that the components on which lies the efficiency of linguistic communication are precisely those responsible for these semantic changes. To investigate this matter, we provide an empirical study of frequency profiles of a few hundreds of linguistic constructions undergoing one or several grammaticalizations throughout the French language history. These frequencies of use are extracted from the textual database Frantext, which covers a period of seven centuries. The S-shaped frequency rise co-occurring with semantic change, well attested in the existing literature, is confirmed. We moreover complement it by a latency part during which the frequency does not rise yet, though the construction is already used with its new meaning. The statistical distribution of the different observables related to these two phenomenal features are extracted. A random walk model is then proposed to account for this two-sided frequency pattern. The latency period appears as a critical phenomenon in the vicinity of a saddle-node bifurcation, and quantitatively matches its empirical counter-part. Finally, an extension of the model is sketched, in which the relationship between the structure of the semantic network and the outcome of the evolution could be discussed.
Complete list of metadata
Contributor : ABES STAR :  Contact
Submitted on : Thursday, June 7, 2018 - 2:39:11 PM
Last modification on : Friday, June 24, 2022 - 3:31:35 AM


Version validated by the jury (STAR)


  • HAL Id : tel-01753835, version 2


Quentin Feltgen. Statistical physics of language evolution : the grammaticalization phenomenon. Physics [physics]. Université Paris sciences et lettres, 2017. English. ⟨NNT : 2017PSLEE039⟩. ⟨tel-01753835v2⟩



Record views


Files downloads