Accueil > Recherche > Professeurs invités > Invités 2013-2014 > Ted BRISCOE

Dernière modification : 7 décembre 2016


Université de Cambridge (Royaume-Uni)
Invité du Lattice – mars 2014

Ted BRISCOE - invité 2013/2014

En mars 2014, le labex TransferS et le Lattice invitent le Pr. Ted Bricoe de l’Université de Cambridge, Royaume-Uni, pour un cycle de conférences sur le thème de la modélisation de l’évolution des langues.


Topics in Evolutionary Linguistics

Until recently the word « natural » in natural language has served only to distinguish the study of human languages from that of logics or programming languages. Moreover, the distinction has been methodologically without much content since generative linguistics treats human language as a static well-formed stringset and employs the same formal tools - formal language theory, denotational semantics - as theoretical computer science. Many theoretical linguists would probably agree, at least methodologically, with Richard Montague´s famous dictum, « I reject the contention that an important theo­re­tical difference exists between formal and natural languages » (Montague), because not to would appear to preclude the possibility of formal linguistics. One direct consequence has been that the study of language change and variation has remained mostly descriptive and largely outside the realm of at least generative linguistics. Inevitably, the focus of research has been on the individual speaker and her idiolect.

Recent accounts of linguistic change, learning and typology (e.g. Hurford, Niyogi, Kirby, Steels, Croft, Deutscher, etc.) model language as a dynamic system (variation + inheritance through learning), in some cases with adaptation (linguistic selection amongst variants). The crucial shift in perspective in these models is to study (populations of) (generative) language learners and users, as opposed to focussing on the indi­vidual speaker and her idiolect at one point in time. The conclusion that language is a dynamic system is a direct and inevitable consequence of this shift of perspective. One which can be seen as an extension rather than rejection of the insights and achievements of generative linguistics. The study of language as a dynamic system will require a very different research paradigm based largely on computational simulation (as mathematical analysis has so far been restricted to extremely simple and artificial dynamic systems). Nevertheless, there is a rich body of theoretical work on which to draw, both in evolutionary biology and in the study of naturally-occurring complex dynamic systems.

Dynamic systems, such as the weather, can be rule-governed (deterministic), complex in the sense of unpredictable (chaotic), but not adaptive. Others, such as the immune system, are adaptive and complex. Complexity in dynamic systems can arise through constraints on variation (self-organisation) as much as through (competing) selection pressures. Constraints on language variation will be a consequence of the capacities and limitations of the language users in whom the linguistic system is embedded - just as the constraints of chemistry places limitations on variation in biological organisms. As language users are also subject to evolution, and pressures for greater learnability, expressivity and interpretability are often in conflict, so (co-evolutionary) interactions are likely, contributing both to the complexity of the phenomena and their study (Laland and Brown, 2011 : Ch7). The use of evolutionary terminology in (diachronic) linguistics is not new (see e.g. McMahon, 1994 : Ch12) but, advances in the understanding of dynamic systems and the availability of computational simulation techniques now make it pos­sible to move beyond the use of evolution, primarily as metaphor, and study language directly from an evolutionary perspective.

The lectures will describe a computational simulation model which characterises languages as complex adaptive systems selected for learnability, interpretability and/or expressivity within a population of language using agents (Briscoe, 2000). This characterisation is a direct consequence of shifting the study of language from synchronic idiolects to evolving populations of classical generative agents learning, producing and interpreting language in a shared arena. As each new generation of learners acquires a grammar from the utterances in the arena of use to which they are exposed, they, in effect, select from this arena the more learnable variants (Hurford, 2012). Or, to turn this on its head, variant language constructions compete for learners, and the more learnable ones tended to be selected more often. However, learnability is counter-balanced by expressivity (e.g. economy of production), interpretability (e.g. working memory limitations and parsability), as well as « extraneous » factors such as population movement and social power or prestige which undoubtedly play a role, leading to competing linguistic selection pressures (competing motivations in diachronic linguistics) and thus a correspondingly dynamic and complex adaptive landscape.



Each presentation will last about 40 minutes leaving 20 minutes for questions and general discussion. I’ll suggest some further reading on each topic after each session and make my slides available to all participants after the sessions.

  1. Introduction — The Model
    1. Generative Linguistics
    2. Universal Darwinism
    3. Linguistic Units of Selection
    4. Linguistic Universals / Tendencies
    5. Gene-Language Co-evolution
    6. Language Agents
  2. First Language Acquisition Play
    1. Desiderata for a Model of FLA
    2. Generalized Categorial Grammar
    3. (Evolutionary) Bayesian Learning
    4. Iterated Learning Models
    5. Priors and fixed points ?
  3. Syntactic Change and Typology Play
    1. (Statistical) Universals
    2. Working Memory and Parsability
    3. S-curves and Power Laws
    4. Populations of Language (Learning) Agents
    5. Uniform Information Density
    6. Language Contact and Social Factors
  4. Co-evolution of Language and Genes Play
    1. Genetic Assimilation / Baldwin Effect
    2. Populations of Evolving Language Agents
    3. Learning Bias vs. Linguistic Bias ?
    4. Learning Costs and (De)Correlation
    5. Timescales in Language Evolution
  5. Grammar vs. Inference Play
    1. Relative Clause Formation Strategies
    2. Ambiguity and Dependency Length
    3. Trade-offs between (en/de)coding and inference
    4. Ambiguity and Prosody
    5. Relative Clause Usage
  6. Second Language Acquisition
    1. Differences FLA and SLA ?
    2. Morphology vs. Syntax in Diachrony
    3. Perceptual Factors in F/SLA
    4. Directionality in Language Change ?
    5. (Un)Folding of Languages
  7. More Putative Linguistic Universals
    1. Universal Grammar vs. Selection Pressures
    2. Nested, Cross-serial or Intersecting Dependencies
    3. Head (Dis)Harmony and Dependency Length
    4. The Final-over-Final Constraint
    5. Convergent Evolution – Functional vs. Arbitrary ?
  8. Conclusions, Issues, and Future Work
    1. Comparison with Other Approaches
    2. Methodology for Evolutionary Linguistics
    3. Desiderata for Simulation Models
    4. Maths vs. Computation
    5. Sociolinguistic Networks of Interaction


Background Reading
Briscoe, E.J. (2000). Evolutionary Perspectives on Diachronic Syntax. In Pintzuk S. et al (eds.), Diachronic Syntax : Models and Mechanisms, Oxford University Press -
Croft, W. (2000). Explaining Language Change, Longman
Deutscher, G. (2005). The Unfolding of Language, Heinemann
Hawkins, J.A. (1994). A Performance Theory of Order and Constituency, Cambridge University Press
Hurford, J. (2012). The Origins of Grammar, Oxford University Press
Kirby, S. (1997). Function, Selection and Innateness, Oxford University Press
Laland, K. and Brown, G. (2011). Sense and Nonsense : Evolutionary Perspectives on Human Behaviour, Oxford University Press
McMahon, A. (1994). Understanding Language Change, Cambridge University Press
Niyogi, P. (2006). The Computational Nature of Language Learning and Evolution, MIT Press
Steels, L. (2000). Language is a Complex Adaptive System, 6th Int. Conf. on Parallel Problem Solving, Springer


haut de page