3rd Conference Abstracts |
Language Evolution and Computation Research Unit
Department of Theoretical and Applied Linguistics
University of Edinburgh
kenny@ling.ed.ac.uk
The notion of the innateness of some part of human language competence is central to the most influential linguistic theories of modern times (Chomsky, 1965, 1980, 1981, 1987). Pinker & Bloom (1990) outline what could be considered to be an orthodox explanation for the origins of this innate, language specific mental organ. They argue that the human language faculty has been shaped by natural selection, assisted by positive interactions between natural selection and learning such as the Baldwin effect (Baldwin, 1896; Hinton & Nowlan, 1987).
Computational simulations of the emergence of communication systems, including language-like syntactically structured communication systems, fall into three main groups: those which suggest natural selection alone is capable of developing and refining innate communication systems (eg Werner & Dyer, 1991; Oliphant, 1996; MacLennan & Burghardt, 1994; Ackley & Littman, 1994; Levin, 1995; Cangelosi & Parisi, 1996; Bullock, 1997; Werner & Todd, 1997; de Bourcier & Wheeler, 1997; Di Paolo, 1997; Noble, 1998), those which suggest that repeated learning interactions alone are capable of developing and refining entirely learned communication systems (eg Oliphant, in press; Kirby, in press a, in press b; Hurford, in press; Batali, 1998, in press; Hutchins & Hazelhurst, 1995; Steels & Vogt, 1997), and those which suggest that natural selection and learning interact to develop communication systems which are part innate and part learned (eg Batali, 1994; Briscoe, 1997; Kirby & Hurford, 1997).
This paper explores the interactions between natural selection and learning in the evolution of simple communication, or signalling, systems (Lewis, 1969). Such systems consist of a set of meaning-form pairs, where both the meanings to be communicated and the communicative forms used to communicate those meanings are unstructured. Optimal communication among a population of individuals using such a communication system requires that the entire population use mutually intelligible communication systems and that these communication systems make an unambiguous mapping between meanings and forms (ie the population's communication systems must be free of synonymy and homonymy).
A simulated population of agents using genetically encoded communication systems was evolved, with breeding based on ability to communicate with other members of the population. No learning took place in this simulation. As suggested by similar computational simulations (eg Werner & Dyer, 1991; Oliphant, 1996; MacLennan & Burghardt, 1994; Ackley & Littman, 1994; Levin, 1995; Cangelosi & Parisi, 1996; Bullock, 1997; Werner & Todd, 1997; de Bourcier & Wheeler, 1997; Di Paolo, 1997; Noble, 1998), an innate, optimal communication system rapidly emerged in the simulated population.
Cultural transmission was then added to this model. Mature individuals were selected for breeding according to communicative success, as before, but their offspring learn a communication system based on the communicative behaviour of the mature population. An individual's genes now function as a starting point for, and constraint on, learning, rather than fully determining an individual's communication system. Previous work on the interaction between natural selection and learning in the evolution of communication (eg Batali, 1994; Briscoe, 1997; Kirby & Hurford, 1997) suggests that, under these circumstances, communication systems which are part innate and part learned should emerge.
This was found not to be the case. Such a simulated population was incapable of developing an optimal communication system from random behaviour. Furthermore, such a population was incapable of maintaining an optimal communication system over time, even if this optimal communication system was both encoded in the genes of the population and learnable by observation.
The combination of selection for communicative success and learning was incapable of developing an optimal communication system due to an overly-plastic phenotype. Learning overrides the genetically encoded communication system of individuals in the population. This has two effects. Firstly, natural selection is effectively disabled - due to learning in the phenotype, there is no selectional pressure for genes encoding optimal communication systems. This phenomenon is known as shielding (Ackley & Littman, 1991). Secondly, suboptimal communication systems, which form the vast majority of the communication systems used by the initial random population, are preserved - learning alone is incapable of developing optimal communication systems.
The combination of selection for communicative success and learning was also incapable of preserving optimal communication systems, even if those communication systems were encoded in the genomes of all members of the initial population. This was due to shielding of genetic information and cultural transmission. The absence of selectional pressure on the population's genes allows those genes to drift through genetic space due to random mutations. Eventually, one individual will inherit genes which are so bad that those genes do influence the learning process, preventing that individual from acquiring the optimal communication system of the population. While this individual will be weeded out by natural selection, its communicative behaviour will be observed by some language learners. These language learners run the risk of failing to acquire the optimal communication system, due to the noise introduced by the suboptimal communicator. If these individuals learn a suboptimal communication system they will also be weeded out by natural selection, but not before their communicative behaviour influences the communicative behaviour of yet more learners. Cultural transmission allows suboptimal communication systems to spread through the population like a virus, until the whole population communicates using a suboptimal system.
These simulations suggest two conclusions on the relative importance of natural selection and cultural transmission in the evolution of the human language faculty. Firstly, extreme plasticity in the phenotype of a communicative agent may disable natural selection, leading to behaviour which is determined by the agent's learning apparatus, rather than a combination of natural selection and learning. Secondly, if optimal learned communication systems are to emerge in a population, the agents in that population must be capable of selective acquisition or production of communicative forms. This linguistic selection at the individual level drives linguistic evolution at the population level, pushing the communication system of the population towards optimal systems of the type seen in Oliphant (in press), Kirby (in press a, in press b), Hurford (in press), Batali (1998, in press), Hutchins & Hazelhurst (1995) and Steels & Vogt (1997).
Ackley, D. and M. Littman (1991). Interactions between learning and evolution. In C. Langton, C. Taylor, J. Farmer and S. Rasmussen (eds.), Artificial Life 2, pages 487-509. Redwood City, CA: Addison-Wesley
Ackley, D. and M. Littman (1994). Altruism in the evolution of communication. In R. Brooks and P. Maes (eds.), Artificial Life 4: Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 40-48. Redwood City, CA: Addison-Wesley.
Baldwin, J.M. (1896). A new factor in evolution. American Naturalist 30, 441-451.
Batali, J. (1994). Innate biases and critical periods: Combining evolution and learning in the acquisition of syntax. In R. Brooks and P. Maes (eds.), Artificial Life 4: Proceedings of the Fourth International Workshop on the Synthesis and Simulation of Living Systems, pages 160-171. Redwood City, CA: Addison-Wesley.
Batali, J. (1998). Computational simulations of the emergence of grammar. In J.R. Hurford, M. Studdert-Kennedy and C. Knight (eds.), Approaches to the Evolution of Language: Social and Cognitive Bases. Cambridge: Cambridge University Press.
Batali, J. (in press). The negotiation and acquisition of recursive grammars as a result of competition among exemplars. In E. Briscoe (ed.), Linguistic Evolution through Language Acquisition: Formal and Computational Models. Cambridge: Cambridge University Press.
Briscoe, E. (1997). Language acquisition: the bioprogram hypothesis and the Baldwin Effect. MS, Computer Laboratory, University of Cambridge.
Bullock, S. (1997). An exploration of signalling behaviour by both analytic and simulation means for both discrete and continuous models. In P. Husbands and I. Harvey (eds.), Fourth European Conference on Artificial Life, pages 454-463. Cambridge, MA: MIT Press.
Cangelosi, A. and D. Parisi (1996). The emergence of a `language' in an evolving population of neural networks. Technical Report NSAL-96-004, Institute of Psychology, National Research Council, Rome.
Chomsky, N. (1965). Aspects of the Theory of Syntax. Cambridge, MA: MIT Press.
Chomsky, N. (1980). Rules and Representations. London: Basil Blackwell.
Chomsky, N. (1981). Government and Binding. Dordrecht: Foris.
Chomsky, N. (1987). Knowledge of Language: Its Nature, Origin and Use. Dordrecht: Foris.
de Bourcier, P. and M. Wheeler (1997). The truth is out there: The evolution of reliability in aggressive communication systems. In P. Husbands and I. Harvey (eds.), Fourth European Conference on Artificial Life, pages 444-453. Cambridge, MA: MIT Press.
Di Paolo, E. (1997). An investigation into the evolution of communication. Adaptive Behaviour 6, 285-324.
Hinton, G. and Nowlan, S. (1987). How learning can guide evolution. Complex Systems 1, 495-502.
Hurford, J. (in press). Expression/Induction Models of Language Evolution: Dimensions and Issues. In E. Briscoe (ed.), Linguistic Evolution through Language Acquisition: Formal and Computational Models. Cambridge: Cambridge University Press.
Hutchins, E. and B. Hazelhurst (1995). How to invent a lexicon: The development of shared symbols in interaction. In N. Gilbert and R. Conte (eds.), Artificial Societies: the computer simulation of social life. London: UCL Press.
Kirby, S. (in press a). Syntax without natural selection: How compositionality emerges from vocabulary in a population of learners. In C. Knight, M. Studdert-Kennedy and J. Hurford (eds.), The Evolutionary Emergence of Language. Cambridge: Cambridge University Press.
Kirby, S. (in press b). Learning, bottlenecks and the evolution of recursive syntax. To appear in E. Briscoe (ed.), Linguistic Evolution through Language Acquisition: Formal and Computational Models. Cambridge: Cambridge University Press.
Kirby, S. and J. Hurford (1997). Learning, culture and evolution in the origin of linguistic constraints. In P. Husbands and I. Harvey (eds.), Fourth European Conference on Artificial Life, pages 493-502. Cambridge, MA: MIT Press.
Levin, M. (1995). The evolution of understanding: a genetic algorithm model of the evolution of communication. BioSystems 36, 167-178.
Lewis, D. (1969). Convention: A Philosophical Study. Cambridge, MA: Harvard University Press.
MacLennan, B. and G. Burghardt (1994). Synthetic ethology and the evolution of cooperative communication. Adaptive Behaviour 2, 161-187.
Noble, J. (1998). Evolved signals: Expensive hype vs. conspiratorial whispers. In C. Adami, R. Belew, H. Kitano and C. Taylor (eds.), Artificial Life 6: Proceedings of the Sixth International Conference on Artificial Life. Cambridge, MA: MIT Press.
Oliphant, M. (1996). The dilemma of saussurean communication. BioSystems 37, 31-38.
Oliphant, M. (in press). The learning barrier: Moving from innate to learned systems of communication. To appear in Adaptive Behaviour.
Pinker, S. and Bloom, P. (1990). Natural language and natural selection. Behavioral and Brain Sciences 13, 707-784.
Steels, L. and P. Vogt (1997). Grounding adaptive language games in robotic agents. In P. Husbands and I. Harvey (eds.), Fourth European Conference on Artificial Life, pages 474-482. Cambridge, MA: MIT Press.
Werner, G. and M. Dyer (1991). Evolution of communication in artificial organisms. In C. Langton, C. Taylor, J. Farmer and S. Rasmussen (eds.), Artificial Life2, pages 659-687. Redwood City, CA: Addison-Wesley.
Werner, G. and P. Todd (1997). Too many love songs: Sexual selection and the evolution of communication. In P. Husbands and I. Harvey (eds.), Fourth European Conference on Artificial Life, pages 434-443. Cambridge, MA: MIT Press.
Conference site: http://www.infres.enst.fr/confs/evolang/