Language Evolution on Stella Fructa: The Effects of Novel Variables on an Iterated Learning Model of Linguistic Evolution by Cultural Transmission
- Jon William Carr, University of Glasgow
- Old College Lorimer Room, University of Edinburgh
It is suggested that human language arose as a function of three adaptive processes: evolution
by natural selection, individual learning, and cultural evolution (Kirby and Hurford 2002). In this
paradigm, biological evolution is intrinsically difficult to explore due to the ephemeral nature of
language; however, the interface between learning and cultural evolution has recently been tested
in a variety of mathematical, computational, and experimental models. These models demonstrate
that systematic linguistic structure can arise in the transmission of language across multiple
language users. Kirby, Cornish, and Smith (2008) introduced an experimental method for studying
the cumulative effect on language of this cultural transmission. Their iterated learning model
represented the first experiment on human participants to suggest that the cultural transmission of
language leads cumulatively to the appearance of linguistic design without any explicit designer.
The present paper presents the results from a repeat of this experiment, which was conducted
with a number of novel variables. Therefore, not only is the experimental method verified, but the
resilience of the approach is tested too. The notable variables introduced in this version of the
experiment are (a) an auditory modality, (b) a modified meaning-space, and (c) a modified signalspace.
Despite these novel variables, learnability increases over the course of the experiment in a
way that is strikingly similar to the results observed by Kirby et al. (2008). The emergence of
compositional structure, however, is less forthcoming. This presents a problem: if the languages
evolve to become easier to learn, yet a strong degree of compositionality does not emerge, then
there must exist at least one other mechanism by which the languages optimize their successful
transmission. The results suggest that this mechanism might lie in a different type of adaptation of
the signal-space (i.e. the sounds and syllable structures employed by the languages). Over the
course of the experiment, the languages tended to become more focused on a smaller set of
syllable patterns, and this adaptation appears to make them easier to learn. Furthermore, as the
signal-space becomes increasingly more focused, the probability of a chance alignment between
signal and meaning becomes increasingly more likely to occur. It is therefore hypothesized that, in
allowing the languages to evolve for perhaps five or ten more generations, a degree of
compositional structure comparable with Kirby et al. (2008) might eventually emerge.
References:
Kirby S, Hurford JR (2002) The emergence of linguistic structure: An overview of the iterated
learning model. In Cangelosi A, Parisi D (eds) Simulating the Evolution of Language (London:
Springer Verlag): 121—148
Kirby S, Cornish H, Smith K (2008) Cumulative cultural evolution in the laboratory: An
experimental approach to the origins of structure in human language. Proceedings of the National
Academy of Sciences 105: 10681—10686
by natural selection, individual learning, and cultural evolution (Kirby and Hurford 2002). In this
paradigm, biological evolution is intrinsically difficult to explore due to the ephemeral nature of
language; however, the interface between learning and cultural evolution has recently been tested
in a variety of mathematical, computational, and experimental models. These models demonstrate
that systematic linguistic structure can arise in the transmission of language across multiple
language users. Kirby, Cornish, and Smith (2008) introduced an experimental method for studying
the cumulative effect on language of this cultural transmission. Their iterated learning model
represented the first experiment on human participants to suggest that the cultural transmission of
language leads cumulatively to the appearance of linguistic design without any explicit designer.
The present paper presents the results from a repeat of this experiment, which was conducted
with a number of novel variables. Therefore, not only is the experimental method verified, but the
resilience of the approach is tested too. The notable variables introduced in this version of the
experiment are (a) an auditory modality, (b) a modified meaning-space, and (c) a modified signalspace.
Despite these novel variables, learnability increases over the course of the experiment in a
way that is strikingly similar to the results observed by Kirby et al. (2008). The emergence of
compositional structure, however, is less forthcoming. This presents a problem: if the languages
evolve to become easier to learn, yet a strong degree of compositionality does not emerge, then
there must exist at least one other mechanism by which the languages optimize their successful
transmission. The results suggest that this mechanism might lie in a different type of adaptation of
the signal-space (i.e. the sounds and syllable structures employed by the languages). Over the
course of the experiment, the languages tended to become more focused on a smaller set of
syllable patterns, and this adaptation appears to make them easier to learn. Furthermore, as the
signal-space becomes increasingly more focused, the probability of a chance alignment between
signal and meaning becomes increasingly more likely to occur. It is therefore hypothesized that, in
allowing the languages to evolve for perhaps five or ten more generations, a degree of
compositional structure comparable with Kirby et al. (2008) might eventually emerge.
References:
Kirby S, Hurford JR (2002) The emergence of linguistic structure: An overview of the iterated
learning model. In Cangelosi A, Parisi D (eds) Simulating the Evolution of Language (London:
Springer Verlag): 121—148
Kirby S, Cornish H, Smith K (2008) Cumulative cultural evolution in the laboratory: An
experimental approach to the origins of structure in human language. Proceedings of the National
Academy of Sciences 105: 10681—10686