<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-1103">
  <Title>Self Organisation in Vowel Systems through Imitation</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusions and Discussion
</SectionTitle>
    <Paragraph position="0"> The first conclusion that can be drawn from the work presented above is that stable sound systems do emerge in a population of artificial agents that play imitation games. Moreover, these systems have discrete clusters in a continuous acoustic space that could be described by (discrete) distinctive features, even though there was no predetermined partition of  the left after 1000 imitation games.</Paragraph>
    <Paragraph position="1"> the acoustic space. In addition, it can be concluded that the shapes of these emergent systems show remarkable similarities to the shapes of the most frequent vowel systems found in human languages.</Paragraph>
    <Paragraph position="2"> It remains to be seen to what extent these results are applicable to human language. It must be admitted that the language capabilities of the simulated agents are a gross simplification of the language capabilities of humans. However, the agents are entirely biologically plausible. This means that they can do nothing that humans could not do in principle. Also, they provide a possible mechanism by which functional constraints on vowel system that were first researched with computers by Liljencrants and Lindblom (Liljencrants and Lindblom, 1972) can emerge from interacting language users.</Paragraph>
    <Paragraph position="3"> The system described here provides a model for predicting certain universals of vowel systems. It does not have to postulate innate distinctive features or innate mechanisms other than the fact that agents communicate with a limited set of sounds. Also the system shows individual variation and language change that do not decrease the agents' ability to analyse each other's sounds. A remarkable property of the simulations that have been presented is that both the learning of speech sounds as well as sound change can be generated by the same mechanism.</Paragraph>
    <Paragraph position="4"> The author thinks that these results justify considering phonological processes in language as self-organising processes. By taking this point of view it also becomes possible to bridge the gap between language as behaviour of individuals and language as a system by using computational models.</Paragraph>
  </Section>
class="xml-element"></Paper>