Title: Language evolution is shaped by the structure of the world: An iterated learning analysis
Abstract: Language evolution is shaped by the structure of the world: An iterated learning analysis Amy Perfors ([email protected]) School of Psychology, University of Adelaide Daniel Navarro ([email protected]) School of Psychology, University of Adelaide Abstract Human languages vary in many ways, but also show strik- ing cross-linguistic universals. Why do these universals ex- ist? Recent theoretical results demonstrate that Bayesian learn- ers transmitting language to each other through iterated learn- ing will converge on a distribution of languages that depends only on their prior biases about language and the quantity of data transmitted at each point; the structure of the world being communicated about plays no role (Griffiths & Kalish, 2005, 2007). We revisit these findings and show that when certain as- sumptions about the independence of languages and the world are abandoned, learners will converge to languages that depend on the structure of the world as well as their prior biases. These theoretical results are supported with a series of experiments showing that when human learners acquire language through iterated learning, the ultimate structure of those languages is shaped by the structure of the meanings to be communicated. Keywords: language evolution; iterated learning; Bayesian modeling; linguistic structure Figure 1: (a) Schematic illustration of the typical iterated learning paradigm, which assumes that learner n acquires language on the basis of the language data produced by learner n − 1. (b) A dif- ferent view of iterated learning recognizes that because individuals produce language in order to communicate about the world, the data available to learners includes meanings in the world as well as the linguistic data produced by the learner before them. Introduction Human languages have rich structure on many levels, from phonology to semantics to grammar. Where does this struc- ture come from? Most researchers agree that linguistic struc- ture is shaped by the structure of our minds – that our brains contain prior biases that favor the acquisition or retention of some linguistic systems over others. As such, debate gen- erally centers around the nature and origin of these biases. Some suggest that the human language faculty is genetically specified, with natural selection operating on genes for lan- guage (e.g., Pinker & Bloom, 1990; Nowak, Komarova, & Niyogi, 2001; Komarova & Nowak, 2001) or else selecting for other capabilities (e.g., Hauser, Chomsky, & Fitch, 2002). Others have suggested that humans easily learn language not because of a language-specific genetically encoded mecha- nism, but because language evolved to be learnable and use- able by human brains (e.g. Zuidema, 2002; Brighton, Smith, & Kirby, 2005; Christiansen & Chater, 2008). While these accounts disagree in many particulars, they agree that the structure of language arises from the structure of the brain. In this paper we argue that language evolution is shaped by the structure of the world in addition to pre-existing cognitive biases. Because language involves communicating about the world, the structure of that world (i.e., the things to be com- municated) can interact with people’s prior biases to shape the languages that develop. We offer theoretical and exper- imental support of this proposition. On the theoretical side, we take as our starting point recent work within the “iterated learning” framework (in which new learners receive their data from previous learners). Previous research has shown that when learners are individually Bayesian, an iterated learning chain converges in the limit to the prior distribution over all possible languages (Griffiths & Kalish, 2005, 2007). How- ever, the proof of this assumes a priori that a language carries no assumptions about the frequencies of events in the world. As we will show, when this assumption is relaxed, the iterated learning process converges to a distribution that depends on the distribution of meaningful events in the world as well as the prior biases of the learner. We experimentally test these theoretical results in a lab-based iterated learning experiment (as in, e.g., Kirby, Cornish, & Smith, 2008) and find that par- ticipants converge on different languages depending on the structure of the space of meanings they are shown. Iterated learning The iterated learning modeling (ILM) framework is widely used in language evolution research (e.g., Kirby & Hurford, 2002; Griffiths & Kalish, 2007; Kirby et al., 2008; Smith, 2009; Reali & Griffiths, 2009). It views the process of lan- guage evolution in terms of a chain of learners (or genera- tions), shown schematically in Figure 1(a). The first learner in the chain sees some linguistic data (e.g., utterances), forms a hypothesis about what sort of language would have gener- ated that data, and then produces their own data, which serves as input to the next learner in the chain. Over time, the lan- guages that emerge from this process become non-arbitrary: Griffiths and Kalish (2005, 2007) (henceforth, GK) demon-
Publication Year: 2011
Publication Date: 2011-01-01
Language: en
Type: article
Access and Citation
Cited By Count: 8
AI Researcher Chatbot
Get quick answers to your questions about the article from our AI researcher chatbot