SISTA Colloquium Series
|Date||Monday, November 14, 2011|
|Details||Light refreshments in Gould-Simpson 9th floor atrium before talk.|
|Speaker||Andrew B. Wedel|
|Affiliation||University of Arizona|
Language As An Evolutionary System: Word Meaning Exerts Selection Pressure On Sound Systems
Languages can be understood as evolutionary systems, changing over time under the influence of both drift and selectional pressures, operating within and between levels of organization (e.g., Ohala 1989, Blevins 2004, Wedel 2006, Chater & Christiansen 2010). The two levels of organization I will talk about are the word and the sound. Every language has a limited set of individually meaningless sounds (for example, p, t, s, m, etc.) that can be used in combination to create distinct, meaningful words. Just as with the genetic code, a particulate system such as this allows the creation of a potentially unlimited set of larger-scale elements (e.g., genes or words) from a limited set of smaller-scale elements (e.g., nucleotides or sounds; Abler 1989). Likewise, an important factor that governs the formation of larger scale elements is the number of distinctive elements at the lower scale and the constraints on how they can be arranged.
The set of sounds that a language makes use of is not static, but changes over time. As an example, the original [p] sound in the common ancestor of Germanic and Romance languages shifted to [f] in Germanic, giving rise to regular correspondences like Father in English and Pater in Latin. Often, sound-change simply alters the quality of sounds without changing the number or arrangement of sounds in the system; an example would be change from trilled [r] to uvular [r] in German. Sometimes, however, change can result in the loss of a sound from the system, which then potentially reduces the number of distinctions at the word level. For example, in Western dialects of American English, the historical [ɔ] vowel found in thought and caught is increasingly pronounced [ɑ] as in lot, which creates new homophones such as caught ~ cot.
The nearly century-old ‘functional-load’ hypothesis (Jakobson 1931, Trubetzkoy 1939, Martinet 1955, Hockett 1965, Surendran & Niyogi 2006) suggests that sounds are proportionally less likely to be lost from a language the more ‘work’ they do at the word level, that is, the more words they distinguish from one another in communication. This is an intuitively attractive idea, but it has been difficult to test until now because (i) there has been no model for how distinctions at the level of individual words could influence the course of sound change in a whole language, and (ii) the requisite data and computational power to test the hypothesis have not been available. In this talk I will present the first statistical analysis of data from a wide range of languages showing that the functional load hypothesis appears to be correct: sounds are significantly less likely to be lost from a language if they distinguish many words at the level of communication. Using computational simulations, I will then illustrate an evolutionary model of a linking mechanism between the word and sound levels that produces this effect.