Colloquium: Testing Language Formation Theories: Computer experiments as linguistic time machines

April 30, 3:30-5pm, Cobb 201
Teresa Satterfield, University of Michigan

Various hypotheses have been offered for the emergence of creole languages and their native-speakers in historical language contact settings, such as a plantation scenario. Accounts such as the Language Bioprogram Hypothesis (Bickerton 1984, and others) argue that these types of creoles arise within one generation when children create a novel L1 due to exposure to their immigrant parents pidgin input. A contrasting view argues that creole grammar is manifested by adults whose unsuccessful attempts at L2 acquisition converge to a new code over the span of several generations (Chaudenson 1992, 1995; Arends 1995, Mufwene 1996, among many others). A long-standing challenge for creole studies has been how to reliably test the claims put forth by these and other proposals, since it is difficult to provide a complete and empirically-grounded account of the origin and evolution of historical creole languages, owing to imperfect records particularly of the substrate languages involved, not to mention the extinction of intermediate forms. Along with this predicament is the added complexity of placing the study of creole formation squarely within the larger disciplinary contexts of L1 and/or L2 acquisition theories. Since the dynamics generated in the interaction between the numerous socio-cultural and cognitive variables is too complex to be solved analytically, the properties are examined within computer models, which provide methods for systematically analyzing the concurrent formation of social patterns and linguistic structures as they develop within that history. First, it is necessary to determine the viability of designing and implementing computer-generated an artificial society that effectively replicates the demographic and socio-communicative features associated with a well-known creole language scenario. For this goal Sranan Tongo is adopted as a case study (Arends 1995, Bruyn 1995, Van den Berg 2000, Winford 2003, among others). Secondly, this experimental environment is invoked to test two prominent theories of classic Atlantic plantation creole formation. Building on Satterfield (1999, 2001, 2005), this talk presents a series of theoretically-based models and compares their behaviors. Artificial agents (computer-simulated robots) are utilized with diverse social and linguistic repertoires to carry out the tasks within the artificial society. A central objective here is to determine, based on the hypotheses outlined, whether the interaction between multiple agents will indirectly result in the emergence of creolized linguistic structures.

While the current study ultimately takes no position on whether one single theory can be viewed as the final truth with respect to creole genesis, it does signal theoretical constraints which may make certain positions more or less plausible, and it offers clear indications for the optimal computational solution. In terms of the models' outcomes, bilingual competence in older children emerges as a key element in creole grammar formation.