Please join us for a meeting of the Language Variation & Change workshop, this Friday, April 26 at 3:30pm, in Rosenwald 301.

Modeling dynamic processes of language emergence in creole genesis
Marlyse Baptista, Jinho Baik, Ken Kollman, and Alton Worthington
University of Michigan

This paper features an agent-based model of language creation and acquisition that may offer insights into dynamic processes, such as transfer, feature recombinations (Mufwene, 2001, 2008; Aboh, 2015) convergence and language shift, believed to be in part responsible for the emergence of creole languages. Our primary purpose is to provide a conceptual framework that allows us to examine hypothetical scenarios of creole genesis. Indeed, creolists agree that a creole’s lexicon originates from its superstrate but no consensus has emerged so far with respect to the source of creoles’ grammatical features. Competing theories of creole genesis are based on assumptions that they involve processes of second language acquisition (Chaudenson, 2001; Mufwene, 2001), or interlanguages (Plag, 2008), or relexification (Lefebvre, 1988) or language creation (Baker, 1996) or simply feature recombinations (Mufwene, 2001, 2008; Aboh, 2015). In response to these long-standing controversies, we attempt to simulate the kind of linguistic interactions that emerge in a multilingual setting when slaves and colonizers first come in contact by designing an agent-based model that is informed by data from 18th-century Haitian Creole (HC) diachronic texts. These texts which are believed to have been written by different scribes, including literate African slaves and native French speakers, reflect much variation. This variation results in part from the mixing of forms from non-standard varieties of French and possibly Fongbe, a Kwa language assumed to have contributed to the genesis of HC.

In our examination of these diachronic texts, we focused on three functional features of HC: the definite determiner la, negation pa, and plural marking yo. The rationale for selecting these specific features is that they reflect much variation and different degrees of stability, attesting to the various degrees of proficiency that the original scribes had in HC. In these texts, the determiner is highly unstable, occurring either in a pre-nominal (la+N) as in French, postnominal position (N+la) as in HC or preceding and following the noun (la+N+la). In contrast, both negation and plural marking are highly stable, appearing consistently pre-verbally for negation (pa+V) and consistently post-nominally for the plural unbound marker (N+yo), as it does in HC. Such unstable variation suggests that the original scribes, some of whom likely to have been African slaves were subject to pressures to shift or not to shift from their L1 to the L2 patterns.

History plays a crucial role in our agent-based model, in the sense that timing and sequence of events, especially importation of slaves in large numbers, affect the dynamics of language acquisition and creation (Singler, 1996). In a variety of simulations, we experimented with aspects of linguistic interactions between the Fongbe and French populations, altered the dynamics of population change, the pressures towards one language (French), and the degree to which agents in the model were willing or pressured to shift to a new set of linguistic features.

The results from the simulations suggest that unstable features emerge due to ambiguous and contradictory cues from the source languages whereas stable features emerge when no such contradiction is present. This supports aspects of Mufwene (2001, 2008), Aboh (2015), Lefebvre (1988), and Baker (1996).