A critique of recent approaches to creole typology.

What is it about?

This paper takes as its starting point Muysken’s (1988a:300) view of a ‘creole’ language as “just a language.” With this statement, Muysken rejects the idea of a creole prototype. Over the past 20 years, that idea has seen several new proponents. We provide a brief overview, before turning our attention to the work of Bakker et al. (2011; et seq.). In this paper, we focus on Bakker et al.'s methodologies, providing detailed explanations of the statistical and phylogenetic computational procedures. We show that the data set used in the statistical procedure, a multiple regression analysis, fails to come anywhere near the minimum number of datapoints required for such an analysis to be meaningful. We show further that the apparent sophistication of the phylogenetic approach, which relies on a combination of computational packages to produce output in the form of reticulate networks, cannot remedy the flaws arising from questionable assumptions about linguistic features, pervasive errors in the data which are fed into the computation (including errors which appear to be systematic), the problematic treatment of gaps in the data, and the overinterpretation of output patterns. Contrary to Bakker et al.’s claims, Muysken’s (1988a) statement has not been disproven by these technologies.

Featured Image

Why is it important?

Bakker et al.'s work relies on statistical and phylogenetic computational procedures which are poorly understood by most linguists in the field of pidgin and creole studies. As a result, their methodologies have gone virtually unchallenged, although several critiques of their data – and by implication, of their conclusions – have been published. This chapter demystifies the methodologies and shows that the work illustrates the principle "garbage in - garbage out".