Published online by Cambridge University Press: 18 December 2008
This paper describes some cases of inter- and intra-informant variability which cannot be explained in terms of social and contextual factors, and which do not seem to require probabilistic models to account for them. Analysis of several Zinacanteco speech taxonomies suggests that what are called variant responses are often only incomplete responses, and that if the cognitive system is represented as a taxonomy, then it is a taxonomy which is made up of a number of partial taxonomies, each of which is produced by a different informant or group of informants or by the same informant on different occasions. A frequently used method of ‘controlled eliciting’ is shown to be inherently incapable of guaranteeing the elicitation of complete taxonomies in every interview. For this reason, it is urged that before constructing a model of a cognitive system it is first necessary to establish what is ‘real’ cognitive variation by recognizing and eliminating that variation which results from methodological indeterminacy. (Ethnoscience, cognition, variability, methodology, controlled eliciting, folk taxonomy, speech, Maya languages.)