Abstract
This paper investigates the contribution of author/idiolect vs. register/type-of-text - as the most salient factors influencing the final shape of a text - towards explaining the variation observed in Czech texts. Since it is almost impossible to explore the effect of these factors on authentic data, we used elicited letters collected in a fully crossed experimental design (representative sample of 200 authors x four elicitation scenarios serving as a proxy to register variation).
The variation encompassed by the elicited texts is analyzed through the lens of a general-purpose multi-dimensional model of Czech. Using triangulation via three established statistical methods and one devised for the purpose of this study, we find that register matters a great deal, explaining 1.5 times as much variation overall as idiolect.
This should be taken into account when designing research in sociolinguistics or variation studies in general.