Abstract
The article summarizes the theoretical foundations and results of a corpus-driven study of register variability in contemporary Czech. The descriptive framework is based on the methodology of multidimensional analysis, as previously applied to various other languages (see Biber 1995).
The starting point is a quantitative analysis of a custom-built genre-diversified corpus in which linguistic features have been identified that are likely to be related to functional and systematic variability on different linguistic levels. Statistical processing using factor analysis then yields a model which identifies (in the case of Czech) 8 dimensions of variation of the texts.
The greatest proportion of variance is explained by the first two dimensions, which can be described as dichotomies distinguishing between dynamic vs. static and spontaneous vs. prepared.