Аннотация:Variation in empirical data has been a perseverant problem for theoretical linguistics, especially syntax. Data inconsistencies among authors allegedly analyzing the same phenomenon are ubiquitous in the syntactic literature (e.g., literature on focus-raising in Hungarian; É. Kiss 1987 vs. Lipták 1998), and partly result from the highly informal methodology of data collection. However, even if adequate controls are used to exclude potential biases, variation might remain. The general practice in syntactic research has been to ignore these „microvariations”-mainly in the lack of any systematic empirical method to detect them. The present paper shows that this practice leads to serious theoretical problems and proposes a new empirical method, cluster analysis, to discover, explore and systematize these variations. It also illustrates how this richer empirical basis gives rise to a more fine-grained theoretical analysis.
