EN
This contribution discusses three ways of operationalising the notion of frequency as it relates to how often an item occurs in a corpus: the proportional frequency of forms (i.e. the percentage of instances in which one or another variant is found) and two ways of looking at absolute frequency. Working with data from unmotivated morphological variation in Czech case forms, we show that different types of data contribute to some extent to the way variation is perceived and implemented by native speakers, but suggest that proportional frequency seems most salient for speakers in forming their impressions and shaping their behaviour.