Abstract
The study reviews the fundamental assumptions of various readability measures and provides a case study demonstrating the properties of some readability measures when applied to Czech. Based on the analogy with the measures of production (MLU, IPSyn), it is argued that automatic readability measures provide a valid estimate of text reading difficulty.
Different ways of validity testing for the readability indices are proposed. The case study shows that different measures based on average sentence and word length are very similar, while measures taking into account the token-type ratio (repetition index) behave differently.
Readability indices for a series of textbooks show gradual increase in estimated difficulty with increasing grade level. The estimates for other texts also fit the expectations based on the text purpose and style.