This study presents results from a one-year longitudinal study of reading comprehension skills in Czech primary school children. Using the two different types of reading comprehension test (specified later), we tested the theoretical model of skills predicting reading comprehension de- velopment in Czech children from Caravolas, Volín &Hulme (2005). 351 primary school children aged approx. 7 – 9 years participated in the study and were tested two times within one academic year. To assess reading comprehension two different tasks were used: 1. timed cloze test where children fill in missing words in short passages of a text and 2. reading aloud of a short text followed by series of questions proposed by a test administrator. For both reading comprehension tests we tested predictive power of a set of skills containing: nonverbal IQ, decoding (words and pseudowords reading), rapid naming of digits, vocabulary, listening comprehension and morpho- syntactic awareness. For both types of reading comprehension test, morpho-syntactic knowledge measured at the onset of the school year proved to be the strongest predictor of reading com- prehension at the end of the same academic year. For the timed cloze format test of reading comprehension also decoding seems to be accompanying predictor of the same importance. Generally the model for timed cloze test proved to have more predictive power than the model for answering questions test.