Discourse represents connected language units and its fundamental feature is linguistic completeness (Barić et al., 1997). Narration is a form of discourse and narrative samples are often used for the analysis of child language. In this paper, the narration was studied on a microstructural level, i.e. syntactic complexity of children narrative discourse was analyzed. The goal of this study was to examine syntactic complexity by using four measures: mean length of communication unit, mean length of clause, clause density, and a number of different types of compound-complex sentences used in narration. The study was conducted on 58 participants (28 six-year-olds and 30 eight-year-olds). The participants were selected from the Croatian FrogStory Corpus (Trtanj et al., 2017), which is a part of the CHILDES, a global bank of child language (McWhinney, 2000). The narrative samples extracted from the corpus were analyzed in the CLAN software (McWhinney, 2000). The data obtained in the software were then examined by using the IBM SPSS 22 statistical analysis software. The findings of the study indicate that there are statistically significant differences between the two age-groups, particularly in the mean length of communication unit and mean length of clause. In contrast, there are no significant differences between six-year-olds and eight-year-olds with respect to the clause density and the number of different types of compound-complex sentences. Since there is still no consensus on the best approach to research syntactic complexity and it is still not known which are the most reliable measures of syntactic complexity, it is necessary to conduct research in various languages. Narration is a significant indicator of cognitive and language development; therefore, studies of this type allow professionals to observe the development of child language and consequently diagnose and treat various disorders in communication development as well as language and speech development in time.