Collocations as a multi word expressions, i. e. multi-member lexical units, are an insufficiently researched area in the Croatian language, especially when it comes to collocational relations in the early language discourse. However, in recent years, more attention has been paid to collocations in Croatian literature and research has been conducted that has contributed to a better understanding of them. The aim of this study was to examine the acquisition of collocational relations in younger grades of primary school students and studentˈs sensitivity to collocations , i. e. to determine to what extent students understand and use collocations at the end of the first educational cycle. Collocation exercises belong to a subgroup of lexical-semantic exercises in Croatian language learning and teaching. Lexical exercises encourage the development of studentsˈ communication competence in general. In the early language discourse students acquire and learn the language in concrete everyday communication, so lexical exercises should be conducted as often as possible. Therefore, for the purposes of this thesis, research was conducted in two 4th grades in the Antun Augustinčić Primary School in Zaprešić. The participants were divided into a control and an experimental group. The research instruments were tests made specifically for the purposes of this research. Both groups of participants solved the same initial and final test. The initial test was solved by a total of 31 participants (N=31), and the final test was solved by a total of 25 participants (N=25). After the initial testing, collocation exercises were held for two 45-minute class periods with the experimental group, while in the control group the exercises were not conducted. The basic aim of the study was to examine whether there is a difference in the understanding and use of collocations between the control and the experimental group at the beginning and at the end of the study, after the collocational exercises were held with the experimental group. The results of the research showed that the examinees of the experimental group solved the final exam better than the examinees of the control group. However, this difference is not statistically significant. If gender variables and final Croatian language grades at the end of the last school year are taken into account, it is evident that girls were slightly more successful than boys and that participants with a higher final grade achieved significantly better results than participants with a lower final grade.