The paper consists of theoretical and research part. The theoretical part is supported by methodological-pedagogical and psychological literature, and presents the basic concepts and considerations in the field of evaluation. The research conducted a qualitative and quantitative critical analysis of publishing examinations of knowledge of the Croatian language for fifth, sixth, seventh and eighth grade written on the basis of the Curriculum for primary school in 2006, by the publisher Profil Klett. The emphasis is on checking the quality of examination tasks with the aim of establishing a methodology for the procedures for determining the validity of measuring instruments in the examination of language competence. Given that evaluation is both the starting and ending point of the educational process, it is important to follow certain elements, methods and criteria in order to approach the measurement of student achievement as objectively as possible. Exam tasks should reveal how students have mastered certain educational outcomes and what cognitive strategies they have used. Bloom's taxonomy is the most common taxonomy for defining goals, outcomes and creating exam tasks, and served as a starting point for this research. The initial hypotheses confirmed the inconsistency of the researched tasks with the prescribed educational achievements, their inadequacy of students' age, saturation of tasks with ambiguities and unnecessary contents, representation of open type of tasks and their spelling and grammatical correctness. The possibility of further research opens up, given that new implementation curricula have been developed in which the outcomes are set somewhat differently than in the 2006 Curriculum.