The study was carried out from 2018 to 2020 with the challenge - how to assess the level of computational thinking. The research design is mixed since the disclosure of mutual influence of the components of the chain 'learning programming -- computational thinking -- evaluating computational thinking' requires the use of both qualitative and quantitative research methods. The conceptualisation of the 'computational thinking' idea is based on the premise of the impact of abstract thinking and computers on human thinking evolution. The structural interpretation of 'computational thinking', consisting of nine components, reflects the presence of a semantic link between teaching programming and the development of abstract computational thinking. Four levels (phenomenological, analytic-synthetic, set-prognostic and axiomatic) of computational thinking have been identified for each of these nine components. The study involved 102 elementary school students who are learning programming in Scratch. The guiding questions and problems we have developed for elementary school students are designed following the characteristics of the four levels of computational thinking. The results of the study showed that the ratio of 'structural components' to different levels of computational thinking, with the corresponding characteristics, allows one to determine the degree of its development or its individual components.