Traditional analyses of reasoning and interactions often describe the building of knowledge by comparing general outcomes of group data (e.g., averages, frequencies and percentages) and as a result, often neglect the underlying individual processes. In this paper, we present a case study of two dyads of children (four to six years old) working on a problem-solving task and aim to capture the moment-to-moment changes in a time serial data set. Particularly, we present the changes in the types of interactions and scientific reasoning skills (SRS) during a single problem-solving task. The three descriptive techniques implemented in the data analysis are: (1) a time-series analysis to track the behaviours from moment to moment; (2) transition matrices to describe behaviour changes; and (3) a hierarchical agglomerative clustering (HAC) with a hybrid clustering technique used to detect dyadic patterns. Our results describe the intra-individual variability of the dyads, supporting the assumption of the non-ergodicity (i.e., neither homogeneous nor static) condition of psychological processes. We show that intra-individual trajectories in the children's behaviours are neither homogeneous nor stationary, but variable over time.