Within a Dynamic System Theory (DST) approach, it is assumed that language is in a constant flux, but that differences in the degree of variability can give insight into the developmental process. This longitudinal case study focuses on intra-individual variability in accuracy rates and complexity measures in Finnish learner language. The study illustrates the use of several useful DST methods and techniques such as min-max graphs and regression analyses to gauge whether different degrees of variability are meaningful, and Monte Carlo analyses to test for significance. Error rates were found to decrease rapidly in most cases except in four notoriously troublesome ones. Both word complexity and sentence complexity, and word complexity and NP complexity develop simultaneously and can be seen as connected growers, but NP complexity and sentence complexity alternate in developing and can be considered competitors. The study clearly shows that the interaction of different complexity measures change over time. Quite surprisingly, no meaningful relationship was found between accuracy and complexity measures over time.