Researching complex dynamic systems: 'Retrodictive qualitative modelling' in the language classroom

While approaching second language acquisition from a complex dynamic systems perspective makes a lot of intuitive sense, it is difficult for a number of reasons to operationalise such a dynamic approach in research terms. For example, the most common research paradigms in the social sciences tend to examine variables in relative isolation rather than as part of a system or network, and most established quantitative data analytical procedures (e.g. correlation analysis or structural equation modelling) are based on linear rather than nonlinear relationships. In this paper I will first summarise some of the main challenges of dynamic systems research in general and then present a concrete research template that can be applied to investigate instructed second language acquisition. This approach involves a special type of qualitative system modelling - retrodictive qualitative modelling' - that reverses the usual research direction by starting at the end - the system outcomes - and then tracing back to see why certain components of the system ended up with one outcome option and not another. By way of illustration I will provide examples from two classroom-oriented research projects in which the language classroom was taken to be the dynamic system, and the system outcome options were the various learner prototypes (e.g. motivated, laid back, passive) observed in the classroom.