Market microstructure models in the genre of Easley et al. are studied in order to examine the extent to which empirically observed autocorrelation in order-flow data is the consequence of “true” state dependence rather than “spurious” state dependence due to heterogeneity. Models with pure heterogeneity are contrasted with models containing both heterogeneity and state dependence. The models were tested on order-flow data from the automated Limit Order Book on the Helsinki Stock Exchange. Models including both heterogeneity and state dependence were found to give the best fit and it also became clear that the state dependence cannot be compensated for by increasing the heterogeneity of the information structure. The study also adds credibility to this modelling strategy and throws light on both the information structure and the amount of informed trading across firms.