Computer models for building usage simulation are needed that produce detailed data about activities of members of an organization for accurate building evaluation. Intermediate activities that interrupt the planned activities play an important role but are often ignored in existing simulations. A model is proposed to predict the occurrence and the frequency of intermediate activities during an office working day. Two prediction methods are presented, namely probabilistic and S-curve. The applicability of each method depends on the characteristics of the intermediate activity. Through an experiment data were collected about intermediate activity behavior in an office. From these data conclusions are drawn on the validity of the prediction methods. Finally for each intermediate activity the formulas are presented including the parameter values that are analyzed from the experiment. The formula can be used in office evaluations that require detailed input data on occupant activities.