In fMRI, voxel measurements are complex-valued at each time point. Historically, complex-valued fMRI time courses are converted to real-valued magnitude-only ones then statistically analyzed for brain activation (Bandettini et al. 1993). This process discards the phase or half of the numbers. Recently Rowe and Logan (2004) introduced a complex-valued magnitude activation model with constant phase. Subsequently, Rowe and Logan (2005) generalized the model to have no restrictions on the phase and derived the same regression coefficients and activation statistic as the usual magnitude-only data model. Published findings by Rowe (2005a) show improved magnitude detection power when the phase is in fact constant. Published findings by Menon (2002) and others indicate that the voxel phase time course should be constant in grey matter but may exhibit task related changes in voxels containing draining veins. The Rowe-Logan complex activation model was generalized by Rowe (2005b) to model and test for magnitude and/or phase activation in complex data.

In this talk I will summarize the previously mentioned models and a new homologous real-valued phase-only activation model. Hypotheses regarding task related magnitude and/or phase changes are evaluated on actual fMRI data using presented activation statistics. It will be shown that the Rowe-Logan complex constant phase model strongly biases against undesirable voxels with task related phase changes resulting in more focal activation.