Given the uneven distribution of linguistic items like words across the space of linguistic variables, binned designs are often present difficulty for neurolinguistic experiments.  We report results from MEG experiments exploiting continuous linguistic variables and regression analyses of MEG activation from ROIs in source space.  The results are robust across experiments, and the methods seem appropriate for MEG language experiments more broadly.