Error, Error Everywhere: A Look at Megastudies of Word Reading

Two primary methods have been used in studies of word reading: small-scale factorial studies and larger scale "megastudies" involving thousands of words. We conducted comparisons between the two, using the standard frequency X regularity interaction in word naming as test case. Whereas the effect replicates across small-scale studies, somewhat different results were observed using item means from 3 megastudies. Correlations between the megastudies are also relatively low. The considerable error variance in the megastudies limits their use in creating mini ("virtual") experiments. The megastudies yield small but more consistent results using regression analyses examining specific factors.