Measurement properties of subjective magnitude estimates of word frequency

The subjective magnitude estimation (SME) procedure was used to obtain estimates of relative word frequency from two adult groups (15 lexicographers, 13 other adults) for 60 words ranging widely in objective frequency. Lexicographers rendered more reliable estimates, and their averaged data correlated more highly with objective log frequency than those of the second group. The objective frequncy of the first stimulus considered in the SME task is not related to a subject's overall accuracy in predicting objective frequency, but accuracy is related to the subject's tendency to perceive frequency ratios as relatively large. Subjective estimates measure something slightly different from what is indexed by currently available objective counts, and may be more valid measures of true word probability.