Three randomizations of a list of twenty-five words were recorded by a single talker on separate days and under identical conditions. One of the recordings was rerecorded three times and these dubbings were acoustically corrected, cut, and spliced into the three different orders. Two groups of twelve subjects each were tested. One group listened to the three independent recordings, the other to the three randomizations of the single reading. Statistically significant differences in listener performance were found among the three independent readings of the word list, but no differences were found among the randomizations of the single reading. These results indicate that repeated readings of a list of words do not necessarily give equivalent forms of a test of speech discrimination.