During the second half of the past decade, psycholinguistic models of speech production have been modified to include a “prearticulatory editing” component, the supposed function of which is to verify the linguistic integrity of impending phoneme strings destined for articulation. Empirical support for psycholinguistic editing comes primarily from studies by the present authors on laboratory-induced verbal slips. All of these studies have depended upon two crucial claims: that laboratory slips are errors of output, and that error-rate differentials are evidence of editing. Recently, however, it has been recognized that neither of these assumptions has been convincingly demonstrated, and consequently the editing models instigated by the earlier studies are being reassessed. The present study tested these two assumptions directly. The output-error assumption was supported by finding larger Galvanic Skin Resistance (GSR) responses for verbal slips than for correct vocalizations. The editing assumption was supported by finding that presumably edited vocal responses (identified by GSR) require more processing time (i.e., vocal response latency) than unedited responses. The reasoning from the empirical observations to conclusions about these assumptions is discussed in detail. Secondarily, a hypothesized social acceptability editing criterion was supported by the observation that neutral verbal slips outnumber taboo counterparts.