An experimental study is reported of the retrieval by mock eyewitnesses of a scripted video crime event, comparing free recall against Change Temporal Order (CTO) recall, in which event retrieval is prompted in reverse order. Contrary to proponents of the technique who suggest that CTO allows greater discovery of script incidental information and increases the amount of information recalled, CTO was found to impair retrieval, leading to fewer script consistent events, reduced recall of correct information, increased confabulations, and lowered accuracy proportional to items retrieved. The disruptive effects of CTO are interpreted as providing further evidence for the role of temporal clustering highlighted in the CMR model of memory. Impairment induced by the CTO technique continued to influence retrieval negatively even during a secondary free recall phase. We suggest that CTO prevents the blocking of confabulations, and that these confabulations may subsequently contribute to forgetting by population dilution.