Friedberg and Adamowicz 1 make several interesting observations in their review of recovery rates reported in White et al. 2 However, we disagree with them that a previous trial of an active behavioural intervention, Deale et al, which reported very similar recovery rates to White et al, 'used similar recovery criteria'. [1][2][3] The White et al 2 recovery criteria were: Chalder Fatigue Questionnaire (CFQ; Likert scoring) and Short Form (SF)-36 physical functioning (PF) scores in normal range (ie, ≤18 and ≥60, respectively); Oxford CFS criteria not met; and clinical global impression (CGI) score of 1 or 2. The Oxford criteria required CFQ (bimodal scoring) ≥6 and SF-36 PF ≤65.By contrast, Deale et al 3 used more stringent criteria for 'complete recovery': SF-36 PF >83, CFQ (bimodal) <4, were employed full-time, and Oxford criteria not met. Only two outcomes reported by Deale et al and White et al are directly comparable. In White et al, 40% obtained CGI scores of 1 or 2, significantly lower than 68% in Deale et al (Fisher's exact test, p<0.05). Fifty-two per cent had SF-36 PF scores ≥60, not much more than 48% in Deale et al, who had SF-36 PF scores >83, a much more stringent threshold.Friedberg and Adamowicz state the operationalised criteria for recovery in White et al were "informed by population data." However, when White et al said they changed their original protocol's threshold score of SF-36 ≥85 to a lower score "as that [original] threshold would mean that approximately half the general working age population would fall outside the normal range," referencing Bowling et al and a mean of 84, they made an important error. 2 Part of their mistake was to use the mean and SD when the scores are not normally distributed. 4