The overall agreement with in vivo ratings was 91% (n = 1598 items, Kappa .812, p < .001). Inter-rater agreement was substantial for both pre- and post-therapy assessments. All participants made a numerical improvement in naming treated items (Fig. 1). The change was statistically significant for 15 participants (Wilcoxon matched samples, one-tailed

test, p < .05), with S.C. in Sunitinib chemical structure the Tavistock study showing no significant change in naming treated items (further details in Hickin et al., 2002). A comparison between the mean pre-intervention score [43.5, standard deviation (SD) 18.12] and the mean post-intervention score (62, SD 22.85) for treated items reveals the large effect size for the group (Cohen’s d of .897). The findings for untreated items are shown in Fig. 2. The change shown is proportional as there were different numbers of unseen items in the two projects (Tavistock study 100; Buckinghamshire study 50). A comparison between the mean pre-intervention raw score (33.84, SD 17.61) and the mean post-intervention score (36.31, SD 19.17) for untreated items reveals an effect

size (Cohen’s d) of .134. While this should be interpreted with care due to the different number of items in the different studies, it is clear the effect size for the group is minimal. Table 4 shows that there was stability in the control tasks across occasions (raw scores for each participant are provided in Appendix 4). A One way Repeated Measures Analysis of Variance (ANOVA) demonstrated no significant difference TSA HDAC cell line between the mean scores at different time points on either task [short term memory (STM)

pointing span, F(2, 22) = .12, p = .88; Sentence comprehension F(2, 22) = .94, p = .40]. The following section relates the categories to which we allocated participants on the basis of background language testing to the change in picture naming with therapy. Table 5 provides mean change on treated items for the four sub-groups with relatively Pregnenolone stronger and poorer semantic and phonological output processing (naming of the whole 200 items is provided in Appendix 5). The sub-groups change on treated items ranges from 14 to 22%, with those having relatively better semantic processing and better phonological output processing making slightly more change on average, although none of the sub-groups stands out. This was confirmed by a 2 × 2 between subjects ANOVA [F(1, 12) < 1, n.s. for effect of semantic impairment, effect of phonological impairment and interaction]. Fig. 3 shows mean change on untreated items for the four sub-groups. The three participants (H.M., T.E., P.P.) with relatively less of a semantic difficulty and more of a phonological output deficit (stage 3) show a pattern of generalisation to untreated items. A 2 × 2 between subjects ANOVA on the untreated items shows: an effect of semantic impairment F(1, 12) = 7.73, p = .017; no effect of phonological impairment F(1, 12) = 3.58, p = .