شرح مشکلات اولیه کنترل توجه در سندرم ایکس شکننده: تمرکز بر روی تغییرات محاسباتی عصبی

Abstract Fragile X syndrome (FXS) is due to the silencing of a single X-linked gene and it is associated with striking attentional difficulties. As FXS is well characterised at the cellular level, the condition provides a unique opportunity to investigate how a genetic dysfunction can impact on the development of neurocomputational properties relevant to attention. Thirteen young boys with FXS and 13 mental-age-matched typically developing controls performed a touch-screen-based search task that manipulated the similarity between targets and distractors and their heterogeneity in size. Search speed, path and errors were recorded as multiple measures of performance. Children did not differ in overall search speed or path when searching amongst distractors, but striking error patterns distinguished children with FXS from controls. Firstly, although clear markers of previously found targets remained on screen, children with FXS perseverated on touching previous hits more than typically developing controls, consistent with the well-documented inhibitory deficits in adults with the disorder. Secondly, they could accurately discriminate single target-distractor pairs, but, when searching a complex display, they touched distractors more often than control children when distractors were similar to targets and especially so when these were infrequent, highlighting difficulties in judging relative size and allocate attentional weight independently of stimulus frequency. Thirdly, their performance was also characterised by inaccuracies in pointing, suggesting additional motor control deficits. Taken together, the findings suggest that fragile X syndrome affects the early development of multiple processes contributing to efficient attentional selection, as would be predicted from an understanding of the neurocomputational changes associated with the disorder.

Results Table 1 represents the total number of targets found, total number of errors, different error types, search speed and path as a function of group, target-distractor similarity and distractor heterogeneity. The table also includes average performance on baseline runs and reports statistically significant main effects and interaction effects. In summary, children with FXS did not search amongst distractors more slowly or less systematically than MA controls (speed and path measure), but the overall pattern of errors produced distinguished the two groups. Children with FXS produced more repetitive errors, and their errors were more influenced by similar than by dissimilar distractors compared to MA controls. They also produced more inaccurate pointing errors. These trends were supported statistically as follows. Table 1. Average number of errors, erroneous touches on distractors and repetitive errors on previously found targets as a function of group, target-distractor similarity and distractor heterogeneity (standard error of the mean is indicated in brackets) Dependent variables Distractor types Baseline Homog. dissimilar Homog. similar Heterog. (infrequent dissimilar) Heterog. (infrequent similar) Statistically significant effects F ratio, p value Search accuracy Accurate responses (per run) FXS 7.23 (.49) 6.85 (.54) 6.0 (.55) 6.38 (.56) 6.84 (.54) Similarity F(1, 24) = 9.973, p = .004 MA 8.0 (.00) 7.92 (.08) 7.46 (.39) 7.61 (.31) 7.77 (.23) Total errors (per run) FXS 4.15 (.17) 6.0 (1.15) 7.54 (1.01) 7.46 (1.16) 7.3 (1.11) Group F(1, 24) = 26.593, p < .001 MA .77 (.34) 1.16 (.59) 3.0 (1.0) 2.0 (.68) 2.92 (.77) Repetitions (per hit) FXS .69 (.41) .40 (.12) .43 (.10) .49 (.16) .34 (.08) Group F(1, 24) = 7.788, p = .01 MA .06 (.03) .09 (.04) .13 (.06) .05 (.04) .23 (.09) Touches on distractors (per run) FXS NA 1.23 (.51) 4.62 (.97) 4.0 (.89) 2.46 (.55) Group F(1, 24) = 15.885, p = .001 MA NA .46 (.24) 2.15 (.77) 1.23 (.62) .85 (.37) Similarity F(1, 24) = 21.125, p < .001 Inaccurate touches (per run) FXS 1.46 (.67) 1.92 (.67) .69 (.23) 1.15 (.59) 1.23 (.44) Group F(1, 24) = 9.763, p = .005 MA .23 (.17) .08 (.08) .15 (.10) .31 (.17) .31 (.17) Search speed (s) FXS 1.33 (.13) 1.84 (.51) 1.54 (.18) 1.74 (.24) 1.81 (.24) MA 1.63 (.15) 1.55 (.18) 2.17 (.24) 1.63 (.21) 1.58 (.19) Search path (cm) FXS 7.76 (.45) 6.67 (.58) 5.18 (.39) 5.53 (.38) 5.58 (.45) Group (baseline condition only) t(24) = 4.863, p < .001 MA 4.99 (.35) 4.89 (.32) 5.32 (.41) 5.57 (.29) 5.01 (.26) Table options 2.1. Total number of targets found and errors There was a statistically significant main effect of target-distractor similarity on the total number of targets found. This was confirmed by a Friedman test, χ2 = 13.088, p = .004. Runs with distractors that were similar to targets resulted in overall fewer targets found compared to those with dissimilar distractors (on average, 6.9 hits and 7.4 hits per run, respectively). The effect of group on the number of targets found also displayed a trend towards significance, F(1, 24) = 4.532, p = .048 (driven by the fact that children with FXS tended to find fewer targets than MA controls), which did not however survive Bonferroni correction. Neither heterogeneity nor any of the interactions had any significant effect on accuracy, highest F value = 1.273, p = .267. Children with FXS and controls did not differ in the total number of accurate touches on baseline runs, t(12) = −1.552, p = .147. In terms of overall errors, children with FXS produced more errors in total than MA controls (on average, 7.0 errors per run and 2.38, respectively). A significantly larger number of errors for children with FXS across all conditions were also confirmed through non-parametrics statistics (Mann–Whitney U Test, p < .01 for all comparisons). The groups also tended to differ in the number of errors on baseline runs, t(13.305, corrected for homogeneity violations) = 2.247, p = .042, with children with FXS committing more errors than MA controls (4.1 errors as opposed to .77 errors, respectively), a comparison that again did not survive Bonferroni corrections. However, we explored whether this trend towards baseline group differences in total number of errors could account for the large group differences in errors over the runs with distractors, and this was not the case. The main effect of Group on total errors remained statistically significant when baseline errors were used as a covariate, F(1, 23) = 13.122, p = .001. None of the other main effects or interactions reached statistical significance, highest F value = 1.834, p = .188. Beyond the overall differences in total number of errors, the current analysis aimed to characterise potential group differences in error types, by subdividing them into repetitive errors, distractor errors and inaccurate touches. A more conservative criterion was therefore employed to control for the increased likelihood of Type I errors, corrected α = .05/3 = .016. 2.2. Repetitions on previously touched targets Mean repetitions on previously found targets per run are represented in Fig. 2. Children with FXS produced more repeats on previously found targets than MA controls (on average, .491 repeats per hit and .129, respectively). These group differences were also tested using non-parametric statistics: children with FXS produced significantly more of these errors than controls in all four conditions, Mann–Whitney U Test, p < .01 for all comparisons. None of the other effects (similarity and homogeneity) and interactions reached significance, highest F(1, 24) = 3.243, p = .084. Children with FXS did not differ from MA controls in the number of repeats per hit on baseline runs, t(12.102, corrected for homogeneity violations) = 1.543, p = .148. Average number of repetitive errors per hit by MA controls and children with ... Fig. 2. Average number of repetitive errors per hit by MA controls and children with FXS, as a function of target-distractor similarity and distractor heterogeneity (standard error of the mean is indicated in brackets). Figure options Repetitive touches on targets were further subdivided into immediate repeats, and returns to previous targets after other locations were visited, to explore differences in these types of errors previously reported for older children with FXS (Wilding et al., 2002) and these are reported in Table 2. Alpha was divided by the number of mutually exclusive types of repetitive errors, corrected α = .05/2 = .025. Table 2 presents mean number of different types of repetitive errors per hit across the different search displays. There was a statistically significant main effect of Group on immediate repetitions per hit, F(1, 24) = 6.654, p = .016, due to children with FXS producing a larger number of these errors per hit (.288, SEM = ±.05 on average) than MA controls (.112, SEM henceforth = ±.05 on average). There was also a marginally significant main effect of Group on returns on previously found targets, F(1, 24) = 6.000, p = .022, due to children with FXS producing a larger number of these errors per hit (.131 ±.03 on average) than MA controls (.017 ±.03 on average). None of the other main effects (similarity and homogeneity) or interactions reached statistical significance, highest F value = 3.537, p = .074. Groups did not differ in the number of immediate repeats or returns in baseline runs, t(24) = 1.462, p = .157 and t(12, corrected for homogeneity violations) = 1.648, p = .125, respectively. Table 2. Average number of immediate repeats on previously found targets per hit and returns to hits as a function of group, target-distractor similarity and distractor heterogeneity (standard error of the mean is indicated in brackets) Type of repetition error Distractor type Baseline Homog. dissimilar Heterog. (infrequent dissimilar) Heterog. (infrequent similar) Homog. similar Immediate repeats (per hit) FXS .60 (.37) .29 (.07) .36 (.11) .19 (.06) .31 (.08) MA .06 (.03) .09 (.04) .05 (.04) .19 (.09) .13 (.06) Returns (per hit) FXS .09 (.06) .11 (.08) .14 (.07) .15 (.06) .12 (.05) MA .00 (.00) .00 (.00) .01 (.01) .05 (.02) .01 (.01) Table options 2.3. Erroneous touches on distractors Mean distractor touches per run are represented in Fig. 3. Children with FXS produced more touches on distractor circles than MA controls (on average, 1.5 and .59 errors of this type per run, respectively). There was also a main effect of target-distractor similarity, due to a greater number of distractor errors when distractors were similar (1.5 per run on average) than when they were dissimilar to targets (.63 per run on average). The main effect of similarity was confirmed using non-parametric statistics: there were statistically significant differences across conditions, Friedman test, χ2 = 13.830, p = .003, and fewer of these errors were committed in the condition with homogeneous dissimilar distractors than in the condition with homogeneous similar distractors, Wilcoxon Signed Ranks Test, p = .001. These differences seemed greater for children with FXS, Wilcoxon Signed Ranks Test, p = .008, than for MA controls, Wilcoxon Signed Ranks Test, p = .049 (with the latter not surviving correction for multiple comparisons). However, when tested using parametric statistics, none of the other effects and interactions reached significance, highest F value = 3.492, p = .074 for the interaction between similarity and group. Average number of distractor touches per run by MA controls and children with ... Fig. 3. Average number of distractor touches per run by MA controls and children with FXS, as a function of target-distractor similarity and distractor heterogeneity (standard error of the mean is indicated in brackets). Figure options To explore further the type of distractor errors produced by all children, these were subdivided into touches on the smallest (most dissimilar) distractors, and those on the medium (similar) distractors across heterogeneous conditions. In these conditions, medium and small distractors were intermixed in different proportions, resulting in either similar or dissimilar distractors being more infrequent, and therefore more salient compared to the conditions in which they were frequent. Therefore, in addition to the absolute number of touches on each distractor type, we calculated the percentage of touches divided by the variable number of distractors of that type present in each display and we thank an anonymous Reviewer for the latter suggestion. Mean number of touches on similar and dissimilar distractors across all conditions and the percentage of these touches moderated by the overall number of each distractor type per display are presented in Table 3. Alpha was divided by the number of mutually exclusive types of distractor errors, corrected α = .025. Table 3. (a) Average absolute number of distractor touches per run on medium (similar) and small (dissimilar) distractors, respectively and (b) proportion of targets of that type touched when search displays contained both types of distractor: heterogeneous displays with infrequent similar distractors or heterogeneous with infrequent dissimilar distractors Type of distractors Distractor condition Homog. dissimilar Homog. similar Heterog. (infrequent similar) Heterog. (infrequent dissimilar) Similar distractors Similar distractors per display 0 24 6 18 Touches on similar distractors (per run) FXS – 4.62 (.98) 2.15 (.45) 2.39 (.68) MA – 2.15 (.77) .77 (.32) 1.08 (.63) Touches on similar distractors (per run, %) FXS – 19.23 (4.1) 35.89 (7.5) 13.25 (3.8) MA – 12.82 (5.4) 5.98 (3.5) 8.98 (3.2) Dissimilar distractors Dissimilar distractors per display 24 0 18 6 Touches on dissimilar distractors (per run) FXS 1.23 (.31) – .30 (.13) 1.62 (.76) MA .46 (.24) – .08 (.08) .15 (.10) Touches on dissimilar distractors (per run, %) FXS 5.13 (2.1) – 1.71 (.74) 26.92 (12.7) MA 1.92 (1.0) – .43 (.43) 2.56 (1.7) The mean number of distractor touches for the homogeneous conditions are also provided for reference (standard error of the mean is indicated in brackets). Table options Let us first consider absolute mean number of distractor touches on similar and dissimilar distractors. In the heterogeneous condition with infrequent similar distractors (6 per display), all children incorrectly touched more frequently the similar compared to the dissimilar distractors, F(1, 24) = 29.170, p < .001, and children with FXS made more distractor touches overall, F(1, 24) = 5.906, p = .023. Critically, similarity of the distractors and Group interacted, F(1, 24) = 6.027, p = .022. This interaction effect was due to children with FXS incorrectly touching similar distractors more frequently (on average 2.15 times per run) than MA controls (on average, .77 times per run), t(24) = 2.496, p = .020 (Mann–Whitney U Test, p = .012), while the two groups did not differ in the number of times they touched dissimilar distractors, t(19.2, corrected for violations of homogeneity) = 1.50, p = .150 (.31 and .08 times per run, respectively, Mann–Whitney U Test, p = .289). In this heterogeneous condition, children with FXS also touched similar distractors more often than dissimilar distractors, t(12) = 4.951, p < .001 (Wilcoxon Signed Rank Test, p = .002), whereas MA controls only exhibited a trend in this direction, t(12) = 2.420, p = .032 (Wilcoxon Signed Rank Test, p = .041), which did not survive correction for multiple comparisons. When dissimilar distractors were infrequent, children with FXS generally touched distractors more than controls, F(1, 24) = 6.421, p = .018 (on average 2.0 and .62 of these errors for children with FXS and MA controls, respectively), but touches on similar and dissimilar distractors did not differ, F(1, 24) = 1.607, p = .205. When one considers the number of distractors touched in the two heterogeneous displays in proportion to their number in each display (e.g., 6 distractors if infrequent, 18 distractors if frequent), a similar pattern of group differences emerges. When similar distractors were infrequent (6 per display), all children erroneously touched a higher proportion of these distractors (by touching 24.3% of distractors of these type, ±4.6) compared to dissimilar distractors (by touching 1.07% of dissimilar distractors in the display ±.43), F(1, 24) = 28.56, p < .001, and children with FXS touched a higher proportion of distractors overall (18.8% ± 3.48) compared to MA controls (6.62% ± 3.48), F(1, 24) = 6.153, p = .021. Additionally, Group and distractor type interacted significantly, F(1, 24) = 6.252, p = .020. This was due to children with FXS touching a significantly higher proportion of similar distractors (35.89% ± 6.54) compared to MA controls (12.82% ± 6.54), t(24) = 2.496, p = .020 (Mann–Whitney U Test, p = .012), and a significantly higher proportion of similar distractors compared to dissimilar ones, t(12) = 4.864, p < .001 (Wilcoxon Signed Rank Test, p = .002), This difference was statistically significant for MA controls, t(12) = 2.404, p = .033 (Wilcoxon Signed Ranks Test, p = .042), but did not survive correction by multiple comparisons. The two groups did not differ in the proportion of dissimilar distractors touched (1.71% ± .6 for children with FXS and .43% ± .6 for MA controls), t(19.200, corrected for heterogeneity of variance) = 1.5, p = .147 (Mann–Whitney U Test, p = .143). In heterogeneous displays in which dissimilar distractors were infrequent, children with FXS simply produced more distractor touches (20.09% ± 4.59) than MA controls (4.28% ± 4.59) overall, F(1, 24) = 5.936, .023. Finally, children with FXS were more affected than controls by the relative proportion of distractors of various sizes, producing more touches on similar distractors when they were infrequent than when they were frequent, t(12) = 2.851, p = .015 (Wilcoxon Signed Ranks Test, p = .007). This difference was not statistically significant for controls, t(12) = 1.298, p = .219. 2.4. Inaccurate touches Fig. 4 illustrates the average number of inaccurate touches per run by group and search condition. There was a main effect of Group on the number of inaccurate touches in the experimental trials, due to children with FXS producing a larger number of these errors per run (1.250 ± .24 on average) than MA controls (.212 ± .24 on average). This was also confirmed through non-parametric statistics, showing statistically significant differences across groups when homogeneous similar distractors were present, Wilcoxon Signed Ranks Test, p = .001. The effect of homogeneity and the other interaction effects did not reach statistical significance, highest F value = 2.678, p = .115. Groups did not differ in terms of inaccurate touches in baseline runs, t(13.488) = 1.793, p = .095. Average number of inaccurate touches per run by MA controls and children with ... Fig. 4. Average number of inaccurate touches per run by MA controls and children with FXS, as a function of target-distractor similarity and distractor heterogeneity (standard error of the mean is indicated in brackets). Figure options 2.5. Analyses of search speed and path Target-distractor similarity, distractor heterogeneity and group membership did not affect search speed differentially in the two groups. None of the interaction effects reached statistical significance, highest F value = 2.919, p = .1. Furthermore, children with FXS and MA controls did not differ in speed to find targets on baseline runs, t(24) = 1.498, p = .147. Similarity, heterogeneity and group did not affect search path significantly, highest F value = 2.177, p = .152, but there was a trend towards statistically significant interaction between similarity and group, F(1, 24) = 4.730, p = .040, which did not however survive Bonferroni corrections. However, children with FXS produced significantly larger search paths than MA controls on baseline runs, t(24) = 4.863, p < .001 (Mann–Whitney U Test, p < .001). We therefore explored whether this difference in search path in runs that did not contain distractors could influence group differences in distractor runs by employing baseline distance between successive touches as a covariate in the earlier analysis of variance. None of the main effects or interaction effects in this additional analysis reached statistical significance, lowest p = .146.