کنترل توجه خود گزارش شده با مقیاس کنترل توجه: ساختار عاملی و ارتباط با علائم اضطراب و افسردگی

Abstract The Attentional Control Scale (ACS) is a self-report questionnaire that has been developed to measure individual differences in attentional control. Despite its fairly widespread use, little is known about the psychometric properties of the scale in adult samples. In the present study, factor structure of the ACS and its relationship with symptoms of anxiety and depression was investigated in a total sample of 728 Icelandic university students. Exploratory factor analysis in sample 1 (n = 361), yielded two factors, labeled focusing and shifting. Confirmatory factor analysis in sample 2 (n = 367) showed a reasonable fit of this two factor model. The two factors correlated strongly (0.73). The two subscales showed different predictive validity in a set of hierarchical regression analyses where the focusing subscale made a significant prediction of anxiety scores when depression scores were controlled for, and the shifting subscale significant prediction of depression scores when anxiety scores were controlled for. These findings are discussed in relation to previous studies on attentional and executive control in anxiety and depression.

. Introduction Neuroticism or trait anxiety is among the general vulnerability factors for the development of psychological disorders. The influence of such vulnerability factors on behavior is modulated by people's voluntary control efforts that is the ability to modulate or control dominant response tendencies in the service of another response (Rothbart & Bates, 1998). Voluntary control of attention, termed attentional control (Derryberry & Reed, 2001), may be particularly important given the existence of biased information processing at the level of attention and interpretation in anxiety (Cisler and Koster, 2010 and Wells and Matthews, 1994) and depression (Joormann, 2009 and Wisco, 2009). The Attentional Control Scale (ACS; Derryberry & Reed, 2002) is a self-report questionnaire that has been developed to measure individual differences in attentional control. Attention arises from several interacting networks, one being the anterior attentional system that serves an executive control function over other attentional processes (Posner & Rothbart, 1998). Because various functions of the anterior system have been proposed, the ACS was developed as a general scale to assess overall differences in voluntary attentional control (Derryberry & Reed, 2001). The scale has been used to study the role of voluntary control efforts in relation to various symptoms of psychopathology. However, despite its fairly widespread use, little is known about the psychometric properties of the ACS. The purpose of the present study was to investigate psychometric properties of the scale and its relationship to symptoms of anxiety and depression in adults. The ACS comprises 20 items that initially appeared as two scales, attentional focusing and attentional shifting in a study by Derryberry and Rothbart (1988). They defined the construct of attentional focusing as “the capacity to intentionally hold the attentional focus on desired channels and thereby resist unintentional shifting to irrelevant or distracting channels” and attentional shifting as “the capacity to intentionally shift the attentional focus to desired channels, thereby avoiding unintentional focusing on particular channels” (Derryberry & Rothbart, 1988, p. 966). In recent years, the two scales have been combined under the heading of Attentional Control Scale using the total score as a measure of people's ability to control attention. According to Derryberry and Reed (2002), factor analyses of the ACS indicate that it consists of “…correlated sub factors related to the abilities (a) to focus attention (e.g., “My concentration is good even if there is music in the room around me”), (b) to shift attention between tasks (e.g., “It is easy for me to read or write while I’m also talking on the phone”), and (c) to flexibly control thought (e.g., “I can become interested in a new topic very quickly when I need to”)” (Derryberry & Reed, 2002, p. 226). To the best of our knowledge, no studies have been published on the factor structure of the ACS in adult samples. Factor structure of the ACS has been evaluated in one study in a sample of eight to 18 year old Dutch children and adolescents (Verstraeten, Vasey, Claes, & Bijttebier, 2010). Results from confirmatory factor analysis in this study supported two factors rather than one with two items (items 9 and 10) being omitted from the analysis. The total score of the scale is internally consistent with reliability estimates ranging from α = 0.71 ( Gyurak & Ayduk, 2007; Verwoerd et al., 2006; cited in Verwoerd, de Jong, & Wessel, 2008) to α = .88 ( Derryberry & Reed, 2001, cited in Derryberry & Reed, 2002). Reliability estimates of the subscales have not been reported in adult samples, but in the study of Verstraeten et al. (2010) it was α = 0.70 for the focusing scale and α = 0.63 for the shifting scale with the scales being moderately correlated (r = 0.41). Similar results have been reported in other samples of children ( Muris, de Jong, & Engelen, 2004). Although little information is available on the psychometric properties of the ACS in adults, information on the validity of the ACS has been accumulating over the years. ACS has been used in a number laboratory experiments that focus on the interplay between effortful control and automatic processes in adults. Derryberry and Reed (2002) found that although participants high in self reported trait anxiety showed the often hypothesized automatic attentional bias towards threat stimuli in a computerized spatial orienting task (dot-probe task), anxious participants also scoring high on the ACS were better able to shift attention away from the threat at later intervals on the task, indicating that good attentional control allowed trait anxious participants to modulate the dominant attentional bias. Derryberry (2002) discusses three unpublished studies conducted in his laboratory that show that trait anxious participants scoring high on the ACS, compared with low ACS scorers, have better control over dominant response tendencies on a stimulus–response compatibility task, that impulsive participants scoring high on the ACS show better performance than low scorers on a stop-signal task and finally that high ACS scorers are better able to inhibit dominant conceptual associations on a priming task. Other investigators have also shown that attentional control, as indexed by the ACS, is linked with physiological reactions such as an eye-blink response to rejection in people that are sensitive to rejection because of low self-esteem (Gyurak & Ayduk, 2007). These results show that ACS scores are related to response control across different domains of behavior (orienting of attention, motor response tendencies, conceptual processing tendencies, physiological reactions) lending support for the notion that ACS taps the executive function of the anterior attentional system that is hypothesized to be involved in the top-down regulation of automatic responding (i.e., Posner and Raichle, 1994 and Posner and Rothbart, 1998). This conclusion is further supported by results showing ACS scores to be correlated with activation of certain areas within the prefrontal cortex (rostral anterior cingulate) when fear related pictures are being processed (Mathews, Yiend, & Lawrence, 2004). As would be expected, the total score of the ACS has moderate negative correlations with self-report measures of trait-anxiety (Derryberry & Reed, 2001, cited in Derryberry & Reed, 2002), neuroticism (Verwoerd et al., 2008) and depressive symptoms (Reinholdt-Dunne, Mogg, & Bradley, 2009). Verwoerd et al. (2008) also found that lower ACS scores predicted increase in diary ratings of intrusive thoughts over four consecutive days after watching an emotional film fragment, supporting the predictive validity of the scale. Support has also been found with self-report questionnaires for the hypothesized interaction between attentional control and predisposing factors when predicting symptoms of psychopathology. Ayduk et al. (2008) found for example in a non-clinical sample that better attentional control (measured with a 12 item abbreviated version of the ACS) attenuated the association between rejection sensitivity and core symptom dimensions of borderline personality disorder among those high in rejection sensitivity. The aim of the present study was to investigate the ACS and its relationship with symptoms of anxiety and depression using an Icelandic translation of the scale. Although the total score is generally used, the scale was originally composed of focusing and shifting subscales, and measures in addition the ability to flexibly control thoughts (Derryberry & Reed, 2002). Because no results on the factor structure of the ACS in adult samples have been published, we apply a split sample procedure in this study to explore the factor structure of the ACS in one sample and confirmatory factor analysis to test the fit of this factor structure in the second sample. We also wanted to investigate if the ACS sub-factors are differentially related to symptoms of anxiety and depression. Studies have shown that the ACS total score is negatively linked with trait anxiety, neuroticism and depressive symptoms (i.e., Ayduk et al., 2008, Derryberry and Reed, 2002 and Reinholdt-Dunne et al., 2009) in adults but this relationship has not been investigated at the sub-scale level. Both focusing and shifting of attention as well as flexible control of thought are impaired in anxious and depressed moods (i.e., Elliott, 1998, Eysenck et al., 2007 and Veiel, 1997), so corresponding sub-scales should all be related to measures of these symptoms. However, because measures of anxiety and depression are generally moderately correlated, the common variance needs to be taken into account to see if sub-factors of attentional control are uniquely related to anxiety and depression. For example, anxiety is characterized by high vigilance (Weierich et al., 2008 and Wells and Matthews, 1994) that may be modulated by the ability to focus attention in face of distraction (i.e., Eysenck et al., 2007). Unique relationship between attentional shifting and depression would be consistent with considerable research evidence showing that depressed affect is related to a compromised shifting or switching abilities on executive control tasks (Channon, 1996, Davis and Nolen-Hoeksema, 2000, Grant et al., 2001 and Merriam et al., 1999).

3. Results A split sample procedure was performed in the total sample, yielding two samples of 361 (sample 1) and 367 (sample 2) subjects each. The total score on the ACS in sample 1 (M = 51.92; SD = 7.51) and sample 2 (M = 52.42; SD = 7.66) did not differ significantly (t(726) = −0.870, p = 0.385). 3.1. Exploratory Factor Analysis of the ACS Data in sample 1 were subjected to a Principal Components Analysis (PCA) with oblimin rotation of components because they were expected to correlate substantially. We selected PCA on traditional grounds but also because we wanted to extract a set of components explaining maximum amount of variance that would then be evaluated in sample 2 using CFA. The Kaiser–Meyer–Olkin measure of sampling adequacy was 0.868 indicating that the data was suitable for factor analysis. However, correlations between individual items were investigated first. All items had significant correlations in the expected direction in most cases except item 9 that did not have a significant correlation with 15 of the items (p > 0.10 in all cases) and had significant but weak correlations (r = 0.11 and 0.12, p < 0.05 in all cases) with the remaining four items of the scale. Given that items with weak or non-significant correlations with other items tend to perform poorly in a factor analysis (i.e., Floyd & Widaman, 1995), item 9 was not included in the remaining analysis. Parallel analysis was used to decide on number of components to be extracted. In parallel analysis the eigenvalues of components extracted in a PCA are compared to eigenvalues of components that emerge from randomly created datasets. Because these random components are based on capitalization on chance fluctuations in scores, they can form the basis for judging which components have eigenvalues exceeding the values that would have emerged by chance, given a chosen probability level (see for example Child, 2006). For the present parallel analysis, an SPSS macro of O’Connor (2000) was used to create random eigenvalues corresponding to the 95th percentile using 100 data sets. The five first random eigenvalues were 1.49, 1.41, 1.33, 1.28 and 1.22. In the PCA, five components emerged with eigenvalues greater than 1, their values being 5.08, 1.60, 1.26, 1.22 and 1.12. Only the first two components in the PCA have eigenvalues greater than their parallel random components, thus two factors were retained. The results can be seen in Table 1. The two factors explained 35.13% of the variance with the first factor explaining 26.72% and the second factor 8.40%. Table 1. Results from exploratory (EFA) and confirmatory (CFA) factor analysis of the Attentional Control Scale items in two split samples. Description of item content EFA results in sample 1 (n = 361) CFA results in sample 2 (n = 367) I II h2 I II 1. Hard for me to concentrate on a difficult task when there are noises around 0.79 −0.10 0.56 0.63 3. When working on something, still get distracted by events around me 0.75 −0.04 0.49 0.71 6. When reading/studying, get easily distracted if people talking 0.75 −0.11 0.54 0.70 5. When concentrating, can focus and become unaware 0.65 −0.03 0.42 0.56 4. Concentration is good, even if there is music in the room around me 0.63 0.04 0.41 0.46 2. When need to concentrate/solve a problem, trouble focusing 0.61 0.17 0.50 0.68 7. When trying to focus, have difficulty blocking out distracting thoughts 0.55 0.14 0.40 0.62 8. Have a hard time concentrating when excited about something 0.46 0.15 0.29 0.57 12. Difficult to coordinate attention between listening/writing 0.33 0.27 0.41 0.37 13. Can become interested in a new topic very quickly when I need to −0.22 0.73 0.30 0.29 10. Can quickly switch from one task to another −0.05 0.66 0.27 0.42 16. Have hard time coming up with new ideas quickly −0.03 0.56 0.44 0.37 19. Easy for me to alternate between two different tasks 0.06 0.55 0.20 0.56 11. Takes me a while to get really involved in a new task 0.09 0.50 0.24 0.54 15. Have trouble carrying on two conversations at once −0.03 0.50 0.30 0.32 14. Easy for me to read/write while talking on the phone 0.13 0.37 0.15 0.33 18. Distracting thought comes to mind, easy for me to shift my attention away from it 0.28 0.35 0.29 0.54 20. Hard to break from one way of thinking to another 0.11 0.34 0.33 0.27 17. After being interrupted/distracted, easily shift attention back 0.13 0.32 0.16 0.44 Note: h2 = communality. Items 1–3, 6–8, 11, 12, 15, 16, 18 and 20 are reversed for scoring. For the EFA, pattern matrix is shown with oblimin rotation and loadings of 0.30 or greater in bold. For the CFA, standardized solution is shown. All loadings were significant. Error terms between items items 5 and 4, 17 and 18, 3 and 6 and 7 and 8 were allowed to correlate (not shown in the table). Table options Nine items have their highest loading on factor 1. All loadings are above the traditional 0.30 level and range from 0.33 to 0.79. These items seem to reflect the ability to control attention by focusing it in face of distraction and factor 1 was therefore labeled attentional focusing. Ten items have their highest loading on factor 2 with all loadings above 0.30 and ranging from 0.32 to 0.73. The content of these items is more concerned with controlling attention by being able to shift it between different tasks (item 19: It is easy for me to alternate between two different tasks) and flexibly controlling content of one's thoughts (item 13: I can become interested in a new topic very quickly when I need to). The second factor was therefore labeled attentional shifting. 3.2. Confirmatory Factor Analysis of the ACS The two factor model with correlated factors that emerged in sample 1 was evaluated with Confirmatory Factor Analysis (CFA) in sample 2. The fit of this model was also compared with the fit of two factor model with orthogonal factors and with a single factor model with all items loading on one factor. This was done to see if the factor structure of the ACS would be better represented by a single or a two factor model and to see if a two factor model would be better represented by correlated or orthogonal factors. The correlated two factor model from sample 1 fitted reasonably well in sample 2 (S–B χ2 = 446.53, df = 151, p < 0.001; RMSEA = 0.073; SRMR = 0.064; CFI = 0.92) although the CFI value fell below the 0.95 criteria. The fit of this model with orthogonal factors was poor (S–B χ2 = 543.82, df = 152, p < 0.001; RMSEA = 0.084; SRMR = 0.15; CFI = 0.89). A single factor model had also a poor fit (S–B χ2 = 518.47, df = 152, p < 0.001; RMSEA = 0.081; SRMR = 0.068; CFI = 0.90). Modification indices for the correlated two factor model indicated that the fit of the model could be improved by allowing the error terms between items 5 and 4, 17 and 18, 3 and 6 and between items 7 and 8 to correlate. This adjusted model met all requirements for a reasonable model fit (S–B χ2 = 311.04, df = 147, p < 0.001 RMSEA = 0.055; SRMR = 0.059; CFI = 0.95) and fitted the data significantly better than the previous model (ΔS–B χ2 = 105.75, Δdf = 4, p < 0.001). The two latent factors correlated strongly or 0.73. The standardized factor loadings in this model are shown in Table 1. The magnitudes of the loadings on the focusing factor are greater than the loadings on the shifting factor although all parameters were significant in this model. Two items, items 13 and 20, have loadings below 0.30 on the shifting factor, or 0.29 and 0.27 respectively. To maximize precision in parameter estimates, the adjusted two factor model was finally fitted to the data combined from both samples. The model had a reasonable fit (S–B χ2 = 514.27, df = 147, p < 0.001; RMSEA = 0.059; SRMR = 0.53; CFI = 0.95). 3.3. Descriptive statistics, internal consistency and relationship with gender and age Internal consistency of the ACS and its subscales was calculated in the total sample. Internal consistency of the total score (α = 0.84) and the focusing factor (α = 0.82) was good but of the shifting factor it was reasonable (α = 0.68). Corrected item-total correlations for the items of the focusing factor ranged from a low 0.36 (item 12) to a high 0.60 (items 2 and 3) with a mean correlation of 0.53. On the shifting factor, item 20 had corrected item-total correlation of 0.23 that is below the traditional 0.30 minimum but the correlations for the other items ranged from 0.31 to 0.43 (item 19) with a mean correlation of 0.35 for all the ten items. In the total sample, mean score on the ACS was 50.17 (SD = 7.49), on the focusing factor 23.77 (SD = 4.54) and on the shifting factor 26.40 (SD = 3.96). The corresponding mean values for men (n = 234) were 50.90 (SD = 6.40), 24.21 (SD = 4.08) and 26.69 (SD = 3.50) but for females (n = 493) they were 49.80 (SD = 7.92), 23.55 (SD = 4.72), and 26.25 (SD = 4.17). T-tests indicated significant gender differences in means for the total score on the ACS (t(556.35) = 2.00, p < 0.05), marginally significant differences on the focusing subscale (t(522.89) = 1.95, p = 0.051) but no significant difference on the shifting subscale (t(536.91) = 1.47, p = 0.14). The total score on the ACS correlated only weakly with age (r = 0.08, p < 0.05). Scores on the focusing subscale had a weak correlation with age (r = 0.10, p < 0.01) but scores on the shifting factor were unrelated to age (r = 0.04, p = 0.30). 3.4. Relationship with anxiety and depression Both focusing and shifting factors of the ACS had moderate and negative correlations with symptoms of anxiety (r = −0.45 and −0.35 respectively, p < 0.001) and depression (r = −0.37 and −0.33 respectively, p < 0.001), indicating that both types of symptoms are associated with decreased ability to focus and shift attention. However, scores on the anxiety and depression scales were strongly correlated in this study (r = 0.64, p < 0.001) and of interest is to see if the ACS subscales are differentially related to these symptoms after accounting for the common variance between them. To investigate this, two hierarchical regression analyses were conducted and the focusing and shifting subscales used to predict anxiety and depression scores. Gender and age were entered on the first step, anxiety was entered on the second step when predicting depression but depression was entered on the second step when predicting anxiety to control for the common variance between these variables. The ACS subscales were then entered simultaneously on the third step. The results are shown in Table 2. Table 2. Results from two hierarchical regression analyses using focusing and shifting subscales of the ACS to predict anxiety and depression scores (n = 727). B SE B Beta Dependent variable: HADS anxiety Step 1 (ΔR2 = 0.009 *) Gender 0.808 0.220 0.100*** Age −0.003 0.016 −0.005 Step 2 (ΔR2 = 0.421 ***) HADS depression 0.739 0.039 0.553*** Step 3 (ΔR2 = 0.052 ***) ACS focusing −0.188 0.028 −0.224*** ACS shifting −0.039 0.031 −0.041 Dependent variable: HADS depression Step 1 (ΔR2 = 0.002) Gender −0.695 0.171 −0.115*** Age 0.014 0.012 0.033 Step 2 (ΔR2 = 0.424 ***) HADS anxiety 0.446 0.024 0.596*** Step 3 (ΔR2 = 0.015 ***) ACS focusing −0.031 0.022 −0.050 ACS shifting −0.072 0.024 −0.100** * p < 0.05. ** p < 0.01. *** p < 0.001. Table options In Table 2 it can be seen that in the prediction of anxiety scores, only the focusing subscale of the ACS makes a significant contribution when depression and demographics have been taken into account. In prediction of depression scores, however, the shifting subscale and not focusing makes a significant contribution above what can be accounted for by anxiety and demographics. This indicates a unique relationship between a compromised ability to focus attention in face of distraction with symptoms of anxiety and a compromised ability to shift attention with symptoms of depression, when the common variance between these constructs has been controlled for.