- Email: [email protected]
The development and initial validation of a new scale to measure explanatory style
Personality and Individual Differences 81 (2015) 1–6
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
The development and initial validation of a new scale to measure explanatory style q Katrina M. Travers a,⇑, Peter A. Creed b, Shirley Morrissey b a b
Australian Institute of Psychology, Brisbane, Queensland, Australia School of Applied Psychology and Griffith Health Institute, Griffith University, Australia
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 24 January 2015 Accepted 27 January 2015 Available online 12 February 2015 Keywords: Explanatory style Reformulated learned helplessness theory Attributional Style Questionnaire New explanatory style scale
a b s t r a c t The reformulated learned helplessness (RLH) theory and its associated construct of explanatory style have been tested extensively using the Attributional Style Questionnaire (ASQ: Peterson et al., 1982) against outcomes such as depression. However, support for the RLH theory is best described as inconclusive. This is because: (a) the causal-locus dimension is poorly defined and the causal-locus items have poor reliability and validity; (b) the ASQ has not been demonstrated to have structural validity; and (c) the definitions of explanatory style are inconsistent across studies. The current study was conducted with the goal of developing a valid and reliable measure of explanatory style. The new measure met simple structure when tested using exploratory and confirmatory factor analyses, and each short subscale had acceptable internal reliability. Further, construct validity was partially demonstrated, but incremental validity was not supported. Implications for theory and practice are discussed. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction The reformulated learned helplessness (RLH) theory (Abramson, Seligman, & Teasdale, 1978) contends that negative psychological consequences (e.g., depression) occur when a person is pessimistic in their explanation of unexpected negative events. This pessimism has been defined as an explanatory style that is generalising (global), persistent (stable), and personal (locus) (Abramson et al., 1978). Peterson et al. (1982) developed a measure of explanatory style (the Attributional Style Questionnaire; ASQ), to investigate this theory. The ASQ, however, has returned inconclusive results in construct validity investigations, meaning that the validity of the RLH theory has been poorly supported. Reasons include: correlations between the ASQ subscales and outcomes such as depression have been non-significant or weak, particularly with the causallocus subscale (Bennett & Elliott, 2002; Dua, 1995; Gibb, Zhu, Alloy, & Abramson, 2002; Luten, Ralph, & Mineka, 1997; Michelson, Bellanti, Testa, & Marchione, 1997; Volpe & Levin, 1998); structural validity of the ASQ has not delineated a parsimonious model-of-fit (Furnham, Sadka, & Brewin, 1992;
q
This article is a Special issue article – ‘‘Young researcher award 2014’’.
⇑ Corresponding author at: Australian Institute of Psychology, Locked Bag 15, Fortitude Valley, Queensland 4006, Australia. Tel.: +61 731122000; fax: +61 732577195. E-mail address: [email protected] (K.M. Travers). http://dx.doi.org/10.1016/j.paid.2015.01.045 0191-8869/Ó 2015 Elsevier Ltd. All rights reserved.
Kleim, Gonzalo, & Ehlers, 2011; Proudfoot, Corr, Guest, & Gray, 2001; Smith, Caputi, & Crittenden, 2013; Xenikou, Furnham, & McCarrey, 1997); and research has been inconsistent in the way that explanatory style has been defined. For example, some studies have defined explanatory style with causal-stability and causalglobality only (Dykema, Bergbower, & Peterson, 1995; Peterson & Vaidya, 2001), or only causal-globality (Bennett & Elliott, 2002). Thus, Abramson, Metalsky, and Alloy (1989) proposed an alternative theory to the RLH theory, named the hopelessness theory of depression, purporting that it was causal-globality and causal-stability that predisposed a person to hopelessness, and thus hopelessness depression (HD). Hopelessness theory has been supported in the literature. For example, Sturman, Mongrain, and Kohn (2006) found a significant correlation between generality (a composite of causal-stability and causal-globality) and a composite of five HD symptoms (r = .24; p = .02); whereas, generality was uncorrelated with non-HD symptoms (r = .03, p = .76). Although Abramson et al.’s (1989) solution to revise the RLH theory to the hopelessness theory has received some support, the mixed results summarised for the RLH theory could also be due to the problematic causal-locus dimension in the ASQ. Thus, a revised causal-locus subscale is required so that the RLH model can be assessed along the three original explanatory style dimensions. The goal of the current study was to develop an alternative measure of explanatory style that incorporated new causal-locus items, which would allow the RLH theory to be measured consistently, and according to its definition.
2
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
1.1. The causal-locus dimension items The causal-locus dimension of explanatory style was originally conceived by Lewin (1936), who was interested in understanding social behaviour from a mathematical perspective (Burnes & Cooke, 2013). Lewin articulated the life-space theory which theorised that a person’s behaviour (B) is a function of the interaction between the person (p) and the environment (situation) (e), or B = f(p,e). This person/situation distinction in life-space theory was influential in conceptualising the personal/impersonal heuristic in Heider’s (1958) attributional theory of social perception. Heider theorised that lay people rely on this heuristic when explaining intentional social behaviour as opposed to unintentional actions (e.g., yawning) (Malle, 2011). According to Malle (2011), Heider’s (1958) personal/impersonal heuristic was misrepresented in subsequent attributional theories including those by Kelley (1967) and Weiner (1986) to refer to the causal-locus dichotomy (internal/external), and it was this misrepresentation that persisted in later attributional theories including the RLH theory by Abramson et al. (1978). This interpretation of the person/situation dichotomy continues to be debated. Weiner (2014), for instance, argued that the causal-locus dimension cannot be considered independent to a controllability dimension. Malle, Knobe, O’Laughlin, Pearce, and Nelson (2000) theorised that the causal-locus dimension is better understood within a folk-conceptual framework. Briefly, the folk-conceptual framework purports that people explain intentional (controllable) behaviour using one of three modes of explanation (causal history of reasons, enabling factors, or reason explanations including desire, belief or value). Finally, Gilbert, Pinel, Brown, and Wilson (2000) questioned the validity of a causal-locus dichotomy, arguing that persons have an unconscious ‘‘psychological immune system’’ (PIS) which naturally externalises causes of sub-optimal events in order to increase self-satisfaction. In this study, we refer back to the original life-space theory by Lewin (1936) and propose re-writing the ASQ causal-locus items so that this dimension is measured by items reflecting each of the three life-space domains (behaviour, person, and environment) in individual items, with each item anchored by internal and external descriptions of the Lewinian domain being measured. This would achieve a more consistent, comprehensive, and interpretable causal-locus dimension within the construct of explanatory style. 1.2. Structural validity of the ASQ Previous attempts to find structural validity of the ASQ using exploratory factor analysis (EFA) have been unsuccessful. Kleim et al. (2011) found one factor, Corr and Gray (1996), Higgins, Zumbo, and Hay (1999), and Proudfoot et al. (2001) found two factors, and Xenikou et al. (1997) found three (interpreted as two positive and one negative factor). Higgins et al. (1999) argued that these previous attempts have been unsuccessful due to the inbuilt ‘‘testlets’’ or common item-stems which generate co-variation among items of the same stem if not allowed to correlate when conducting an EFA. Higgins et al. (1999) suggested that error terms in confirmatory factor analysis be allowed to correlate in order to address this. We addressed this issue by reformatting the ASQ so that each item-stem is answered by only one item, thereby eliminating the co-variation due to the testlets. We also included only the negative item-stems. 1.3. Objectives Thus, this investigation was conducted to achieve four aims: (a) to generate new causal-locus items reflecting the underlying life-
space theory, (b) to create a unique item-stem for each item, (c) to establish initial internal reliability and structural validity for the new scale, and (d) to establish initial construct and incremental validity of the scale. To achieve these aims, a two-phase study was completed. Phase 1 addressed the first three aims, and Phase 2 addressed the final aim. 2. Phase 1 2.1. Materials and methods 2.1.1. Participants Three participant samples volunteered for Phase 1. Sample 1 included six academic psychologists with doctorate level training (five females; mean age = 35.5 years, SD = 11.1 years). Sample 2 included 320 participants: 248 participants (77.5%) were first-year psychology students (189 females), and 72 (22.5%) were community members (49 females). The age range was 18–59 years (M = 25.82, SD = 10.25). Sample 3 consisted of 409 participants: 258 (63.1%) were first-year university students (181 female) and 151 (36.9%) were community members (89 females). The age range was 16–76 years (M = 26.43, SD = 12.25). 2.1.2. Procedure Thirty new items (six each for causal-global, causal-stable, and the three life-space domains) were generated deductively. These new items were then added to the 12 original negative causal-global and causal-stable items, resulting in 42 items, allowing for redundant items during item and factor analysis (Hinkin, Tracey, & Enz, 1997). Further, 36 new negative item-stems were generated and added to the existing six negative item-stems, thereby allowing each item to be preceded with a unique item-stem. These itemstems were randomly allocated to each dimension (including the three life-space domains). The original 7-point Likert-type response was retained (Hinkin et al., 1997). Finally, all items were randomly ordered throughout the scale to avoid explanatory style dimensions being measured by consecutive items. Focus-group methodology (Greenbaum, 1993) was used to assess content validity of the new items and stems, using Sample 1 participants. See Fig. 1 for sample items of each dimension, including each of the three Lewinian domains for the causal-locus dimension. Sample 2 participants were then administered the 42-item scale for item analysis and EFA procedures. The scale was distributed to participants in class who were given the option of completing it immediately, or at a later time. Community respondents were distributed surveys using snowballing procedures. The scale resultant from item analysis and EFA procedures was then administered to Sample 3 participants using identical distribution procedures, in order to confirm the initial structural validity results using confirmatory factor analysis (CFA) procedures. All participants were treated in accordance with the university’s Human Research Ethics Committee guidelines. 2.2. Results 2.2.1. Sample 2 item analysis Item distributions, inter-item, and item-total correlations were used to assess for problematic items. Items with poor distributions or correlations at r 6 .20 were removed. This resulted in the deletion of 12 items. 2.2.2. Sample 2 exploratory factor analysis Firstly, the remaining 30 items (ten each for causal-globality, causal-stability, and causal-locus), were analysed using Horn’s (1965) parallel-analysis to determine the correct number of factors
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
3
Fig. 1. Sample items for causal-globality, causal-stability, and causal-locus (represented by the three Lewinian domains of personality, behaviour, and environment) for the new explanatory style scale.
to be retained (Velicer, Easton, & Fava, 2000). Results indicated that three factors were to be extracted. The items were then analysed using principal axis factor analysis (PAF) with varimax rotation (de Winter & Dodou, 2012; Worthington & Whittaker, 2006). An orthogonal rotation was selected since the explanatory style dimensions have returned low inter-correlations in previous investigations (Asner-Self & Schreiber, 2004; Jeavons & Greenwood, 2007; Peterson & Villanova, 1988; Peterson et al., 1982). The KMO measure was adequate (.75) and the Bartlett’s Test of Sphericity was significant (<.001). Items were retained if they returned factor loadings P.30 with no cross loadings, resulting in the deletion of four causal-locus, three causal-stability, and one causalglobality item. A second analysis produced three factors, representing causal-globality (nine items), causal-stability (seven items) and causal-locus (six items). To ensure an equal number of items for each factor, the six causal-globality and causal-stability items with the highest factor loadings were retained, resulting in a final 18 item solution, meeting simple structure, and accounting for 38.7% of the variance. The three factors were interpreted as ‘‘causal-globality’’, ‘‘causal-stability’’, and ‘‘causal-locus’’. The 18 items, their factor loadings and summary statistics are reported in Table 1. 2.2.3. Sample 3 confirmatory factor analysis Thirteen cases were deleted due to missing data, leaving 396 cases for analysis. We used maximum likelihood estimation (AMOS; V17) to test a model that included three latent variables (causal-globality, causal-stability, and causal-locus), each with six observed variables. All latent variables were allowed to correlate.
Model fit was assessed by chi-square, the Goodness-of-Fit Index (GFI) and the Comparative-Fit-Index (CFI). Cut-off values of >.90 were considered acceptable (Byrne, 2001). The Root Mean Square of Approximation (RMSEA) index was also used, with cut-off values <.05 considered excellent (Byrne, 2001). The fit statistics for the initial model were v2(126, N = 396) = 185.27, p < .001, GFI = .95, CFI = .96, RMSEA = .04. This indicated an acceptable fit, although the standardised regression weights indicated that four observed variables had non-significant factor loadings <.3. These were removed. Subsequent examination resulted in an improved fit, v2(72, N = 396) = 114.60, p < .001, GFI = .96, CFI = .97, RMSEA = .04. So that each factor had an equal number of items (four), the items with the lowest eigenvalue in the causal-globality and causal-stability subscales were removed, resulting in a final fit of v2(50, N = 396) = 68.6, p < .05, GFI = .97, CFI = .98, RMSEA = .03. Standardised regression weights ranged from .50 to .82 (causal-stability), .39 to .72 (causal-locus) and .43 to .79 (causal-globality). The inter-correlations among the latent variables were nonsignificant to weak indicating that the latent variables are largely independent of one another, adding support for the construct validity of the new scale. The lack of latent variable inter-correlation also excluded a 2nd order CFA analysis to investigate whether all three latent variables can be represented as a single, higher level factor (Byrne, 2001). Summary statistics and internal consistency (Cronbach’s alpha and split-half) for each subscale and the overall scale are reported in Table 2. The internal reliability coefficients remain within acceptable levels for four item subscales.
4
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
Table 1 Principal axis factor estimates (rotated) and summary statistics (N = 320). Factor
1
Negative causal-stability You fail an exam You do not do as well in your exams as expected You save all year for a good holiday but when the time comes, you find you did not save enough and you cannot go You forget to pick up your partner (boyfriend/girlfriend) from work You go out on a date and it goes badly You unintentionally hurt someone’s feelings Negative causal-globality You get retrenched from your job You get a job but find you do not get along with your supervisor/boss You have been looking for a job unsuccessfully for some time You start your own business but it does not take off and you have to liquidate You are successful in getting a position you want but find that your colleagues are hard to work with The company you work for goes through a restructure, and your position is downgraded Negative causal-locus You break a diet Despite wanting to stick to it, you break your new year’s resolution You finally get to university and you fail your first semester You try your hand at a craft project (woodwork, folk art) and it does not turn out so well You just cannot seem to learn how to use the computer You meet a potential boyfriend/girlfriend and you burp Eigenvalues (unrotated) Percentage of variance explained (unrotated; rotated) Subscale summary Statistics (M, SD) Internal reliability coefficient (Cronbach alpha)
2
3
ICE
.73 .62 .61 .58 .56 .54
.08 .22 .20 .21 .03 .01
.08 .11 .06 .02 .14 .06
.47 .39 .40 .39 .34 .35
.09 .07 .03 .15 .15 .10
.71 .62 .58 .58 .56 .52
.07 .08 .16 .11 .18 .21
.47 .41 .36 .39 .40 .33
.04 .00 .03 .13 .05 .14
.03 .00 .12 .09 .02 .03
.70 .68 .60 .56 .55 .44
.42 .39 .39 .34 .32 .21
3.7 17.5; 13.1 24.2 (6.4) .80
2.9 12.7; 12.9 26.3 (6.6) .77
2.1 8.5; 12.8 33.6 (5.3) .78
Note: For total scale, % variance explained = 38.7; M = 84.1 (SD = 11.3); a = .75; ICE = Initial Communality Estimates.
3. Phase 2 3.1. Materials and methods 3.1.1. Participants Two hundred and seventy-three participants volunteered for Phase 2. One hundred and thirteen (41.4%) were first-year university students (68 female) and 160 (58.6%) were community members (115 female). Age range was 16–51 (M = 20.7, SD = 5.5).
3.1.2. Materials Participants completed five scales. Firstly, the 12-item explanatory style scale devised from Phase 1 was administered with the Expanded Attributional Style Questionnaire – short form (EASQS; Whitley, 1991) to establish initial concurrent validity. The EASQ-S has reported acceptable concurrent validity with the Expanded Attributional Style Questionnaire (EASQ; Peterson & Villanova, 1988), acceptable alpha coefficients for its causal-globality (.75) and causal-stability (.76) subscales, but an unacceptable alpha coefficient for its causal-locus subscale (.47) (Peterson, 1991). To test initial incremental and convergent validity with depressive symptomatology and ill-health, the General Health Questionniare-12 item (GHQ-12; Goldberg & Williams, 1988) and the Short Form 12-item Health Survey (SF-12; Ware, Kosinski, & Keller, 1996) were administered. The GHQ-12 has good internal reliability
Table 2 Summary statistics and internal reliability coefficients for subscales and total score (N = 396). Subscale Causal-globality Causal-stability Causal-locus Total Score
M 19.20 15.30 23.50 57.90
SD 4.30 5.20 3.30 8.30
Alpha .70 .78 .68 .69
Split-half .68 .75 .73 .73
2
3 [⁄⁄]
.26
(a = .89; Hardy, Shapiro, Haynes, & Rick, 1999) and good convergent validity (Hardy et al., 1999). The SF-12 reported acceptable alpha coefficients with this sample (a = .71) and has demonstrated acceptable convergent validity with a single item measure of overall positive wellbeing (r = .29; Luo et al., 2003). To establish initial divergent validity against neuroticism, the Short Eysenck Personality Questionnaire Revised – Neuroticism Scale (Short EPQR-N; Eysenck & Eysenck, 1991) was administered. The Short EPQR-N has a strong alpha coefficient of .82, moderate test–retest reliability (r = .59; Birley et al., 2006) and strong concurrent validity with the long version of the neuroticism scale (rho = 0.96; Birley et al., 2006). 3.1.3. Procedure The scale administration and storage procedures adopted in Section 2.1.2 were also adopted here. 3.2. Results 3.2.1. Incremental validity The incremental validity of the new scale as an improved measure, compared to an existing scale (EASQ-S), was assessed using hierarchical regression analysis as per recommendations by Hunsley and Meyer (2003). The criterion variable was depressive symptomatology (GHQ-12). At Step 1, gender accounted for 2% of the variance in depressive symptomatology (see Table 3). At Step 2, the EASQ-S total and subscale scores accounted for a further significant 7% unique variance, and at Step 3, the new scale (total and subscale scores) accounted for a further non-significant 1% of the variance in depressive symptomatology. Thus, the new scale does not have incremental validity in this sample.
4 .09 .05
.72** .74** .41**
Note: Split-half reliabilities are adjusted using Spearman–Brown Prophesy Formula. ** p < .01.
3.2.2. Construct validity Concurrent validity with the EASQ-S indicated moderate associations with each subscale except causal-locus (causal-globality, r = .34, p < .05; causal-stability, r = .46, p < .01; and total scale, r = .48, p < .01). Convergent validity was partially demonstrated for each subscale except causal-locus with the GHQ-12
5
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6 Table 3 Hierarchical regression results for the incremental validity analysis of the new explanatory style scale. Criterion = GHQ-12 (depressive symptomatology) (N = 273). R2 for model
Predictors included 1 2 3
Gender Gender, EASQ-S, EASQ-S(glob), EASQ-S(stab), EASQ-S(loc) Gender, EASQ-S, EASQ-S(glob), EASQ-S(stab), EASQ-S(loc) NS (total), NS (glob), NS (stab), NS (loc)
F for model *
R2 change
F for R2 change
.02 .09
F(1,228) = 4.83 F(5,224) = 4.32**
.02 .07
F(1,228) = 4.83* F(4,224) = 4.12**
.10
F(8,221) = 3.05**
.01
F(3,221) = .94
Note: NS = new scale; glob = causal-globality; stab = causal-stability; loc = causal-locus. * p < .05. ** p < .01
(causal-globality, r = .15, p < .05; causal-stability, r = .13, p < .05; and total scale, r = .13, p < .05) and the SF-12 (causal-globality, r = .27, p < .01; causal-stability, r = .22, p < .01; and total scale, r = .23, p < .01). Divergent validity with the short EPQR-N was partially demonstrated for each subscale except causal-globality (causal-stability, r = .02, p = .84; causal-locus, r = .13, p = .16; and total scale, r = .06, p = .54). See Table 4 for a summary of results. Since the causal-locus subscale did not directly correlate with the GHQ-12 or SF-12, we tested its indirect associations using moderation analysis as recommended by Baron and Kenny (1986). Causal-locus as a predictor variable was entered at Step 1. Causal-globality and causal-stability as moderator variables, was entered at Step 2, and the interaction terms were entered at Step 3. None of the interaction terms were significant, indicating that causal-locus is not associated with depression or ill-health, either directly or indirectly.
4. Discussion The current study set out to devise a new explanatory style scale which would overcome problems associated with earlier scales, such as the ASQ, by creating new causal-locus items reflecting the life-space theory (Lewin, 1936), and redesigning the item format in order to minimise correlated errors due to the testlets. In this process, we tested the initial structural validity and construct validity of the new scale in a non-clinical sample. In Phase 1, the new items were tested for content validity and initial structural validity using EFA and CFA procedures. This phase was successful in devising sufficient content valid causal-locus items, as well as sufficient stems for each item, thereby eliminating potential covariance reported to be due to the ‘‘testlet’’ format. This phase was also successful in demonstrating initial support for the structural validity for the new scale using EFA procedures, and confirming the initial structural validity using CFA procedures. Finally, this phase was successful in demonstrating the initial internal reliability (Cronbach’s alpha and split-half reliability) of the new scale. In Phase 2, the initial partial construct validity for the new scale was demonstrated, by testing the new scale’s associations with a similar scale (EASQ-S), depressive symptomatology (GHQ-12), ill-health (SF-12), and with neuroticism (short EPQR-N). Finally, Table 4 Correlation coefficients of the new explanatory style scale with the EASQ-S, GHQ-12, SF-12, and short EPQR-N (N = 273). Subscale Causal-globality Causal-stability Causal-locus Total Score * **
p < .05. p < .01.
EASQ-S *
.34 .46** .00 .48**
GHQ-12 *
.15 .13* .08 .13*
SF-12 **
.27 .22** .02 .23**
Short EPQR-N .22* .02 .13 .06
incremental validity results indicated that the new scale does not account for any further variance in depressive symptomatology compared to existing measures of explanatory style. This is the first study that has attempted to address the structural validity of a measure of explanatory style by reformatting the ASQ in this way, and is thus the first study that has been successful in achieving acceptable structural validity for a measure of explanatory style using EFA techniques, and confirming the structure with CFA techniques. Further, the initial partial concurrent, convergent, and divergent validity results found in Phase 2 replicates previous research into the construct validity of explanatory style with other outcome variables, in particular generally finding no significant direct or indirect associations between the causallocus dimension and outcome variables (Luten et al., 1997; Michelson et al., 1997; Volpe & Levin, 1998). Finally, this is the first study to report incremental validity for a measure of explanatory style. The inability to demonstrate construct validity for the causallocus dimension, despite improved internal reliability for this dimension, is interpreted as evidence that the causal-locus dimension is problematic at the conceptual, rather than methodological, level. That is, the current results are interpreted as support for alternative theories of the locus-causal dimension, including Weiner’s (2014) assertion that the causal-locus dimension cannot be considered independent to the controllability dimension, and Malle, Knobe, O’Laughlin, Pearce, and Nelson’s (2000) folk-conceptual framework. An alternative interpretation is that the current results add support to the growing argument that the locus-causal dimension, as it is currently conceptualised, should not be included in attributional theories as asserted by Gilbert et al. (2000) and Abramson et al. (1989). Since the structural, content, and construct validity of the new scale has been partially established, and improved alpha coefficients have been shown for all dimensions of the new scale, future research needs to: (a) establish external reliability of the new scale (test–retest reliability), and (b) examine the utility of interpreting results at the dimensional level only. Further, research aiming to find an improved measure of explanatory style would benefit from: (c) considering alternative theories of the causal-locus dimension other than the internal/external dichotomy as is currently theorised in RLH theory such as Malle et al. (2000) and Weiner (2014), and (d) examine whether the incremental validity of the new scale can be achieved when the causal-locus dimension is conceptualised using an alternative theory of causal-locus. Finally, investigations examining the relationship between explanatory style and outcomes could also consider attributional theories that do not include the internal/external conceptualisation of the causal locus dimension, such as the hopelessness theory of depression or Gilbert et al.’s (2000) psychological immune system theory. Overall, the present study was successful in highlighting that previous research examining how causal explanations influence outcomes such as depressive symptomatology and ill health have
6
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
been problematic and limited due to a lack of parsimony across findings. Further, the present study successfully demonstrated that a major reason for the lack of parsimonious results in these investigations is due to the causal-locus dimension being ill-defined in the early stages of its conceptualisation. It is believed that, with the guidance of the recommendations suggested in this paper, our understanding of how social cognition influences psychological outcomes will be significantly increased.
References Abramson, L. Y., Metalsky, G. I., & Alloy, L. B. (1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96, 358–372. doi: 0033-295X/89/$00.75. Abramson, L. Y., Seligman, M. E., & Teasdale, J. D. (1978). Learned helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87, 49–74. http://dx.doi.org/10.1037//0021-843x.87.1.49. Asner-Self, K. K., & Schreiber, J. B. (2004). A factor analytic study of the attributional style questionnaire with Central American immigrants. Measurement and Evaluation in Counseling and Development, 37, 144–152. Retrieved from http:// www.highbeam.com/doc/1G1-125640880.html. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173. Bennett, K. K., & Elliott, M. (2002). Explanatory style and health: mechanisms linking pessimism to illness. Journal of Applied Social Psychology, 32, 1508–1526. http://dx.doi.org/10.1111/j.1559-1816.2002.tb01449.x. Birley, A. J., Gillespie, N. A., Heath, A. C., Sullivan, P. F., Boomsma, D. I., & Martin, N. G. (2006). Heritability and nineteen-year stability of long and short EPQ-R neuroticism scales. Personality and Individual Differences, 40, 737–747. http:// dx.doi.org/10.1016/j.paid.2005.09.005. Burnes, B., & Cooke, B. (2013). Kurt Lewin’s Field Theory: A review and reevaluation. International Journal of Management Reviews, 15, 408–425. http:// dx.doi.org/10.1111/j.1468-2370.2012.00348.x. Byrne, B. M. (2001). Structural equation modelling with AMOS: Basic concepts, applications, and programming. New Jersey: Lawrence Erlbaum Assoc. Corr, P. J., & Gray, J. A. (1996). Structure and validity of the Attributional Style Questionnaire: A cross-sample comparison. Journal of Psychology, 130, 645–658. http://dx.doi.org/10.1080/00223980.1996.9915038. de Winter, J. C. F., & Dodou, D. (2012). Factor recovery by principal axis factoring and maximum likelihood factor analysis as a function of factor pattern and sample size. Journal of Applied Statistics, 39, 695–710. http://dx.doi.org/10.1080/ 02664763.2011.610445. Dua, J. (1995). Retrospective and prospective psychological and physical health as a function of negative affect and attributional style. Journal of Clinical Psychology, 51, 507–518. doi: 10.1002/1097-4679(199507)51:4<507::AIDJCLP2270510407>3.0.CO;2-P. Dykema, J., Bergbower, K., & Peterson, C. (1995). Pessimistic explanatory style, stress, and illness. Journal of Social and Clinical Psychology, 14, 357–371. http:// dx.doi.org/10.1521/jscp.1995.14.4.357. Eysenck, H. J., & Eysenck, S. B. G. (1991). Manual of the eysenck personality scales (EPS adult). London: Hodder & Staughton. Furnham, A., Sadka, V., & Brewin, C. R. (1992). The development of an occupational Attributional Style Questionnaire. Journal of Organisational Behaviour, 13, 27–39. http://dx.doi.org/10.1002/job.4030130104. Gibb, B. E., Zhu, L., Alloy, L. B., & Abramson, L. Y. (2002). Attributional styles and academic achievement in university students: A longitudinal investigation. Cognitive Therapy and Research, 26, 309–315. http://dx.doi.org/10.1023/ A:1016072810255. Gilbert, D. T., Pinel, E. C., Brown, R. P., & Wilson, T. D. (2000). The illusion of external agency. Journal of Personality and Social Psychology, 79, 690–700. http:// dx.doi.org/10.1037//0022-3514.79.5.690. Goldberg, D., & Williams, P. (1988). A users guide to the General Health Questionnaire. Windsor, Berks: NFER-Nelson. Greenbaum, T. L. (1993). The handbook for focus group research. New York, NY: Macmillan. Hardy, G. E., Shapiro, D. A., Haynes, C. E., & Rick, J. E. (1999). Validation of the General Health Questionnaire-12 using a sample of employees from England’s Health Care Services. Psychological Assessment, 11, 159–165. http://dx.doi.org/ 10.1037//1040-3590.11.2.159. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Higgins, N. C., Zumbo, B. D., & Hay, J. L. (1999). Construct validity of attributional style: Modeling context-dependent item sets in the Attributional Style Questionnaire. Educational and Psychological Measurement, 59, 804–820. http://dx.doi.org/10.2224/sbp.1999.27.3.221. Hinkin, T. R., Tracey, J. B., & Enz, C. A. (1997). Scale construction: Developing reliable and valid measurement instruments. The Council on Hotel, Restaurant, and Institutional Education, 21, 100–120. http://dx.doi.org/10.1177/ 109634809702100108.
Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185. http://dx.doi.org/10.1007/BF02289447. Hunsley, J., & Meyer, G. J. (2003). The incremental validity of psychological testing and assessment: Conceptual, methodological, and statistical issues. Psychological Assessment, 15, 446–455. http://dx.doi.org/10.1037/10403590.15.4.446. Jeavons, S., & Greenwood, K. M. (2007). World assumptions, coping, and attributions. Personality and Individual Differences, 42, 1427–1437. http:// dx.doi.org/10.1016/j.paid.2006.11.010. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.). Nebraska symposium on motivation (Vol. 15, pp. 192–240). Lincoln: University of Nebraska Press. Kleim, B., Gonzalo, D., & Ehlers, A. (2011). The Depressive Attributions Questionnaire (DAQ): Development of a short self-report measure of depressogenic attributions. Journal of Psychopathology and Behavioral Assessment, 33, 375–385. http://dx.doi.org/10.1007/s10862-011-9234-9. Lewin, K. (1936). Principals of Topological Psychology. New York, NY: McGraw Hill. Luo, X., George, M. L., Kakauras, I., Edwards, C. L., Pietrobon, R., Richardson, W., et al. (2003). Reliability, validity, and responsiveness of the short form 12-item survey (SF-12) in patients with back pain. Spine, 28, 1739–1745. http:// dx.doi.org/10.1097/01.BRS.0000083169.5867.96. Luten, A. G., Ralph, J. A., & Mineka, S. (1997). Pessimistic attributional style: Is it specific to depression versus anxiety versus negative affect. Behaviour Research and Therapy, 35, 703–719. http://dx.doi.org/10.1016/s0005-7967(97)00027-2. Malle, B. (2011). Time to give up the dogmas of attribution: An alternative theory of behaviour explanation. In M. Zanna & J. Olsen (Eds.). Advances in Experimental Social Psychology (Vol. 44, pp. 297–352). San Francisco, CA: Elsevier Inc. Malle, B. F., Knobe, J., O’Laughlin, M. J., Pearce, G. E., & Nelson, S. E. (2000). Conceptual structure and social functions of behaviour explanations: Beyond person–situation attributions. Journal of Personality and Social Psychology, 79, 309–326. http://dx.doi.org/10.1037/0022-3514.79.3.309. Michelson, L. K., Bellanti, S., Testa, M., & Marchione, N. (1997). The relationship of attributional style to agoraphobia severity, depression, and treatment outcome. Behaviour Research and Therapy, 35, 1061–1073. http://dx.doi.org/10.1016/ s0005-7967(97)10011-0. Peterson, C. (1991). On shortening the expanded Attributional Style Questionnaire. Journal of Personality Assessment, 56, 179–183. http://dx.doi.org/10.1207/ s15327752jpa5601_16. Peterson, C., Semmel, A., von Baeyer, C., Abramson, L., Metalsky, G., & Seligman, M. (1982). The Attributional Style Questionnaire. Cognitive Therapy and Research, 6, 287–300. http://dx.doi.org/10.1007/BF01173577. Peterson, C., & Vaidya, R. S. (2001). Explanatory style, expectations, and depressive symptoms. Personality and Individual Differences, 31, 1217–1223. http:// dx.doi.org/10.1016/s0191-8869(00)00043-x. Peterson, C., & Villanova, P. (1988). An expanded Attributional Style Questionnaire. Journal of Abnormal Psychology, 97, 87–89. http://dx.doi.org/10.1037//0021843x.97.1.87. Proudfoot, J. G., Corr, P. J., Guest, D. E., & Gray, J. A. (2001). The development and evaluation of a scale to measure occupational attributional style in the financial services sector. Personality and Individual Differences, 30, 259–270. http:// dx.doi.org/10.1016/s0191-8869(00)00043-x. Smith, P., Caputi, P., & Crittenden, N. (2013). Measuring optimism in organisations: Development of a Workplace Explanatory Styles Questionnaire. Journal of Happiness Studies, 14, 415–432. http://dx.doi.org/10.1007/s10902-012-9336-4. Sturman, E., Mongrain, M., & Kohn, P. (2006). Attributional style as a predictor of hopelessness depression. Journal of Cognitive Psychotherapy: An International Quarterly, 20, 447–458. http://dx.doi.org/10.1891/jcpiq-v2014a008. Velicer, W. F., Easton, C. A., & Fava, J. L. (2000). Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In R. D. Goffin & E. Helmes (Eds.), Problems and solutions in human assessment: Honoring Douglas N. Jackson at Seventy (pp. 41–71). Boston: Kluwer. Volpe, N., & Levin, R. (1998). Attributional style, dreaming and depression. Personality and Individual Differences, 25, 1051–1061. http://dx.doi.org/ 10.1016/s0191-8869(98)00073-7. Ware, J. E., Kosinski, M. M., & Keller, S. D. (1996). A 12-item Short-Form Health Survey: Construction of scales and preliminary tests of reliability and validity. Medical Care, 34, 220–233. http://dx.doi.org/10.1097/00005650-19960300000003. Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer Verlag. Weiner, B. (2014). The attribution approach to emotion and motivation: History, hypotheses, home runs, headaches/heartaches. Emotion Review, 6, 353–361. http://dx.doi.org/10.1177/1754073914534502. Whitley, B. E. Jr., (1991). A short form of the expanded Attributional Style Questionnaire. Journal of Personality Assessment, 56, 365–369. http:// dx.doi.org/10.1207/s15327752jpa5602_14. Worthington, R. L., & Whittaker, T. A. (2006). Scale development research: A content analysis and recommendations for best practices. The Counselling Psychologist, 34, 806–838. http://dx.doi.org/10.1177/0011000006288127. Xenikou, A., Furnham, A., & McCarrey, M. (1997). Attributional style for negative events: A proposition for a more reliable and valid measure of attributional style. British Journal of Psychology, 88, 53. http://dx.doi.org/10.1111/j.20448295.1997.tb02620.x.
Contents lists available at ScienceDirect
Personality and Individual Differences journal homepage: www.elsevier.com/locate/paid
The development and initial validation of a new scale to measure explanatory style q Katrina M. Travers a,⇑, Peter A. Creed b, Shirley Morrissey b a b
Australian Institute of Psychology, Brisbane, Queensland, Australia School of Applied Psychology and Griffith Health Institute, Griffith University, Australia
a r t i c l e
i n f o
Article history: Received 24 April 2014 Received in revised form 24 January 2015 Accepted 27 January 2015 Available online 12 February 2015 Keywords: Explanatory style Reformulated learned helplessness theory Attributional Style Questionnaire New explanatory style scale
a b s t r a c t The reformulated learned helplessness (RLH) theory and its associated construct of explanatory style have been tested extensively using the Attributional Style Questionnaire (ASQ: Peterson et al., 1982) against outcomes such as depression. However, support for the RLH theory is best described as inconclusive. This is because: (a) the causal-locus dimension is poorly defined and the causal-locus items have poor reliability and validity; (b) the ASQ has not been demonstrated to have structural validity; and (c) the definitions of explanatory style are inconsistent across studies. The current study was conducted with the goal of developing a valid and reliable measure of explanatory style. The new measure met simple structure when tested using exploratory and confirmatory factor analyses, and each short subscale had acceptable internal reliability. Further, construct validity was partially demonstrated, but incremental validity was not supported. Implications for theory and practice are discussed. Ó 2015 Elsevier Ltd. All rights reserved.
1. Introduction The reformulated learned helplessness (RLH) theory (Abramson, Seligman, & Teasdale, 1978) contends that negative psychological consequences (e.g., depression) occur when a person is pessimistic in their explanation of unexpected negative events. This pessimism has been defined as an explanatory style that is generalising (global), persistent (stable), and personal (locus) (Abramson et al., 1978). Peterson et al. (1982) developed a measure of explanatory style (the Attributional Style Questionnaire; ASQ), to investigate this theory. The ASQ, however, has returned inconclusive results in construct validity investigations, meaning that the validity of the RLH theory has been poorly supported. Reasons include: correlations between the ASQ subscales and outcomes such as depression have been non-significant or weak, particularly with the causallocus subscale (Bennett & Elliott, 2002; Dua, 1995; Gibb, Zhu, Alloy, & Abramson, 2002; Luten, Ralph, & Mineka, 1997; Michelson, Bellanti, Testa, & Marchione, 1997; Volpe & Levin, 1998); structural validity of the ASQ has not delineated a parsimonious model-of-fit (Furnham, Sadka, & Brewin, 1992;
q
This article is a Special issue article – ‘‘Young researcher award 2014’’.
⇑ Corresponding author at: Australian Institute of Psychology, Locked Bag 15, Fortitude Valley, Queensland 4006, Australia. Tel.: +61 731122000; fax: +61 732577195. E-mail address: [email protected] (K.M. Travers). http://dx.doi.org/10.1016/j.paid.2015.01.045 0191-8869/Ó 2015 Elsevier Ltd. All rights reserved.
Kleim, Gonzalo, & Ehlers, 2011; Proudfoot, Corr, Guest, & Gray, 2001; Smith, Caputi, & Crittenden, 2013; Xenikou, Furnham, & McCarrey, 1997); and research has been inconsistent in the way that explanatory style has been defined. For example, some studies have defined explanatory style with causal-stability and causalglobality only (Dykema, Bergbower, & Peterson, 1995; Peterson & Vaidya, 2001), or only causal-globality (Bennett & Elliott, 2002). Thus, Abramson, Metalsky, and Alloy (1989) proposed an alternative theory to the RLH theory, named the hopelessness theory of depression, purporting that it was causal-globality and causal-stability that predisposed a person to hopelessness, and thus hopelessness depression (HD). Hopelessness theory has been supported in the literature. For example, Sturman, Mongrain, and Kohn (2006) found a significant correlation between generality (a composite of causal-stability and causal-globality) and a composite of five HD symptoms (r = .24; p = .02); whereas, generality was uncorrelated with non-HD symptoms (r = .03, p = .76). Although Abramson et al.’s (1989) solution to revise the RLH theory to the hopelessness theory has received some support, the mixed results summarised for the RLH theory could also be due to the problematic causal-locus dimension in the ASQ. Thus, a revised causal-locus subscale is required so that the RLH model can be assessed along the three original explanatory style dimensions. The goal of the current study was to develop an alternative measure of explanatory style that incorporated new causal-locus items, which would allow the RLH theory to be measured consistently, and according to its definition.
2
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
1.1. The causal-locus dimension items The causal-locus dimension of explanatory style was originally conceived by Lewin (1936), who was interested in understanding social behaviour from a mathematical perspective (Burnes & Cooke, 2013). Lewin articulated the life-space theory which theorised that a person’s behaviour (B) is a function of the interaction between the person (p) and the environment (situation) (e), or B = f(p,e). This person/situation distinction in life-space theory was influential in conceptualising the personal/impersonal heuristic in Heider’s (1958) attributional theory of social perception. Heider theorised that lay people rely on this heuristic when explaining intentional social behaviour as opposed to unintentional actions (e.g., yawning) (Malle, 2011). According to Malle (2011), Heider’s (1958) personal/impersonal heuristic was misrepresented in subsequent attributional theories including those by Kelley (1967) and Weiner (1986) to refer to the causal-locus dichotomy (internal/external), and it was this misrepresentation that persisted in later attributional theories including the RLH theory by Abramson et al. (1978). This interpretation of the person/situation dichotomy continues to be debated. Weiner (2014), for instance, argued that the causal-locus dimension cannot be considered independent to a controllability dimension. Malle, Knobe, O’Laughlin, Pearce, and Nelson (2000) theorised that the causal-locus dimension is better understood within a folk-conceptual framework. Briefly, the folk-conceptual framework purports that people explain intentional (controllable) behaviour using one of three modes of explanation (causal history of reasons, enabling factors, or reason explanations including desire, belief or value). Finally, Gilbert, Pinel, Brown, and Wilson (2000) questioned the validity of a causal-locus dichotomy, arguing that persons have an unconscious ‘‘psychological immune system’’ (PIS) which naturally externalises causes of sub-optimal events in order to increase self-satisfaction. In this study, we refer back to the original life-space theory by Lewin (1936) and propose re-writing the ASQ causal-locus items so that this dimension is measured by items reflecting each of the three life-space domains (behaviour, person, and environment) in individual items, with each item anchored by internal and external descriptions of the Lewinian domain being measured. This would achieve a more consistent, comprehensive, and interpretable causal-locus dimension within the construct of explanatory style. 1.2. Structural validity of the ASQ Previous attempts to find structural validity of the ASQ using exploratory factor analysis (EFA) have been unsuccessful. Kleim et al. (2011) found one factor, Corr and Gray (1996), Higgins, Zumbo, and Hay (1999), and Proudfoot et al. (2001) found two factors, and Xenikou et al. (1997) found three (interpreted as two positive and one negative factor). Higgins et al. (1999) argued that these previous attempts have been unsuccessful due to the inbuilt ‘‘testlets’’ or common item-stems which generate co-variation among items of the same stem if not allowed to correlate when conducting an EFA. Higgins et al. (1999) suggested that error terms in confirmatory factor analysis be allowed to correlate in order to address this. We addressed this issue by reformatting the ASQ so that each item-stem is answered by only one item, thereby eliminating the co-variation due to the testlets. We also included only the negative item-stems. 1.3. Objectives Thus, this investigation was conducted to achieve four aims: (a) to generate new causal-locus items reflecting the underlying life-
space theory, (b) to create a unique item-stem for each item, (c) to establish initial internal reliability and structural validity for the new scale, and (d) to establish initial construct and incremental validity of the scale. To achieve these aims, a two-phase study was completed. Phase 1 addressed the first three aims, and Phase 2 addressed the final aim. 2. Phase 1 2.1. Materials and methods 2.1.1. Participants Three participant samples volunteered for Phase 1. Sample 1 included six academic psychologists with doctorate level training (five females; mean age = 35.5 years, SD = 11.1 years). Sample 2 included 320 participants: 248 participants (77.5%) were first-year psychology students (189 females), and 72 (22.5%) were community members (49 females). The age range was 18–59 years (M = 25.82, SD = 10.25). Sample 3 consisted of 409 participants: 258 (63.1%) were first-year university students (181 female) and 151 (36.9%) were community members (89 females). The age range was 16–76 years (M = 26.43, SD = 12.25). 2.1.2. Procedure Thirty new items (six each for causal-global, causal-stable, and the three life-space domains) were generated deductively. These new items were then added to the 12 original negative causal-global and causal-stable items, resulting in 42 items, allowing for redundant items during item and factor analysis (Hinkin, Tracey, & Enz, 1997). Further, 36 new negative item-stems were generated and added to the existing six negative item-stems, thereby allowing each item to be preceded with a unique item-stem. These itemstems were randomly allocated to each dimension (including the three life-space domains). The original 7-point Likert-type response was retained (Hinkin et al., 1997). Finally, all items were randomly ordered throughout the scale to avoid explanatory style dimensions being measured by consecutive items. Focus-group methodology (Greenbaum, 1993) was used to assess content validity of the new items and stems, using Sample 1 participants. See Fig. 1 for sample items of each dimension, including each of the three Lewinian domains for the causal-locus dimension. Sample 2 participants were then administered the 42-item scale for item analysis and EFA procedures. The scale was distributed to participants in class who were given the option of completing it immediately, or at a later time. Community respondents were distributed surveys using snowballing procedures. The scale resultant from item analysis and EFA procedures was then administered to Sample 3 participants using identical distribution procedures, in order to confirm the initial structural validity results using confirmatory factor analysis (CFA) procedures. All participants were treated in accordance with the university’s Human Research Ethics Committee guidelines. 2.2. Results 2.2.1. Sample 2 item analysis Item distributions, inter-item, and item-total correlations were used to assess for problematic items. Items with poor distributions or correlations at r 6 .20 were removed. This resulted in the deletion of 12 items. 2.2.2. Sample 2 exploratory factor analysis Firstly, the remaining 30 items (ten each for causal-globality, causal-stability, and causal-locus), were analysed using Horn’s (1965) parallel-analysis to determine the correct number of factors
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
3
Fig. 1. Sample items for causal-globality, causal-stability, and causal-locus (represented by the three Lewinian domains of personality, behaviour, and environment) for the new explanatory style scale.
to be retained (Velicer, Easton, & Fava, 2000). Results indicated that three factors were to be extracted. The items were then analysed using principal axis factor analysis (PAF) with varimax rotation (de Winter & Dodou, 2012; Worthington & Whittaker, 2006). An orthogonal rotation was selected since the explanatory style dimensions have returned low inter-correlations in previous investigations (Asner-Self & Schreiber, 2004; Jeavons & Greenwood, 2007; Peterson & Villanova, 1988; Peterson et al., 1982). The KMO measure was adequate (.75) and the Bartlett’s Test of Sphericity was significant (<.001). Items were retained if they returned factor loadings P.30 with no cross loadings, resulting in the deletion of four causal-locus, three causal-stability, and one causalglobality item. A second analysis produced three factors, representing causal-globality (nine items), causal-stability (seven items) and causal-locus (six items). To ensure an equal number of items for each factor, the six causal-globality and causal-stability items with the highest factor loadings were retained, resulting in a final 18 item solution, meeting simple structure, and accounting for 38.7% of the variance. The three factors were interpreted as ‘‘causal-globality’’, ‘‘causal-stability’’, and ‘‘causal-locus’’. The 18 items, their factor loadings and summary statistics are reported in Table 1. 2.2.3. Sample 3 confirmatory factor analysis Thirteen cases were deleted due to missing data, leaving 396 cases for analysis. We used maximum likelihood estimation (AMOS; V17) to test a model that included three latent variables (causal-globality, causal-stability, and causal-locus), each with six observed variables. All latent variables were allowed to correlate.
Model fit was assessed by chi-square, the Goodness-of-Fit Index (GFI) and the Comparative-Fit-Index (CFI). Cut-off values of >.90 were considered acceptable (Byrne, 2001). The Root Mean Square of Approximation (RMSEA) index was also used, with cut-off values <.05 considered excellent (Byrne, 2001). The fit statistics for the initial model were v2(126, N = 396) = 185.27, p < .001, GFI = .95, CFI = .96, RMSEA = .04. This indicated an acceptable fit, although the standardised regression weights indicated that four observed variables had non-significant factor loadings <.3. These were removed. Subsequent examination resulted in an improved fit, v2(72, N = 396) = 114.60, p < .001, GFI = .96, CFI = .97, RMSEA = .04. So that each factor had an equal number of items (four), the items with the lowest eigenvalue in the causal-globality and causal-stability subscales were removed, resulting in a final fit of v2(50, N = 396) = 68.6, p < .05, GFI = .97, CFI = .98, RMSEA = .03. Standardised regression weights ranged from .50 to .82 (causal-stability), .39 to .72 (causal-locus) and .43 to .79 (causal-globality). The inter-correlations among the latent variables were nonsignificant to weak indicating that the latent variables are largely independent of one another, adding support for the construct validity of the new scale. The lack of latent variable inter-correlation also excluded a 2nd order CFA analysis to investigate whether all three latent variables can be represented as a single, higher level factor (Byrne, 2001). Summary statistics and internal consistency (Cronbach’s alpha and split-half) for each subscale and the overall scale are reported in Table 2. The internal reliability coefficients remain within acceptable levels for four item subscales.
4
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
Table 1 Principal axis factor estimates (rotated) and summary statistics (N = 320). Factor
1
Negative causal-stability You fail an exam You do not do as well in your exams as expected You save all year for a good holiday but when the time comes, you find you did not save enough and you cannot go You forget to pick up your partner (boyfriend/girlfriend) from work You go out on a date and it goes badly You unintentionally hurt someone’s feelings Negative causal-globality You get retrenched from your job You get a job but find you do not get along with your supervisor/boss You have been looking for a job unsuccessfully for some time You start your own business but it does not take off and you have to liquidate You are successful in getting a position you want but find that your colleagues are hard to work with The company you work for goes through a restructure, and your position is downgraded Negative causal-locus You break a diet Despite wanting to stick to it, you break your new year’s resolution You finally get to university and you fail your first semester You try your hand at a craft project (woodwork, folk art) and it does not turn out so well You just cannot seem to learn how to use the computer You meet a potential boyfriend/girlfriend and you burp Eigenvalues (unrotated) Percentage of variance explained (unrotated; rotated) Subscale summary Statistics (M, SD) Internal reliability coefficient (Cronbach alpha)
2
3
ICE
.73 .62 .61 .58 .56 .54
.08 .22 .20 .21 .03 .01
.08 .11 .06 .02 .14 .06
.47 .39 .40 .39 .34 .35
.09 .07 .03 .15 .15 .10
.71 .62 .58 .58 .56 .52
.07 .08 .16 .11 .18 .21
.47 .41 .36 .39 .40 .33
.04 .00 .03 .13 .05 .14
.03 .00 .12 .09 .02 .03
.70 .68 .60 .56 .55 .44
.42 .39 .39 .34 .32 .21
3.7 17.5; 13.1 24.2 (6.4) .80
2.9 12.7; 12.9 26.3 (6.6) .77
2.1 8.5; 12.8 33.6 (5.3) .78
Note: For total scale, % variance explained = 38.7; M = 84.1 (SD = 11.3); a = .75; ICE = Initial Communality Estimates.
3. Phase 2 3.1. Materials and methods 3.1.1. Participants Two hundred and seventy-three participants volunteered for Phase 2. One hundred and thirteen (41.4%) were first-year university students (68 female) and 160 (58.6%) were community members (115 female). Age range was 16–51 (M = 20.7, SD = 5.5).
3.1.2. Materials Participants completed five scales. Firstly, the 12-item explanatory style scale devised from Phase 1 was administered with the Expanded Attributional Style Questionnaire – short form (EASQS; Whitley, 1991) to establish initial concurrent validity. The EASQ-S has reported acceptable concurrent validity with the Expanded Attributional Style Questionnaire (EASQ; Peterson & Villanova, 1988), acceptable alpha coefficients for its causal-globality (.75) and causal-stability (.76) subscales, but an unacceptable alpha coefficient for its causal-locus subscale (.47) (Peterson, 1991). To test initial incremental and convergent validity with depressive symptomatology and ill-health, the General Health Questionniare-12 item (GHQ-12; Goldberg & Williams, 1988) and the Short Form 12-item Health Survey (SF-12; Ware, Kosinski, & Keller, 1996) were administered. The GHQ-12 has good internal reliability
Table 2 Summary statistics and internal reliability coefficients for subscales and total score (N = 396). Subscale Causal-globality Causal-stability Causal-locus Total Score
M 19.20 15.30 23.50 57.90
SD 4.30 5.20 3.30 8.30
Alpha .70 .78 .68 .69
Split-half .68 .75 .73 .73
2
3 [⁄⁄]
.26
(a = .89; Hardy, Shapiro, Haynes, & Rick, 1999) and good convergent validity (Hardy et al., 1999). The SF-12 reported acceptable alpha coefficients with this sample (a = .71) and has demonstrated acceptable convergent validity with a single item measure of overall positive wellbeing (r = .29; Luo et al., 2003). To establish initial divergent validity against neuroticism, the Short Eysenck Personality Questionnaire Revised – Neuroticism Scale (Short EPQR-N; Eysenck & Eysenck, 1991) was administered. The Short EPQR-N has a strong alpha coefficient of .82, moderate test–retest reliability (r = .59; Birley et al., 2006) and strong concurrent validity with the long version of the neuroticism scale (rho = 0.96; Birley et al., 2006). 3.1.3. Procedure The scale administration and storage procedures adopted in Section 2.1.2 were also adopted here. 3.2. Results 3.2.1. Incremental validity The incremental validity of the new scale as an improved measure, compared to an existing scale (EASQ-S), was assessed using hierarchical regression analysis as per recommendations by Hunsley and Meyer (2003). The criterion variable was depressive symptomatology (GHQ-12). At Step 1, gender accounted for 2% of the variance in depressive symptomatology (see Table 3). At Step 2, the EASQ-S total and subscale scores accounted for a further significant 7% unique variance, and at Step 3, the new scale (total and subscale scores) accounted for a further non-significant 1% of the variance in depressive symptomatology. Thus, the new scale does not have incremental validity in this sample.
4 .09 .05
.72** .74** .41**
Note: Split-half reliabilities are adjusted using Spearman–Brown Prophesy Formula. ** p < .01.
3.2.2. Construct validity Concurrent validity with the EASQ-S indicated moderate associations with each subscale except causal-locus (causal-globality, r = .34, p < .05; causal-stability, r = .46, p < .01; and total scale, r = .48, p < .01). Convergent validity was partially demonstrated for each subscale except causal-locus with the GHQ-12
5
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6 Table 3 Hierarchical regression results for the incremental validity analysis of the new explanatory style scale. Criterion = GHQ-12 (depressive symptomatology) (N = 273). R2 for model
Predictors included 1 2 3
Gender Gender, EASQ-S, EASQ-S(glob), EASQ-S(stab), EASQ-S(loc) Gender, EASQ-S, EASQ-S(glob), EASQ-S(stab), EASQ-S(loc) NS (total), NS (glob), NS (stab), NS (loc)
F for model *
R2 change
F for R2 change
.02 .09
F(1,228) = 4.83 F(5,224) = 4.32**
.02 .07
F(1,228) = 4.83* F(4,224) = 4.12**
.10
F(8,221) = 3.05**
.01
F(3,221) = .94
Note: NS = new scale; glob = causal-globality; stab = causal-stability; loc = causal-locus. * p < .05. ** p < .01
(causal-globality, r = .15, p < .05; causal-stability, r = .13, p < .05; and total scale, r = .13, p < .05) and the SF-12 (causal-globality, r = .27, p < .01; causal-stability, r = .22, p < .01; and total scale, r = .23, p < .01). Divergent validity with the short EPQR-N was partially demonstrated for each subscale except causal-globality (causal-stability, r = .02, p = .84; causal-locus, r = .13, p = .16; and total scale, r = .06, p = .54). See Table 4 for a summary of results. Since the causal-locus subscale did not directly correlate with the GHQ-12 or SF-12, we tested its indirect associations using moderation analysis as recommended by Baron and Kenny (1986). Causal-locus as a predictor variable was entered at Step 1. Causal-globality and causal-stability as moderator variables, was entered at Step 2, and the interaction terms were entered at Step 3. None of the interaction terms were significant, indicating that causal-locus is not associated with depression or ill-health, either directly or indirectly.
4. Discussion The current study set out to devise a new explanatory style scale which would overcome problems associated with earlier scales, such as the ASQ, by creating new causal-locus items reflecting the life-space theory (Lewin, 1936), and redesigning the item format in order to minimise correlated errors due to the testlets. In this process, we tested the initial structural validity and construct validity of the new scale in a non-clinical sample. In Phase 1, the new items were tested for content validity and initial structural validity using EFA and CFA procedures. This phase was successful in devising sufficient content valid causal-locus items, as well as sufficient stems for each item, thereby eliminating potential covariance reported to be due to the ‘‘testlet’’ format. This phase was also successful in demonstrating initial support for the structural validity for the new scale using EFA procedures, and confirming the initial structural validity using CFA procedures. Finally, this phase was successful in demonstrating the initial internal reliability (Cronbach’s alpha and split-half reliability) of the new scale. In Phase 2, the initial partial construct validity for the new scale was demonstrated, by testing the new scale’s associations with a similar scale (EASQ-S), depressive symptomatology (GHQ-12), ill-health (SF-12), and with neuroticism (short EPQR-N). Finally, Table 4 Correlation coefficients of the new explanatory style scale with the EASQ-S, GHQ-12, SF-12, and short EPQR-N (N = 273). Subscale Causal-globality Causal-stability Causal-locus Total Score * **
p < .05. p < .01.
EASQ-S *
.34 .46** .00 .48**
GHQ-12 *
.15 .13* .08 .13*
SF-12 **
.27 .22** .02 .23**
Short EPQR-N .22* .02 .13 .06
incremental validity results indicated that the new scale does not account for any further variance in depressive symptomatology compared to existing measures of explanatory style. This is the first study that has attempted to address the structural validity of a measure of explanatory style by reformatting the ASQ in this way, and is thus the first study that has been successful in achieving acceptable structural validity for a measure of explanatory style using EFA techniques, and confirming the structure with CFA techniques. Further, the initial partial concurrent, convergent, and divergent validity results found in Phase 2 replicates previous research into the construct validity of explanatory style with other outcome variables, in particular generally finding no significant direct or indirect associations between the causallocus dimension and outcome variables (Luten et al., 1997; Michelson et al., 1997; Volpe & Levin, 1998). Finally, this is the first study to report incremental validity for a measure of explanatory style. The inability to demonstrate construct validity for the causallocus dimension, despite improved internal reliability for this dimension, is interpreted as evidence that the causal-locus dimension is problematic at the conceptual, rather than methodological, level. That is, the current results are interpreted as support for alternative theories of the locus-causal dimension, including Weiner’s (2014) assertion that the causal-locus dimension cannot be considered independent to the controllability dimension, and Malle, Knobe, O’Laughlin, Pearce, and Nelson’s (2000) folk-conceptual framework. An alternative interpretation is that the current results add support to the growing argument that the locus-causal dimension, as it is currently conceptualised, should not be included in attributional theories as asserted by Gilbert et al. (2000) and Abramson et al. (1989). Since the structural, content, and construct validity of the new scale has been partially established, and improved alpha coefficients have been shown for all dimensions of the new scale, future research needs to: (a) establish external reliability of the new scale (test–retest reliability), and (b) examine the utility of interpreting results at the dimensional level only. Further, research aiming to find an improved measure of explanatory style would benefit from: (c) considering alternative theories of the causal-locus dimension other than the internal/external dichotomy as is currently theorised in RLH theory such as Malle et al. (2000) and Weiner (2014), and (d) examine whether the incremental validity of the new scale can be achieved when the causal-locus dimension is conceptualised using an alternative theory of causal-locus. Finally, investigations examining the relationship between explanatory style and outcomes could also consider attributional theories that do not include the internal/external conceptualisation of the causal locus dimension, such as the hopelessness theory of depression or Gilbert et al.’s (2000) psychological immune system theory. Overall, the present study was successful in highlighting that previous research examining how causal explanations influence outcomes such as depressive symptomatology and ill health have
6
K.M. Travers et al. / Personality and Individual Differences 81 (2015) 1–6
been problematic and limited due to a lack of parsimony across findings. Further, the present study successfully demonstrated that a major reason for the lack of parsimonious results in these investigations is due to the causal-locus dimension being ill-defined in the early stages of its conceptualisation. It is believed that, with the guidance of the recommendations suggested in this paper, our understanding of how social cognition influences psychological outcomes will be significantly increased.
References Abramson, L. Y., Metalsky, G. I., & Alloy, L. B. (1989). Hopelessness depression: A theory-based subtype of depression. Psychological Review, 96, 358–372. doi: 0033-295X/89/$00.75. Abramson, L. Y., Seligman, M. E., & Teasdale, J. D. (1978). Learned helplessness in humans: Critique and reformulation. Journal of Abnormal Psychology, 87, 49–74. http://dx.doi.org/10.1037//0021-843x.87.1.49. Asner-Self, K. K., & Schreiber, J. B. (2004). A factor analytic study of the attributional style questionnaire with Central American immigrants. Measurement and Evaluation in Counseling and Development, 37, 144–152. Retrieved from http:// www.highbeam.com/doc/1G1-125640880.html. Baron, R. M., & Kenny, D. A. (1986). The moderator–mediator variable distinction in social psychological research: Conceptual, strategic, and statistical considerations. Journal of Personality and Social Psychology, 51, 1173–1182. http://dx.doi.org/10.1037/0022-3514.51.6.1173. Bennett, K. K., & Elliott, M. (2002). Explanatory style and health: mechanisms linking pessimism to illness. Journal of Applied Social Psychology, 32, 1508–1526. http://dx.doi.org/10.1111/j.1559-1816.2002.tb01449.x. Birley, A. J., Gillespie, N. A., Heath, A. C., Sullivan, P. F., Boomsma, D. I., & Martin, N. G. (2006). Heritability and nineteen-year stability of long and short EPQ-R neuroticism scales. Personality and Individual Differences, 40, 737–747. http:// dx.doi.org/10.1016/j.paid.2005.09.005. Burnes, B., & Cooke, B. (2013). Kurt Lewin’s Field Theory: A review and reevaluation. International Journal of Management Reviews, 15, 408–425. http:// dx.doi.org/10.1111/j.1468-2370.2012.00348.x. Byrne, B. M. (2001). Structural equation modelling with AMOS: Basic concepts, applications, and programming. New Jersey: Lawrence Erlbaum Assoc. Corr, P. J., & Gray, J. A. (1996). Structure and validity of the Attributional Style Questionnaire: A cross-sample comparison. Journal of Psychology, 130, 645–658. http://dx.doi.org/10.1080/00223980.1996.9915038. de Winter, J. C. F., & Dodou, D. (2012). Factor recovery by principal axis factoring and maximum likelihood factor analysis as a function of factor pattern and sample size. Journal of Applied Statistics, 39, 695–710. http://dx.doi.org/10.1080/ 02664763.2011.610445. Dua, J. (1995). Retrospective and prospective psychological and physical health as a function of negative affect and attributional style. Journal of Clinical Psychology, 51, 507–518. doi: 10.1002/1097-4679(199507)51:4<507::AIDJCLP2270510407>3.0.CO;2-P. Dykema, J., Bergbower, K., & Peterson, C. (1995). Pessimistic explanatory style, stress, and illness. Journal of Social and Clinical Psychology, 14, 357–371. http:// dx.doi.org/10.1521/jscp.1995.14.4.357. Eysenck, H. J., & Eysenck, S. B. G. (1991). Manual of the eysenck personality scales (EPS adult). London: Hodder & Staughton. Furnham, A., Sadka, V., & Brewin, C. R. (1992). The development of an occupational Attributional Style Questionnaire. Journal of Organisational Behaviour, 13, 27–39. http://dx.doi.org/10.1002/job.4030130104. Gibb, B. E., Zhu, L., Alloy, L. B., & Abramson, L. Y. (2002). Attributional styles and academic achievement in university students: A longitudinal investigation. Cognitive Therapy and Research, 26, 309–315. http://dx.doi.org/10.1023/ A:1016072810255. Gilbert, D. T., Pinel, E. C., Brown, R. P., & Wilson, T. D. (2000). The illusion of external agency. Journal of Personality and Social Psychology, 79, 690–700. http:// dx.doi.org/10.1037//0022-3514.79.5.690. Goldberg, D., & Williams, P. (1988). A users guide to the General Health Questionnaire. Windsor, Berks: NFER-Nelson. Greenbaum, T. L. (1993). The handbook for focus group research. New York, NY: Macmillan. Hardy, G. E., Shapiro, D. A., Haynes, C. E., & Rick, J. E. (1999). Validation of the General Health Questionnaire-12 using a sample of employees from England’s Health Care Services. Psychological Assessment, 11, 159–165. http://dx.doi.org/ 10.1037//1040-3590.11.2.159. Heider, F. (1958). The psychology of interpersonal relations. New York: Wiley. Higgins, N. C., Zumbo, B. D., & Hay, J. L. (1999). Construct validity of attributional style: Modeling context-dependent item sets in the Attributional Style Questionnaire. Educational and Psychological Measurement, 59, 804–820. http://dx.doi.org/10.2224/sbp.1999.27.3.221. Hinkin, T. R., Tracey, J. B., & Enz, C. A. (1997). Scale construction: Developing reliable and valid measurement instruments. The Council on Hotel, Restaurant, and Institutional Education, 21, 100–120. http://dx.doi.org/10.1177/ 109634809702100108.
Horn, J. L. (1965). A rationale and test for the number of factors in factor analysis. Psychometrika, 30, 179–185. http://dx.doi.org/10.1007/BF02289447. Hunsley, J., & Meyer, G. J. (2003). The incremental validity of psychological testing and assessment: Conceptual, methodological, and statistical issues. Psychological Assessment, 15, 446–455. http://dx.doi.org/10.1037/10403590.15.4.446. Jeavons, S., & Greenwood, K. M. (2007). World assumptions, coping, and attributions. Personality and Individual Differences, 42, 1427–1437. http:// dx.doi.org/10.1016/j.paid.2006.11.010. Kelley, H. H. (1967). Attribution theory in social psychology. In D. Levine (Ed.). Nebraska symposium on motivation (Vol. 15, pp. 192–240). Lincoln: University of Nebraska Press. Kleim, B., Gonzalo, D., & Ehlers, A. (2011). The Depressive Attributions Questionnaire (DAQ): Development of a short self-report measure of depressogenic attributions. Journal of Psychopathology and Behavioral Assessment, 33, 375–385. http://dx.doi.org/10.1007/s10862-011-9234-9. Lewin, K. (1936). Principals of Topological Psychology. New York, NY: McGraw Hill. Luo, X., George, M. L., Kakauras, I., Edwards, C. L., Pietrobon, R., Richardson, W., et al. (2003). Reliability, validity, and responsiveness of the short form 12-item survey (SF-12) in patients with back pain. Spine, 28, 1739–1745. http:// dx.doi.org/10.1097/01.BRS.0000083169.5867.96. Luten, A. G., Ralph, J. A., & Mineka, S. (1997). Pessimistic attributional style: Is it specific to depression versus anxiety versus negative affect. Behaviour Research and Therapy, 35, 703–719. http://dx.doi.org/10.1016/s0005-7967(97)00027-2. Malle, B. (2011). Time to give up the dogmas of attribution: An alternative theory of behaviour explanation. In M. Zanna & J. Olsen (Eds.). Advances in Experimental Social Psychology (Vol. 44, pp. 297–352). San Francisco, CA: Elsevier Inc. Malle, B. F., Knobe, J., O’Laughlin, M. J., Pearce, G. E., & Nelson, S. E. (2000). Conceptual structure and social functions of behaviour explanations: Beyond person–situation attributions. Journal of Personality and Social Psychology, 79, 309–326. http://dx.doi.org/10.1037/0022-3514.79.3.309. Michelson, L. K., Bellanti, S., Testa, M., & Marchione, N. (1997). The relationship of attributional style to agoraphobia severity, depression, and treatment outcome. Behaviour Research and Therapy, 35, 1061–1073. http://dx.doi.org/10.1016/ s0005-7967(97)10011-0. Peterson, C. (1991). On shortening the expanded Attributional Style Questionnaire. Journal of Personality Assessment, 56, 179–183. http://dx.doi.org/10.1207/ s15327752jpa5601_16. Peterson, C., Semmel, A., von Baeyer, C., Abramson, L., Metalsky, G., & Seligman, M. (1982). The Attributional Style Questionnaire. Cognitive Therapy and Research, 6, 287–300. http://dx.doi.org/10.1007/BF01173577. Peterson, C., & Vaidya, R. S. (2001). Explanatory style, expectations, and depressive symptoms. Personality and Individual Differences, 31, 1217–1223. http:// dx.doi.org/10.1016/s0191-8869(00)00043-x. Peterson, C., & Villanova, P. (1988). An expanded Attributional Style Questionnaire. Journal of Abnormal Psychology, 97, 87–89. http://dx.doi.org/10.1037//0021843x.97.1.87. Proudfoot, J. G., Corr, P. J., Guest, D. E., & Gray, J. A. (2001). The development and evaluation of a scale to measure occupational attributional style in the financial services sector. Personality and Individual Differences, 30, 259–270. http:// dx.doi.org/10.1016/s0191-8869(00)00043-x. Smith, P., Caputi, P., & Crittenden, N. (2013). Measuring optimism in organisations: Development of a Workplace Explanatory Styles Questionnaire. Journal of Happiness Studies, 14, 415–432. http://dx.doi.org/10.1007/s10902-012-9336-4. Sturman, E., Mongrain, M., & Kohn, P. (2006). Attributional style as a predictor of hopelessness depression. Journal of Cognitive Psychotherapy: An International Quarterly, 20, 447–458. http://dx.doi.org/10.1891/jcpiq-v2014a008. Velicer, W. F., Easton, C. A., & Fava, J. L. (2000). Construct explication through factor or component analysis: A review and evaluation of alternative procedures for determining the number of factors or components. In R. D. Goffin & E. Helmes (Eds.), Problems and solutions in human assessment: Honoring Douglas N. Jackson at Seventy (pp. 41–71). Boston: Kluwer. Volpe, N., & Levin, R. (1998). Attributional style, dreaming and depression. Personality and Individual Differences, 25, 1051–1061. http://dx.doi.org/ 10.1016/s0191-8869(98)00073-7. Ware, J. E., Kosinski, M. M., & Keller, S. D. (1996). A 12-item Short-Form Health Survey: Construction of scales and preliminary tests of reliability and validity. Medical Care, 34, 220–233. http://dx.doi.org/10.1097/00005650-19960300000003. Weiner, B. (1986). An attributional theory of motivation and emotion. New York: Springer Verlag. Weiner, B. (2014). The attribution approach to emotion and motivation: History, hypotheses, home runs, headaches/heartaches. Emotion Review, 6, 353–361. http://dx.doi.org/10.1177/1754073914534502. Whitley, B. E. Jr., (1991). A short form of the expanded Attributional Style Questionnaire. Journal of Personality Assessment, 56, 365–369. http:// dx.doi.org/10.1207/s15327752jpa5602_14. Worthington, R. L., & Whittaker, T. A. (2006). Scale development research: A content analysis and recommendations for best practices. The Counselling Psychologist, 34, 806–838. http://dx.doi.org/10.1177/0011000006288127. Xenikou, A., Furnham, A., & McCarrey, M. (1997). Attributional style for negative events: A proposition for a more reliable and valid measure of attributional style. British Journal of Psychology, 88, 53. http://dx.doi.org/10.1111/j.20448295.1997.tb02620.x.