Neuroticism: moderator or mediator in the relation between locus of control and depression?

Transcription

1 Personality and Individual Differences 37 (2004) Neuroticism: moderator or mediator in the relation between locus of control and depression? Dave Clarke * School of Psychology, Massey University, Albany Campus, Private Bag , North Shore Mail Centre, New Zealand Received 13 February 2003; received in revised form 30 July 2003; accepted 24 August 2003 Available online 21 October 2003 Abstract The present investigation examined empirically the influence of neuroticism on the relation between locus of control and depression, first as a moderator with specified interaction effects, and secondly as a mediator in the path from locus of control to depression. LevensonÕs internality, powerful others and chance (IPC) locus of control scales, the EysencksÕ EPQ-R neuroticism scale and a depression inventory were completed by 162university students aged years (mean ¼ 27.9, SD ¼ 10.3). There were no significant differences between sexes on any mean score. Controlling for sex and age, externality, neuroticism and depression were significantly correlated (p < 0:001), but only chance locus of control and neuroticism predicted depression. In the absence of significant interaction and quadratic effects, multiple regression analyses did not support neuroticism as a moderator, but as a partial mediator in the relationship between chance locus of control and depression. Path analysis showed that the total effect size was very large. The influence of neuroticism was discussed in terms of the depressive paradox (self-blame and uncontrollability co-existing in depressed individuals), the multidimensional aspects of locus of control, and implications for treating depression. Ó 2003 Elsevier Ltd. All rights reserved. Keywords: Depression; Locus of control; Neuroticism 1. Introduction Recently the personality traits of locus of control and neuroticism have been examined together for their impact on various clinical conditions (Beautrais, Joyce, & Mulder, 1999; Horner, 1996; Murray, Hay, & Armstrong, 1995; van den Heuvel, Smits, Deeg, & Beekman, 1996; Wise & * Tel.: x9075; fax: address: d.clarke@massey.ac.nz (D. Clarke) /$ - see front matter Ó 2003 Elsevier Ltd. All rights reserved. doi: /j.paid

2 246 D. Clarke / Personality and Individual Differences 37 (2004) Mann, 1993). Locus of control involves a general enduring belief in the controllability of outcomes of events in oneõs life (Rotter, 1966). According to Rotter, it ranges from a belief that such outcomes are controllable (internal locus of control orientation or internality), to a belief that outcomes are generally beyond personal control (external locus of control orientation or externality). Neuroticism is a personality trait that has provided a useful integrative psychobiological construct in depression research (Murray et al., 1995). Eysenck (1967) proposed that it originates from the sensitivity of the limbic autonomic nervous system that tends to be highly reactive to environmental and psychological stimuli. People with high neuroticism are easily startled and agitated. For example, Horner (1996) used RotterÕs (1966) locus of control scale and the neuroticism scale of the Eysenck personality questionnaire (EPQ; Eysenck & Eysenck, 1975) with 173 community dwelling adults to show that an interaction of stressful life experiences, very high externality and very high neuroticism could predict reported physical illness. Neither external locus of control or neuroticism independently had a significant, direct effect on illness. For psychological conditions, the two variables are also salient. Murray et al. (1995) found with an all-female sample that external locus of control of moods and neuroticism were related to seasonal affective disorder, but extraversion, psychoticism and the EPQ Lie scale were not. Similarly, Beautrais et al. (1999) demonstrated that externality and neuroticism could reliably distinguish between young suicide attempters and non-attempters, but extraversion could not. In a large sample of 3107 elderly participants in a Dutch longitudinal study, van den Heuvel et al. (1996) found that women tending toward externality and neuroticism reported more depressive symptoms when experiencing various impairments in cognitive functioning. For men experiencing memory impairment, only neuroticism affected levels of depression. There are, however, some problems with the conceptual and empirical aspects of the associations between locus of control, neuroticism and depression. These problems include the depressive paradox, the multidimensional nature of locus of control, and the functions of moderators and mediators The depressive paradox In their meta-analysis of data from a wide range of empirical studies of the relation between locus of control and depression, Benassi, Sweeney, and Dufour (1988) highlighted the depressive paradox between two major cognitive theories of depression. SeligmanÕs helplessness theory postulates that depressed people perceive events as uncontrollable, while BeckÕs negative schematic theory implies that they have thoughts of self-deprecation and self-blame. Both theories have received substantial empirical support. The paradox seems to involve locus of control orientation because depressed individuals perceive events beyond their personal control (external), while at the same time blaming themselves for failures (internal). A number of hypotheses have been generated to explain the paradox (Peterson, 1979). The first hypothesis is that there is no paradox at all: depression is not related to locus of control. Any relationships reported might have been spurious. But Benassi et al. (1988) have clearly shown that there is a consistent, moderately strong relationship between external locus of control orientation and depression. Greater belief that outcomes of events are generally beyond personal control is associated with greater depression. A second hypothesis is that there are two groups of depressed

3 D. Clarke / Personality and Individual Differences 37 (2004) persons: one group believes in uncontrollability, the other uses self-blame. A third hypothesis is that there are two types of outcomes: depressed persons perceive positive outcomes as uncontrollable or resulting from luck, and blame themselves for negative outcomes. A fourth hypothesis equates self-blame with internality. None of these four hypotheses has received empirical support (Abramson & Sackeim, 1977; Benassi et al., 1988; Peterson, 1979). There is some empirical support, however, for a fifth hypothesis: depressed individuals are particularly illogical in their attributions. They have a tendency to blame themselves for events over which they perceive no control. Peterson, for example, has shown with undergraduate research participants that scores on the Beck depression inventory (BDI) were significantly and positively correlated with perceived helplessness and guilt simultaneously. Almost all subjects were Ôillogical,Õ and depressives (those with high scores on the BDI) were particularly so. (p. 623) Although neuroticism is not explicitly mentioned in the literature, the reviews, empirical findings and discussions denote that the paradox and the way that externality functions may involve a personality construct such as neuroticism. Because of emotional instability, a neurotic person may have illogical attributions. Neuroticism also involves worry, guilt and preoccupation with security (Eysenck & Eysenck, 1991). As Horner (1996) has noted, the perception of uncontrollability appears to be exacerbated for individuals who are already worried and anxious (p. 197). Thus, externality is an antecedent, cognitive condition that can lead to depression (Benassi et al., 1988), and neuroticism may influence at least partially that pathway from externality to depression. Internal locus of control, however, may play no role in depression Dimensional aspects of locus of control All of the above studies with both neuroticism and locus of control assessed locus of control orientation on a Rotter-based scale whereby responses are scored on one internal external dimension: from internality to externality. But there is some empirical evidence (Lefcourt, 1991) that locus of control is multidimensional. A person can believe to a certain degree in oneõs own ability to control events while also believing that powerful others control the events in oneõs life, or in the power of luck or chance happenings. Lefcourt, for example, has shown that on LevensonÕs (1981) internality (I) powerful others (P) and chance (C) (IPC) scales, while P and C were moderately correlated with each other, correlations of the P and C scales with I ranged weakly from negative to positive. Conceptually, little research has tried to understand why internality and externality are on different dimensions, but it seems important in empirical studies to treat locus of control as a multidimensional rather than a unidimensional construct. In their meta-analysis, Benassi et al. (1988) examined the possibility that the connection between externality and depression may be due to the possibility that depressed individuals participating in research might have responded to the negativity in external locus of control items rather than to the external content in the items. The research that supported that possibility was based on RotterÕs forced-choice items. However, the relationship between externality and depression still existed even when Likert-type rating scales such as LevensonÕs IPC scales were used in other studies, and when forced-choice items were balanced for mood level. Rating scales do not force respondents to choose between two items of a pair, one of which may be negative in tone. Benassi et al. (1988) also compared the uni- and multidimensional measures of locus of control. The strength of the relationship between externality and depression was significantly weaker for

4 248 D. Clarke / Personality and Individual Differences 37 (2004) Rotter-type locus of control scales than for multidimensional scales such as LevensonÕs IPC scales. It was also significantly weaker for BeckÕs depression inventory (BDI) and the MMPI/ depression subscale, than for the Center for Epidemiological Studies-Depression (CES-D) and Zung measures of depression. Although depression is two to three times more common in women than in men (Davison & Neale, 2001), effect sizes were not significantly related to the percentage of females in the studies reviewed by Benassi et al. When internality, powerful others and chance locus of control have been investigated separately, depression was associated with powerful others locus of control and a lack of internality (Benassi et al., 1988). More recent studies found that chance locus of control predicted depression (e.g., Lamanna, 2001) and, inversely, subjective well-being (DeNeve & Cooper, 1998), rather than internality or powerful others Moderators and mediators To date, neuroticism has been considered as a moderator in the relationship between external locus of control and depression (Horner, 1996). A variable functions as a moderator when it alters the direction and/or strength of the effect of a predictor variable on a criterion variable (Baron & Kenny, 1986). A moderating effect is determined statistically by the presence of an interaction effect of the predictor and the moderator on the criterion. For example, in her review of the literature, Horner concludes that neuroticism is the most likely personality variable that moderates the influence of locus of control. It is stable over time and situation, and a person with high neuroticism is vulnerable to clinical conditions even when stressors are not present. She then demonstrated that neuroticism and externality interacted in a quadratic fashion to moderate the effects of stressors on physical illness. Only very high neuroticism and very high external locus of control in the presence of stressors predicted illness. Neuroticism and externality were not significant predictors of illness. An alternative consideration is that neuroticism may function as a mediator in the relationship between externality and clinical conditions. For example, Wise and Mann (1993) suggested path analysis for clarifying the nature of the mediating effect of neuroticism between locus of control and alexithymia, a clinical construct denoting a personõs inability to experience and relate emotional feelings, and related to depression. Because of their small number of subjects, they were unable to do path analysis. A variable functions as mediator to the extent that it accounts for the relation between the predictor and the criterion (Baron & Kenny, 1986). Whereas moderators specify when certain effects will occur, mediators explain how or why such effects occur. Fig. 1 illustrates the difference between neuroticism as a moderator and neuroticism as a mediator. In Fig. 1a, there are three causal paths that lead to depression: the influence of externality as a predictor (Path a), the influence of neuroticism as a moderator (Path b), and the interaction of these two (Path c). The moderating effect is present if Path c is significant. Paths a and b may be significant also, but are not directly relevant for testing moderation. It would also be expected that N would not be correlated with both Ext and Dep to provide a clearly interpretable interaction effect. Thus, if N is a moderator, then there would be a synergistic (multiplicative) effect on depression: externality and neuroticism together would account for additional variation in depression beyond externality and neuroticism as independent variables.

5 D. Clarke / Personality and Individual Differences 37 (2004) Ext N a b Dep a Ext N b Dep Ext x N c c (a) Moderated Model (b) Mediated Model Fig. 1. Models derived from Baron and Kenny (1986) of moderated (a) and mediated (b) effects of neuroticism (N) on the relation between external locus of control (Ext) and depression (Dep). In Fig. 1b, however, N would be correlated with both Ext and Dep. Mediation implies that Ext (the predictor) influences N (the mediator) which, in turn, influences Dep (the criterion variable). According to Baron and Kenny (1986): Moderator variables are typically introduced when there is an unexpectedly weak or inconsistent relation between a predictor and a criterion variable (e.g., a relation holds in one setting but not in another, or for one subpopulation but not for another)....mediation, on the other hand, is best done in the case of a strong relation between the predictor and the criterion variable (p. 1178). Baron and Kenny have provided several guidelines for investigating empirically the effects of a mediator on a predictive relationship. When individual differences such as neuroticism are involved, a mediator represents a property or stable characteristic of a person that transforms the predictor variables in some way. For neuroticism to function as a mediator in the path from externality to depression, two conceptual relationships are assumed: (1) externality leads to depression, rather than vice versa, and (2) depression is not a cause of neuroticism. Support for the two assumptions is provided from the findings of Greenspoon and Sasklofske (2001) who demonstrated with a sample of 407 children that externality and neuroticism were significantly related to the development of internalised psychopathology. Similarly, Sumi and Kanda (2002) showed in a prospective design with 138 Japanese male college students that neurotic perfectionism was a predictor of depression, and not vice versa The present study The purpose of the present investigation was to use Baron and KennyÕs (1986) criteria for testing empirically if neuroticism has a moderating or a mediating influence on the relation between externality and depression. From the literature reviewed above, several guidelines are suggested. First, the strengths of the relationships among the variables should be measured to (a) ascertain which variables are related to depression, and (b) determine if neuroticism fulfills the desired conditions for a moderating or a mediating effect. For (b), if the relation between externality and depression is weak or inconsistent, then a moderating effect may be present; if the

6 250 D. Clarke / Personality and Individual Differences 37 (2004) relationships among all three are strong, then a mediating effect is suspected. Secondly, the multidimensional role of locus of control as a predictor of depression should be investigated before testing for moderating or mediating effects of neuroticism. Thirdly, it can be assumed that neuroticism is the personality trait most likely to affect the relation between externality and depression. Fourthly, the measure of depression should be appropriate for a non-clinical as well as a clinical population. 2. Method 2.1. Sample A class of 165 students who were enrolled in an introductory psychology course at the Albany campus of Massey University completed the questionnaires; 162of them provided sufficient data for the current analyses. There were 129 women and 33 men. The age distribution was moderately skewed (0.94), with no outliers. Ages ranged from 17 to 57 years, with an average age of 27.9 years (SD ¼ 10.3; median ¼ 24). Seventy-two percent of the sample identified themselves as Caucasian or New Zealand European, 14% as Polynesian, and 9% as Asian. Most of the participants (83%) were in the average and above average socioeconomic groups. The sample thus consisted predominately of young, female, Caucasian adults with moderate to high socioeconomic status Materials Locus of control RotterÕs (1966) internal external locus of control scale has been the most widely used and cited measure (Benassi et al., 1988), but has been criticized extensively with regard to its presumed unidimensionality, its inherent social desirability response bias, and the difficulties and complications created by its forced-choice format, such as its negatively biased items (Benassi et al., 1988; Lefcourt, 1991). The internality, powerful others and chance (IPC) scales (Levenson, 1981) consist of 24 items with three subscales corresponding to three dimensions of locus of control. For the present study, participants rated each item on a 7-point Likert-type scale ranging from strongly disagree (1) to strongly agree (7), with neutral (4) as the midpoint. In contrast to RotterÕs dichotomous measure of locus of control, LevensonÕs scales provide a multidimensional measure of the concept, have no reported social desirability bias, are more statistically independent, have good psychometric properties, and make a distinction between personal and ideological statements by having all items phrased in the first person as opposed to describing people in general (Lefcourt, 1991). The internality (I) scale measures the extent to which one believes that one has control over oneõs life. The items seem to describe the concept of self-determination (e.g., My life is determined by my own actions. ). The powerful others (P) scale concerns the belief that other persons control the events in oneõs life (e.g., Getting what I want requires pleasing those people above me. ). The chance (C) scale measures the degree to which one believes that fate or luck affects oneõs experiences and outcomes (e.g., To a great extent my life is controlled by accidental happenings. ). Scores can range from 8 to 56 on each scale. For the present sample, the coefficients of internal consistency (CronbachÕs alpha), were 0.63 (I), 0.74 (P) and 0.82(C).

7 Neuroticism The neuroticism (N) scale was selected from the revised Eysenck personality questionnaire (EPQ-R; Eysenck & Eysenck, 1991) to measure the trait of emotional instability, worry and guilt. Scores range from 0 to 24. CronbachÕs alpha for the group was Depression The depression symptom inventory was developed from previous research (Bell, LeRoy, & Stephenson, 1982) with medical inpatients and outpatients, and cross-sectional studies such as the Stirling County and the Midtown Manhattan surveys to ascertain levels of depression in the community. It was used rather than the Beck depression inventory (BDI; Beck, Ward, Mendelson, Mock, & Erbaugh, 1961) because the BDI was designed to assess the severity of depression in a clinically diagnosed group (Davison & Neale, 2001), whereas the depression symptom inventory is also applicable for a non-clinical population. Each item is rated on a five-point scale, from often to never, with a maximum score of 72. For the present study, item one was re-worded from Do you feel in good spirits? to Do you have a feeling of well-being?, to avoid Maori connotations relating to the spirits of their ancestors; and item 18 from How does the future look to you? to Does the future seem uncertain to you? to make sense in terms of the rating scale. The modified depression scale had an internal consistency coefficient of 0.83 for the present sample Procedure D. Clarke / Personality and Individual Differences 37 (2004) The strengths of the intercorrelations were assessed by Pearson correlation coefficients. A simultaneous regression analysis was undertaken to determine the unique contribution of each variable in the prediction of depression. With the significant predictor variables, the analyses then tested whether neuroticism assumed the role of moderator or mediator in the relation between externality and depression. For the moderating effect of neuroticism, hierarchical regression analysis tested for linear, quadratic and interaction effects on the relation between externality and depression, as suggested by Baron and Kenny (1986). The order of entry was externality, neuroticism (N), their interaction (externality neuroticism), and the quadratic terms (N 2 and externality N 2 ). Quadratic effects were examined because Horner (1996) in her conclusions suggested that nonlinear aspects of personality may be more salient for illness research than linear ones. Before testing for the interaction effects, the scores on the appropriate IPC scales and N were centered by subtracting their respective sample means from all individualsõ scores, thus producing revised sample means of zero. The procedure, suggested by Aiken and West (1991), was to eliminate the possible effects of multicollinearity between first order terms of the main effects and the higher order interaction and quadratic terms. To test for the mediating effect of neuroticism in the relation of externality to depression, three regression equations must be computed and four conditions must be met (Baron & Kenny, 1986). For the variables in the present study, the three regression equations are: (1) Neuroticism ¼ B 1 Externality + Control Variables, (2) Depression ¼ B 2 Externality + Control Variables, (3) Depression ¼ B 3 Externality + B 4 Neuroticism + Control Variables,

8 252 D. Clarke / Personality and Individual Differences 37 (2004) where Bs represent the unstandardised regression coefficients. With these three equations, the four conditions are: (a) B 1 must be significant; (b) B 2 must be significant; (c) B 4 must be significant; and (d) B 3 must be smaller than B 2.IfB 3 is reduced to a non-significant effect, full mediation is demonstrated. If B 3 is reduced, but still significant, then partial mediation is demonstrated. If a mediating effect is present, then path analysis would be conducted to estimate the direct and indirect effects of externality on depression, with neuroticism as the mediating variable (Pedhazur, 1997). 3. Results Prior to inferential analyses, the data for each scale were examined for outliers and for assumptions of normality. For the IPC and N scales, and the depression inventory, the distributions of data were within the normal ranges and raw data were used in the computations. The level for significance was set at p < 0:001 for a conservative estimate of the significance of relationships among the variables. A one-way MANOVA revealed no significant differences between menõs and womenõs mean scores on the IPC locus of control scales, Neuroticism and depression, WilksÕ lambda ¼ 0.94, F ð5; 157Þ ¼2:04, p > 0:05. Therefore, menõs and womenõs data were combined for succeeding analyses Intercorrelations The intercorrelations among the variables, their means and standard deviations are shown in Table 1. The zero-order correlations appear below the diagonal; partial correlations in which the effects of sex and age have been controlled are above the diagonal. As depicted in Table 1, there was almost no change in significance or value after controlling for sex and age. Depression was significantly related to powerful others locus of control (r ¼ 0:35), chance locus of control (r ¼ 0:47) and neuroticism (r ¼ 0:63). Neuroticism significantly correlated with powerful others and chance locus of control (r ¼ 0:33, each). The two external locus of control variables, powerful others and chance, were strongly correlated (r ¼ 0:70). Table 1 Means, standard deviations and correlations for LevensonÕs IPC locus of control scores, Eysenck neuroticism scale scores and depression inventory scores (N ¼ 162) Variables Internality )0.02 ) )0.01 Powerful Others ) * 0.32* 0.36* Chance ) * 0.33* 0.47* Neuroticism * 0.33* 0.64* Depression ) * 0.47* 0.63* Mean SD *p < 0:001. Pearson product moment correlations appear below the diagonal; partial correlations controlling for sex and age, above the diagonal. All tests are two-tailed.

9 Internality was not significantly related to any of the other variables. A simultaneous regression analysis with I, P, C and N as predictors of depression revealed that only C ðb ¼ 0:35Þ and N ðb ¼ 0:54Þ accounted for significant, unique variance in depression scores, F ð4; 158Þ ¼ 36:11, p > 0:001. Internality (b ¼ 0:06) and powerful others locus of control (b ¼ 0:07) contributed negligibly (p > 0:05), and were thus omitted in the tests for moderating and mediating effects on the relation between externality and depression. The purpose of the present investigation was to ascertain the unique contribution and importance of the predictor variables on depression. Although P correlated moderately with depression (r ¼ 0:35), it was not included in testing for the moderating and mediating effects, because of its multicollinearity with C and hence instability and problematic interpretation of the results of multiple regression analysis (Polit, 1996, Chap. 4) Moderating effects of neuroticism Table 2presents the results of hierarchical regression analysis for the main, interaction and quadratic effects on depression. Only the main effects, chance locus of control (b ¼ 0:27) and neuroticism (b ¼ 0:52), remained significant at each step, accounting for unique variance in depression scores, before and after the introduction of the interaction and quadratic terms. Chance locus of control accounted for 7% and neuroticism 29% of the variance in depression scores Mediating effects of neuroticism D. Clarke / Personality and Individual Differences 37 (2004) The first step in testing the mediating effect of neuroticism was to demonstrate that chance locus of control is significantly related to neuroticism. The zero-order correlation of 0.33 noted in Table 1, after controlling for sex and age, provides initial support for this effect. The results of the regression analyses predicting depression as a function of chance locus of control and mediated by neuroticism appear in Table 3. All four of the conditions specified by Baron and Kenny (1986) were met at or beyond the level of significance: (a) chance locus of control significantly accounted for variations in neuroticism (B 1 ¼ 0:25), F ð1; 161Þ ¼29:47, (b) the relationship between chance locus of control and depression was initially significant (B 2 ¼ 0:57), F ð1; 161Þ ¼44:91, (c) neuroticism significantly accounted for variations in depression (B 4 ¼ 0:88), F ð2; 160Þ ¼71:84, and (d) with neuroticism in the equation, the unstandardised regression coefficient for chance locus of control on depression was smaller than B 2 (B 3 ¼ 0:35). Further, because Table 2 Results of hierarchical regression analysis of powerful others locus of control, chance locus of control, neuroticism, interaction and quadratic effects on depression (N ¼ 162) Variable (R 2 ¼ 0:48) B SE B b Partial R 2 F at entry Chance 0.33* * 44.91* Neuroticism 0.86* * 71.84* Chance neuroticism ) ) * Neuroticism * Chance neuroticism * Intercept 27.52* 2.62 *p < 0:001.

10 254 D. Clarke / Personality and Individual Differences 37 (2004) Table 3 Summary of regression analyses for direct and indirect effects of chance locus of control on depression via neuroticism (N ¼ 162) Variable B SE B b Partial R 2 Regression of neuroticism on chance locus of control (R 2 ¼ 0:11) Constant Chance 0.25* Regression of depression on chance locus of control (R 2 ¼ 0:22) Constant 30.77* 2.24 Chance 0.57* Regression of depression on neuroticism and chance locus of control (R 2 ¼ 0:47) Constant 26.91* 1.89 Chance 0.35* Neuroticism 0.88* Indirect 0.18 Spurious 0.10 *p < 0:001. B 3 was still significant, neuroticism had a partial mediating effect on the prediction of depression from chance locus of control. The results of the third regression in Table 3 show the proportion of the total effect of chance locus of control (0.47) on depression, consisting of the direct effect (0.29) and the indirect effect through neuroticism (0.18). Therefore, as a mediator, neuroticism accounted for 38% of the total effect of chance locus of control on depression. For multiple and multiple partial correlations, Cohen (1992) has tabled a standardised measure of association (f ) which is independent of the original measurement units. An f 2 of 0.35 is considered a large effect size. For the third regression, an R 2 of 0.47 yields an f 2 of Hence, the total effect of chance locus of control on depression via neuroticism was very large. The statistical power to detect significance at the 0.01 level was at least 90% for large effect sizes Neuroticism 0.33* 0.54* 0.53 Chance Locus of Control 0.29* Depression Note: Residuals appear incircles. *p< Fig. 2. Path model and standardised regression coefficients depicting the role of neuroticism in mediating the effects of chance locus of control on depression.

11 D. Clarke / Personality and Individual Differences 37 (2004) Fig. 2presents the path analysis for the mediating model (Pedhazur, 1997). The standardised path coefficients (bõs) show the direction and magnitude of the effects of chance locus of control and neuroticism on depression. The coefficients depicting the paths from chance locus of control to neuroticism (0.33), from neuroticism to depression (0.54) and from chance locus of control to depression (0.29) were significant (p < 0:001). 4. Discussion 4.1. Locus of control, neuroticism, and depression Because internal locus of control was not related to powerful others and chance locus of control, the present study provided further evidence that locus of control is a multidimensional rather than a unidimensional construct. In contrast, powerful others and chance locus of control were highly correlated. The two variables could be considered as different manifestations of external locus of control, belief that events in oneõs life are controlled by factors beyond oneõs influence or control. Internality was also not related to neuroticism or depression. Because the coefficient of internal consistency was the lowest of the three scales (0.63), it is possible that error variance could have masked the extent of the association of internality with the other variables. For example, the correlation between internality and chance locus of control ()0.23) was significant beyond the 0.01 level, above the criterion for the tests of significance in the present study (p < 0:001). The correlations with powerful others and depression were almost zero. Thus, it is likely that internality is negatively correlated with chance rather than powerful others locus of control or depression. An increase in self-determination or internality is associated with a decrease in belief in fate or luck but does not seem to affect depression Moderating versus mediating effects of neuroticism The data from the present investigation provided no empirical support for the moderating influence of neuroticism on the relation between externality and depression. The correlations among the three variables were highly significant and moderately strong, suggesting a mediating influence instead. Confirmatory hierarchical regression analysis showed that in the absence of interaction and quadratic effects, neuroticism was not functioning as a moderator, specifying when the relation is most likely to occur. On the other hand, there was strong support for the concept of neuroticism as a mediator in the path from externality and depression. Specifically, the component of externality that functioned in the path was chance rather than powerful others locus of control, as found in recent studies (DeNeve & Cooper, 1998; Lamanna, 2001). Neuroticism could explain partially how chance locus of control can lead to depression: the stronger the belief that fate controls the outcomes of personal events, the greater the stimulation of worry and guilt, and the deeper the depression that is manifested (Eysenck & Eysenck, 1991; Greenspoon & Sasklofske, 2001; Sumi & Kanda, 2002). Because the relationships among the three variables were essentially the same after controlling for sex and age, their influences seem to apply equally for both male and female adults of all ages.

12 256 D. Clarke / Personality and Individual Differences 37 (2004) In contrast, Horner (1996) found that neuroticism moderated the relation between external locus of control and physical illness non-linearly with stressors. Similarly, van den Heuvel et al. (1996) showed that neuroticism and externality moderated the relation between cognitive functioning and depression in older adults. However, both of these studies used unidimensional measures of locus of control and did not examine mediating effects. The mediating effect of neuroticism may be applicable only for chance locus of control leading to depression. Whether neuroticism has an actual causal moderating or mediating effect needs to be ascertained with experimental and longitudinal data (van den Heuvel et al., 1996). The evidence that does exist from prospective research (Greenspoon & Sasklofske, 2001; Sumi & Kanda, 2002) suggests that externality and neuroticism predict depression. Yet to be demonstrated is that by introducing an uncontrollable chance event, chance locus of control stimulates people who worry, with both chance locus of control and neuroticism leading to depression The depressive paradox The paradox that individuals who perceive the outcomes of events as uncontrollable, yet somehow still feel responsible, may be explained partially by the mediating effect of neuroticism. First, it is more likely that chance rather than powerful others locus of control leads to depression. People who perceive that the outcomes of events are due to chance or luck may try to deal with their feelings of helplessness neurotically with worry and guilt (Peterson, 1979). Such a reaction exacerbates the tendency of chance locus of control to lead to depression. Secondly, it is irrelevant if the outcomes are positive or negative (Benassi et al., 1988). For example, individuals who perceive success as due to chance, may initially feel elated, but then feel guilty at succeeding when others have not, worry that they may not be so lucky next time, and thus become depressed. Because their nervous systems are attuned to be highly reactive to environmental and psychological stimuli (Eysenck, 1967), they can react also from their own perceptions that luck or fate determines their successes and failures. If they believe that powerful others affect the outcomes for them, then they are less likely to be depressed than chance believers, because they can attribute the responsibility to other people and not feel guilt and worry about the outcomes. According to the mediated model, if their belief in fate is changed, then both neuroticism and depression are affected. For example, BeckÕs therapies have helped depressed people break their maladaptive, illogical and cyclical thought patterns of negative beliefs, and hopelessness (Beck, 1995). Cognitive-behavioural therapies may thus work partially through neuroticism, because lowering belief in chance locus of control depressive symptoms may also lower the neurotic reactions leading to depression. In a review of studies of psychotherapies for depression, Jarrett (1990) concluded that cognitive, behavioural and interpersonal therapies were effective only in the acute treatment of adult, non-bipolar, non-psychotic depressed outpatients. Hence, early intervention would be recommended to affect both externality and neuroticism. Based on prospective research (Greenspoon & Sasklofske, 2001; Sumi & Kanda, 2002), it was assumed in the present study that externality leads to depression, rather than vice versa, and that depression is not a cause of neuroticism. The relationships among externality (the predictor), neuroticism (the mediator) and depression (the criterion) may not necessarily be causal. If chance locus of control, neuroticism and depression are merely covariates, then from the present findings, the covariance between chance locus of control and depression is diminished when

13 D. Clarke / Personality and Individual Differences 37 (2004) neuroticism is controlled (James & Brett, 1984). To demonstrate an actual causal mediating effect, longitudinal data would be required (van den Heuvel et al., 1996). For therapy, the implication of covariation then is that dealing with the neuroticism may be helpful in weakening the link between chance locus of control and depression for vulnerable individuals. Acknowledgements This research was funded by a Massey University Research Fund grant to D. Clarke. Much appreciation is given to Dr Ross Flett and to the reviewers for comments and suggestions on an earlier version of this article. References Abramson, L. Y., & Sackeim, H. A. (1977). A paradox in depression: uncontrollability and self-blame. Psychological Bulletin, 84(5), Aiken, L. S., & West, S. G. (1991). Multiple regression: testing and interpreting interactions. Newbury Park, CA: Sage. 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, Beautrais, A. L., Joyce, P. B., & Mulder, R. T. (1999). Personality traits and cognitive styles as risk factors for serious suicide attempts among young people. Suicide and Life-Threatening Behavior, 29(1), Beck, A. T., Ward, C. M., Mendelson, M., Mock, J. E., & Erbaugh, J. K. (1961). Inventory for measuring depression. Archives of General Psychiatry, 4, Beck, J. S. (1995). Cognitive conceptualization. In Cognitive therapy: basics and beyond (pp ). New York: Guilford Press. Bell, R. A., LeRoy, J. B., & Stephenson, J. J. (1982). Evaluating the mediating effects of social support upon life events and depressive symptoms. Journal of Community Psychology, 10, Benassi, V. A., Sweeney, P. D., & Dufour, C. L. (1988). Is there a relation between locus of control orientation and depression? Journal of Abnormal Psychology, 97(3), Cohen, J. (1992). A power primer. Psychological Bulletin, 112(1), Davison, G. C., & Neale, J. M. (2001). Abnormal Psychology (8th ed.). New York: John Wiley & Sons. DeNeve, K. M., & Cooper, H. (1998). The happy personality: A meta-analysis of 137 personality traits and subjective well-being. Psychological Bulletin, 1992, Eysenck, H. J. (1967). The biological basis of personality. Springfield, IL: Charles C. Thomas. Eysenck, H. J., & Eysenck, S. B. G. (1975). Manual of the Eysenck personality questionnaire. London: Hodder & Stoughton. Eysenck, H. J., & Eysenck, S. B. G. (1991). Manual of the Eysenck personality scales (EPS adult). London: Hodder & Stoughton. Greenspoon, P. J., & Sasklofske, D. H. (2001). Toward an integration of subjective well-being and psychopathology. Social Indicators Research, 54(1), Horner, K. L. (1996). Locus of control, neuroticism, and stressors: combined influences on reported physical illness. Personality and Individual Differences, 21(2), James, L. R., & Brett, J. M. (1984). Mediators, moderators, and tests for mediation. Journal of Applied Psychology, 69, Jarrett, R. B. (1990). Psychosocial aspects of depression and the role of psychotherapy. Journal of Clinical Psychiatry, 51(6), Lamanna, M. D. (2001). The relationships among emotional intelligence, locus of control and depression in selected cohorts of women. Dissertation Abstracts International: Section B (Sciences and Engineering), 61(10-B), 5569.

14 258 D. Clarke / Personality and Individual Differences 37 (2004) Lefcourt, H. M. (1991). Locus of control. In J. P. Robinson, P. R. Shaver, & L. S. Wrightsman (Eds.), Measures of personality and social psychological attitudes (2nd ed., pp ). San Diego, CA: Academic Press. Levenson, H. (1981). Differentiating among internality, powerful others and chance. In H. M. Lefcourt (Ed.), Research with the locus of control construct (Vol. 1, pp ). New York: Academic Press. Murray, G. W., Hay, D. A., & Armstrong, S. M. (1995). Personality factors in seasonal affective disorder: Is seasonality an aspect of neuroticism? Personality and Individual Differences, 19(5), Pedhazur, E. J. (1997). Multiple regression in behavioral research: explanation and prediction. Fort Worth, TX: Harcourt Brace College Publishers. Peterson, C. (1979). Uncontrollability and self-blame in depression: investigation of the paradox in a college population. Journal of Abnormal Psychology, 88(6), Polit, D. (1996). Multiple regression. In Data analysis and statistics for nursing research (pp ). Stamford, CT: Appleton & Lange. Rotter, J. B. (1966). Generalized expectancies for internal versus external control of reinforcement. Psychological Monographs, 80, Sumi, K., & Kanda, K. (2002). Relationship between neurotic perfectionism, depression, anxiety, and psychosomatic symptoms: a prospective study among Japanese men. Personality and Individual Differences, 32(5), van den Heuvel, N., Smits, C. H. M., Deeg, D. J. H., & Beekman, A. T. F. (1996). Personality: a moderator of the relation between cognitive functioning and depression in adults aged 55 85? Journal of Affective Disorders, 41, Wise, T. N., & Mann, L. S. (1993). Is alexithymia distinct from health locus of control? International Journal of Psychiatry in Medicine, 23(4),