Since the introduction of the seven-factor model of personality [
Changes in HA and SD scores before and after remission suggest that these TCI subscale scores can be influenced by the mood of the subject when filling in the questionnaire. However, significant differences in TCI scores between patients with depression and normal controls, even during remission, indicate that depression severity may consist of two components: one derived from the state-dependent effects of depression and the other from the effects of the intrinsic association between personality traits and depression.
This issue may be further clarified by longitudinal studies. Such studies can predict the onset of depression or depression severity at later stages of followup by using baseline TCI subscale scores after controlling for baseline depression severity. This may rule out the state dependency effects of TCI scores on current depression severity. For example, Naito, Kijima, and Kitamura [
Anxiety disorders have also been studied in terms of their relationship with TCI personality traits. People with panic disorder score high in HA and low in SD [
Although the abovementioned longitudinal studies elegantly controlled for the state dependency of TCI scores on depression severity by means of multiple regression analysis, they may not be free from flaws. Firstly, in such regression analyses, depression severity at Time 2 is primarily explained by that at Time 1, and only the remaining variance of the severity at Time 2 is explained by TCI scores. Hence, the portion of the covariance between depression and personality traits at Time 1 that is related to their intrinsic association may be treated as a part of state-dependent effects, resulting in possible underestimation of the real predictive power of personality traits on later depression severity.
Second, these analyses are based on the assumption that depression severity is purely a “state” measure. There have been arguments that depression, like anxiety, may consist of trait and surplus components. The trait component is a temporally stable component reflecting enduring characteristics of individuals. The surplus component, on the other hand, is a variable one that reflects the current mood state. The state we observe is thus an amalgamation of these two components. Ritterband and Spielberger [
Third, multiple regression analyses assume that all the variables are free of errors. This is an implausible assumption. Errors may blur what should otherwise be clearly observed. Thus, the addition of error variables in SEM may be a way to reduce such bias.
Fourth, past investigations have generally measured only depression. However, the oft-reported association between depression and anxiety suggests that caution is required when interpreting results concerning the association between depression and TCI subscale scores, because it may be confounded by anxiety scores.
Finally, many previous studies simultaneously entered TCI temperament and character subscale scores into the regression equation. However, a basic assumption of the psychobiology model of Cloninger, Svrakic, and Przybeck [
Taking into account these criticisms of longitudinal studies regarding the prediction of depression from baseline TCI scores, we conducted a study in which university students were twice administered questionnaires containing both depression and anxiety measures, separated by a five-month interval.
In a structural equation model, we created latent variables comprised of trait components of both depression and anxiety (Figure
Structural regression path of the T1 HADS and T2 HADS dimensions and the TCI subscales. NS: novelty seeking; Ham: harm avoidance; RD: reward dependence; P: persistence; SD: self-directedness; C: co-operativeness; ST: self-transcendence.
University students in Kumamoto, Japan were solicited for participation in a two-wave study. Questionnaires were distributed to new students in May after they enrolled in college (T1) and again five months later (T2). At Time 1, 525 questionnaires were distributed out of which 240 (46%) usable questionnaires were returned. Of these students, 184 (77%) students responded at Time 2. The data of these 184 students were used for subsequent analyses. There were no differences between the students who responded at two occasions (
The TCI [
The Hospital Anxiety and Depression Scale (HADS [
We calculated the mean and standard deviation of each variable used in this study and examined the correlations between each pair of variables. We then constructed a path model using SEM based on our research hypotheses (Figure
Covariances were added according to greater modification indices if such additions fit clinical and research assumptions. The fit of the model with the data was examined in terms of chi-squared (CMIN), comparative fit index (CFI), and root mean square error of approximation (RMSEA). According to conventional criteria, a good fit would be indicated by CMIN/df < 2, CFI > 0.97, and RMSEA < 0.05, and an acceptable fit by CMIN/df < 3, CFI > 0.95, and RMSEA < 0.08 [
This project was approved by the Ethical Committee of Kumamoto University Graduate School of Medical Sciences.
The correlations between all variables used in the present study are shown in Table
Correlations between variables used in the path analysis (
1 | 2 | 3 | 4 | 5 | 6 | 7 | 8 | 9 | 10 | 11 | 12 | 13 | |
---|---|---|---|---|---|---|---|---|---|---|---|---|---|
(1) Gender (male—1; female—2) | — | ||||||||||||
(2) Age | −0.09 | — | |||||||||||
(3) NS | −0.08 | 0.03 | — | ||||||||||
(4) HA | 0.06 | −0.02 | −0.19* | — | |||||||||
(5) RD | 0.38** | 0.07 | 0.14 | 0.08 | — | ||||||||
(6) P | 0.09 | 0.00 | −0.19* | −0.05 | 0.15* | — | |||||||
(7) SD | −0.02 | −0.06 | −0.25** | −0.36** | 0.02 | 0.09 | — | ||||||
(8) C | 0.36** | 0.07 | 0.01 | 0.07 | 0.57** | 0.13 | 0.24** | — | |||||
(9) ST | 0.09 | −0.07 | 0.20** | −0.23** | 0.09 | 0.15* | −0.14 | 0.13 | — | ||||
(10) T1 HADS-D | 0.07 | 0.03 | 0.10 | 0.35** | −0.08 | −0.05 | −0.54** | −0.18* | 0.13 | — | |||
(11) T1 HADS-A | −0.07 | 0.01 | −0.01 | 0.24** | −0.30** | −0.15* | −0.43** | −0.40** | 0.01 | 0.53** | — | ||
(12) T2 HADS-D | 0.02 | 0.05 | 0.11 | 0.24** | −0.11 | 0.06 | −0.43** | −0.16* | 0.15* | 0.56** | 0.34** | — | |
(13) T2 HADS-A | −0.10 | −0.01 | 0.10 | 0.28** | −0.28** | −0.11 | −0.43** | −0.32** | 0.03 | 0.48** | 0.56** | 0.61** | — |
Mean | 1.67 | 18.3 | 26.3 | 34.9 | 31.6 | 16.6 | 39.0 | 50.5 | 17.1 | 5.4 | 4.3 | 5.5 | 4.6 |
Standard deviation | 0.47 | 0.9 | 6.5 | 7.4 | 6.0 | 3.7 | 8.1 | 7.8 | 6.4 | 3.6 | 2.8 | 3.7 | 3.2 |
Cronbach alpha | — | — | 0.72 | 0.78 | 0.72 | 0.62 | 0.78 | 0.82 | 0.80 | 0.78 | 0.61 | 0.76 | 0.69 |
NS: novelty seeking; HA: harm avoidance; RD: reward dependence; P: persistence; SD: self-directedness; C: cooperativeness; ST: self-transcendence. *
T1 and T2 HADS-D and HADS-A were significantly correlated with HA and inversely with SD and C. In addition, T1 and T2 HADS-A were inversely correlated with RD. T1 HADS-A was correlated inversely with P, and T2 HADS-D was correlated with ST (Table
Because there were significant correlations between many of the variables examined in this study, we created an original path model (Figure
Our final path model can be regarded as good: CMIN/df = 2.01, CFI = 0.991, and RMSEA = 0.074 (Figure
Final SEM model NS: novelty seeking; Ham: harm avoidance; RD: reward dependence; P: persistence; SD: self-directedness; C: cooperativeness; ST: self-transcendence. Paths without significance (<0.05) are not shown in the Figure (but were not deleted).
A primary finding of this longitudinal study was that it was not the surplus component of the mood measurement but rather trait depression and anxiety that were mainly associated with TCI subscale scores. Thus, trait depression was associated with high HA and low SD, as well as high ST. Trait anxiety was similarly associated with not only high HA and low SD, but also with low RD, P, and C.
People with high HA and low SD may be characterised by depressive and anxious traits that are part of their personalities. Such people may be more likely to develop clinical mood and anxiety disorders under stressful life situations. Previous studies have found links between high HA and low SD on the one hand and mood and depressive disorders on the other, but these findings may be biased by the fact that trait and surplus components were measured simultaneously as state components.
In this study, we observed a significant link between trait depression and high ST. This link has not been consistently reported. However, some investigators have found that people with bipolar disorder were characterised by high ST in addition to high HA and low SD [
Unlike trait depression, trait anxiety was also associated with low RD and C. It has been reported that low RD and C are characteristics of anxiety disorders such as obsessive compulsive and phobic disorders. The results of the present study are in line with these findings. People who are low in RD are practical and cold and can be withdrawn and detached. People who are low in C are socially intolerant, critical, revengeful, and destructive and may thus be more anxious when relating to others. We consider high HA and low SD to be personality traits that are related to dysphoric mood in general, including depression and anxiety, whereas low RD and C are specific to anxiety.
Strengths of the present study include a longitudinal research design and the statistical separation of the trait and surplus components of depression and anxiety. Past studies have usually treated scores of mood measures as state indicators. In our view, they fail to distinguish the temporary reactive components of dysphoric mood from the stable ones. The former are more likely to be induced and maintained by immediate environmental factors such as stressful life events and enacted social support. In contrast, the latter is more likely to be dispositional. Our finding that TCI profiles were associated with trait components is in agreement with this notion. We expect that reanalysing the data produced by past investigations using our current statistical methods may cast more light on this issue.
Limitations of the present study should be noted. First, our sample size was modest. Given that SEM is a statistical method that requires a large sample size, further studies using a larger sample size are necessary. However, the good fit of the model to the relatively small current sample is encouraging. Second, we used a university student population. University students between 18 to 30 years of age constitute the study population. A few students were out of range of youth in this study. Nevertheless, the age range was narrow, and we should not extrapolate the data to older populations, particularly since TCI subscale scores vary with age [
The links between TCI subscales and trait depression and anxiety are the result of an inherent association combined with the effects of mood state on the self-report of personality. This issue is difficult to disentangle using the research design employed in the present study. One possible way to cast light on the distinction between the two components is to use multiple raters of each individual’s personality, for instance the participant him- or herself as well as those who know him or her well (e.g., family members and friends). This approach is beyond the scope of the present study but would benefit from further study.
Taking into consideration these shortcomings, the present research showed that TCI profiles were associated with the trait components of depression and anxiety rather than their surplus components. Depression and anxiety traits shared several of the same TCI characteristics, including high HA and low SD, but differed in specific details, for instance with low C and RD being associated with anxiety only.