There is more to mental illness than negative affect: comprehensive temperament profiles in depression and generalized anxiety

Our main (mental illness-related) hypothesis was that, when an FET-based test of temperament (STQ-77) is used to assess temperament profiles, the following patterns in these profiles should differentiate between Control (healthy), Depression, Anxiety and Comorbid groups:

The effect of age on the expression of illness in temperament profiles was also examined in our study. It is well known that aging results in a progressive decline in physiological function, possibly influencing temperament and illness risk [56]. In normative samples the most consistent age-related differences in temperament traits are a decline in tempo-related characteristics [57] and sensation seeking [58]. Our age-related hypothesis was that age differences would be most significant in temperament traits related to dynamics of behaviour (tempo, plasticity and impulsivity) and sensation seeking in healthy people. It was not clear, however, if, or how, age affects the dynamical traits of temperament in anxious and depressed patients, and whether or not any age-related slow down amplifies any slow-down reported in patients with mental illness. It was also not clear if, or how, Sensation Seeking couples with mental illness and age. If age is a factor compromising the speed of behavioural integration and sensation seeking even more than mental illness, then we should observe statistically significant differences between the age groups of patients suffering from the same mental illness. The age groups were chosen for the study due to their correspondence to specific age-related functional tasks and stages of life: age 18–24 (transition from dependence on parents to independent living); age 25–45 (establishment of social, financial and psychological independence; building a professional career and setting up a family); age 46–65 (focus on career and exploration of personal interests; raising of children in their stage 1–2); age 66–85 (transition to retirement and subsequent retirement).

The intake records of 687 Canadians aged 18–85, clients, patients in treatment and associates of a private psychiatric and psychological practice, Psychological Services 4018, were examined for this study. The practice serves Hamilton, Niagara Falls, Haldimand-Norfolk, Mississauga and Toronto areas, and the sample represents these five distant locations. The practice has its own Late Life Memory Clinic operating under the Haldimand War Memorial Hospital (Dunnville, Ontario) that provides testing and screening for dementia in clients and patients over 60. The final sample excluded subjects with dementia and protocols with poor validity scores (STQ Validity score > 13).

The demographic data of the sample (N = 687, M/F = 294/393, Mean age/SD = 43.75/19.27) is given in Table 1.

All participants of this study signed a consent form allowing the use of their intake forms for research purposes. Clients of our practice completed their intake forms (including questionnaires) during their initial contact with the Practice. Those patients whose protocols were selected for MDD, GAD or comorbid group were in an active state of illness, regardless of previous or subsequent history of illness. It is important to note that this study focused on the coupling between the acute state of mental illnesses and temperament profiles and did not extend to the topic of temperament differences in the subsequent course of illness. Thus we did not differentiate between the protocols of clients who did or did not experience subsequent remission of their illnesses. Diagnoses in the mental illness groups were made using DSM-5 criteria based upon a structured DSM clinical interview, file review to obtain additional past history (if available), observation throughout the course of treatment (subjects were generally followed over the course of many weeks to years) plus the use of additional confirmatory tests. Cut-off scores for tests applied are listed in Table 1. These additional confirmatory tests varied between sites.. Subjects having a past history or current symptoms of Bipolar Disorder were excluded from this study.

Control subjects consisted of those clients who presented to the practice requesting counselling for problems such as marital and/or family dysfunction, problematic children or elderly parents, vocational choices, situational stress at work, etc. as well as healthy volunteers and their family members. Control subjects were all assessed to ensure that they did not meet the criteria for any DSM-5 diagnosis. Three control participants reported symptoms consistent with Adjustment Disorder (i.e., according to the DSM-5, a condition that lasted no longer than a month and should not be considered as a consistent behavioural pattern).

During either intake testing (for patients and clients) or research (for healthy participants), each person completed the Compact Structure of Temperament Questionnaire (STQ-77) [18, 50, 51]. The complete validation history can be found in [18, 50, 51, 54]. The STQ-77 consists of 77 statements, assigned to 12 temperament scales (6 items each) (corresponding to the FET model) and a validity scale (5 items, addressing social desirability bias), which are listed below. Results within the range of 15–20 on the validity scale were considered invalid as the respondents were likely to demonstrate a positive impression bias in their responses. Subjects responded according to a 4-point Likert scale format: (1) “strongly disagree,” (2) “disagree,” (3) “agree,” (4) “strongly agree.” The structure of this test, a description of the 12 temperament scales and their reliability indices as calculated for this data are given in Fig. 1.

In all ANOVA processing, post-hoc comparisons were performed using both the Tukey and Fisher LCD tests with an alpha level of .05. The partial Eta-squared values (η ² ) were also calculated as an additional metric of effect size for all significant ANOVA contrasts.

The dataset supporting the conclusions of this article is available from McMaster University, Faculty of Health Sciences server: http://​fhs.​mcmaster.​ca/​cilab/​DataDeprAnx.​xls.

The most significant age effects (at p < 0.001) in the Control (healthy) group were found for the scales of Sensation seeking, Motor-Physical Tempo and Social-verbal Tempo, as well as a moderate effect (at p < 0.01) for the scale of Intellectual Endurance (Table 4, Fig. 6). The last effect was, however, not reflective of linear relationships and followed an inverted U shape, with the highest Intellectual Endurance (ability for sustained attention) reported not by the youngest group, but by the healthy participants in AgeGroups 2 and 3 (aged 25–45 and 46–65).

Overall, Age showed no significant interaction effects with the factor Diagnosis suggesting that mental illness affects people of different ages in similar ways. The only strong (at p < 0.001) effect was found in differences between Comorbid and MDD patients aged 25–65 on the scale of Social-verbal Tempo: Comorbid patients reported low scores on these scales in both Age 2 and Age 3 groups, whereas MDD patients in the Age 2 group (25–45 years old) reported significantly higher Social Tempo, while the Age 3 (46–65) group in MDD patients reported Social Tempo which was even lower than in the Comorbid group (F(1,184) = 11.61; η ²  = .059). This age-related drop in the scores of MDD patients likely contributed to weak (at p < 0.05) Age-Diagnosis interaction effects on this scale in MDD vs Control (F(2,332) = 3.81; η ²  = .022, in 3 age groups) and MDD vs GAD (F(1,173) = 6.62; η ²  = .037, in 2 age groups) comparisons.

There were a few other weak (at p < 0.05) Age-Diagnosis interaction effects: on the scale of Intellectual Endurance in comparing Comorbid and MDD patients (F(1,184) = 4.04; η ²  = .021) and in compariing GAD and MDD patients (F(1,183) = 6.11; η ²  = .034), patients in the Age 3 group estimated their ability to stay focused (i.e. Intellectual Endurance) significantly lower than other same-age participants, including patients with other diagnosis. Weak Age-Diagnosis interaction effects were found in comparisons between Comorbid and Control groups on the Sensation Seeking scale (SS) (F(1,264) = 4.29; η ²  = .039) and on the Empathy scale, EMP (F(1,264) = 6.32; η ²  = .023) - younger participants with mental illness had lower scores on these scales than their healthy peers. In comparisons between GAD and Control groups on the Social Endurance scale, anxious participants in the Age 2 Group reported higher scores than anxious participants in either Age 1 or 3 Groups, and this effect was especially profound in men (F(2,391) = 4.29; η ²  = .021, 3 age groups, Fig. 3).

The results related to age differences in temperament traits are summarized in Fig. 6 and Table 4.

In terms of sex differences, healthy men reported significantly higher scores in Physical Endurance compared to women, and lower Social Tempo (Table 4, Fig. 7b). There were no significant (at p < 0.01) Sex x Diagnosis interactions for Depression and Anxiety. There were weak Sex x Diagnosis interactions for the Comorbid group on scales of Sensation Seeking (F(1,78) = 4.277; p = .040; η ²  = .021), Self Confidence (F(1, 225) = 5.690; p = .018; η ²  = .028) (lower in women), and the scales of Social Tempo (F(1,78) = 6.572; p = .011; η ²  = .031), Neuroticism (F(1,78) = 5.615; p = .019; η ²  = .027) (higher in women).

As noted in the Introduction, previous attempts to find differential effects of Depression and Anxiety on temperament profiles relied mostly on emotionality traits (such as Positive-Negative Affect, Neuroticism). and this approach proved unsuccessful for differentiating between these major disorders. We argued that the use of a temperament model differentiating between physical, social and intellectual aspects of behavior, as well as between energetic, dynamical and orientational aspects, should result in a more nuanced discrimination of mental disorders. Indeed, when the FET framework was used to study temperament profiles, MDD, GAD and Comorbid depression and anxiety all showed distinct patterns of coupling with temperament profiles which can be distinguished one from another (Figs. 2, 3, 4, 5 and 7a, Tables 2 and 3). The structure of the FET model appeared very efficient in matching the “signature patterns” of temperament profiles of the clinical samples to the symptoms of the corresponding mental disorders as described in the DSM/ICD.

First, in line with our hypothesis and the DSM symptom of fatigue, MDD patients did indeed report lower scores on all three endurance-related scales compared to Controls and to subjects with Anxiety. Second, in line with the symptom of psychomotor retardation, MDD patients also had lower scores for all of the dynamic traits (Motor Tempo, Social Tempo and Plasticity).

This is consistent with the theory that diminished activity in serotonergic systems is linked to depression [56], and with the predictions of the FET model, since serotonin systems are major regulators of the energetic maintenance of behaviour [19, 20], and therefore there should be a decline in FET energetic traits. The serotonin system has long been implicated in the endurance aspects of activities although the relationship is complicated [59‐61].

In contrast to these findings, Depression had less effect on the scales related to behavioural orientation to specific reinforcers (especially the scales of Empathy and Sensitivity to Probabilities). This finding points to possible difficulties in CBT if the focus is on cognitive constructs in depressed patients’ perception, i.e. orientational and not executive aspects of behaviour.

These differential effects show the benefits of distinguishing between endurance- and behavioural orientation-related traits when mapping symptoms of mental disorders.

In contrast to the FET model, the Cloninger model and the Clark-Watson model of Positive/Negative Affects lack endurance- and speed-related components (an exception is the scale of Persistence in the Cloninger model). Both Persistence in the Cloninger model and Plasticity in the FET model relate to facility in changing behavioral programs, each residing on the opposite poles of this characteristic. Differences in the items selected for these two scales were, perhaps, the reasons why studies using the Cloninger model did not show differences on the Persistence scale in MDD and GAD patients, in comparison to controls.

In line with the second part of our hypothesis and consistent with the DSM/ICD symptoms, GAD was associated with high Neuroticism scores. Moreover, Anxiety was also associated with significantly lower Social Endurance and Social Tempo (compared to Controls), and this is consistent with the commonly reported symptom of social withdrawal [10]. Compared to Controls, Anxiety patients also had significantly lower scores on the scale of Sensitivity to Probabilities (i.e. a low ability to comprehend the context and causal relationships between events). The differences in effects between MDD and GAD show the benefits of the activity-specific approach in temperament, distinguishing between traits related to physical, social and mental aspects of behavior. No other temperament model uses this differentiation but in the development of joint taxonomies for mental illnesses and temperament it might be crucial. The lower scores on Motor-Physical scales in patients with Depression might be considered as a signature in MDD profiles, while lower scores on the Social Endurance and Tempo might be indicators of Anxiety.

From a clinical standpoint, Comorbid depression and anxiety presents as a more serious disorder than MDD or GAD, with more severe symptomatology, a prolonged course and diminished treatment response [55]. Thus it was hypothesized that subjects with Comorbid depression and anxiety should report the lowest scores in relation to all aspects of activities. The results indeed showed significantly lower scores (compared to Controls) on 10 out of 12 scales (with the exception of Empathy and Sensation Seeking) in patients with a Comorbid diagnosis. The Comorbid MD and GAD group showed its own differential pattern in temperament profiles. Plasticity, sustained attention (measured as Intellectual Endurance), probabilistic processing (measured as Sensitivity to Probabilities) and impulse control (measured as Impulsivity) – are traditionally linked to the frontal cortex, and exactly these four traits showed very significant effects (Comorbid compared to Control). A differentiation between traits regulated by the frontal cortex (regulating implicit, probabilistic and complex aspects of behaviour), from traits regulating physical and verbal aspects of behavior was not proposed in either the Clark-Watson model or Cloninger model, but apparently such differentiation has practical value. The degree of negative change in these aspects of behavior differentiates between MDD, GAD and Comorbid depression and anxiety.

In our study, patients with a history of Bipolar Disorder, past or present, were specifically excluded and Comorbid subjects were treated separately. The very lage increase in Impulsivity in the Comorbid group is thought to represent a synergistic interaction between Depression and Anxiety. This is opposite to what would be expected in the presence of fatigue or psychomotor retardation in Comorbid patients, but consistent, however, with the FET hypothesis, which suggests that when executive capacities decrease, the plasticity of behaviour should be negatively affected, and impulsivity should increase. This in fact was observed. These results can be compared to the prediction of another temperament model, Gray’s Reinforcement Sensitivity Theory (RST). The RST describes two regulatory systems: the Behavioural Activation System (BAS) and the Behavioural Inhibition System (BIS) [16]. According to this model impulsivity occurs whenever there is an excess of BAS activation over BIS, while anxiety-neuroticism occurs whenever there is an excess of BIS over BAS. In Gray’s model, impulsivity cannot be a symptom of anxiety because anxiety and impulsivity arise in mutually exclusive states of BAS-BIS balance. Our results were not in line with the predictions of the RST, but were in line with the FET model. Moreover, several reports have also described elevated impulsivity in anxious and depressed patients [11, 37‐39]. Thus an elevation of Impulsivity and a decrease in Plasticity may represent new symptoms of Comorbid depression and anxiety which warrant consideration for subsequent versions of the DSM/ICD. More general features differentiating between Major Depression and Comorbid depression and anxiety may be declines in the functions of cortical areas emerging as compromised plasticity, an inability to focus, impaired probabilistic thinking, and elevated impulsivity. This warrants further study.

Finally, consistent with the literature [31‐33, 37‐39], all three illness groups reported significantly higher scores on Neuroticism, but the Anxiety group reported the greatest scores on this scale. This converges with findings of higher Harm Avoidance in studies using the Five-Factor [12, 13, 31], Akiskal [34, 35], Mehrabian [9], and Cloninger models [14, 30, 42, 43] and Negative Affectivity using the Clark and Watson model [6, 10, 22, 24, 25]. This shows quite clearly that elevated Neuroticism by itself is a non-specific indicator of the three illnesses considered here although the degree of Neuroticism might be diagnostic. Very high Neuroticism might be a “signature” symptom of GAD or Comorbid depression and anxiety, and moderate Neuroticism might be merely a symptom of the presence of mental illness in general.

There are subtleties in the effects that these three disorders have on the Emotionality traits. As noted, Comorbid depression and anxiety was associated with much higher Impulsivity than the other two disorders while Anxiety was associated with much higher Neuroticism than the others two diagnoses. Depression and Comorbid depression and anxiety were both associated with lower scores in Self Confidence but the decline was greatest in the latter disorder.

The ability of the FET framework to differentiate between these three most common mental disorders demonstrates the value of the activity-specific approach in temperament research that could be employed for future classifications. Moreover, the current study has implications for modifying the classification OF and criteria for MD and GAD:

The analysis of age differences in the temperament profiles associated with three mental disorders was conducted using four adult age groups. Surprisingly, there were practically no Age x Diagnosis interaction effects (i.e. there were no age differences in the way in which mental illness affected biologically-based aspects of behavioural regulation (temperament traits) of our participants). It appears that people of various ages suffer from mental illness in a similar fashion, and this was true for all three diagnoses. This is also in keeping with clinical findings that the expression of illness remains fairly consistent across the full age range.

There were, however, effects related to the factor of Age per se. Older age in normal subjects was associated with significantly lower scores for Physical Tempo, Social Tempo, Intellectual Endurance, Sensation Seeking, Social Tempo, and an increase in Neuroticism (Fig. 6). Lower Sensation Seeking in older age groups has been noted previously [58]. For Physical Tempo, Intellectual Endurance, lower scores were prominent only in the elderly age group, which parallels the fact that significant physical and intellectual decline does not occur until quite late in life [56, 57]. It is well known that there are differences in the presentation of anxiety and depressive illness in older age groups compared to young. The low occurrence of Depression in our youngest subsample (age 18–24) and the low occurrence of GAD in our oldest subsample (age 65–84) might be important observations by themselves.

As this was a cross sectional, rather than a longitudinal study, these results are only indirect evidence of the absence of noticeable differences across four age cohorts in the way in which mental illness affects temperament profiles.

Gender appears to play a fairly minor role in the expression of these effects. Regardless of the presence of depression or anxiety, men reported significantly higher scores for Physical Endurance compared to women, consistent with well known physical (constitutional) differences between the sexes [58, 62, 63]. Women on the other hand reported significantly higher scores for Social Tempo than men (Fig. 7b). There were no significant Gender x Diagnosis interactions for Anxiety or Depression. This was quite interesting given the commonly held view that women are more emotionally vulnerable than men. In fact the only significant Gender x Diagnosis interactions occurred for Comorbid depression and anxiety involving Sensation Seeking, Self Confidence (lower in women) and Social Tempo, Neuroticism (higher in women). It is only in the presence of the more severe disorder that men and women begin to separate out significantly. This suggests that healthy men and women do not fundamentally differ in terms of their emotionality.

Sex differences in response to mental illness appeared to be related mostly to those traits on which men and women differ in their younger years, that is Physical Endurance (higher for men) and Social Tempo (higher in women). Indeed in Depression, younger men showed a greater decline in Physical Endurance compared to women, while women showed a greater decline in Social Tempo compared to men. These declines may be a direct result of the depression on behaviour regulation but one might speculate that they may also be due to the impact of depression on a person’s self judgment, which might make them overly critical of losses in the most prominent traits that participate in determining their sense of self. Men tend to see themselves as physically strong while women tend to view themselves are being more sociable and so these traits may be more vulnerable to negative judgment especially in younger people whose sense of self might not have stabilized.

Further studies will hopefully continue to assess the relationship between temperament profiles and symptoms of mental disorders, however the authors hope that current teams working on improving the DSM and ICD will consider the possibility of presenting both, mental disorders and healthy temperaments along a continuum, using the same components (formal descriptors or scales). We believe that the FET model, being supported by results in current psychophysiology [1, 19, 20, 47] provides a good start for developing a universal matrix for clinical and healthy psychological taxonomies of individual differences.