Modeling human temperament and character on the basis of combined theoretical approaches

Volunteers gathered from around the country gathered data from their region in the frame of standardizing the three instruments. The study included subjects from the general population who satisfied the following inclusion criteria:

The study sample included 734 subjects from the general Greek population (436 females; 59.4% and 298 males; 40.6%). Mean age was 40.80 ± 11.48 years (range 25–67 years). The mean age for females was 39.43 ± 10.87 years (range 25–65 years) while the mean age for males was 42.82 ± 12.06 years (range 25–67 years).

All subjects provided written informed consent and the protocol was approved by the Ethics committee of the Faculty of Medicine, Aristotle University of Thessaloniki, Greece.

The protocol included the gathering of sociodemographic data and the application of the TEMPS-A [25, 43‐45], the TCI [36, 46‐49] and the NEO-PI-3 R [34, 35]. On the basis of this dataset all of them were officially validated in the Greek language [50‐52] and their psychometric properties can be found in the related publications. It is also important to note that the collection of the data has been completed by 2008, before the current economic crisis began, and originally was used for the validation of these instruments.

The statistical analysis included the following:

Both EFA analyses led to the recognition of first-order factors. Following this, the factor scores were calculated for each subject by multiplying factor loadings with values of each variable and summing the results. The factor scores underwent EFA again to investigate for the presence of second-order factors. The same procedure was performed for a third time to identify third-order factors. The above analysis was performed twice, separately one time for each EFA analysis mentioned above.

Data contained in this study are available from the authors on request.

Age and gender distribution in the study sample, compared to the general Greek population according to the 2009 census is shown in Additional file 1: Table S1. The distribution of occupation in the study sample (data were available for 533 subjects) is shown in Additional file 1: Table S2. The results indicate that the study sample is representative of the country’s active population with some over-representation of younger ages in terms of the country’s population of the year 2008, when the data were collected.

The first EFA analysis returned four first-order factors and explained 68% of observed variance. The first factor included TEMPS depressive, cyclothymic, irritable and anxious temperaments, NEO-PI-3 N, and TCI HA and SD (the last with an opposite sign to the rest). The second factor included NEO-PI-3 E and O and TCI NS and RD. The third included TEMPS-hyperthymic temperament, NEO-PI-3 C and TCI ST. The fourth included TEMPS irritability, and with an opposite sign NEO-PI-3 A and TCI CO (Additional file 1: Table S3). The EFA with factor scores revealed the presence of two second-order factors; the first includes factor 1 and 3 and the second includes 2 and 4. It explained 50% of observed variance (Additional file 1: Table S4).

The second EFA analysis returned twelve factors and explained 63% of total variance, which is similar to the first EFA. The results are shown in Additional file 1: Table S5. The first factor included all TEMPS temperaments, NEO-PI-3 N1-4, N6, E6, and TCI HA1-4, SD1-5. The second included TEMPS-hyperthymic temperament, NEO-PI-3 C1-5, E3-4 and TCI Perseveration. The third included NEO-PI-3 A4, N5, C6, and TCI NS2-4, and SD5, the fourth included NEO-PI-3 E1, A1-6, C3, and TCI CO4, the fifth TEMPS-hyperthymic temperament, NEO-PI-3 O6 and TCI ST1-3, the sixth included NEO-PI-3 O1-3, O5-6 and E5, the seventh included TCI RD3, SD5 and CO1-3, the eighth included NEO-PI-3 E2, E6, RD2, the ninth included NEO-PI-3 O4 and TCI NS1, RD1, the tenth included TCI RD3, SD2, the eleventh was a residual factor (with highest loadings for NEO-PI-3-A1, Ε3 and Ν4) with very low loadings while the twelfth included TCI C5. The EFA with factor scores revealed the presence of five second-order factors and explained 40% of variance (Additional file 1: Table S6). An additional EFA on factor scores returned two third-order factors and explained 40% of observed variance (Additional file 1: Table S7).

Results of the EFA analyses and the hierarchical model which emerges are shown in Figs. 1 and 2.

The correlation matrix is shown in Additional file 1: Table S8.

The FSLRA results are shown in Additional file 1: Table S9. Highest explained percentage of variability (R²) was 52% while the lowest was 7%. Individual TEMPS subscales were predicted by NEO-PI-3 and TCI subscales, with explained variability 33–44% and 37–49%, respectively. The NEO-PI-3 subscales were predicted by TEMPS and TCI subscales with explained variability 4–52% and 22–54%, respectively. The TCI subscales were predicted by TEMPS and NEO-PI-3 subscales with explained variability 10–54% and 7–51%, respectively.

The development of the model was based on the second EFA since it is more detailed and also essentially includes conclusions from the first EFA.

The literature suggests that aspects of human personality extending beyond temperament usually include attitudes, beliefs, goals, and values. These elements seem to develop out of evolutionarily conserved temperament systems. Personality also includes higher-level cognitive functioning relatively unique to human beings (including language, abstract thought, meta-cognition, etc.). In terms of clinical utility, temperament is mostly studied in relationship to bipolar disorder [50‐64].

The model developed with the current study suggests that the basic psychological structure in humans comprises two separate super-modules placed at the top of a hierarchical structure. One reflects the perceived components of ‘self’ and the second reflects the interaction of these components with the internal representation of the environment in interaction with its properties which have been internalized and embedded in the character of the person. Both components are ‘internal’ by definition, since they reflect subjective experiences and processing of internalized descriptions and reconstructions of the environment with emphasis on the social environment. An important element is the frequent admixture of emotional and cognitive processes in the same module. Although one of these processes seems to dominate the respective module, very often a component of the other process also exists.

It is very interesting that Social Emotionality (SE) and Social Emotionality and Behavior (SEB) seem to bridge Self (S) and Self–Environment Interaction (SEI). SE contributes to S but also to SEB which in turn contributes to SEI. This ‘bridge’ denotes that the two components of ‘Self’ are kept functionally together by social emotion and the corresponding behavior corresponding mainly to parts of ‘openness’, ‘extraversion’ and ‘reward dependence’ but also with parts of ‘novelty seeking’ and ‘ (Fig. 2). In this sense cognitive processes within the S super module probably correspond to inherent pre-existing cognitive templates (biases) while the cognitive processes within the self–environment interaction) probably correspond to meta-cognition.

So far, existing models reflect processes within emotions and within cognition separately, but almost never an interaction of these two modes of psychological function. The early theories focused on activity and affective functioning which were considered to be developmentally stable. Later attention and self-control were added [65]. These later psychological functions emerged later both in evolution and also during individual development and they are probably shaped also by the environment [66, 67]. It is highly likely that the brain circuitry which serves human psychological function is extremely complex with extensive feedback, as well as with simultaneous parallel and serial processing which makes linear analysis and solutions inadequate and relatively naïve [68‐70]. In this sense, the arrows used in Figs. 1 and 2 should be considering only as marking the progress from lower to higher levels of modules rather than that of direction in the flow of information which should be considered to be largely bi-directional.

There are three dominant models of temperament and personality today and there exist significant theoretical and also essential differences between these three theoretical approaches and consequently the respected questionnaires. McCrae and Costa proposed the five-factor model (Big Five) [71] which includes neuroticism, extroversion agreeableness, openness and conscientiousness and constitutes a further development of Eysenck’s theory. The older concept of ‘psychoticism’ was substituted by agreeableness and conscientiousness while openness has some degree of overlap with extroversion [72]. Their work is largely based on the classical psycholexical study by Gordon Allport and Henry Odbert [73, 74].

The work of Robert Cloninger is characterized by an attempt to intimately connect temperamental characteristics with individual differences in genetics, neurotransmitter systems, and behavioral conditioning. He described novelty seeking (anger), harm avoidance (fear), reward dependence (attachment) and persistence (ambition) [22, 36]. His research suggests that temperament components can be assessed as early as preschool age [23] and remain moderately stable throughout a person’s lifespan except for changes from behavioral conditioning [75]. A main strength of Cloninger’s Temperament and Character model [36] is that each temperament dimension was identified and characterized as a relatively ‘pure’ and independently inherited trait that can be ascribed to the basic emotions of fear, anger, attachment and ambition or determination. Fear and anger are the most basic emotions, regulating, respectively, inhibition and initiation of behavior. In the current model harm avoidance is perceived as part of ego-resiliency, that is part of the mechanisms of self to regulate feelings of danger while novelty seeking as part of mechanisms controlling activity. Reward dependence is conceptualized as a mechanism which bridges the two others, while persistence is part of ‘ego strength’.

Hagop Akiskal has conceived temperament as the affective predisposition or reactivity, based on the original descriptions by Kraepelin (1921) of fundamental states (manic or hyperthymic, irritable, cyclothymic, anxious and depressive. These are close to the classic descriptions of Kraepelin, Kretschmer and Schneider [25]. The model of Hagop Akiskal [25, 43‐45] concerns exclusively the affective temperament modules and has been conceived while evaluating and observing mood patterns in clinical practice. Empirical research has confirmed the hypothesized four-dimensional factor structure of affective temperament and is in agreement with those previously proposed on clinical populations. Temperament traits according to this model also correspond to fear and anger and it is not surprising that all these temperament traits are included mainly within the ‘self’ module and more particularly within the ‘ego resiliency’ group while the Hyperthymic temperament is also part of the ‘ego-strength’. While Hyperthymic trait is exclusively within the ‘self’ module, Irritability contributes to the ‘self–environment’ module by participating in mechanisms controlling activity, but this contribution is rather weak. The place of these traits within the current model confirms that Akiskal’s model captures the basic affective style and mood pattern as well as identifies individuals with high risk for mood disorders [76‐78], suicidality but also various types of psychopathology (REF: Pompili M, Rihmer Z, Akiskal H, et al. Temperaments mediate suicide risk and psychopathology among patients with bipolar disorders. Compr Psychiatry 2012;53(3): 280–5). Some studies are in accord with the ego-resiliency vs. ego strength sub-organization of affective temperaments proposed in the current model [44, 79].

While personality refers to goals, coping styles, defensive styles, motives, self-views, life stories, and identities [80], basic personality traits (e.g., extraversion or neuroticism) are essentially parts of temperament [81]. Apart from these, there are three major systems of learning and memory which play a major role in the shaping of human behavior: associative conditioning of habits and skills, declarative learning of facts, and autonoetic learning of a personal lifetime narrative (autobiography) [82‐84].

The current model expands the field to include elements probably corresponding to meta-cognition mechanisms and complex interactions between affective and cognitive control on activity. It was developed exclusively after research on mentally healthy persons so it has no direct relevance to psychopathology. Future research on patient populations might provide with valuable insight concerning the areas of dysfunction in the structure of this model.

According to this model (Fig. 2), the Self (S) comprised mainly emotional (EmoS) and thought mechanisms (CogS) which seem to be relatively distinct, highly intrinsic and independent from the environment. There seems to be a significant possibility they reflect the most genetically determined traits. On the other hand, emotional functions dominate the self–environment interaction (SEI) as well as the bridging between the two super-modules, that is Self (S) and Self–Environment Interaction (SEI). The influence of ethical values (EV and EB) seems to constitute a distinct element probably influenced significantly by forces outside the person but still they are internalized. Then it is emotional function related to social tendencies (SE) which stems out of the Self and receives the influence of control mechanisms (ECC) leading to the development of a block of social emotion and behavior (SEB) which in turn is fused with ethical values (EEB) to create the SEI. Control mechanisms (ECC) seem to constitute from two distinct modules, one emotional which has to do with emotional motivation (EM) and on cognitive which probably reflects some kind of meta-cognition (SD).

The gross structure of this model suggests that at the core of psychological function are the internal emotional and cognitive processes which through social emotionality and meta-cognition determine the externalized behavior which is further shaped by internalized social factors in the form of ethical values. It is interesting that both meta-cognitive modules (ECC and EEB) are not purely cognitive but they include a strong emotional component (EM and IIE).

The presence of two super-modules at the top is in accord with previous studies, which reported similar structure but with different functions. The first such study named these super-modules as ‘alpha’ and ‘beta’ since their psychological meaning was unclear. ‘Emotional stability’ was recognized in one of them as the analog of ER, while in the other module traits of extraversion and creativity were identified probably reflecting EIE, EC and EM among others but in a very different hierarchical structure [85]. It is interesting that Digman et al. interpreting the alpha factor (which among others included emotional stability) and shares some elements with the S super-module as a ‘Social factor’ by theorizing that emotional stability and health are the direct consequence of social environment. These authors interpreted the beta factor with shares elements with the SEI as ‘personal growth and self-fulfillment related to self-actualization’ which is generally not in contrast to the findings of the current study [85]. Other authors interpreted alpha as ‘stability’ and beta as ‘plasticity’ [86] or ‘ego control’ and ‘ego resiliency’ [26, 87]. Most models suggest the presence of a module of extraversion/positive emotionality, orienting sensitivity, and affiliativeness, and of a second model reflecting negative affect versus effortful control content [88]. In general these models recognize the presence of a function of ‘effortful control’ which is similar to the EEB module of the current module while ‘Orienting Sensitivity’ could share features with SEB, but the distinction of positive vs negative affectivity modules is not in accord with our findings. The module corresponding to ‘affiliativeness’ is probably IIE and it is located at the lowest level instead of the top [89‐91].

Also the literature concerning the three major theories taken together in the current paper suggests that the four-temperament model of Akiskal [26], the cube model of Cloninger [24], the five-factor model represented by the NEO-PI [92], the seven-factor model of Tellegen [93] and Cattell’s 16 factor model [94] may in fact represent different levels of an hierarchical structure of normal and pathological personality with a two-superfactor solution at the top [26, 87], a limited number of temperaments in the middle (named under many labels, but significantly overlapping) [95] and many characters [10‐15] at the bottom. According to most conceptualizations, ‘Temperament’ corresponds to the ‘higher’ levels, while ‘personality’ and ‘character’ to the ‘lower’ [96]. In another approach, fear and anger could be used in a bidimensional model to describe affective temperament traits [97, 98].

An important characteristic of the current model is that it does not accept this hierarchical separation of ‘temperament’ vs. ‘character’ and locates both of them across all hierarchical levels and modules.

Several limitations of the study should be mentioned. First of all, there are limitations related to the application of linear methods. The method the current model was derived (orthogonal EFA) produces modules which do not correlate to each other. Thus at each level the modules are independent. However, the connection between levels is presumed to be reciprocal although the exact power of each direction remains elusive. Thus the arrows in Figs. 1 and 2 should not be considered as representing the direction of the flow of information and effect but rather they point to the corresponding higher-level module. This reflects a weakness of linear methods and might not correspond to reality, which probably reflects non-linear dynamical systems [49].

Second, the study has limitations related to the interpretation of the structure. In the literature there is a fundamental question whether modules exist at all or it would be rather more suitable to approach psychological function as a complete indivisible pattern (trait vs. profile) [65, 99]. The question is further complicated by the fact that the modules identified in the current model (and in all models) correspond to functions not to anatomical circuits, neurotransmitters or genes. Attempts to correlate temperament isolated traits and genes were promising but so far unsuccessful probably because of complex genetic mechanisms and environmental influence [64, 99‐107].

Third, the study has limitations with respect to the questionnaires used. The scales included in the current study reflect aspects of temperament, character and personality but they do not reflect all theoretical or empirical approaches. There is the possibility the model was biased towards the theories underlying these questionnaires rather than true psychological structure per se.

Future models should utilize non-linear approaches possibly with the use of network analysis and the training of neural networks. Still all these attempts will always be limited by the fact that input will be restricted to inner experience as it is perceived and described consciously by the individual.