gms | German Medical Science

Researchers believe that there is a close connection between health disturbances and emotional adjustment problems. Particularly the combination of suppressed anger and animosity could, within a given social context, contribute to an elevated cardiovascular risk [8], [9]. These studies clearly show that the inability to express emotions is per se not pathological, but only when the social situation is stressful and the need to express emotions is suppressed. Pennebaker [10] also refers to this as “active inhibition“ and King and Emmons [11] as “ambivalence over emotional expressiveness“.

The theory developed by King and Emmons [12] stipulates that emotional ambivalence reflects the conflict between one’s need to express emotions or the requirements of the situation to do so, respectively, and the desire not to display subjective emotions. This inner ambivalence – as to displaying emotions in certain situations or not – is considered stressful and regarded as an origin of abnormal processes, especially if it is not only a short-term adjustment of the behavior, but a habitual cognitive mechanism. People using such mechanisms have difficulties communicating their emotional needs and also have more problems in relationships that could possibly entail social consequences. These, in turn, could increase the stress level and hence lead to diminished social support and lower self esteem [2]. Emotional ambivalence can manifest itself in various ways: Persons using such mechanisms (1) might want to express emotions, but do not feel capable of doing so, or (2) might express emotions they are not convinced of, or (3) might regret having expressed emotions at a later point in time.

King and Emmons developed an instrument to diagnose and operationalise emotional ambivalence: “The Ambivalence Over Emotional Expressiveness Questionnaire (AEQ)“ [12], [13]. It includes 28 items, allowing to label various aspects of cognitive ambivalence or conflicts pertaining to emotional expression. The items verbalize the desire to show a certain emotion. They also verbalize doubts regarding one’s own ability to express emotions and the fear of negative consequences resulting from the expression of the emotion. The authors’ first principal component analysis (PCA, criterion eigenvalue >1) yielded two factors: one factor included items expressing cognitive ambivalence about the expression of positive emotions, the second factor included items expressing ambivalence about the expression of negative emotions. Based on a confirmatory factor analysis showing a high correlation of items with both factors and high intercorrelations of both factors (0.71), King and Emmons [12] concluded that their instrument is one-dimensional and offers the general factor Emotional Ambivalence. The English version of the AEQ is said to have high internal consistency (Cronbach’s α 0.89 [12], [14]) and test-retest reliabilities (Pearson correlation coefficient 0.78 after 6 weeks [12]).

After having translated the 28 items of the AEQ [12] into German, Traue [1] developed a German short form (AEQ-G18, formerly referred to as FEMKO). They conducted their first PCA with a sample consisting of 164 university employees. This analysis had an eigenvalue criterion of >2 and, contrary to the original version, a two-factorial outcome. Therefore, items that showed no clear or only a very low loading towards their respective factor were excluded, resulting in a version with 18 items and the two scales effect ambivalence and competence ambivalence. Effect ambivalence refers to fear of repercussions from the expression of negative emotions, whereas competence ambivalence refers to doubts and insecurities regarding the ability to show positive emotions. The two-factorial structure was confirmed by additional studies with 248 college students. A third study involved 115 general medicine patients (mean age 37.8). In all samples, both scales and the overall scale had satisfactory to high reliability coefficients [2].

The AEQ-G18 has so far not been tested on a representative sample of the German population. Therefore, we attempted to replicate the two-factorial structure of this questionnaire with such a sample. In addition, such an approach will additionally verify the statistical parameters for the scales, the influence of socio-demographic factors as well as provide normative values for the AEQ-G18. Based on our findings [2], we had suspected a higher degree of emotional ambivalence in women. Since there is also a lack of information concerning the impact of education on emotional ambivalence as well as possible differences between people living either in East or West Germany, we also explored these aspects.

Furthermore, we tested some aspects of the validity of the AEQ-G18 based on the short form of the Patient Health Questionnaire PHQ-D [15], the Depression Screener DEP-2 [3], the Profile of Mood States POMS [4], the revised Beck Depression Inventory BDI [5], and the Health Survey Questionnaire SF-36 [7]. Based on the existing findings for the AEQ-G18, we hypothesized that ambivalence is positively correlated with depressiveness and reduced psychological state of health (depression, fatigue, and anger) and negatively correlated with health-related quality of life.

Based on a request of the University Leipzig, the data used for this study was obtained in the fall 2002 by the polling institute USUMA Berlin as part of a representative multipurpose survey. The representativeness of the sample was verified by USUMA through comparisons with other samples and with information provided by the Federal Statistical Office. Trained interviewers visited the study participants in their homes. The study participants were given questionnaires that were to be completed without any assistance. Participation was voluntary and each of the participants was handed a data protection guarantee, signed by the interviewer. This survey is based on persons living in private households in Germany over the age of 14.

The survey’s response rate was 66.5%. Of the 2066 persons surveyed, we only used information provided by German citizens. The socio-demographic characteristics of the remaining 2043 participants are listed in Table 1 [Tab. 1]. Since the survey was structured for a comparison between East and West Germany, the amount of East German participants is disproportionate to the overall demographics (1009 participants in East Germany compared to 1034 in West Germany).

The AEQ-G18 [1] includes 18 out of the 28 items of the original English AEQ [12], [16], offering five response options (0 = “never“ to 4 = “always“). In addition to the two scales effect ambivalence and competence ambivalence, we also calculated a total score. The ten items of the scale effect ambivalence pertain to consequences of expressed emotions and refer mostly to negative emotions, whereas the eight items concerning the scale competence ambivalence describe the ability to express feelings.

For a clinically normal sample of 164 participants that had been recruited from professional and private contacts of university employees, the internal consistencies (Cronbach’s α) were 0.80 for the scale competence ambivalence, 0.77 for the effect ambivalence and 0.77 for the total scale [2]. Intercorrelations between the scales were at around 0.2.

A sample of 248 medical students showed somewhat higher internal consistencies: 0.85 for the scale competence ambivalence, 0.79 for effect ambivalence, and 0.88 for the total scale [2].

Even a sample of 115 patients in a general medicine practice confirmed the results for Cronbach’s α: 0.84 for the scale competence ambivalence, 0.76 for the effect ambivalence and 0.87 for the total scale [2]. To our knowledge the test-retest-reliability has not been verified yet.

The PHQ is used to diagnose depression and anxiety. In the German short version of the PHQ [15], the depression module PHQ-9 [6], [17] has 9 items. These are the only items we used for this study. The items offer four different response options. The depression score is calculated by adding up the individual score for each item (internal consistency: Cronbach’s α=0.89).

The revised Beck Depression Inventory by Schmitt & Maes [5], [18] is a shortened version of the original and includes 20 items [19]. A scale with six different response options tries to capture the respondent’s current attitude towards life. The sum of all items forms the total BDI score (internal consistency: Cronbach’s α=0.90).

Participants also answered a depression screener [3] consisting of two questions based on a five-point scale geared towards depressive core symptoms that had been developed by the “U.S. Preventive Task Force“ [20]. The depression score DEP-2 is calculated by adding the individual scores and should be considered reliable (correlation of both items r=0.80).

This questionnaire [4] is used to determine a person's mood and attitude. We used the German short form of the POMS with 35 items and 7 response options for each item. The 35 items constitute the following 4 scales (internal consistency Cronbach’s α between 0.89 and 0.95): depression/anxiety (14 items), fatigue (7 items), vigor (7 items), and anger (7 items) [21].

Internationally, the SF-36 [22] is the most frequently used method to determine health-related quality of life. Its 36 items with varying response options can be allocated to the following dimensions: ability to function physically, physical role functions, (absence of) physical pain, overall perception of health, vitality, ability to function socially, emotional role behavior and mental well-being. This study used a revised form of the SF-36 [7], [23]. Internal consistencies (Cronbach’s α) ranged between 0.81 and 0.94 for the subscales.

The data was analyzed by using the software package for statistics SPSS for Windows 10.0. The analysis included all questionnaires that had only one item missing. The missing item was replaced by a mean score. The effect size was calculated as the difference between the groups’ means and the pooled standard deviation. According to Cohen [24], effect sizes of >0.20 represent a weak, >0.50 an average, and >0.80 a strong effect.

Descriptive statistics for each item of the AEQ-G18 are given in Table 2 [Tab. 2]. The items were positively skewed with item means varying from 1.03 to 1.71. The participants used the entire response spectrum from 0 to 4 for all items.

A principal component analysis (PCA) led to a solution with these eigenvalues (only the first six of the eighteen values are listed) in descending order: 6.93 - 1.38 - 1.13 - 0.91 - 0.84 - 0.76. The Kaiser-criterion (eigenvalues >1) suggested three factors (explained variance 52.4%), with the first factor accounting for 38.5% of the variance, the second factor for 7.6%, and the third factor for 6.3%. This result did not confirm the two-factorial structure that is postulated for the AEQ-G18 (cf. Deighton & Traue [2]). Comparably, the Scree-test did not support the postulated solution, but indicated a one-factorial result. However, a two-factorial structure seems plausible, when comparing the above mentioned eigenvalue progression within the context of a parallel analysis [25] with an eigenvalue progression that can be expected by factorizing correlation matrices that are generated with random numbers. We arrived at the factor loadings listed in Table 2 [Tab. 2] after orthogonal rotation (explained variance 46.1%), when entering two factors to be extracted in accordance with the original version.

First and foremost, it needs to be stated that, except for item no. 4, all other seven items of the scale competence ambivalence clearly loaded on the first factor (cf. highlighted values in Table 2 [Tab. 2]). But two items (no. 1 and 11) also showed substantial loadings of >0.30 on the second factor.

The results are less clear for the ten items of the scale effect ambivalence. Of these, three items seemed to load more on the first factor, three other items primarily loaded on the second factor with the remaining four items noticeably loading on both extracted components.

The following sections describe the results for the total scale (total score) and the two scales effect ambivalence and competence ambivalence. Due to the apparent one-factorial results, we additionally developed a short form AEQ-G10 (see below), and, for the sake of clarity, included the results in the following tables.

Table 3 [Tab. 3] includes the descriptive statistics for the scales of the AEQ-G18 along with information regarding internal consistency (Cronbach’s α). The correlations for both scales of the AEQ-G18 with r=0.75 (Spearman) were high and significant (p<0.001). The respective correlations with the total score were r=0.93 for competence ambivalence and r=0.94 for effect ambivalence.

The three-factorial covariance analysis with the factors “gender,” “East-West affiliation” and “educational background”, and the covariate “age” showed that the factor “education” had a significant main effect on the total score and on the scale competence ambivalence, whereas the factor “gender“ had a significant main effect on the scale effect ambivalence (cf. Table 4 [Tab. 4]). Women had higher scores on the scale effect ambivalence. Male and female participants without college education indicated more competence ambivalence and more emotional ambivalence overall (total score). The other factors were not statistically significant. Overall, however, the effect sizes indicated only weak effects.

We were able to replicate the expected positive correlations (cf. Table 5 [Tab. 5]) between emotional ambivalence and depressivity and a negative attitude/mood (depression, fatigue and anger) as well as the expected negative correlations between emotional ambivalence and quality of life and positive attitude/mood (vigor). The obtained positive correlations were higher compared the negative correlations.

We developed a short version by gradually reducing the amount of items, until the questionnaire had a clearly one-factorial structure. The item reduction was based on the item total correlation coefficients, beginning with the items with the lowest coefficient. After 6 items had been removed, the survey was clearly one-factorial. The amount of items was further reduced by eliminating items with similar content (items 10 and 14; items 13 and 17). The version without items 13 and 14 achieved the highest values for Cronbach’s α (0.87). The PCA with these 10 items (cf. Table 2 [Tab. 2]) resulted with one eigenvalue >1 accounting for 46.1% of the variance.

For the short form AEQ-G10, we conducted a three-factorial covariance analysis with the factors “gender,” “East-West affiliation” and “educational background”, and the covariate “age” and were only able to determine a significant main effect for the educational factor (cf. Table 4 [Tab. 4]). Participants without any college education displayed more emotional ambivalence than those, who had at least some college education. However, these differences were within the range of weak effects (cf. Table 4 [Tab. 4]).

The total score of the AEQ-G10 correlated significantly (p<0.001) with the respective total score of the long form AEQ-G18 (r=0.96) as well as the scales effect ambivalence (r=0.86) and competence ambivalence (r=0.96).

Table 6 [Tab. 6], Table 7 [Tab. 7], Table 8 [Tab. 8] and Table 9 [Tab. 9] list the percentile norms for all scales of the AEQ-G18 and the AEQ-G10, both for the total samples and differentiated by age and gender.

The present study used a representative sample of German citizens to examine the German version of the Ambivalence Over Emotional Expressiveness Questionnaire AEQ-G18 [1] with regard to its factorial structure, the statistical value of the scales, and the impact of socio-demographic factors. It should be noted, however, that the East-German participants were disproportionably represented in the sample.

Notably, we were only partially – and mostly within the context of competence ambivalence – able to confirm the two factors postulated by Traue et al. [1], [2]. In our sample, the two factors had a correlation of r=0.75, which was significantly higher compared to correlations obtained in preliminary studies [2]. This result suggests that both scales capture similar aspects of emotional ambivalence (comparable to the American studies [12]) and hence alludes to an underlying general factor. At the same time, it furthermore indicates that the structure of emotional ambivalence does change with different samples. Therefore, it would be beneficial to conduct additional studies, especially with clinical samples.

The simultaneously presented short form AEQ-G10 was optimized to be one-dimensional. Therefore, it is easier to use for test batteries or screenings, because the common variance with the AEQ-G18 is about 92%. The collected representative data point more towards a one-factorial rather than a two-factorial solution. The patient data indicated such a tendency that might be clinically relevant. The limitation to one factor possibly neglects the option for a clinically relevant internal differentiation. Preliminary research was able to show that somatization is more closely related to effect ambivalence whereas deficits in the social network of the examined patients more closely reflect competence ambivalence. These effects, however, are plausible from a theoretical perspective [2].

All tested scales of the AEQ-G18 as well as the short form AEQ-G10 had satisfactory reliabilities (internal consistency). The socio-demographic variables had little impact, if any, on the severity of the emotional ambivalence. There was a weak gender-related effect for the scale effect ambivalence, which seems to be more distinct in women. These findings correlated with the results of Deighton & Traue [2] (study involving university employees and patients of a general medicine practice). The study involving medical students did not show any gender-related differences [2].

Medical-psychological research has shown that a higher educational background constitutes a psycho-social, health-protective factor. Findings regarding the correlation of the AEQ with various mental and physical factors tend towards a slightly better health status and less ambivalence [2]. The socio-economic factors had an overall weak impact on our sample, but it should be noted that the educational background appeared to have the strongest influence on emotional ambivalence. Participants with a higher educational background had less emotional ambivalence, which fits into the previously outlined context.

The depression screeners DEP-2 [3] and the Profile of Mood States POMS [4], the revised Beck Depression Inventory BDI [5], the short form of the Patient Health Questionnaire PHQ-9 [6] and the SF-36 health survey questionnaire [7] showed indications for the validity of the AEQ-G18 and the short form AEQ-G10. The hypothesis was confirmed: emotional ambivalence positively correlates with depression and a reduced psychological state of health (depression, fatigue, and anger) and negatively with health-related quality of life and a positive attitude (vigor). The correlation coefficients for emotional ambivalence, depression and anger, respectively, were higher than those for the quality of life. The level of the correlation coefficients between emotional ambivalence and depression determined in this study confirmed the results by Deighton and Traue [2].

Based on the correlations between emotional ambivalence and mental and physical complaints, therapeutic interventions that address the emotional state of the patient offer specific ways to influence dysfunctional processes of emotion regulation [26]. Greenberg & Safran [27] have subdivided such interventions into the following four groups: emotional discharge, emotional insight, emotionally adaptive behavior, and exposition.

Although additional studies to validate our results remain to be done, our study indicates that the Ambivalence over Emotional Expressiveness Questionnaire AEQ-G18 [1], [2] and its short form AEQ-G10 are instruments that allow measurements of emotional ambivalence in a clinically relevant, valid, and time-efficient manner.

Emotional ambivalence within the context of a conflict between the need to express emotions and their simultaneous suppression constitutes a relevant factor for the development and persistence of mental and psychical complaints. It is possible and meaningful to operationalise the construct of “emotional ambivalence“ with both German versions of the Ambivalence over Emotional Expressiveness Questionnaire, AEQ-G18 [1] and its short form AEQ-G10.

None declared.