The evidence discussed here provides important information that has contributed to the development of precise definitions of anxiety types according to their structural components and functions in controlling stimulation (see Figure 2A):

Reactive Arousal Type of anxiety is composed of:

Regulative Apprehension Type of anxiety is composed of:

Mixed Type of anxiety represents balanced apprehension and arousal elements and balanced reactive and regulative functions of stimulation processing. Speculatively, it covers specific patterns of stimulation processing of both primary types (arousal and apprehension) and their activation might be situation-dependent.

Figure 2B summarizes definitions of the three types of depression according to their structural components and the functions they play in stimulation processing.

Reactive Valence Type of depression goes with:

Regulative Anhedonic Type of depression includes:

Mixed Type of depression is defined through the relative balance of the valence and anhedonia elements, and balanced reactive and regulative functions of stimulation processing. It most probably comprises specific patterns of stimulation processing of both primary (valence or anhedonic) types, and their activation might depend on the specific situation.

Even though some of the structural components overlap across various affective types, they do not always mean the same. Somatic reactivity is a component very specific for anxiety (cf. Watson, 2000), thus it appears in both types of anxiety. However, it has different causes and expressions. In arousal anxiety it is a primary element, while in the apprehension type it is not a crucial one, it is caused by worrisome thoughts and is rather expressed as reduced somatic reactivity. Next, attentional control is present in both regulative types, i.e., apprehension anxiety and anhedonic depression. In apprehension anxiety it appears as an effect of worrisome thoughts and primarily indicates impaired inhibition functioning, while in anhedonic depression it appears as an effect of emotional-motivational deficits and negatively influences prolonged and sustained attention. Finally, negative affect is a part of the structure in both depression types. Nonetheless, hostility and anger are typical for valence depression, while for anhedonic depression it is guilt and shame.

With reference to dominant functions in controlling stimulation processing, one should expect similarities in patterns of stimulation processing (e.g., in attentional processing) across types, i.e., reactive types, regulative types and functionally balanced types, and differences within types, i.e., between reactive and regulative types (see Figure 2).

A psychometric study was conducted with the aims of revising the postulated anxiety and depression types (Fajkowska, 2013, 2015) and of constructing an instrument, the Anxiety and Depression Questionnaire (ADQ), which corresponds to the six affective types. Consequently, we proposed four scales of the ADQ to directly measure arousal anxiety (ADQ-ArA), apprehension anxiety (ADQ-ApA), valence depression (ADQ-VD), and anhedonic depression (ADQ-AD). However, in line with the theory (Fajkowska, 2013, 2015), the mixed types of anxiety and depression should be regarded, respectively, as the ratio of arousal anxiety to apprehension anxiety and valence depression to anhedonic depression, and they have the status of secondary types (cf. balance of the nervous processes as a secondary trait/scale, Strelau et al., 1999).

The first development stage involved the generation of items (experts, n = 4; Ph.D. students of psychology, n = 4) that will form the four scales of the ADQ. This generation was guided by methodological requirements underlying the construction of personality inventories (cf. Zawadzki, 2006) and operational definitions of each affective type. Additionally, about five percent of the total number of items was taken, mostly in slightly modified versions, from other inventories [e.g., Attentional Control Scale (ACS), (Fajkowska and Derryberry, 2010), Mood and Anxiety Symptom Questionnaire (MASQ), (Watson, 2000), The Penn State Worry Questionnaire, (Meyer et al., 1990)].

The linguistic analysis of items, as well as the assessment of the content validity (sorting items into types and subscales, experts, n = 4), led to the development of an item pool for each scale. Arousal anxiety consists of 64 items grouped into subscales of Somatic Reactivity (SR, 35 items), Panic/Phobia (PP, 18), and Attentional Vigilance/Avoidance (AVA, 11), while apprehension anxiety has 89 items in subscales of Worrisome Thoughts (WT, 22), Attentional Control (AC, 44), Attentional Avoidance (AA, 13), and Somatic Reactivity (rSR, 10; “r” means that items indicate reduced somatic activity). It should be noted that although Attentional Avoidance is not mentioned in the definition of apprehension anxiety, we introduced this scale experimentally as some studies report significant connections between attentional avoidance and apprehension (Laguna et al., 2004).

Valence depression consists of 71 items clustered around Negative Affect (NA, 50) and Attentional Avoidance (AA, 21), and for anhedonic depression there are 139 statements in four subscales of Emotional-Motivational Deficits (EMD, 87), Positive Affect (PA, 12), Negative Affect (NA, 26), and Attentional Control (AC, 14). Attention was paid during all stages of item generation to keep a balanced keying within each of the scales.

Since the theory assumes that affective types are personality traits rather than purely clinical disorders, the study was conducted on two general, non-clinical samples. Both samples matched the demographic structure of the Polish population.

The first sample (N = 1,109) consisted of 546 males (49.2%) and 563 females (50.8%) with a mean age of 39.19 (SD = 13.59; range 18–65 years). Participants filled out the two scales of the ADQ measuring anxiety types: ADQ-ArA and ADQ-ApA.

39.7% held university degrees, 36.6% finished high school, 15.9% vocational school, and 7.8% elementary school. The participants were of various professional and educational backgrounds, including university students, high-school students, working people, white-collar workers, part-time workers, unemployed, and pensioners. 49.3% of the participants came from villages and small towns, 20% from cities, and 30.7% from big cities and metropolises. 26.1% reported that they had experienced anxiety disorders in the past, and 15.3% suffered from anxiety at the moment of the study. Moreover, 10.1% of respondents benefited from psychological and psychiatric help, and 3.4% were hospitalized because of severe anxiety. The respondents specified different phobias, panic attacks, separation anxiety, and anxiety coexisting with other disorders and states (e.g., depression, traumatic experiences and addictions, reaction to rape, violence, and unsuccessful social and family relations).

The second sample contained 1,086 participants (549 [50.6%] females and 537 [49.4%] males) who were on average 39.01 years old (SD = 13.33; range 18–65 years). They completed the ADQ-VD and the ADQ-AD scales, assessing depression types.

36.6% of the participants completed university education, 36.1% finished high school, 20.3% vocational school, and the remaining 7% elementary school. They represented different professions and schools, including university students, high-school students, working people, white-collar workers, part-time workers, and pensioners. 48.8% of the sample represented inhabitants of villages and small towns, 20.4% of cities, and 30.8% of big cities and metropolises. 22.9% of individuals acknowledged suffering from depression in the past and 12% in the present. In addition, 12.3% received help from psychological and psychiatric services, and 4.5% underwent hospitalization because of depression. They reported reactive depression (e.g., because of loss of a close person or job, difficult family relationships), bipolar depression, major depression (with dominating sadness, lack of values in life, low self-esteem, suicidal thoughts) and depression concomitant with other disorders (e.g., alcoholism, psychosis, borderline personality).

This study was approved by the ethics committee of the Institute of Psychology, Polish Academy of Sciences. Participants provided their written informed consent before the on-line procedure was activated. Participants were recruited from an on-line research panel and every person who completed the full procedure on-line received points that were exchangeable for rewards. The order of the questionnaires was randomized across subjects. Altogether they contained 153 (anxiety questionnaires) and 210 (depression questionnaires) agree-disagree items that allow the assessment of arousal anxiety and apprehension anxiety in the first sample, and valence and anhedonic depression in the second sample.

All of the items included in the arousal anxiety (ADQ-ArA) and valence depression (ADQ-VD) scales are summed to score (1 point per diagnostic item Agree/Disagree), though extra calculations are required for the AC subscales of apprehension anxiety (ADQ-ApA) and for the anhedonic depression (ADQ-AD) scales. To calculate the scores on these scales, the obtained score should be subtracted from the maximum possible score in the given scale, because we are interested in evaluating decreased attentional control (while the items measure the strength of attentional control).

We performed analyses on the discriminatory power of items (Youle's Phi coefficients; ϕ, Phi), confirmatory factor analysis (CFA), intercorrelations of subscales (Pearson's r) and internal consistency (Cronbach's α coefficients) in the first and second sample separately.

Discriminatory power of items was used as the criterion for excluding the preliminarily selected items from the ADQ. For all scales of the ADQ, the number of items that was kept depended on their Youle's Phi coefficients (ϕ,). We decided to apply the value ϕ ≥ 0.30 as it is a correlation between the item score and the overall scale score reduced by this item, which is usually lower than the correlation between item and the total scale score, and values 0.30 and above indicate good and very good discrimination (see Drwal and Brzozowski, 1995). The number of remaining items in each subscale of each ADQ scale was sufficient for further statistical analysis (ArA = 42, ApA = 62, VD = 36, AD = 69).

Confirmatory factor analysis (CFA) was used to verify the factor structure of the set of introduced variables (Kline, 2005). More precisely, we tested the theoretically suggested structures of the affective types. We expected that the models including the number of factors derived from the theory will show a better fit than other (e.g., one-factor) models. The analyses were performed in Mplus (Muthén and Muthén, 1998), which enables models of binary data to be built (see Górniak, 2000; Hox, 2002). Because χ² values for the models fit across all scales of the ADQ were significant (suggesting a lack of fit between the hypothesized models and the data) and due to the sensitivity of χ² in large samples, other fit indices were assessed and reported, namely: Comparative Fit Index (CFI); Root Mean Square Error of Approximation (RMSEA); Tucker Lewis Index (TLI) (Kline, 2005).

In arousal anxiety (ADQ-ArA) we compared a one-factor solution with a three-factor solution. The fit indexes reflected the improvement in fit of the three-factor model [RMSEA = 0.051; CFI = 0.92; TLI = 0.92; χ²/df = 3.93; χ(816, N = 1,109)2 = 3211.13, p < 0.001] over the alternative one [RMSEA = 0.054; CFI = 0.91; TLI = 0.91; χ²/df = 4.24, χ(819, N = 1,109)2 = 3474.33, p < 0.001]. All items loaded significantly onto their respective factors (loadings ranging from 0.66 to 0.84 on the SR subscale, from 0.43 to 0.87 on the PP subscale, and between 0.29 and 0.86 on the AVA subscale). None of the test positions were eliminated from further analysis.

To assess the factor structure of the ADQ-ApA, the fit of one-factor and four-factor models was examined. As predicted, the model fit for the four factors [RMSEA = 0.055; CFI = 0.87; TLI = 0.87; χ²/df = 4.33; χ(1,823, N = 1,109)2 = 7908.17, p < 0.001] was better than for one factor [RMSEA = 0.057; CFI = 0.86; TLI = 0.85; χ²/df = 4.56; χ(1,829, N = 1,109)2 = 8354.69, p < 0.001]. Given that the four-factor model did not reach the fit parameters (CFI, TLI) over 0.90, items with the lowest factor loadings were removed. Indeed, the fit indexes improved [RMSEA = 0.053; CFI = 0.92; TLI = 0.92; χ²/df = 4.16; χ(1,420, N = 1,109)2 = 5918.256, p < 0.001]; however, the left items were not differentiated in their meaning as the factor analysis favors items similar in their content. Thus, for the sake of better psychological and theoretical rationality of items, we decided to keep all of them for subsequent elaboration of the ADQ-ApA. As a result, factor loadings for the WT items ranged from 0.49 to 0.89, for the AC varied from 0.36 to 0.85, for the AA extended from 0.43 to 0.86, and for the rSR oscillated from 0.47 to 0.87.

In case of the ADQ-VD we tested one-factor and two-factor models. Contrasting with the first model, the latter one had a good fit [respectively, RMSEA = 0.071; CFI = 0.88; TLI = 0.87; χ²/df = 6.46; χ(816, N = 1,109)2 = 3211.133, p < 0.001; and RMSEA = 0.057; CFI = 0.92; TLI = 0.92; χ²/df = 4.56; χ(594, N = 1,086)2 = 3838.70, p < 0.001]. The lowest factor loading was 0.58 and the highest 0.86 on the NA factor, while on the AA factor we had a range of factor loadings from 0.59 to 0.84.

The hypothetical model of the structure of the ADQ-AD was grounded on four factors, but we examined three competing models of one-factor, three-factors (EMD, NA—boosted by the items of PA, treated as reversed, and of AC), and four-factors (EMD, NA, PA, and AC). The findings suggested that the four-factor solution was the best one [RMSEA = 0.049; CFI = 0.93; TLI = 0.93; χ²/df = 3.63; χ(2,271, N = 1,086)2 = 8254.56, p < 0.001] compared to the one- and three-factor models [respectively, RMSEA = 0.051; CFI = 0.92; TLI = 0.92; χ²/df = 3.82; χ(2,277, N = 1,086)2 = 8717.44, p < 0.001; and RMSEA = 0.050; CFI = 0.92; TLI = 0.92; χ²/df = 3.71; χ(2,274, N = 1,086)2 = 8441.85, p < 0.001]. The final assessment concerned the factor loadings. For the EMD they ranged from 0.64 to 0.89, for the NA from 0.74 to 0.93, for the PA from 0.63 to 0.93, and for the AC from 0.32 to 0.91. Two items of the AC subscale with the lowest factor loadings were kept after linguistic correction.

With reference to the relevant data (Fajkowska, 2013, for a review see: Watson, 2000), we expected positive correlations (Pearson's r) among the ADQ-ArA subscales. Indeed, the analyses revealed a positive relationship between the SR subscale and the PP subscale (0.32, p < 0.01), the SR and the AVA subscales (0.28, p < 0.01), and the PP and the AVA subscales (0.18, p < 0.01).

Supported by the neuropsychological models of emotions (e.g., Heller, 1993a,b; Heller and Nitschke, 1998; for a review see: Fajkowska, 2013) in the case of the ADQ-ApA, we predicted (a) positive relations among subscales WT, AA, and rSR, and (b) negative relations between the AC subscale and WT, AA, and rSR subscales. The obtained results (Pearson's r) confirmed these speculations to some extent. As predicted, the WT subscale correlated positively with the AA subscale (0.23, p < 0.01) and negatively with the AC subscale (−0.44, p < 0.01). However, contrary to expectations rSR correlated negatively with the WT and AA subscales (−0.44, p < 0.01 and −0.23, p < 0.01, respectively). Many psychophysiological studies reveal that anxious apprehension, unlike other anxious states, are not associated with a greater response of the autonomic system but rather with autonomic rigidity (e.g., Hoehn Saric et al., 1989; Thayer et al., 1996). However, in the long-term perspective it was observed that in addition to worry, physical symptoms and elevated physiological arousal often accompany anxious apprehension (e.g., Nitschke et al., 1999; Laguna et al., 2004). These outcomes are in line with our findings. Thus, the elevated autonomic responsiveness relates to decreased attentional control (positive correlation of rSR with AC, 0.32, p < 0.01).

With regards to the appropriate data presented in the literature (see Fajkowska, 2013 for a review), we anticipated a positive but weak correlation between subscales of the ADQ-VD, i.e., between the NA and AA subscale. The findings (Pearson's r) are in accordance with predictions (0.35, p < 0.01). The attentional insensitivity (or avoidance) to valence of emotional and social material is connected with negative affectivity, which is typical for e.g., anxiety or non-melancholic types of depression (see Fajkowska, 2013).

Turning to the relations among subscales of the ADQ-AD, the results of other studies suggest that we should expect positive and moderate associations between EMD and NA, as well as AC and PA, and negative ones between EMD, AC, and PA (see Fajkowska, 2013), and non-significant relations between NA and PA, defined here as affective traits (see Watson, 2000). The collected data (Pearson's r) partially confirmed these expectations. It was found that negative affect relates negatively to positive affect (−0.59, p < 0.01). The bipolar relation between NA and PA reflected in this study might be explained by the fact that both affects were explored with items representing very intensive negative and positive states. In other words, a high or very high level of NA implies a low or very low level of PA, and vice versa (Watson and Tellegen, 1985; Watson, 2000).

The data revealed a positive relation between EMD and NA (0.71, p < 0.01) and a negative relation between EMD and PA (−0.65, p < 0.01), which Watson (2000) also documented in his research. Motivational deficits (among others understood as the loss of appetitive behaviors and interest in pursuing goal-directed activities) relate to the very specific for this type of depression (a) marked reduction in experiencing pleasure and extremely low PA, and (b) high NA, nonspecific for it (Watson, 2000).

In addition, several studies have shown that effective attentional control is subjected to positive mood, while negative affect has an adverse effect on it (see Fajkowska, 2013 for a review), which is congruent with the results of our study (0.65, p < 0.01 and −0.63, p < 0.01, correlations between AC and PA, AC and NA, respectively). The association between EMD and AC should be negative, and it was (−0.63, p < 0.01). There is a conflict between the intentional and effortful, effective attentional control (cf. Fajkowska and Derryberry, 2010) and emotional-motivational deficits—difficulties in pursuing goals and tasks, putting effort into realizing them, and troubles in undertaking and initiating activities.

The results showed high Cronbach's alphas for all scales of the ADQ (42-item ADQ-ArA = 0.93; 62-item ADQ-ApA = 0.91; 36-item ADQ-VD = 0.93; 69-item ADQ-AD = 0.82). High and moderate internal consistency was also observed across all subscales of each ADQ scale, namely: arousal anxiety (ADQ-ArA: 19-item SR = 0.91; 15-item PP = 0.86; 8-item AVA = 0.53), apprehension anxiety (ADQ-ApA: 16-item WT = 0.90; 29-item AC = 0.91; 8-item AA = 0.53; 9-item rSR = 0.79), valence depression (ADQ-VD: 21-item NA = 0.92; 15-item AA = 0.87), and anhedonic depression (ADQ-AD: 33-item EMD = 0.95; 12-item PA = 0.91; 12-item NA = 0.92; 12-item AC = 0.76). The two shortest subscales—AVA from the ADQ-ArA and AA from the ADQ-ApA—showed the lowest internal consistency.

Based on the results from the analysis of discriminatory power of items and from the CFA, we proposed an experimental version of the ADQ. However, a few final corrections were made. Some items were excluded (e.g., with the lowest factor loadings), some linguistically corrected, and some moved to different subscales, and the subscale of AA was removed from the ADQ-ApA due to having very weak psychometric parameters (results cumulated from all analyses). The intercorrelational analysis among the subscales of the ADQ-ApA revealed that the rSR subscale is not valid. Thus, supported by these results and the appropriate results from the other studies (e.g., Laguna et al., 2004), we decided to transform this subscale into one assessing elevated somatic reactivity in apprehension anxiety.

Moreover, we added some filler items to the arousal anxiety (ADQ-ArA) and valence depression (ADQ-VD) scales to balance keying within each of the scales. Generally, it was not always possible to find a sufficient number of well-balanced items.

The final form of the ADQ (please see the Supplementary Material) is composed of four scales, directly measuring arousal anxiety, apprehension anxiety, valence depression, and anhedonic depression, and indirectly measuring mixed types of anxiety and depression. There is a dichotomous response format for all items (Agree/Disagree). The scoring system is identical as previously described.

ADQ-Arousal Anxiety (ADQ-ArA): 45 items (including 4 fillers), 3 subscales

ADQ-Apprehension Anxiety (ADQ-ApA): 48 items, 3 subscales

ADQ-Valence Depression (ADQ-VD): 40 items (including 4 fillers), 2 subscales

ADQ-Anhedonic Depression (ADQ-AD): 64 items, 4 subscales

Table 1 demonstrates the socio-demographic characteristics of the validation, non-clinical sample (N = 1,632). The sample matched the demographic structure of the Polish population. Participants who provided an unusually high number of identical answers on any questionnaire were removed from analyses. This procedure was used for each questionnaire separately. For ADQ and EPQ-R we removed participants with M+2SD of identical answers, and for the remaining questionnaires we removed those who had zero variance in their answers, as the M+2SD procedure turned out to be ineffective. As a result, from 4 to 11.8% (M = 7.6%) of the participants were removed from the original sample (N = 1,632), hence the differences in Ns across analyses.

Except demographic questions, participants were asked about whether they suffered (now or in the past) from anxiety or depression, and if “yes” they were questioned about psychotherapy, pharmacotherapy, hospitalization, professional diagnosis, and causes of these disorders. Respectively, 23.5 and 19.9% of individuals admitted that they experienced anxiety or depression disorders in the past. 13.3% of them reported that they had suffered from anxiety and 11.1% from depression in the moment of the study. Furthermore, owing to anxiety or depression, correspondingly 9.5 and 11.7% of respondents reported having used psychological and psychiatric help. 4.1 and 5% had been hospitalized because of severe anxiety or depression, respectively. The participants stated different phobias, panic attacks, separation anxiety, and anxiety coexisting with other disorders and states, reactive depression, bipolar depression, major depression, and depression associated with other disorders.

The ethics committee of the Institute of Psychology, Polish Academy of Sciences approved this on-line study and the consent procedure elaborated to it. The procedure of participants' recruitment and data collection were the same as in Study 1.

Respondents across two separate sessions completed a battery of on-line self-report techniques, randomized across subjects, and across sessions:

To provide a general statistical description of items and scales of the ADQ we elaborated means, standard deviations, Cronbach's α coefficients, the discriminatory power of items (Youle's Phi coefficients; ϕ), intercorrelations of subscales (Pearson's r) on the total sample, the means, standard deviations, and t-tests showing sex differences, and the prevalence of affective types in women and men.

The content validity of the ADQ was assessed with inter-rater agreement Fleiss' kappa (κ), while the construct validity was evaluated with confirmatory factor analysis (CFA). Pearson's r and t-tests were used to examine, respectively, the structure, convergent, and divergent validity of the test and theory-consistent group differences. In addition, the test-retest (r_tt) correlations were used to test the stability of the ADQ.

Table 2 summarizes the means, standard deviations, and Cronbach's α coefficients of the total sample. It shows that the internal consistencies of each scale of the ADQ are very high, with Cronbach's α coefficients ranging from 0.92 (for the ADQ-VD scale) to 0.96 (for the ADQ-ApA scale and the ADQ-AD scale). Excepting the AVA subscale of the ADQ-ArA, the αs are also high for the subscales of each ADQ scale (ranges from the 0.93 for the EMD subscale of the ADQ-AD to 0.73 for the AC subscale of the ADQ-AD). Apparently, the lowest numbers of items within subscales can explain the lowest α's (cf. the AVA from the ADQ-ArA or the AC from the ADQ-AD). According to the Spearman-Brown formula, these scales would achieve reliability of around 0.80 with 13 (instead of 5) and 12 (instead of 8) items, respectively.

Table 3 demonstrates the means, standard deviations, as well as t-test results between the sexes, separately evaluated for each scale of the ADQ. It informs that women scored significantly higher on both types of anxiety. There were no significant sex differences on valence and anhedonic depression.

We extracted the “pure types” by controlling the level of the other three affective types. For example, we identified arousal anxiety when the individuals scored above the median in ADQ-ArA and below the median in the other three types (ADQ-ApA, ADQ-VD, ADQ-AD). Mixed types, on the other hand, were built of individuals who scored above the median on both types of anxiety (ADQ-ArA, ADQ-ApA) or depression (ADQ-VD, ADQ-AD). Interestingly, as Table 4 indicates, men reported all types of depression more frequently than women, while women declared arousal and mixed types of anxiety more frequently than men.

The discriminatory power coefficients of items from the four ADQ scales are reported in Table 5. Similarly to the construction stage, we calculated the Youle's Phi coefficients and proposed the value ≥0.30 as indicative of good and very good discrimination.

The results suggest high item discrimination coefficients, ranging from 0.33 (the AVA subscale) to 0.68 (the SR subscale) in the ADQ-ArA, from 0.30 (the AC subscale) to 0.69 (the WT subscale) in the ADQ-ApA, from 0.37 (the NA subscale) to 0.63 (the NA subscale) in the ADQ-VD, and from 0.34 (the EMD subscale) to 0.69 (the NA subscale) in the ADQ-AD.

In order to check the obtained intercorrelations among subscales composing adequate scales of the ADQ in study 1, we analyzed them in study 2. Table 6 demonstrates that generally all of the results are confirmed. However, comparing these findings to the findings from study 1, all correlation coefficients, across all scales of the questionnaire, increased.

In the case of content validity, we examined the extent to which the particular scales of the ADQ represent all proposed facets of arousal anxiety (ADQ-ArA), apprehension anxiety (ADQ-ApA), valence depression (ADQ-VD), and anhedonic depression (ADQ-AD) constructs. Thus, three experts evaluated whether items of the ADQ-ArA and ADQ-AD assess the defined content of arousal anxiety and anhedonic depression, and another three experts decided if the test positions of the ADQ-ApA and ADQ-VD cover the content of apprehension anxiety and valence depression, respectively. Precisely, the raters were instructed to assign the items measuring the adequate construct (e.g., arousal anxiety) to the distinguished facets of that construct (e.g., SR, PP, AVA). The inter-rater agreement was measured with Fleiss' kappa (Fleiss, 1971) using an online calculator (Geertzen, 2012). Kappa (κ) ranges from 0 to 1, with higher values showing greater inter-rater reliability of agreement. For all tested versions of the ADQ the κ-values were very high: ADQ-ArA, κ = 0.93; ADQ- ApA, κ = 0.90; ADQ-VD, κ = 0.94; ADQ-AD, κ = 0.98. Then the raters' sorting was compared to the scoring keys in order to replace, remove or linguistically correct problematic items.

The CFA was used to evaluate the factorial validity and authorize the results from Study 1. Along with these findings we tested (Mplus; Muthén and Muthén, 1998) the same models; however, they were formed on the corrected versions of the ADQ elaborated according to the results from Study 1 (e.g., they have a different number of items because some of them were removed from the original versions, different number of subscales). Again, χ² values for the models fit across all version of the ADQ were significant, thus other fit indices were reported (Kline, 2005).

The results of the CFA for all scales of ADQ are presented in Tables 7, 8. As can be seen, the three-factor model showed the best fit in case of the ADQ-ArA with all indices reaching satisfactory levels. The data clearly demonstrated that all the items are proper markers of the expected single factor. More precisely, for the SR factor range from 0.42 to 0.87, for the PP factor from 0.54 to 0.82, and for the AVA from 0.49 to 0.74.

In case of the ADQ-ApA, the three-factor model had better fit parameters than the one-factor model. The factors loadings for items of the three-factor model are high: from 0.51 to 0.87 for the WT subscale, from 0.41 to 0.85 for the AC subscale, and from 0.61 to 0.84 for the SR subscale.

Again, the two-factor model for the ADQ-VD seemed to be a better fit than the one-factor solution, and all of the items loaded high on the adequate factor: for NA from 0.53 to 0.80, and for AA from 0.55 to 0.81.

The best solution for the ADQ-AD is the four-factor model and the factor loadings of items for this model reach a satisfactory level. More specifically, for the EMD subscale from 0.51 to 0.85, for the PA subscale from 0.59 to 0.86, for the NA subscale from 0.70 to 0.88, and for the AC subscale from 0.52 to 0.88.

Additionally, in order to place the proposed affective types among other related personality constructs, we assessed the convergent and divergent validity with well-recognized measures. We predicted that state-like arousal anxiety and trait-like apprehension anxiety are both positively related to state anxiety and trait anxiety (STAI). However, the correlation between state anxiety and arousal anxiety should be higher than the correlation between trait anxiety and arousal anxiety, and opposite relations should be identified for apprehension anxiety. As can be seen from Table 9, the obtained data confirmed these predictions.

The review of results from other sources showed that a moderate negative correlation with extraversion and a moderate and high positive correlation with neuroticism (EPQ-R[S]) is usually obtained for both state and trait anxiety (STAI) (see Fajkowska, 2013). Thus, similar relations between extraversion, neuroticism, arousal anxiety, and apprehension anxiety could be expected, which is actually reflected in the results showed in Table 9.

According to the tripartite model of anxiety and depression proposed by Clark and Watson (Burns and Eidelson, 1998; Watson, 2000), anxiety relates to Negative Affect (NA) but is not connected with Positive Affect (PA), while depression is associated with both affects by correlating negatively with PA and positively with NA. The results of our studies only partially support this model. Both types of anxiety moderately and positively related to NA; however, they also correlated moderately and negatively with PA (see Table 9). But both types of depression, low and negatively (valence depression) and moderately and negatively (anhedonic depression) related to PA, and moderately and positively to NA (see Table 11). Nonetheless, there are some studies matching our (but not Clark and Watson's) findings (e.g., Burns and Eidelson, 1998; Fajkowska and Marszał-Wiśniewska, 2009).

Fajkowska and Derryberry (2010) provided evidence that anxious and depressive subjects scored significantly lower on effortful attentional control than non-anxious and non-depressive individuals. The results from our study are congruent with their findings. As Tables 9, 10 show, apprehension anxiety and anhedonic depression are negatively correlated with attentional control.

The Mood and Anxiety Symptom Questionnaire (MASQ; Watson, 2000) can be used to assess anhedonic depression. The MASQ Anhedonic Depression Subscale displays good convergent validity with the Beck Depression Inventory (BDI; cf. Kendall et al., 1987). Therefore, we predicted a higher and positive correlation between anhedonic depression and depression measured by BDI, and low or moderate between valence depression and depression measured by BDI. This forecast is supported by the data presented in Table 10.

Generally, in most studies low extraversion and high neuroticism are found in clinical and nonclinical depression (e.g., Watson et al., 1999; Kotov et al., 2010; Fajkowska, 2013). Thus, it implies that we should expect that both types of depression would be negatively related to extraversion and positively to neuroticism. Indeed, the data presented in Table 10 support this hypothesis.

The structure of both types of depression refers to the NA; however, its content is depression-type specific (cf. definition of valence and anhedonic depression). We predicted that valence depression should correlate higher with hostility than anhedonic depression, while anhedonic depression would be more strongly related to sadness and guilt. Fear should not differentiate depressions. The obtained data supported these speculations (cf. Table 11).

In addition, the very low PA is a part of the structure of anhedonic depression, thus we expected stronger negative relations between anhedonic depression and the Basic Positive Scales (Joviality, Self-assurance, and Attentiveness) than between valence depression and these scales. Again, these expectations are reflected in our empirical data (cf. Table 11).

Finally, as the emotional-motivational deficit defines anhedonic depression we also predicted stronger relations between Other Affective States, especially those referring to low energetic states (Fatigue, Serenity), and anhedonic depression than between Other Affective States and valence depression. As Table 11 shows, the predictions were confirmed.

We assumed that reactive types, that is arousal anxiety (ArA) and valence depression (VD), should be more weakly associated with adaptive and nonadaptive cognitive strategies of emotion regulation than regulative types, that is apprehension anxiety (ApA) and anhedonic depression (AD). Data presented in Table 12 qualified our predictions.

The next step in assessing construct validity was to analyze the theory-consistent group differences. Gray (1981) proposed two systems of controlling behavioral activity, that is the behavioral inhibition system (BIS) and the behavioral activation system (BAS). The BIS is thought to regulate aversive motives, in which the goal is to move away from something unpleasant, while the BAS is understood to regulate appetitive motives, in which the goal is to move toward something desirable (Carver and White, 1994).

It claims that the amygdala provides inputs to the BIS and may relay its outputs to the hypothalamus and autonomic nervous system, thereby mediating anxious arousal. Sustained activation of the BIS may therefore account for some features of anxiety and be related to panic (cf. Barlow, 2004, p. 210). Thus, we assumed that high arousal-anxious individuals would score higher on the BIS than low arousal-anxious ones. And it is clear from the results that high arousal-anxious subjects (n = 265) are higher (M = 3.05, SD = 0.47) on the BIS (Carver and White, 1994; Müller and Wytykowska, 2005) scale than low arousal-anxious participants (n = 313; M = 2.50, SD = 0.47), t₍₅₇₆₎ = 14.00, p < 0.001, d = 1.18. However, as data from one study (Moser et al., 2013) indicated, we should expect higher BIS in apprehension anxiety (measured by STAI) than in arousal anxiety (measured by MASQ). Their study showed that apprehension anxiety correlates three times higher with BIS than arousal anxiety because the former one is most closely associated with error monitoring. The findings from our study employing ADQ-ArA and ADQ-ApA measures are in accord with the cited studies: high apprehension-anxious individuals (n = 134) are higher (M = 2.88, SD = 0.39) on the BIS scale than high arousal-anxious participants (n = 133; M = 2.72, SD = 0.44), t₍₂₆₅₎ = 3.49, p < 0.001, d = 0.44. The high-apprehension subjects (n = 298) obtained higher scores on BIS than low-apprehension individuals (n = 301; M = 3.12, SD = 0.46, and M = 2.47, SD = 0.37, respectively; t₍₅₉₇₎ = 17.16, p < 0.001, d = 1.40).

In addition, high activity of the BIS means a higher level of sensitivity to nonreward, punishment, and novel experience, which results in a natural avoidance of such environments in order to prevent negative experiences such as fear, anxiety, frustration, and sadness. Thus, it should be predicted that individuals with a high level of valence depression would show a higher level of BIS comparing to individuals with a low level of valence depression and high-anhedonic depressive, because the negative affect building this type of depression relates to the aforementioned range of negative emotions. Indeed, results of the analysis aimed at these differences showed that individuals with high valence depression (n = 296), measured with ADQ-VD, revealed a higher BIS level (M = 2.92, SD = 0.43) than subjects with low valence depression (n = 319; M = 2.51, SD = 0.42), t₍₆₁₃₎ = 10.84, p < 0.001, d = 0.87, and participants with high valence depression (n = 122) scored higher on the BIS scale than those with high anhedonic depression as assessed by the ADQ-AD (n = 124; M = 2.87, SD = 0.40, and M = 2.75, SD = 0.47, respectively), t₍₂₄₄₎ = 2.13, p < 0.05, d = 0.30).

Bijttebier et al. (2009) summarized the studies that examined the relationship between sensitivity of the BIS and BAS systems and a broad range of psychiatric disorders. Among others, they found that low BAS sensitivity characterized anhedonic depression. Along with these results we assessed the differences in BAS level and three BAS-related scales: BAS Drive, BAS Fun Seeking, and BAS Reward Responsiveness (Carver and White, 1994; Müller and Wytykowska, 2005) between high anhedonic-depressive (n = 263) and low anhedonic-depressive (n = 300) individuals, and between high anhedonic-depressive (n = 124) and high valence-depressive (n = 122). Anhedonic depression and valence depression were assessed by the ADQ-VD and ADQ-AD, respectively. As expected, the results showed that high anhedonic-depressive individuals scored significantly lower on the BAS (M = 2.52, SD = 0.50) than low anhedonic-depressive (M = 2.86, SD = 0.43), t₍₅₆₁₎ = 8.54, p < 0.001, d = 0.72, and high valence-depressive individuals (M = 2.54, SD = 0.48, and M = 2.77, SD = 0.43, respectively), t₍₂₄₄₎ = 3.89, p < 0.001, d = 0.50. The high anhedonic-depressive participants were significantly lower (M = 2.85, SD = 0.54) on the BAS Drive scale than low anhedonic-depressive participants (M = 5.41, SD = 1.82), t₍₅₆₁₎ = 7.87, p < 0.001, d = 0.68, and they were lower on the BAS Fun Seeking (M = 2.58, SD = 0.55) and BAS Reward Responsiveness (M = 2.98, SD = 0.51) scales than low anhedonic-depressive individuals (M = 2.88, SD = 0.45, t₍₅₆₁₎ = 6.99, p < 0.001, d = 0.58 and M = 3.27, SD = 0.38, t₍₅₆₁₎ = 7.76, p < 0.001, d = 0.66, respectively). Also, the high anhedonic-depressive participants were lower on BAS drive, BAS Fun Seeking, and BAS Reward Responsiveness scales than high valence-depressive individuals [M = 2.47, SD = 0.55 vs. M = 2.72, SD = 0.56, t₍₂₄₄₎ = 3.54, p < 0.001, d = 0.45; M = 2.62, SD = 0.53 vs. M = 2.82, SD = 0.49, t₍₂₄₄₎ = 3.21, p < 0.001, d = 0.39; M = 3.03, SD = 0.45 vs. M = 3.20, SD = 0.46, t₍₂₄₄₎ = 2.86, p < 0.01, d = 0.37, respectively).

The test-retest (r_tt) reliabilities were evaluated on smaller groups (randomly selected from the total sample) that filled out the ADQ scales again after 5 weeks from the initial study. According to the results, the r_tt reliabilities (see Table 13) are high for all scales of the ADQ, varying from 0.70 (ADQ-ArA) to 0.89 (ADQ-ApA). Moreover, coefficients for the tests measuring state-like arousal anxiety (ADQ-ArA) and valence depression (ADQ-VD) are lower (0.70, 0.79, respectively) than for the tests assessing trait-like apprehension anxiety (ADQ-ApA) and anhedonic depression (APQ-AD); 0.89 and 0.88, respectively. Coefficients of 0.70 are considered satisfactory for personality states (Spielberger, 1983). Additionally, the test-retest reliability of most subscales of each ADQ scale is high or moderate, except for poor reliability (0.45) of the subscale of the ADQ-ArA version that assesses attentional vigilance/avoidance (AVA).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.