The Affective Neuroscience Personality Scales (ANPS) in Slovenia: Validation of the Six Primary Scales Across Three ANPS Versions

1.1 The Primary Emotional Systems and the Affective Neuroscience Personality Scales (ANPS)

Over the past few decades, there has been a substantial increase in the understanding of the neurobiological underpinnings of personality, specifically of its emotional foundation (Davis & Panksepp, 2018; DeYoung et al., 2021; Panksepp, 1998; Yarkoni, 2015). One of the leading contributions was the Affective Neuroscience Theory (Panksepp, 1998). The author and his colleagues performed experiments on animals (vertebrates, mainly mammals) that involved pharmacological manipulations, deep electric brain stimulation, and lesion studies combined with detailed subject observation. They identified seven primary emotional systems that are grounded in the subcortical regions of all mammals (the systems are capitalized to prevent confusion with similar terms for emotions used in the common vernacular). These systems are SEEKING, LUST, CARE, PLAY, ANGER, FEAR, and SADNESS. Each of them has a separate neural pathway during activation (e.g., SEEKING involving the nucleus accumbens, ventral tegmental area, mesolimbic and mesocortical outputs, lateral hypothalamus, and the periaqueductal gray; Montag & Davis, 2018; Panksepp, 2011), specific neuromodulators that activate and regulate it (e.g., SEEKING – dopamine, ANGER – substance P, FEAR – glutamate, CARE – oxytocin), and expression in observable behaviors (e.g., distress call vocalizations – SADNESS, fleeing – FEAR, biting – ANGER) that causally follow the activation of each system (Panksepp, 2006). While SEEKING, LUST, CARE, and PLAY represent the positive primary emotional systems, ANGER, FEAR, and SADNESS represent the negative emotional systems (Davis & Panksepp, 2011).

The primary emotional systems are thought to have developed as “survival guides” and have been crucial in allowing mammals to navigate environments filled with evolutionary pressures that have endangered them throughout the course of their evolution (Panksepp, 1998). The SEEKING system drives animals to search the environment for the resources required to fulfill their survival needs (a drive accompanied by feelings of enthusiasm, curiosity, and motivation). The ANGER system enables animals to defend themselves and their resources from others. When activated, the system evokes feelings of irritation, anger, frustration, and rage. The FEAR system helps animals escape and avoid physical threats that could result in bodily harm or death. The system associates with anxiety, worry, and fear. The LUST system drives animals to search for and identify potential mates and thus reproduce. In the course of their later evolution, mammals acquired a more social orientation which resulted in an affective system that drives them to nurture their offspring (CARE). The system is associated with a need for nurturing, feelings of empathy and affection, and preference to care for others, as well as with a powerful separation distress that fosters caregiver-offspring contacts and is accompanied by distress vocalizations, feelings of sadness, and panic in the offspring (SADNESS). Finally, a system that promotes social bonding and learning, especially between young animals, is accompanied by feelings of joy and happiness (PLAY). This neuro-evolutionary psychobiological point of view offers an insight into the functional characteristics of motivational/emotional systems in human phylogenesis (Cory & Gardner 2002; Darwin 1965; McLean, 1990) and has the potential to contribute to a better understanding of human personality, as well as to promote treatment of mental disorders (Coenen et al., 2011; Davis & Panksepp, 2011).

To investigate individual differences in these evolutionarily ancient drives, Davis et al. (2003) constructed a self-report measure – the 110-item ANPS. They capture six of the primary emotional systems, with the LUST system omitted due to concerns regarding social desirability bias which might threaten the validity of the respective scale. However, the authors added the spirituality scale as an additional dimension for its documented relevance in the clinical treatment of addiction problems (Jarusiewicz, 2000). Since its original version, the ANPS has been subject to several revisions, with the 112-item version 2.4 (Davis & Panksepp, 2011) being the most prominent one in current use. It contains slight adjustments over the ANPS intended to enhance the psychometric properties of the scales. Indeed, the ANPS and their variants have suffered from several psychometric issues, including occasional difficulties to support the factorial structure according to the most commonly cited fit indexes (Hu & Bentler, 1999). To improve upon some of these drawbacks, several short versions of the instrument, targeting the original six primary emotional systems, have been created. These include the 33-item BANPS (Barrett et al., 2013), the 36-item ANPS-S (Pingault et al., 2012a), and the ANPS-AR (Montag & Davis, 2018) which accounts for observer ratings.

It is also of note that the ANPS, all designed as trait measures of Panksepp’s primary emotions (Montag et al., 2021), have been biologically validated. Behavioral genetic studies with twins yielded moderate (e.g., for ANGER) to high (e.g., PLAY) heritability estimates for the primary emotional systems as assessed by the ANPS (Montag et al., 2016a,b). Molecular genetic research further identified links of several genetic markers (e.g., DARPP-32, OXTR, DA, COMT) with the negative ANPS systems (Felten et al., 2011; Harro et al., 2019; Montag et al., 2011; Reuter et al, 2009). Likewise, brain imaging studies found associations of brain activity in discrete areas with specific scores of the “ANPS negative primary systems” (e.g., Deris et al., 2017; Unterrainer et al., 2017), and Reuter et al. (2009) reported that higher ANGER scores relate to lower gray matter density in the left amygdala.

The ANPS and its later versions have been translated and validated in several countries and various languages, including French (ANPS 2.4: Pahlavan et al., 2008; Pingault et al., 2012b; ANPS-S: Pingault et al., 2012a), Spanish (ANPS 2.4: Abella et al., 2011), Turkish (ANPS: Özkarar-Gradwohl et al., 2014), Norwegian (ANPS 2.4., BANPS, ANPS-S: Pedersen et al., 2014), Polish (ANPS 2.4: Cwojdzińska & Rybakowski, 2015), Italian (ANPS: Pascazio et al., 2015; ANPS 2.4: Giacolini et al., 2017), Brazilian Portuguese (BANPS: Esposito et al., 2016), German (ANPS: Reuter et al., 2017), Japanese (ANPS 2.4: Narita et al., 2017), Persian (BANPS: Amiri & Azad-Marzabadi, 2017), Mandarin Chinese (ANPS 2.4: Sindermann et al., 2018), Portuguese (BANPS: Gurfinkel et al., 2018), and Serbian (ANPS 2.4: Montag et al., 2019). Montag et al. (2021) introduced the latest revision of the instrument – the ANPS 3.1 and invited researchers to start using it. It differs from the previous versions in that it features a six-point rating scale and includes an experimental social dominance scale while retaining the spirituality scale. The potential significance of the social dominance scale for the assessment of human personality was proposed due to the identification of the trait in other animal species and the associated neuromodulators such as testosterone and cortisol (van der Westhuizen & Solms, 2015).

Following this suggestion, the current study aimed to validate the ANPS 3.1 in Slovenia where none of the ANPS versions has been translated/adapted to date. As the 3.1 version also includes items of the short ANPS versions, we simultaneously targeted the validation of the BANPS and ANPS-S.

1.2 Primary Emotional Systems, Personality Traits, Emotional Experiences, and Clinical Correlates

From the very beginnings of the development of the ANPS, the primary emotional systems that arise from subcortical brain regions have been thought to, at least in part, represent the neurobiological substrate of the five robust personality traits (Davis et al., 2003; Davis & Panksepp, 2011) and a likely evolutionary foundation of human personality (Davis & Panksepp, 2011; Montag & Panksepp, 2017). Specifically, Panksepp’s primary emotional systems have been considered as bottom-up emotional underpinnings of the lexically and statistically derived higher-order traits of the FFM or the Big Five (Davis & Panksepp, 2018; Marengo et al., 2021; Montag et al., 2021).

Albeit convincing neuropsychological evidence for the links of neural circuitry to the ANPS scores is still insufficient (Montag et al., 2021), self-report data obtained through measures of personality traits and the ANPS (asking individuals more directly about feelings and their expressions in behavior than personality inventories) show consistent and substantial relations of the six primary emotional systems with four robust personality traits across different countries. Specifically, SEEKING associates with openness, PLAY with extraversion, both low ANGER and high CARE with agreeableness, and FEAR, SADNESS, and ANGER with neuroticism (Abella et al., 2011; Barrett et al., 2013; Davis & Panksepp, 2011; Montag et al., 2019; Montag & Panksepp, 2017; Pahlavan et al., 2008) (Figure 1). Montag and Panksepp (2017), for example, reported the following associations between the theoretically related ANPS and the FFM traits: SEEKING and openness (r = 0.47), both FEAR and SADNESS and neuroticism (r = 0.75 and r = 0.68, respectively), CARE and agreeableness (r = 0.50), ANGER and both neuroticism (r = 0.65) and agreeableness (r = −0.48), and PLAY and extraversion (r = 0.46). A recent meta-analysis of 13 studies using the NEO and the Big Five inventories (Marengo et al., 2021) also suggested that ANGER associates with both agreeableness (r = −0.47) and neuroticism (r = 0.42), CARE with agreeableness (r = 0.39), FEAR and SADNESS with neuroticism (r = 0.73 and r = 0.67, respectively), PLAY with extraversion (r = 0.64), and SEEKING with openness (r = 0.47). These mainly differential relationships support the convergent validity of the ANPS scores. In contrast, modest or nonsignificant associations of the ANPS scores with conscientiousness across the studies are consistent with the divergent validity of the ANPS. While the ANPS items that refer to the six primary emotional systems scales were designed to capture individual differences in conscious expression of subcortical primary emotions, conscientiousness represents the most cerebrally-focused personality dimension of the FFM/Big Five measures and thus, links poorly to the primary emotions (Marengo et al., 2021).

Figure 1

The theoretically congruent associations between the primary emotions, Big Five traits, and positive and negative experiences.

Other important personality theories have been evaluated in regard to the ANPS (see Montag et al., 2021 for a review), such as Cloninger’s biosocial theory. Cwojdzińska and Rybakowski (2015), for example, found novelty seeking associated with PLAY, low harm avoidance with SEEKING, CARE with both reward dependence and persistence, FEAR and SADNESS with both harm avoidance and reward dependence, and ANGER with low persistence. Validity evidence of the ANPS was also documented with scores on questionnaires stemming from the subjective well-being tradition (e.g., positive and negative affect), or instruments targeting specific emotions (e.g., anxiety, anger). Employing such self-report measures with young adults from a general population, Pingault et al. (2012b) revealed relations of the ANPS scores referring to positive (particularly PLAY) and negative primary emotional systems (especially FEAR and SADNESS) with positive and negative affectivity scores. Further support for the validity of the ANPS comes from clinical studies. Pedersen et al. (2014) validated three versions of the ANPS in a large sample of persons with various personality disorders (PDs) and found similar results on construct validity and internal reliability to studies in community samples. Compared to a large control group, depressed patients reported higher levels of SADNESS and FEAR, as well as lower levels of SEEKING and PLAY, but the controls who scored higher on depression showed a similar pattern of associations with the ANPS scores and reported higher levels of ANGER (Montag et al., 2017). Along with SADNESS most strongly predicting depression, the scale scores were associated with somatization, anxiety, substance abuse disorder (Fuchshuber et al., 2019a), and childhood trauma (Fuchshuber et al., 2019b).

1.3 The Present Study

Our goal was to simultaneously validate three versions of the ANPS in Slovenia, i.e., the most recent ANPS 3.1 that has not been validated yet, and two of the short versions, namely the BANPS and the ANPS-S. Since both short versions include the six primary emotional scales, we considered only these six scales for the ANPS 3.1. Specifically, we aimed to assess the construct validity of the three ANPS versions (their factorial structure and intercorrelations among the scales), the correlations between the corresponding scales across the three versions, the internal consistency of the scales, and the convergent validity of the scale scores against the Big Five personality traits and the experiences of positive and negative emotions. Likewise, we examined the associations of the scales forming general scores of positive affect and negative affect. To support the convergent validity of the three ANPS versions, PLAY should be associated with extraversion, ANGER with neuroticism and (negatively) with agreeableness, SEEKING with openness, and both FEAR and SADNESS with neuroticism. Likewise, the ANPS general positive affect and general negative affect, as well as each of their correspondent scale scores should share variance with the frequency of individuals’ positive emotional experiences and negative emotional experiences, respectively.

Based on the obtained psychometric properties, we further aimed to opt for the relatively most appropriate version of the ANPS in the Slovenian social-cultural context and ascertain whether the mean scale scores can be directly compared between males and females by testing the measurement invariance across sex. Except for the ANPS-S (Orri et al., 2018), the measurement invariance has not been evaluated in previous ANPS studies, which established consistent, but modest sex differences, with women scoring higher on FEAR, SADNESS, and CARE (Pahlavan et al., 2008; Pedersen et al., 2014; Pingault et al., 2012b); other score differences (i.e., for ANGER and PLAY) were sample-specific. However, a lack of evidence on measurement invariance may cast doubts on the outcomes of group comparisons because we cannot be certain that the target constructs are measured in the same way across different groups (e.g., men and women).

2.1 Participants and Procedure

We recruited participants through an online survey platform. E-mails were sent to all three Slovenian public universities and the study was promoted on social media. The only requirement to take part was 18 years of age. Potential participants were first informed about the aim of the study, the approximate time needed to complete the survey and their rights (e.g., anonymity, consensual participation, possibility to terminate participation at any time, data analysis only on a group level). The introductory page also mentioned a raffle that was employed in the study (one participant would win a 25€ gift card) as a modest incentive for completing the entire survey. The study was approved by the Ethics Committee of the Faculty of Arts, University of Ljubljana (No. 248-2021).

Individuals who intended to participate gave their consent by clicking “next page” to enter the survey. Out of 843 potential participants who provided informed consent (M _age = 29.85 years, SD = 12.54; 39.5% men and 60.5% women; 51.6% students, 35% employed full time, 1.8% employed part time, 3.7% unemployed, and 8.9% indicated “other” status). The highest attained level of education was as follows: high school (44.2%), Bachelor’s degree (22.5%), Master’s degree (18%), and 15.3% of the remaining individuals reported either a nine-year elementary school, secondary technical school, or a PhD. We performed the analyses only for the participants with complete data. Of the 843 initial participants, 502 completed the survey and constituted our final sample. It included 308 women (60.7%) and 194 men (38.6%). The age of the final sample ranged from 18 to 30 years (M = 28.99 years, SD = 12.59); 274 (54.6%) were students, 15.4 (30.7%) were employed full time, 9 (1.8%) were employed part time, 17 (3.4%) were unemployed, and 48 (9.6%) specified “other” as their current employment status. Most participants completed a four-year high school (231 or 46%), followed by those who had a Bachelor’s degree (106 or 21.1%), a Master’s degree (94 or 18.7%), and finished either a nine-year elementary school, secondary technical school, or a PhD (17 or 14.2%). The group comparison analysis between the initial sample and the final sample did not reveal statistically significant differences across the respective demographic characteristics (age: F(1, 1345) = 1.510, p = 0.22; employment status: χ ² = 15.49, df = 16, p = 0.49; level of education: χ ² = 41.028, df = 13, p = 0.26). Note that the structure of the final sample by sex was nearly identical to the structure of the initial sample.

2.2 Measures

The survey consisted of demographic questions (age, gender, employment status, and level of completed education) and three self-report questionnaires. To avoid a potential bias in the order of presentation, the rank order of appearance of the questionnaires was automatically rotated.

2.3 Data Analyses

To examine the construct validity of the three ANPS versions, we performed confirmatory factor analyses in R (R Core Team, 2021). Model fit was assessed according to the following indexes: CFI (Bentler, 1990), TLI (Tucker & Lewis, 1973), RMSEA (Steiger & Lind, 1980), and SRMR (Hu & Bentler, 1999). We relied on the cut-off values most often cited in the literature as suggesting good model fit (Hu & Bentler, 1999): CFI and TLI of 0.90 (or above), RMSEA between 0.05 and 0.08, and SRMR below 0.08. The robust maximum likelihood estimator was used to evaluate the model fit of the measures. Using IBM SPSS Version 25 (IBM, 2017), internal consistency of the scales was estimated by alpha coefficients, whereas correlations between the corresponding scale scores across the three ANPS versions, intercorrelations among the scale scores within each ANPS version, and convergent validity of the scales across the versions were estimated by the Pearson product-moment coefficients of correlation. Based on the results of these analyses, we selected the relatively most appropriate ANPS version and tested the factorial model for measurement invariance across sex through a multigroup confirmatory factor analysis. We examined configural, metric, and scalar invariance, using ∆CFI as a measure of invariance; ∆CFI smaller or equal to 0.01 was considered to indicate measurement invariance at each level (Cheung & Rensvold, 2002). If the model fit worsened substantially at the next step of invariance testing (i.e., ΔCFI > −0.01), we reviewed the modification indexes to examine noninvariant items. We then removed the equality of constraint in items that contributed most to an inadequate model fit. Partial invariance was achieved if we were able to decrease the differences between CFIs and RMSEAs within the accepted cut-offs and the parameter remained unconstrained in the following models of measurement invariance assessment. Lastly, we ran independent samples t-tests (two-sided) to examine sex differences and estimated the effect sizes of the differences by Hedge’s g.

3.1 Confirmatory Factor Analysis of the ANPS 3.1, the BANPS, and the ANPS-S

3.2 Correlations Across the Three ANPS Versions in the Slovenian Sample

As shown in Table 2, all the correlations between the corresponding scale scores of the three ANPS versions were high. Within this range of effect size, the associations were relatively weaker between the CARE scales, particularly between the BANPS and the ANPS-S. It is of note once again that the CARE scale of the ANPS 3.1 and the ANPS-S had low loadings for two and three items in our sample, respectively. However, a similar pattern of correlations between the corresponding scales of multiple ANPS versions was revealed in a large Norwegian clinical sample (Pedersen et al., 2014).

3.3 Intercorrelations, Internal Reliability, and Convergent Validity

The multi-trait-multi-method (MTMM) matrix for each of the three ANPS versions (Tables 3, 4, and 5 for the ANPS 3.1, BANPS, and ANPS-S, respectively) contains intercorrelations between the six primary scale scores and between the general positive affect and general negative affect scores, correlations between the scale scores and each of the two General Affect scores, as well as evidence for internal reliability and convergent validity of all these scale scores.

Except for the FEAR and the SADNESS scales, sharing a substantial amount of variance across the three versions, the correlations between the primary scales were low (some of them non-significant) and a few were moderate. The association between the two general affect scores was also low, suggesting the dissimilarity of the constructs. Furthermore, the correlations of each of the three positive emotions scales with the general positive affect and of each of the three negative emotions scales with the general negative affect were high. The internal reliability of the primary scale scores was acceptable, with a somewhat lower alpha coefficient for the BANPS CARE scale (0.67) and the highest one for the ANPS 3.1 FEAR scale (0.91). The alphas for the general affect scores were 0.80 and above across the versions.

The associations of the primary emotions scale scores with the frequency of recent positive and negative emotional experiences (over the last 4 weeks) and with the theoretically congruent Big Five traits provide evidence for the convergent validity of the scales across the three ANPS versions. Overall, the convergence of general negative affect with negative experiences appeared stronger (rs over 0.60) than of general positive affect with positive experiences (moderate effect sizes), whereas the significant associations of general positive affect with negative experiences and of general negative affect with positive experiences were all negative and low or moderate. The correlations of separate negative and positive primary emotional scale scores with negative emotional experiences (from 0.33 for the ANPS-S ANGER to 0.65 for the BANPS SADNESS) and positive emotional experiences (from 0.20 for the BANPS SEEKING to 0.36 for the BANPS/3.1 PLAY), respectively, were obtained within a similar range. In addition, there were a few non-significant correlations between the presumably unrelated or negatively related scale scores.

In relation to the Big Five, PLAY showed the highest correlations with extraversion, i.e., 0.57 (BANPS/3.1) and 0.49 (ANPS-S), ANGER correlated with both neuroticism (from 0.49 for ANPS-S to 0.55 for ANPS 3.1) and with agreeableness (from −0.41 for ANPS-S to −0.49 for 3.1), SEEKING with openness (from 0.57 for 3.1 to 0.61 for BANPS), and both FEAR and SADNESS with Neuroticism (from 0.74 for BANPS to 0.81 for 3.1 and from 0.61 for 3.1 to 0.67 for BANPS, respectively). In terms of effect sizes, all these associations were at least moderate. Likewise, the general affect scores and the four robust personality traits were substantially related as indicated by moderate (e.g., between ANPS-S positive affect and openness) to high correlations (up to 0.80 and 0.81 of BANPS and 3.1/ANPS-S negative affect with neuroticism, respectively).

3.4 Measurement Invariance of the BANPS Across Sex

To the best of our knowledge, only Orri et al. (2018) reported on the measurement invariance of an ANPS measure (i.e., the ANPS-S) across sex. As the BANPS suggested relatively best validity evidence among the three versions under our study, we examined whether the six scales of the BANPS primary emotional systems are measured in the same way in men and women. To assess the measurement invariance, we tested the hypotheses regarding equality of factor loadings (metric invariance) and equality of item intercepts (scalar invariance). The configural models were adequate (Table 6) and served as baseline models to evaluate the change in overall model fit after imposing specific parameter constraints of subsequent metric and scalar invariance models. Imposing the constraint of equal loadings (metric invariance) did not statistically significantly change the fit of the six-factor model in the BANPS (ΔCFI = −0.03). Nevertheless, when we tested the hypothesis regarding equal indicator intercepts (full scalar invariance), there was a significant change in overall model fit (ΔCFI = −0.015), and only partial scalar invariance was shown by releasing several items.

The process of determining partial scalar invariance was as follows: Upon first inspection of the modification indexes, we found that freeing intercepts for the item BANPS10 (“I often have the feeling that I am going to cry.”) (MI = 28.5) would significantly improve model fit. After doing so, CFI change was still above 0.01. The modification indexes suggested that the intercept for the item BANPS7 (“I rarely become sad.”) was the next source of model misfit. After freeing the intercepts for this item (MI = 24.6), we reevaluated the CFI and RMSEA change of the partial scalar model with the metric model and found a CFI and RMSEA change of 0.008 and 0.001, respectively. Both items BANPS10 and BANPS7 are the constituents of the SADNESS scale. Muthén and Asparouhov (2014) proposed that the results can be considered trustworthy when no more than 25% of the parameters are non-invariant. Accordingly, our results suggest complete configural and metric invariance of the BANPS, as well as partial scalar invariance across sex. It means that the BANPS in the Slovenian sample allows for mean-level comparisons between men and women despite slight differences in the intercepts of observed indicators. The results of the t-tests (Supplement, Table S4) suggest statistically significant sex differences of modest effect sizes in the levels of FEAR, CARE, and SADNESS, with higher scores for women than men across the three scales.

4.1 The Score Structure of the Three ANPS Versions

Using confirmatory factor analyses, we first tested the proposed six-factor structure of the primary emotional systems scales for each of the ANPS versions separately. Since the ANPS 3.1 includes two additional scales, we omitted these scale-items from data collection to compare the six-factor models across the versions. The BANPS showed the relatively best construct validity based on the fit indexes. While the RMSEA and SRMR were within the acceptable range across the three measures, the TLI and CFI were fairly low for the ANPS 3.1, and the two indexes for both BANPS and ANPS-S were slightly outside of the common cut-off criteria of 0.90. However, they were notably higher than for the ANPS 3.1.

An overview of the factor loadings (Tables S1–S3 in the Supplementary material) also suggests an advantage of the BANPS over the other two versions. Except for the item 21 (equal to the item 57 in the ANPS 3.1: “I am usually not interested in solving problems and puzzles just for the sake of solving them.”), all other BANPS factor loadings were above 0.40. The lower factor loading of the item 21 is likely due to double negations, which might have caused confusion in the Slovenian language. We recommend a rewording of this item into an affirmative statement in future iterations of the measure in Slovenia, which might improve model fit. In contrast, the factor loadings of six ANPS-S items (of 36) were under 0.40 (a half of them from the CARE scale) and so were 19 items of the ANPS 3.1 (of 84 used in our survey), with half of them belonging to the SADNESS scale. We may conclude that the CARE scale of the ANPS-S and the SADNESS scale of version 3.1 would need additional revision in the Slovenian cultural context.

Our results regarding the construct validity of different ANPS versions are similar to the only other study examining the ANPS 2.4 (as opposed to version 3.1 in our study), the BANPS, and the ANPS-S simultaneously (Pedersen et al., 2014). In a sample of patients with PDs, these authors also reported an advantage of the BANPS over both the ANPS-S and the ANPS 2.4. However, the BANPS fit indexes in their study appear slightly better than ours, perhaps partly due to a potentially more pronounced emotionality in the clinical sample relative to our community sample.

To our knowledge, the present study marks the first psychometric validation of the ANPS 3.1. Considering a reduced number of items that presumably reflect six factors (instead of 112 items of the full instrument that includes additional Spirituality and Social Dominance factors), we obtained a poorer model fit of the 3.1 version compared to each of the two short versions. The lower values of the TLI and CFI, metrics for incremental fit, are likely attributable to the increased complexity of the model which is a consequence of the larger number of items (84). Larger measures tend to be more susceptible to local misspecification issues due to heightened model complexity. In parallel to Pedersen et al. (2014) and Pingault et al. (2012b) who highlighted deficiencies in the two incremental fit indexes for the ANPS 2.4, caution is needed when employing the 84-item ANPS 3.1 in Slovenia. However, the full 3.1 version is yet to be tested for the proposed eight-factor model. While acknowledging its limitations, it is also of note that the reliability and convergent validity evidence suggest good psychometric properties of the six primary scales of the instrument. We recommend an improvement of the ANPS 3.1 items of the six scales and utilization of shorter measures to assess the basic emotions, at least in Slovenia.

4.2 Correlations Between the Corresponding Scales Across the Three Versions and Scale Intercorrelations Within Each Version

Correlations between the corresponding scale scores of the three ANPS versions were all high (predominantly over 0.80), suggesting a strong overlap between the scale scores that were supposed to measure the same primary emotional systems expression. This was not surprising because all of the ANPS-S items and a majority of the BANPS items are subsumed in version 3.1.

Mostly low and a few moderate intercorrelations (with one exception) between different scale scores within each of the ANPS versions under our study are in line with the theory (Panksepp, 1998, 2011) that diverse primary emotions evolved as distinct responses to environmental challenges, as well as with the previous research outcomes (e.g., Davis et al., 2003; Pedersen et al., 2014; Pingault et al., 2012a). An inspection of the number and the magnitude of these divergent correlations between the scale scores within each ANPS version (Tables 3–5) again suggests an advantage of both short versions over version 3.1. However, we revealed high correlations between the FEAR and the SADNESS scale scores across the ANPS versions. This finding has been noted since the first iterations of ANPS development (Davis et al., 2003; see also Pahlavan et al., 2008; Pedersen et al., 2014; Pingault et al., 2012b) and has been discussed by Montag et al. (2022) who proposed several explanations. First, the elevated correlations may stem from the comorbidity of anxiety, a characteristic shared by both fear and sadness. Second, the shared sensitivity to pain may contribute to the considerable FEAR-SADNESS overlap, aligning with the shared neurobiological wiring of pain activation in the brain (MacDonald & Leary, 2005; Panksepp, 1998). Third, the close association between physical and physiological pain systems and the high comorbidity between anxiety and mood disorders (e.g., 80% of adults with depression reported significant anxiety symptoms; Gorman, 1996) may further explain the clear FEAR-SADNESS correlations. To address this issue in the ANPS improvements, excluding cognitive aspects from the SADNESS scale and adding items related to flight tendencies to the FEAR scale could mitigate their overlap.

Across the ANPS versions, we also found substantial associations between each of the primary positive emotional systems scores and each of the primary negative emotional systems scores with the composite scores of general positive affect and general negative affect, respectively, adding to the validity of the superordinate scores.

4.3 Reliability and Convergent Validity Evidence

All of the primary emotional systems scale scores and the composite General Affect scores had acceptable internal reliability for each of the ANPS versions, with relatively the lowest estimates for the BANPS CARE scale and the ANPS-S SEEKING scale. In general, the alpha coefficients appear somewhat higher for the 3.1 version, possibly due to a greater number of items per scale. To improve the (still acceptable) reliability of the respective scales of the two short forms, we do not suggest an increase in the number of items, but a detailed reliability analysis which, however, exceeds the aims of this paper. For example, a slightly lower reliability of the SEEKING scale was noticed in our study and had also been reported previously (Pingault et al., 2012a). This may suggest that the scale contains items that measure two somewhat different aspects of seeking behavior, i.e., curiosity and enjoyment in problem-solving.

Convergent validity evidence was demonstrated by the correlations of self-perceived expression of the primary emotions with four of the Big Five personality traits, as well as by the associations of the positive and negative scale scores (including the respective composite general affect scores) with the self-reported positive and negative emotional experiences, respectively. However, the convergence between negative emotions (FEAR, ANGER, SADNESS, and general negative affect) and recent negative emotional experiences was notably stronger (moderate to high correlations) than those between the positive emotions (SEEKING, PLAY, CARE, and general positive affect) and recent positive emotional experiences (modest to moderate effect size). This result is likely due to a greater similarity between the items assessing negative experiences and the negative primary emotional systems. The experiences of sadness, fear, and anger are all contained in the SPANE (Diener et al., 2009), while the positive experiences related directly to CARE, PLAY, and SEEKING are not. The ANPS and their composite scores are also considered trait-like measures of emotion (Montag et al., 2021), while the SPANE scores represent a more “state-like” measure that is less stable over time as it refers to the frequency of emotions felt in the past four weeks.

Consistent with the ANT hypothesis that four of the robust personality traits might be based on the activity of subcortical affective systems (Davis & Panksepp, 2018; Montag & Davis, 2018), we found evidence for substantial and differential links between the ANPS (versions 3.1, B, and S) and the respective Big Five scale scores. Our results concur with those observed in other studies using different versions of the ANPS and different personality measures (e.g., Giacolini et al., 2017; Montag et al., 2019; Montag & Panksepp, 2017), including the meta-analysis (Marengo et al., 2021). Precisely, we established associations of SEEKING with openness, PLAY with extraversion, FEAR/ANGER/SADNESS with neuroticism, CARE with agreeableness, and low ANGER with agreeableness. Activation of the SEEKING system is marked by heightened motivation and enthusiasm, fostering a proclivity for novelty and exploration (Davis & Panksepp, 2011; Panksepp, 1998). This aligns with openness, as seen in individuals of an exploratory nature and attracted to novelty (Costa & McCrae, 1992). Heightened ANGER levels often accompany marker traits associated with low agreeableness, manifested in increased aggression and conflict (Bettencourt et al., 2006). The CARE system, characterized by a tendency to nurture and attend to the needs of others aligns (Panksepp, 1998) with characteristics of agreeableness (e.g., altruism, cooperation, and consideration; John et al., 2008). Although to a lesser extent, we found CARE positively related to extraversion, too (see also Marengo et al., 2021), presumably because both traits directly refer to the interpersonal aspect of personality. Elevated FEAR levels correlate with an active behavioral inhibition system (Gray, 1987), leading to heightened anxiety and avoidance – the key components of neuroticism (Barlow et al., 2014). Increased SADNESS and grief levels are also captured by neuroticism, along with a heightened irritability and anger proneness (John et al., 2008), exhibited in heightened levels of ANGER. The PLAY system which activates feelings of joy and happiness (Davis & Panksepp, 2011; Panksepp, 1998) represents the biological basis of extraversion, the trait described largely by an inclination towards positive emotionality and social interaction (John et al., 2008).

In addition to these convergences, we found extraversion moderately associated with SEEKING (also reported by Marengo et al., 2021), presumably due to their shared neurochemistry involved in reward processing (Smillie, 2013). Similar to previous findings (e.g., Barrett et al., 2013; Marengo et al., 2021), we found FEAR and SADNESS more strongly related to Neuroticism than ANGER. This might be due to a conceptual and anatomical overlap between the SADNESS and the FEAR systems and their relation to pain (MacDonald & Leary, 2005; Panksepp, 1998), while the ANGER system has been shown to have distinct neurochemistry (Panksepp, 1998). In support of the convergent validity of the three ANPS versions, we further revealed considerably lower shares of common variance between the primary emotional systems and conscientiousness, the trait which reflects the most cerebrally focused personality trait among the Big Five.

4.4 Generalizability Evidence of the BANPS Across Sex and Mean-Score Sex Differences

Taken together, the BANPS showed relatively better psychometric properties (in particular construct validity) among the three ANPS versions employed in the Slovenian sample. Therefore, we further assessed its measurement invariance across sex to establish whether the six-factor model is sufficiently equivalent between men and women to claim that the constructs have the same meaning in both groups which would allow us to test for the sex differences in the mean scale scores. To our knowledge, measurement invariance of the BANPS across different groups has not been reported yet, whereas Orri et al. (2018) examined it for the ANPS-S, both across sex and over a 4-year time interval. Much like their results, the configural model in our study showed a good fit for the RMSEA and SRMR, while the CFI score was slightly lower. The results of multiple-group CFAs supported the hypothesis of metric invariance of the BANPS across sex (the latent factors were equally associated with the items in both groups). Testing for scalar invariance suggested partial scalar invariance. As less than 25% of the intercepts was nonvariant (Muthén & Asparouhov, 2014) and due to the common consensus that equality of factor loadings is more crucial for establishing measurement invariance than equality of item intercepts (Meade & Kroustalis, 2006), our results still allow valid comparisons of the mean scale scores across sex.

The consequent group comparison analysis showed modest sex differences, with women reporting higher levels of FEAR, CARE, and SADNESS. Similarly, studies using the ANPS-S (Orri et al., 2018; Pingault et al., 2012a) and the ANPS 2.4 (Abella et al., 2011; Orri et al., 2018; Pingault et al., 2012b) established the strongest differences (“favoring” women) in these scale scores. These outcomes are in line with demonstrations of a greater inclination for nurturing behaviors among women (Davis et al., 2003; Derntl et al., 2010), as well as a higher prevalence of depressive and anxious feelings in women than men (McLean & Anderson, 2009; Parker & Brotchie, 2010). However, sex differences in the primary emotions also seem to be subject to cross-cultural differences (Özkarar-Gradwohl et al., 2014).

4.5 Limitations, Conclusions, and Future Directions

There are several limitations to this study. Similar to previous ANPS validation studies (Barrett et al., 2013; Davis et al., 2003; Pingault et al., 2012a), the present study featured a community sample consisting of predominantly emerging adults (most of them students) and young adults. Consequently, we were not able to definitively prove that the structure of the six primary emotional systems remains stable across subsequent developmental periods in adulthood. Regardless we assume the likelihood of significant change in mean levels of emotional expression and/or significant changes in rank-order stability of individuals’ traits with age is low. We aimed to compare the validity evidence among three versions of the ANPS that include massively overlapping items. Therefore, we only applied one questionnaire with 89 items (those of version 3.1 referring to the six primary emotional systems and five items of the BANPS that differ from both versions 3.1 and S). We did so to avoid response fatigue and potential confusion among the respondents, which could have contributed to greater measurement error. In line with the ANPS 3.1 scoring, we also used a six-point rating scale across the items (note that the ANPS-S and the BANPS include a five-point scale). Administering the short ANPS forms as separate questionnaires and original rating scales may have, at least to some extent, affected the score structure as well.

Despite some issues with the incremental fit indexes of the BANPS, the troublesome factor loading of item 21 (likely due to the double negation), and relatively low (although still falling within the lower ranges of acceptability) (George & Mallery, 2003) internal consistency of the CARE scale, we consider the Slovenian adaptation of this short form for further research in the respective social-cultural context acceptable. In contrast, several modifications are warranted to the ANPS-S and, in particular, to the ANPS 3.1 (after validating its full version with two additional factors).

Regarding the future of the ANP scales, they would benefit from ameliorations focused on their psychometric properties (e.g., reducing the overlap between FEAR and SADNESS scales, increasing the reliability of the CARE and SEEKING scales in the short ANPS forms), shortenings of the full forms, incorporation of the LUST scale, and creations of alternative “state” instead of trait-measures (Montag et al., 2022). Examples of the LUST scale have already been constructed (Fuchshuber et al., 2022), but the full ANPS 3.1 that additionally includes factors of spirituality and social dominance still awaits validation. Nonetheless, an increasing number of studies have recently employed the ANPS measures across a variety of domains, as mentioned in the Introduction. This might be partly due to an advantage of the ANPS over other measures of personality traits. Most notably, the primary emotions are neurobiologically grounded and are thus likely to provide the evolutionary foundation of the robust personality traits (Davis & Panksepp, 2018; Marengo et al., 2021; Montag et al., 2021). Consequently, they pave the way for studies, which might help elucidate human personality on a neuroanatomical, neurochemical, neurophysiological, and molecular genetic level. The ANPS has also received attention from the growing movement of neuropsychoanalysis (Panksepp & Solms, 2012; Solms & Panksepp, 2012) and their employment suggested results that are consistent with certain psychoanalytic hypotheses, such as those of drive theory (the primary emotions are thought to represent the instinctual drives) and attachment theory (the SADNESS and the CARE systems are assumed to provide foundations of attachment) (Solms & Turnbull, 2011). The psychometrically improved ANPS measures can thus offer valuable assessment tools in diverse areas of research on the human mind.