The Guilt Phenomenon. An Analysis of Emotions Towards God in Highly Religious Adolescents and Young Adults

: In his model of religiosity, Huber postulates a “qualitative leap” between the groups of the “religious” and the “highly religious”. Correspondingly, the data from the Empirica Youth Survey 2018 underline that the topic of guilt and forgiveness is in itself only really present in the “highly religious”. Thus, this article aims to provide a detailed analysis of the relation between emotions towards God and the centrality of religiosity. One of the results of the exploratory factor analysis concludes that emotions towards God comprise three aspects within Protestant “highly religious” adolescents and young adults: a factor for positive emotions, one for negative emotions, and a third for emotions of guilt, release and fear. In this article, we focus on the factor that drives the experience of guilt (and release and fear) and conclude that it is a phenomenon only found within the “highly religious” and not the “religious” Protestant adolescents and young adults. We explicitly incorporate the journal’s main foci in two regards: First, we focus on the particularities of the group of “highly religious” people as identiﬁed by the Centrality of Religiosity Scale (CRS) along with the interactions between the theoretical concept of centrality of religiosity and the content of religiosity. Secondly, we brieﬂy compare “highly religious” with “religious” adolescents and young adults.


Introduction
Religious experiences produce deeply contradictory feelings such as guilt, fear, joy, or release, which have always played a role in religious research. Huber and Richard (2010) state that although "religiosity is closely linked to emotional experiences [ . . . ] [r]esearch about emotions in the context of religion is rare" (p. 21). In recent years, however, emotions have become increasingly compelling concerning theological research, as many contributions on emotions in the three-volume handbook Evangelical Spirituality indicate (Zimmerling 2017(Zimmerling , 2018(Zimmerling , 2020. A tradition of researching religious feelings exists in the psychology of religion and the research of god-images, which, at the latest since Rizzuto's groundbreaking work, constitutively includes feelings towards God (Schaap-Jonker 2018, p. 8). In recent research on God images, also named God representations, it is assumed that "the God image has both affective and cognitive . . . aspects" (Schaap-Jonker et al. 2008, p. 502) which interact with each other: "an individual's thoughts of God, containing some experience, have effects on their experiences of God, while experiences of God affect the cognitions about God" (ibid.). Since the 1990s, important research foci in the psychology of religion have been reviewed in theology (Klein and Streib 2011, p. 198): Initially, the emphasis was laid on the positive connections between religiosity and psychological stability, physical health and longevity (Zwingmann et al. 2017, p. 11). Recently, certain contents cannot be derived theoretically, and a theory formation and operationalization that is limited to only one of the two parameters is incomplete, according to Huber (Huber 2003, p. 193). The validity of the centrality scale could be confirmed several times, e.g., on the basis of high correlations with religious self-assessment or with the significance of religion in everyday life (Huber and Huber 2012, pp. 711, 715). With the help of the centrality scales and on the basis of threshold values, it is possible to distinguish between highly religious, religious and non-religious persons who differ from one another not only gradually but also qualitatively (Huber 2008, p. 6). Thus, Huber (2003Huber ( , 2007Huber ( , 2008 postulates a "qualitative leap" between the groups of the "religious" and the "highly religious". In the case of the "highly religious", religious contents are represented in a more differentiated way and have a stronger influence on general experience and behaviour than in the case of the "religious". According to Huber and Klein, this dialectic also leads to a clearer mental representation of theological concepts in "highly religious" people: "The more religious someone is, the more a sediment of the denominational doctrine seems to be present" (Huber and Klein 2009, p. 263). They show that the classical reformatory paradigm of faith's primacy in opposition to deeds can only be found unambiguously in the "highly religious". Against this background, we examine the mental representation of Luther's doctrine of justification by faith in this article. We assume that this teaching is more clearly pronounced in the group of the "highly religious" than in the group of the "religious". Huber and Richard (2010) showed, that "[f]rom a theological point of view the emotion of guilt is of special interest" (p. 38) because guilt and release can be felt simultaneously. The reason for this can be found in Luther's doctrine of justification: "In Luther's concept, the Christian is always both sinner and justified, simul iustus et peccator" (Huber and Klein 2009, p. 254). It is a logical conclusion that this can be found analogously on the level of feelings: guilt and forgiveness can be felt at the same time.
The aim of this article is to demonstrate the relationship between the centrality of one's religiosity and emotions towards God, especially the emotions guilt and release. The analysis consists of a post-hoc analysis, a factor analytical approach, of the data collected as part of the Empirica Youth Survey 2018. This study aims at gaining detailed insight into the lives and beliefs of young, "highly religious" Protestants as well as understanding how this particular group's beliefs are reflected in their everyday life 1 . In this paper, we do a three-step analysis: first, we run an explorative factor analysis to examine to what extent the EtG are driven by common factors within the "highly religious" Protestant adolescents and young adults. In a second step, we replicate this with only the "religious" sample to find out how these two groups differ concerning the EtG. In a last step, we run a multiple regression with the "highly religious" sample, predicting one of the factors (the guilt-factor) using the demographics age and gender as well as the denominations.

Description of Data and Sample
The data are from the Empirica Youth Study 2018 (n = 3187). Our main object of research in this study was "highly religious" Protestant adolescents and young adults (from 14 to 29 years). Using the centrality score that Huber established (explained further in 2.2), 2386 participants (74.9%) could be identified as "highly religious", 650 (25.1%) as "religious".
In this article, the term "Protestant" can be understood as those who associate themselves with Mainline Protestant churches, the Pietistic Fellowship within the mainline Protestant churches (Landeskirchliche Gemeinschaft) or one of the many Free churches in Germany. We distinguished between the subjective feeling of belonging and actual formal belonging (as in church membership), because it had been shown (Faix et al. 2018, p. 21;Faix and Künkler 2018, p. 161) that the feeling 1 For a detailed description of the methodology, the methodical procedures, the sample characteristics and the results of the Empirica Youth Survey 2018 see: . A further insight into the Empirica Youth Survey 2018 and for sociological, theological and pedagogical perspectives onto the results, see: ). of belonging is more important than actual membership. Thus, in the following, we focus on subjective self-association.
The selective sample consists of three subsamples: (a) computer-assisted survey at two religious youth events "Christival" 2016 and "Jugendkirchentag" 2016, (b) computer-assisted online survey through promotion and invitations on websites and blogs, (c) computer-assisted online survey through invitations via e-mail, Twitter and Facebook.
Basic descriptive statistics of the sample of the "highly religious" adolescents and young adults such as gender, mean age and denomination can be found below, in Table 1 as well as the descriptive statistics of the "religious" comparison group. Most of the descriptive statistics such as gender and age are relatively equally distributed between the "religious" and the "highly religious", which means that they are comparable. However, concerning denominations, a dissimilar distribution becomes apparent: there is a much higher proportion of people that are part of a Free church within the "highly religious" group. According to Pearson's Chi-squared test (p-value 0.000), this is not due to coincidence.

The Use of the CRS
We used Huber's well-established Centrality of Religiosity Scale (CRS) to measure religiosity (Huber and Huber 2012). While varying scales comprise between 5 and 20 items (Huber and Huber 2012, p. 717), we used the version with five items (CRS-5). These items cover the five dimensions of religiosity and consist of one question each 2 : Each of the five items was coded on a scale of 1 to 5. According to Huber and Huber (2012), this allows for a distinction between different groups of religious centrality: People with a mean smaller than 2.0 were classified as "non-religious", which led to them being excluded from this sample. A mean of 2.1-3.9 indicated that a person is "religious" and people with a mean of 4.0-5.0 were classified as "highly-religious". 2 The Empirica Youth Survey was done in Germany and in German. For the purpose of this article, we use the translations of the items here.
A principal component analysis was performed to determine to which extent the Centrality of Religiosity Scale was driven by an underlying common component. The Kaiser-Meyer-Olkin-criterion is 0.809, which presents a sufficiently high and good correlation between the items in order to perform the principal component analysis. It is hereby confirmed that the dataset is appropriate. (Kaiser 1974). According to criterions such as the Kaiser criterion, the Scree-Plot, MAP-Tests as well as Horn's parallel analysis (Guttman 1954;Kaiser 1974), one common component can be assumed. Therefore, the different items of the Centrality of Religiosity Scale can be used as one-the centrality of religiosity.
Regarding the CRS-subscales (Table 2), both the "highly religious" and "religious" indicate the smallest mean on the experience dimension (3.80/2.66). The highest mean of the "highly religious" can be found on the ideology dimension of the CRS (4.87), followed by private practice (4.80). In comparison, the highest CRS subscale means of the "religious" are public practice (3.47) and intellect (3.39).

Measuring Emotions towards God
In the Empirica Youth Study 2018, we decided to measure the EtG essentially on the basis of Murken et al. (2011). Table 3 shows all items on Murken's Scale (Murken et al. 2011, p. 85) and shows which of them were used in our analysis: The reason why not all items were taken from Murken's scale is that neither images of God nor specific feelings towards God were the main focus of the Empirica youth study. In order to do justice to the explorative character of this study, we had to compromise in order to keep the questionnaire manageable. However, we added the emotion of "release from guilt" from the EtG by Huber and Richard, because from our field knowledge we assumed that the release from guilt has a high priority in a Protestant, highly religious context. In sum, we measured emotions towards God with the following items using a 5-stage response format (1 = never, 2 = rarely, 3 = sometimes, 4 = often, 5 = very much): How often do you experience . . . Looking at positive emotions towards God (Table 4), both "highly religious" as well as "religious" adolescents and young adults indicate the highest mean when it comes to experiencing gratitude (4.49/3.75) and the smallest concerning experiencing release (3.74/2.83). Before examining the negative EtG, it is important to note that positive emotions towards God are experienced more strongly than negative emotions. Guilt is the emotion with the highest mean of all the negative EtG. Furthermore, when comparing the "highly religious" and "religious" group, they differ most in the aspect of guilt when it comes to negative EtG (3.21/2.59). Table 4. Emotions towards God Scale of the sample.

Religious N (%) or M (SD)
Emotions towards God (

Analytic Strategy
As an analytic strategy, we used an exploratory approach to examine the correlation between emotions towards God and the Centrality of religiosity by doing a three-step analysis: first, we run an explorative factor analysis to examine to what extend the EtG are driven by common factors within the "highly religious" Protestant adolescents and young adults. In a second step, we replicate this with only the "religious" sample to find out how these two groups differ concerning the EtG. In a last step, we run a multiple regression with the "highly religious" sample predicting one of the factors (the guilt-factor) using the demographics age and gender as well as the denominations.

Results
In the following paragraphs, we show the results of the factor analysis of the group of "highly religious". Furthermore, we explain briefly how the factor analysis of the "religious" adolescents and young adults differs from the previous. After the factor analysis, we point out what influences "highly religious" adolescents' and young adults' experience of the guilt-factor and to what extent. We do so by presenting the outcomes of the multiple linear regression.

Results of the Factor Analysis of the "Highly Religious"
Before conducting a factor analysis, we take a look at the correlations between (a) emotions towards God and the Centrality of Religiosity Scale and (b) emotions towards God compared with one another (Table 5). Note. This contains the whole sample ("highly religious" and "religious").
There are large correlations between emotions of protection, gratitude, release, love and guilt towards God and the Centrality of Religiosity Scale. Guilt is the only negative EtG that correlates with the CRS by more than r = 0.050. When looking at the correlations between the different EtG, it is striking that, in most cases, positive EtG correlates negatively with the negative EtG and the other way around. Nonetheless, there is an exception: guilt correlates more strongly with positive EtG. For instance, the correlation between release and guilt is r = 0.365. There is even a small positive correlation between fear and release (r = 0.080), which is an exception to the negative correlations between fear and other positive EtG. All in all, this implies an underlying factor that connects guilt, fear and release with one another. We examine this implication further.
Against the background of these correlations and the examined theories on emotions towards God (see Section 2.2.2), we ran an exploratory factor analysis. The goal was to identify to which extent the different emotions towards God are driven by common factors. After conducting a principal factor analysis, all criteria (eigenvalues, screeplot, cumulative variance) (Table 6) suggested that three factors for the "highly religious" adolescents and young adults should be used. Table 6. Criteria for the factor analysis of the "highly religious" adolescents and young adults (N = 2103). We used an Iterated Principal Factor analysis because it neither over nor underestimates the commonalities through repeated estimations. We further used an orthogonal Varimax rotation with a horst-modification which normalizes the loadings with the cumunalities to prevent an overestimation of larger communalities (Kopp and Lois 2012, p. 90;Horst 1965, p. 428). Furthermore, we applied the Kaiser-Meyer-Olkin measure of sampling adequacy, the Horn's parallel analysis and tested the Cronbach's alpha. Cronbach's alpha as measurement of internal reliability of the EtG-Scale for the "highly religious" is 0.5896 and therefore acceptable.

Component
The factor analysis of the "highly religious" sample shows that there are three factors: The factor for positive EtG contains protection (0.69), release (0.48), love (0.71), and gratitude (0.49) and the factor for negative EtG contains fear (0.32), trouble (0.83) and disappointment (0.72). The third factor encompasses fear (0.40), guilt (0.73) and release (0.30). Though it is mainly driven by guilt (0.73), there is a cross-loading between the positive and the guilt-factor on the emotion of release and a crossloading between the negative and the guilt-factor on the emotion of fear. Even though release and fear both have loadings on two factors, we continued without excluding it (see Section 4). Thus, we continued the analysis with the cross-loadings of release and fear. The uniqueness scores are in a good range (all under 0.8 and mostly around 0.5). Moreover, the score of univocality is good. The difference between estimated scores and the implicated correlations is >0.8 and the model has good reliability scores (standard deviation between 0.77 and 0.87).
The results of the factor analysis show the factorial structure of the EtG in the "highly religious": a factor for positive emotions (protection, love, gratitude and release), one for negative emotions (rage, disappointment and fear) and another factor that is driven by guilt, release and fear. One common dimension seems to underlie experiencing fear, guilt and release. The structure of the three factors of EtG of the "highly religious" adolescents and young adults are represented in Table 7. We consider larger loadings when they are >0.5 (Kopp and Lois 2012, p. 90). The loadings (Table 7) of different emotions on the factors underline what the correlation matrix (Table 3) already suggested: the third factor, despite containing release and fear, is mainly driven by guilt (0.73). That is why we called it the guilt-factor.

Results of the Factor Analysis of the "Religious"
As said, the previous factor analysis was only conducted with the "highly religious" sample. In order to understand whether the factorial structure was only a "highly religious" phenomenon we applied the same procedure to the "religious" group. We did a factor analysis with the "religious" adolescents and young adults in the same way as with the "highly religious". When the factor analysis is done for the "religious" adolescents and young adults, only two factors are recommended by the eigenvalues, the screeplot and the cumulative variance (Table 8). Table 8. Criteria for the factor analysis of the "religious" adolescents and young adults (N = 649) *. We also tested for the internal reliability of the EtG-Scale of the "religious" sample with Cronbach's alpha, with the result of 0.7243. An IPF and orthogonal Varimax rotation with a horst-modification result in two factors: positive and negative EtG with the EtG guilt loading on both factors with a very small difference (>0.01). We decided to exclude the emotion guilt for the "religious" because of the problematic cross-loading with a mere difference of >0.01 (see Table 9 for the results).

Results of the Multiple Linear Regression
After the factor analysis, we examined the effects of other additional variables on experiencing the guilt-factor that was extracted for the "highly religious" sample using a multiple linear regression. The standard demographics, gender and age were assumed to influence how emotions towards God are experienced. We also tested for the effect of denominations. The descriptive statistics shown in Table 1 allowed for the comparison of the adolescents and young adults with one another because they were equally distributed, except for the denominations. Furthermore, we analysed correlations and significance by conducting t-tests (p-values = 0.00) and correlation coefficients to verify that age, gender and denomination could be of importance in the multiple linear regression.
As Table 10 shows, age has a highly significant effect on how the guilt-factor is experienced. The older the adolescents and young adults, the less they experience the combination of guilt, release and fear. The highly significant results of the multiple linear regression after the factor analysis further emphasize that "highly religious" men experience the guilt-factor more frequently than "highly religious" women do. These results even increase slightly upon taking denominations into account in the multiple linear regression model. In comparison to people from the mainline Protestant churches, adolescents and young adults from the Pietistic Fellowship within the mainline Protestant churches experience slightly more guilt. Moreover, the "highly religious" adolescents and young adults from Free churches experience the guilt-factor the most. This outcome is highly significant as well.

Discussion
When comparing the factor analysis of the "highly religious" with the "religious" of this sample, it must be kept in mind that the sample contains only a small number of "religious" people in comparison to "highly religious" ones-who were the main object of research. In conclusion, statements about the "religious" adolescents and young adults in this sample should be made with caution, due to the lack of reliability. For example, the "religious" adolescents and young adults in the Empirica sample may be more religious than other religious people their age, which is very likely because they participated at religious events or read a religious magazine where the Empirica Youth Survey 2018 was advertised. In this regard, it is also interesting to pick up on the means of the CRS in Table 2: both the "religious" and "highly religious" have high CRS means. The "highly religious" adolescents and young adults from this sample have a CRS mean of 4.49, which is very high compared to Huber and Richard (2010, p. 29) and may partly be an outcome of the sampling strategy itself (for example: a big part of the participants completed the survey at an event for religious adolescents and young adults). This would also make sense regarding the comparatively high CRS mean of 3.31 of the "religious" adolescents and young adults.
Another limitation of this article is that we did not control for effects of a person's personality, for example, the Big-5, as Huber and Richard (2010, p. 28) do in their research. Moreover, it is important to note that our findings are only applicable to the "highly religious" Protestant adolescents and young adults. Because the different denominations in Germany are a specific German phenomenon, it is difficult to generalize the outcomes for other countries.
Further, it should be noted that more restrictive handling of cross-loadings could lead to different outcomes. Gorsuch (1983) critiqued that cross-loadings with a difference under 0.30 should not be interpreted. However, his critique has mainly been applied to smaller samples. In our analysis, we excluded the emotion of guilt from the factor analysis of the "religious" because of the crossloading with a difference <0.01. The other two crossloadings within the "highly religious" sample had larger differences. However, had we applied a more restrictive handling of crossloadings, the factor structure would be different. Furthermore, we interpreted smaller loadings than >0.5. Therefore we need to be aware that the guilt-factor is mainly driven by guilt-release and fear have relatively small loadings. However, due to the high number of participants, it is unlikely that the cross-loading emerged by coincidence.
In summary, we conclude that there is a difference between the "religious" and "highly religious" adolescents and young adults in how their negative and positive emotions are correlated, and that there are not two (positive and negative) but three factors for "highly religious" adolescents and young adults. The third factor in the EtG contains guilt, release and fear of God and is mainly driven by guilt. Comparing the exploratory factor analyses of "religious" and "highly religious" Protestant adolescents and young adults, we conclude that the emotions guilt, release and fear are not driven by the same factorial variable for the "religious" adolescents and young adults but are experienced as positive and negative emotions. In addition, experiencing guilt cannot be assigned clearly to only the positive or the negative factor when it comes to the "religious" adolescents and young adults. In comparison, fear, guilt and release can be traced back to one common factor in the case of the "highly religious". This suggests that the correlation between the 'emotions towards God' -fear, guilt and release-is a phenomenon that can only be found within the "highly religious" and disappears in a "religious" sample. The results tie in with Huber and Klein's findings, who found both positive and negative emotions towards God among "highly religious" people (Huber and Klein 2009). Huber and Klein explain this as a theological sediment of Luther's doctrine of justification, in which people are both sinners and redeemed. The classically reformatory paradigm of faith's primacy in opposition to deeds could only be found unambiguously in the highly religious, or as Huber and Klein (2009, p. 263) state: "The more religious someone is, the more a sediment of the denominational doctrine seems to be present". In our results, these characteristics are even more evident. They therefore confirm the findings of Huber and Klein that there is a qualitative difference between "religious" and "highly religious" people, that the latter have a more complex mental representation of the structure of religious emotions than "religious" people and that this is can be interpreted as a theological sediment. Our results also show the interaction of cognitive and affective elements in the God image, as described above. Interpreted as theological sediments, they also confirm the influence of the specific religious culture on the God image. As Schaap-Jonker (2018, p. 12) points out, "one cannot study God representations without considering what culturally specific systems, as bearers of religious traditions, (also) constitute the content of these representations". However, it is also possible to interpret this findings as an reflection of a general psychological phenomenon. In the introduction, we briefly referred to the QGR scale, which also includes the findings of three subscales with respect to feelings towards God. Schaap-Jonker (2018, p. 10) interprets these within a relational psychological framework. In the tradition of Kirkpatrick's correspondence thesis (Kirkpatrick 1992), early relational experiences contribute to the formation of attachment styles which then contribute to the formation of the God image. Against this background, the QGR scale can also be explained psychologically in its tripolarity: "they reflect an attachment relationship with God that may be characterized by closeness, love and affection to a supporting, patient, and protecting God who is unconditionally open, and/or by feelings of fear of being rejected or punished by a God who judges and exerts power, and/or by angry and disappointed feelings because of a God who does not care, leaving people to their own devices" (Schaap-Jonker 2018, p. 134). This logic, which can also be found in Murken et al. (2011) as described, could also explain the three factors in our findings. These would then reflect a general psychological phenomenon, in which anxiety represents an adaptive trait. However, then it remains unclear why fear, trouble and disappointment can only be attributed to one factor among the "religious" and why in the EtG of the "religious" negative emotions directed at God (trouble and disappointment) coincide with negative emotions that concern the self in relation with God (fear). At least, the thesis of a stronger differentiation of the God-representations among the highly religious would then also be confirmed.
Regarding our findings of the multiple regression with the "highly religious" sample predicting the guilt-factor, our findings of the effect of age differ from other studies. In our sample, age has a negative effect on experiencing the guilt factor. Others, for example Braam et al., find the opposite with their sample of older adults and they assume that to feel more guilt could also be a matter of age, for example, through a longer life people could grow more aware of their faults and mistakes in life (Bramm et al. 2008, p. 234). They also find that age is positively correlated with having an image of a punishing God (Bramm et al. 2008, p. 231). It is not surprising that the male respondants are more likely to experience the guilt-factor. This tendency corresponds to the fact that, overall in the Empirica youth survey, female respondents seem to have a statistically significant image of a more comforting God and the male respondents a more angry God-image (Faix et al. 2018, p. 64).
The adolescents and young adults from the Mainline Protestant churches are the group which experiences the guilt dimension the least. Members of Free churches, on the other hand, experience the guilt-factor the most in comparison with the other denominations. Within the denominations, only the outcomes for those from Free churches are statistically significant. There might be several reasons for this finding: in order to exemplify this, we will elaborate on the role of Luther's doctrine of justification. With the emergence of liberal theology and exegesis in the history of religion in the 19th century, criticism of Luther's doctrine of justification became louder on the Protestant side, and Luther's interpretation of man being a sinner was especially criticized. A person does not have to know about her or his guilt first in order to be redeemed but can also live from the Gospel alone (Schleiermacher [1799(Schleiermacher [ ] 1984Schweitzer 1951). This distinction is examined by the religious psychologist William James in his work "The Varieties of Religious Experience: A Study in Human Nature" by distinguishing between those 'born once and twice' (James 1902, pp. 152-58). The 'once-born' live their faith in an unbroken (healthy) life without a sense of guilt and damnation, sustained by the saving gospel alone. The 'twice-born' have had the disruptive (pathological) experience of guilt and despair and now need (as Luther postulated) redemption and salvation. For James, these two different religious experiences are of fundamental importance, since religious feelings reveal the original religious event and shape "in the most intimate layer of personal self-consciousness, an event that concerns the immediacy of self-consciousness" (James 1902, p. 159). For James, these religious experiences were decisive in the development of his own faith and the understanding of his own humanity. James examines religious experiences, in particular the different feelings that arise from them in detail and relates them to the conversion and rebirth of a believer (James 1902, p. 57). Conversion, however, is experienced only by those born twice, because it relates to feelings of guilt and loss (James 1902, p. 189). If we follow James here, our results correspond with his description of the 'twice-born', because only they feel such strong and ambivalent emotions like fear, guilt and liberation towards God. For James, these experiences play a central role, because the negative experiences can help to develop a more mature faith and a more mature personality. Just as repentance and rebirth are part of faith for those born twice, pain and fear are as much a part of life as hope and healing (James 1902, p. 189). Charles Taylor incorporates these thoughts in his book "Varieties of Religion Today: William James Revisited" and postulates that the mere experience of negative feelings such as guilt and fear and overcoming them represent a deeper and truer experience than the purely positive experiences of the 'once-born', and thus form an authentic faith (Taylor 2003, p. 33). Therefore, the decisive point in one's own religiosity is not mainly the dogmatic cognitive understanding of faith, but the subjective experience and the feelings experienced in the process. Regarding our results, this could mean that sinfulness is less emphasized in the doctrine of the Mainline Protestant church context and that the "lostness of man" plays a subordinate role compared to the Free church context. This is also apparent in the example of conversion, which plays a much more central role in the Free church context than in the context of the Evangelical Church Germany (Evangelische Kirche Deutschland-EKD) (Kißkalt 2003, p. 171). Accordingly, in large parts of the Free church context, the ideology of the 'twice-born' is dominant, in which Christianity requires the acknowledgment of guilt and conversion. Huber and Richard (2010) come to similar conclusions in their studies, finding that committed Free church members (as opposed to committed church members, church members and non-denominational members) feel a higher degree of positive (release, etc.) and at the same time negative (guilt, etc.) feelings (Huber and Richard 2010, p. 29). The more "highly religious" people believe that they experience God and believe in him, the more positive or negative their feelings are (Huber and Richard 2010, p. 36).

Conclusions
For the "highly religious" three factors underlie emotions towards God: a positive factor that contains protection, love, gratitude and release; a negative factor that contains trouble and disappointment and a third factor, which is mainly driven by guilt but further contains release and fear.
Comparing the "highly religious" and "religious" we find that the strong correlation between the EtG fear, guilt and release is a phenomenon that can only be found within the "highly religious".
Moreover, we find that gender, age and denomination have significant effects on how the guilt-factor is experienced.
In comparison to Murken et al. (2011, pp. 77, 83), who found three factors for the Emotions towards God, positive emotions, negative emotions towards one's self and negative emotions towards God, we exploratively discovered the three factors of positive emotions, negative emotions and guilt.
Unlike Huber, we do not only find a connection between release and guilt for the "highly religious" adolescents and young adults, we find a third factor that contains fear in addition to release and guilt. This is alarming since Zwingmann et al. and others have shown the correlation between religious feelings and mental health (Zwingmann et al. 2017;Winter et al. 2009). The negative feelings of guilt and fear, as part of relational aspects of religion which "seem to be most strongly linked to mental health" (Schaap-Jonker 2018, p. 8), can especially bring about negative developments in the development of faith and personality and must, therefore, be viewed critically. On the other hand, it is important to take James' and Taylor's classification into account who speak of overcoming and processing negative feelings in order to achieve a more mature faith. In James' case, this maturing of faith was reflected in his time of science and enlightenment in the increasing conflict between faith and unbelief. Taylor points out that, more than 100 years later, this conflict is even more important (Taylor 2003, p. 55). Thus, for further reflection, the question of how people with negative religious emotions such as guilt and fear can be helped to understand, process and integrate these feelings for a more mature faith arises. Funding: The Empirica Youth Survey 2018 on whose data this research is based was funded by the 'Amt für Jugendarbeit der Evangelischen Kirche von Westfalen', 'Wertestarter-Stiftung für Christliche Wertebildung' and the 'Heinz-Horst Deichmann-Stiftung'.