A FURTHER ELABORATION OF THE POST-CRITICAL BELIEF SCALE : EVIDENCE FOR THE EXISTENCE OF FOUR DIFFERENT APPROACHES TO RELIGION IN FLANDERS-BELGIUM

To measure religious attitudes in a secularized context, Hutsbaut (1996; 1997) constructed the Post-Critical Belief scale, which was considered to contain three subscales (Orthodoxy, External Critique and Historical Relativism). In a first study it is demonstrated that the two bipolar dimensions of the model of Wulff (1991; 1997) (Inclusion vs. Exclusion of Transcendence and Literal vs. Symbolic) chararcterized the internal structure of the Post-Critical Belief scale. In a twodimensional representation of the Post-Critical Belief scale, four instead of three subscales emerged (Orthodoxy, External Critique, Relativism and Second Naïveté). Each of these subscales relates to one of the four quadrants in Wulff's model. Based on these findings a revised Post-Critical Belief scale was constructed, in order to elaborate each of the four subscales. This revised version was put to the test in a second and third study. The internal structure was demonstrated to be stable and replicable. The consequences of the revision are illustrated via the relationships between the subscales and racism.

ments, regardless of their specific content, thus presenting themselves as either indiscriminately pro-or anti-religious.An explanation for this might be that, in highly secularized countries -like Belgium and most other European countries -religion no longer holds an important social function.In contrast to the USA, it is seen as a private matter (Dobbelaere, 1996).When there is no reason to be extrinsically religious, a strong relationship is to be expected between the intrinsic value of one's belief and subscribing to items expressing the social benefits of church attendance or the instrumentality of prayer.Furthermore, since it is no longer obvious to be religious in a secularized context, it is likely that people who intrinsically value their belief will also question it.
Recently, Hutsebaut (1996;1997) constructed the Post-Critical Belief scale, a new multidimensional instrument for measuring religious attitudes in a secularized context.In the present article, we present three subsequent studies that shed a new light on the validity and the interpretation of the subscales of the Post-Critical Belief scale.Before introducing these three studies, the Post-Critical Belief scale and its theoretical framework will be presented.
The Post-Critical Belief scale Theoretical model.Hutsebaut (1996;1997) based the construction of the Post-Critical Belief scale on Wulff's (1991;1997) model -one of the dominant models in the psychology of religion nowadayswhich integrates various approaches to religion into one comprehensive framework.In the epilogue of his book, Wulff (1991;1997) argued that the various approaches to religion can be located in a twodimensional space (see Figure 1).The vertical axis specifies the degree to which the objects of religious interest are granted participation in a transcendent reality (inclusion vs. exclusion of transcendence).The horizontal axis indicates whether religion is interpreted literally or symbolically (literal vs. symbolic).These two dimensions define four basic attitudes toward religion: Literal Affirmation, Literal Disaffirmation, Reductive Interpretation and Restorative Interpretation (see Figure 1).

Insert Figure 1 about here
The upper left quadrant, Literal Affirmation, represents a position in which the literal existence of the religious realm is affirmed.This position is most clearly embodied by religious fundamentalism.
Elements of this posture, however, also appear among those who are not particularly conservative but nevertheless do not want to hear that all Bible stories are legends.According to Wulff (1991;1997), this position, which is partly reflected in scales like Barron's (1963) Fundamentalist Belief scale, can only be sustained by rejecting those who doubt the validity of the conservative view.Therefore, it should not be surprising that, according to Wulff (1991;1997), literal believers tend to score higher on measures of prejudice and lower on cognitive development, and can be characterized as rigid and low in ability to adapt.
The lower left quadrant, Literal Disaffirmation, represents a position in which the existence of the religious realm is rejected, but in which the possibility that religious language has a symbolic meaning is lost out of sight.So, like in the first quadrant, religious language is understood in a literal way.The difference, however, lies in the rejection of what is written or said.If anything is made absolute, it is rational, formal principles of knowledge, or scientific method.According to Wullf, this position is partly reflected in what is identified in the literature as the indiscriminately anti-religious orientation (Allport & Ross, 1967), and in Barron's (1963) Fundamentalist Disbelief scale.People occupying this quadrant tend to be less dogmatic and more intellectual than many of the religious subjects, but also less fair-minded, less capable to evaluate ideas, rather rigid and low in ability to adapt.
The lower right quadrant, Reductive Interpretation, represents a position in which the existence of the religious realm is rejected, and in which a privileged perspective of the true but hidden meaning of religion's myths and rituals is claimed.This quadrant is derived from the work of Ricoeur (1970), who proposed that in modern hermeneutics, in order to clear away from religious symbols the excrescence of idolatry and illusion, a Reductive Interpretation would be necessary.In this respect, Ricoeur refers to the so-called Masters of Suspicion (Marx, Freud, & Nietzsche).Wulff (1991;1997) draws on findings obtained with closely related scales, such as Batson's (1976) Quest scale and Barron's (1963) Enlightenment Disbelief scale, to fill out a portrait of persons in this quadrant, and concludes that these persons are complex, socially sensitive and insightful, relatively unprejudiced and original.
The upper right quadrant, Restorative Interpretation, represents a position in which the existence of the religious realm is affirmed, but in which one tries to encompass and go beyond all possible reductive interpretations in order to find the symbolic meaning of the religious language.Again, this quadrant is derived from the work of Ricoeur (1970), who proposed that in modern hermeneutics, in order to make it possible for the object of suspicion to be restored to an object of understanding and faith, a Restorative Interpretation would be necessary.In this respect, Ricoeur introduced the concept of Second Naïveté.According to Wulff (1991;1997), characterizing persons who occupy this position is somewhat more difficult, for until recently, they have been largely neglected in empirical research.
Moreover, given its profoundly individual character, this quadrant is probably also the most difficult to operationalize with standardized questionnaires.Nevertheless, this posture might be represented by Fowler's (1981) fifth stage (conjunctive faith).Research with scales designed to operationalize this stage might therefore cast light on this position.
Construction of the Post-Critical Belief scale.Inspired by the ideas of Ricoeur (1970) and Wulff's integrative framework on the various approaches to religion (Wulff, 1991), Hutsebaut (1996) formulated 24 statements which were assumed to capture the various approaches to religion within a Roman Catholic (but secularized) context.Especially the concept of Second Naïveté was focused on, since it plays a central role in Wulff's model, but had been barely empirically investigated.When these statements were presented to a sample of adult subjects, three predominantly unipolar factors emerged after applying factor analysis followed by a VARIMAX rotation.The three rotated factors were called Orthodoxy, External Critique and Historical Relativism.Based on the item content and the external relationships of the factors, Orthodoxy was interpreted as a measure of Literal Affirmation, External Critique as a measure of Reductive Interpretation and Historical Relativism as a of measure Restorative Interpretation.For instance, Orthodoxy correlated positively with literal thinking and religious certainty, Historical Relativism correlated positively with quest belief and openness for complex questions, and External Critique correlated positively with autonomy and negatively with religious certainty.In subsequent studies, factor analysis confirmed the existence of these three factors (Hutsebaut, 1997;Desimpelaere et al., 1999) and supported the interpretation of Orthodoxy and Historical Relativism.
The findings of Desimpelaere et al. (1999), however, raised doubt about the interpretation of External Critique.Both Orthodoxy and External Critique were found to be positively related to Perry's (1970) psycho-epistemological style of Dualism.This style is characterized by a preference for unambiguity: There is only one right answer for each problem.These findings suggested that both Orthodoxy and External Critique are rooted in a literal mode of thinking and that the External Critique items, in terms of Wullf's model, were measuring Literal Disaffirmation instead of Reductive Interpretation.This was the focus of the first study.In this study, we applied various methods of analysis in order to get a better view on the internal structure of this scale and its possible relation to Wullf's model.These led to a revision of the Post-Critical Belief scale.In order to represent more fully the religious attitude domain, several new items were constructed.The validity of the revised version of the Post-Critical Belief scale is investigated in the second and the third study.

Perspectives and problems of the
Racism and the Post-Critical Belief scale.The problem of the interpretation of External Critique is being investigated by focusing on the relationships between religious attitudes and racism.This concept was chosen as a marker for the discriminant validity of the subscales of the Post-Critical Belief scale, because the relationship between religiosity and racism is probably the most important paradox within the psychology of religion.Whereas all world religions proclaim brotherly love, history is littered with moments in which religion has provided a justification for, or has given cause to, all kinds of atrocities directed towards people from a different race or culture.A number of historians and theologians concluded from this that religion should contrarily be considered as a catalyst for prejudice, racism, etc., and a lot of psychological and sociological research has been carried out to investigate whether this is true (for a recent overview: see Duriez & Hutsebaut, 2000).If the Post-Critical Belief scale can be interpreted in terms of Wullf's model (1991;1997), straightforward predictions can be made with regard to the relationship with racism.Orthodoxy, characterized by a rather fundamentalist closed-mindedness, should correlate positively with racism.Historical Relativism, characterized by both an open and symbolic approach and an adherence to the core value of universal brotherly love that is been proclaimed by Christianity, should correlate negatively with racism.The correlations with External Critique, though, depend on its position in Wullf's framework.If External Critique measures Reductive Interpretation, which is related to social sensitiveness and insightfullness (Wulff, 1991;1997), then a negative correlation with racism is to be expected.However, if External Critique measures Literal Disaffirmation, a positive correlation with racism is to be expected.Intolerance for ambiguity, which is characteristic of literal thinking, is related to a higher susceptibility to racism (Sidanius, 1985).Thus, a negative correlation between External Critique and racism would support Hutsebaut's original interpretation (Hutsebaut, 1996;1997) while a positive correlation would support the interpretation of Desimpelaere et al. (1999).Whether and to which extent the correlations with the dimensions of the Post-Critical Belief scale and racism support the interpretation of the internal structure of the Post-Critical Belief scale is being investigated in all three studies.

Introduction
In a first study, we probed the internal structure of the original 24-item version of the Post-Critical Belief scale.Furthermore, we focused on the relationships between the different subscales of the Post-Critical Belief scale and racism.

Method
Sample.The questionnaire was distributed by undergraduate students of the Faculty of Psychology of the Katholieke Universiteit Leuven (Belgium) in (Roman Catholic) schools and via organizations, relatives and friends.In total, they distributed 1500 questionnaires, of which 77% were returned.Some of them, however, were only partially filled out.All respondents were Flemish speaking Belgians; 56.5% were female.The age groups were as follows: 16-18 (21.9%), 19-25 (23.4%), 26-35 (7.8%), 36-45 (10.4%), 46-65 (26.7%) and 65+ (9.8%).The highest educated group of people (18.6%) consisted of people who had obtained a university degree or were university students, the second group of people (31.1%) had obtained a higher non-university degree or were attending higher non-university studies, the third group of people (29.9%) quit studying after completing secondary school or were currently last-year secondary school students, the fourth group of people (7.7%) quit studying after having completed a technical education (TSO), the fifth group of people (4.2%) quit studying after having completed a vocational education (BSO), and the lowest educated group of people (5.8%) consisted of people whose highest diploma was a primary school diploma.The educational level was high compared to that of the Flemish population.The level of Sunday mass attendance was also high: whereas in the general population only about 10% regularly attends (Office of Church Statistics, Brussels, personal communication, September, 2000), 44.2% persons indicated to attend weekly or at least often, 29.3% indicated to attend on special occasions only, and 24.9% indicated to never attend.

Results
Multidimensional Scaling on the Post-Critical Belief scale.We analyzed the internal structure of the Post-Critical Belief scale via multidimensional scaling.Multidimensional scaling unveils the "hidden structure" of items (or stimuli in general) by representing them as points in a geometrical space in such a way that the geometrical distances between the points reflect the psychological dissimilarities between the items as well as possible (Borg & Groenen, 1997;Davison, 1983;Kruskal & Wish, 1978).
We carried out a non-metric multidimensional scaling2 on the dissimilarity data between the 24 items of the original Post-Critical Belief scale. 3As a dissimilarity measure, we used the Euclidean distances between the standardized items.The two-dimensional solution had a Kruskal stress of .09,explained 96% of the variance in the optimally transformed dissimilarities, and can be said to fit the data well (Borg & Groenen, 1997, p. 38). 4 One way to interpret a configuration generated by multidimensional scaling is the neighborhood interpretation (Kruskal & Wish, 1978, p. 43).In this interpretation one looks for regions within the configuration that contain a distinguishable type of items.In a two-dimensional representation of the 24 Post-Critical Belief items, three regions could be identified, each of them containing items supposed to belong to one single subscale (see Figure 2).The upper left quadrant, quadrant 1, encompasses the Orthodoxy items (except for item 01-).The lower left quadrant, quadrant 2, encompasses the External Critique items (except for item E5-).Both the upper and the lower right quadrants, quadrants 3 and 4, encompass the Historical Relativism items.The Historical Relativism items are spread out over the two right quadrants.A content analysis of the items indicated interpretative differences for the items that are situated in the upper right quadrant, quadrant 4, and the items situated in the lower right quadrant, quadrant 3 (see Figure 2 and Appendix).The items of quadrant 4 clearly capture an inclusion of tran-scendence, while the items in quadrant 3 capture a more relativist stance.Therefore, conceptually, the items of quadrant 4 can be interpreted in terms of Restorative Interpretation and can be seen as an measuring what Ricoeur (1970) called Second Naïveté.Taking a relativistic stance towards religion fits the Reductive Interpretation in Wullf's model.Because of the content of the items in quadrant 3 we will refer to this quadrant as Relativism.The two items that were reversed in the original use of the Post-Critical Belief scale (item O1-and item E5-) (Hutsebaut, 1996;1997;Desimpelaere et al., 1999) emerged opposite of the other items belonging to the same subscales.This could be expected, since we decided to stay close to the data and did not reverse these items.Because these items emerged in a meaningful part of the configuration without reversion, it is meaningful to consider them to belong to a different Post-Critical Belief subscale: item O1-can be considered to belong to External Critique, whereas item E5-can be considered to belong to Historical Relativism.

Insert Figure 2 about here
If External Critique constitutes an operationalization of Literal Disaffirmation, as hypothesized by Desimpelaere et al. (1999), the dimensions of the two-dimensional representation can be interpreted perfectly in terms of Wulff's model.The first bipolar dimension can be interpreted as Inclusion vs. Exclusion of Transcendence (see Figure 2): The items that were intended to measure the affirmation of the transcendent realm and the items that were intended to measure disaffirmation get clearly separated in this dimension.The second bipolar dimension can clearly be interpreted as Literal vs. Symbolic (see Figure 2): The items measuring literal (dis)affirmation of the transcendent realm and the items measuring symbolic (dis)affirmation get separated in this dimension.The present two-dimensional representation is considered as a point of reference for the internal structure analyses throughout this article.
Principal Component Analysis on the Post-Critical Belief scale.The discrepancy between the results of the multidimensional scaling revealing two bipolar dimensions as predicted by the model of Wullf (1991;1997) and the discovery of three (predominantly) unipolar factors by factor analyses after VARIMAX rotation (Hutsebaut, 1996;1997;Desimpelaere et al., 1999) calls for an explanation.
Therefore, we subjected the responses to the 24-item Post-Critical Belief scale to a principal component analysis.In line with the original outlook on the Post-Critical Belief scale, Cattell's scree test (Cattell, 1966) pointed to three components. 5A three-componential solution accounted for 45% of the total variance.In order to investigate the relationship with the two-dimensional solution of the multdimensional scaling, the three-component solution was rotated towards the two-dimensional solution using an orthogonal Procrustes rotation (Schönemann, 1966; see also McCrae et al., 1996).After .98% of the variance in the optimally transformed dissimilarities.The two dimensional solution was retained because this solution was theoretically well interpretable and already accounted for nearly all of the variance. 5The eigenvalues for the first six components, after extraction, were 5.51, 2.96, 2.32, 1.08, 1.02 and 0.90 respectively.doing this, two bipolar components and one unipolar component showed up (see Table 1).The first component could be interpreted as Inclusion vs. Exclusion of Transcendence: Both the Orthodoxy items and the Second Naïveté items loaded positively onto this dimension, whereas both the External Critique items and the Relativism items loaded negatively onto this dimension.The second component could be interpreted as Literal vs. Symbolic: Both the Orthodoxy items and the External Critique items loaded negatively onto this dimension whereas both the Relativism items and the Second Naïveté items loaded positively onto this dimension.Tucker's Phi indices suggested high congruence between the first two components and the two dimensions found by multidimensional scaling: .99 for Inclusion vs. Exclusion of Transcendence and .95for Literal vs. Symbolic.All items loaded positively on the third component.

Insert Table 1 about here
Post-Critical Belief subscales.For each of the three original subscales, scale scores were computed by averaging the item scores of the items theoretically belonging to these subscales. 6 Based on the mutidimensional scaling and the principal component analysis, Historical Relativism was split into two separate subscales: Relativism and Second Naïveté.Estimates of internal consistency (Cronbach alpha's) were .55 for Relativism (Mean=4.57,SD=1.28) (4 items), and .73 for Second Naiveté (Mean=3.92,SD=1.58) (4 items).The latter rose up to .82 when one item (H6) was left out of the analysis.We decided to do so.Relativism was negatively related to Orthodoxy (r(=-.23,p<.0001), positively related to External Critique (r=.31, p< .0001)and slightly positively related to Second Naiveté (r=.07, p<.05).Second Naiveté was positively related to Orthodoxy (r=.46,p<.0001)and negatively related to External Critique (r=-.51,p<.0001).
Relationships with racism.Racism was positively related to both Orthodoxy (r=.35, p<.0001) and External Critique (r=.14, p<.005), and was slightly negatively related to Historical Relativism (r=-.10,p<.05).However, taking into account the results of the previous analyses, it became apparent that the slightly negative correlation between Historical Relativism and racism was due to the existence of a negative correlation between Relativism and racism (r=-.16,p<.0005).Second Naiveté and racism were unrelated (r=-.03,n.s.). 76 Item O1-and item E5-were reversed as was done in the original scale. 7The relationships between the four religiosity scales and racism are not affected by the educational level of the subjects.In regression analyses with the religiosity scales, educational level and the interaction effects between the religiosity scales and educational level as predictors and racism as criterion, all interaction effects

Conclusion
Two bipolar dimensions were identified that could be interpreted in terms of Wulff's model: Inclusion versus Exclusion of Transcendence and Literal versus Symbolic (Wulff, 1991;1997).This representation sheds new light on the Post-Critical Belief subscales.Contrary to the original interpretation (Hutsebaut, 1996;1997), External Critique is situated at the literal side of the second dimension.Furthermore, Historical Relativism appears to be comprised of two subscales, Second Naïveté 8 and Relativism, which correspond to Wulff's Restorative and Reductive Interpretation respectively. 9 The stability of the configuration was supported via method convergence.After orthogonal Procrustes rotation of the component solution to the multidimensional scaling solution, the first two components were highly similar to the two dimensions.However, the third component cannot be interpreted in terms of any religious attitude dimension.After orthogonal Procrustes rotation all items loaded positively onto this third component, even items with a clearly contradictory content (e.g.E2 and S2).The existence of differences between subjects in response style, especially acquiescence, forms a plausible explanation for the emergence of this component.If there exist systematic differences between subjects in the tendency to endorse items in the Post-Critical Belief scale, irrespective of their specific content, it will exert a systematic bias on the correlations between the items (see also Green, Goldman & Salovey, 1993): Positive correlations will become more positive and negative correlations will be suppressed.Multidimensional scaling is only marginally affected by a response style that systematically influences all correlations since it represents the pattern of relationships between the items and not the absolute size of these relationships.
Factor analysis, on the contrary, does represent the absolute size of the correlations and, hence, will be affected by a systematic response tendency.A VARIMAX-rotation will then produce unipolar factors, which represent the positive correlations.However, if the factor structure is rotated to a structure that is unaffected by this response tendency -as is done here by a rotation to the two-dimensional multidimensional representation -an extra factor, representing interindividual differences in the systematic response tendency, should show up.
The renewed interpretation of External Critique is supported by its positive relationship to racism.turned out to be non-significant: t(502)=1.80,n.s., for the interaction between Orthodoxy and ecudational level; t(502)=-0.83,n.s., for the interaction between External Critique and ecudational level; t(502)=1.57,n.s., for the interaction between Relativism and ecudational level; and t(502)=-0.43,n.s., for the interaction between Second Naiveté and ecudational level.
8 In previous research, this dimension was also labeled Historical Awareness (Duriez, Hutsebaut & Roggen, 1999) and Symbolic Belief (Fontaine et al., 2000;Luyten et al., 1998).We prefer the term Second Naiveté, because it explicitly points to the concept of Ricoeur (1965) which we tried to operationalize. 9This interpretation receives support from the relations with frequency of Sunday mass attendance.These were .48(p<.0001) for Orthodoxy, -.48 (p<.0001) for External Critique and .36(p<.0001) for Historical Relativism.However, these relations were -.08 (p<.01) for Relativism and .57(p<.0001) for Second Naiveté.
The differential relationship between Second Naïveté and racism on the one hand and Relativism and racism on the other hand supports the distinction between both aspects of Historical Relativism.Even though the Relativism scale is far less reliable than the Second Naïveté scale, a negative relationship was found between Relativism and Racism, whereas no relationship was found between Second Naïveté and racism.
Results of the first study thus provide clear empirical support for the interpretation of the Post-Critical Belief scale in terms of Wullf's bipolar four-quadrant model (Wulff, 1991;1997).

Introduction
The renewed interpretation of External Critique and the disentanglement of Historical Relativism in Relativism and Second Naïveté rise questions about the construct representation of the 24 items of the original Post-Critical Belief scale.Relativism, for instance, was constituted of four items only and appeared to have low internal consistency.Therefore, a revised 33-item version of the Post-Critical Belief scale was developed.First, we decided to change the formulation of item O1-because some of the participants of the first study reported difficulties filling out this item.Probably this was due to the unintended complexity of this item.The renewed formulation of this item can be found in the Appendix (O9).Second, for similar reasons we decided to change some of the External Critique items (E1, E4 and E5-).
Probably, the difficulty with these items was that they did not clearly enough capture a stance of unbelief.
We decided to substitute them by a couple of new items (E9-E12, see Appendix).In this way we also aimed to broaden the range of the concept.Third, we found it necessary to further elaborate both Relativism and Second Naiveté.Concerning Relativism, we decided to include extra items (R5-R8, see Appendix), further elaborating what we thought was typical of this position.To avoid confusion, we relabeled the original Relativism items (see Appendix).Concerning Second Naiveté, we decided to leave out item H6, relabel the retained items (see Appendix) and elaborate the scope of the concept's operationalization by including some new items (S4-S8, see Appendix) so that it would more closely resemble the concept of Ricoeur (1970).Empirical evidence for the validity of this revised version, based on internal structure analyses and the relationship with racism, is presented in this second study.

Method
Sample.The questionnaires were distributed by undergraduate students of the Faculty of Psychology of the Catholic University Leuven (Belgium).In total, they distributed 300 questionnaires, of which 76% were returned.Some of them, however, were only partially completed.All respondents were Flemishspeaking Belgians; 52.2% were female.The age groups were: 16-18 (4.4%), 19-25 (52.6%), 26-35 (7.0%), 36-45 (8.3%) and 46-65 (27.7%).The highest educated group (89.5%) consisted of people who had obtained a university degree or were either university students, the second group (6.6%) consisted of people who had completed or were attending a non-university higher education.The level of Sunday mass attendance was comparable to that in our first study: 40.1% claimed to attend weekly or at least often, 33.3% claimed to attend on special occasions only and 25.9% claimed to never attend.

Results
Multidimensional Scaling on the Post-Critical Belief scale.A non-metric multidimensional scaling on the Euclidean distances between the 33 standardized items of the revised Post-Critical Belief scale was performed.The two-dimensional solution (see Figure 3) fitted the data well.It had a Kruskal stress of .11and explained .94% of the variance in the optimally transformed dissimilarities. 10In order to optimally compare the internal structure of the revised version of the Post-Critical Belief scale to the internal structure of the original version, an orthogonal Procrustes rotation (Schönemann, 1966) towards the internal structure of this original version was applied.After rotation, Tucker's Phi indices indicated a high congruence for both dimensions: .96for Inclusion vs. Exclusion of Transcendence and .92for Literal vs. Symbolic. 11  Insert Figure 3

about here
As in Study 1, the four quadrants of the two-dimensional representation could straightforwardly be interpreted (see Figure 3).The upper left quadrant, quadrant 1, encompasses the Orthodoxy items.
The lower left quadrant, quadrant 2, encompasses the External Critique items and one of the supposed Relativism items (R8).The lower right quadrant, quadrant 3, encompasses the Relativism items (except R2 and R8) and one of the supposed Second Naiveté items (S4).Finally, the upper right quadrant, quad-n.s.).

Conclusion
Based on the congruence measures between the internal structure of the 33-item revised Post-Critical Belief scale and the 24-item original version, we can conclude that the internal structures are (almost) identical.The two dimensions fit Wulff's Inclusion vs. Exclusion of Transcendence dimension and Literal vs. Symbolic dimension (Wulff, 1991;1997) The relationships with racism, where in line with the findings of the first study for three of the four Post-Critical Belief subscales.Again, a positive relationship with Orthodoxy, a negative relationship with Relativism and no relationship at all with Second Naïveté was found.However, the expected positive relationship between External Critique and racism was not observed.Nevertheless, despite the failure to observe a direct relationship between External Critique and racism, a well interpretable internal structure, which strongly relates to the structure in Study 1, was revealed in this second study.

Introduction
In this third study, we tried to replicate the results of the second study.Our main question is how stable the two bipolar religious attitude dimensions are.Furthermore, it is investigated whether the failure to find a relationship between External Critique and racism is generalizable to other samples.This could point to specific effects of the newly constructed External Critique items.

Method
Sample.Participants were 338 first year psychology students at the Katholieke Universiteit Leuven (Belgium), ranging in age from 17 to 24 years old (Mean= 18).All of them were Flemish speaking 14 Correlations with frequency of Sunday mass attendance were .49(p<.0001) for Orthodoxy, -.66 (p<.0001) for External Critique, -.21 (p<.005) for Relativism and .63(p<.0001) for Second Naiveté.
Belgians; 82% were female.Participation was obligatory and all participants received course credit.
Full confidentiality and anonymity were guaranteed.Of all participants, 9% indicated to attend Sunday mass weekly or at least often, 67.2% indicated to attend only at special occasions, and 23.3% indicated to never attend.
Procedure.Participants completed the revised 33-item Post-Critical Belief scale, as presented in Study 2. Participants also completed the racism scale.The internal consistency (Cronbach alpha) of this scale was .84 (Mean=1.96,SD=0.61).

Results
Multidimensional Scaling on the Post-Critical Belief scale.A non-metric multidimensional scaling analysis on the Euclidean distances between the standardized items of the 33-item revised Post-Critical Belief scale was carried out.A two-dimensional solution (see Figure 4) was adequate to represent the internal structure. 15It had a Kruskal stress of .13 and explained 90% of the variance in the optimally transformed dissimilarities.After an orthogonal Procrustes rotation (Schönemann, 1966) towards the dimensional structure of the original version, Tucker's Phi indices suggested high congruence for both dimensions: .97 for Inclusion vs. Exclusion of Transcendence and .93 for Literal vs. Symbolic.This configuration was also highly similar to the configuration in the second study (Tucker's Phi of .90 for Inclusion vs. Exclusion of Transcendence and of .93 for Literal vs. Symbolic).16

Insert Figure 4 about here
As in Study 1 and 2, the four quadrants of the two-dimensional representation could straightforwardly be interpreted (see Figure 4).The upper left quadrant, quadrant 1, encompasses the Orthodoxy items (except 09).The lower left quadrant, quadrant 2, encompasses the External Critique items.The lower right quadrant, quadrant 3, encompasses the Relativism items (except R7).Finally, the upper right quadrant, quadrant 4, encompasses the Second Naiveté items, one of the Orthodoxy items (O9) and one of the Relativism items (R2).Thus, in total, 31 out of the 33 items emerged in the hypothesized quadrant.The two items shifting to an adjacent quadrant are different items than the items shifting to an adjacent quadrant in the previous study.This suggests that these small shifts are resulting from sampling fluctuation.
Principal Component Analysis on the Post-Critical Belief scale.As in the two previous studies, the 33-item revised Post-Critical Belief scale was subjected to a principal component analysis.Again, Cattell's scree test (Cattell, 1966) pointed to a three-componential solution, which accounted for 45% of the total variance. 17When orthogonally rotating this three-componential solution to the twodimensional solution of the first study, the two bipolar and the unipolar component showed up again (see Table 3).Tucker's Phi indices suggested high congruence between the first two components and the two dimensions generated by multidimensional scaling in the first study: .97 for Inclusion vs. Exclusion of Transcendence and .93 for Literal vs. Symbolic.If the three-componential solution of the present study is rotated to the two-dimensional scaling solution of the second study, the congruence is even higher (.99 for Inclusion vs. Exclusion of Transcendence and .96for Literal vs. Symbolic).
Relationships with racism.Again, we expected both Orthodoxy and External Critique to be positively related to racism (Hypothesis 1 and 2), Relativism to be negatively related to racism (Hypothesis 3) and Second Naiveté to be unrelated to racism (Hypothesis 4).Results supported all hypotheses.Racism was positively related to both Orthodoxy (r=.29, p<.0001) and External Critique (r=.18, p<.0005) and negatively related to Relativism (r=-.16,p<.005).o relationship with Second Naïveté was found (r=-.10,n.s.). 18

Conclusion
17 The eigenvalues for the first six components after extraction were 8.07, 4.04, 2.24, 1.52, 1.33, 1.17. 18A 10-item version of the Marlowe-Crowne Social Desirability scale (Crowne & Marlowe, 1960) was also included in this study.Results showed that the relationships between racism and the different dimensions of the Post-Critical Belief scale largely remained unaltered when controlling for socially desirable answering: .32Again, a two-dimensional representation of the revised Post-Critical Belief items could be interpreted perfectly in terms of Wulff's (1991;1997)   19 This time the relationships with racism clearly supported this interpretation.Racism was negatively related to Relativism, but was not related at all to Second Naïveté.Furthermore, racism was positively related to both Orthodoxy and External Critique.This last finding shed some light on the results of the second study, where this relationship was not found.An explanation in terms of the newly constructed External Critique items is unlikely in the light of the relationship found in the present study.Therefore, the absence of this relationship in the second study is probably due to sample characteristics.

General Discussion
Across the three studies presented in this article, the internal structure of both the original and the revised Post-Critical Belief scale is highly stable and replicable.This conclusion is supported by the congruence measures between the solutions of the various studies.Very few items shift to adjacent quadrants, and since the shifting of items occurred randomly across the three studies, this phenomenon has to be attributed to sampling fluctuation.The better construct representation of the religious attitude domain distinguishes the original version from the revised version.
Concerning the internal structure, method convergence was observed.After orthogonal Procustes rotation, the first two components of a three-component solution turned out to be highly related to the two dimensions of the geometrical representation.The difference is the emergence of a general component with all items loading positively on it in the principal component analyses.This component did not emerge as a dimension in the multidimensional scaling.As explained earlier, interindividual differences in a response style, such as acquiescence, form a plausible explanation.
In the present research, we investigated whether responses to the items of the Post-Critical Belief scale can be interpreted in terms of Wullf's model (Wulff, 1991;1997).Responses to these items can indeed be interpreted in terms of Wulff's two dimensions (Inclusion vs. Exclusion of Transcendence and Literal vs. Symbolic).Orthodoxy consistently measured Literal Affirmation.External Critique consistently measured Literal Disaffirmation, as was suggested by Desimpelaere et al. (1999).Furthermore, it was consistently found that Historical Relativism needed to be split op into two different di-(p<.0001)for Orthodoxy, .17(p<.005) for External Critique, -.17 (p<.005) for Relativism, and -.10 (n.s) for Second Naiveté. 19Correlations with frequency of Sunday mass attendance were .26(p<.0001) for Orthodoxy, -.38 (p<.0001) for External Critique, -.03 (n.s.) for Relativism and .39(p<.0001) for Second Naiveté.mensions: Relativism, which forms an operationalization of the Reductive Interpretation, and Second Naïveté, which forms an operationalization of Restorative Interpretation.
The renewed interpretation of the Post-Critical Belief subscales has important consequences.Lack of differentiation, as was the case in the original version, leads to partly meaningless and even faulty interpretations of the relationships between religiosity and racism.Contrary to what we previously thought, we could not find any evidence for the existence of an explicit belief stance that truly counters racism.Apparently, religion cannot be said to provide a real antidote for racism in Flanders (for a more thorough discussion: see Duriez & Hutsebaut, 2000).Recent research indicates that the renewed outlook on the Post-Critical Belief scale is also relevant for the investigation of the relationship between religiosity and other concepts like value orientations (Duriez, Fontaine & Luyten, in press;Fontaine, Luyten & Corveleyn, 2000;Fontaine, Duriez, Hutsebaut & Luyten, 2000), authoritarianism and social dominance orientation (Duriez & Van Hiel, 2000), and economic and cultural conservatism (Duriez, Luyten, Snauwaert & Hutsebaut, 2000).
We can conclude that there clearly is validity evidence for a multidimensional conceptualization of religious attitudes in the Flemish secularized context.Interindividual differences in religious attitudes can be interpreted in terms of Wulff's (1991;1997)  Therefore, the Post-Critical Belief scale offers new perspectives for the empirical study of interindividual differences in religiosity.However, further research should point out whether and to which extent this scale can be generalized to other cultural groups and religious denominations.
English translation of the items of the original Post-Critical Belief scale (Version 1) and of the revised Post-Critical Belief scale (Version 2), ordered according to the original and revised subscales (Orthodoxy (O), External Critique (E), Historical Relativism (H), Second Naïveté (S) and Relativism (R)), including the labels by which these items are referred to in this article.Note that, although in our study, all items were administered in Flemish, we present them in English.To obtain these English versions the original items were first translated from Flemish into English.Second, they were independently translated back into Flemish in order to compare the original items and the doubly translated versions.Wulff's (1991;1997)

Label Version Items
Post-Critical Belief scale.Based on the empirical research, we can conclude that the Post-Critical Belief scale is a promising scale for the measurement of religious attitudes in a secularized context.The three dimensions of this scale are empirically distinguished and are differentially related to external variables.The empirical research, however, also indicated two problems that form the focus of the present research.First, the internal structure of the Post-Critical Belief scale does not fit Wullf's model very well.According to Wullf (1991; 1997), the domain of religion is characterized by two bipolar dimensions (Inclusion vs. Exclusion of Transcendence and Literal vs. Symbolic) that give rise to four approaches to religion.Instead of two bipolar dimensions, three unipolar factors are identified in the Post-Critical Belief scale.This raises doubts about the use of Wullf's model for the interpretation of the three dimensions of the Post-Critical Belief scale.Second, although the items of the External Critique dimension were constructed to measure Reductive Interpretation, external correlations suggest that this dimension actually measures Literal Disaffirmation.The internal structure of the Post-Critical Belief scale.In the present investigation, we dealt with the first problem by extensively investigating the internal structure of the Post-Critical Belief scale.
. Again, Orthodoxy measured Literal Affirmation, External Critique measured Reductive Interpretation, Relativism measured Reductive Interpretation and Second Naiveté measured Restorative Interpretation. 14But although the internal structure of the revised version is highly similar to the original version, the revised version can appeal to a better construct representation of the four separate quadrants.
theoretical framework.The present conclusion is supported by extensive analyses of the internal structure of the Post-Critical Belief scale as well as by the external relationships between the various subscales of this Post-Critical Belief scale and racism.
model.Orthodoxy measured Literal Affirmation, External Critique measured Reductive Interpretation, Relativism measured Reductive Interpretation and Second Naiveté measured Restorative Interpretation.

Table 1 .
Component Loadings in Study 1 for the Original Version of the Post-Critical Belief Scale with Three Components After Orthogonal Procrustes Rotation to the Reference Configuration a Reference Configuration = Two-dimensional representation of original version of Post-Critical Belief Scale in Study 1 a

Table 2 .
Component Loadings in Study 2 for the Revised Version of the Post-Critical Belief Scale with Three Components After Orthogonal Procrustes Rotation to the Reference Configuration a Reference Configuration = Two-dimensional representation of original version of Post-Critical Belief Scale in Study 1 a

Table 3 .
Component Loadings in Study 3 for the Revised Version of the Post-Critical Belief Scale with Three Components After Orthogonal Procrustes Rotation to the Reference Configuration a