Scientific Information System Network of Scientific Journals from Latin America, the Caribbean, Spain and Portugal Itziar Etxebarria, Pedro Apodaca Both Freud and Hoffman are Right: Anxious-Aggressive and Empathic Dimensions of Guilt The Spanish Journal of Psychology, vol. 11, núm. 1, may, 2008, pp. 159-171, Universidad Complutense de Madrid España Available in: http://www.redalyc.org/articulo.oa?id=17211115 The Spanish Journal of Psychology, ISSN (Printed Version): 1138-7416 [email protected] Universidad Complutense de Madrid España How to cite Complete issue More information about this article Journal's homepage www.redalyc.org Non-Profit Academic Project, developed under the Open Acces Initiative The Spanish Journal of Psychology Copyright 2008 by The Spanish Journal of Psychology 2008, Vol. 11, No. 1, 159-171 ISSN 1138-7416 Both Freud and Hoffman are Right: Anxious-Aggressive and Empathic Dimensions of Guilt Itziar Etxebarriaand Pedro Apodaca Universidad del País Vasco (Spain) The purpose of the study was to confirm a model which proposed two basic dimensions in the subjective experience of guilt, one anxious-aggressive and the other empathic, as well as another dimension associated but not intrinsic to it, namely, the associated negative emotions dimension. Participants were 360 adolescents, young adults and adults of both sexes. They were asked to relate one of the situations that most frequently caused them to experience feelings of guilt and to specify its intensity and that of 9 other emotions that they may have experienced, to a greater or lesser extent, at the same time on a 7- point scale. The proposed model was shown to adequately fit the data and to be better than other alternative nested models. This result supports the views of both Freud and Hoffman regarding the nature of guilt, contradictory only at a first glance. Keywords:guilt,empathy,anxiety,emotion El objetivo de este estudio fue confirmar un modelo que proponía dos dimensiones básicas de le experiencia subjetive de la culpa, una ansiosa-agresiva y la otra empática, además de otra dimensión asociada pero no intrínseca a la culpa, es decir, la dimensión de las emociones negativas asociadas. Participaron en el estudio 360 adolescentes, jóvenes y adultos de ambos sexos. Se les pidió que relataran una de las situaciones que les hacían sentirse culpables con mayor frecuencia y que especificaran su intensidad y la de otras 9 emociones que hubieran podido experimentar, en mayor o menor grado, al mismo tiempo en una escala de 7 puntos. El modelo propuesto se ajustaba a los datos y se observó que era mejor que otros modelos alternativos anidados. Este resultado apoya los puntos de vista tanto de Freud como de Hoffman con respecto a la naturaleza de la culpa, sólo aparentemente contradictorios. Palabras clave:culpa, empatía, ansiedad, emoción This study is part of a research project funded by the Universidad del País Vasco (Project code: 1/UPV00227.231-H-14897/2002). Correspondence concerning this article should be addressed to Itziar Etxebarria, Departamento de Procesos Psicológicos Básicos, Universidad del País Vasco, Aptdo. 726, 2008 San Sebastián (Spain). Phone: 34-943-015739. Fax: 34-943-015670. E-mail: [email protected] 159 160 ETXEBARRIAAND APODACA After a long period of neglect, the study of guilt has once into self-punishment. Furthermore, given its extremely again become the object of much attention over the last unpleasant nature, the individual tends to set in motion decade. From the 1990s onwards, a number of theoretical numerous defense mechanisms (rationalizations, projections, and empirical studies have focused on fundamental aspects reaction-formations, etc.). As a result, feelings of guilt end of this emotion such as the cognitive content of the up manifesting themselves in a wide variety of different experience and its actual antecedents (Frijda, 1993; Kroon, ways, to the extent that the individual him/herself often fails 1988), its interpersonal roots and functions (Baumeister, to recognize them for what they are, experiencing them as Stillwell, & Heatherton, 1994; Berndsen, van der Pligt, anxiety and floating aggressiveness (most commonly directed Doosje, & Manstead, 2004; Jones, Kugler, & Adams, 1995), at him/herself, but often turned outwards also) whose origin the difference between guilt and shame and the implications the person is unable to identify. In any case, these feelings of such differences (Abe, 2004; Buss, 1980, 2001; Lewis, of guilt constitute an underlying factor in much pathology 1971; Lewis, 2000; Smith, Webster, Parrott, & Eyre, 2002; (depression, obsessive-compulsive disorders, etc.). Tangney, 1995; Tangney & Dearing, 2002; Tracy & Robins, Klein (1973) agrees with most of Freud’s analyses, but 2004), its early development (Barret, 1998; Ferguson & also identifies a different type of guilt. In addition to Stegge, 1995; Kochanska, Gross, Lin, & Nichols, 2002; Zahn- persecutoryguilt, very similar in nature to the Freudian guilt Waxler & Kochanska, 1990; Zahn-Waxler & Robinson, 1995) described above, this author also identifies depressiveguilt. and its relation with psychopathology (Bybee & Quiles, 1998; This type of guilt does not consist so much of anxiety Harder, 1995; Tangney, Burggraf, & Wagner, 1995), etc. regarding the feared object as of sadness and regret at having Nevertheless, hardly any research has been carried out on harmed the beloved object. More than feeling accused, an an equally important question: the emotional quality of the individual experiencing this kind of guilt feels responsible subjective experience of guilt. What is the nature of the most for the harm done to the beloved object, harm that needs to common experience of guilt? What emotional components be redressed. Another author such as Fromm (1947/1985) are involved? Are there clearly differentiated types? proposes a new distinction: that between authoritarianguilt Although empirical research is scarce, theoretical and humanistic guilt. The first is confused with fear of approaches to the emotional quality of the subjective authority (parents, state, Church) which is either external experience of guilt abound. However, these theoretical or internalized within the individual’s own consciousness approaches vary widely, resulting in a somewhat confusing and implies aggressiveness, stemming from frustration, which theoretical panorama. Our aim when designing this study is constantly directed towards oneself. The second is defined was to provide empirical data to help clarify the as the feeling derived from not having acted in accordance aforementioned panorama. The following is a brief overview with one’s own dignity, of not having risen to the occasion. of the most notable theoretical approaches. Learning theories have also striven to shed light on the Firstly, we should highlight those theories postulated by nature of guilt. In relation to this approach, we can highlight Freud, who dedicated many of his analyses to unraveling the work of Eysenck and Mosher. Eysenck (1964/1977) the nature and effects of feelings of guilt. According to Freud conceived guilt as a conditioned emotional disturbance. (1923/1961, 1930/1961), at first, the feeling of guilt Guilty reactions are simply due to a Pavlovian conditioning (Schuldgefühl) is nothing more than the anxiety felt by a process: if punishment is applied before, or at the start, of child in response to the loss of parents’love when s/he fails the transgression, then resistance to the transgression is to behave as they would like. After the development of the generated; however, if punishment is applied once the superego as the result of the internalization of parental transgression has been committed, then the result is feelings values, the feeling of guilt has a twofold origin: anxiety of guilt. Mosher (1965) also associates guilt with anxiety, over external authority and anxiety caused by the severe defining it as a “generalized expectancy for self-mediated vigilance of the superego. When the individual acts or feels punishment (i.e., negative reinforcement [sic]) for the the impulse to act in a way which contravenes the mandates violating, anticipating the violation of, or failure to attain of the superego, the superego berates that person with harsh internalized standards of proper behavior” (p. 162). criticism; in order to avoid these recriminations, the Alongside the theories mentioned above, special mention individual ends up capitulating to the dictates of the should be made of that postulated by Hoffman (1982, 1998), superego. According to Freud, the aggressiveness that the one of the few authors that continued to work in this field superego unleashes onto the ego, perceived by the ego as during the 1980s and whose work still strongly influences a feeling of guilt, originates not only from the real severity current thinking. From Hoffman’s perspective, guilt (or more of the individual’s parents; it also stems from the individual’s specifically, the guilt that he terms true or interpersonal own aggressiveness, from both his/her original aggressiveness guilt, in order to distinguish it from Freudian guilt) is and that provoked by the frustration of his/her desires, which intrinsically linked with empathy. When an individual feels the subject has been forced to repress. In addition to acting empathic pain and sees him/herself as being responsible for as a powerful inhibitory factor, feelings of guilt also produce another person’s pain, then the empathic experience tends the need for punishment, a need which tends to be translated to transform itself into guilt. This constitutes a fundamental ANXIOUS-AGGRESSIVE AND EMPATHIC DIMENSIONS OF GUILT 161 prosocial motive, since it fosters reparatory behavior and considers it to be a fundamental element in depressive guilt all types of positive behaviors that transcend to the victim. –empathic guilt. Finally, among the more recent approaches, we should To what extent are these theoretical approaches valid? highlight those postulated by Baumeister et al. (1994). Based How faithful a reflection are they of real subjective on a comprehensive review of empirical research, as well experiences of guilt? Despite the number of studies on the as on their own studies, these authors postulate an emotional correlates of the tendency to experience guilt or interpersonal conception of guilt. From this perspective, trait guilt, studies focusing on the emotional correlates of guilt is rooted, on the one hand, in the human capacity to the experience of guilt in response to a specific event, or feel others’pain, and on the other, in anxiety over rejection state guilt, are still scarce. Nevertheless, as we will see by others. Thus, guilt appears mainly in relation to actions below, existing studies support the fundamental presence of or omissions which cause real or possible pain to others, empathy, anxiety, aggressiveness and sadness in experiences particularly those closest to us, and plays a decisive role in of guilt. the control and reparation of the said actions. Frijda (1993), With regard to the empathic component, after analyzing in his analysis of the cognitive content and actual antecedents child and adult autobiographical reports of emotional of guilt experiences, proposes a similar view. experiences, Tangney, Marschall, Rosenberg, Barlow, and The wide range of different conceptions of guilt, briefly Wagner (1996) found that when relating episodes in which outlined above, in conjunction with the need identified by they had experienced feelings of guilt, both groups often several authors to establish distinctions between different expressed other-oriented empathy. An experimental study types, as well as the fact that the distinctions do not always carried out by Hoffman (Thompson & Hoffman, 1980) also coincide, suggest that we are faced with a complex family supports the presence of an empathic component in the of emotions. Nevertheless, if we compare the various feeling of guilt. Similarly, significant positive correlations approaches, we see that different authors coincide in have been found between the tendency to experience other- highlighting certain types of guilt. In this way, in spite of oriented empathy and the tendency to experience guilt in terminological differences, Freudian approaches, Klein’s sample groups of various ages (Tangney, 1991; Tangney, ideas regarding persecutory guilt, Fromm’s perspective Wagner, Burggraf, Gramzow, & Fletcher, 1991). Finally, regarding authoritarian guilt, and even Eysenck and other studies which show a close relationship between the tendency learning theorists’proposals regarding conditioned anxiety to experience feelings of guilt and the use by parents of responses resulting from educational practices are, in fact, inductive disciplinary techniques –techniques which activate fairly similar. In all cases we are dealing with a feeling of empathy and foster its development– also provide indirect guilt that is confused with anxiety over the reaction of more empirical support for the association between guilt and or less internalized others, a feeling of guilt that, according empathy (Etxebarria, 1994; Hoffman, 1994, 2000). to some authors (particularly those from the psychoanalytic As for the other emotional components whose field), is also marked by contained aggressiveness turned implication in the experience of guilt has been highlighted inwards against the ego. Alongside this type of guilt, Klein by diverse theories, a study carried out by Baumeister, Reis, stresses the existence of depressive guilt, a concept which and Delespaul (1995), using an experience sampling method prefigures the true or interpersonal guilt later proposed by over the course of a week in order to determine the Hoffman. In this case, guilt is associated with empathy over frequency with which people experience guilt and the another’s pain. Baumeister and colleagues propose a kind emotions that are associated with this experience, supports of synthesis of these two aforementioned conceptions, the presence of anxiety, aggressiveness and sadness in guilt emphasizing the participation of both empathy and anxiety episodes. Similarly, studies which analyze various correlates over external rejection in the experience of guilt. of the Guilt Inventory have found that both stateguilt and In sum, guilt has basically been associated with two traitguilt are associated with anxiety, hostility and depression emotions: empathy and anxiety. Everything seems to point (Jones & Kugler, 1993; Jones et al., 1995). to the existence of some experiences of guilt in which the In a number of studies with adult subjects, Watson and predominant emotion is empathy, and others in which the Clark (1992) also found strong correlations between guilt prevailing feeling is that of anxiety, and that, to a large and hostility, sadness, and fear. Nevertheless, in these studies, extent, the discrepancy between different theories is due to most data was based on relatively long-term measures of the fact that the researchers are focusing on different types the affect, rendering the results more relevant in relation to of experiences of guilt. Furthermore, some authors also trait guilt than in connection with state guilt. In any case, coincide in that anxious guilt often involves a fair amount in the first of these studies, the authors used measurements of contained aggressiveness, directed inwards against oneself. of current, momentary affect in two variables, anxiety (STAI Alongside the above mentioned emotional components, a A-State) and anger/hostility (STAS), finding a very high number of authors have also highlighted the role played by correlation between the two. This result deserves special a depressive component in all experiences of guilt: Freud attention in relation to the perspectives that attribute to mentions it in his analysis of anxious guilt, and Klein anxious guilt an important aggressive component. 162 ETXEBARRIAAND APODACA However, the relationship between guilt and aggressive factor); and another that would basically aggressiveness is far from clear. Different studies support correspond to the guilt described by Hoffman, involving the a negative relationship between trait guilt and a tendency feeling of sympathy (empathicfactor). Asadness component toward outwardly directed anger and hostility (Lutwak, would intervene in both factors, although with less weight Panish, Ferrari, & Razzino, 2001; Mosher, 1979; Tangney, than the aforementioned emotional components. Wagner, Fletcher, & Gramzow, 2001). According to Tangney In addition to attempting to confirm the existence of (1995), this is due to the fact that, although the tendency these two factors, we also aimed to determine the extent to towards guilt, like the tendency towards shame, is associated which components of other emotions such as disgust, fear, with the activation of anger, people who tend to experience shame and sense of worthlessness, were involved, and shame-free guilt (as opposed to those who tend to experience present, in experiences of guilt. The above review suggests shame) also tend to handle it in a constructive manner: they that, although in some cases they may be associated with show constructive intentions, attempts to discuss the issue experiences of guilt and can even be fairly intense, these in a non-aggressive way with the person at whom their anger emotional components do not constitute essential elements. is directed, cognitive reevaluations of the other person’s role Therefore, we supposed that these emotions would tend to in the situation of anger, etc. Nevertheless, the negative be grouped in a factor different from the two previously- relationship between trait guilt and a tendency towards anger described ones: namely the associated negative emotions and hostility could also be interpreted as a reflection of the factor. Although clearly distinguishable from the anxious- tendency of guilt-prone people to direct their aggressiveness aggressive factor and the empathic factor, given the highly against themselves, something which, in the long term, may aversive nature of its components, this third factor would not prove particularly adaptive. show a fairly close association with the anxious-aggressive Although shame-free guilt may be associated with factor, while showing no, or a very low, covariation with constructive handling of anger, the presence of an the empathic factor. intropunitive component is tangible in many experiences of In this type of research, both cultural and social elements guilt (Kroon, 1988). This is particularly true when the play an important role, thus necessitating some prior individual is unable to resolve the situation that has given clarifications regarding the study presented here. Said study rise to the experience of guilt through confession, reparation aimed to confirm the three-factor model of guilt described or asking for forgiveness, and the experience of guilt remains above. The objective was to see whether indeed this model therefore in a state of unresolved tension (Lindsay-Hartz, properly encompassed the underlying dimensions of guilt de Rivera, & Mascolo, 1995). The presence of an (outwardly as experienced in everyday life. Thus, the best idea seemed and inwardly directed) aggressive component has also been to be to ask a large number of people to report their habitual identified in cases of so-called chronic guilt. Based on the experiences of guilt without giving them any prior definition distinction between predispositional guilt (predisposition to of said emotion. In this way, the aim was not to limit what feel guilty in response to specific events) and chronicguilt people habitually understood by guiltwith any academic a (an ongoing condition not associated with immediate events), priori. Nevertheless, a problem may arise here: how can we Bybee and Quiles (1998) found that individuals suffering be sure that people would report experiences of guilt rather from more chronic guilt also showed higher levels of than those of other closely-related emotions, such as shame, hostility and more intropunitive behavior. This data indicates for example? In the English-speaking context, it has been the presence of an other-oriented and self-aggressive stated on more than one occasion that lay-people often component in some types of guilt. Nevertheless, more studies confuse guilt and shame (Tangney, 1995; Tangney & focusing on state guilt, such as those carried out by Dearing, 2002). This may indeed be a problem in a study Baumeister et al. (1995), are required in order to enable involving English-speaking participants; however, in the conclusions to be drawn regarding the presence of aggressive Spanish context, people do not tend to have so much components in habitual experiences of guilt. difficulty distinguishing between culpaand vergüenza(the Based on the theoretical review and the empirical studies Spanish terms for guilt and shame, respectively). Arecent mentioned above, in this current study we formulated the study by Pascual, Etxebarria and Pérez (2007) suggests that hypothesis that, fundamentally, the components of anxiety, the limits between the Spanish terms culpa and vergüenza aggressiveness, empathy and sadness would be involved in are less diffuse than the limits between the English terms habitual experiences of guilt, although the anxious and guilt and shame. Similarly, other authors (Hurtado de aggressive components would tend to be more intense in Mendoza, 2007; Hurtado de Mendoza & Parrott, 2002) have some cases, and the empathic component more intense in found that the distance between vergüenza–to be even more others. In other words, we hypothesized that two basic precise, the Spanish vergüenza, as opposed to the clearly differentiated factors would be involved in Guatemalan vergüenza– and culpais greater than the distance experiences of guilt: one that would correspond, to a large between shameand guilt. extent, with the guilt described by Freud, mainly covering However, we should not overlook another problem here, the feelings of anxiety, anger and anger at oneself (anxious- perhaps even more important than the one outlined above. ANXIOUS-AGGRESSIVE AND EMPATHIC DIMENSIONS OF GUILT 163 When we ask people about their experience of culpa, what Instrument and Procedure we are verifying is whether the dimensions proposed by our model underlie the experiences that Spanish people consider Participants responded to a questionnaire in which they to be experiences of culpa, not whether said dimensions were asked to describe, in a certain amount of detail, one underlie experiences of guiltper se. Furthermore, we cannot of the situations which most frequently caused them to suppose an exact correspondence between the Spanish term experience feelings of guilt (culpa). After reporting why this culpa and the English term guilt (Hurtado de Mendoza, made them feel guilty, they had to specify the intensity of 2007; Hurtado de Mendoza & Parrott, 2002; Pascual et al., the feeling of guilt on a 7-point scale (1 = no guilt, 7 = very 2007). Does this mean that we are studying a culturally- intense guilt). Then, there was a list of 9 emotional specific emotional experience? Avariant of guilt that, as experiences where they had to indicate whether they such, would not be worth using to confirm a general model experienced one or more of these feelings, in addition to of this emotion? Not at all; the experience of culpa is not guilt, in the situation they had mentioned. They specified a culturally-specific emotional experience or, to be more the intensity of these feelings on a scale similar to the precise, it is no more so than the experience of guilt of previous one at the side of each of the listed emotions. The English-speaking participants in studies of this kind. list was as follows: “I feel sorry for another person”, “I feel Although there is a certain degree of (often unconscious) nervous, anxious”, “I feel angry with myself”, “I feel ethnocentrism that prevents us from seeing it, many other irritated, angry”, “I feel worthless”, “I feel sad”, “I feel studies carried out in other cultures, including English- disgusted”, “I feel frightened”, “I feel ashamed” and, finally, speaking ones, pose the same problem. As some authors “any other emotion.” have pointed out (Russell, 1991; Russell, Fernández-Dols, The students answered the questionnaire in their Manstead, & Wellenkamp, 1995), when we study guilt, classrooms during class. The young adults and adults Schuldgefühl or culpa, what we are studying, at least with responded to the questionnaire individually at home. this type of method, is what the people from the corresponding culture understand by the term, rather than Data Analysis a universal emotion whose psychological essence is, by coincidence, embodied in the English term guilt, the German To test the hypotheses relating to the structure of the term Schuldgefühl or the Spanish term culpa. Having said subjective experience of guilt (culpa), we used a this, we believe the use of self reports, even with all their confirmatory factor analytic (CFA) procedure, which is based limitations, continues to be a useful tool for analyzing on the analysis of covariance structures, using the AMOS subjective questions such as the one that is the focus of this program (Arbuckle & Wothke, 1999). paper. However, we should not forget that “[...] the data do not Having made these clarifications, and despite the fact confirm a model, they only fail to disconfirm it, [...] when that the Spanish term does not correspond exactly with the the data do not disconfirm a model, there are many other English term guilt or the German term Schuldgefühl, our models that are not disconfirmed either” (Cliff, 1983, p. hypothesis, as stated earlier, is that the factors postulated 116-117). Furthermore, the Chi-square goodness-of-fit test by the theory would be confirmed by the data obtained in is the main procedure of parameter significance for the the Spanish context. analysis of covariance structures. This test, although valuable, is problematic because it is a direct function of sample size, it decreases as the model becomes less parsimonious and it Method increases as the covariances of the matrix become higher (Apodaca & Páez, 1992). Currently, it is more useful when Participants regarded as a measure of fit rather than as a statistic test (Joreskog & Sorbom, 1993). Participants were 360 people from three age groups: one Based on this, we decided to test different nested models. group of 156 adolescents aged between 15 and 19 (81 girls Our aim was to demonstrate that our theoretical model fitted and 75 boys, mean age = 16.87 years), another group of 96 better than other models. Therefore, we tried to test our young adults aged between 25 and 33 (49 women and 47 model in relative terms (as better than other alternative men, mean age = 28 years), and a third group of 108 adults models) rather than in absolute terms (through a test of aged between 40 and 50 (54 women and 54 men, mean age statistical significance). = 44.69 years). The predominating religious tendency was Models. The main model regarding the structure of the Catholic or generally Christian, although the vast majority subjective experience of guilt (theoretical model - M) t were non churchgoers. They voluntarily participated in the hypothesized a priori that (a) it would be described by three study and were treated in accordance with the Ethical factors, two basic factors (the “Empathic” and “Anxious- Principles of Psychologists and Code of Conduct(American aggressive” factors) and another factor of “Associated Psychological Association, 1992). negative emotions”; (b) the first factor –the Empathic one– 164 ETXEBARRIAAND APODACA would show very low correlations with the other two factors Model identification and latent factor scales definition. while the second and the third factors would show a high The Empathic factor in models of two and three factors correlation; (c) the “sadness” emotion would have two (M , M, M , M ) with only two indicators and the 2 t 31 32 nonzero loadings (one in the Empathic and the other in the sadness item also loading on the Anxious-aggressive factor Anxious-aggressive factors) while the rest of the emotions could cause the under-identification of the model. To solve would have only one nonzero loading; (d) specifically, the this problem the loading of sympathy is set to 1 (as the Empathic factor would include sympathy and sadness, the main component of the Empathic factor) and the loading Anxious-aggressive factor would include sadness, anxiety, of sadness is set to 0.5 (as a secondary component of the anger and anger at oneself, and the Associated negative Anxious-aggressive factor). These decisions are congruent emotions factor would include fear, disgust, shame and a with the substantive criteria mentioned above. For the sense of worthlessness. definition of latent factor scales the loadings of one item This theoretical model was compared with other nested per factor are set to 1. models to determine whether or not its fit was better. These The final sample size was 271 in order to ensure that models were as follows: (a) M - Independence, (b) M - we were working with the complete data (as recommended 0 1 One-factor model, (c) M - Two-factor model, (d) M - in structural equation methods) and to prevent a nonpositive 2 31 Three-factor model-Alternative 1, and (e) M - Three-factor definite covariance matrix. 32 model-Alternative 2. The theoretical model shows a number of Goodness-of-fit criteria. The evaluation of the model characteristics that limit the power of the analyses. The was based on multiple criteria, mainly the substantive difficulties basically arise from the limited number of meaningfulness of the model, but also Chi-squared likelihood indicators of each dimension, which is a reflection of the ratio statistics and other practical indexes of fit (absolute, theoretical views on the nature of the subjective comparative and cross-validation fit indexes). A more experience of guilt. However, the biases found in the detailed description of the indexes now follows: simulation studies carried out by Marsh, Kit-Tai, Balla, 1. The Chi-square test of the overall differences and Grayson (1998) indicate an acceptable precision level between nested models. This difference is calculated with Ns and number of indicators per factor similar to by simple subtraction of the chi-square statistics of those considered in our study. Furthermore, the strategy the nested models. This difference is distributed as of testing nested models lends additional consistency to chi-square with degrees of freedom equal to the the estimates. Nevertheless, it will be necessary to analyze difference in degrees of freedom between the the confidence interval for the RMSEAand the ECVI, in models. order to assess the precision of the estimates under these 2. The Goodness-of-fit index (GFI) is an absolute index circumstances. of fit because it compares the hypothesized model Finally, one last statistical consideration should be with no model at all. mentioned, regarding the normality of the distributions. 3. The Root mean square error of approximation We analyzed the asymmetry and kurtosis coefficients in (RMSEA) is considered an important informative each of the variables and, as shown in Table 2, the criterion. This index is expressed per degrees of distributions were reasonably close to normal (Muthén & freedom and is therefore sensitive to the complexity Kaplan, 1985) and are within the limits recommended by of the model and useful for comparing alternative various authors for these analyses (Cuttance & Ecob, (non-nested) models. Steiger (1989) suggests the use 1987). of confidence intervals to assess the precision of the RMSEAestimates. 4. The comparative fit index (CFI), proposed by Bentler Results (1990), is useful for comparing nested models. It compares the hypothesized model with a base model Intensity of the Different Emotional Components that is usually the independence model. This index is Involved in the Subjective Experience of Guilt sensitive to the complexity of the model. Both Cheung and Rensvold (2002) and Elosua (2005) suggest that The analysis of the means and standard deviations of when the difference between two nested models in the different emotional components clearly showed that, as this coefficient is higher than .01, the most restricted we had hypothesized, the components that intervened most model should be rejected. intensely in the experience of guilt were sympathy, sadness, 5. Fit indexes for comparing non-nested models. anxiety, anger and anger at oneself. The other emotions Akaike’s information criteria (AIC) and the expected (fear, disgust, shame and sense of worthlessness) were found cross-validation index (ECVI) are useful for ordering with a much lower intensity. See Table 1. models from best fitting to worst fitting (Maruyama, These results were consistent with our model, in which 1998). these last components were considered to be collateral. ANXIOUS-AGGRESSIVE AND EMPATHIC DIMENSIONS OF GUILT 165 Table 1 Mean,Standard Deviation,Asymmetry,and Kurtosis for each of the Variables Age Group N M SD Asymmetry Kurtosis Sympathy 299 4.18 1.89 –.27 –1.03 Sadness 314 4.37 1.83 –.30 –.92 Anxiety 306 4.04 1.88 –.05 –1.14 Anger 304 4.01 1.83 –.03 –1.09 Anger at oneself 329 4.78 1.65 –.45 –.67 Fear 279 2.57 1.90 .96 –.35 Disgust 277 2.40 1.77 1.13 .19 Shame 289 3.12 1.90 .48 –.95 Sense of worthlessness 286 2.88 1.91 .71 –.71 Table 2 Intercorrelations for all the Variables Sympathy Sadness Anxiety Anger Anger Fear Disgust Shame Sense of at oneself worthlessness Guilt .167** .338** .345** .312** .405** .119 .200** .145* .304** Sympathy — .306** .118 .067 .047 .062 .139* .112 .031 Sadness — .218** .349** .371** .340** .349** .295** .404** Anxiety — .412** .387** .381** .257** .186** .321** Anger — .540** .270** .484** .277** .397** Anger at oneself — .232** .247** .258** .395** Fear — .335** .310** .385** Disgust — .328** .459** Shame — .331** Sense of worthlessness — * p< .05. **p < .01. Testing the Three-Factor-Model The correlation matrix to be analyzed is presented in Table 2. We tested the validity of the theoretical three-factor model in two ways. Firstly, by evaluating its degree of fit to the data by means of the different fit indexes and the consistency of the estimated parameters with what had been hypothesized. And secondly, by comparing that model with other nested models. Figure 1 shows the theoretical model (M) with the t standardized estimates of the factor loadings and the intercorrelations between the factors. The fit indexes are shown in Table 3. The diagram shows factor loadings that are consistent with the proposed hypothesis. As we can see, the Empathic dimension mainly includes the feeling of sympathy, with the feeling of sadness also being present but with less weight. The Anxious-aggressive dimension mainly includes anxiety, anger and anger at oneself, with sadness also present but again, with less weight. There is a very low covariation between the two dimensions, a finding which is consistent with the theoretical Figure 1. M - Three-factor model (Theoretical model); standardized t proposals that identify two clearly differentiated and independent estimates. 166 ETXEBARRIAAND APODACA Table 3 Fit Indexes for Nested Models Model χ2(df) GFI RMSEA(lo-hi) CFI AIC ECVI (lo-hi) M 586.4(36) .558 .238 (.22-.25) .000 604.4 2.24 (1.96-2.54) 0 M 103.5(27) .927 .102 (.08-.12) .861 139.5 .52 (.41-.65) 1 M 83.1(26) .939 .090 (.07-.11) .896 121.1 .45 (.36-.56) 2 M 64.9(24) .953 .079 (.06-.10) .926 106.9 .40 (.32-.50) t M 30.9(21) .976 .042 (.00-.07) .982 78.9 .29 (.26-.36) 31 M 118.8(25) .920 .118 (.10-.14) .830 158.8 .59 (.48-.73) 32 Note. M0 - Independence model; M = One-factor model; M = Two-factor model; M = Three-factor model = Theoretical model; 1 2 t M = Three-factor model-Alternative 1; M = Three-factor model-Alternative 2; GFI = goodness of fit index; RMSEA= root mean 31 32 square error of approximation and 90% confidence interval; CFI = comparative fit index; AIC = Akaike’s information index; ECVI = expected cross-validation index and 90% confidence interval. factors. As regards the third factor, the Associated negative As we can see, the three-factor model or theoretical emotions factor, the coefficients in Figure 1 show that it has model (M) is clearly superior in its fit than the one-factor t good internal consistency, as well as correlations with the other and two-factor models. The one-factor model (M ) presents 1 factors in accordance with what had been hypothesized. really poor fit indexes: both the absolute index (GFI) and As shown in Table 3, the set of fit statistics corresponding the comparative one (CFI) are below the limits considered to the M model is satisfactory and supports the proposed acceptable. As regards the two-factor model (M ), the fit t 2 model. Both the GFI and the CFI are satisfactory although indexes are somewhat better but they are also below what moderate. The RMSEAindex is within the limits of what is considered acceptable. Figure 2 shows the estimates can be considered acceptable. The confidence intervals to corresponding to the two-factor model. assess the precision of the RMSEAand ECVI estimates are This two-factor model is especially interesting because presented in order to show the stability of these estimates. it groups in a single factor two of the factors of the As we can see, the confidence interval for the RMSEA theoretical model, the Anxious-aggressive factor and the oscillates between .06 and .10. The highest limit is above what is considered good fit or reasonable errors of approximation, but its range is sufficiently narrow to rely on the precision of that fit index. The interval of the ECVI is also narrow enough to rely on its precision. In conclusion, we can say that the fit between the theoretical three-factor model and data is moderately good, although it is clear that modifying some parameters of the model would improve that fit considerably. This, in fact, is something we will do at a later stage. As stated above, the second level of analysis to test the validity of the model is its comparison with other nested models. Table 3 shows the fit indexes of these models. Table 4 shows the test of significance for the differences between models. Table 4 Hierarchical ∆χ2Test between Nested Models Model ∆χ2 ∆df p M -M 20.4 1 < .01 1 2 M -M 18.2 2 < .01 2 t M -M 53.9 1 < .01 32 t Note. M = One-factor model; M = Two-factor model; M = 1 2 t Three-factor model = Theoretical model; M = Three-factor 32 model-Alternative 2. Figure 2. M2- Two-factor model; standardized estimates. ANXIOUS-AGGRESSIVE AND EMPATHIC DIMENSIONS OF GUILT 167 Associated negative emotions factor. As mentioned above, the emotions included in the Associated negative emotions factor can only be considered collateral or not part of the most intrinsic elements of guilt. Nevertheless, from an empirical perspective, they present correlation patterns that could call their merely collateral character into question. Indeed, as shown in Figure 1, the correlation of the Associated negative emotions factor with the Anxious- aggressive factor is .81. We tested a two-factor model with the aim of helping to clarify these empirical correlation patterns and understand what exactly these associated negative emotions represent in relation to the experience of guilt from a conceptual perspective. The comparison between the two models (M - M ) t 2 supports the need to distinguish between the Associated negative emotions factor and the Anxious-aggressive factor since the three-factor model is clearly superior to the two- factor model in all fit indexes, as well as in the test of significance of the differences (see Table 4). Therefore, the results supported the idea that the Associated negative emotions factor is distinguishable from Figure 3. M - Three-factor model-Alternative 1 (free error the intrinsic or substantive factors of the subjective experience 31 covariances). of guilt and has an adequate level of internal consistency. This factor is closely related to the Anxious-aggressive factor, while only moderately to the Empathic factor. Nevertheless, of guilt. Therefore, we can conclude that the data clearly further analyses are required regarding the internal structure support the central corpus of the theory on the subjective of this factor, since correlations were observed between the experience of guilt. errors of some of its emotions and those of the emotions However, in our strategy of comparing nested models, characteristic of the anxious-aggressive dimension of guilt. it seemed necessary to contemplate some other alternative These inter-correlations reflect the existence of covariation model that altered the configuration of the first two factors. structures that cannot be satisfactorily explained by the three That model was the ‘M ’model, which affirmed that the 32 dimensions of the model. They seem to point to a difference feeling of sadness was only an indicator of the Empathic in the nature of emotions such as fear and disgust, and shame factor. This contradicts one of our main hypotheses, which and sense of worthlessness. posed that sadness was an emotion common to the Empathic The M model explores precisely this possibility by and Anxious-aggressive dimensions. As can be observed in 31 freeing the intercorrelations between the errors of some Table 3, the alteration of this parameter had a strong negative emotions. Figure 3 shows the estimates obtained. The impact on the model’s fit. Table 4 shows that there are diagram shows interesting correlations between the emotions significant differences between the fit of the proposed of the Anxious-aggressive factor and two of the emotions theoretical model and the alternative model (M ). Thus, 32 of the Associated negative emotions factor. The fit indexes we can also conclude that sadness is an important element corresponding to this model, shown in Table 3, are highly in both the Anxious-aggressive and the Empathic dimensions satisfactory but of only relative value, because they have of the subjective experience of guilt. been achieved following empirical post hoccriteria obtained with the same sample. Further research should try to clarify these relations with a wider range of emotions, alternative Discussion factorial structures of a hierarchical nature and, especially, different samples. The theoretical model proposing the existence of two Nevertheless, from the fit indexes obtained by the M fundamental factors in the subjective experience of guilt, 31 model we can indeed draw an important conclusion: the one Empathic and the other Anxious-aggressive, as well as proposed model has a very high fit level in the first two a third factor, the Associated negative emotions factor, was factors, which are precisely those that, from a theoretical shown to adequately fit the data and to be better than other viewpoint, were considered intrinsic to or constituent of the alternative models. subjective experiences of guilt. The source of the distortion The results obtained confirm the presence of two of the fit lies in the emotions of the third factor, which we fundamental factors in experiences of guilt (‘culpa’): an consider related but extrinsic to the subjective experience empathic factor and an anxious-aggressive factor, both of