DOI 10.2466/08.02.07.PR0.111.4.233-252 ISSN 0033-2941
DEVELOPMENT AND PSYCHOMETRIC PROPERTIES OF THE SCALE FOR SELF-CONSCIOUSNESS ASSESSMENT1
KOSTAS MYLONAS, PANAYIOTIS VELIGEKAS, AIKATERINI GARI, AND DIONYSIA KONTAXOPOULOU
University of Athens, Greece
Summary.—This scale development employed Duval and Wicklund’s (1972), Carver’s (1979), and Zaborowski’s (1987) theories on self-consciousness. The aim of the study was to create a new method to assess the self-consciousness construct, in an effort to operationally express self-consciousness, while circumventing existing metric and other impediments. Initially, 38 pilot interviews were conducted with undergraduate psychology students, and two studies followed, one on 494 partici-pants and one on 248 participants. Exploratory factor analysis models, equivalence testing, followed by a third confirmatory factor analysis study on a separate sample of 216 participants, resulted in a final 24-item scale. A four-factor structure of two public and two private self-consciousness dimensions emerged. The Scale for Self-Consciousness Assessment (SSCA) can be of use in various areas of psychologi-cal research, possibly in concurrent use with other constructs of interest, due to its theoretical and research importance and its adequate psychometric properties.
The concept of self-consciousness entails coding, processing, and in-tegrating information about the self (Wicklund, 1975; Cramer, 2000). Ac-cording to Ito (1998), consciousness comprises three different levels: wake-fulness, awareness, and self-consciousness. Consciousness in humans is directed to the self so that an individual is “aware of what is going on in his or her internal world” (Ito, 1998, p. 191). The process of self-conscious-ness is made up of both content and form. Content refers to the information that is being attended to, while one is preoccupied with self. Form refers to the manner in which such information is processed (Zaborowski, 1987; Cramer, 2000).
Self-consciousness as a term has been used by James (1892), Cool-ey (1907), Mead (1914), Vygotsky (1925/1999), and more recently by Du-val and Wicklund (1972), Wicklund (1975), Fenigstein (1997), Fenigstein, Scheier, & Buss (1975), Zaborowski (1987), and other researchers. Gener-al theories on the functioning of self-consciousness have been supported empirically, namely, Duval and Wicklund’s theory (1972), Hull and Levy’s theory (1979), Carver’s theory (1979; Scheier & Carver, 1985), and Za-borowski’s theory (1987). Wicklund’s theory places emphasis on the con-tent of self-consciousness, Hull’s theory gives priority to form, while Za-borowski’s theory assumes a dialectical interaction of form and content. Wicklund, Carver, and others have argued that self-consciousness entails 1Address correspondence to Kostas Mylonas, Department of Psychology, University of Ath-ens, 157 84 Ilisia, Athens, Greece or e-mail ([email protected]).
K. Mylonas, et al.234
processes involved in matching a person’s momentary condition against one’s own standards. Hull argued that self-consciousness entails process-es of coding information pertaining to the self. Zaborowski’s theory seems to account for a broader range of phenomena than Wicklund’s and Hull’s theories. Wicklund has coined the notion of salient self-components upon which a person’s attention is focused, while Zaborowski employs the con-cept of content centers of self-consciousness. Zaborowski’s theory admits the operation of different standards in external self-consciousness and places the emphasis on the standard of inner justice which functions in both external and internal self-awareness. So, according to different theo-ries, emphasis has been placed on different—independent or interactive—facets of self-consciousness.
Zaborowski (1980, 1987) distinguished between internal and external self-consciousness; the internal facet relates to egocentrism, individual-ism, negative emotional responses toward the self (e.g., sense of guilt), low self-esteem, etc. Defensive self-awareness and ill disposition stem-ming from it may be an even more negative expression of this internal facet. In contrast, the external facet is an objective, socialized processing of the self. An intermediate form between the two facets is reflective self-consciousness (appraisal and evaluation of self and others, assessment of duties and rights and recollection of needs and emotions). Carver and Scheier (1998) have demonstrated that individuals who mostly attend to their own inner thoughts and feelings are high in private self-conscious-ness, while those who mostly view themselves as social objects tend to see themselves according to others’ view of them and are sensitive to oth-ers’ reactions to their behavior. Finally, the terms “self-consciousness” and “self-awareness” have been used interchangeably by Zaborowski (1987) and later on by other researchers (Silvia & Gendolla, 2001; Wickens & Sta-pel, 2008, 2010). For Duval and Wicklund (1972) though, self-awareness is the state of self-focused attention, whereas the trait is called self-con-sciousness, with objective self-awareness being the ability to become the object of one’s own attention (Heinemann, 1979).
Fenigstein, et al. (1975) devised a 23-item scale (Self-Consciousness Scale, SCS) to measure individual differences in self-consciousness. They supported the distinction between self-consciousness and self-awareness. Self-awareness refers to a state of self-directed attention, while self-con-sciousness refers to dispositional self-directed attention. Factor analysis of the Fenigstein, et al. scale (1975) revealed that self-consciousness consisted of three factors: public self-consciousness, private self-consciousness, and social anxiety. Fenigstein, et al. argued that “the private self-consciousness factor was concerned with attending to one’s inner thoughts and feelings. The public self-consciousness factor was defined by general awareness of
Scale for Self-conSciouSneSS aSSeSSment 235
the self as social object that has an effect on others. The social anxiety fac-tor was defined by a discomfort in the presence of others” (p. 523).
Although the SCS has been widely used by many researchers, sup-porting the three-factor structure (e.g., Buss, 1980; Bernstein, Teng, & Garb-in, 1986), and has demonstrated construct validity in a variety of contexts (e.g., Carver & Glass, 1976), some investigators have supported a four-factor structure and some others a five- or a six-factor structure (Ander-son, Bohon, & Berrigan, 1996; Chan 1996; Cramer, 2000). Burnkrant and Page (1984) applied confirmatory factor analysis models to the original SCS items, and concluded that a four-factor structure better fit the data. Five items, regarding all dimensions, were omitted. They revealed that the factor of private self-consciousness could be divided into two sepa-rate factors: the “self-reflectiveness” and “internal state awareness.” These subscales have been used in many studies (e.g., Piliavin & Charng, 1988; Conway & Giannopoulos, 1993; Watson & Biderman, 1993; Reeves, Wat-son, Ramsey, & Morris, 1995). Other researchers, such as Mittal and Bal-asubramanian (1987), have suggested that the public self-consciousness factor could also be divided into two separate factors: “style conscious-ness” and “appearance consciousness.” Cramer (2000), through confirma-tory factor analysis, supported the four-factor structure as the most stable, comprehensive, and replicable factor structure.
The Self-Consciousness Scale (SCS) has been translated and used in many countries (Australia, France, French-speaking Canada, Germany, Hong Kong, The Netherlands, Italy, Portugal, Spain, Sweden, and Tur-key). It has also been employed in Greek settings by the first three authors of the present study (Veligekas & Mylonas, 2001; Veligekas, Mylonas, & Gari, 2001; Veligekas, Mylonas, Gari, Ploubidis, & Mantzavinou, 2003) with samples of university students and track and field athletes. Meth-ods of back-translation were implemented for these studies with three psychologists translating this scale from English to Greek and two Eng-lish literature teachers back-translating the scale into English (van de Vij-ver & Leung, 1997). By employing exploratory and confirmatory analysis models, it was shown that at least six of the original items were psycho-metrically weak for the specific cultural setting and possibly culturally bi-ased (van de Vijver & Leung, 1997; Poortinga, 1989) and could not be reli-ably implemented. It was also shown that a two-factor model, with social anxiety removed from the latent variables, provided a better fit for both samples. Thus, several operational definition problems might be active for some of the items and/or for the dimensions, at least in respect to the Greek cultural reality.
Following these initial psychometric analyses, it seemed appropriate to attempt assessing self-consciousness through a novel Scale for Self-Con-
K. Mylonas, et al.236
sciousness Assessment as constructed on the basis of the psychometric out-comes and the operational definition clarifications reached via the two 2001 studies. This scale was constructed as follows: in respect to the international and the Greek evidence, the main effort was directed so as to integrate the operational definitions provided by Fenigstein, et al. (1975) and the theoreti-cal aspects proposed by Duval and Wicklund (1972) and also by Zaborows-ki (1987) and Carver (1979), taking cultural issues into account as well (Lal-wani, Shrum, & Chiu, 2009). The final product of such an attempt should avoid possible bias in terms of culture (Poortinga & van de Vijver, 1987; van de Vijver & Tanzer, 1997) and reflect the main theoretical structures as evi-dent in the literature. It was expected that at least a two-factor structure (pri-vate and public self-consciousness) would hold for this new scale, but the presence of more specific facets could also be possible. Four studies (a pilot and three main ones) are presented hereafter.
Pilot StudyThe first step toward creating the Scale for Self-Consciousness As-
sessment was the interview techniques implemented at the item construc-tion stage. These interviews were carried out according to the theoreti-cal directions given by Duval and Wicklund (1972), Wicklund (1975), and Zaborowski (1987). The interviewees were 38 undergraduate psychology students from the University of Athens. The questions used for these in-terviews were constructed in respect to: (a) the following construct defi-nitions: “private self-consciousness represents a self-focused attention to reflect on covert, hidden and personal aspects that are not easily acces-sible to others, e.g., private motives, feelings and beliefs. Public self-con-sciousness has a propensity to attend to those self aspects that are also ex-hibited to the public, e.g., appearance and mannerisms” (Chang, 1988, p. 635), (b) the operational definitions proposed by Fenigstein, et al.(1975) and by Zaborowski (1987), and (c) the factor structure evidence distin-guishing between the two private self-consciousness and two public self-consciousness dimensions (Burnkrant & Page, 1984; Mittal & Balasubra-manian, 1987).
Through the interview stage, the conceptual facets along with the cognitive procedures related to self-consciousness were addressed and the outcome was a set of 104 items. Those items were then administered to a pilot sample of university students (N = 54) and were evaluated for their basic statistical properties. Some of those items (22 in number) were eliminated on the basis of those properties, mostly due to metric and methodological discrepancies such as very high intercorrelations and/or extremely low shared variance. The remaining 82 items (intermediate ver-sion) resulted in the scale to be tested under a first psychometric attempt through Study 1.
Scale for Self-conSciouSneSS aSSeSSment 237
Study 1Sample and Procedure
The aim of this first study was to use principal components analysis to explore the factor structure of the 82-item questionnaire, constructed in the pilot study. The sample for this first study consisted of 494 participants (39% men, 61% women). Of these, 319 were university students (65%) and 175 were non-student adults (35%). The university students (M age = 22 yr.) were recruited from several departments of the University of Athens (Philosophy, Psychology, Theology, Physics, Medicine, and Physical Edu-cation & Sports) and from the National Metsovion Polytechnic School of Athens. The non-student adults (M age = 34 yr.) were employed mainly as clerks (56%); 11% were doctors or lawyers, 9% were in other various occu-pations and 10% were unemployed. A large part of the non-student adult sample (43%) came from rural areas, and the remaining 57% from the ur-ban area of Athens. All participants responded to the 82-item question-naire on a 5-point Likert-type scale with anchors 5: Always true for me and 1: Never true. Results
For the 494 participants, a hierarchical cluster analysis for all 82 items was employed in an attempt to identify homogeneous sets of data. The results indicated several clusters of variables but 20 items were not part of any of these homogeneous sets and they were excluded from further analysis. Three items were verbally similar to other items and were ex-cluded as well, in order to avoid possible sources of collinearity. Thus, 59 items (revised version) remained in the item pool and they were analyzed further through exploratory factor analysis models. The participant/vari-able ratio for this analysis was approximately 8, satisfying the large sam-ple prerequisite (Kline, 1993; Streiner, 1994; Bryant & Yarnold, 1995). For these analyses and all exploratory factor analyses in all studies, principal component and orthogonal rotation of axes methods were employed (cut-off loading for inclusion of items in the factors was .45), as the aim at this stage was to arrive at a scale with dimensions as clear but also as stable and independent of each other as possible.
A four-dimensional structure appeared plausible with 40.38% of the variance being explained. The first component (explaining 11.6% of the variance) was related to public self-consciousness and consisted of 13 items on appearance. The second component (10.5% of the variance) was related to private self-consciousness and consisted of 13 items on self-re-flection estimates. The third component (9.8% of the variance) was related to private self-consciousness as well, but consisted of 16 items reflecting
K. Mylonas, et al.238
self-knowledge. Finally, the fourth component (8.5% of the variance) was related to public self-consciousness but consisted of nine items on social fit. In all, 50 out of the 59 items participated in this factor structure.
A number of expected cross-loading items somewhat blurred the con-tent of the four dimensions, but the general outcome resulted into two “public” and two “private” self-consciousness dimensions, with minor irregularities, thus the overall result of this series of exploratory analy-sis seemed promising and indicated further exploration of the 59-item re-vised version questionnaire for the second sample of 248 university stu-dent respondents.
Study 2Sample and Procedure
The aim of this second study was to explore the factor structure of the 59-item version of the scale and to replicate the structure found in the first study. The sample of the second study consisted of 248 participants (40% men, 60% women) with adequate representation of rural areas. The par-ticipant/variable ratio was approximately 4, suitable for principal compo-nent analysis (Kline, 1993; Streiner, 1994; Bryant & Yarnold, 1995). All par-ticipants were university students of the Physical Education and Sports Department who responded to the shorter revised version of 59 items on the same 5-point Likert-type scale as in Study 1. Results
Principal component analysis for a four-dimensional solution, fol-lowed by orthogonal rotation of the axes, explained 37.5% of the vari-ance. The first component (explaining 12.0% of the variance) was related to private self-consciousness and consisted of 19 items on self-reflection estimates. The second component (11.2% of the variance) was related to public self-consciousness and consisted of 11 items reflecting appearance. The third component (7.8% of the variance) was related to public self-con-sciousness as well but consisted of 10 items reflecting social peer fit, the sense that others (social environment and friends) evaluate and approve or disapprove of specific behavior. Finally, the fourth component (6.5% of the variance) was related to private self-consciousness and consisted of six items reflecting self-knowledge. A number of expected cross-loading items again somewhat blurred the identities of the four principal compo-nents, but for this sample too, the general outcome resulted intwo “pub-lic” and two “private” self-consciousness dimensions.
Comparing the structures between Study one (N1 = 494) and Study two (N2 = 248), there were several indications of differences in the struc-tures as derived for the 59-item version of the Scale for Self-Consciousness Assessment. Different items loaded on different components and even
Scale for Self-conSciouSneSS aSSeSSment 239
some of the cross-loading items were not the same across the two solu-tions. Tucker’s phi (φ) coefficient was employed to further explore these differences. These congruence coefficients reached levels of identity (.90 or higher) only for the first N1 and second N2 pair of components (φ = .95), meaning that only one of four components could be considered equiva-lent across the two solutions. The remaining three component pairs were similar but not identical (φ indices ranged from .83 to .85). An unexpected similarity of the third N1 component with the fourth N2 component (.83) and with the first N2 component (.79) was also observed. These unexpect-ed results were at the similarity and not the identity level, but they certain-ly raised some questions about the content identity of the third N1 compo-nent and/or the salience of the first N2 component. This, along with other research questions, had to be explored further in the next stages.
Finally, a target rotation (Procrustean method, as described by van de Vijver and Leung, 1997) was implemented. As expected, the Tucker’s φ co-efficients for the target-rotated solution did not reach equivalence levels for three of the four dimensions.
To gain more insight and possibly reach a more stable set of items, through a similar approach with the one described in the previous para-graphs and comparing across the two studies, more items were identi-fied as being problematic or even metrically inappropriate. Using mini-mum loading criteria along with Tucker’s φ coefficients and Procrustean factor solutions, items were compared across the solutions for N1 and N2. Through successive attempts (the intermediate results are not presented for reasons of brevity) a common solution was achieved for both sam-ples, with 42 items in the structure. For this, Tucker’s φ coefficients for the four dimensions, as re-computed for the target-rotated solution (Table 1) reached identity (.96, .93, .90, and .92) with 37 items loading on at least one dimension. This structure was tested in the next stage.
At this last stage concerning these two samples, we returned to the first study’s sample (N = 494) and computed a series of confirmatory fac-tor analysis (CFA) models to further eliminate items that were biased or prevented a good factorial fit. Thus, confirmatory factor analysis was em-ployed at this stage only as yet another criterion for item elimination, be-fore proceeding to a proper confirmatory factor analysis stage computed for a new sample (Study 3). A 33-item scale was reached as the best pos-sible fit at this stage, and was achieved through successive CFA attempts. The main profit from these attempts (the results are summarized in Table 2 but are not presented here in full for reasons of brevity) was that three possibly cross-loading items were excluded from further analysis along a fourth item suspect of multicollinearity. The χ2 criterion for this 33-item scale was statistically significant (as expected, as this criterion is a very
K. Mylonas, et al.240
TAbLE 1target-rotated Factor Structure For the 42-item Scale
4-factor Structure, 42-item Solution F1 F2 F3 F4 h2
I am concerned about what other people think about me. .85 .09 .04 −.08 .74I am interested in the impression I make on other people. .85 .01 −.01 −.07 .73I am interested in how other people evaluate me. .82 .08 .01 .00 .68I am concerned about the impression I make on other peo-
ple. .81 .02 .04 .03 .66I am interested in how I present myself to other people. .79 .18 .13 .00 .67I am interested in what other people think about me. .79 −.02 .08 −.02 .63I am concerned about how other people evaluate me. .68 .10 −.02 −.12 .49I am interested in how other people evaluate me regarding
my job. .60 .25 .21 .08 .47I am interested in my physical appearance at my workplace. .56 .22 .22 .10 .42Prior to my actions I check my motives. −.08 .70 .14 .16 .54Prior to my actions I check my pursuits. −.01 .68 .03 .17 .49Prior to my actions I check my specific traits. .23 .66 .08 .19 .53Prior to my actions I check my needs. −.06 .66 .08 .09 .45Prior to my actions I check my desires. .04 .62 .25 .06 .45I think about the way that I act. −.02 .60 .04 .42 .54I act carefully and with precision. .00 .53 .01 .20 .32When I perform a task, I know from the very beginning the
extent to which my goals will be achieved. −.21 .48 .01 .24 .33I evaluate myself after every action. .06 .47 −.02 .34 .34I expect a lot of myself in any action. .03 .45 −.01 .24 .26I am fully aware of my behavior in my social relationships. .01 .45 .13 .41 .39I criticize myself on the basis of previous experience. .10 .40 .02 .34 .29Prior to my actions I check my mood state. .12 .37 .17 .22 .23I reflect on any personal success’ consequences upon me. .06 .36 .14 .18 .19When with friends, I am interested in being helpful. −.05 .02 .81 .01 .66When with friends, I am interested in being understanding. .18 .13 .77 .01 .64When with friends, I am interested in being pleasant. .24 .16 .72 .06 .61When with friends, I am interested in being honest. −.02 .10 .66 .13 .46I am interested in my friends’ criticism about my behavior. .48 −.01 .57 −.07 .56I am interested in my friends’ criticism regarding a mistake
I have made. .05 .45.42 .08 .51When with friends, I am interested in showing sensitivity. .16 .08 .48 .23 .32In my collaborations, I am interested in maintaining friendly
relationships. .21 .24 .47 .02 .32I can describe my emotions. .21 −.03 .22 .79 .72I am aware of my desires and I can describe them. .09 .14 .25 .71 .59I can easily describe my thoughts. .04 −.08 .23 .67 .51I think about myself and I feel I have a deep knowledge of
myself. −.09 .14 .10 .60 .40I could thoroughly describe myself to someone else. .30 .01 .20 .56 .44All of my actions are accompanied by emotions I can easily
understand. .03 .22 .24 .52 .38(continued on next page)
Scale for Self-conSciouSneSS aSSeSSment 241
austere one) but the goodness-of-fit indices were more acceptable in com-parison to the other models. RMSEA was smaller, meaning that this so-lution was homing in on the expected theoretical structure. The Tucker-Lewis estimate was positive and the difference between the χ2 criteria was statistically significant, portraying some small departure from the “null-model” 36-item scale towards a better, more clearly identifiable four-di-mensional self-consciousness structure.
The four factors comprising the 33 items appear in Table 1 (44.2% of the variance explained): the first factor consisted of 9 items explaining 14.1% of the variance and was named Public Self-consciousness or “ap-pearance”. The second factor consisted of 12 items explaining 13.1% of the variance and was named Private Self-consciousness or “self-reflective-ness”. The third factor consisted of 6 items explaining 9.5% of the vari-ance and was named Public Self-consciousness or “social fit”. Finally, the fourth factor consisted of 6 items explaining 7.5% of the variance and was named Private Self-consciousness or “self-knowledge”. The internal con-sistency indices for the four dimensions were satisfactory (.92, .84, .83, and .74, respectively). When the scores within each factor were averaged, a possible social desirability effect was noticeable, apparent mainly for the third factor; this produced higher respective mean scores and negative skewness. Although a moderate skewness existed for the first factor as well, these scores did not seem to be largely affected by this response style. However, at this stage such effects were not tested, as this question along with the main question on factor confirmation were to be addressed by the last study.
TAbLE 1 (cont’d)target-rotated Factor Structure For the 42-item Scale
4-factor Structure, 42-item Solution F1 F2 F3 F4 h2
I do not act at random; my actions are in line with my principles. −.14 .31 .02 .32 .22
I reflect a lot on every action before acting. .06 .25 −.08 .12 .09I reflect on any consequences of personal failure. .11 .28 −.03 .10 .10My behavior is affected by my deep understanding
of my emotional state and its changes. .21 .28 .34 .08 .24I reflect on my mistakes. −.01 .17 −.02 .03 .03Note.—The item “I am fully aware of my behavior in my social relationships” was excluded from further analysis. All four cross-loading items were closely inspected and along with an initial 37-item attempt during main Study 3 it was found that multicollinearity problems were present due to this item. Additionally, the item’s loadings on two factors were both relatively low and very close in magnitude. Thus, the CFA models explored during Study 2 refer to a 36-item Scale and its shorter 33-item form. The cutoff loading for Study 2 solution was .35. The communality estimates refer to the Procrustean solution loadings.
K. Mylonas, et al.242
Study 3Sample and Procedure
The sample for this final study consisted of 216 participants (34% men, 66% women) all university students (M age = 20 yr.). These partici-pants were recruited from several Departments of the University of Ath-ens (Philosophy, Physics, Psychology, Medicine, Political Science and Pub-lic Administration, and from the National Metsovion Polytechnic School of Athens). The final aim was to confirm the factor structure found in pre-vious attempts through confirmatory factor analysis models for this final sample of University students. All participants responded to the 42-item questionnaire rated on a five-point Likert-type scale (5: Always true for me, 4: Many times true for me, 3: Sometimes true and sometimes not true, 2: Few times true for me, 1: Never true). Results
Although there were 42 items administered, the analysis regarded 33 of them, as these were shown to participate in the four-factor structure found in the two previous studies. The variable ratio exceeded 5, satisfy-ing the criterion set by Bryant and Yarnold (1995), but ratio of participants to free parameters, more suitable for CFA, reached only 3 and was less than satisfactory (Streiner, 1994; Tabachnick & Fidell, 2001). Although this is a limitation of the current study, this is the first time SSCA data have been subjected to such an analysis, which should in any case provide ini-tial insight for further research in respect to the scale’s structure. Sever-al models were tested after having ruled out collinearity problems, after
Null model36 items (3 cross-loadings)a 1,633.59 591 2.76 .063 .84 .8636 items (no cross-loadings)b 1,760.14 594 2.96 .067 .82 .84 NA NA NA33 items (no cross-loading)c 1,321.28 495 2.67 .064 .85 .87 .05 312.3* 96
Note.—
*Statistically significant at the .001 level. NA: non-applicable, since (χ20 ÷ df0) – (χ2
i ÷ dfi) < 0. aFor this model all 36 items which loaded on at least one factor (see Table 1) were considered. Three items were allowed to cross-load. A fourth item had been already eliminated due to multicollinearity problems. bFor this model, no cross-loadings were allowed for the same 36 items, that is, we related each of the items only with the highest loading factor. cFor this mod-el, all three items which cross-loaded on the factors were totally excluded from the analysis. p < .00001 for all models.
χ20
χ20 1
df0
df0
df1
χ21
TLI =
−
−
Scale for Self-conSciouSneSS aSSeSSment 243
having ascertained overidentification of all models to be tested and find-ing no missing values. The Pearson r indices among the 42 items were also compared to Kendall T and Spearman ρ indices, as the initial scoring scale is a 5-point Likert-type one and there was a need to ascertain that Pearson r indices could be safely employed in these models. Indeed, Pearson r in-dices did not differ at a statistically significant level (the comparison was carried out on Fisher z transformations of initial indices and through for-mula 1) with the Kendall and Spearman coefficients, thus it was justified to use Pearson r indices for model estimation:
[1]
where z1 and z2 are the Fisher z transformations of initial indices, N1 = N2 = 216; the z-criterion is evaluated for its statistical significance un-der the standard normal distribution. Six CFA models were tested.
Model 1: Independence model.—The independence model (all variables uncorrelated) was easily rejected. The χ2 criterion reached 4,561.34 and was statistically significant (p < .00001) for 861 degrees of freedom. Fol-lowing this, a series of models were tested for which either the number of factors was under question or the associations and their modifications, or both (for a brief summary of the outcomes for all models see Table 3).
Model 2: Single factor, 33-item model.—The possibility of a unifactori-al solution was tested. This model was rejected as well, as the χ2 criterion reached 3,342.77 and was statistically significant (p < .00001) for 495 de-grees of freedom.
Model 3: Two-factor, 33-item model.—For this model all the private self-consciousness items were aggregated into a single factor and all public self-consciousness items were aggregated into another single factor and the two dimensions were tested for their presence in the data. The χ2 cri-terion reached 1,838.35 and was statistically significant (p < .00001) for 494 degrees of freedom. This model was rejected as well.
Model 4: Four-factor, 33-item model.—For this model, each item was re-lated to its respective factor as indicated through Study 2 (loadings in Ta-ble 1). This was the target model as the four-factor structure for the 33 items was expected on the basis of all previous exploratory analyses to fit the data best. Although χ2 was significant, RMSEA was .071 indicat-ing a very large drop in comparison to previous models. Indices indicat-ed that this model was better, although this model did not fit the data well. For this reason, and in an attempt to remedy possible inconsisten-cies which caused the lack of fit, modification suggestions were computed and these suggested six additional direct paths from latent variables to the
z1
N1 − 3 N2 − 31 1+
− z2z =
K. Mylonas, et al.244
observed variables and a number of interrelated item-specific variances. These modifications were tested as well but the χ2 was still significant al-though RMSEA dropped to .042. One might argue that this model should be accepted, even though the fit was far from perfect. One might indeed consider this model final and stop testing at this point, but the fact that such a model contained six direct links of factors to items which clearly belonged to other dimensions and would have to be considered as cross-loading in the final structure—let alone the correlated item-specific vari-ances—called for more extensive actions. It was thus decided to treat these six items as possible disturbances and exclude them from further analysis, recomputing the models for a 27-item scale.
Model 5: Four-factor, 27-item model.—Each of the 27 items was related to its respective factor as indicated through Study 2 (loadings in Table 1). The fit indices were not as good as those for the modified 33-item model 4, but were certainly better than the fit indices for the unmodified model 4. Thus, it was decided to pursue this further and compute modification
modeli 276.70 231 < .05 ( = .021) 1.198 .030 .90 .97 .818d-i 743.99 258*All Δχ2 values are statistically significant at the .00001 level. aAll variables uncorrelated. bAll 33 items loading on a single latent variable. c33-item solution assuming two factors, a Pri-vate Self-Consciousness and a Public Self-Consciousness. d33-item hypothetical solution rep-licating the zero cross-loadings factor structure found during study 2 (see also Tables 1 and 2). eModifications imposed on (d) in respect to direct paths from latent variables to observed ones and in respect to correlations between item-specific variances. fReduced four-factor 27-item model (six items eliminated to avoid suggested cross-loadings). gModifications im-posed on (f) in respect to direct paths from latent variables to observed ones. hReduced four factor 24-item model (three more items eliminated to avoid suggested cross-loadings). iMod-ifications imposed on (h) in respect to correlations between item-specific variances only.
Scale for Self-conSciouSneSS aSSeSSment 245
suggestions for the 27-item model as well. Three more direct paths from latent variables to those observed were suggested (albeit not in accord with original expectations), so the model had to be redefined according to these suggestions and the fit indices were recomputed. Although, some gains in fit indices were observed, the three cross-loading items were ex-cluded from further analysis and a 24-item scale was tested.
Model 6: Four-factor, 24-item model.—Once more, each of the 24 items was related to its respective factor as was indicated through Study 2 (load-ings in Table 1). There was some improvement in fit statistics. When mod-ification suggestions were computed, no direct paths were suggested al-though a number of correlations between item-specific variances were suggested. Some of them referred to covariances between items of the same factor (e.g., within the Self-Reflectiveness, Private Self-Conscious-ness factor) but some (53%) of these correlated covariances ran across dif-ferent factors showing some non-orthogonality between these factors. This final modified model was tested provisionally accepting a manage-able amount of non-orthogonality; then the fit indices along with the stan-dardized indices for the correlations of item specific variances were re-computed. Fit statistics were much better. No modification indices could be suggested for this final model (Fig. 1). Considering all, this model was the best fitting model and was the least complicated in terms of interpre-tation, as by computing the standardized estimates for the correlations between item-specific variances it was evident that these were very small and did not indicate alarming intercorrelation among factors (median esti-mated correlation = .12, the highest estimate for correlation of item specific variances across different factors being .19).
Estimates of reliability were computed for these four factors (the first consisting of 8 public self-consciousness items, the second consisting of eight private self-consciousness items, the third consisting of four pub-lic self-consciousness items and, the fourth consisting of four private self-consciousness items). The standardized Cronbach’s α indices for these fac-tors were .92, .80., .70, and .74, respectively. Following Hair, Black, Babin, and Anderson (2010), we also applied Equation 2 to gain better insight in respect to construct reliability, and extending this, to construct convergent validity. These estimates were .91, .80, .72, and .70 for the first to fourth fac-tors, respectively. Where, L is the standardized factor loading of each ob-served variable on the factor, n is the number of loadings, and e is the error variance associated with each observed variable:
[2]∑Li
∑Li ∑ei
n
n n
2
2i = 1
i = 1 i = 1
⏝ ⏝ ⏝ ⏝
⏝ ⏝
+
CR =
K. Mylonas, et al.246
Pu 1
Pu 2
Pr 1
i38
i29
i21
i35
i23
i37
i42
i14
i30i26 i16
i3
i36
i18
i41
i39
i32
i1i5i7i9i10
5154
6644 69 45
70
58
6782
61
38
6865
4860
4184
26
32
13
11
27
9 53
58
77
64
12
819
52
74
79 51
66
811
301
27
4
40
20
29
28
46
5532 63
86
−100
−100
−100
−100
−16
−10
56
13
1570
81
−11
−4
86
82
77
89
84
85
74
i12
Pr 2
Fig. 1. Final factor structure for the 24-item Scale for Self-Consciousness Assessment. Pu 1 = Public Self-Consciousness factor (1st factor), “appearance”, 8 items, Cronbach α = .92, CR = .91. Pr 1 = Private Self-Consciousness (2nd factor) “self-reflectiveness in respect to ac-tions”, 8 items, Cronbach α = .80, CR = .80. Pu 2 = Public Self-Consciousness factor (3rd fac-tor), “social fit,” 4 items, Cronbach α = .70, CR = .72. Pr 2 = Private Self-Consciousness (4th factor), “self-knowledge,” 4 items, Cronbach α = .74, CR = 70. χ2 = 276.70, df = 231, p = .02, RM-SEA = .03. All parameter values appear as the actual coefficients × 100.
i15
13
5
Scale for Self-conSciouSneSS aSSeSSment 247
Another way of gaining insight in respect to convergent validity is to compute the average variance extracted (average communality) which is the sum of all squared loadings over the total number of items in the so-lution. This is expected to be .50 or higher; for our solution it reached .45 which is acceptable, in terms of average explained variance. Furthermore, average variance extracted for each pair of factors (six pairs in all) was computed in order to estimate the discriminant power among the dimen-sions, as each of the variance extracted indices should exceed the squared correlation of the two factors in the pair. In all pairs, the average variance extracted was much higher than the respective squared correlation of the factors. All results in respect to convergent validity and discriminant pow-er along with construct reliability indicated good factorial validity overall.
Obviously, due to the successive reduction of items, the content with-in each factor for this final solution did not fully coincide with the out-comes reached through the exploratory stages and mainly through Study 2. The first factor in this final solution remained almost the same with the one found during Study 2, so it was still named “appearance”; however, the second factor suffered a number of item losses and was now named “self-reflectiveness in respect to actions,” or in short “personal actions”; the third factor also remained unchanged in content and was still named “social fit”; finally, the fourth factor of “self-knowledge” remained the same although the link between actions and emotions (“All my actions are accompanied by emotions I can easily understand”) was not a part of the fourth factor in this final solution.
diScuSSion In the present series of studies, two complementary theories were
mainly addressed, the Duval and Wicklund theory (1972), and the Za-borowski theory (1987), in devising a Scale for Self-Consciousness As-sessment. Zaborowski’s internal and external Self-Consciousness dimen-sions were found in the data, split in the final factor structure in two facets each. Accordingly, the results concurred with the Fenigstein, et al. (1975) research, corroborating evidence in distinguishing between the two pri-vate self-consciousness and two public self-consciousness dimensions. The two Private Self-Consciousness facets are distinguished on the ba-sis of their general or specific use. For the “self-reflectiveness in respect to actions” facet, the internal-self characteristics are explored and re-con-sidered at certain levels by the individual under specific circumstances or conditions. Prompted by these conditions the individual addresses him-self in terms of motivation, goals, desires, and needs in conjunction with the ways of action and their evaluation, as if the self is being reflected on an inner mirror. The second facet (“self-knowledge”) is a general aware-
K. Mylonas, et al.248
ness self-approach with the focus on personal emotions, thoughts, and de-sires and the extent to which those are apparent to the individual.
The Public Self-Consciousness facets are even more clearly distin-guishable, since the “appearance” facet is the extent to which one can conceptualize the way others think and judge them. The other facet (“so-cial fit”) refers to the social relationships the individual has, in respect to the individual’s levels of attempts to be socially accepted by peers and friends. Such a perception of this facet might suggest that social desirabil-ity effects may be present, an issue initially raised during the last stages in Study 2. This issue was addressed in Study 3 as a set of measures that could be employed to test for such social desirability effects and was in-cluded in the data. In particular, the 19 “Lie scale” items from the Eysenck Personality Questionnaire (EPQ; 1975) were employed for Study 3 and from those an aggregate “lie” score for each of the 216 respondents was derived as a proxy measure of social desirability.
Correlations (η) between each item and this aggregate “lie” score were computed, as this would indicate which items might be affected by so-cial desirability in terms of shared variance. For nine out of the 24 items in the final factor structure these correlations ranged between .20 and .25 (all remaining η indices were lower than .20). Although .25 does not imply strong associations with social desirability, it was worth the effort to ex-amine these items further and depict their position in the final factor struc-ture. The highest index (.244) was observed for the 12th item in the scale (see also Fig. 1). This private self-consciousness item (“Prior to my actions I check with my pursuits”) is the item for which the largest number of error covariance associations (seven such associations) were observed in the final factor structure. Other private self-consciousness items associat-ed with social desirability were “I act carefully and with precision”, “Prior to my actions I check with my specific traits” and, “I think about myself and I feel I have a deep knowledge of me”. For these, the error covariances present in the factor structure were just one for each item. Finally, public self-consciousness items sharing variance with social desirability were “I am interested in my physical appearance at my work-place” (one error co-variance), “I am concerned about what other people think about me” (one error covariance), “I am interested in the impression I make to other peo-ple” (three error covariances), “I am interested in how other people evalu-ate me” (one error covariance) and, “When with friends, I am interested in being honest” (one error covariance). Having examined these associa-tions, it might be the case that social desirability interferes, although not to a large extent, possibly producing some of the item-specific variances associations as an underlying factor. Certainly, such a conclusion was not directly tested in this study and requires further research.
Scale for Self-conSciouSneSS aSSeSSment 249
In relating Private and Public Self-consciousness, although distin-guishable in terms of metric attributes, they seem to comprise “two sides of the same coin”, being in line with George Kelley’s idea (1955, 1963) that everybody has the vital need to retain a view for oneself and then gain val-idation for it, which means this view needs to be affirmed by others under specific “times” and “places.” The two sides of the same coin are also in line with another fundamental idea of personality theory, that everybody has the potential for conscious reflexivity—reflecting on both the actions of others and on their own actions as well. Such interpretations of private and public self-consciousness might also explain the small positive corre-lations estimated in the final CFA model between private self-conscious-ness factors and the second public self-consciousness facet (“social fit”) with these correlations possibly inflated due to underlying social desir-ability effects. Such effects have not been addressed, and the current find-ings are in partial agreement with the findings reported by Wickens and Stapel (2010) where public and private self-consciousness were found to be correlated at a medium level.
On psychometric grounds, the initial pool of items was gradually re-duced following specific criteria and factor analysis models’ outcomes to shorter versions which were homing on the factor structure, as described in the Results section. Using covariance structure modelling, final deci-sions were reached on the appropriateness of the items finally involved in the scale. All four facets of the new self-consciousness scale seem to be much more in line with Burnkrant and Page (1984) in regard to “self-re-flectiveness in respect to actions” as part of the private self-consciousness and with Mittal and Balasubramanian (1987) in regard to “appearance” as a dimension of public self-consciousness, enriching the operational definition of the term. Although the 24-item version seems to be the best structured and psychometrically solid version in respect to the series of versions considered in these studies, further support is needed through studies of self-consciousness correlates in order to further test for valid-ity levels and for hypotheses related to the factor structure of the Scale for Self-Consciousness Assessment.
REFERENCES
anderSon, e. m., Bohon, l. m., & Berrigan, l. P. (1996) Factor structure of the Private Self-Consciousness scale. Journal of Personality Assessment, 66, 144-152.
BernStein, i. h., teng, g., & garBin, c. P. (1986) A confirmatory factoring of the self-consciousness scale. Multivariate Behavioral Research, 21, 459-475.
Bryant, F. B., & yarnold, P. r. (1995) Principal components analysis and exploratory and confirmatory factor analysis. In L. G. Grimm & P. R. Yarnold (Eds.), Reading and understanding multivariate statistics. Washington, DC: American Psychological Association.
K. Mylonas, et al.250
Burnkrant, r. e., & Page, t. J. (1984) A modification of the Fenigstein, Scheier, and Buss Self-Consciousness Scales. Journal of Personality Assessment, 48, 629-637.
BuSS, a. h. (1980) Self-consciousness and social anxiety. San Francisco, CA: Freeman.carver, c. S. (1979) A cybernetic model of self attention processes. Journal of Personality
and Social Psychology, 37, 1251-1281.carver, c. S., & glaSS, d. c. (1976) The Self-Consciousness Scale: a discriminant va-
lidity study. Journal of Personality Assessment, 40, 169-172.carver, c. S., & Scheier, m. F. (1998) On the self-regulation of behavior. New York: Cam-
bridge Univer. Press.chan, d. W. (1996) Self-consciousness in Chinese college students in Hong Kong. Per-
sonality and Individual Differences, 21, 557-562.chang, l. (1988) Factor interpretations of Self-Consciousness Scale. Personality and
Individual Differences, 24, 635-640.conWay, m., & giannoPouloS, c. (1993) Self-esteem and specificity in self-focused at-
tention. Journal of Social Psychology, 133, 121-123.cooley, c. h. (1907) Social consciousness. The American Journal of Sociology, 12, 675-694.cramer, k. m. (2000) Comparing the relative fit of various factor models of the self-
consciousness scale in two independent samples. Journal of Personality Assessment, 75(2), 295-307.
duval, S., & Wicklund, r. a. (1972) A theory of objective self-awareness. New York: Aca-demic Press.
eySenck, h. J., & eySenck, S. B. g. (1975) Manual of the Eysenck Personality Questionnaire. London: Hodder and Stoughton.
FenigStein, a. (1997) Self-consciousness and its relation to psychological mindedness. In M. McCallum & W. E. Piper (Eds.), Psychological mindeness: a contemporary un-derstanding. Mahwah, NJ: Erlbaum. Pp. 105-131.
FenigStein, a., Scheier, m. F., & BuSS, a. h. (1975) Public and private self-consciousness: assessment and theory. Journal of Consulting and Clinical Psychology, 43, 522-527.
hair, J. F., Black, W. c., BaBin, B. J., & anderSon, R. E. (2010) Multivariate data analysis. (7th ed.) Upper Saddle River, NJ: Prentice Hall.
heinemann, W. (1979) The assessment of private and public self-consciousness: a Ger-man replication. European Journal of Social Psychology, 9, 331-337.
hull, J. g., & levy, a. S. (1979) The organizational functions of the self: an alternative to the Duval-Wicklund model of self-awareness. Journal of Personality and Social Psychology, 44, 1097-1109.
ito, m. (1998) Consciousness from the viewpoint of the structural-functional relation-ships of the brain. International Journal of Psychology, 33, 191-197.
JameS, W. (1892) The stream of consciousness. First published in Psychology, Chapter XI. Cleveland & New York: World.
kelley, g. (1955) The psychology of personal constructs. Vols. 1 & 2. New York: W. W. Norton.
kelley, g. (1963) A theory of personality. New York: W. W. Norton.kline, P. (1993) The handbook of psychological testing. London: Routledge.lalWani, a. k., Shrum, l. J., & chiu, c-y. (2009) Motivated response styles: the role
of cultural values, regulatory focus, and self-consciousness in socially desirable responding. Journal of Personality and Social Psychology, 96(4), 870-882.
Scale for Self-conSciouSneSS aSSeSSment 251
mead, g. h. (1914) 1914 class lectures in social psychology. In D. L. Miller (Ed.), The individual and the social self. Chicago, IL: Univer. of Chicago Press. Pp. 27-105.
mittal, B., & BalaSuBramanian, S. k. (1987) Testing the dimensionality of the Self-Consciousness Scale. Journal of Personality Assessment, 57, 53-68.
Piliavin, J. a., & charng, h. (1988) What is the factorial structure of private and public Self-Consciousness Scales? Personality and Social Psychology Bulletin, 14, 587-595.
Poortinga, y. h. (1989) Equivalence of cross-cultural data: an overview of basic issues. International Journal of Psychology, 24, 737-756.
Poortinga, y. h., & van de viJver, F. J. r. (1987) Explaining cross-cultural differences: bias analysis and beyond. Journal of Cross-Cultural Psychology, 18, 259-282.
reeveS, a. l., WatSon, P. J., ramSey, a., & morriS, r. J. (1995) Private self-consciousness factors, need for cognition, and depression. Journal of Social Behavior and Personal-ity, 10, 431-443.
Scheier, m. F., & carver, c. S. (1985) The self-consciousness scale: a revised version for use with general populations. Journal of Applied Social Psychology, 15, 687-699.
Silvia, P. J., & gendolla, g. h. e. (2001) On introspection and self-perception: does self-focused attention enable accurate self-knowledge? Review of General Psychol-ogy, 5, 241-269.
Streiner, d. l. (1994) Figuring out factors: the use and misuse of factor analysis. Cana-dian Journal of Psychiatry, 39, 135-140.
taBachnick, B. g., & Fidell, l. S. (2001) Using multivariate statistics. (4th ed.) Boston, MA: Allyn & Bacon.
van de viJver, F. J. r., & leung, k. (1997) Methods and data analysis for cross-cultural re-search. Thousand Oaks, CA: Sage.
van de viJver, F. J. r., & tanzer, n. k. (1997) Bias and equivalence in cross-cultural as-sessment: an overview. European Review of Applied Psychology, 47, 263-279.
veligekaS, P., & mylonaS, k. (2001) Epaneksetase klimakas aftosinedesias se digma tu hellinikou plethesmou. aftosinedesia ke simvouleftike [Revision of the Self-Conscious-ness Scale in a Greek sample: self-consciousness and counseling.] [Review of Coun-selling and Guidance], 56 & 57, 53-71.
veligekaS, P., mylonaS, k., & gari a. (2001) Revision of the Self-Consciousness Scale: track and field athletes. The 10th World Congress of Sport Psychology Proceedings, Skiathos, May 28–June 2. Pp. 83-85.
veligekaS, P., mylonaS, k., gari, a., PlouBidiS, g., & mantzavinou, k. (2003) Assess-ment of self-consciousness in the Greek population and applications in athletic settings. Collection of articles from The XI European Congress of Sport Psychology, Co-penhagen, July 22–27. ISbN 87 89 361 96 2. [CD]
vygotSky, l. S. (1925/1999) Consciousness as a problem in the psychology of behavior. New York: Peter Lang Publishing.
WatSon, P. J., & Biderman, m. d. (1993) Narcissistic Personality Inventory factors: splitting and self-consciousness. Journal of Personality Assessment, 61, 41-57.
WickenS, c. J., & StaPel, d. a. (2008) I versus we: the effects of self-construal level on diversity. Social Cognition, 26, 368-377.
WickenS, c. J., & StaPel, d. a. (2010) Self-awareness and saliency of social versus indi-vidualistic behavioral standards. Social Psychology, 41(1), 10-19.
Wicklund, r. a. (1975) Objective self-awareness. In L. Berkowitz (Ed.), Advances in ex-perimental social psychology. Vol. 8. New York: Academic Press. Pp. 233-275.
K. Mylonas, et al.252
zaBoroWSki, z. (1980) Z pogranicza psychologii spolecznej i psychologii osobowoci [At the in-terface of social and personality phychology]. Warsaw, Poland: Panstwowe Wydawnict-wo Naukowe.
zaBoroWSki, z. (1987) A theory of internal and external self-consciousness. Polish Psy-chological Bulletin, 18, 51-61.
