The Comparison of Nomothetic (Highly Structured) and Idiographic (Open-ended) Measures of Multifaceted Self-Concepts. Lawrence A. Roche & Herbert W. Marsh Faculty of Education, University of Western Sydney, Macarthur. Abstract Renewed interest in self-concept as a useful theoretical construct can be attributed partly to advances made by researchers focussing on the development of reliable rating scales designed to measure multiple facets of self concept. The Self Description Questionnaires (Marsh, 1990), for example, consist of a fixed set of items that are assumed to be generally appropriate to all respondents. An implicit assumption in this nomothetic paradigm is that the underlying structure of self-concept is similar across individuals. In contrast, an idiographic paradigm emphasised in phenomenological approaches assumes that the structure of self-concept is idiosyncratic to each individual. Advocates of this paradigm argue that the items, or at least the scoring of items should be specific to each individual. In this study SDQ III scores from 48 subjects are compared with self- identified discrepancies between self-generated lists of attributes describing their "actual", "ideal" and "ought" selves regarding each of five distinct facets of self-concept. ______________ Paper presented at the AARE Annual Conference, Fremantle, W.A., 1993. Nomothetic and idiographic multifaceted self-concept p. 2 Introduction The Idiographic-Nomothetic Debate In self-concept research -- and personality research more generally -- there is a clear distinction drawn between highly structured, nomothetic measures of "universal" traits or characteristics and more open-ended, idiographic measures of individualised characteristics (Allport, 1962). Nomothetic measures like the Self Description Questionnaires (Marsh, 1988a; 1990; 1992) consist of a fixed set of items that are assumed to be generally appropriate and relevant to all respondents. An implicit assumption is that the underlying structure of self-concept is similar across individuals. In marked contrast, an idiographic approach emphasised in phenomenological approaches assumes that the structure of self-concept is idiosyncratic to each individual. Advocates of this approach (eg. Runyan, 1983) argue that the "items" or "constructs" (cf. Kelly, 1955), or at least the way such items are scored should be specific to each individual. Historically, the two approaches are viewed as antagonistic and typically involve very different research methodologies (see, for example, Eysenck, 1954). Over recent years, however, there has been an increasing awareness that the two approaches should be viewed as complementary (Epstein, 1980; Hermans, 1988; Marsh, 1990; Pelham, 1993). Quite apart from the widespread perception that idiographic and nomothetic techniques are appropriate for two distinct kinds of research goals, it is likely that both approaches can contribute to general questions about the structure of self-concept (or other aspects of personality). The contrasting implicit assumptions referred to above may both be "correct" in that very narrowly defined dimensions may be idiosyncratic to individuals, whereas common dimensions emerge when the stimulus situations are more broadly defined. Psychometric properties and methodological rigour have typically been regarded as critical elements favouring the use of nomothetic measures (Hoge and McCarthy, 1983; Paunonen and Jackson, 1985) and this has led to a general decline in the use of idiographic measures. Multifaceted Self-Concept from a Nomothetic perspective Reviews of self-concept research prior to the 1980's noted a lack of theoretical models for defining and interpreting the construct of self-concept and the poor quality of instruments used to assess it (Wells & Marwell, 1976; Wylie, 1974; 1979). The proposal of a multifaceted, hierarchical model of self- concept by Shavelson, Hubner and Stanton (1976) stimulated considerable empirical research and theoretical refinement leading to the development of the Self Description Questionnaire (SDQ) instruments by Marsh (1988a, 1990). These instruments are designed to measure multiple facets or dimensions of self-concept for pre-adolescents (SDQ I), for early and middle adolescents (SDQ II), and for late adolescents and young adults (SDQ III). Nomothetic and idiographic multifaceted self-concept p. 3 After decades in a theoretical, methodological and empirical wilderness, self-concept research was now able to establish that: (a) self-concept is multifaceted, with quite small correlations between dimensions, and this multifaceted structure can be reliably measured; (b) the multidimensional self-concept responses display strong convergent and divergent validity, being substantially correlated with such criteria as self-concepts in matching dimensions inferred by significant others, and matching academic achievement indicators (for verbal and mathematics self-concepts); (c) self-concept dimensions are systematically and logically related to a variety of other constructs such as gender, locus of control, self-attributions for academic success and failure, physical fitness and participation in sports, and interventions designed to enhance self-concept (see Marsh, 1990, for a review of this research). Such strong support for the construct validity of SDQ responses has led to widespread acceptance and use, as well as endorsement from other leading self-concept researchers. Wylie (1989) evaluated the SDQs as being the strongest and most researched self-concept instruments, and Hattie (1992) concluded that they were "the best set of measures available" (p. 83). Self-Concept from an Idiographic Perspective A wide variety of techniques have been developed to infer self-concept from an idiographic perspective that does not impose a structure on respondents. In the most open-ended formats, respondents are asked to give 20 answers to the question "who are you?" (Bugental and Zelen, 1950), or to respond to incomplete statements such as "I am ..." or "The person I would like to be ..." (Smith, 1992). The most salient feature of this approach is the unstructured, projective format, avoiding the potentially restrictive imposition of the researcher's preconceived notions regarding the structure and content of relevant self-concept components. Whereas many content analysis schemes have been developed for interpreting or scoring the responses, one of the most common is to simply evaluate the favourability or unfavourability of each response. As noted by Wylie (1974, p. 245), however, the problem with this approach is that the favourability is determined by the researcher instead of the respondent so that "the score would, in principle, not be consistently phenomenological." Based on her review Wylie (1974, p.246) concluded that there was no basis for support of the reliability or validity of schemes for coding responses based on this approach. Other approaches incorporating an idiographic component require respondents to judge a fixed set of items according to how they actually are and how they would like to be, to obtain a correlation or discrepancy between the two sets of ratings (Eg., Katz and Zigler, 1967). Although the set of items is fixed, this allows subjects to establish their own standards for each item. The value of this idiographic approach is undermined by results which indicate that the actual ratings alone tend to be more reliable and valid than the differences between actual and ideal ratings (Wylie, 1974; Marsh, 1990; Nomothetic and idiographic multifaceted self-concept p. 4 1993). Another paradigm with a similar idiographic aspect allows subjects to rate the importance of each content- specific area or facet of self concept in addition to rating how they actually see themselves for each facet (Hoge and McCarthy, 1984; Pelham and Swann, 1989). Typically, this paradigm seeks to address the issue of whether specific areas of self-concept are better able to predict global esteem when weighted by their importance -- an intuitively appealing contention argued by James (1890/1963) and reiterated often since (eg Coopersmith, 1967; Harter, 1986; Rosenberg, 1982; Wylie, 1974). Research reviewed by Marsh (1986;1993), however, provided no support for the hypothesis, despite strong intuitive and theoretical appeal, again casting doubt on the idiographic component of this approach. The waning popularity of idiographic approaches seems to stem largely from the lack of methodological rigour and psychometric soundness characterising such measures, in contrast to the nomothetic measures which became available during the 1980's (Paunonen and Jackson, 1985). It is also ironic to note that idiographic approaches predominantly focus on providing measures or predictions of general self-concept that have failed to take into account the multidimensionality of the construct that has been emphasised in the nomothetic approach. Hattie (1992) reviewed the classifications developed by researchers for open-ended self-concept responses and extracted seven major sets of categories. Subjects tended to report attributes relating to domains of self concept labelled academic, family, peer, confidence in self, physical, description of self, and higher-order notions of self. Despite this evidence for a differentiated or multifaceted self-concept, idiographic measures continue to be framed in terms of a global construct. A central aspect of the rationale for such completely open-ended formats is to avoid imposing a restrictive structure. It is likely, however, that too little overt "cueing" as to the kind of information sought by the researcher will itself produce restrictive structures from less tangible (and therefore less containable) sources. Such sources of restriction include situational cues and the respondent's expectations, motivations, mood and creativity. While these influences are potential contaminants regardless of the level of structure provided, their effect is likely to be considerably exaggerated in the absence of other relevant cues. The resurgence of cognitive approaches to dynamic self processes (see Markus and Wurf, 1987) has led to a renewed interest in idiographic measures incorporating actual/ideal discrepancies (e.g., Higgins, 1987; Higgins, Klein and Strauman, 1985; Markus and Nurius, 1986). Higgins' (1987) self discrepancy theory provides conceptual advances over previous research. He expanded the typology of discrepancy scores by considering "actual", "ideal" and "ought" selves from the standpoint of one's "own" self and significant others. In this expanded framework, one's own/actual self is contrasted with, say, four potential standards (ie., own/ideal (the person one would ideally like to be); own/ought (the Nomothetic and idiographic multifaceted self-concept p. 5 person one feels obliged to be); other/ideal (the person a relevant other would ideally like one to be); and other/ought (the person a relevant other feels one should or ought to be)). Higgins also introduced new approaches to inferring discrepancies that are based on an idiographic format. Subjects are asked to list up to ten attributes describing each different standpoint/self combination (e.g., attributes, or characteristics describing who you actually are -- own/actual; who your mother would ideally like you to be -- other/ideal). Discrepancy is inferred from the number of mismatching attributes (antonyms) minus the number of matching (same or synonymous) attributes in two different lists when the lists are compared by external scorers. While the use of discrepancy scores has received frequent criticism on psychometric and methodological grounds (eg. Hattie, 1992; Marsh, 1993), this technique appears to offer some potential as a means of linking idiographic and nomothetic approaches to measurement. Research using this measure (known as the "Selves questionnaire") by Higgins and his colleagues (eg. Strauman and Higgins, 1988; Strauman, Vookles, Berenstein, Chaiken, & Higgins, 1991) has provided evidence for an association between particular types of discrepancy (eg. own/actual-own/ideal or own/actual-own/ought) and distinct affective outcomes (dejection-related emotions versus agitation-related emotions). Although not his primary focus, Higgins also related his measures to overall self esteem. His preliminary analyses suggested that own/actual-own/ideal discrepancies contributed to the prediction of esteem beyond what could be explained by own/actual responses alone. Because he did not actually estimate own/actual scores he used Anderson's (1968) norms of adjective likability to determine whether attributes listed for own/actual self were positive or negative, and defined own/actual self as either the number or percentage of negative attributes listed. The use of this crude "normative" approach to inferring own/actual self is ironic in a paradigm that is claimed to rely upon idiographic measures. It is perhaps not surprising that these own/actual self estimates were only modestly related to self-esteem and much less correlated with esteem than measures derived from well-constructed nomothetic scales (e.g., Marsh, 1986). The focus of Higgins' research has been to expand and explicate processes that are part of a self-system rather than to develop psychometrically sound measurement procedures. A critical resultant limitation is a lack of evidence in support of the reliability and validity of his approaches. Aims of the Research The purpose of the present investigation is to expand the idiographic approach to self-concept measurement proposed by Higgins (1987) and to use multitrait-multimethod analysis (Marsh, 1988b) to compare it with responses to the SDQ III as well as other nomothetic discrepancy measures analogous to those Higgins examined. It should be noted that the intention Nomothetic and idiographic multifaceted self-concept p. 6 is not to pit "idiographic" against "nomothetic" methods to determine which is superior - an oversimplified and misleading endeavour at best, as Hermans (1988) points out. The aim is rather to embark on essentially "within construct" research (Marsh, 1990), to establish whether this idiographic measure displays convergent and discriminant validity when compared to the best nomothetic instrument available, and to examine whether a more "pure" idiographic measure than Higgins used is able to contribute to the prediction of general self-esteem beyond an (also more "pure") idiographic measure of global self-concept favourability. These "within construct" measurement issues should be seen as a logical prerequisite to the "between construct" research in which Higgins and his colleagues have been engaged. There were three aspects of Higgins' approach which were specifically targeted for refinement and evaluation: 1. Higgins specifically focused on a general self-concept by not restricting the content considered by subjects in selecting their attributes. An extension of this approach would be to ask subjects to select attributes within a specific context or domain (e.g., mathematics, sports/physical, social contact, verbal). In this way idiographic scores could be developed that were domain- specific. The aim of this expansion was to develop idiographic measures that parallel particular SDQ scales. It was predicted on the basis of extensive (predominantly nomothetic) research by Marsh (1990; 1992) and others (see Hattie, 1992), that these domain-specific idiographic measures would show greater convergent and discriminant validity than a completely open-ended, general measure when compared with nomothetic (SDQ) scales. 2. Higgins initially used external scorers to evaluate matching and mismatching attributes according to Roget's Thesaurus (Higgins et al., 1985). A later modification involved asking respondents to rate the extent to which each attribute applied to the particular standpoint/self combination under consideration (Higgins, 1987). This allowed scorers to distinguish between "true matches" and "mismatches of extent" where synonymous terms were rated as being "slightly" versus "extremely" applicable (eg., describing one's actual self as attractive (slightly) and one's ideal self as attractive (extremely)). Since Higgins stressed the need for open-endedness in eliciting attributes, a phenomenologically more "pure" measure of the discrepancy would appear to be provided by the subject's own judgement of whether attributes in the different lists represent matches or mismatches according to the meaning intended by their descriptions. Consequently, in the present investigation each subject was asked to evaluate the matches and mismatches among their lists. 3. An important limitation in Higgins' own/actual (general) self-esteem score was the crude normative basis of determining the desirability of each attribute. A potentially superior Nomothetic and idiographic multifaceted self-concept p. 7 (and again more consistently idiographic) approach adopted in this study is to simply ask the subjects themselves to evaluate the desirability of each attribute in their list. It was hypothesised, in contrast to Higgins' findings, that this new idiographic approach to infer actual self-desirability from own-actual self attributes would account for most of the relationship between esteem and own/actual-own/ideal discrepancy scores. More specifically, predictions were made that (a) idiographic actual self-concept desirability would be more highly correlated with general esteem than idiographic actual-ideal discrepancy; and (b) that the partial correlation between idiographic actual-ideal discrepancy and general esteem would not be significant when idiographic actual self- concept desirability was partialed out. These predictions could be seen as challenging either the idiographic measurement technique used by Higgins, or the whole discrepancy approach to self-concept, or both. They are actually aimed more specifically at replicating, with idiographic measures, a failure to support the discrepancy model as applied to general esteem or self concept which has been repeatedly exposed using nomothetic techniques (eg. Hoge & McCarthy, 1983; Marsh, 1993; Wylie, 1979). Whether this failure is due to methodological problems associated with the variety of measures employed to assess discrepancy (Hattie, 1992), or a weakness in the model itself, (or both), is not of central concern here. Rather, the intention is to raise questions related to each of these possibilities. In addition, and as a further rationale for the hypothesis, it may be argued that the more refined idiographic measure of self-concept desirability used here represents an alternative (idiographic) measure of global self-esteem - reflecting as it does an affective evaluation (desirability) of presumably important or relevant attributes. Thus, it is predicted that the two measures of general esteem will display sufficient convergent validity to preclude significant additional variance in the nomothetic scale from being predicted by actual-ideal discrepancy. Instrument Development and Administration Procedures The Macarthur Individualised Self (MISELF) Description Package The Macarthur Individualised Self (MISELF) Description Package was developed from a thorough review of the procedures used by Higgins and his colleagues (see Higgins, 1987 for a review), as well as taking into account the developments noted above. In addition to such idiographic measures, nomothetic rating items corresponding to each of the idiographic standpoint/self discrepancies (eg., actual-own/ideal) were developed. One general (completely open-ended) and four domain-specific idiographic measures were collected, while the nomothetic ratings were made for each of the 12 facets of self-concept measured by the SDQ III, excluding the general esteem scale. The SDQ III scales themselves were also incorporated into the package. Pilot administrations led to further refinements aimed at streamlining the required administration time and Nomothetic and idiographic multifaceted self-concept p. 8 enhancing the clarity of the directions. (More detail regarding the contents and administration of the MISELF package are provided in the procedures section below.) Participants Responses to the MISELF Package were collected from 48 subjects who were each offered $20 for their participation. One respondent did not complete the SDQ III component of the package, resulting in 47 cases being available for comparisons involving the SDQ III. Subjects were all first year students at UWS-Macarthur, predominantly enrolled in an undergraduate primary teaching degree program in the Faculty of Education. There were 39 females and 9 males in the sample, with ages ranging from 18 (the modal age) to 52. The mean age was 22; 56% of subjects were under 20 years old. Procedures The MISELF was administered idividually or in small group settings of 2-5 respondents, and took approximately two hours to complete. The package was designed so that the administration would, at the outset, closely approximate Higgins' open-ended paradigm: Subjects were first given general instructions including definitions of actual self ("your beliefs concerning the attributes you think you actually possess"); ideal self ("your beliefs concerning the attributes you would like ideally to possess; your ultimate goals for yourself"); and ought self ("your beliefs concerning the attributes you should or ought to possess; your obligations, duties, or normative rules or prescriptions for yourself"). They were then asked to "list up to 10 attributes of the type of person you actually are". On the next page, subjects were asked to "list up to 10 attributes of the type of person you think you would ideally like to be". Next, subjects were instructed to cover or fold back their "ideal" list, and to "list up to 10 attributes of the type of person you think you should or ought to be". While Higgins asked subjects to consider the standpoints of three significant others (eg., mother, father, closest friend) he only used the most important or relevant other person in calculating discrepancies. To streamline the procedure, therefore, subjects were asked to list attributes for only one significant other standpoint, by selecting the most important or relevant other person (initially regarding general self and then later for each of the four specific facets), and only listing the other/ideal and other/ought self attributes for that most relevant person. Having indicated their choice of most important or relevant other person, subjects were asked to once again cover or fold back their "own/ought" list of attributes, then "list up to 10 attributes of the type of person your chosen other person would ideally like you to be"; and "the type of person your chosen other person thinks you should or ought to be". Nomothetic and idiographic multifaceted self-concept p. 9 Subjects then repeated this attribute-listing process for four domain-specific areas of self concept (mathematics ability; sports and physical ability; social interaction and relationships; and reading and verbal ability). Next, they made the attribute comparisons (noting matches and mismatches) between their own/actual list and each of the other four lists, for each domain-specific area. It was emphasised that the basis of judgement was the respondent's own intended meaning, rather than any strict adherence to synonyms and antonyms. In many cases this encompassed an implicit rating of extent, as respondents occasionally noted that a pair of attributes that did not appear to represent a mismatch (eg., actual: average reader; ideal: fast reader) were actually opposites according to their own point of view. The final idiographic aspect of the package required subjects to evaluate the desirability or favourability of their own-actual attributes (again for the general self list and each of the four domain-specific areas) using a 9-point scale. Nomothetic ratings were then obtained for 12 summary items reflecting each domain-specific facet of self-concept from the SDQ: Using nine-point scales, subjects firstly rated their own/actual self-concept for each facet; then the importance of each facet to the subject; followed by ratings of the discrepancy between their own/actual facet ratings and the other relevant self-standpoint combinations (own/ideal, own/ought, other/ideal and other/ought) as included in the idiographic aspect of the package. Finally, subjects completed the SDQ III instrument. As mentioned above, the SDQ III measures 12 domain-specific facets: Physical appearance; physical ability; mathematics ability; same-sex relations; opposite-sex relations; parent relations; verbal ability; general academic; problem solving; religion/spirituality; honesty; emotional stability; as well as a general self-esteem scale. Outcomes of the Investigation A rich data base has been provided as a result of the unique combination of open-ended and nomothetic measures, and in focusing on convergent and divergent validity comparisons, only a sample of the potential outcomes can be presented here. A more comprehensive analysis, involving an increased sample size, consideration of key demographic influences (particularly gender), and more powerful statistical techniques, is in progress (Roche and Marsh, 1993). Idiographic discrepancy scores were computed by subtracting the number of matches from the number of mismatches reported for comparisons between each subject's own/actual list and each of the other four lists (own/ideal, own/ought, other/ideal, other/ought standpoint/self combinations), for each domain-specific area. A modified score was computed taking into account the number of attributes in each list (which increases the number of comparisons available to a subject), however this score was very highly correlated with the original raw score, and so the results presented here use Nomothetic and idiographic multifaceted self-concept p. 10 the raw scores, consistent with Higgins (1987; Strauman et. al., 1991). Relations Among Idiographic Discrepancy Measures To answer the question of whether the general and domain- specific open-ended measures display good convergent and divergent validity within the idiographic discrepancy paradigm based on Higgins (1987), a multitrait-multimethod matrix was constructed, treating the five facets (general, mathematics, sports/physical ability, relationships and verbal) as traits and the four standpoint/self open-ended discrepancy measures as methods (see Table 1a). Higgins does not claim that the various standpoint/self discrepancies are alternative measures of the same construct, but does report significant and substantial correlations between them, which, as he points out, should be expected since they each have the "actual" self in common. It would not, however, be appropriate to consider this analysis to be a critical test of Higgins' approach, since two "methods" may fail to correlate significantly because of actual differences in the underlying standpoint/self discrepancies rather than, or in addition to, any weaknesses in the measurement procedure. The issue being explored here is the relative convergence and distinctiveness exhibited by the general and domain-specific discrepancies when measured idiographically. The criterion for convergent validity is that the correlations between matching traits (facets) measured using the various methods (standpoint/self combinations) should be substantial. These correlations (shown in bold in Table 1a) vary from .10 to .86 (mean r =.62). Of the 30 convergent validities, 28 are significant at the .05 level (r > .28). The two nonsignificant convergent validities both relate to other/ought measures of the general discrepancy, as does one of the lowest significant validities. Thus the other/ought measure of the general discrepancy has largely failed to converge with (or relate to) the other measures of general discrepancy used by Higgins. For the other measures and traits, however, the criterion of convergent validity appears to be reasonably satisfied (see Table 1b for a summary). The first criterion of discriminant validity is that the convergent validities (mean r =.62) should be higher than correlations between nonmatching measures (mean comparison r = .25) in the same row and column of the rectangular submatrix (different facets - different methods correlations). This criterion was satisfied for 225 of 240 comparisons overall, with the failures all relating to comparisons involving the other/ought method with each of the three other methods (Table 1b). The general discrepancy displayed the weakest discriminant validity. This may be partly "justified" in that, as a "global" facet, it may be expected to correlate more highly with discreet "sub-facets" than such facets correlate among themselves. Such an argument does not, however, account for the relatively lower convergent Nomothetic and idiographic multifaceted self-concept p. 11 validities (mean r = .41) associated with these general discrepancy measures. It should be noted that Higgins, Klein, and Strauman (1985) reported somewhat higher correlations among their (general) standpoint/self measures (from r =.53 to r =.76; mean r= .65; p<.001 for 52 cases), which are comparable to the convergent validities for the more specific facets reported here (mean r=.66). These high intercorrelations, in the absence of other criteria for comparison (such as discriminant validation, do not constitute strong evidence for valid measurement. Thus, while some aspect of the method used for this study may have resulted in a poorer measure of self-discrepancy (at least regarding the other/ought perspective), there are several alternative explanations. Moreover, the methods used here have been successful for the domain-specific measures, suggesting perhaps that the general measure is less robust, or that the domain-specific measures would display even greater validity if Higgins' supposedly superior measurement conditions had been applied. Finally, these "higher" correlations should not be overemphasised. Only one of the six coefficients reported by Higgins -- the other/ought- own/ideal correlation (r =.59), was significantly different (at the .05 level) to the corresponding correlation in this study (r=.10; z=2.81; p<.01). The possible implications of this discrepant result will be discussed in more detail later. The second criterion of discriminant validity is that the convergent validities (mean r =.62) should be higher than correlations among the different traits for the same methods involved in each convergent validity (mean comparison r =.25). This criterion was satisfied for 556 of 600 comparisons overall, with the failures relating once again to comparisons involving the other/ought method with each of the three other methods, particularly the own/ideal measure (Table 1b). This is perhaps not surprising, considering that these two methods appear to represent "maximally different" measures (relative to this matrix), ie., different selves from different standpoints. While the own/ought-other/ideal comparison also involves this double distinction, it appears reasonable to argue that the "ought" self, defined as "...your obligations, duties, or normative rules or prescriptions" is more closely associated with what is "expected", particularly from important others (even when considering one's own standpoint). In summary, there is reasonable support for the convergent and divergent validity of these specific facets of self-concept discrepancy using the MISELF package. Of particular interest is the overall improvement in convergent validity when specific contexts are delimited (mean r = .66 for the domain- specific facets, compared to mean r = .41 for the general, most open-ended discrepancy). Relations Between Idiographic and Nomothetic Measures While Higgins clearly saw value in distinguishing between the different "selves" considered above, he appeared to consider Nomothetic and idiographic multifaceted self-concept p. 12 the own/actual-own/ideal discrepancy to be the most closely aligned with notions of general self concept and self-esteem (see Higgins, 1987 for a historical and theoretical justification). Consistent with this view, the actual-ideal discrepancy was the most highly correlated of the idiographic discrepancy scores (r =.25) with the general self scale from the SDQ III. In order to investigate the degree of convergence and divergence between idiographic and nomothetic measures of both actual-ideal discrepancies and own/actual self-concept, a second multitrait-multimethod matrix was constructed, comparing four methods (two actual-ideal discrepancy measures (idiographic and nomothetic); and two own/actual self-concept measures (idiographic and nomothetic)) by five self concept facets as traits (see Table 2a). The idiographic discrepancy score is, as described above, the number of mismatches between the "actual self" and "ideal self" attribute list, less the number of matches between the same two lists. The idiographic self-concept score is the mean desirability rating (on a 9-point scale) given to the attributes in the "actual self" attribute list. The nomothetic discrepancy scores are ratings of comparison between a subject's "own actual rating" and "your own ideal" regarding each facet. The nomothetic self concept scores consist of the mean response to the 10 or 12 items in each SDQ scale. This analysis is somewhat complicated by the fact that the MISELF package did not include a perfectly consistent set of idiographic and nomothetic discrepancy measures. In relation to the completely open-ended (general) idiographic measure, the SDQ III contains a general self-esteem scale, however there was no directly comparable single nomothetic discrepancy rating item. One solution to this issue is to use a total or overall nomothetic discrepancy score consisting of the mean of each subject's specific discrepancy scores across all 12 SDQ facets. Correlations based on this total measure are included in Table 2a, in place of a single general discrepancy item. As can be seen from the correlation matrix, the total score tends to exhibit relatively high "convergent" validities (bold type in Table 2a) at the expense of inflated comparison correlations (since the scores being compared have been included in calculating the overall measure), thus having a detrimental effect on divergent validity. This does pose a problem for the interpretation of the results based on the submatrices affected. However, this measure was conceptually closest to a general item, and, as noted in relation to the previous analysis, such "general" scores may be expected to correlate more highly with the specific facets which it presumably subsumes. Furthermore, empirically, the total self-concept score from the SDQ III (on which these nomothetic discrepancy items are based) usually relates more closely to the general esteem scale than any specific scale (Marsh, 1990; 1992). Another disparity in matching measures relates to the idiographic "social interaction and relationships" measure, which potentially encompasses the three SDQ scales relating to Nomothetic and idiographic multifaceted self-concept p. 13 relationships (opposite-sex, same-sex, and parent relations). Extensive research with the SDQ III (Marsh, 1990; 1992) has found that these scales measure relatively distinct (but correlated) components of self concept. Thus, the effect of combining these facets into a more general "social interaction and relationships" context for the idiographic measure is to produce a facet that is certainly more specific than the general, context-free approach, but presumably less specific or well-defined than the other facets which more closely reflect particular, empirically distinct SDQ III facets. Perhaps not surprisingly, given the wording of the idiographic prompt, an earlier MTMM analysis (not presented here) revealed that the parent relations scale did not appear to correlate significantly with any of the idiographic relationship discrepancy measures, whereas there was clearly some convergence with same- and opposite-sex relations. Nomothetic measures for both opposite-sex and same-sex relations are therefore included in the matrix (Table 2a), with a weighting of 0.5 applied to them in all comparisons. Again, it is emphasised that this composite of two relatively distinct scales effectively reduces the specificity of this facet, but maximises the congruence (both conceptually and empirically) between the idiographic and nomothetic measures. The convergent validities (correlations between matching facets measured by the various methods) were again significant and substantial, though generally more moderate than those in Table 1. These correlations vary (in absolute terms) from .25 to .77 (mean r =.54; see Tables 2a & 2b). Of the 30 convergent validities, 29 are significant at the .05 level (r > .28). The physical abilities (mean r =.66) and mathematics (r =.59) facets showed the highest mean convergent validities, while the "most convergent" methods were the idiographic actual self-desirability ratings related to the nomothetic actual self SDQ scales (mean convergent r =.63). The lowest convergent validities were associated with the general (mean r =.47) and the relationships (mean r =.44) facets, the two least specific facets. The only nonsignificant convergent validity (r=-.25) related to the comparison between idiographic general actual-ideal discrepancy and the nomothetic general self-esteem measure. Again, these are "maximally different" measures, and such an outcome may seem unsurprising, except that Higgins' own research suggests that substantially higher correlations between such discrepancy scores and measures of general esteem could have been expected. Higgins (1987) does not report zero-order correlations between actual-ideal discrepancy and general esteem. However, he reports a partial correlation of r(38)=-.47 (with the common association to the rather weak measure of "global self-concept negativity" partialed out). While the zero-order correlation may be expected to be somewhat higher, it is important to keep in mind that with such small samples (41 in the study to which Higgins (1987) refers), differences between coefficients in the two studies need to be considerably larger than this to reach statistical significance. It is possible, however, that the more purely Nomothetic and idiographic multifaceted self-concept p. 14 phenomenological approach to evaluating matches and mismatches (self-scoring by subjects) used in this study produces less reliable results, at least for the "general" self, than using external scorers to evaluate matches and mismatches. Alternatively, as suggested above, the general measures may simply be less robust or less distinct than the domain- specific measures. This issue will be pursued in the discussion below. Comparing the convergent validities (mean r =.54) with correlations between nonmatching measures in the same row and column of the rectangular submatrix (mean comparison r =.21) indicates that the first criterion of discriminant validity is consistently met for at least 95 percent of comparisons across all self-concept facets and between all methods (236 out of 240 comparisons overall; see Table 2b). The second criterion of discriminant validity was satisfied for 87 percent of comparisons overall (524 out of 600). Almost half of the failures relate to comparisons between the idiographic and nomothetic measures of actual-ideal discrepancy (see Table 2b). In particular, the "relationships" and "verbal" facets, exhibiting the lowest convergence between these two measures, also displayed the least discrimination with regard to different facets measured using the same methods, each achieving only 10 out of 20 successful comparisons. While the presumably inflated comparison coefficients in the nomothetic discrepancy triangular submatrix would have adversely affected this result, the convergent validities are so low as to be almost equally indistinct from coefficients in the other, "non- inflated" idiographic discrepancy triangular submatrix. The divergent validity shown by the idiographic discrepancy measure in relation to the nomothetic actual self measure was also relatively modest (84 successful comparisons out of 100). For the comparison of these methods (which both included bona fide general measures), the weakest traits were the two least specific facets -- the general and the "relationships" facets, with 13 and 14 successful comparisons (out of 20) respectively. Similarly, divergent validities for the nomothetic discrepancy measure in relation to both the nomothetic and idiographic actual self measures were also modest (89 successful comparisons out of 100 for each). These two comparisons (which both involved the "total score" proxy for a general measure of nomothetic discrepancy) also identified the general and the "relationships" facets as possessing lower divergent validity, along with the verbal facet (with 85, 82.5, and 85 percent successful comparisons respectively). The comparison correlations for this criterion (correlations among the facets for each of the measures) were highest for the two measures of actual-ideal discrepancy (mean r =.26 for the idiographic; and r =.38 for the nomothetic method, incorporating the total score as a proxy for a general score). The use of a total score has apparently inflated the mean Nomothetic and idiographic multifaceted self-concept p. 15 correlation among facets for the nomothetic discrepancy, introducing what may be best described as a "within-method method effect", demonstrating a potential danger associated with the use of proxy variables (Pedhazur & Schmelkin, 1991). Overall, apart from the weaknesses identified above, there is good support (from Tables 2a and 2b) for the convergent and divergent validity of the specific facets of actual self- concept and self-concept discrepancy using both the idiographic and nomothetic measures available in the MISELF package. Most noteworthy is the increase in convergent validity associated with the more specific contexts (mean r = .59 for the most domain-specific facets -- mathematics ability, physical ability, and verbal ability; compared to a mean r of .46 for the general and relationships facets). This outcome provides strong endorsement for the use of more domain-specific open-ended measures, as well as for the procedures employed (including the self-evaluation of matches and mismatches by subjects). It is most unlikely that this result is an artifact of failing to provide a general item measuring nomothetic discrepancy, since the two lowest convergent validities for the general facet relate to comparisons not involving the proxy measure, in which other facets showed higher convergent validities. Where the proxy is used, the convergent validity improves both in an absolute sense and in relation to the other facets. This suggests that the proxy variable may, if anything, have in fact overestimated the convergent validity of the nomothetic general discrepancy. Another particularly significant result was the relatively strong convergent validity demonstrated by the idiographic desirability ratings and the nomothetic SDQ scales as measures of own/actual self-concept facets. The correlation of .55 between the general idiographic desirability ratings and the general esteem scale of the SDQ is substantially greater than the correlations of -.33 to -.35 found by Moretti and Higgins (1987; reported in Higgins, 1987) between crude normative estimates of attribute negativity and the Rosenberg (1965) Self-Esteem scale, on which the SDQ III general esteem scale is based. This outcome substantiates the claim that this more idiographically consistent measure is a better predictor of esteem than the normative measure used by Higgins. Furthermore, correlations between the more domain-specific idiographic ratings and matching SDQ III facets are higher again (ranging from .63 for mathematics to .77 for verbal), but lower for the relationships facet, which is a blend of two relatively distinct facets). This result again supports the hypothesis that the provision of domain-specific contexts for idiographic measures improves the validity of the construct. Assessing the Contribution of Idiographic Discrepancy Scores to General Esteem Table 3 presents zero-order correlations among the general and total measures of actual-ideal discrepancy and actual self- concept measured both idiographically and nomothetically. Nomothetic and idiographic multifaceted self-concept p. 16 Total scores for the idiographic measures were calculated by taking the mean of the four specific domains available for this study. Comparable (sub)total scores were computed for the nomothetic measures using only the mean of the five corresponding facets (ie. mathematics, physical, verbal, and same- and opposite-sex relations). These facets were based on single items for the nomothetic discrepancies, and SDQ III scale scores for the nomothetic actual self-concept. In addition, totals combining all 12 nomothetic facets were included for both actual self and actual-ideal discrepancy. The results show that the best single predictor of nomothetic general self is the nomothetic total self-concept score (r =.64), followed by idiographic general self-desirability (r =.55). In contrast, both the total and general idiographic discrepancy measures failed to correlate significantly with this nomothetic criterion of general esteem. Correlations between both nomothetic discrepancy scores (the subtotal comparable to the idiographic total measure, and the total which was used as a proxy for a nomothetic general measure) and esteem were both significant, but again lower than the corresponding nomothetic subtotal and total actual scores. This result reconfirms the common finding that actual self scores are more highly correlated with a nomothetic measure of general esteem. Partial correlations were computed in order to test whether the idiographic actual-ideal discrepancies as measured in this study contribute to the prediction of nomothetic general self- esteem beyond general own/actual desirability, as claimed by Higgins (1987), on the basis of his less authentically idiographic measures. In contrast to Higgins, but in line with the hypothesis presented earlier, the partial correlation between actual-ideal discrepancy and general esteem with idiographic desirability partialed out was zero. When the idiographic actual/ideal discrepancies for the specific facets were entered into a regression equation along with the general discrepancy and desirability to predict (nomothetic) general esteem, there was no significant improvement in R-square. There was, however, an almost significant positive partial correlation between verbal actual/ideal discrepancy and esteem with the other variables partialed out (partial r(40) = .27, p<.1). While this correlation is in the opposite direction to that predicted by the discrepancy model, and may not represent a robust finding, it is not without similar precedent in the literature (Marsh, 1993). It is conceivable that discrepancies between self and ideal relating to certain facets do not impact negatively on self-esteem (as is generally assumed under the discrepancy model), but rather show a positive relation as a result of a reciprocal effect of actual-self perceptions on ideals. While the results obtained here certainly do not constitute support for this speculation, the possibility does once again highlight the importance of taking a multifaceted approach to the study of self-concept. (Further regression analyses using this data have been conducted, but are beyond the scope of Nomothetic and idiographic multifaceted self-concept p. 17 this paper and will be presented elsewhere, see Roche and Marsh, 1993, in preparation). It may be argued that the more refined idiographic measure of self-concept desirability used here represents an alternative (idiographic) criterion for global self-esteem - reflecting as it does an affective evaluation (desirability) of presumably important or relevant attributes. As predicted and reported above, the two measures of general esteem display sufficient convergent validity to preclude significant additional variance in the nomothetic scale from being predicted by idiographic actual-ideal discrepancy. Referring again to Table 3, it can be seen that once again, the best single predictor of idiographic general self-desirability is the nomothetic total self-concept score (r =.60), followed by nomothetic general self (r =.55). In contrast to the findings above using the nomothetic criterion, however, both the total and general idiographic discrepancy measures do correlate significantly with this idiographic criterion (r =.34 & r =.48 respectively). Indeed, the general discrepancy score correlates slightly more highly with idiographic general self than does the total idiographic actual self score (r =.45). Both nomothetic discrepancy scores (subtotal and total) exhibit correlations with this idiographic measure of esteem which are quite comparable to the corresponding relations between this criterion and the nomothetic subtotal and total actual scores. Finally, when nomothetic esteem is partialed out, the general idiographic discrepancy retains a positive partial correlation with the idiographic general self criterion (partial r(45)=-.40, p<.01. This result appears, in contrast to the above findings, to at least offer some support for the use of general discrepancy scores. As anticipated, these apparently contradictory findings have raised several questions about both the general discrepancy model and the methods used to evaluate it. A resolution to these questions is outside the scope of this paper, but it can be concluded, in summary, that the refined idiographic measure of general self used here accounts entirely for the correlation between idiographic actual-ideal discrepancy and nomothetic esteem. As a criterion of esteem, idiographic general self can be predicted virtually equally as well by nomothetic discrepancy scores or nomothetic actual self scores, and only marginally less well by idiographic discrepancy scores. Summary and Implications This investigation aimed to expand the idiographic approach to self-concept measurement proposed by Higgins (1987) and to use multitrait-multimethod analysis (Marsh, 1988b) to establish whether such an idiographic approach would exhibit convergent and discriminant validity when compared with responses to the SDQ III as well as other nomothetic discrepancy ratings specifically developed to match the idiographic measures. The Nomothetic and idiographic multifaceted self-concept p. 18 Macarthur Individualised Self (MISELF) Description Package was primarily developed to refine and evaluate Higgins' paradigm with respect to three basic issues discussed below: Firstly, to investigate the validity of domain-specific compared to "context-free" or completely open-ended idiographic measures; secondly, to maintain the integrity of the idiographic paradigm by allowing subjects to evaluate their own self- descriptive attribute lists for "matches" and "mismatches" which are used to compute the idiographic discrepancy scores; and finally to evaluate the contribution of the idiographic self-ideal discrepancy score to the prediction of general self-esteem after partialing out an idiographic measure of actual self-concept desirability. In relation to the first major goal, the findings from the first multitrait-multimethod analysis comparing the four idiographic standpoint/self measures provide reasonable support for the convergent and divergent validity of both the general and specific facets of self-concept discrepancy as measured by the different standpoint/self combinations. In line with the prediction, the more domain-specific measures demonstrated stronger validity. As mentioned earlier, while a "general" facet may be anticipated to correlate more highly with discreet "sub-facets" than such facets correlate among themselves (producing lower discriminant validity), this cannot account for the relatively lower convergent validities associated with these general discrepancy measures. In contrast, the generally higher correlations among standpoint/self measures reported by Higgins, et al., (1985) may suggest that some aspect of the method used for this study has resulted in a poorer measure of (at least the general) self-discrepancy. However, in the absence of other criteria for comparison, higher intercorrelations do not constitute good evidence for valid measurement. In fact, refinements to the scoring of matches and mismatches made by Higgins and his colleagues appear, if anything, to have reduced these intercorrelations slightly (eg Strauman et al., 1981). Moreover, the methods used here have yielded strong support for the domain-specific measures, suggesting that the general measure may be less robust (in relation, for example, to time constraints, or the method of allowing subjects to self- evaluate matches and mismatches), or that the domain-specific measures would display even greater validity under Higgins' measurement conditions. These possibilities are not mutually exclusive, and both seem to be plausible when consideration is given to the cue restrictions imposed by completely open-ended formats, as mentioned in the introduction. It should be noted, in addition, that this apparent failure to adequately reflect correlations obtained by Higgins' general discrepancy approach applies only to certain standpoint/self comparisons, in particular, those involving the other/ought perspective, and that only one of these correlations was significantly different (at the .05 level) to the result obtained here. Generally good support was also found for the convergent and divergent validity of the specific facets of actual self- Nomothetic and idiographic multifaceted self-concept p. 19 concept and actual-ideal self-discrepancy using both the idiographic and nomothetic measures available in the MISELF package. Potential problems associated with using a total score in place of a general nomothetic discrepancy measure were discussed in the results section. While this strategy clearly imposes limitations on the interpretability of the results regarding comparisons with this measure, one result is to produce a considerably more stringent test of divergent validity, which was particularly well satisfied by two of the most specific (mathematics and physical ability) facets. The other three method comparisons are free of this complication, and again, the superior validity exhibited by the more specific facets would appear to strengthen the claim that greater domain-specificity in open-ended measures would improve their validity in self-concept research. Concerning the second issue, regarding the adoption of the "subject self-evaluation" approach to attribute match/mismatch estimation, the results discussed above also appear to vindicate this procedure. Apart from the uncertainty regarding correlations involving the other/ought discrepancy measure, allowing subjects to evaluate their own self- descriptive attribute lists for "matches" and "mismatches" produced scores that demonstrated good convergent and divergent validity in relation to each other and, in the case of actual-ideal discrepancies, in relation to corresponding nomothetic measures including scales from the SDQ III (which exhibits very strong construct validity) and nomothetic discrepancy ratings derived from the SDQ III scales. Subject self-evaluation would also seem to be more appropriate to maintain the integrity of the idiographic paradigm being used. If the attributes listed by a person are more salient, or have greater meaning than a checklist or fixed questionnaire containing items to be rated (as Higgins (1987) rather convincingly argues), then it would appear to be most legitimate to allow the person to use this greater personal meaning, including an appreciation of both the quality and extent of each attribute's applicability, in evaluating whether such attributes represent matches, mismatches or nonmatches. While self-evaluation therefore appears to be quite defensible, to properly assess the relative efficacy of having subjects evaluate their own self-descriptive attribute lists for "matches" and "mismatches" would require a more direct comparison with Higgins' (original and modified) scoring methods. Since no ratings of the extent to which attributes applied are available, it is not possible to calculate the modified scores for the present data (although some approximation may be possible, as subjects were encouraged to use words or phrases, which often gave an indication of extent). Scoring according to the original method (using only the Thesaurus to evaluate synonyms or antonyms) is currently being undertaken, and may be presented in a related study (Roche and Marsh, 1993, in preparation). Nomothetic and idiographic multifaceted self-concept p. 20 Finally, the results supported the prediction regarding the third issue, that the idiographic actual-ideal discrepancy score would not contribute to the prediction of nomothetic general self-esteem after partialing out idiographic general self-desirability. As emphasised in making and confirming this prediction, this raises methodological and theoretical questions regarding the use of discrepancy scores and the testing of the discrepancy model, but does not decisively discredit either. Indeed, using the idiographic general self- desirability as the criterion of general esteem provided an interesting contrast in results. Discrepancy scores became roughly equivalent to comparable actual self scores as single predictors of esteem. While this finding requires considerable further corroboration and expansion, it at least appears to offer some support for the use of discrepancy scores. There are several limitations which need to be emphasised in relation to the above summation of results. Firstly, given the limited size and dubious representativeness of the sample, it would be premature to treat the reported correlations as robust indicators of the relative validity of the various measures. Thus, while the pattern of correlations suggest, for example, that certain domain-specific idiographic measures may show greater convergent and discriminant validity than a general idiographic measure, the results are suggestive, but certainly not conclusive. Other problems relate to the testing session itself. Firstly, there is the issue of the time constraint placed on respondents. While a majority of respondents were able to list between 8 and 10 attributes for most lists in the maximum allowable time, there were clear differences in the rate at which this task was accomplished. Presumably, allowing more time would have enabled more attributes to be listed by the slowest subjects, offering potential improvements in the validity of the measure. It should also be noted that there were differences in the apparent difficulty of completing the various lists. Secondly, the single order of administration may have influenced the results. This applies most critically to the order of listing attributes from different standpoint/self combinations (with own/actual always first, through to other/ought last); to the order of repeating this process for different facets (general first, verbal last); and to repeating that order for assessing the matches and mismatches, then repeating it again to assess the desirability of each actual attribute. The subsequent nomothetic measures are also subject to influence from this particular order, but presumably considerably less so than the idiographic measures would have been had they followed such "context-setting" nomothetic items. The order used was considered appropriate to minimise the contamination of each kind of measure with previous measures, although to properly achieve and assess this would require several separate testing sessions and a number of different administration orderings. Apart from overcoming these problems, future research using idiographic measures of self-concept and self-discrepancies Nomothetic and idiographic multifaceted self-concept p. 21 should focus on evaluating the contribution of specific domains, as well as specific standpoint/self discrepancies, to the prediction of esteem measured from both an idiographic and nomothetic perspective, as well as their relations to other validational criteria. Such "within construct" research should preferably be conducted as a prerequisite to research relating these measures to other constructs. Questions raised by the results presented here highlight the need to evaluate more closely the nature of discrepancies and their presumed unidirectional influence on self-esteem. Higgins himself has expanded his "list of selves" to include the "unfulfilled positive potential", incorporating notions of the "can" and the "future" selves (cf. Markus and Nurius, 1986). Merging a multifaceted (domain-specific) approach into this research should facilitate a more comprehensive investigation of these issues. The refinement of Higgins' paradigm represents an exciting advancement for the use of idiographic measures of self- concept. As this project apparently represents a unique attempt to measure distinct domains or components of self- concept using an idiographic framework, it is a particularly significant contribution to the field. The MISELF instrument has potential for use in a variety of settings, particularly educational institutions. This kind of package may have considerable implications for the measurement of self-concept and esteem across a variety of cultural contexts, since it allows individuals to define and describe the multiple domains of self in their own words, according to their own background and experiences. As such, it may represent a measure that is particularly sensitive to differences in gender, ethnicity and social class. These differences are obviously important, yet can be difficult to assess using traditional nomothetic questionnaires, which may even be constructed to minimise or cancel them out altogether. In what could be a parody of Marsh's (1990; 1992) dictum, that "self-concept cannot be adequately understood if its multidimensionality is ignored", Kenway and Willis (1990) assert that "self-esteem research ... cannot be adequately understood unless placed in the context of a particular theory, a particular culture and a particular time" (p. 237). The approach adopted with the MISELF allows this context to be largely defined by the respondent, within a "metatheoretical framework" that is relevant to the researcher. Nomothetic and idiographic multifaceted self-concept p. 22 References Allport, G. W. (1962). The general and the unique in psychological science. Journal of Personality, 30, 405- 422. Anderson, N. H. (1968). Likableness ratings of 555 personality-trait words. Journal of Personality and Social Psychology, 9, 272-279. Bugental, J., & Zelen, S. (1950). Investigations into the self-concept. The W. A. Y. technique. Journal of Personality, 18, 483-498. Coopersmith, S. (1967). The antecedents of self esteem. San Francisco: Freeman. Epstein, S. (1980). The self concept: A review and the proposal of an integrated theory of personality. In E. Staub (Ed.), Personality: Basic issues and current research (pp. 81-132). Englewood Cliffs, NJ: Prentice Hall. Eysenck, H. J. (1954). The science of Personality: Nomothetic! Psychological Review, 61, 339-342. Harter S. (1986). Processes underlying the construction, maintenance, and enhancement of the self-concept in children. In J. Suls & A. G. Greenwald (Eds.), Psychological perspectives on the self, vol. 3 (pp. 137- 181). Hillsdale, NJ: Lawrence Erlbaum. Harter S. (1993). Causes and consequences of low self esteem in children and adolescents. In R. F. Baumeister (Ed.), Self esteem: The puzzle of low self-regard (pp. 87-116). New York: Plenum Press. Hattie, J. (1992). Self concept. Hillsdale, NJ: Lawrence Erlbaum. Hermans, H. J. M. (1988). On the integration of nomothetic and idiographic research methods in the study of personal meaning. Journal of Personality, 56, 785-812. Higgins, E. T. (1987). Self-discrepancy: A theory relating self and affect. Psychological Review, 94, 319-340. Higgins, E. T. Klein, R. & Strauman, T. (1985). Self concept discrepancy theory: A psychological model for distinguishing among different aspects of depression and anxiety. Social Cognition, 3, 51-76. Hoge, D. R., & McCarthy, J. D. (1983). Issues of validity and reliability in the use of real-ideal discrepancy scores to measure self-regard. Journal of Personality and Social Psychology, 44, 1048-1055. Hoge, D. R., & McCarthy, J. D. (1984). Influence of individual and group identity salience in the global self- Nomothetic and idiographic multifaceted self-concept p. 23 esteem of youth. Journal of Personality and Social Psychology, 44, 1048-1055. James, W. (1890/1963). The principles of psychology. New York: Holt, Rinehart & Winston. Katz, P. & Zigler, E. (1967). Self-image disparity: A developmental approach. Journal of Personality and Social Psychology, 5, 186-195. Kelly, G. A. (1955). The psychology of personal constructs. New York: Norton. Kenway, J. & Willis, S. (1990). Conclusion. In J. Kenway & S. Willis (Eds.), Hearts and minds: Self-esteem and the schooling of girls. London: Falmer Press. Markus, H., & Nurius, P. (1986). Possible Selves. American Psychologist, 41, 954-969. Markus, H., & Wurf, E. (1987). The dynamic self-concept: A social psychological perspective. Annual Review of Psychology, 38, 299-337. Marsh H. W. (1986). Global self-esteem: Its relation to specific facets of self-concept and their importance. Journal of Personality and Social Psychology, 51, 1224- 1236. Marsh, H. W. (1988a). The Self Description Questionnaire (SDQ): A theoretical and empirical basis for the measurement of multiple dimensions of preadolescent self- concept: A test manual and a research monograph. San Antonio, TX: The Psychological Corporation. Marsh, H. W. (1988b). Multitrait-multimethod analyses. In J. P. Keeves (Ed.), Educational research methodology, measurement and evaluation: An international handbook (pp. 570-580). Oxford: Pergamon. Marsh, H. W. (1990). A multidimensional, hierarchical model of self concept: Theoretical and empirical justification. Educational Psychology Review, 2, 77-172. Marsh, H. W. (1992). Self Description Questionnaire III (SDQ III): Manual. Campbelltown, NSW: Faculty of Education, UWS-Macarthur. Marsh H. W. (1993). Relations between global and specific domains of self: The importance of individual importance, certainty and ideals. Journal of Personality and Social Psychology, 65, (in press). Paunonen, S. V., & Jackson, D. N. (1985). Idiographic measurement strategies for personality and prediction: Some unredeemed promissory notes. Psychological Review, 92, 486-511. Nomothetic and idiographic multifaceted self-concept p. 24 Pedhazur, E. J., & Schmelkin, L. (1991). Measurement, design, and analysis: An integrated approach. Hillsdale, NJ: Lawrence Erlbaum. Pelham, B. W. (1993). The idiographic nature of human personality: Examples of the idiographic self-concept. Journal of Personality and Social Psychology, 64, 665-677. Pelham, B. W., & Swann, W. B., Jr., (1989). From self- conceptions to self-worth: On the sources and structure of global self-esteem. Journal of Personality and Social Psychology, 57, 672-680. Roche, L. A., & Marsh, H. W. (1993, in progress). Forget a resolution to the idiographic-nomothetic debate - try a revolution: Combining the power of open-ended and structured rating-scale measures of self concept. Rosenberg, M. (1965). Society and the adolescent self-image. Princeton, NJ: Princeton University Press. Rosenberg, M. (1982). Psychological selectivity in self- esteem formation. In M. Rosenberg & H. B. Kaplan (Eds.), Social psychology of self concept. (pp535-546). Arlington Heights,IL: Harlan Davidson. Runyan, W. M. (1983). Idiographic goals and methods in the study of lives. Journal of Personality, 51, 413-437. Smith, I. D. (1992). Measuring student self concept. Paper presented at the Australian Association for Research in Education annual conference, Geelong. Shavelson, R. J., Hubner, J. J., & Stanton, G. C. (1976). Validation of construct interpretations. Review of Educational Research, 46, 407-441. Strauman, T. J., & Higgins, E. T. (1988). Self-discrepancies as predictors of vulnerability to distinct syndromes of chronic emotional distress. Journal of Personality, 56, 685-707. Strauman, T. J., Vookles, J., Berenstein, V., Chaiken, S., & Higgins, E. T. (1991). Self-discrepancies and vulnerability to body dissatisfaction and disordered eating. Journal of Personality and Social Psychology, 61, 946-956. Wells, L. E., & Marwell, G. (1976). Self-esteem: Its conceptualization and measurement. Beverly Hills: Sage. Wylie, R. C. (1974). The self concept (rev. ed., vol. 1). Lincoln: University of Nebraska Press. Wylie, R. C. (1979). The self concept (Vol. 2): Theory and research on selected topics. Lincoln: University of Nebraska Press. Nomothetic and idiographic multifaceted self-concept p. 25 Wylie, R. C. (1989). Measures of self concept. Lincoln: University of Nebraska Press. Table 1a: MTMM correlation matrix with four discrepancy measures as methods by five self concept facets astraits. Method: [Actual-Ideal Discrep-Idiographic] [Actual-Ought Discrep-Idiographic] [Other-Ideal Discrep- Idiographic] [Other-Ought Discrep- Idiographic] Traits: GAID-I MAID-I SAID-I RAID-I VAID-I GAOU-I MAOU-I SAOU-I RAOU-I VAOU-I GAOI-I MAOI-I SAOI-I RAOI-I VAOI- I GAOO-I MAOO-I SAOO-I RAOO-I VAOO-I GAID-I 1000 MAID-I 171 1000 SAID-I 196 602** 1000 RAID-I 360* 451** 370** 1000 VAID-I 047 103 067 252 1000 GAOU-I 573** 266 162 313* 245 1000 MAOU-I 253 546** 289* 256 -033 259 1000 SAOU-I 038 204 512** 223 -048 129 263 1000 RAOU-I 205 423** 246 819** 325* 222 037 106 1000 VAOU-I 000 153 185 410** 823** 257 -190 057 539** 1000 GAOI-I 687** 057 125 274 274 538** 095 007 212 194 1000 MAOI-I 150 769** 519** 526** 191 181 491** -002 422** 215 108 1000 SAOI-I 094 305* 565** 230 -244 068 215 675** 028 -138 150 158 1000 RAOI-I 320* 285* 233 708** 252 214 048 -036 704** 401** 325* 404** 165 1000 VAOI-I 120 190 348* 392** 721** 181 -224 139 533** 787** 285* 244 025 421** 1000 GAOO-I 097 216 250 331* 472** 268 -004 293* 499** 617** 314* 1967 -008 189 550** 1000 MAOO-I 254 643** 339* 351* -025 137 640** -124 251 -092 168 791** 119 309* -071 -156 1000 SAOO-I 228 288* 498** 182 -177 148 274 540** 016 -163 181 143 739** 108 -064 -178 321* 1000 RAOO-I 259 -143 -283 344* 376** 216 -106 -213 466** 345* 489** -014 -254 648** 325* 229 -005 -298* 1000 VAOO-I 069 221 188 299* 853** 191 -152 037 434** 833** 273 243 -081 381** 865** 536** -074 -133 335* 1000 Note: *=Signif. LE .05 **=Signif. LE .01 (2-tailed). Coefficients in bold represent convergent validities. Decimal points omitted to save space. GAID-I = general actual-ideal discrepancy (idiographic); GAOU-I = general actual-ought discrepancy (idiographic); GAOI-I = general actual-other/ideal discrepancy (idiographic); GAOO-I = general actual-other/ought discrepancy (idiographic); MAID-I = mathematics actual-ideal discrepancy (idiographic); MAOU-I = mathematics actual-ought discrepancy(idiographic); MAOI-I = mathematics actual-other/ideal discrepancy (idiographic); MAOO-I = mathematics actual-other/ought discrepancy (idiographic); SAID-I = sports actual-ideal discrepancy (idiographic); SAOU-I = sports actual-ought discrepancy (idiographic); SAOI-I = sports actual-other/ideal discrepancy (idiographic); SAOO-I = sports actual-other/ought discrepancy (idiographic); RAID-I = relationship actual-ideal discrepancy (idiographic); RAOU-I = relationship actual-ought discrepancy (idiographic); RAOI-I = relationship actual-other/ideal discrepancy (idiographic); RAOO-I = relationship actual-other/ought discrepancy (idiographic); VAID-I = verbal actual-ideal discrepancy (idiographic); VAOU-I = verbal actual-ought discrepancy (idiographic); VAOI-I = verbal actual-other/ideal discrepancy (idiographic); VAOO-I = verbal actual-other/ought discrepancy (idiographic); Table 1b: Summary of MTMM analysis of correlation matrix in Table 1a. TRAITS SUMMARY METHODS SUMMARY TOTAL GAID-I MAID-I SAID-I RAID-I VAID-I ID-I/OU-I ID-I/OI-I ID-I/OO- I OU-I/OI-I OU-I/OO-I OI-I/OO-I Proportion successful Proportion successful Crit 1 (significant /6) .667 1.000 1.000 1.000 1.000 (significant /5) 1.000 1.000 .800 1.000 .800 1.000 .933 Crit 2a (successful /48) .750 1.000 1.000 .938 1.000 (significant /40) 1.000 1.000 .775 1.000 .900 .950 .938 Crit 2b (successful /120) .750 .983 .958 .942 1.000 (significant /100) .960 .990 .760 .980 .930 .940 .927 Mean correlations Mean correlations mean convergent validity .41 .65 .59 .61 .78 convergent validity .65 .69 .49 .64 .55 .67 .62 comparison r (Cr 2a) .21 .26 .25 .18 .22 .21 .25 comparison r (Cr 2b) .24 .25 .25 .22 .22 .23 .23 Table 2a: MTMM correlation matrix with four methods (two actual-ideal discrepancy measures (idiographic andnomothetic); and two own/actual self-concept measures (idiographic and nomothetic)) by five self concept facets as traits. Method: [Actual-Ideal Discrepancy-Idiographic] [ Actual-Ideal Discrepancy - Nomothetic ] [ Actual Self - Idiographic ] [ Actual Self - Nomothetic ] Traits: GAID-I MAID-I SAID-I RAID-I VAID-I AID-Nm MaAIDN PhAIDN OsAIDN SsAIDN VeAIDN GenSCI MatSCI SptSCI RelSCI VerSCI GenSCN MatSCN PhySCN OsRSCN SsRSCN VerSCN GAID-I 10000 MAID-I 1712 10000 SAID-I 1956 6023** 10000 RAID-I 3599* 4514** 3698** 10000 VAID-I 0465 1031 0674 2516 10000 AID-Nm 5264** 3526* 4158** 3529* 0535 10000 MaAIDN 3937** 5078** 2606 2287 -1293 6459** 10000 PhAIDN 2508 1665 5247** 1918 -1740 5283** 3476* 10000 OsAIDN 3649* -0468 0173 3335* -1564 5887** 2275 1635 10000 SsAIDN 3311* 2353 1515 2647 1319 5464** 1642 1873 3601* 10000 VeAIDN 0729 2312 2667 1123 3009* 5231** 2697 1808 3654* 3522* 10000 GenSCI -4758** -2348 -0382 -3356* -2927* -5430** -2285 0292 -4539** -3975** -3055* 10000 MatSCI -2301 -6876** -4998** -3410* -0533 -5033** -6003** -2797 -2268 -3387* -4008** 1855 10000 SptSCI -1192 -4874** -7546** -2350 -0028 -3778** -2017 -5887** -1909 -2667 -3596* 1141 4368** 10000 RelSCI -3103* -3457* -1239 -5758** -0623 -4284** -1768 0372 -5933** -3293* -3943** 5157** 2669 1854 10000 VerSCI -0568 0616 -0439 -0063 -5395** -3818** 0086 0143 -2841 -2206 -5448** 3941** -0101 1268 1881 10000 GenSCN -2516 -1668 -1787 -3123* 0122 -4994** -1627 -1644 -3433* -2008 -1081 5492** 1257 0875 3552* 2647 10000 MatSCN -2834 -5281** -2338 -2274 -1693 -4156** -5932** -0623 -0513 -3639* -2094 2379 6264** 1670 0943 0071 1685 10000 PhySCN -0061 -4162** -5910** -2463 0034 -3677* -3100* -7493** -1324 -1132 -2643 -0733 4271** 7578** 0453 0696 0839 1747 10000 OsRSCN -0893 -0044 -0960 -3103* 0027 -1563 1214 1143 -6221** 0188 -2335 2610 0038 1489 4236** 2866 3023* -1223 0707 10000 SsRSCN -2351 -2702 -2431 -3460* -1355 -4616** -0575 -1109 -4561** -5006** -4287** 4387** 3262* 3652* 4612** 3013* 4091** 2046 2366 5129** 10000 VerSCN -1941 -0236 -1867 -0435 -5564** -3730** -0362 0655 -1443 -2046 -5205** 3876** 01681403 2389 7725** 2148 1556 -0059 2270 4124** 10000 Note: *=p<.05; **=p<.01 (2-tailed). Coefficients in bold represent convergent validities. A mean nomothetic discrepancy score based on all 12 facets is included in place of a general nomothetic actual-ideal discrepancy score. Decimal points omitted to save space. GAID-I,MAID-I,SAID-I,RAID-I,VAID-I = general, maths, sports, relationships, & verbal actual-ideal discrepancy (idiographic); AID-Nm=mean nomothetic actual-ideal discrepancy; MaAIDN,PhAIDN,OsAIDN,SsAIDN,VeAIDN= Maths, Physical, Opposite-sex & Same-sex relations, & Verbal actual-ideal discrepancy (nomothetic); GENSCN MATSCN PHYSCN OSRSCN SSRSCN VERSCN= General, Mathematics, Physical, Opposite-sex & Same-sex relations, & Verbal self-concept (nomothetic (SDQ) scale) GENSCI,MATSCI,SPTSCI,RELSCI,VERSCI=Overall, General, Maths, Sports, Relationships, & Verbal self concept favourability (idiographic). Table 2b: Summary of MTMM analysis of correlation matrix in Table 2a. TRAITS SUMMARY METHODS SUMMARY TOTAL SUMMARY GENL MATH PHYS RELS VERB AI-I/AI-N AI-I/SC-I AI- I/SC-N AI-N/SC-I AI-N/SC-N SC-I/SC-N Proportion successful Proportion successful Cr 1 (significant /6) .833 1.000 1.000 1.000 1.000 (significant /5) 1.000 1.000 .800 1.000 1.000 1.000 .967 (29/30) Cr 2a (successful /48) .958 1.000 1.000 .958 1.000 (successful /40) .950 1.000 .950 1.000 1.000 1.000 .983 (236/240) Cr 2b (successful /120) .842 .933 .950 .792 .850 (successful /100) .670 .960 .840 .890 .890 .990 .873 (524/600) Mean correlations Mean correlations Mean convergent validity .474 .590 .661 .444 .539 Convergent validity .432 .606 .451 .548 .585 .629 .542 comparison r (Cr 2a) .219 .194 .162 .266 .220 .190 .208 Separate methods mean comparison r (Cr 2b) .262/.379 .262/.242 .262/.185 .379/.242 .379/.185 .185/.242 Combined methods mean comparison r (Cr 2b) (.321) (.252) (.224) (.311) (.282) (.214) .267 Note: Cr 1= Criterion 1 (Convergent validity); Cr 2a= Criterion 2a (Heterotrait-heteromethod discriminant validity); Cr 2b= Criterion 2b (Heterotrait-homomethod discriminant validity); AI-I, AI-N = Idiographic & Nomothetic Actual-ideal discrepancy; SC-I, SC-N = idiographic (mean attribute desirability), & nomothetic (SDQ scale) actual self-concept measures: Comparison between methods indicated by '/'; TOTAL = overall matrix results. Table 3: Correlations between idiographic and nomothetic measures of actual-ideal discrepancy and own/actualself-concept using general and total scores. T4AID(I) GAID-(I) T4AID(N) TAID-(N) T4SC-(I) GenSC(I) T4SC-(N) TotSC(N) GenSC(N) Discrepancy scores T4AID(I) 1.0000 GAID-(I) .2655 1.0000 T4AID(N) .3508* .4626** 1.0000 TAID-(N) .4012** .5264** .8911** 1.0000 Actual scores T4SC-(I) -.6652** -.2812 -.7121** -.6697** 1.0000 GenSC(I) -.3368* -.4758** -.3936** -.5430** .4512** 1.0000 T4SC-(N) -.5881** -.2670 -.6305** -.6134** .7793** .3879** 1.0000 TotSC(N) -.4872** -.3655* -.5342** -.6976** .6663** .5970** .8076** 1.0000 GenSC(N) -.2161 -.2516 -.3119* -.4994** .3155* .5492** .3806** .6403** 1.0000 Notes: *=p<.05; **=p<.01 (2-tailed). The general scores are based on a unitary (ie., non-domain specific) construct; Total scores are based on a mean of multiple facets. Where four facets are indicated, the score is based on the four facets investigated using the idiographic method, (or the five corresponding facets for the nomothetic measures). Coefficients in bold represent relations with the nomothetic [GenSC(N)] and idiographic [GenSC(I)] criteria of general esteem. T4AID(I)= Total (of 4 domain-specific facets) actual-ideal discrepancy (idiographic); GAID-(I)= general actual- ideal discrepancy (idiographic); T4AID(N)= Total (of 4 domain-specific facets) actual-ideal discrepancy (nomothetic); TAID-(N)= Total (of all 12 facets) actual-ideal discrepancy (nomothetic); T4SC-(I)= Total (of 4 domain-specific facets) actual self concept desirability (idiographic); GenSC(I)= General actual self-desirability or esteem (idiographic); T4SC-(N)= Total (of 4 domain-specific facets) actual self concept (nomothetic (SDQ)); ToSC-(N)= Total (of all 12 facets) actual self concept (nomothetic (SDQ)); GenSC(N)= General self-esteem (nomothetic), based on SDQ general self scale.