Congruence Between a Theoretical Continuum of
Masculinity and the Rasch Model:
Examining The Conformity to Masculine Norms Inventory
Larry H. Ludlow
Associate Professor
Lynch School of Education
Boston College
140 Commonwealth Avenue Chestnut Hill,
MA 02467
617-552-4221
email: Ludlow@bc.edu
and
James R. Mahalik
Associate Professor
Lynch School of Education
Boston College
140 Commonwealth Avenue
Chestnut Hill, MA 02467
617-552-4077
email: Mahalik@bc.edu
Journal of Applied Measurement, 2, 205-221, 2002
Abstract
The purpose of this study was to examine the psychometric structure of the Conformity to Masculine Norms Inventory (CMNI) in relation to the Rasch model. The CMNI was specifically constructed to measure a set of unidimensional constructs. As such, the items were intended to define a uniform spread of locations along each construct. The CMNI measures conformity to twelve masculine norms: winning, emotional control, risk-taking, violence, dominance, playboy, self-reliance, primacy of work, power over women, disdain for homosexuals, physical toughness, and pursuit of status. Three hundred forty-eight men participated in the study. In addition to examining global Rasch characteristics and the unidimensionality of each of the 12 scales, a detailed Rasch rating scale analysis is provided for the Violence Scale with unusual response patterns discussed in terms of their clinical usefulness. The results across all 12 scales reveal an excellent congruence between the theoretically derived construct of conformity to masculine norms and the theoretically defined objectives of the Rasch rating scale model.
Congruence Between a Theoretical Continuum of Masculinity and the Rasch Model: Examining The Conformity to Masculine Norms Inventory
Social norms are rules and standards that guide and constrain our social behavior (Cialdini & Trost, 1999). We learn which norms operate in a given social context through observing what most people do in a social situation (e.g., descriptive norms of not talking in an elevator), being told what is approved or disapproved behavior (e.g., injunctive norms when one is "shushed" to be quiet in the library), and observing how popular or admired people act (e.g., cohesive norms that lead teenagers to emulate rap stars). In response, persons often conform to social norms and experience the benefits of group acceptance and avoid negative group reactions, even though they may risk individuation. Other persons choose not to conform to social norms thus maintaining control over their lives but risking group rejection (Aronson, 1999; Asch, 1956; Cialdini & Trost, 1999).
Gender role norms, or those rules and standards that guide and constrain our behavior as men and women, are special types of social norms. Similar to how general social norms influence people to engage in specific social behavior, gender norms also operate when we observe what most men or women do in social situations, are told what is approved or disapproved behavior for men or women, and observe how popular or admired men or women act (Mahalik, 2000).
Just as persons conform or not to general social norms, so too do men and women also conform, or do not conform, to their respective gender role norms (Mahalik, 2000). When men and women conform to gender roles, they often experience group acceptance or avoid group rejection. However, persons who constrain themselves to conform to gender roles may violate their own internal convictions about how to act, lose a sense of themselves as an individual, or may engage in things that are not healthy as part of conforming to gender role norms (Cialdini & Trost, 1999). Likewise, men and women who choose not to conform to gender roles may experience greater authenticity and control in their lives but are likely to experience negative social consequences (Mahalik, 2000). Although violating gender norms has negative consequences for both genders, Pleck (1981, 1995) posits that the consequences are more severe for men than women.
Although it might seem, therefore, that conforming to masculine norms would be adaptive for males in that it leads to group acceptance, research indicates that the rigid enactment of masculinity is associated with a number of other problems. Specifically, these problems include lower self-esteem (Cournoyer & Mahalik, 1995; Davis, 1988), lower capacity for intimacy (Cournoyer & Mahalik, 1995; Sharpe & Heppner, 1991), higher anxiety and depression (Cournoyer & Mahalik, 1995; Davis, 1988; Good & Mintz, 1990; Sharpe & Heppner, 1991), abuse of alcohol (Blazina & Watkins, 1996), rigid interpersonal behavior (Mahalik, 2000), as well as higher levels of general psychological symptomology (Good et al., 1995; Hayes & Mahalik, 2000).
Although these findings are important, especially for mental health professionals who work with men, they must be balanced with the understanding that research on masculinity tends to suffer from measures that have weakly defined and poorly operationalized constructs. Specific to masculinity related constructs, pencil-and-paper measures tend to (a) estimate a global index of masculinity (e.g., the Bem Sex Role Inventory; Bem, 1974) overlooking the multiplicity of societal messages that constitute traditional masculinity in U.S. society, (b) assess traditional masculinity only (e.g., the Gender Role Conflict Scale; O'Neil et al., 1986) without assessing men's possible rejection of these messages, (c) are not be based on a theoretical model (e.g., Male Role Norms Scale; Thompson & Pleck, 1986), or (d) have poor psychometric properties including low internal consistency ratings and questionable construct validity.
To address these issues, the Conformity to Masculine Norms Inventory (CMNI) was developed to (a) examine conformity to 12 masculinity norms including winning, emotional control, risk-taking, violence, dominance, playboy, self-reliance, primacy of work, power over women, disdain for homosexuals, physical toughness, and pursuit of status, (b) estimate men's locations along a continuum ranging from extreme conformity to extreme non-conformity, and (c) reflect premises of a model of conformity to masculine norms (Mahalik, 2000).
Premises of the Conformity to Masculinity Model
Mahalik (2000) argues that there are many masculinity norms in U.S. society to which an individual man may, or may not conform. For example, for one man, masculinity norms about putting work first and being emotionally controlled may be salient, but others such as being violent and a playboy may not. For another man, masculinity norms about being a winner, physically tough and self-reliant may be salient, but others such as having power over women may not.
He also posits that a man's conformity to a given masculine norm ranges along a continuum of conformity from extreme conformity to extreme non-conformity. Thus, it is important to estimate the degree of conformity or non-conformity to understand how rigid or flexible an individual man is regarding a particular norm. That is, we would expect that the extreme conformer or non-conformer would maintain his position on the continuum in almost all situations. For example, if a man was an extreme conformer in response to the "Emotional Restriction" norm, we would expect him to continue that position even in places where emotional expression was acceptable (e.g., with a loved one who is being tender with him or a supportive group or therapeutic setting). Similarly, if a man was an extreme non-conformer in response to the "Violence" norm, we would expect him to continue in that position even in a situation where it was acceptable to conform to the norm of "Violence." For example, if the extreme non-conformer was drafted into military service, one would expect him to resist the draft or act as a conscientious objector.
Unlike the extreme positions on the conformity continuum, males who take moderate positions regarding masculinity norms are likely to be less rigid and more flexible in their conformity or non-conformity. For example, although the moderate conformer to the norm of "Emotional Restriction" may have some ambivalence toward expressing feelings, given a safe place to be emotional, he may do so in that context. Similarly, although the moderate non-conformer on "Violence" may tend to reject it, he may have some ambivalence about rejecting that norm if drafted or seeing the need for violent resistance to some injustice.
Finally, Mahalik (2000) argues that conformity along this continuum to any given norm is comprised of affective, behavioral, and cognitive components. For example, consider a young man playing on a sports team. He may work very hard to win but does not really believe that "winning is the only thing" nor does he feel ashamed or depressed when his team loses. In this case, the young man would be understood as conforming with behavioral norms around winning but not conforming with cognitive or affective norms around winning. However, another young man on his team works very hard at winning, believes "winning isn't everything, it's the only thing", and feels ashamed and depressed whenever he loses at competition. In this case, properly understanding an individual male's conformity to the masculine norm of winning entails understanding variability in behavioral, affective, and cognitive components of conformity to that norm.
Thus, developing a scale that assesses the affective, behavioral, and cognitive components of multiple masculinity norms along a range of conformity (i.e., from extreme non-conformity to extreme conformity) would operationalize these premises of Mahalik's (2000) conformity to masculine norms model.
The purpose of this paper is to explain how a Rasch modeling approach to inventory construction was used to define a continuum of masculinity messages from extreme conformity of the masculinity norm, to moderate conformity, to moderate non-conformity, and to extreme non-conformity. The Rasch approach that we refer to is one of first conceptually and theoretically defining the constructs to be measured, the writing of items to operationally define those constructs, and the gathering of empirical data to test the extent to which the scales function as intended (Rasch, 1960).
Method
Participants
Three hundred forty-eight men participated in the study. These men were recruited from college campuses in the northeast, mid-Atlantic, southeast and mid-west. They volunteered through classroom recruitment and psychology human subjects pools. Men averaged 19.84 years of age (SD=3.18), were mostly Caucasian (282 Caucasian, 12 African American, 27 Asian or Asian American, 11 Hispanic/Latino, 5 were bi-racial, and 10 described themselves as "Other"). Most men reported they were single (341 single, 3 married, 1 divorced, and 3 in committed same sex relationships). They were mostly heterosexual (335 heterosexual, 12 homosexual, and 1 bi-sexual). All were in college or graduate school.
Inventory Development
The construction of the CMNI is based on Mahalik's (2000) model of conformity to masculine gender norms. This model posits four conformity statuses each reflecting a different point on the continuum of conformity to non-conformity with regard to masculine norms, i.e., extreme conformity, moderate conformity, moderate non-conformity, and extreme non-conformity. All of these conformity statuses are believed to have benefits and costs. In general, the benefits of the two conformity statuses entail group acceptance whereas the individual costs involve personal restriction, particularly for extreme conformity. The benefits of the two non-conformity statuses entail experiencing personal authenticity whereas the costs involve group rejection, particularly for extreme non-conformity.
The development phase of this research consisted of the following. Two focus groups consisting of both men and women who were masters and doctoral students in counseling psychology met to discuss the conceptual aspects of masculine identity and its potential for measurement. They met every week for 90 minutes over an 8-month period to discuss and refine the categories of the norms, and to construct items to assess the continuum of conformity to the norm. Furthermore, the group members were given assignments to discuss these issues with friends and family. In subsequent meetings the issues and ideas around each norm were refined until there was consensus that it reflected a masculine norm in U.S. society.
Trial items were then written and submitted to college students in two intact classrooms who were requested to answer the items and critique the wording and scoring of the items. The CMNI was then piloted on a third set of college students and items were subsequently revised, discarded, left alone, or newly written before full-scale data collection with the finished instrument began.
Instructions to Respondents and Scoring
The CMNI is composed of 144 items answered on a 4-point forced-choice scale. Respondents are asked to respond to "how much do you agree with the following statements" by indicating SD for 'Strongly Disagree', D for 'Disagree', A for 'Agree' or SA for 'Strongly Agree'." Each item for all sub-scales assesses either an affective, behavioral, or cognitive component for each of the four conformity statuses (e.g., an affective-extreme conformity item). Thus, each of the twelve sub-scales contains 12 items.
The 144 items can be summed in different ways to examine various components of conformity to masculinity norms. For example, after reverse scoring the non-conformity items, the CMNI yields a total score reflecting the sum of all items on the inventory. In addition, the items can be scored to obtain overall affective, behavioral and cognitive conformity scores; overall scores for the conformity statuses of extreme conformity, conformity, non-conformity and extreme non-conformity; and summary scores for each of the twelve scales.
Analysis
A series of classical test theory (CTT) analyses were initially conducted with this sample. These analyses addressed the traditional concepts of internal consistency, test-retest reliability, construct validity, and concurrent validity (see Mahalik, Locke, Ludlow, Diemer, Scott, Gottfried & Freitas, 2000). In addition, and more relevant for the present purpose, a confirmatory Rasch rating scale analysis (Andrich, 1978; Wright & Masters, 1982) was performed on each scale. With the Rasch rating scale model, an individual's "degree of conformity" parameter ( ), a set of scoring category "threshold" parameters ( ), and an item's "level of conformity" parameter ( ) are estimated. These analyses were confirmatory to the extent to which the hypothesized continuum for each scale was actually achieved. That is, (a) each "variable map" (a graphical representation of persons and items located along a common continuum) was analyzed to determine the extent to which the empirical results for each scale corresponded to the theoretical structure proposed by the developers; (b) the spread and clustering of the items and men was inspected and interpreted; (c) the distribution of the scoring category step estimates was interpreted; (d) the distributions of the residuals were tested from a variety of perspectives; and (e) specific instances of person and item misfit were identified and interpreted. Finally, a clinical judgment regarding the utility of the continuum from extreme conformity to extreme non-conformity was provided. The analyses were run with SCALE (Masters, Wright, & Ludlow, 1981/2000).
Results
Classical test theory results
Four studies examined the CTT psychometric properties of the CMNI sub-scales (Mahalik et al., 2000). Study 1 yielded strong internal consistency estimates for all sub-scales and low to moderate correlations among the sub-scales. Studies 2 and 3 indicated that the sub-scales were significantly and positively correlated to other masculinity measures, psychological distress, social dominance and aggression, and significantly and negatively correlated to favorable attitudes toward psychological help seeking. Study 4 indicated that the CMNI scores had high positive test-retest estimates over a 2-3 week period.
Summary of Rasch model analyses on all scales
This section addresses the overall fit of the 12 CMNI scales to the Rasch model. Although space considerations prohibit a detailed analysis that would include, among other things, the variable map for each scale, it is possible to summarize some of the more familiar and typically reported results. A detailed analysis of the Violence scale is presented below.
Table 1 presents the mean measures and fit statistics for the men, their separation statistics, and the number of men with fit statistics greater than 3.0. The scales are ordered by the men's mean measures. That is, the negative measures indicate that these college men were least likely to conform to the masculine norms represented through the Playboy and Power scales. The Status and Winning scales, in contrast, had the highest mean measures-these college men gave their highest responses to these scales. Finally, the separation statistics suggest that the men may by grouped into at least three scoring levels on each of the 12 scales.
The number of men who misfit the scales was remarkably similar. This finding is partly attributable to the fact that many of the same men misfit multiple scales. For example, three men misfit six scales, one man misfit eight scales, and one man misfit nine scales and six of his nine fit statistics were greater than 5.0. This observation has obvious clinical consequences-who are these men and what is it about them that produced such unexpected but consistent responses?
Insert Tables 1 and 2 about here
Table 2 summarizes relevant statistics for the items on the 12 scales. The scales follow the same ordering as Table 1. Here we observe that the scales (a) possess an acceptable overall fit across their respective items, (b) all possess excellent separation in their item location estimates, and (c) all have excellent comparability in their threshold estimates. This latter finding is particularly gratifying since it indicates that the scoring categories are functioning as intended across all 12 scales.
One additional set of analyses was performed as a check on the unidimensional structure within each of the 12 scales (Ludlow, 1985, 1986; Mead, 1975). First, a person-by-item standardized residual matrix was generated from each Rasch analysis. Then each residual matrix was subjected to a principal component analysis. Each analysis extracted two-, three-, and four components. The first two components were extracted as a check on the hypothesis that the items within each scale separated into positively and negatively worded items. The three-component solutions served as checks on the hypothesis that items separated into cognitive, affective, and behavior items. The four-component solutions checked on the hypothesis that the items separated into extreme conformity, moderate conformity, moderate non-conformity, and extreme non-conformity items.
A summary of the principal component results for each scale is presented in Table 3. First, the determinants of the residual matrices are all non-zero and are larger than the determinants for the raw data matrices (e.g., on the Violence scale the residual matrix determinant was .03 compared to .015 for the raw data). Since all determinants are derived from correlation matrices they are directly comparable and because the determinant is a generalized variance estimate, these findings indicate greater unique variance, hence less shared variance, in the residuals.
Insert Table 3 about here
Second, the KMO statistics for the residual matrices are all in the "unacceptable" region according to Kaiser's criteria (1974). This is a positive result in the present situation since it indicates there is too little shared variation to be meaningfully partitioned through a principal component analysis. In contrast, the KMO for the Violence scale raw data was (.87).
Third, the magnitudes of the eigenvalues, the lack of a discernible break in the scree plots (not shown), and the magnitude of the percent of variance accounted for by each component are all consistent with results obtained when analyzing residuals from data simulated to fit the Rasch model (Ludlow, 1983). Finally, plots of the first two components for all solutions were consistent with the general circular patterns found for residuals simulated to fit the Rasch model (Ludlow, 1983). Given these results, there is no evidence in any of the solutions that the residuals for these items correspond to any of the hypothesized principal components. In other words, there is strong evidence that the 12 CMNI scales possess the critical and fundamental property of unidimensionality.
Rasch model results for the Violence scale
The Violence scale is presented in detail as it was typical in structure to the other eleven scales yet provoked some individual response patterns of interest for interpretation. Conformity to masculine norms around violence is operationalized as endorsement of items about liking to fight and observe violence, getting involved in violence, and believing that one is cowardly if one walks away from violence or that violence is sometimes necessary. Non-conformity is operationalized as endorsement of items about being uncomfortable or disgusted by violence, not engaging in violence, and believing that violence is rarely or never justified. The items are presented in Table 4. The original item names are used in order to be consistent with other publications of the CMNI.
Insert Table 4 about here
The variable map in Figure 1 reveals the operational definition of what it means to measure the endorsement of violence. At the top of the map are the men (X's) who scored highest on the scale. They gave the strongest endorsements of violence. They are the men most likely to say that they are "the first to start a fight." The items at the top of the map are the lowest scored ones, e.g. V26. Those items are the hardest to strongly endorse and they define the extreme conformity end of the scale. Note that the three "hardest" items all address the same aspect of violence that is personal and physical--fighting.
Insert Figure 1 about here
At the bottom of the map are the men with the lowest scores. They expressed the least level of endorsement for violence. Their location is below the highest scored items--those items that were the easiest to endorse in terms of some level of violence. Thus, these are the men who are least likely to endorse physical violence even if "violent action is necessary." The extreme non-conformity items tend to define this end of the scale. The moderate conformity and moderate non-conformity items tend to define the mid-range of the scale.
The person separation was 2.5 and the reliability of the person separation was .86 (Wright & Masters, 1982). The analogous item separation was 10.6 and the reliability of item separation was .99. These statistics indicate a wide spread to the men and the items. Such a spread is desirable because it provides support for (a) the extent to which the men are being measured at different levels along the scale, and (b) the extent to which the items are addressing a wide range of issues.
The mean estimate for the men was (-.78), indicating a relatively low level of endorsement of violence. In fact, there are men at the extreme lower end of the scale who are not likely to endorse any level of violence as described by these items. This suggests that additional items could be written to better define this end of the scale for these men. Likewise, there are a few men with extreme violence endorsement positions who are also not well measured by the particular items at the upper end of the scale.
The item locations represent the average score on an item. This means that a man located above an item, say V26 "start a fight", has a relatively high probability of either agreeing or strongly agreeing with V26 but it is not exactly clear which of those two response categories is the more likely. To find the most probable response on each item requires Table 5. In this table the items are ranked from the least level of violence endorsement (bottom of the figure) to the highest level of violence endorsement (top of the figure). The curving strings of 1's, 2's, and 3's correspond to the location where the most probable response of men scoring in one category to the next changes. The estimated locations of the men on the Violence scale are represented by the "Person Frequency Distribution" display.
Insert Table 5 about here
For the purpose of scale development the important information in this table is that the scoring category threshold estimates are well-separated from one another. This means that the scoring categories of SA, A, D, and SD are functioning as intended. That is, they are all eliciting relatively large numbers of responses from the men. In fact, each category is the most probable response to each item for someone along the continuum (except for V26 "start a fight"). The scoring category threshold estimates for the Violence scale are typical of all the scales: (-2.49, -.21, 2.69). These estimates, particularly the estimate corresponding to the transition from the third to the fourth scoring category (2.69), mean that it is relatively difficult for these men to strongly endorse these Violence items.
Table 6 is now of interest because it contains the fit statistics for the items. Different authors have preferences for different fit statistics. The standard ones reported in the Rasch literature are usually derived from the residuals (differences between the observed responses and the expected responses) and, as such, they tend to agree on which items or people seem to exhibit some aspect of "misfit" to the Rasch model. The present fit statistics are the standardized, weighted mean square residual t-statistics ("FIT": Wright & Masters, 1982, p. 101). They are affected by sample size and they do not have a well-known distribution (Smith, 1991), even though, theoretically, they were developed to approximate a unit normal distribution. Nevertheless, a good starting point is to look at the largest positive valued fit statistics. The negative valued fit statistics are not usually too interesting because even though they occur when the residual variation on an item is less than what is expected under the model, there is rarely an obvious explanation for why that happened-which is precisely the case with these data.
Insert Table 6 about here
Large positive valued t-statistics result from numerous unexpected responses having occurred. On one hand, this occurs when a man with a high estimate of violence endorsement gives a weaker level of endorsement than expected (e.g., gives an "agree" when a "strongly agree" was expected). Or, such a man has disagreed when expected to agree to an item. On the other hand, these statistics occur when a man with a lower level of violence endorsement gives an unexpectedly high level of endorsement to an item. These three types of unexpected responses led to items V97 (t = 3.07) and V45 (t = 4.36) being flagged as potentially misfitting the model.
These items are interesting. They are both in the middle of the person and item distribution continuum. V97 asks if you like to watch violent games. An analysis of the residuals on this item revealed that there were many men with low estimates who gave surprising "strongly agree" responses to this item. Apparently, some men who are not comfortable with personal physical violence do not experience much of a problem watching sport games that are inherently violent.
V45 asks if you are uncomfortable seeing violence. Here the residuals showed no consistent trend. Some men with low estimates gave higher than expected responses (they were not too uncomfortable watching violence) while other men with high estimates gave lower than expected responses (they did feel somewhat uncomfortable watching violence).
The information in Table 6 is presented three ways because fit statistics may exhibit patterns related to their order of presentation or magnitude of calibration. The "SERIAL ORDER" section shows no apparent start-up or end-of-instrument effect. There is also no apparent "CALIBRATION ORDER" effect. Furthermore, the "FIT ORDER" section indicates that the reverse coded items, as a group, are not problematic.
Table 7 reveals the response patterns for three individual men with fit statistics greater than 3.0. The table contains their raw summary score, computed strength of endorsement estimate (MEASURE), and fit statistic (FIT). In addition, the table contains their observed responses (the first row of numbers), expected responses (the second row), and standardized residuals (the third row). This table provides a level of analysis with considerable clinical utility. Now it is possible not only to measure a man's position on the conformity continuum but to also identify inconsistent response patterns.
Insert Table 7about here
Person 46 (FIT= 3.67), with a low level of violence endorsement (MEASURE = -1.41), was disgusted by any kind of violence ("strongly agreed" when "agreed" was expected-the seventh column in the rows of responses), yet enjoyed watching violent games ("strongly agreed" when "disagreed" was expected-the ninth column), and "strongly agreed" instead of "disagreed" with being willing to fight if necessary-the eleventh column. Person 109 (FIT= 3.42) had an even lower estimate (MEASURE = -2.11) yet "strongly disagreed" instead of "agreed" with being uncomfortable seeing violence-the sixth column. Speculatively, these response patterns fit a man who is strongly against violence in society, but may see sports as an acceptable context in which to act violently, e.g. really likes watching football.
Person 130 (FIT = 3.19), in contrast to the two others, had a high level of violence endorsement (MEASURE = 3.29) yet "strongly agreed" instead of "strongly disagreeing" with being uncomfortable seeing violence, and "disagreed" instead of "agreed" with liking to fight-the eighth column.
Discussion
The CMNI was developed as a tool for use by psychologists and counselors to examine masculinity issues with individual men. In terms of clinical uses, psychologists have advocated increasingly for integrating a gendered perspective into assessment and treatment with clients (e.g., Brown, 1986; Gilbert & Scher, 1999; Good, Gilbert, & Scher, 1990; Mahalik 1999a, 1999b). Thus, by using the CMNI a counselor may explore with a client how his position on any of the masculinity norms contributes stress to his relationships, work, and health. For example, it should benefit the clinical work to explore how a man's depression may be connected to a pattern of isolation brought about through conformity to masculine norms around self-reliance. Similarly, it might be useful to explore how a man's substance abuse may be affected by conformity to masculine norms around risk-taking.
The person-level Rasch analyses highlight an important aspect of the CMNI. Specifically, reviewing individual scores (however they are reported or estimated) shows that men who otherwise do not conform to the masculinity norm around violence may, in fact, like watching violence and may not necessarily be uncomfortable while watching it. As these two issues are both affective components of conformity to violence, this finding helps illustrate the potential clinical utility to such "surprising" responses. That is, part of the interpretation protocol when using the CMNI with individuals is to examine whether the affective, behavioral and cognitive components are at similar levels within the scale. For example, is an individual consistent or inconsistent in his attitudes, behaviors and cognitions about the masculinity norm of violence or other masculinity norms? The results from this analysis indicate that some undergraduate men in the sample who largely did not conform with the violence norm did conform affectively in the sense that watching violence was enjoyable and they were not uncomfortable about it.
In the clinical office, these types of apparent discrepancies among item responses are likely to be rich sources of exploration. That being said, the clinician should be interested in exploring with the client the effects of conflicting affective, behavioral and cognitive components of conformity on the various scales. For example, consider a man who has high scores on the behavioral items of "Primacy of Work" but low scores on the affective and cognitive items of that scale. For such an individual, despite his feelings that he does not like putting work first and his beliefs that work should not be the most important thing to him, he may feel and believe, nonetheless, that he must spend long hours on the job. Exploring this difference with him might help both clinician and client better understand the forces that contribute to stress for him. For example, does he feel he has to work long hours to avoid being fired or to live up to some ideal about being a good provider or that he feels he is worthwhile only when he is successful?
This opportunity for clinical exploration is a direct consequence of the finding that the CMNI scales are well fit by the Rasch model. That is, since the theory-based items define scales that are congruent with the objectives of a meaningful Rasch continuum, individual responses that deviate from those expected are likely to be of some importance in understanding the person. Thus, "misfitting responses" can serve as markers or flags that may be clinically useful for revealing internal conflicts, inconsistencies in behavior, and sources of personal stress and confusion over gender roles.
References
Andrich, D. (1978). A rating formulation for ordered response categories. Psychometrika, 43, 561-573.
Aronson, E. (1999). The social animal (8th edition). New York: Worth Publishers.
Asch, S. (1956). Studies of independence and conformity: A minority of one against a unanimous majority. Psychological Monographs, 70 (9, White No. 416).
Bem, S.L. (1974). The measurement of psychological androgyny. Journal of Consulting and Clinical Psychology, 42, 155-162.
Blazina, C., & Watkins, C.E., Jr. (1996). Masculine gender role conflict: Effect on men's psychological well-being, chemical substance usage, and attitudes toward help-seeking. Journal of Counseling Psychology, 43, 461-465.
Brown, L.S. (1986). Gender role analysis: A neglected component of psychological assessment. Psychotherapy, 23, 243-248.
Cialdini, R.B., & Trost, M.R. (1999). Social influence: social norms, conformity, and compliance. In D. Gilbert, S. Fiske, and G. Lindzy (Eds.) The handbook of social psychology, volume 2 (pp. 151 - 192). Boston: McGraw-Hill, Inc.
Cournoyer, R.J., & Mahalik, J.R. (1995). Cross-sectional gender role conflict examining college-aged and middle-aged men. Journal of Counseling Psychology, 42, 559-568.
Davis, F. (1988). Antecedents and consequences of gender role conflict: An empirical test of sex-role strain analysis. (Doctoral dissertation, Ohio State University). Dissertation Abstracts International, 48/11, 3443.
Gilbert, LA. & Scher, M. (1999). Gender and sex in counseling and psychotherapy. Needham Heights, MA: Allyn & Bacon.
Good, G.E., Gilbert, L.A., & Scher, M. (1990). Gender aware therapy: A synthesis of feminist therapy and knowledge about gender. Journal of Counseling Development, 68, 376-380.
Good, G.E., & Mintz, L.B. (1990). Gender role conflict and depression in college men: Evidence for compounded risk. Journal of Counseling and Development, 69, 17-21.
Good, G.E., Robertson, J.M., O'Neil, J.M., Fitz, L.F., DeBord, K.A., Stevens, M., Bartels & Braveman, D.G. (1995). Male gender role conflict: Psychometric properties and relations to distress. Journal of Counseling Psychology, 42, 3-10.
Hayes, J.A., & Mahalik, J.R. (2000). Gender role conflict and psychopathology in a clinical population. Psychology of Men and Masculinity, 1, 116-125.
Kaiser, H.F. (1974). An index of factorial simplicity. Psychometrika, 39, 31-36.
Ludlow, L.H. (1983). The analysis of Rasch model residuals. Unpublished doctoral dissertation. University of Chicago.
Ludlow, L.H. (1985). A strategy for the graphical representation of Rasch model residuals. Educational and Psychological Measurement, 45, 851-859.
Ludlow, L.H. (1986). Graphical analysis of item response theory residuals. Applied Psychological Measurement, 10, 217-229.
Mahalik, J.R. (1999a). Interpersonal psychotherapy with men who experience gender role conflict. Professional Psychology: Research and Practice, 30, 5-13.
Mahalik, J.R. (1999b). Incorporating a gender role strain perspective in assessing and treating men's cognitive distortions. Professional Psychology: Research and Practice, 30, 333-340.
Mahalik, J.R. (2000). A model of masculine gender role conformity. Symposium - Masculine gender role conformity: Examining theory, research, and practice. Annual meeting of the American Psychological Association, Washington D.C.
Mahalik, J.R., Locke, B., Diemer, M., Ludlow, L., Scott, R., Gottfried, M., & Freitas, G. (2000). Development of the Conformity to Masculine Gender Roles Scale. Symposium - Masculine gender role conformity: Examining theory, research, and practice. Annual meeting of the American Psychological Association, Washington D.C.
Masters, G.N., Wright,B.D. & Ludlow,L.H. SCALE: A Rasch program for Rating Scale Data. MESA Psychometric Laboratory, University of Chicago, 1981. (Expanded edition, Boston: Boston College, 2000).
Mead, R.J. (1975). Analysis of fit to the Rasch model. Unpublished doctoral dissertation. University of Chicago.
O'Neil, J.M., Helms, B., Gable, R., David, L. & Wrightsman, L. (1986). Gender Role Conflict Scale: College men's fear of femininity. Sex Roles, 14, 335-350.
Pleck, J.H. (1981). The myth of masculinity. Cambridge, MA: MIT Press.
Pleck, J.H. (1995). The gender role strain paradigm: An update. In R.F. Levant & W.S. Pollack (Eds.), A new psychology of men (pp. 11-32), New York: Basic Books.
Rasch, G. (1960, 1980). Probabilistic models for some intelligence and attainment tests. Copenhagen: Danish Institute for Educational Research, 1960. (Expanded edition, Chicago: The University of Chicago Press, 1980).
Sharpe, M.J., & Heppner, P.P. (1991). Gender role, gender role conflict, and psychological well-being in men. Journal of Counseling Psychology, 38, 232-330.
Smith, R.M. (1991). The distributional properties of Rasch item fit statistics. Educational and Psychological Measurement, 51, 541-565.
Thompson, Eh.H., & Pleck, J.H. (1986). The structure of male role norms. American Behavioral Scientist, 29, 531-543.
Wright, B.D. & Masters, G.N. (1982). Rating scale analysis. Chicago: MESA Press.
|
SCALE |
Measures (b) Mean (S.E.) |
Fit (t) Mean (S.D.) |
Separation of Measures |
N of Men with t > 3.0 |
|
1 Playboy |
-1.74 (1.35) |
-.13 (1.34) |
2.30 |
7 |
|
2 Power |
-1.08 (1.53) |
-.18 (1.58) |
2.49 |
12 |
|
3 Emotion |
-1.06 (2.02) |
-.19 (1.43) |
3.24 |
8 |
|
4 Violence |
-.78 (1.34) |
-.24 (1.6) |
2.49 |
15 |
|
5 Work |
-.57 (.95) |
-.19 (1.61) |
1.56 |
13 |
|
6 Reliance |
-.51 (1.27) |
-.22 (1.54) |
2.01 |
9 |
|
7 Risk |
.12 (1.36) |
-.22 (1.52) |
2.08 |
10 |
|
8 Toughness |
.29 (1.15) |
-.22 (1.60) |
1.99 |
12 |
|
9 Disdain |
.30 (1.68) |
-.22 (1.51) |
2.99 |
11 |
|
10 Dominance |
.35 (1.17) |
-.15 (1.43) |
1.92 |
10 |
|
11 Winning |
.51 (1.70) |
-.23 (1.56) |
2.86 |
25 |
|
12 Status |
.76 (1.15) |
-.25 (1.64) |
1.82 |
33 |
TABLE 1: Summary of the Rasch CMNI results for the men
Note: N=348
|
SCALE |
Fit Mean (S.D.) |
Separation Of Items |
N of items with t >3.0 |
Threshold Estimates (S.E.) |
||
|
1 Playboy |
-.28 (1.74) |
8.61 |
0 |
-2.19 (.04) |
.18 (.05) | 2.01 (.12) |
|
2 Power |
-.48 (2.67) |
10.09 |
0 |
-3.12 (.05) |
.16 (.05) | 2.96 (.13) |
|
3 Emotion |
-.6 (3.28) |
9.66 |
1 |
-3.90 (.06) |
-.07 (.05) | 3.97 (.11) |
|
4 Violence |
-.4 (2.43) |
10.61 |
2 |
-2.49 (.05) |
-.21 (.04) | 2.69 (.08) |
|
5 Work |
-.26 (3.43) |
11.15 |
2 |
-3.04 (.06) |
.12 (.04) | 2.92 (.09) |
|
6 Reliance |
-.48 (2.42) |
16.39 |
0 |
-3.65 (.06) |
.09 (.05) | 3.56 (.08) |
|
7 Risk |
-.62 (3.63) |
15.05 |
2 |
-3.91 (.08) |
-.19 (.05) | 4.09 (.07) |
|
8 Toughness |
-.09 (1.45) |
9.24 |
0 |
-3.11 (.08) |
-.02 (.04) | 3.08 (.07) |
|
9 Disdain |
-.48 (3.03) |
11.87 |
2 |
-3.00 (.07) |
-.02 (.04) | 3.02 (.06) |
|
10 Dominance |
-.19 (2.59) |
10.47 |
2 |
-3.43 (.09) |
.09 (.04) | 3.34 (.07) |
|
11 Winning |
-.52 (2.72) |
11.94 |
1 |
-3.46 (.08) |
-.01 (.04) | 3.47 (.07) |
|
12 Status |
-.61 (4.14) |
14.70 |
2 |
-3.29 (.08) |
-.20 (.05) | 3.49 (.06) |
TABLE 2: Summary of the Rasch CMNI results for the 12 scales
|
|
|
|
Component 1 |
Component 2 |
Component 3 |
Component 4 |
||||
|
Scale |
Det |
KMO |
Eigenvalue |
% Var. |
Eigenvalue |
% Var. |
Eigenvalue |
% Var. |
Eigenvalue |
% Var |
|
1 Playboy |
.02 |
.05 |
2.36 |
19.7 |
1.87 |
15.6 |
1.16 |
9.6 |
1.02 |
8.5 |
|
2 Power |
.04 |
.03 |
1.88 |
15.7 |
1.49 |
12.5 |
1.37 |
11.4 |
1.23 |
10.3 |
|
3 Emotion |
.06 |
.06 |
2.11 |
17.6 |
1.52 |
12.7 |
1.16 |
9.7 |
1.16 |
9.7 |
|
4 Violence |
.03 |
.05 |
2.05 |
17.0 |
1.80 |
15.0 |
1.52 |
12.6 |
1.09 |
9.1 |
|
5 Work |
.03 |
.03 |
1.73 |
14.4 |
1.54 |
12.8 |
1.45 |
12.1 |
1.25 |
10.4 |
|
6 Reliance |
.03 |
.05 |
2.34 |
19.5 |
1.72 |
14.4 |
1.27 |
10.6 |
1.11 |
9.2 |
|
7 Risk |
.03 |
.06 |
2.25 |
18.8 |
1.57 |
13.1 |
1.42 |
11.9 |
1.17 |
9.7 |
|
8 Toughness |
.02 |
.05 |
2.48 |
20.6 |
1.42 |
11.8 |
1.28 |
10.7 |
1.11 |
9.3 |
|
9 Disdain |
.04 |
.09 |
2.44 |
20.3 |
1.62 |
13.5 |
1.31 |
10.9 |
1.20 |
10.0 |
|
10 Dominance |
.03 |
.03 |
2.11 |
17.6 |
1.33 |
11.1 |
1.32 |
11.1 |
1.16 |
9.7 |
|
11 Winning |
.08 |
.05 |
1.98 |
16.5 |
1.43 |
11.9 |
1.34 |
11.1 |
1.16 |
9.7 |
|
12 Status |
.03 |
.04 |
2.16 |
18.0 |
1.48 |
12.4 |
1.29 |
10.8 |
1.15 |
9.6 |
Legend:
Det: Determinant of the correlation matrix
KMO: Kaiser-Meyer-Olkin measure of sampling adequacy
TABLE 3 Summary of the principal component analyses of the CMNI residuals
Score (N) Estimate Error
33( 1) 4.1 0.71 ENDORSEMENT OF THE MESSAGE X|
.
32( 0) 3.7 0.65 |
.
31( 1) 3.3 0.60 X|
.
30( 1) 2.9 0.57 X|
.
29( 0) 2.6 0.55 |
.
28( 1) 2.3 0.53 X|
|
27( 0) 2.1 0.52 |
|
| V26 I am always the first to start a fight. (SA=3)
26( 3) 1.8 0.51 XXX|
|
25( 5) 1.6 0.50 XXXXX|
| V39 If you walk away from a fight, you are
| a coward. (SA=3)
24( 1) 1.3 0.49 X|
| V67 I like fighting. (SA=3)
23( 7) 1.1 0.49 XXXXXXX|
.
22( 7) 0.8 0.49 XXXXXXX|
|
21( 20) 0.6 0.48 XXXXXXXXXXXXXXXXXXXX| V12 If there is going to be violence, I
| find a way to avoid it. (SD=3)
20( 17) 0.4 0.48 XXXXXXXXXXXXXXXXX|
| V18 Violence is almost never justified.(SD=3)
|
19( 23) 0.1 0.48 XXXXXXXXXXXXXXXXXXXXXXX|
|
18( 20) -0.1 0.47 XXXXXXXXXXXXXXXXXXXX|
|
17( 27) -0.3 0.47 XXXXXXXXXXXXXXXXXXXXXXXXXXX| V97 I like watching violent games. (SA=3)
| V45 I am uncomfortable seeing violence. (SD=3)
16( 20) -0.5 0.47 XXXXXXXXXXXXXXXXXXXX| V32 I believe that violence is never
| justified. (SD=3)
15( 28) -0.7 0.47 XXXXXXXXXXXXXXXXXXXXXXXXXXXX|
|
| V58 I am disgusted by any kind of violence. (SD=3)
14( 25) -1.0 0.47 XXXXXXXXXXXXXXXXXXXXXXXXX| V127 No matter what the situation I would never
| act violently. (SD=3)
13( 22) -1.2 0.47 XXXXXXXXXXXXXXXXXXXXXX| V136 I am willing to get into a physical
| fight if necessary. (SA=3)
12( 15) -1.4 0.47 XXXXXXXXXXXXXXX| V141 Sometimes violent action is necessary. (SA=3)
|
11( 8) -1.6 0.48 XXXXXXXX|
.
10( 12) -1.9 0.49 XXXXXXXXXXXX|
|
9( 12) -2.1 0.49 XXXXXXXXXXXX|
.
8( 14) -2.4 0.51 XXXXXXXXXXXXXX|
|
7( 5) -2.6 0.52 XXXXX|
.
6( 9) -2.9 0.54 XXXXXXXXX|
.
5( 6) -3.2 0.57 XXXXXX|
.
4( 6) -3.6 0.61 XXXXXX|
.
3( 1) -4.0 0.67 X|
.
2( 3) -4.5 0.78 XXX|
.
1( 2) -5.3 1.05 REJECTION OF THE MESSAGE XX|
Figure 1 Map showing positions of men and items on the "Violence" scale
ITEM STATUS (COMPONENT) DESCRIPTION V67. Extreme Conformity (Affect) I like fighting V26. (Behavior) I am always the first to start a fight V39. (Cognition) If you walk away from a fight, you are a coward V97. Moderate Conformity (Affect) I like watching violent games V136. (Behavior) I am willing to get into a physical fight if necessary V141. (Cognition) Sometimes violent action is necessary V45.* Moderate Non-Conformity (Affect) I am uncomfortable seeing violence V12.* (Behavior) If there is going to be violence, I find a way to avoid it. V18.* (Cognition) Violence is almost never justified. V58.* Extreme Non-Conformity (Affect) I am disgusted by any kind of violence V127.* (Behavior) No matter what the situation I would never act violently. V32.* (Cognition) I believe that violence is never justified. Legend: Strongly Disagree = 0, Disagree = 1, Agree = 2, Strongly Agree = 3.
An * indicates a reverse coded item. Items that required reverse coding were written as "negative" statements because they reflected the natural manner in which the statements are usually expressed.
Table 4 The Violence Scale on the Conformity to Masculine Norms Inventory
ITEM SEQ MOST PROBABLE RESPONSES
NAME NUM ----------------------- ITEM
-5 -4 -3 -2 -1 0 1 2 3 4 LOGITS VALUE FIT
------------------------------------------------------------------------------------------------------------------ ----- ---
V26 3 1 2 3 1.9 -0.5
V39 5 1 2 3 1.5 -1.3
V67 8 1 2 3 1.2 -3.8
V12 1 1 2 3 0.6 -2.9
V18 2 1 2 3 0.3 -0.6
V97 9 1 2 3 -0.3 3.1
V45 6 1 2 3 -0.4 4.4
V32 4 1 2 3 -0.5 1.1
V58 7 1 2 3 -0.9 -2.3
V136 11 1 2 3 -1.0 1.1
V127 10 1 2 3 -1.0 -0.8
V141 12 1 2 3 -1.3 -2.2
------------------------------------------------------------------------------------------------------------------ ----- ---
-5 -4 -3 -2 -1 0 1 2 3 4 LOGITS
CATEGORY PRINTED IS MOST PROBABLE UP TO NEXT HIGHEST CATEGORY
PERSON FREQUENCY DISTRIBUTION
-----------------------------
1 1 1 1 2 2 2 2 2 2 2 1 2
2 3 1 6 6 9 5 4 2 2 8 5 2 5 8 0 7 0 3 7 0 7 7 1 5 3 1 1 1 1
---------------------------------------S------------M-------------S------------------------------------------------------
LOGITS LOGITS
Legend:
"S" and "M": standard deviation and mean of the estimates, respectively.
TABLE 5 Most probable response for each man to each item of the "Violence" scale
SERIAL ORDER CALIBRATION ORDER FIT ORDER
--------------------------------- ------------------------------------ -------------------------------------------------
| SEQ |CALIBRATION| | | SEQ |CALIBRATION| | | SEQ |CALIBRATION| |
| NUM NAME |VALUE ERROR| FIT | | NUM NAME |VALUE ERROR| FIT | | NUM NAME |VALUE ERROR| MS SE FIT |
--------------------------------- ------------------------------------ -------------------------------------------------
| 1 V12 0.61 0.10 -2.95 | | 3 V26 1.92 0.11 -0.51 | | 8 V67 1.18 0.10 0.74 0.07 -3.76 |
| 2 V18 0.26 0.10 -0.59 | | 5 V39 1.52 0.10 -1.30 | | 1 V12r 0.61 0.10 0.80 0.07 -2.95 |
| 3 V26 1.92 0.11 -0.51 | | 8 V67 1.18 0.10 -3.76 | | 7 V58r -0.92 0.10 0.83 0.08 -2.31 |
| 4 V32 -0.48 0.10 1.06 | | 1 V12 0.61 0.10 -2.95 | | 12 V141 -1.32 0.10 0.83 0.08 -2.20 |
| 5 V39 1.52 0.10 -1.30 | | 2 V18 0.26 0.10 -0.59 | | 5 V39 1.52 0.10 0.90 0.08 -1.30 |
| 6 V45 -0.40 0.10 4.36* | | 9 V97 -0.35 0.10 3.07* | | 10 V127r -1.03 0.10 0.93 0.08 -0.82 |
| 7 V58 -0.92 0.10 -2.31 | | 6 V45 -0.40 0.10 4.36* | | 2 V18r 0.26 0.10 0.95 0.07 -0.59 |
| 8 V67 1.18 0.10 -3.76 | | 4 V32 -0.48 0.10 1.06 | | 3 V26 1.92 0.11 0.96 0.08 -0.51 |
| 9 V97 -0.35 0.10 3.07* | | 7 V58 -0.92 0.10 -2.31 | | 4 V32r -0.48 0.10 1.08 0.08 1.06 |
| 10 V127 -1.03 0.10 -0.82 | | 11 V136 -0.99 0.10 1.12 | | 11 V136 -0.99 0.10 1.09 0.08 1.12 |
| 11 V136 -0.99 0.10 1.12 | | 10 V127 -1.03 0.10 -0.82 | | 9 V97 -0.35 0.10 1.25 0.08 3.07* |
| 12 V141 -1.32 0.10 -2.20 | | 12 V141 -1.32 0.10 -2.20 | | 6 V45r -0.40 0.10 1.37 0.08 4.36* |
----------------------------- -------------------------------- --------------------------------------------
Legend:
SEQ NUM: sequence number of the item.
ERROR: standard error of estimate for an item.
VALUE: logit difficulty estimate.
r: indicates a reverse scored item.
TABLE 6 Item estimates and fit statistics presented in various orders for the "Violence" scale
SEQ PERSON RESPONSES, EXPECTED RESPONSES AND STANDARDIZED RESIDUALS FOR
NUM NAME SCORE MEASURE ERROR MS SE FIT MISFITTING PEOPLE. ITEMS LISTED IN SERIAL ORDER. MISFIT |T|=2.00
----------------------------------------------------------------------------------------------------------------------------------
46 1 1 1 12 -1.41 0.47 2.98 0.37 3.67 0 0 0 0 1 2 0 0 3 1 3 2
1 1 0 1 0 1 1 1 1 1 1 2
-1-1 0-1 1 1-2 0 2 0 2 0
109 2 2 1 9 -2.11 0.49 2.19 0.39 3.42 1 0 0 0 1 3 1 0 0 1 0 2
1 1 0 1 0 1 1 0 1 1 1 1
0-1 0-1 1 3 0 0-1 0 -1 1
130 4 4 3 31 3.29 0.60 2.18 0.41 3.19 3 3 3 3 3 0 3 1 3 3 3 3
2 2 2 3 2 3 3 2 3 3 3 3
1 0 1 0 1-4 0-2 0 0 0 0
|
TABLE 7 Clinically interesting unexpected response patterns on the "Violence" scale