Abstract
My Multiple Selves (MMS) is an open-ended self-concept measure that leads respondents to systematically explore who they are before selecting and rank-ordering their most important selves. A sample of college students (n = 204) completed either the MMS or the frequently used Twenty Statements Test (TST). As predicted, results demonstrated that MMS responses included elements underdetected by the TST: other people, school crowd labels, aspirations for the future, and undesirable selves. Furthermore, the MMS revealed that when other people are included in self, they are rank ordered by participants as the most important element. Shortcomings of the MMS are discussed, and directions for future research outlined.
Open-ended approaches have long been a popular means of assessing an individual’s self-concept. These methods, the most often used of which is the Twenty Statements Test (TST; Kuhn & McPartland, 1954), allow measurement of self-definition from the respondent’s personal perspective. However, the TST and related devices (e.g., Who am I? [WAI], Bugental & Zelen, 1950) may not capture fully the individual’s relatively enduring set of personally important beliefs about self. The TST elicits the most salient answers to the question, Who am I? In so doing, it assesses the “spontaneous self-concept” (McGuire & Padawer-Singer, 1978) and may not elicit important self-definitions that are not quickly and easily accessed from memory. Thus, we propose a new open-ended self-concept measure, My Multiple Selves (MMS), which assesses what we term the considered self-concept, that set of self-beliefs that a person provides after being prodded to think deeply and broadly about who one is and then asked to identify which of these many self-construals are most central to his or her self-concept.
Self-Concept as a Hypothetical Construct
Self-concept is typically defined as the set of thoughts and feelings that the person has about self. The multiple self-construals, or selves, making up a person’s self-concept are not “ . . . a laundry list (of) . . . randomly scattered elements . . . ” (M. Rosenberg, 1979, p. 17). Instead, the individual makes her or his multifaceted self intelligible by developing an integrated, and differentiated, self-concept. Thus, self-concept is a structured set of self-definitions. The pieces fit together into a particular pattern; this pattern or whole is, along with the discrete parts, who I am (M. Rosenberg, 1979; S. Rosenberg, 1997). According to McAdams (1997), people also achieve unity by constructing a life story: where I came from, where I am now, and where I will go in the future. Thus, who I am is not just what I am now, but also includes my past and what I hope or fear I will be in the future. Neisser (1988) terms this the “extended self” (p. 36).
Hierarchy is one form of mental organization that is particularly important for all or most people. James (1890) noted that people arrange self-definitions along a hierarchical scale, from those most important to those of lesser importance. The notion of psychological importance is widely recognized by psychologists (e.g., hierarchical arrangement or elaboration S. Rosenberg & Gara, 1985) and sociologists (e.g., centrality M. Rosenberg, 1979), as well as by those concerned with personal selves (S. Rosenberg, 1997) and those focusing on social identities (Ashmore et al., 2004; Deaux, 1991).
TST as All-Purpose Measure of Self-Concept
The TST has become the most widely used open-ended self-concept measure. From a humble introduction (see Bugental and Zelen’s, 1950, caution regarding the exploratory nature of the WAI) and a not so well received childhood and adolescence (see Wylie’s, 1974, critical review of the accumulated literature), the WAI/TST (hereafter TST) has become the instrument of choice for those wishing to assess self-concept from the respondent’s point of view (for reviews and critiques, see Gordon, 1968; Grace & Cramer, 2003; Rees & Nicholson, 1994; Spitzer et al., 1970; Wells & Marwell, 1976; Wylie, 1974; Zurcher, 1977).
The TST is appropriate for studying situation-specific self-concept reporting (which was the original intent of its creators Hickman & Kuhn, 1956; cf. Wylie, 1974), and there is considerable evidence that the TST is capable of detecting how context affects self-definition (cf. McGuire & McGuire, 1984). Furthermore, it is likely that highly important self-beliefs are often activated in everyday life and may be relatively easily brought to mind in most assessment situations.
However, use of the TST to assess comprehensively enduring and personally important self-beliefs may be questioned. As with any self-report measure, the TST assumes that respondents are willing and able to reveal their self-beliefs. There is a substantial literature on socially desirable responding as a threat to the validity of self-report instruments (Cook & Selltiz, 1964; Holtgraves, 2004; Paulhus, 1991). Most respondents tend to accentuate the positive and reduce the negative in self-reports, and this bias is present in TST responses (McGuire et al., 1979, p. 81).
Concerning respondents’ ability to report self-beliefs, there are three issues with the TST. First, self-concept likely involves much tacit knowledge; information stored in memory that guides thought, feeling, and action, but which is not easily accessed. People likely need to consciously explore their memory to identify tacit self-beliefs, but the TST does not encourage such reflection. Second, self-esteem maintenance motivation (Baumeister, 1998) may cause people to avoid thinking about the undesirable aspects of self; therefore, these self-construals are likely less accessible without prompting, which the TST does not provide. Third, people likely have a huge store of knowledge about self and identifying the most important selves may be difficult without help from the assessment procedure, help the TST does not provide. Consequently, participants may use simplifying heuristics such as distinctiveness (McGuire & McGuire, 1984).
As the foregoing considerations suggest, ability to report one’s self-definitions depends on the self-knowledge and motives that the person brings to the situation, and the assessment procedure itself. The standard TST instructions are: There are twenty numbered blanks on the page below. Please write twenty answers to the simple question “Who am I?” in the blanks. Just give twenty different answers to this question. Answer as if you were giving the answers to yourself, not to somebody else. Write the answers in the order that they occur to you. Don’t worry about logic or “importance.” Go along fairly fast for time is limited. (Kuhn & McPartland, 1954, p. 69)
Participants are instructed to ignore the importance of self-definitions and complete the task quickly. This likely causes respondents to report salient self-views and discourages them from thinking deeply about self. Although appropriate for assessing situational variation in self-beliefs and testing the distinctiveness hypothesis, this approach is not fully adequate for assessing the individual’s enduring and personally significant ideas about self.
Furthermore, the sentence completion response format of the TST likely underplays self-definitions that do not logically follow from the implicit stem “I am ____,” which suggests the most appropriate answers are adjectives. Thus, the TST may elicit simple physical descriptors and personality traits (e.g., “I am . . . thin, kind”). It may not occur to respondents that other important people in their lives (e.g., the respondent’s mother or boyfriend) are acceptable answers to the stem “I am ____” (e.g., “I am my Mother”) even if they regarded these other people as important parts of self. Similarly, the “I am ___” construction may limit other self-extensions. For example, while nationality and ethnicity are likely viewed as appropriate responses (e.g., “I am an American”), respondents may not consider informal social labels applied by others and accepted as self-defining (e.g., school crowd labels such as jock). Furthermore, the TST does not invite respondents to consider future aspirations (e.g., “I hope to be a veterinarian”). Finally, the TST does not provide information on self-concept structure. Participants are instructed not to consider importance, yet most theories of self suggest that hierarchal arrangement is a central feature of self-concept organization. Serial position of TST responses has been used to index psychological centrality (e.g., Kuhn & McPartland, 1954), but the evidence supporting this approach is mixed (cf. Gordon, 1968; Wylie, 1974).
Recent Research With the TST
To date, the TST remains a widely used open-ended self-concept assessment (see del Prado et al., 2007; Grace & Cramer, 2003; Vignoles et al., 2006). Some attempts have been made to legitimize a more diverse range of responses. For example, Becker et al. (2012) used a modified version of the TST in which participants were encouraged to consider social groups they belong to, and relationships with other people. However, as described below, this is only a small subset of possible responses to open-ended self-concept assessments. Furthermore, in Becker et al. (2012), responses were only rated by participants on specific dimensions relevant to the research questions (perceived centrality, positive affect, and four measures of distinctiveness), and no comparisons were made between this modified version and the original TST instructions. Thus, recent research has yet to directly address the issues raised above regarding the inability of the TST to asses a diverse range of important self-beliefs.
Other Self-Concept Measures
More recent self-concept assessments tend to focus on specific structural aspects of self, such as self-complexity and self-concept clarity. In contrast to the TST, measurement of these constructs does not include open-ended assessments of self-concept content. For example, Self-complexity is assessed by having participants sort a set of experimenter-defined trait words into piles (Rafaeli-Mor et al., 1999). Self-concept clarity is assessed with close-ended, Likert-type ratings of experimenter-defined scale items (Campbell et al., 1996). While both measures are well suited for their intended purpose, neither are open-ended measures of self-concept content, and neither assess the importance or psychological centrality of specific self-descriptors.
Similarly, there are a variety of self-concept scales designed to measure specific self-concept dimensions such as relational, individual, and collective self-aspects (RIC; Kashima & Hardie, 2000), and independent and interdependent self-construals (SCS; Singelis, 1994). However, these are also close-ended, Likert-type type ratings of experimenter-defined scale items designed to measure a specific dimension as opposed to an open-ended assessment of self-concept content and structure.
Assessing the Considered Self-Concept
The foregoing argument suggests the need to develop an open-ended measure of self-concept that invites respondents to consider a variety of self-aspects and directs them to rank order their most important selves. The MMS is offered to fill this need and employs the following strategy: (a) legitimize diverse contents; (b) facilitate careful and conscious exploration of memory; (c) request identification and hierarchical ordering of most important selves.
The MMS instructions begin by legitimizing a wide variety of content: “Each of us is really several different selves in the same person.” The many possible types of selves are briefly explained, and examples of several different selves are presented. Next, the MMS presents a set of open-ended probes asking the person about his or her self-definitions in a wide variety of content areas (see Method section for details). After respondents have filled in self-definitions for each content area, they are asked to identify those self-construals that are most personally important (see Method section). The final result of this procedure is the individual’s hierarchically arranged set of most important selves. The MMS assesses both considered self-concept content and structure in a way that allows respondents to uncover tacit knowledge, consider negative aspects of self, and to consider aspects of self that might not appear as acceptable or logical completions of the stem “I am ____.”
Current Research and Predictions
The primary goals of the current research were to (a) compare directly the MMS with the TST; (b) test for potential methodological effects due to the order of MMS probes and the number of response slots available; (c) test if legitimizing diverse self-content, a key element of the MMS, would allow those who initially completed the TST to add types of selves not well detected by the TST; and (d) explore perceptions of others who are introjected into self.
Goal 1
We predicted that relative to the TST, MMS respondents would include a higher proportion of four types of selves that are not well detected by the TST: (a) Other people; (b) School crowds; (c) Aspirations; (d) Negatively evaluated. In contrast, we predicted that the TST would yield more personality traits than the MMS.
Other people
James (1890) recognized that people can extend the self beyond their body. One way of doing this is to make others part of self. Freud and other psychoanalytic theorists used many terms to describe this process, including “introjection,” “internalization,” and “incorporation” (Compton, 1985; Lancaster & Foddy, 1988; Saravay et al., 1994). Object relations theories suggest children internalize images of parents as “selfobjects” (Kohut, 1977). McGuire and McGuire (1982) found that children and adolescents frequently mentioned other people when asked “Tell us about yourself” (p. 84). Many other researchers have recognized the importance of making others part of self (Andersen & Chen, 2002; Aron & Aron, 1986; Markus & Kitayama, 1991; Ogilvie & Ashmore, 1991; M. Rosenberg, 1979). Which others can be expected to be included in self? The most likely are parents, especially mothers who are generally the primary caretaker (Bowlby, 1969; Kohut, 1977; Markus & Cross, 1990; Ogilvie & Ashmore, 1991; Stern, 1985), and romantic partners (Aron & Aron, 1986). Thus, the self-concepts of our late adolescent participants are likely to contain childhood identifications and current love interests.
School crowds
The social system in which one is embedded will likely influence self-definition. Our respondents are college students, and the college student culture has a rich vocabulary for designating various types of students (Ashmore et al., 2002). These labels (e.g., nerd, brain) are used to make sense of others and to define self (Lemay & Ashmore, 2004). Therefore, it is expected that participants will incorporate college student type labels into their self-concept.
Aspirations
Late adolescents are looking to the future as they develop their personal identity (Erikson, 1968) and life story (McAdams, 1997). Thus, it is expected that they will include in their self-concept possible selves (Markus & Nurius, 1986). That is, who I am now includes who I hope to be or who I fear I might become in the future.
Negatively evaluated
Participants were expected to report a substantial proportion of positive qualities. However, it is assumed that people are motivated to be as accurate as possible in self-perception (Jussim et al., 1995), which should lead participants to include some negative qualities as important parts of self. Furthermore, it is predicted that the proportion of positive and negative self-contents would correlate with self-esteem for participants completing the MMS but not for participants completing the TST.
Goal 2
To test for potential methodological effects due to the order of MMS probes and the number of response slots available, participants were assigned to complete either the TST or one of four versions of the MMS (described in the Method section). We still predicted that relative to the TST, all four MMS conditions would produce higher proportions of others in self, self in school crowds, aspirations, and negative selves.
Goal 3
To address the third goal, respondents in the TST condition completed two additional questionnaires (see Method section) that legitimized the same diverse domains as the MMS and allowed them to think more deeply about their self-concept. We expected that the final self-descriptor lists based on these follow-up activities would show increases from Initial TST responses in the same categories expected to differentiate the MMS from the Initial TST: (a) Other people; (b) School crowds; (c) Aspirations; (d) Negative selves.
Goal 4
To address the fourth goal, participants in all conditions rated the other people that appeared in their list of 16 most important selves on several scales (e.g., “I love this person”). No predictions were made regarding Goal 4 beyond the obvious expectation that introjected others would be rated relatively positively.
Method
Participants
Two hundred four (120 female, 84 male) college students from a state university in the Northeastern United States took part in this study for partial credit in an Introductory Psychology course. Participants varied in age from 16 to 53 (M = 18.75, SD
Materials
My Multiple Selves (MMS)
As noted above, the MMS (available from the first author) directs the participant to open-endedly describe multiple facets of self across 12 content areas. The MMS participant instructions and probes for each content area are presented in the appendix. The content areas are (1) important people in the person’s life, (2) roles, (3) skills, (4) liked activities, (5) inner self (traits, emotions, beliefs), (6) physical, (7) future selves, (8) less desirable side, (9) social categories that others put on the person, (10) personal background, (11) current situation, and (12) other selves. The important people probe (1) had eight follow-up items, asking the person to list mother, father, another family member, a person outside the family who had a big impact on them, boy/girlfriend/lover/spouse/fiancé/partner, close friend, someone who did the person wrong, and groups of people important to the person. For roles (2), skills (3), liked activities (e.g., including school activities and interpersonal relationships such as hobbies, sports, social groups, etc.), (4), inner self (5), physical (6), future (7), less desirable side (8), and other selves (12), there were three blank lines for the person to fill in. Item 9 (“Social categories people put me in”) presented a list of 59 college crowd labels derived from previous research (Ashmore et al., 2002). Respondents were asked to check those that others categorized them into, to pick two of these that were most often applied to them, and, for these two, to indicate whether each was an important part of self and also whether they would use a different label for the same identity. Item 10 had spaces for the participant to fill in sex, ethnicity, religion, and social class. “My current situation” (11) had 7 subprobes: major, where grew up, what do for a living, primary extracurricular activity, primary hobby, club, and political party or group. Item 12 (other selves) asked participants if we missed any important aspect that they would include as an important aspect of their self-concept. Participants were then free to list three additional descriptors, and there were no constraints on these items. Thus, if participants wished to provide additional items for any of the previous probes or even descriptors that were not included in any of the previous probes, they were free to list them here (e.g., other hobbies, interests or social groups, etc.). After listing multiple self-definitions, the participant is instructed to put a star next to at least 20, and no more than 25, identities the respondent feels are most personally important. Finally, the person is asked to list, in rank order of personal importance, the 16 most important identities. This final list of 16 self-selected rank ordered identities is subjected to the content coding process described below.
To test for potential methodologic effects due to the order of probes or the number of response slots, four versions of the MMS were created by crossing two factors: (1) Order of the important people in my life and the inner me probes; and (2) Number of response slots for each probe. Because one of our primary hypotheses was that the MMS would produce higher proportions of other people compared with the TST we were concerned that placing this probe first could artificially increase the number of other people listed. Thus, order 1 presented the important people probe first and the inner me probe fifth; Order 2 presented the inner me probe first and the important people probe fifth. Number of response slots was either variable (described above), or four response slots for all probes (the important people probe had four prompts: parents, another family member, boy/girlfriend/lover/spouse/fiancé/partner, and close friend).
The Twenty Statements Test (TST)
The current study used the standard TST procedure described in detail in the introduction. This is a single sheet of paper with 20 blank lines and instructions that encourage speed and discourage reflection and consideration of the relative importance of responses (Kuhn & McPartland, 1954, p. 69). Two open-ended questionnaires followed the TST: “Further Thoughts” and “Final 16.”
Further Thoughts
Further Thoughts explained (as does the MMS) that “Each of us is really several different selves in the same person . . . The aim of this questionnaire is to get a list of the different facets of you.” The facets used on the MMS were then listed and described. The respondent was then asked to list “20 more answers to the question, Who am I?”
Final 16
The Final 16 questionnaire instructed participants to look back at their TST and Further Thoughts responses and identify the 16 most important selves and put these in rank order.
Rating My Most Important Selves
Participants rated each self from their top 16 selves (MMS or TST Final 16) on Importance (This self is an important/not important part of my overall self-concept) on a 1 to 5.
Other People in the List of My Most Important Selves
Participants rated each other person included in their final 16 list on seven dimensions which tap emotional closeness and the degree of influence the other person may have had: This person is part of me; This person influenced me; I would feel the loss of this person; I like this person; I like myself when I am with this person; I love this person; This person is a role model for me. All ratings were made on a 5-point scale (1 = Not at all true to 5 = Completely true).
The Rosenberg Self-Esteem Scale (SES)
M. Rosenberg’s (1965) Self-Esteem Scale is a 10-item questionnaire designed to measure global feelings of self-worth or self-acceptance; each item is rated on a 1 (Strongly Agree) to 4 (Strongly Disagree) scale. The SES has high internal consistency and test–retest reliability, and there is considerable evidence of both convergent and discriminant validity (Blascovich & Tomaka, 1991).
Procedure
Participants were run in groups of 10 to 30, but each participant was attended to individually and worked at his or her own pace. Participants were randomly assigned one of the four forms of the MMS or the TST. After completion of the MMS or TST (TST, Further Thoughts, and Final 16), all respondents completed the Rating My Most Important Selves and the Other People in the List of My Most Important Selves forms. Finally, the SES and a basic Demographics survey were administered.
Content Analysis
The content analysis scheme (available from the first author) was conceptually derived, accords well with major elaborated schemes for coding TST responses (see Cousins, 1989; Gordon, 1968; Hartley, 1970; Kuhn, 1960; McPartland, 1965; Rhee et al., 1995), and can be used to code self-definitions obtained by any open-ended self-concept measure. The system (outlined in the far left-hand column of Table 1) describes the self at three levels: (a) Individual, an individual person with qualities that distinguish self from other people; (b) Social, a person in relation to other people; (c) Abstract, an individual in relation to the universe as a whole. The first two meta-categories are widely recognized as personal identity and social identity (Deaux, 1996; M. Rosenberg, 1979; Sedikides & Brewer, 2001). The third meta-category recognizes that people can define self in terms that transcend the individual and other people (e.g., “I am a pebble in the river of life”). Gordon (1968) terms these “abstract identifications” (p. 125; though he also includes under this rubric “ideological and belief references” such as liberal that we categorize differently) and Zurcher (1977) uses the label “oceanic” (p. 47).
Proportion of Responses in Each of 16 Type of Self-Description Categories, Total Individual, and Total Social.
Note. Means in the five experimental conditions (four MMS and Initial TST) not sharing a superscript are different at p < .05. Means in the TST condition (Initial TST, Further Thoughts, Final 16) not sharing a subscript are different at p < .05. MMS = My Multiple Selves; TST = Twenty Statements Test; Mean Prop = mean proportion of total responses coded into each category; % Res = percentage of respondents using each category at least once.
The personal identity or individual self-description meta-category was further partitioned into physical (e.g., “tall”) and seven categories of psychological contents: abilities (e.g., “skilled at math”), interests (e.g., “like to shop”), traits (e.g., “outgoing”), emotions (e.g., “happy”), beliefs (e.g., “pro-environment”), changeable (e.g., “outgoing with some people”), complex (e.g., “multifaceted”), and aspirations (e.g., “want to get PhD”). Following Thoits and Virshup (1997) and Sedikides and Brewer (2001), social self-description was partitioned into (1) relational self which includes (a) self and specific other individuals (other person [e.g., “my mom”] and one’s feelings for another, abbreviated as “feelings other” [e.g., “love my dad”]); (b) social role (e.g., “son”); (2) collective self (membership in a group), which includes (a) demographic group (e.g., “male”); (b) school crowds that others put the person into (e.g., “playboy”); (c) chosen group (e.g., “chess club”).
MMS responses were also coded for evaluation: positive, negative, or neutral (or valence uncertain). There is considerable evidence that evaluation is a primary connotative meaning dimension (Osgood, 1969; Osgood et al., 1957; Rosenberg, 1977 on evaluation and coding TST protocols, see Gordon, 1968).
MMS respondents’ final lists and TST respondents’ Initial TST, Further Thoughts, and Final 16 lists were independently coded by five undergraduate research assistants. Coders were trained to use the coding scheme described above and given a detailed coding guide with descriptions of each coding category and example items. Coders practiced on pilot data until an acceptable level of interrater reliability was reached. All coders were blind to experimental condition, the study’s hypotheses, and to identifying information about respondents. Coders did not know which MMS probe produced responses. Average intercoder percent agreement for MMS lists of 16 most important selves and the TST Final 16 lists were: Type of self-description categories 75.1% and evaluation categories 81.5%. For the Initial TST responses, parallel values were 79.8% and 86.8%, respectively. For Further Thoughts responses, parallel values were 71.5% and 87.5%, respectively. Thus, the content analysis was implemented with acceptable intercoder agreement.
Results
Goal 1: Comparison of MMS with TST
Other people, school crowds, and aspirations
Mean proportions in each category and percentage of participants including at least one self in each category are reported in Table 1. A 1 × 5 multivariate analysis of variance (MANOVA) computed across the five experimental conditions (four MMS plus Initial TST) for the 16 type of self-description indices was significant, F(64, 730) = 7.27, p < .001. Tukey’s HSD was computed across the five experimental conditions for each of the 11 categories showing a significant univariate effect within the MANOVA; those means not sharing a superscript are significantly different at p < .05. The nonsignificant univariate effects were: Physical, interests, emotions, feelings for other person, demographic group. As predicted, the MMS (all four versions) produced larger mean proportions than the TST in the Other person, school crowd (though the Order 1/Variable Slots condition was not statistically significant), and aspirations categories. Also, as predicted, the Initial TST produced more trait responses than the four MMS conditions. A second 1 × 5 MANOVA computed for the two sum indices Individual and Social was also significant, F(8, 394) = 14.79, p < .001. The MMS yielded more Social responses, and the Initial TST more Individual responses (though the Order 2/Fixed Slots version was not significantly different from the Initial TST).
Negatively evaluated
Mean proportions for the evaluation codings are reported in Table 2. A 1 × 5 MANOVA for the 4 MMS conditions plus the Initial TST was significant, F (8, 394) = 5.14, p < .001. Univariate effects were significant for positive, F (4,199) = 7.20, p < .001, and negative, F (4, 199) = 7.20, p < .001, but not neutral. All four MMS conditions produced a significantly higher proportion of negative selves. Also, all four MMS conditions yielded a lower portion of positive selves (although the difference was not statistically significant for Order 2/Fixed Slots). As predicted, the MMS allowed participants the opportunity to acknowledge self-elements that were undesirable and to report a self-concept containing negative, neutral, and positive content. As predicted, for respondents completing the MMS, self-esteem correlated with positive (.267, n = 161, p = .001), negative (−.338, n = 161, p < .001), and POS-NEG DIFF (.353, n = 161, p < .001). These correlations were not significant in the TST condition.
Proportion of Responses Coded as Positive, Negative, or Neutral.
Note. Entries are mean proportion of total responses coded into each category. Means in the five experimental conditions (four MMS and Initial TST) not sharing a superscript are different at p < .05. Means in the TST condition (Initial TST, Further Thoughts, Final 16) not sharing a subscript are different at p < .05. MMS = My Multiple Selves; TST = Twenty Statements Test.
Goal 2: Methodological Effects
The four versions of the MMS did yield different results (Table 1). The Order 1/Variable Slots yielded the most other person in self and the Order 2/Fixed Slots yielded the least. A 2 × 2 MANOVA for the four MMS conditions showed only a significant main effect for Slots, F(16,143) = 4.65, p < .001, with significant univariate effects for emotions, beliefs, other person in self, social role, and chosen group. Having a larger number of slots for the Important Others probe does yield more other people in self in the final list of most important selves. However, it is important to note that the proportion of other person in self is significantly higher in all MMS conditions compared with the TST. Regarding the proportion of selves for the evaluation codings (Table 2) there were no significant main effects and no significant interaction in the 2 (Order) × 2 (Slots) MANOVA for the MMS conditions.
Goal 3: Legitimizing Content on TST
Other people, school crowds, and aspirations
The far-right-hand columns of Table 1 display the mean proportion of selves listed in each type of self-description category (as well as the percentage of participants including at least one self for that category) for those assigned to the TST condition (Initial TST, Further Thoughts, Final 16). A repeated measures MANOVA across the three TST conditions was significant, F(32, 132) = 6.66, p < .001; univariate effects were significant for Physical, interests, traits, beliefs, changeable, aspirations, other person, school crowd, and Abstract. Those means not sharing a subscript are significantly different at p < .05. As predicted, in comparing the Final 16 with the Initial TST, there were significantly higher proportions for other person and school crowd, but the expected increase in aspirations was not statistically significant. Thus, just as the MMS makes it possible for people to see and acknowledge parts of self underdetected by the TST, those completing the latter can be nudged to recognize these same contents when asked to take a considered look at themselves.
Negatively evaluated
The repeated measures MANOVA for positive, negative, and neutral codings (Table 2) for the three TST conditions was significant, F(4, 160) = 6.77, p < .001, as was that for the POS − NEG difference score, F(2, 82) = 11.31, p < .001. For TST condition, encouraging participants to look further at who they are after completing the Initial TST led to an increase in the proportion of negative selves and a decrease in the proportion of positive selves (though this latter difference was not statistically significant) in their Final 16 lists.
Findings for Goals 1 and 3 converge to suggest that affording participants the opportunity to consider their self-concept (either via the MMS or the Further Thoughts follow-up to the Initial TST) yields a more complex and nuanced picture than when they are simply instructed to report the first thoughts that come to mind (standard TST procedure).
Goal 4: Perceptions of Others Introjected into Self
Table 3 reports which others are included in self and their hierarchical placement in the respondents’ self-concept. Parents as a unit plus mother and father mentioned separately were the most frequently mentioned introjected others. Introjected parents were high in the respondents’ self-systems, with median ranks of Parents 1, Mother 2, Father 3; Significant other median rank was 4. The eight far-right columns of Table 3 report mean ratings of each type of other person in self. For all eight scales, parents is either the most positively or second most positively rated introjected other, although mean differences are generally small and not significant (likely because ratings are toward the extreme ends of the rating scale, with the exception of others who hurt the person).
Specific Others in Self: Number/Proportion, Median Rank, Ratings.
Note. Only MMS condition participants are included. Ratings (far-right eight columns), Importance response scale was 1 = Very important, 5 = Not at all important; for the next seven ratings the response scale was 1 = Not at all true to 5 = Completely true. Means not sharing a superscript are different at p < .05.
Table 4 presents the percentage of participants whose first listed self (the single most important self) was coded into each of the 16 types of self-description categories. For all four MMS conditions, other person in self was the category most often mentioned as the respondent’s most important self, though it was tied with traits in the Order 2/Variable Slots condition. Similarly, for respondents in the TST condition, 23.8% included an introjected other as their most important self in their Final 16 list.
Percentage of Respondents Whose First Listed Self Was Coded Into Each of 16 Type of Self-Description Categories.
Note. Entries are percentage of participants whose first listed self was coded into each major category. MMS = My Multiple Selves; TST = Twenty Statements Test.
Discussion
The current research demonstrates that there is more to self-concept than revealed by the TST. When participants are asked to systematically explore their beliefs and feelings about self, they include not just traits and roles, as has been regularly demonstrated with the TST, but also previously underdetected contents: other people, self in school crowds, aspirations, and negative selves. Of these underdetected contents, other people as part of self is the most important. Not only do the present results show that people do incorporate others into their identity, but also that parents and significant others are the most common introjects, and that introjected others are seen as extremely high up in participants’ self-system.
MMS Versus TST With Further Thoughts and Final 16
Goal 3 of the current research was to legitimize content on the TST using a “Further Thoughts” and “Final 16” follow-up to the TST, and results indicate that this procedure yielded significantly higher proportions for other person, school crowd, and negative selves, compared with the Initial TST. This suggests that some of the limitations in the TST can be overcome by using this more reflective procedure. This opens up the possibility for researchers to use the Further Thoughts and Final 16 procedure in conjunction with the TST, instead of the full MMS. This may be beneficial in some cases. The Further Thoughts and Final 16 procedures take less time than the full MMS and would be a viable option if time constraints are an issue. In addition, the TST, Further Thoughts, and Final 16 procedure would allow researchers to assess both the “spontaneous self-concept” via responses on the TST and the more considered aspects of the self-concept using the Further Thoughts follow-up.
That being said, it is important to note that while the Further Thoughts, and Final 16 procedure did lead to higher proportions for other person, school crowd, and negative selves, compared with the Initial TST, all three of these proportions were still lower than those reported on the full MMS. Thus, at this point, it is not entirely clear if the Further Thoughts and Final 16 procedure still leads to underreporting of these features, or if perhaps the MMS artificially inflates reporting in some of these categories. Future research (described bellow) should explore this in more detail.
Limitations and Future Directions
There are two limitations in the MMS. First, while the instructions are intended to legitimize a wide variety of content they may also create a social desirability set such that respondents feel that they should include in their final list some selves from all the probe categories. However, it is important to note that only the final list of 16 responses, freely selected by participants are subjected to the coding system. Thus, while the MMS asks participants to generate responses to all MMS probes, participants are then free to choose only the responses that they feel are personally important. In addition, there is systematic variation, with other person in self, social roles, and traits garnering relatively high proportions consistently. Thus, although participants might feel that they should include diverse self-contents, they do not appear to feel that they should do so to the same degree across all MMS probes. Therefore, differences between the MMS and the TST cannot be due entirely to demand characteristics.
Second, the MMS builds a ceiling into the proportions reported in Table 1. For example, in the Fixed Slots conditions where each MMS probe had four blank lines, respondents could have a maximum of four other people, four abilities, four roles, and so on. If a participant thought of self primarily in terms of traits (or roles, or other people), he or she could not show this on the MMS, with the maximum proportion in the Fixed Slots conditions being .25 (4 of 16). This is likely to have happened for the Variable Slots conditions (with three response slots for inner self, the ceiling for traits was .187). Thus, the Variable Slots version of the MMS may underassess traits as part of self-concept.
These major shortcomings suggest two goals for future research. The first goal is to assess whether participants feel that it is socially desirable to include a wide variety of content areas in their list of most important selves. The second goal is to develop MMS versions that eliminate the potential for ceiling effects by placing no restrictions on the number of responses a participant can list for each probe.
Additional important tasks for future research are to identify why others are introjected into self, how these introjections function, and how these introjections differ from merely listing a role. For example, is there a difference between listing another person as part of the self (e.g., “my father”), as opposed to listing a role that specifies a relationship between two people (e.g., “I am a son”)? Perhaps roles represent relatively more objective “facts” about an individual while choosing to include important others as part of the self may reveal more about the nature of the relationship. Regarding why others are introjected into self and how these introjections function, factors may include dependence on others, especially during childhood (McGuire & McGuire, 1982), or the introjection of others for whom the individual feels highly responsible (Lancaster & Foddy, 1988). People may also introject others to expand their self; to appropriate the personal resources, mental point of view, and identity of a significant other to behave more adaptively and successfully (Aron & Aron, 1986). Characteristics of introjected others may include model warmth and success (two crucial factors leading children to imitate adults; Bandura & Walters, 1963).
A final important direction for future research is to assess the correlates of, and the outcomes associated with, the various types of self-content, especially those underdetected in previous research. Is introjection of a specific other associated with relationship satisfaction? What current life circumstances as well as personality and other constructs covary with including school crowd labels in self? Does including aspirations in self-concept predict psychological or physical health outcomes?
Conclusion
Overall, the current research suggests that while the TST may be an appropriate measure of the spontaneous self-concept, important aspects of the self may not be adequately detected by an assessment procedure that relies primarily on salience (as the TST does). In contrast, more reflective procedures such as the MMS and Further Thoughts activity reveal self-features that are underdetected by the TST, but are still extremely important to participants’ views of themselves. We do not intend the MMS (or Further Thoughts) to replace the TST. Rather we see the MMS, Further Thoughts, and the TST as complementary assessments methodologies that can all be used by researchers, depending on their research goals and questions, to provide a more comprehensive assessment of the self.
Footnotes
Appendix
Acknowledgements
We thank Frances K. Del Boca, Marc Beebe, and the many undergraduates who worked in our lab during data collection and analysis. Finally, we acknowledge Seymour Rosenberg whose work on self as a multiplicity led us to develop the initial version of the MMS.
Authors’ Note
This research was initiated and conducted under the direction of Dr. Richard D. Ashmore. We thank our mentor for his guidance and insight throughout the entire research process.
Declaration of Conflicting Interests
The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The author(s) received no financial support for the research, authorship, and/or publication of this article.
