Two measures of self-regulation for young adults and late adolescents in the academic and social domains

Abstract

In this paper we present domain-specific measures of academic and social self-regulation in young adults. We base our scales on Baltes and colleagues’ Selection, Optimization, and Compensation (SOC) model, and establish the factor structure of our new measures using data collected from a sample of 152 college students. We then compare the predictive validity of our scales to that of a domain-general version of the original SOC questionnaire. Our results support the internal and external validity of the academic SOC scale, although support for the social SOC scale is more tentative than support for our academic measure. We discuss these scales as useful supplements to the existing SOC questionnaire.

Keywords

compensation intentional self-regulation optimization scale development scale validation selection SOC

Intentional self-regulation is a core facet of human functioning that reflects an individual’s conscious attempts to influence bidirectional interactions with his or her environment (Gestsdottir & Lerner, 2008). Brandtstädter (1998) has called such bidirectional interactions “developmental regulations” and, when these interactions benefit both the individual and his or her context, “adaptive developmental regulations.” Adaptive developmental regulations are a hallmark of positive human development (Brandtstädter, 2006; Gestsdottir, Urban, Bowers, Lerner, & Lerner, 2011) and self-regulation has accordingly been associated with positive developmental outcomes across the lifespan (for reviews, see Geldhof, Little, & Colombo, 2010; McClelland, Ponitz, Messersmith, & Tominey, 2010).

Despite the integral role of context in adaptive developmental regulations, researchers often treat self-regulation as a domain-general attribute that individuals possess equally across situations. In this paper we extend the measurement of self-regulation beyond such domain-general conceptualizations; we present and validate two domain-specific measures of self-regulation based on Baltes and colleagues’ Selection, Optimization, and Compensation model (SOC; e.g., Baltes & Baltes, 1990; Freund & Baltes, 2000).

Our scales target self-regulation in the academic and social domains. We chose these domains due to the importance of self-regulation for both academic and social functioning in late adolescents and young adults (e.g., Schunk & Zimmerman, 1994; Vohs & Ciarocco, 2004) and because of the breadth of research done in these domains. Constraining measures of self-regulation to specific domains also necessarily limits the extent that these measures can be utilized, and choosing widely studied domains for which the importance of self-regulation is well-established, will increase the utility of domain-specific measures.

The SOC model

The SOC model is an action-theoretical approach to self-regulation that describes the relations between aspects of goal pursuit and developmental outcomes (e.g., Freund & Baltes, 2000). The SOC model emphasizes developmental regulations and acknowledges that multiple interacting levels of the environment continuously influence ontogenetic development, just as the developing individual likewise influences his or her own environment. SOC accordingly sees development as the result of dynamic interactions between an individual and all levels of his or her context (see also Lerner & Busch-Rossnagel, 1981).

Components of SOC

The SOC model specifies four processes that facilitate goal attainment, which the SOC model acknowledges are contingent on resource availability across the lifespan. Distributing a limited set of resources across a limitless set of goals leads one to apply too few resources to any given goal, forcing individuals to select meaningful goals from a larger pool of possibilities. The SOC model calls this elective selection (henceforth selection; Baltes & Baltes, 1990; Freund & Baltes, 2000). Selection prevents the over-dispersion of goal-relevant resources (e.g., Lerner, Freund, De Stefanis, & Habermas, 2001) and ensures that individuals have the resources needed to reach selected goals. In the academic context, selection includes choosing and prioritizing classes and coursework relevant to reaching individual academic goals. Selection in the social context similarly involves the selection and prioritization of social relationships (e.g., friendships).

Goal selection is only adaptive, however, when selected goals are actually pursued. In the SOC model, optimization refers collectively to the acquisition, refinement, and application of goal-relevant means. Optimization facilitates the attainment of selected goals and accordingly resembles more traditional conceptualizations of self-regulated behavior (e.g., focus, persistence). In the academic domain, optimization represents behaviors that specifically help one attain academic goals (e.g., focusing on schoolwork), while social optimization can be conceptualized as actions that strengthen social bonds (e.g., keeping commitments to a friend).

The processes of selection and optimization facilitate higher-level functioning (Freund, Li, & Baltes, 1999), but the processes only respond to developmental gains (i.e., increases in ability due to developmental changes). Developmental declines can cause a mismatch between an individual’s ability and the demands placed by the environment (e.g., Bäckman & Dixon, 1992), and the SOC model discusses two additional processes that respond to such loss.

When individuals implement new or additional means to regain a previously held level of performance, they enact what the SOC model refers to as compensation. For example, academic compensation includes the flexible application of a new means to reach a specific academic goal after an initially utilized means is no longer available, like seeking tutors when self-guidance no longer earns the `A'. Social compensation can similarly be thought of as finding new ways to maintain and develop social relationships when relationship-threatening difficulties alter relationship dynamics in some way (e.g., very different behaviors are required to optimize a friendship before vs. after one friend moves across the country).

Compensation renews functioning toward an already selected goal. At the same time, developmental declines can make previously held goals realistically unattainable. When developmental declines make a previously held goal unattainable, or cause the cost of compensation (e.g., increasing practice even more) to exceed the benefits of goal attainment, selecting a new goal becomes more appropriate than continued goal striving (e.g., becoming a ski instructor). The SOC model differentiates between goal selection driven by developmental gains versus developmental declines, calling the latter loss-based selection. Domain-specific loss-based selection therefore closely mirrors domain-specific selection, but emphasizes re-selection as a response to loss.

SOC across the life span

As a life-span theory, the SOC model is intricately linked to developmental gains and losses. Children experience fewer developmental declines than adults, and the individual SOC components are not thought to fully differentiate until adolescence (e.g., Gestsdottir & Lerner, 2007; Gestsdottir, Lewin-Bizan, von Eye, Lerner, & Lerner, 2009). In fact, loss-based selection may occur so rarely in childhood and adolescence that it has been omitted from many studies of adolescent SOC (e.g., Gestsdottir & Lerner, 2007; Gestsdottir et al., 2009; but see Gestsdottir, Bowers, von Eye, Napolitano, & Lerner, 2010). The SOC processes are still anticipated to be important for positive development throughout adolescence, however (e.g., Lerner et al., 2001), and continue in importance throughout adulthood.

Measuring SOC in young adults and late adolescents

Much of the existing SOC research relies on a measure developed by Baltes and colleagues (see Baltes, Baltes, Freund & Lang, 1999; Freund & Baltes, 2002). Their SOC questionnaire consists of 48 items (12 per SOC component) and is administered using a forced-choice format to reduce the correlation between SOC and measures of social desirability. The SOC questionnaire is framed by a developmental perspective that specifies loss as the loss of a previously successful means. Additionally, most researchers administer the SOC questionnaire in a domain-general format. While the SOC processes are context-dependent, a domain-general questionnaire allows researchers to avoid the complexity of measuring SOC in all potentially relevant contexts.

The SOC questionnaire was designed for use in older adults, and accurately measuring SOC in younger versus older adults may require scales with different measurement characteristics. Below we discuss some key differences between the original SOC questionnaire and our new measures.

Relevance to younger samples

The SOC model emphasizes loss as stemming directly from developmental declines. Accordingly, the SOC questionnaire measures both compensation and loss-based selection by emphasizing the loss of previously available means. Losses caused by developmental declines are relatively common in older populations, but are much less prevalent in younger populations (e.g., adolescents, young adults). The loss-based components of the SOC questionnaire (and indeed of the SOC model as a whole) therefore take on qualitatively different meanings in younger versus older populations.

The traditional conceptualization of loss can therefore be extended to include any failed attempt to optimize a goal, regardless of whether that failure was caused by developmental declines. For instance, Lerner and colleagues (2001) noted that “[y]outh must develop the ability to accept and learn from failures and from loss… These are compensation actions” (p. 37). Our new measures therefore emphasize that goal-relevant means are often “lost” when an individual realizes those means are not appropriate for reaching a target goal in a specific context, in addition to being lost due to developmental declines in ability.

Forced-choice format

The original SOC questionnaire is administered using a forced-choice format to reduce the correlation between SOC and measures of social desirability (Freund & Baltes, 2002). For example, one compensation item asks participants to choose between “When things don’t work the way they used to, I look for other ways to achieve them,” and “When things don’t work the way they used to, I accept things the way they are.” Selecting the first option indicates high compensation, while selecting the second option indicates low compensation.

A forced-choice format may indeed eliminate the correlation between SOC and social desirability (Freund & Baltes, 2002), but this format also presents several psychometric difficulties, such as limiting measurement precision. Furthermore, Uziel (2010) has argued that social desirability may capture a form of social self-regulation. Our new scales therefore deviate from the forced-choice format and instead present items using a Likert-type scale.

Domain generality

As discussed earlier, the SOC constructs are conditioned on context, and it is appropriate for researchers to tailor measurement of SOC to the contexts most germane to their research. It is not feasible to create domain-specific measures of SOC for every possible domain, however, and domain-specific measurement may be irrelevant in studies that specify domain-irrelevant outcomes (e.g., subjective well-being, successful aging). Designed partly to address these concerns, the SOC questionnaire is therefore domain-general, with researchers given the option to make the questionnaire domain-specific by altering the instructions given to participants (i.e., asking participants to answer the items as they relate to school, work, etc.). Domain-general scales cannot provide the level of nuance that domain-specific scales can offer, even when the domain-general items are set in a specific context. Domain-specific measures such as those that we present in this paper are therefore needed to complement the existing SOC questionnaire.

Methods

Participants

We recruited 152 undergraduate students from the psychology department subject pool of a large university in the Midwest region of the US. Participants were more or less equally divided across gender (54% female) and were predominantly Caucasian (5% African American, 11% Asian/Pacific Islander, 74% Caucasian, 6% Hispanic, 3% other). Participants had a mean age of 19.85 years (SD = 2.00, Range: 18.08, 29.83).

Measures

Domain-specific SOC

We created domain-specific measures of SOC by first reviewing the literature and determining how various aspects of the SOC constructs (i.e., selection, optimization, compensation, and loss-based selection) could be operationalized in the academic and social domains. We then generated a list of items that represented these facets and selected approximately four items per facet to be included in an initial item pool. Participants responded to all items from this initial pool, with respondents asked to “indicate how much each of the following items describes you and your beliefs,” on a seven-point Likert-type scale. Response options 1, 4, and 7 were labeled not at all, somewhat, and very much, respectively.

We examined bivariate relations (Kendall’s tau-b) among items from each facet and removed items that did not strongly relate to other same-facet items. Using the remaining items as a reduced item pool, we then selected those items and facets that most strongly represented the SOC constructs. We presented this reduced item pool to outside content experts who had previously published on SOC (e.g., Bowers, Gestsdottir, Geldhof, Nikitin, & von Eye, 2011; Napolitano, Bowers, Gestsdottir, & Chase, 2011). These experts categorized all remaining items as either strongly representing the intended SOC processes or poorly representing the intended SOC processes. All items that the outside content experts felt strongly represented the intended constructs were retained in our final set of items and were included in our analyses. This list of items is presented in Appendix 1. Our final item pool contained 28 academic items (five selection, 10 optimization, eight compensation, and five loss-based selection) and 28 social items (four selection, nine optimization, 10 compensation, and five loss-based selection).

Domain-general SOC

We used a short form of the SOC questionnaire (e.g., Freund & Baltes, 2002), administered using its forced-choice format, to measure domain-general SOC. We anticipated moderate positive relations between our new measures and the SOC questionnaire, particularly for measures of selection and optimization. Because our new measures allow for loss in perceived goal-relevant means while the original SOC questionnaire does not, we additionally anticipated weaker correlations between the original SOC questionnaire’s and our new measures of compensation and loss-based selection.

Individual subscales from the SOC questionnaire displayed extremely low reliability (Cronbach’s α for selection = .51; optimization = .20; compensation = .00; loss-based selection = .17). A follow-up confirmatory factor analysis (CFA) suggested a lack of differentiation between the selection and optimization constructs, a lack of differentiation between the compensation and loss-based selection constructs, and produced several non-significant factor loadings. These results suggest that participants in our study only differentiated between aspects of self-regulation related to developmental gains rather, than developmental declines. Both constructs displayed acceptable reliability (i.e., ω = .75 and .73 for the selection/optimization and compensation/loss-based selection constructs respectively). Composite reliability (ω) is a reliability estimate computed from factor analysis results and can be interpreted exactly as Cronbach’s α. In fact the two are mathematically equivalent under the assumption of essential tau equivalence. When this assumption is violated, as occurs when items have heterogeneous factor loadings, ω becomes a more consistent estimator of a scale’s actual reliability than α.

The factor structure suggested by the CFA departs from the expected structure of the SOC items. Given the very poor reliability indices of the four original SOC subscales versus the sound reliabilities of the two CFA constructs, analyzing the latent constructs suggested by CFA would be statistically preferable to analyzing the four unreliable scales. In order to compare our new scales with the original SOC questionnaire, however, we present two sets of results, one that considers the latent constructs identified by our CFA and another that includes scale scores corresponding to the hypothesized SOC processes.

Criterion items

Drawing on previous work by Gestsdottir and colleagues (e.g., Gestsdottir et al., 2009), we validated our SOC scales against criterion items that represent the five Cs of positive youth development (confidence, competence, character, caring, connection; see Lerner et al., 2005; Roth & Brooks-Gunn, 2003). We measured the five Cs of positive youth development with an adaptation of Neeman and Harter’s (1986) Self-Perception Profile for College Students, including additional scales where appropriate. The Self-Perception Profile for College Students presents dually worded items such that half of the options of a single Likert-type response scale indicate a positively worded choice, while the other half indicate a negative wording of the same choice. Like the forced-choice format of the original SOC questionnaire, the format of the Self-Perception Profile for College Students attempts to minimize correlations with social desirability. Because emerging evidence suggests that measures of social desirability in fact measure a form of self-regulation (Uziel, 2010), we placed one wording per item on a seven-point Likert-type scale (see also Wichstrøm, 1995).

Competence

We measured competence using the academic competence and social competence subscales of the Self-Perception Profile for College Students. Each scale contained four items, with example items being, “I do very well at my studies,” for academic competence and, “I feel I am socially accepted by many people,” for social competence. Both scales displayed acceptable reliability in this study (i.e., ω = .71 and .82 for the academic and social competence scales, respectively).

Connection

We used the close friendships scale from the Self-Perception Profile for College Students to measure connection, meaning our measure of connection specifically emphasized connection to one’s peers rather than a general measure of connection to one’s context. This choice mirrors our social SOC items, which also emphasize friendships. The close friendships scale contained four items, with an example item being, “I am able to make close friends I can really trust.” The close friendships scale displayed acceptable reliability in our study (ω = .82).

Character

The five Cs model of positive youth development describes character as having positive values and moral commitment (Roth & Brooks-Gunn, 2003), and we measured character using the morality subscale of the Self-Perception Profile for College Students. This scale contained four items, with an example being, “I usually do what is morally right.” An initial CFA (discussed later) indicated that only three of the four items should be used in our study, and these items displayed moderately low inter-item reliability (ω = .69), most likely due to having only three items in the scale.

Caring

We measured caring using the empathic concern subscale of Davis’ (1980) larger self-report empathy measure. The scale contains seven items, with an example item being, “When I see someone being taken advantage of, I feel kind of protective toward them.” This scale displayed acceptable reliability in our study (ω = .83).

Confidence

We measured confidence using Rosenberg’s (1979) self-esteem scale. This measure included 10 items, with an example being, “I feel that I am a person of worth.” The self-esteem scale displayed acceptable reliability in our study (ω = .93).

Social desirability

The forced-choice format of the SOC questionnaire was designed to reduce the relations between SOC and social desirability. Because our measures of domain-specific SOC do not adhere to the forced-choice format, we included the denial subscale of the 13-item version of the Marlowe-Crowne social desirability scale discussed by Ballard (1992). This subscale consists of eight items that are scored as true or false, an example item being, “I sometimes feel resentful when I don’t get my way.” The true/false scoring of these items is integral to their implementation as the items represent behaviors, thoughts, or feelings that most people feel/do occasionally. While graded endorsement of these items of the social desirability may reflect actual behavior, the social desirability literature suggests that categorical denial of these behaviors, thoughts, or feelings necessarily represents social desirability. We coded the social desirability items such that higher scores indicated higher levels of social desirability, and this scale displayed acceptable reliability in the present study (ω = .80).

Procedures

Participants received course credit for participation, with study participation strictly optional for all participants. Participants provided written informed consent then completed a computerized questionnaire in the presence of the lead author or one of his research assistants. Each questionnaire contained items from all of the scales described earlier, with a variant of the three-form planned missingness approach (e.g., Graham, Hofer, & MacKinnon, 1996) imposed to reduce participant fatigue. The traditional three-form planned missingness approach is implemented by dividing items from each subscale evenly (or as evenly as possible) into four groups (i.e., Groups X, A, B, and C). Researchers then create three questionnaire forms that contain all items from group X plus all items from two of the remaining groups (i.e., XAB, XAC, XBC). The three-form approach imposes approximately 25% missingness known a priori to be missing completely at random, meaning this missingness can be easily recovered using modern missing data techniques (e.g., Graham et al., 1996).

In our two-group variant of the three-form approach we created two sets of forms that targeted either academic or social SOC. We divided criterion measures according to the traditional three-form approach, with the domain-specific SOC items differentially distributed across the two groups. Forms targeting social SOC contained all social SOC items, with three-form planned missingness imposed on all other items (including the academic SOC scale). Similarly, forms targeting the academic domain contained all academic SOC items, with all remaining items divided among the three forms. We randomly ordered all items and further divided each of the six forms into A and B formats, where A and B formats contained the same items but in reverse order of each other. Based on the above planned missingness design and the number of participants who received each form of the questionnaire, we anticipated approximately 15% missingness which would be known to be missing completely at random due to the study design.

Analyses

We estimated CFA and structural equation models (SEM) using robust weighted least squares in Mplus version 6.11, treating all indicators as categorical. Robust weighted least squares produce unbiased parameter estimates when data are missing completely at random after conditioning on all predictors (Asparouhov & Muthén, 2010). Data from this study contained 18.13% missingness—slightly more than the amount of missingness anticipated by our planned missingness design—suggesting that the missing completely at random assumption was reasonable. We verified the missing completely at random assumption by implementing Little’s (1988) missing completely at random test using code provided by Enders (2012). The results of this analysis supported the missing completely at random assumption (χ² _{(df = 4996)} = 4918.912, p = .779), suggesting that missingness in our dataset did not likely impact the following results.

We next ran a series of confirmatory factor analyses (CFA) and structural equation models (SEM) that: (a) verified the factor structure of our domain-specific SOC measures; and (b) examined the relations between all indices of SOC and the indicators of positive youth development specified earlier. Model fit was gauged by the root mean square error of approximation (RMSEA), the Comparative Fit Index (CFI), and the Tucker–Lewis Index (TLI). RMSEA can be roughly interpreted as the amount of model misfit per degree of freedom, with values less than .08 considered acceptable. CFI and TLI values roughly represent the degree that a model fits better than an alternative null model, with values greater than .90 considered acceptable (see Marsh, Hau, & Wen, 2004 for a more detailed discussion of goodness of fit cutoff criteria).

Results

Initial CFAs

Our first analyses examined the factor structure of our SOC measures, with separate CFA models specified for the academic SOC variables, the social SOC variables, and for the original SOC questionnaire. An initial CFA of the academic SOC items produced acceptable model fit (χ² _{(df = 331)} = 570.886, p < .001; RMSEA: .07, 90% Confidence Intervals [CIs] [.06, .08]; CFI: .96; TLI: .95), although the estimated latent correlation between the selection and optimization constructs was greater than 1.00, suggesting that selection and optimization factors did not differentiate in our sample. Further, the modification indices suggested a strong cross loading for one compensation item onto the selection construct. We therefore combined the academic selection and optimization constructs (Δχ² _(df=3) = 12.22, p > .001) and removed the item with a dual factor loading. All academic SOC constructs displayed acceptable reliability in this model (selection/optimization: ω = .92; compensation: ω = .89; loss-based selection: ω = .79).

A similar model that examined the social SOC scale produced acceptable model fit (χ² _{(df = 358)} = 567.891, p < .001; RMSEA: .06, 90% CIs [.05, .07]; CFI: .92; TLI: .91), although the latent correlation between social optimization and compensation was estimated as greater than 1.00, again suggesting a lack of differentiation. These latent constructs were combined in all subsequent models (Δχ² _(df=3) = 15.15, p > .001), and all constructs displayed acceptable reliability in the subsequent model (selection: ω = .78; optimization/compensation: ω = .90; loss-based selection: ω = .77).

Similar to the results obtained for the domain-specific SOC measures, a CFA of the original domain-general SOC scale indicated acceptable model fit (χ² _{(df = 48)} = 51.844, p < .001; RMSEA: .02, 90% CIs [.00, .06]; CFI: .96; TLI: .94), with two latent correlations estimated above 1.00 and suggesting a lack of construct differentiation. Several factor loadings were also not significant in this model. Items with non-significant loadings were dropped from the model and the original SOC questionnaire’s compensation and loss-based selection constructs were combined into a single construct (Δχ² _(df=3) = 4.193, p > .001), as were the original questionnaire’s measures of optimization and selection (Δχ² _(df=2) = 3.784, p > .001).

The aforementioned CFA models all indicated a general lack of differentiation between the SOC constructs in this sample. The academic SOC measure indicated that optimal selection of academic goals is equivalent to optimizing those same goals. Similarly, optimization and compensation in the social domain were indistinguishable, indicating that participants who are good at optimizing friendships are similarly good at maintaining friendships when obstacles arise. Domain-general SOC also did not strongly differentiate, with participants differentiating only between facets of SOC related to developmental gains (i.e., selection and optimization) and facets relevant to developmental declines (i.e., compensation and loss-based selection). As Table 2 shows, even the same-domain facets of SOC that appeared to differentiate in our models remained highly correlated (e.g., social selection correlated .92 with social optimization/compensation). The exception to this trend was that the domain-specific loss-based selection items tended to correlate negatively with same-domain selection, optimization, and compensation.

We next examined the factor structure of all of the criterion scales, including our measure of social desirability, with all constructs modeled simultaneously. The initial model produced acceptable model fit (χ² _{(df = 758)} = 1100.693, p < .001; RMSEA: .06, 90% CIs [.05, .06]; CFI: .91; TLI: .90), although one item from the character scale displayed a non-significant factor loading. This item was dropped from all subsequent models.

Criterion relationships

Criterion correlations

We next ran two models that examined all constructs simultaneously. We first ran a CFA that included all latent constructs in a single model. This model displayed acceptable model fit (χ² _{(df = 5122)} = 5276.828, p <.001; RMSEA: .01, 90% CIs [.00, .02]; CFI: .94; TLI: .94), with factor loadings from this model presented in Table 1 and all latent correlations presented in the lower triangle of Table 2. We then ran a parallel structural equation model in which all latent constructs were regressed on the social desirability construct, such that the residual latent correlations represented the relations between all latent constructs after controlling for social desirability. These partial correlations are presented above the diagonal of Table 2.

Table 1.

Standardized SOC factor loadings from the full CFA model

Academic SOC
Selection/optimization
asel6	0.71
asel10	0.63
asel13	0.56
asel14	0.61
asel17	0.72
aopt1	0.40
aopt2	0.59
aopt5	0.82
aopt6	0.68
aopt12	0.73
aopt17	0.62
aopt20	0.74
aopt21	0.74
aopt23	0.70
aopt26	0.62
Compensation
acomp1	0.75
acomp3	0.77
acomp6	0.74
acomp7	0.71
acomp15	0.76
acomp22	0.60
acomp24	0.83
Loss-based selection
albs3	0.61
albs4	0.71
albs14	0.69
albs19	0.60
albs20	0.64
Social SOC
Selection
ssel6	0.36
ssel8	0.54
ssel13	0.69
ssel17	0.80
ssel18	0.80
Optimization/compensation
sopt2	0.48
sopt4	0.62
sopt11	0.61
sopt13	0.59
sopt17	0.53
sopt21	0.58
sopt23	0.62
sopt24	0.63
sopt28	0.63
scomp1	0.64
scomp4	0.43
scomp6	0.53
scomp7	0.52
scomp10	0.58
scomp11	0.52
scomp17	0.54
scomp20	0.55
scomp27	0.79
scomp29	0.54
Loss-based selection
slbs2	0.73
slbs4	0.50

slbs12	0.61
slbs23	0.58
slbs24	0.72
Original SOC questionnaire
Selection/optimization
sel4	0.68
sel6	0.53
opt3	0.82
opt4	0.58
Compensation/loss-based selection
comp2	0.30
comp3	0.58
lbs2	0.91
lbs3	0.70
Competence
Academic competence
acomp1	0.71
acomp2	0.82
acomp3	0.33
acomp4	0.55
Social competence
scomp1	0.62
scomp2	0.76
scomp3	0.81
scomp4	0.71
Connection
connect1	0.76
connect2	0.79
connect3	0.68
connect4	0.67
Character
char1	0.42
char2	0.56
char4	0.93
Caring
care1	0.76
care2	0.70
care3	0.65
care4	0.45
care5	0.65
care6	0.57
care7	0.73
Confidence
conf1	0.83
conf2	0.74
conf3	0.78
conf4	0.70
conf5	0.73
conf6	0.69
conf7	0.78
conf8	0.73
conf9	0.71
conf10	0.80
Social desirability
sd1	0.63
sd2	0.75
sd3	0.39
sd4	0.55
sd5	0.59
sd6	0.27
sd7	0.73
sd8	0.70

Table 2.

Full and partial latent correlations between all modeled constructs

	1.	2.	3.	4.	5.	6.	7.	8.	9.	10.	11.	12.	13.	14.	15.
1. Academic selection/optimization	1.00	.82***	−.12	.58***	.55***	−.06	.68***	.67*	.83***	.46***	.39***	.58***	.41***	.43***	n.a.
2. Academic compensation	.80***	1.00	.15	.46***	.53***	.71***	.40**	.52*	.43***	.34***	.44***	.53***	.44***	.37***	n.a.
3. Academic LBS	−.28***	.06	1.00	−.07	−.12	.09	−.25	−.27	−.20	−.13	−.06	−.23*	−.03	−.22*	n.a.
4. Social selection	.61***	.48***	−.15	1.00	.92***	.05	.14	.69***	.60***	1.05^†	.67***	.66***	.56***	.66***	n.a.
5. Social optimization/compensation	.62***	.55***	−.24***	.92***	1.00	−.19*	.01	.74***	.43***	.72***	.57***	.70***	.68***	.51***	n.a.
6. Social LBS	−.13	.05	.71***	.01	−.24**	1.00	−.00	−.45*	−.17	−.17	−.13	−.12	−.24**	−.23**	n.a.
7. Original selection/optimization	.74***	.43***	−.37**	.22	.15	−.07	1.00	.60*	.26	.10	.11	.33*	−.05	.23	n.a.
8. Original compensation/LBS	.64**	.53**	−.29	.69***	.73***	−.46**	.59**	1.00	.60*	.74***	.18*	.81**	.68***	.62***	n.a.
9. Academic compensation	.86***	.46***	−.35***	.63***	.52***	−.23*	.40***	.58*	1.00	.50***	.49***	.34**	.34**	.61***	n.a.
10. Social compensation	.53***	.38***	−.25**	1.04^†	.75***	−.22*	.22	.73***	.57***	1.00	.63***	.55***	.57***	.78***	n.a.
11. Connection	.48***	.46***	−.18	.69***	.62***	−.17	.23	.39*	.56***	.66***	1.00	.39***	.41***	.55***	n.a.
12. Character	.64***	.55***	−.34***	.68***	.73***	−.17	.43**	.79***	.45***	.60***	.46***	1.00	.48***	.55***	n.a.
13. Caring	.44***	.46***	−.11	.58***	.70***	−.27**	.04	.69***	.39***	.59***	.44***	.51***	1.00	.39***	n.a.
14. Confidence	.54***	.41***	−.35***	.68***	.58***	−.28**	.36**	.61***	.68***	.80***	.61***	.61***	.43***	1.00	n.a.
15. Social desirability	.45***	.17	−.41***	.23*	.34**	−.17	.42**	.11	.48***	.33**	.32**	.35**	.19	.40***	1.00

Note. * p < .05; ** p < .01; *** p < .001; † p < .001; constraining these coefficients to 1.00 significantly decreased model fit (i.e., p for both Δχ² tests < .001).

LBS = loss-based selection; bottom triangle represents raw latent correlations. Top triangle represents partial latent correlations, controlling for social desirability. Because social desirability is controlled for in the top triangle, the covariances between social desirability and all other constructs is replaced by n.a.

Because the structure of the original SOC questionnaire departed from its theoretically anticipated structure, we examined a parallel set of CFA models that replaced latent constructs derived from the original SOC questionnaire, with theoretically appropriate, but statistically unreliable, single-item indicators created from each of the four theoretically anticipated scale scores. We provide correlations between these scale scores and all latent constructs in Table 3.

Table 3.

Full and partial correlations between scale scores from the original SOC questionnaire and all other modeled constructs

	Social desirability not controlled				Social desirability controlled
	S	O	C	LBS	S	O	C	LBS
Academic selection/optimization	.25**	.48***	.14	.14*	.19*	.48***	.17*	.14
Academic compensation	.03	.32***	.07	.24**	−.01	.31***	.09	.22**
Academic LBS	−.16	−.16	−.19*	−.03	−.10	.09	−.22*	.02
Social selection	.07	.10	.20*	.19*	.02	.07	.22*	.17
Social optimization/compensation	−.04	.11	.16*	.26**	−.12	.08	.18*	.24**
Social LBS	.04	−.02	.05	−.14	.08	−.00	.05	−.12
Academic compensation	.06	.37***	.04	.15	−.03	.36***	.07	.11
Social compensation	.01	.12	.02	.18*	−.06	.04	.04	.16
Connection	.04	.11	−.08	.19*	−.03	.08	−.07	.17
Character	.06	.20*	.19*	.17	−.01	.17	.22*	.14
Caring	−.07	.07	.05	.15	−.13	.04	.06	.13
Confidence	.11	.17*	.15*	.23**	.04	.14	.19*	.21**
Social desirability	−.18	−.11	.04	−.11	n.a.	n.a.	n.a.	n.a.

Note. * p < .05; ** p < .01; *** p < .001; LBS = loss-based selection.

Criterion regressions

The latent correlations from our full CFA models indicate strong positive relations among the domain-specific and domain-general SOC measures, and indicate moderate to strong positive correlations between the measures of SOC and the five Cs of positive youth development. The exception to this general trend was that domain-specific loss-based selection measures weakly and negatively correlated with the other SOC constructs and with the positive youth development measures, while the two domain-specific loss-based selection constructs correlated positively with each other.

We ran a series of SEMs to clarify the latent relationships presented in Tables 2 and 3, and to determine if our domain-specific measures of SOC predicted the outcome measures above and beyond the variance explained by the original SOC questionnaire. We ran 10 separate models that included the latent constructs derived from the original SOC questionnaire and a parallel set of models that instead considered scale scores derived from the original SOC questionnaire. Each model included the original SOC measure, either the academic or the social SOC scales as predictors, and one of the five Cs of positive youth development. We controlled all constructs for social desirability and all models displayed acceptable fit (i.e., RMSEAs < .08; CFIs ≥ .90; TLIs ≥ .90). Non-significant regression coefficients were pruned from the models one at a time (p for all single degree of freedom Δχ² tests > .001).

Table 4 presents results from our latent regression models when the original SOC questionnaire was modeled using latent constructs, while Table 5 presents the results from parallel analyses of the scale-score representations of the original SOC questionnaire’s subscales. As these tables show, scale-score representations of the original SOC questionnaire had very little predictive ability after controlling for our new measures of SOC. The analytically derived latent constructs from the original SOC questionnaire were more predictive than the theoretically anticipated scale scores, and our discussion focuses on those results.

Table 4.

Criteria regressed on the academic and social SOC and original SOC scale represented as latent constructs Criteria regressed on the academic SOC and original SOC scale represented as latent constructs

		Academic competence	Social competence	Connection	Character	Caring	Confidence
Original SOC	Selection/optimization	−0.319**	.	.	.	−0.783**	.
	Compensation/LBS	.	0.605***	.	0.826***	1.234***	0.526***
Academic SOC	Selection/optimization	1.489***	.		.	.	.
	Compensation	−0.553***	.	0.380***	.	.	.
	LBS	.	.	.	.	.	.
Criteria regressed on the social SOC and original SOC scale represented as latent constructs
		Academic competence	Social competence	Connection	Character	Caring	Confidence
Original SOC	Selection/optimization	.	.	.	.	−0.546**	.
	Compensation/LBS	.	.		0.593***	0.960***	.
				.
Social SOC	Selection	0.494***	1.304***	0.640***	.	.	0.527***
	Optimization/compensation	.	−0.488***	.	.	.	.
	LBS	.	−0.246**	.	.	.	−0.207**

Note. * p < .05; ** p < .01; *** p < .001; LBS = loss-based selection.

Table 5.

Criteria regressed on the academic and social SOC and original SOC scale represented as scale scores Criteria regressed on the academic SOC and original SOC scale represented as scale scores

		Academic competence	Social competence	Connection	Character	Caring	Confidence
Original SOC	Selection	−0.262**	.	.	.	.	.
	Optimization	.	.	.	.	.	.
	Compensation	.	.	.	.	.	.
	LBS	.	.	.	.	.	.
Academic SOC	Selection/optimization	1.411***	0.398***		.	.	.
	Compensation	−0.599***	.	0.384***	0.569***	0.442***	0.399***
	LBS	.	.	.	0.389***	.	−0.234**
Criteria regressed on the social SOC and original SOC scale represented as scale scores
		Academic competence	Social competence	Connection	Character	Caring	Confidence
Original SOC	Selection	.	.	.	.	.	.
	Optimization	0.326***	.	.	.	.	.
	Compensation	.	.	−0.196*	.	.	.
	LBS	.	.	.	.	.	.
Social SOC	Selection	0.552***	1.359***	0.694***	0.646***	.	1.497*
	Optimization/compensation	.	−0.529***	.	.	0.67***	−1.013^†
	LBS	−0.254**	−0.312***	.	−0.206*	.	−0.490**

Note. * p < .05; ** p < .01; *** p < .001; † n.s. per Wald statistic but p < .05 per LRT test; LBS = loss-based selection.

The original domain-general SOC questionnaire only weakly predicted domain-specific competence when same-domain SOC was included as a predictor in the model. Similarly, domain-general SOC did not significantly predict connection when either domain-specific SOC scale was included in the model, but did significantly predict character and caring. Neither domain-specific measure of SOC significantly predicted either character or caring when domain-general SOC was included in the model, however. Finally, confidence was predicted by domain-general SOC when academic SOC was included as a predictor in the model, but domain-general SOC did not predict confidence when social SOC was modeled. These results show that domain-specific SOC predicted same-domain competence as well as connection and, in the social domain, confidence above and beyond the relations between these outcomes and domain-general SOC as measured by the original SOC questionnaire. Domain-general SOC predicted character and caring above and beyond domain-specific SOC, however, indicating that our domain-specific measures of self-regulation complement, not replace, the domain-general SOC questionnaire.

Discussion

This study presented two domain-specific measures of SOC designed to measure self-regulation in young adults and late adolescents. We then provided support for our scales’ internal and external validity. These new scales therefore supplement the measurement of SOC processes in young adults and can therefore be used in conjunction with existing scales (e.g., the existing SOC questionnaire) to further our understanding of self-regulation during this period of the life span. We summarize our findings below.

Internal validity

We established internal validity by asking content experts on our own team to generate an initial item pool and by consulting outside content experts when selecting our final item pool. We then tested the internal consistency of our scales, with confirmatory factor analysis displaying acceptable model fit for both the academic and social SOC scales. These analyses indicated only weak differentiation of the selection, optimization, and compensation constructs in our sample, with some constructs not being differentiable and the correlation between differentiable and non-differentiable SOC constructs remaining quite high. This general lack of differentiation was also apparent for the constructs indicated by the domain-general SOC questionnaire, contrasting previous research on older adults (e.g., Freund & Baltes, 2002) but mirroring research on adolescents. For instance, Gestsdottir and colleagues (2009) were, “unable to provide evidence against or in support of … differentiation,” when examining adolescents as old as 16 (p. 591). These findings therefore suggest the need to further examine the structure of SOC across adolescence and early adulthood to better determine whether this lack of differentiation represents an actual developmental process, is a function of the existing self-report measures, or simply represents a function of the samples examined in these studies.

Future research that clarifies how the SOC constructs differentiate within specific domains may also help explain why our scales displayed different patterns of differentiation across domains. Future research will therefore clarify whether the inconsistent differentiation that we observed in this study represents, among other possibilities: (a) an unwanted artifact of our scale; (b) domain-specific developmental phenomena; (c) an artifact of our limited sample; or (d) developmentally stable domain-specific factor structures. A more detailed understanding of how and whether the constructs differentiate in various domains will also clarify how and whether researchers should compare different aspects of self-regulation across various domains.

Contrasting the general lack of differentiation found for the selection, optimization, and compensation constructs, our domain-specific loss-based selection constructs displayed weak negative correlations with other same-domain SOC constructs and showed a strong positive correlation with each other. This finding was unexpected, given the positive correlation between loss-based selection and the other SOC constructs previously found in both adolescent and adult samples (e.g., Freund & Baltes, 2002; Gestsdottir et al., 2010). Our study also found positive relations between loss-based selection and the other SOC constructs when all constructs were measured using the original SOC questionnaire, suggesting that loss-based selection as measured by our domain-specific scales is qualitatively different than the same construct measured by the original SOC questionnaire.

Aside from domain specificity and a Likert-type response scale, the primary difference between our domain-specific measures of loss-based selection and that of the original SOC questionnaire is that our measures included failure as a mechanism of loss. We argued earlier that adolescents and young adults likely experience few developmental declines, making responses to failure a more appropriate instantiation of “loss” in this age group. A proclivity toward seeing obstacles as insurmountable (i.e., a failed goal that requires loss-based selection) is more akin to reduced personal agency than to the adaptive coping mechanism that loss-based selection is described as in the SOC literature. If researchers continue to treat failure as an instantiation of “loss” relevant to adolescents and young adults, then our results suggest that future work should more fully explore developmental changes in loss-based selection across the lifespan, both at the quantitative level and as its definition changes across the lifespan. Alternatively, these results may indicate that “failure” and “loss” represent qualitatively distinct constructs in this age group, and subsequent work should determine how, precisely, responses to failure relate to the SOC framework.

External validity

After establishing internal validity we established the external validity of our domain-specific scales through a series of structural equation models. These models directly compared the predictive power of our domain-specific SOC measures against that of the original SOC questionnaire by including both as predictors of the five Cs of positive youth development and controlling all constructs for social desirability. Controlling for social desirability was important to ensure that any differences between our domain-specific measures and the original SOC questionnaire were not due to social desirability bias, which the SOC questionnaire was designed to reduce. Our SEMs supported the external validity of our domain-specific scales while highlighting the fact that measures of domain-specific self-regulation should complement, not replace, domain-general measures. The original SOC questionnaire did not significantly predict competence, connection, or confidence (i.e., self-esteem) after controlling for our domain-specific measures, while our domain-specific measures did not significantly predict character or caring after controlling for the original SOC questionnaire constructs.

While our latent regression models generally supported the validity of our domain-specific SOC measures, two aspects of our results deserve further discussion. First, several significant predictive relations are in the opposite direction to that anticipated by theory. Academic compensation negatively predicted academic competence, for instance. While at first glance these relations appear to contradict the validity of our domain-specific scales, it is likely that the negative relations are an artifact of the very strong positive correlations among same-domain SOC constructs. As the correlations in Table 2 show, academic competence positively correlated with academic compensation, meaning the negative relation just described only existed after controlling for academic selection/optimization, which was a much stronger predictor of academic competence than was academic compensation.

The second aspect of our latent regressions that deserves attention is the fact that our domain-specific measures of selection were very strong predictors of positive youth development. Previous research examining the original SOC questionnaire in adolescents has tended to find weak relations between positive youth development and selection, a finding also supported by the relations between positive youth development and the original SOC questionnaire in our present study. This discrepancy suggests that participants’ ratings of their own general ability to select goals do not predict positive development, whereas successful goal selection within a (presumably important) domain does. The especially strong relations between domain-specific selection and same-domain competence therefore imply that effective selection within a domain implies competent goal selection in that domain. Future research on the relations between domain-specific versus domain-general goal selection and competence is therefore warranted.

Limitations

Despite this study’s general support for the validity of our domain-specific SOC measures, our study is not without limitations. The most significant limitation is that we only obtained data from a small, homogeneous, convenience sample of college students recruited from the Midwest region of the US. The generalizability of our domain-specific SOC measures therefore remains unexplored. Future work is needed to determine the validity of these scales across various social contexts and across various periods of the lifespan.

Relatedly, these scales only represent a first step toward measuring domain-specific SOC. Further theoretical work must pin down what exactly constitutes each SOC process in specific domains of interest, with greater theoretical refinement making room for much more refined measures. The scales proposed in this paper, especially the social SOC scale, will therefore benefit from continued refinement of the relevant theoretical concepts.

One additional limitation is that we compared our domain-specific SOC scales to the original SOC questionnaire, where the latter was administered as domain-general. This limitation makes it impossible to disentangle whether the differences between our scales and the original SOC questionnaire are due to differences in domain specificity or are due to differences in item format. While the difficulties associated with both domain-general measures of self-regulation and with a forced-choice format are detailed earlier, a greater understanding of the empirical effects of each could greatly enhance our understanding of self-regulation as a whole.

Final remarks

Self-regulation has been associated with positive developmental outcomes across the lifespan (Geldhof et al., 2010; McClelland et al., 2010) and as a result has been the subject of much psychological inquiry. Our understanding of self-regulation remains limited by domain-general operationalizations, and in this paper we presented two domain-specific measures of self-regulation based on Baltes and colleagues’ SOC model. Our intent was to encourage a deeper consideration of the context when studying self-regulatory processes and to help researchers align the measurement of self-regulation with the widely accepted importance of context.

Footnotes

Acknowledgments

Special thanks to Tara Anderson, Edmond Bowers, Brent McPherson, Patrick Miller, Chris Napolitano, Sunthud Pornprasertmanit, Danielle Sorell, Samantha Tavernaro, and Lyndsey Wheeler, who assisted with the preparation of this manuscript. This manuscript is based on research performed by the first author at the University of Kansas.

Funding

This manuscript was supported by dissertation awards to the first author from the University of Kansas and the Society of Multivariate Experimental Psychology.

Appendix 1

References

Asparouhov

Muthén

B. O.

(2010). Weighted least squares estimation with missing data (Technical report). Retrieved from http://www.statmodel.com/download/GstrucMissingRevision.pdf.

Bäckman

Dixon

R. A.

(1992). Psychological compensation: A theoretical framework. Psychological Bulletin, 112, 259–283.

Ballard

(1992). Short forms of the Marlowe-Crowne social desirability scale. Psychological Reports, 71, 1155–1160.

Baltes

P. B.

Baltes

M. M.

(1990). Psychological perspectives on successful aging: The model of selective optimization with compensation. In Baltes

P. B.

Baltes

M. M.

(Eds.), Successful aging: Perspectives from the behavioral sciences (pp. 1–34). New York, NY: Cambridge University Press.

Baltes

P. B.

Baltes

M. M.

Freund

A. M.

Lang

(1999). The measurement of selection, optimization, and compensation (SOC) by self report: Technical report 1999 (Research Report No. 66). Berlin, Germany: Max Planck Institute for Human Development.

Bowers

E. P.

Gestsdottir

Geldhof

G. J.

Nikitin

von Eye

(2011). Developmental trajectories of intentional self regulation in adolescence: Implications for positive and problematic development among diverse youth. Journal of Adolescence, 34, 1193–1206.

Brandstädter

(1998). Action perspectives on human development. In Damon

(Series Ed.) &

Lerner

(Vol. Ed.), Handbook of child psychology: Vol. 1. Theoretical models of human development (pp. 1029–1144). New York, NY: Wiley.

Brandtstädter

(2006). Action perspectives on human development. In Damon

Lerner

R. M.

(Editors-in-Chief) &

Lerner

R. M.

(Vol. Ed.), Handbook of child psychology. Vol. 1: Theoretical models of human development (6th ed., pp. 516–568). Hoboken, NJ: Wiley.

Davis

M. H.

(1980). A multidimensional approach to individual differences in empathy. JSAS Catalog of Selected Documents in Psychology, 10, 85.

10.

Enders

C. K.

(2012). Little’s (1988) Test of missing completely at random (MCAR) missing data [Computer software]. Retrieved from http://psychology.clas.asu.edu/files/LittlesTestofMCARMissingData.sas

11.

Freund

A. M.

Baltes

P. B.

(2000). The orchestration of selection, optimization, and compensation: An action-theoretical conceptualization of a theory of developmental regulation. In Perrig

W. J.

Grob

(Eds.), Control of human behavior, mental processes and consciousness: Essays in honor of the 60th birthday of August Flammer (pp. 35–58). New York, NY: Erlbaum.

12.

Freund

A. M.

Baltes

P. B.

(2002). Life-management strategies of selection, optimization, and compensation: Measurement by self-report and construct validity. Journal of Personality and Social Psychology, 82, 642–662.

13.

Freund

A. M.

K. Z. H.

Baltes

P. B.

(1999). Successful development and aging: The role of selection, optimization, and compensation. In Brandtstädter

Lerner

R. M.

(Eds.), Action & self-development: Theory and research through the life span (pp. 401–434). Thousand Oaks, CA: SAGE.

14.

Geldhof

G. J.

Little

T. D.

Colombo

(2010). Self-regulation across the lifespan. In Lamb

M. E.

Freund

A. M.

(Vol. Eds.) and

Lerner

R. M.

(Editor-in-Chief), Social and emotional development. Vol 2: The handbook of lifespan development (pp. 116–157). Hoboken, NJ: Wiley.

15.

Gestsdottir

Bowers

von Eye

Napolitano

C. M.

Lerner

R. M.

(2010). Intentional self regulation in middle adolescence: The emerging role of loss-based selection in positive youth development. Journal of Youth and Adolescence, 39, 764–782.

16.

Gestsdottir

Lerner

R. M.

(2007). Intentional self-regulation and positive youth development in early adolescence: Findings from the 4-H study of positive youth development. Developmental Psychology, 43, 508–521.

17.

Gestsdottir

Lerner

R. M.

(2008). Positive development in adolescence: The development and role of intentional self-regulation. Human Development, 51, 202–224.

18.

Gestsdottir

Lewin-Bizan

von Eye

Lerner

J. V.

Lerner

R. M.

(2009). The structure and function of selection optimization and compensation in middle adolescence: Theoretical and applied implications. Journal of Applied Developmental Psychology, 30, 585–600.

19.

Gestsdottir

Urban

J. B.

Bowers

E. P.

Lerner

J. V.

Lerner

R. M.

(2011). Intentional self-regulation, ecological assets, and thriving in adolescence: A developmental systems model. In Lerner

R. M.

Lerner

J. V.

Bowers

E. P.

Lewin-Bizan

Gestsdottir

Urban

J. B.

(Eds.), Thriving in childhood and adolescence: The role of self-regulation processes. New Directions for Child and Adolescent Development, 133, 61–76.

20.

Graham

J. W.

Hofer

S. M.

MacKinnon

D. P.

(1996). Maximizing the usefulness of data obtained with planned missing value patterns: An application of maximum likelihood procedures. Multivariate Behavioral Research, 31, 197–218.

21.

Lerner

R. M.

Busch-Rossnagel

N. A.

(1981). Individuals as producers of their development: A lifespan perspective. New York, NY: Academic Press.

22.

Lerner

R. M.

Freund

A. M.

De Stefanis

Habermas

(2001). Understanding developmental regulation in adolescence: The use of the selection, optimization, and compensation model. Human Development, 44, 29–50.

23.

Lerner

R. M.

Lerner

J. V.

Almerigi

J. B.

Theokas

Phelps

Gestsdottir

von Eye

(2005). Positive youth development, participation in community youth development programs, and community contributions of fifth-grade adolescents: Findings from the first wave of the 4-H study of positive youth development. The Journal of Early Adolescence, 25, 17–71.

24.

Little

R. J. A.

(1988). A test of missing completely at random for multivariate data with missing values. Journal of the American Statistical Association, 83, 1198–1202.

25.

Marsh

H. W.

Hau

Wen

(2004). In search of golden rules: Comment on hypothesis-testing approaches to setting cutoff values for fit indexes and dangers in overgeneralizing Hu and Bentler’s (1999) findings. Structural Equation Modeling, 11, 320–341.

26.

McClelland

M. M.

Ponitz

C. C.

Messersmith

Tominey

(2010). Self-regulation: The integration of cognition and emotion. In Lerner

(Series Ed.) &

Overton

(Vol. Ed.), Handbook of life-span development. Vol. 1: Cognition, biology and methods (pp. 509–553). Hoboken, NJ: Wiley & Sons.

27.

Napolitano

C. M.

Bowers

E. P.

Gestsdottir

Chase

P. A.

(2011). The development of intentional self-regulation in adolescence: Describing, explaining, and optimizing its link to positive youth development. In Lerner

R. M.

Lerner

J. V.

Benson

J. B.

(eds.), Advances in child development and behavior: Positive youth development. Boston, MA: Elsevier.

28.

Neeman

Harter

(1986). Manual for the self-perception profile for college students. Denver, CO: University of Denver.

29.

Rosenberg

(1979). Conceiving the self. New York, NY: Basic Books.

30.

Roth

J. L.

Brooks-Gunn

(2003). What exactly is a youth development program? Answers from research and practice. Applied Developmental Science, 7, 94–111.

31.

Schunk

D. H.

Zimmerman

B. J.

(Eds.). (1994). Self-regulation of learning performance: Issues and educational applications. Hillsdale, NJ: Lawrence Erlbaum Associates.

32.

Uziel

(2010). Rethinking social desirability scales: From impression management to interpersonally oriented self-control. Perspectives on Psychological Science, 5, 243–262.

33.

Vohs

K. D.

Ciarocco

N. J.

(2004). Interpersonal functioning requires self-regulation. In Baumeister

R. F.

Vohs

K. D.

(Eds.), Handbook of self-regulation: Research, theory, and applications (pp. 392–410). New York, NY: Guilford Press.

34.

Wichstrøm

(1995). Harter’s self-perception profile for adolescents: Reliability, validity, and evaluation of the question format. Journal of Personality Assessment, 65, 100–116.