Original Paper

The Adolescent Self-Regulatory Inventory: The Development and Validation of a Questionnaire of Short-Term and Long-Term Self-Regulation

Kristin L. Moilanen1

(1) Department of Psychology, University of Pittsburgh, 210 South Bouquet Street, Pittsburgh, PA 15260, USA

Kristin L. MoilanenEmail: klm73@pitt.edu

Received: 19 August 2005 Accepted: 1 December 2005 Published online: 25 July 2006

Abstract

This manuscript presents a study in which the factor structure and validity of the Adolescent Self-Regulatory Inventory (ASRI) were examined. The ASRI is a theoretically-based questionnaire that taps two temporal aspects of self-regulation (regulation in the short- and long-term). 169 students in the 6th, 8th, and 10th grades of a small, Midwestern school district completed self-report questionnaires focused on self-regulation, parenting behaviors, and psychological adjustment. 80 parents also participated. Confirmatory factor analyses demonstrated that the internal consistency of the long-term and short-term factors was satisfactory. Requirements for concurrent and construct validity were met. The ASRI also demonstrated incremental validity, as the inclusion of the long-term factor with a comparison questionnaire significantly increased the proportion of explained variance in adolescent-reported parental warmth, externalizing, and prosocial behavior. The ASRI has the potential to move research on self-regulation in adolescence in a viable new direction.

Keywords Self-regulation - Self control - Parenting - Internailizing - Externalizing - Academic achievement - Prosocial behavior

One important protective factor that may help to prevent youth from engaging in risky behavior or help adolescents avoid outcomes associated with risky behavior is self-regulation (Jessor and Jessor, 1977). Self-regulation is the ability to flexibly activate, monitor, inhibit, persevere and/or adapt one’s behavior, attention, emotions and

cognitive strategies in response to direction from internal cues, environmental stimuli and feedback from others, in an attempt to attain personally-relevant goals (Barkley, 1997; Demetriou, 2000; Finkenauer et al., 2005; Lengua, 2003; Novak and Clayton, 2001; Thompson, 1994). One major shortcoming of most studies in this area involves the techniques employed to measure self-regulation. In these studies, researchers have primarily focused on regulation in the immediate or short-term context, and have been unable to examine theoretically-vital longer-term components of self-regulation that develop in the teenage years (Demetriou, 2000). This gap is addressed in the current manuscript, the primary goal of which is to examine the validity of a new measure of long and short-term self-regulation.

Distinguishing between short- and long-term temporal frames is crucial because of changes taking place in adolescence. Theorists posit that adolescents can regulate their actions and emotions in the immediate or short-term context in order to attain long-term goals (Demetriou, 2000). Barkley (1997) suggests that the understanding of time is a key element of self-regulation: adolescents are able to plan or prepare for events that are both near and distant in time, while children are limited to events that are temporally near. Growth of the prefrontal cortex in adolescence is thought to be responsible for developments in hindsight and forethought (Barkley, 1997). These changes in reflective abilities enable teenagers to evaluate their past actions and developing plans before hatching new schemes. Furthermore, teenagers’ orientation toward the future tends to increase with age, and appears to be complete only in late adolescence (Nurmi, 1991).

Although theory posits that the span of time over which an individual self-regulates increases with age, these claims have remained largely unexamined and have been widely neglected in adolescent self-regulation research. This may be due to widespread reliance on relatively few existing measures of self-regulation, primarily paper-and-pencil questionnaires. Shorter-term aspects of self-regulation are typically assessed in existing measures. Short-term self-regulation is operationalized as impulse, attentional or emotional control in the “heat of the moment,” or regulation in the immediate context. For example, children or adolescents might meet a momentary goal to remember a phone number by using a strategy such as rehearsal. Alternately, this can also take the form of squelching inappropriate behavior or emotions before they are enacted, such as inhibiting fidgeting. In contrast, long-term self-regulation involves the control of impulses or direction of effort over a longer period of time. This duration may last several weeks, months or years. For example, an adolescent might save their wages from an after-school job for many weeks or months in order to buy an expensive digital music device. Long-term self-regulation may also involve substantial planning, such as plotting a course of study in college in order to meet career goals. This aspect of regulation is not consistently represented in measures of self-regulation used in adolescence research.

Although a variety of instruments have been used in research with teenagers, existing questionnaires tend to be developmentally-limited in any of the following three ways. First, items may focus exclusively on regulation in an immediate or short-term context. For example, the “inhibit” subscale in adolescent self-report version of the Behavior Rating Inventory of Executive Function includes items like “I have problems waiting my turn” and “I interrupt others” (Gioia et al., 2000). Second, items may not specify a temporal context or may ambiguous in this regard. For example, items from the Self-Regulation Scale do not clearly specify a time frame (e.g., “After an interruption, I don’t have any problem resuming my concentrated style of working;” Luszczynska et al., 2004). Finally, individual items may be collapsed into scales without regard to their temporal nature. For example, the inhibitory control subscale of the Early Adolescent Temperament Questionnaire includes items tapping momentary control (e.g., “When I’m excited, it’s hard for me to wait my turn to talk”) and control over extended time (e.g., “I could easily change a bad habit if I wanted to;” Capaldi and Rothbart, 1992). Designing a measure capable of yielding information outside of a short-term context is a necessary step toward a full exploration of the developmental nature of self-regulation in adolescence.

In addition to being couched in a temporal context, self-regulation is comprised by several systems, which are the skills that make regulation possible. These systems are thoroughly described by Barkley’s hybrid model (1997, 2004), which focuses primarily on the regulation of behavior but can also be applied to other domains of regulation. In this theory, self-regulatory actions, referred to as “motor control systems” (also known as “motor control/fluency/syntax;” Barkley, 1997, 2004), are the observable outputs of executive function, which occurs in conjunction with spontaneous behavior inhibition. In other words, inhibiting an action or reaction for a span of time will permit children to think about possible actions purposefully (e.g., in accordance with hindsight, forethought, moral reasoning) before acting in accordance with their experiences, beliefs, and goals. Barkley (1997, 2004) includes 8 motor control systems in this model. These included monitoring (“sensitivity to response feedback”) and persevering (“goal-directed persistence”), which for the purposes of the current study are operationalized consistently with Barkley’s model (1997, 2004). The six remaining systems of Barkley’s model were grouped into 3 broad dimensions in order to reduce item redundancy. These included activating (“executing of goal-directed responses” and “execution of novel/complex motor sequences”), adapting (“behavioral flexibility” and “task re-engagement following disruption”), and inhibiting (“inhibition of task-irrelevant responses” and “control of behavior by internally-represented information;” Barkley, 1997, p. 191). By adolescence, youth should have mastered these components sufficiently to be able to use each interchangeably and appropriately for the given context. Thus, these 5 processes were represented in the self-regulation measure designed for this study.

This study is also based on the theoretical position that the multiple domains of self-regulation are interrelated. In the face of difficulty, children and adolescents use multiple strategies to simultaneously regulate their affect, behavior, and attention (Committee on Integrating the Science of Early Childhood Development, Shonkoff, and Phillips, 2000; Diamond and Aspinwall, 2003; Kopp, 1982). For example, in order for an adolescent to comply with parents’ requests to study for a test instead of going out with friends, the teenager must be able to calm himself in the face of tension between his own and his parents’ desires (emotion regulation), focus his attention on test preparation (attention regulation), and use physical strategies to avoid temptation, such as turning off his mobile phone (behavioral regulation). This perspective better reflects the complexity necessary to explain self-regulation in adolescence and adulthood. By adolescence, distinctions between different types of stimuli being regulated may be largely artificial, as these multiple aspects of regulation may reflect a basic, underlying ability manifested in several ways: teenagers and adults should be capable of regulating their thoughts, feelings, attention and behavior in a planful, goal-oriented way. Thus, the measure designed for the current study taps multiple aspects of self-regulation, including the behavioral, attentional, emotional and cognitive domains.

Finally, this measure also focuses on the products of self-regulation, meaning the results of self-regulatory success or failure (e.g., making mistakes after working too fast; Humphrey, 1982), versus the strategies or specific techniques for regulation (e.g., the degree to which a teenager uses cognitive restructuring in order to manage emotion: Silk et al., 2003). In summary, this new measure taps teenagers’ self-regulatory success or failure through the inclusion of 5 components (i.e., monitoring, activating, adapting, persevering and inhibiting), 4 domains (emotional, behavioral, attentional and cognitive), and 2 temporal contexts (i.e., long and short-term) of self-regulation.

Evaluating the validity of the ASRI

Quite a bit is known about how self-regulation in childhood and adolescence is associated with other constructs, particularly parenting and adjustment. Such thorough knowledge facilitates the examination of new measures; specifically, the validity of the new measure is bolstered if the relations between the constructs are consistent with existing research. Thus, seven additional constructs were included to examine the concurrent and incremental validity of the ASRI. These included three dimensions of parenting and four dimensions of adjustment, and the linkages between these concepts and overall self-regulation are briefly explicated below.

Parenting

Three parenting behaviors that have been previously linked to self-regulation were included in the current study, including parental warmth, psychological control, and behavioral control. High levels of warmth and behavioral control are thought to encourage the development of self-regulatory capacities, while high levels of psychological control are believed to discourage their development (Barber and Harmon, 2001; Baumrind, 1991; Brody and Ge, 2001). Parental warmth hypothetically reduces negative arousal, which can interfere with individuals’ self-regulatory abilities (Brody and Ge, 2001). It is through the imposition of external behavioral control that children internalize parental and societal values and norms for behavioral control (Olson et al., 1990). Finally, psychological control impacts self-regulation through violation of the child’s self-system (Barber and Harmon, 2001; Baumrind, 1991). Parents who control their children using psychological manipulations undermine their attempts at independent regulation, which deprives children of both the experience of autonomous regulation and the opportunity to gain understanding of when self-regulation is necessary. Two recent studies with other measures of self-regulation have supported these theoretical assertions. In a sample of Dutch youth ages 10 to 14, Finkenauer and colleagues (2005) revealed that higher levels of parental acceptance and lower levels of psychological control were indicative of better self-control. In a sample of Australian high school-aged youth, Purdie and colleagues (Purdie et al., 2004) linked parental involvement to academic and prosocial self-regulation.

Adjustment

High levels of short-term self-regulation have been linked to well-adjusted behavior in prior cross-sectional and longitudinal work with children, adolescents and adults (Barkley, 1997; Eisenberg et al., 1993; Eisenberg et al., 1995). This hypothesis has been supported widely in the growing body of studies focused on specific indicators of adjustment (e.g., Brody and Ge, 2001; Finkenauer et al., 2005; Tangney et al., 2004). As such, four dimensions of adjustment were included in the current study, including externalizing, internalizing, prosocial behavior, and academic performance. As with parenting, each has been examined in previous studies featuring different measures of self-regulation.

Low levels of self-regulation have typically been connected to higher levels of externalizing and internalizing problem behavior in childhood and adolescence (Brody and Ge, 2001; Eisenberg et al., 2005; Finkenauer et al., 2005; Galambos and Maggs, 1991; Olson et al., 1999; Tangney et al., 2004). This link has been consistently revealed in studies of antisocial behavior or delinquency (Vazsonyi et al., 2001), substance use (Brody and Ge, 2001; Colder and Chassin, 1997; Tangney et al., 2004) and to overall hostility, anger and aggression (Finkenauer et al., 2004). Depressed children and adolescents also tend to report lower levels of self-regulation than non-depressed youths (Finkenauer et al., 2005; Lengua, 2003).

High levels of self-regulation have also been linked to positive behavior, including prosocial behavior and academic achievement. Prosocial behaviors are any voluntary positive social behaviors intended to benefit another person (Eisenberg and Fabes, 1998). In one study, children who were better able to regulate their positive and negative affect were also more prosocial (Rydell et al., 2003). In another study, prosocial adolescents were more likely to report greater feelings of self-efficacy in regulating participation in risky behavior, managing negative emotions, and expressing positive emotions (Bandura et al., 2003). High levels of self-regulation have been associated with better grades in school and better performance on standardized tests (Mischel et al., 1989; Tangney et al., 2004). Across studies, high levels of self-regulation appear to be associated with higher levels of prosocial behavior and better school performance.

The current studyThe primary goal of the current study was to evaluate a new measure of self-regulation for adolescents. Two related goals were addressed. First, the factor structure and reliability of the ASRI were examined to determine whether short-term and long-term self-regulation form distinct but correlated factors. Youth self-reports and parent reports about their children were compared for consistency in responses across reporters. Second, the concurrent, construct and incremental validity of the ASRI were examined by comparing its performance to another existing measure of self-regulation and by examining its associations with constructs theoretically and empirically tied to self-regulation. These other study constructs included measures of internalizing, externalizing, prosocial behavior, and academic performance. It was hypothesized that the ASRI would be correlated with the comparison self-regulation questionnaire, and that higher levels of short- and long-term self-regulation would be linked to lower levels of maladjustment and higher levels of positive adjustment. Thus, it was hypothesized that youth with better self-regulation will report better school grades, higher levels of prosocial behavior and fewer externalizing and internalizing problems. Three dimensions of parenting were also included for the same purpose. It was anticipated that higher levels of both forms of self-regulation would be associated with higher levels of parental warmth and firm control and lower levels of psychological control. Finally, it was hypothesized that the inclusion of the ASRI long-term factor would predict more variance than the short-term ASRI factor or the existing measure of self-regulation alone.

MethodParticipants

Study participants were adolescents from a suburban school district in southern Michigan, which is comprised of predominantly European American youths (99% of the high school and 98.7% of the middle school are European American; National Center for Education Statistics, 2005). It is likely that relatively few of the participants are in poverty: 8% of students in the middle and high schools qualify for reduced-cost or free school lunch (the average in the state of Michigan is 30%; National Center for Education Statistics, 2005).

Adolescent sample (N=169)

Adolescent participants were students in the 6th, 8th or 10th grade (mean age=13.79 years, S.D.=1.79, range=11–17 years). The resulting sample included youth of both genders (57% female) who were primarily of European American descent (89%) and lived with both biological parents (75%). Few participants (14%) reported ever having been diagnosed with or ever having taken medication for ADD/ADHD.

Parent subsample (n=80)

The parents of a subset of the adolescent sample also participated in the study by providing reports on their own and/or their child’s self-regulation. Most but not all parents provided both reports; 73 provided reports on their own self-regulation and 68 reported on their child’s self-regulation. Additional family demographic information was also collected from parents. Participating parents tended to be female (87%), in their mid-40s (mean age=43.18 years, S.D.=6.29, range=29–59 years), European-American (99%), married (85%), and have attended college or technical school or earned a college degree (84%). Most participants reported their family’s yearly income as $70,000 or higher (58%). Most families reported having more than one child in the household (M=3.02 children, S.D.=1.74, range=1–14).

Study and demographic constructs of youth with and without parent reports were compared to determine if the samples with and without parent reports differed. As in the full adolescent sample, youth in the limited sample also tended to be female (55%), of European American descent (89%), and from two-biological parent families (80%). Youth with parent data tended to be younger than youth without parent data, F (1, 167)=11.32, p < .001. This is unsurprising, given that youth with parent data were more likely to be in the 6th or 8th grades, while youth without parent data were more likely to be in the 10th grade, χ2(2)=17.11, p < .001. Youth with and without parent data did not differ significantly in terms of gender, (χ2(1)=.26, p=.64), ethnicity, (χ2(5)=2.66, p=.75), family structure (2 parents versus all other arrangements), (χ2(1) =.70, p=.27), or ADHD diagnosis, (χ2(1)=.40, p=.65). Comparisons were also conducted for the main study variables (parenting practices, self-regulation, and adjustment indicators). There were no significant differences between the full and limited samples on any of the study constructs.

Table 1 Study variable descriptive statistics

Variable M SD Possible range Actual range

Short Term Self-Regulation

Youth Self-Report 3.15 .56 1–5 1.67–4.73

Parent Self-Report a 3.68 .48 1–5 2.33–4.87

Parent Report about Child b 3.05 .75 1–5 1.60–4.73

Long Term Self-Regulation

Youth Self-Report 3.58 .57 1–5 2.00–5.00

Parent Self-Report a 4.03 .41 1–5 3.07–4.87

Parent Report about Child b 3.50 .75 1–5 1.80–5.00

Comparison Self-Regulation Scale

Youth Self-Report 3.29 .57 1–5 2.00–4.87

Parent Self-Report a 3.88 .53 1–5 2.61–4.71

Parent Report about Child b 3.50 .72 1–5 2.03–4.90

Parental Acceptance 2.51 .46 1–3 1.00–3.00

Parental Psychological Control 1.67 .42 1–3 1.00–3.00

Parental Behavioral Control 2.17 .39 1–3 1.00–3.00

Prosocial Behavior 4.13 .62 1–6 2.50–5.83

Academic Performance 3.68 2.49 1–12 1.00–12.00

Variable M SD Possible range Actual range

Externalizing 1.36 .33 1–5 1.00–2.97

Internalizing 2.92 .82 1–5 1.13–4.87Note. N=169 unless otherwise noted. a n=73. b n=68.

Procedures

Parents and adolescents were recruited separately for the current study.

Adolescents were informed of the general focus of the study through in-class presentations, and were advised that their parents would receive a packet in the mail. This packet included cover letters from the researcher and the school district, a parental consent form, a copy of the youth assent forms, a short parent questionnaire that included demographic questions, and questionnaires on their own and their teen’s self-regulation. Mailing delays necessitated providing a second recruitment packet to each eligible student with instructions to give the packet to their parents. Parents were asked to retain one copy of the consent form and to return the other copy and completed questionnaires to the school. Postcards reminding parents to return the forms and questionnaires were mailed to parents approximately 1 week following the mailing of the study packet. Parents also received debriefing forms in this packet. The debriefing form was folded so that study goals were not visible until the form was unfolded, and “Do not read until after completing the study” was printed on the outer side of the page. Parents were asked not to discuss the project objectives or their responses with their children until after study completion. Six hundred ninety recruitment packets were sent to parents, and one hundred seventy signed consent forms were returned. Two packets were returned in the mail. The final sample consisted of one hundred sixty-nine students (24% response rate).

Teachers initiated the adolescent portion of the study two weeks after students received the second parent study packet. This was conducted over several days in “homeroom.” Each student participant was provided with a study packet, which included two copies of the assent form, several questionnaires and an envelope with their name printed on the outside (no names were included on the questionnaires). Students returned their study packet to their teacher, who stored the packets in a secure location between homeroom periods. Students were instructed to seal and write “done” on the envelope once all questionnaire items were complete. Completed

packets were returned to the researcher each day after homeroom. Teachers provided participants with the study debriefing form.

Measures

Descriptive statistics for all study variables are presented in Table 1.

Self-regulation

The Adolescent Self-Regulatory Inventory (ASRI) was used to measure the self-regulation of teens and their parents. Seventy-seven items of original design were initially developed, and several doctoral developmental psychology students and faculty categorized each item’s temporal context (i.e., long-term or short-term), domain of self-regulation (i.e., emotion, attention, behavior, or cognition) and component (i.e., monitoring, activating, adapting, persevering, or inhibiting). Any item that was not clearly categorized as intended was reworded or discarded. Student and faculty reviewers were also asked for suggestions on item wording, in order to ensure that each item was as clear and unambiguous as possible. Several items were revisions of items from existing measures (e.g., Brown et al., 1999; Capaldi and Rothbart, 1992; Humphrey, 1982). The resulting 40-item questionnaire was piloted with undergraduate psychology students under the age of 20, in order to make the pilot sample more closely resemble the population in which the questionnaire would be used. Two pilot studies (N=50 for each study) were conducted in order to generate data that were used to run exploratory factor analyses and to check the scale reliabilities. Items that did had low inter-item correlations others from the same subscale were eliminated from further consideration. Results provided a basis for the selection of the final set of 36 items included in the current study (please see Appendix).

Adolescents completed the ASRI-A, which includes items about their own self-regulation. Parents completed the ASRI-AD about their own self-regulation and ASRI-P/C about their child’s self-regulation. These three questionnaires consist of the same items, which assess the degree to which adolescents or adults are able to activate, monitor, maintain, inhibit and adapt their emotions, thoughts, attention, and behavior. Respondents rate how true each item is for them, ranging from 1 (not at all true for me) to 5 (really true for me). Following reliability analyses, 13 items remained in the short-term ASRI (adolescent self-report α=.70; parent self-report α=.69; parent report about child α=.88), and 14 items remained in the long-term ASRI (adolescent self-report α=.82; parent self-report α=.72; parent report about child α=.91). These subscales do not include the remaining 9 ASRI items that performed adequately in the pilot studies but had low item-total correlations in the current study.

Comparison self-regulation questionnaire

An existing 31-item scale of self-regulation from another study was also included (Novak and Clayton, 2001). This questionnaire was designed to tap emotional, cognitive, and behavioral self-regulation in middle and high school-aged students (Sample item: “I get fidgety after a few minutes if I am supposed to sit still”). Respondents rate how true each item is for them, ranging from 1 (not at all true for me) to 5 (really true for me). A scale score was calculated by averaging participants’ responses to the 31 items. Parents and teens must have responded to at least 24 items in order to have a scale score computed (98% of adolescents met this criterion. 94% of participating parents were eligible for computation of a self-report scale score, as were 97% of parents’ reports about their teens). The scale demonstrated adequate internal consistency in all three versions (adolescent self-report α=.88; parent self-report α=.91; parent report about child α=.95).

Parenting

The 30-item version of the Children’s Reports of Parental Behavior Inventory (CRPBI: Schaefer, 1965; Schludermann and Schludermann, 1988) was used to measure three parenting dimensions. Each subscale has previously demonstrated adequate internal consistency (α range=.63–.76) and test-retest reliability (range=.79–.89; Schludermann and Schludermann, 1988). The original CRPBI-30 includes two sets of 30 questions about mothers and fathers; to save time, participants were asked to respond to one set of gender-neutral items for the parent with whom they spend the most time. Participants responded to each item on a 3-point response scale, ranging from 1 (Not like my parent) to 3 (A lot like my parent). Parental acceptance was based on 10 items designed to measure the degree to which a parent warmly accepts their child (α=.90; sample item: “My parent is a person who smiles at me very often”). Psychological control was comprised of 10 items designed to measure the degree to which a parent controls or manipulates the adolescent’s feelings or thoughts (α=.80; sample item: “My parent is a person who says, if I cared for him/her, I would not do things that cause him/her to worry”). Behavioral control was derived from 10 items designed to measure the degree to which a parent controls the adolescent’s actions (α=.80; sample item: “My parent is a person who lets me do anything I like to do”). All three scales were constructed by averaging the adolescent’s responses, and participants must have provided answers for at least 75% of items in order to have a scale score calculated (97.1% of adolescents met this standard).

Academic performance

Academic performance was measured with a single item designed to elicit how well the teen had performed over the past year of school (Center for Human Resource Research, 2002). Adolescents were asked to select the choice that best described their average report card grades, with response options ranging from 1 (A) to 12 (E 

or F). Responses were recoded so that the higher numeric values corresponded with better school grades. 97.6% of participants provided a response to this question.

Prosocial behavior

The measure of prosocial behavior was chosen for its brevity and focus on actual behavior versus behavior in hypothetical situations (e.g., Carlo and Randall, 2002). Six items tapping into volunteering, helping, raising or donating money, sharing, and doing favors for others (e.g. “Did someone a favor or lent someone money;” Swisher et al., 1985) were used to assess frequency of prosocial behaviors. Response options ranged from 1 (Never) to 6 (Happens almost every day or more). A scale score was constructed by averaging responses to the items (α=.61). Participants must have answered 5 of the 6 items in order to be included in analyses (98.8% of participants met this criterion).

Internalizing behavior

Two subscales from the distress dimension of the Weinberger Adjustment Inventory (WAI; Weinberger, 1989) were combined to measure internalizing problem behavior. Both subscales have been previously used with children and adults, and have demonstrated adequate reliability (α=.95 for the distress dimension, which includes two additional subscales not included here; Weinberger and Schwartz, 1990). Participants responded to the 15 items on a 5-point response scale, ranging from 1 (False) to 5 (True). Preliminary analyses indicated that the two subscales were moderately correlated (r=.61). In order to minimize the number of study variables, these two subscales were aggregated into an internalizing scale by averaging the 15 items (α=.89; sample item: “In recent years, there have been a lot of times when I’ve felt unhappy or down about things.”). Scale scores were constructed only for participants who completed at least 75% of the items (95.3% of participants met this standard).

Externalizing behavior

Externalizing problem behavior was measured with a 33-item scale developed by Galambos and colleagues (Galambos et al., 2003), based on measures used previously by Brown and colleagues (Brown et al., 1986) and Kaplan (1978). Participants responded to each item with a 5-point scale, ranging from 1 (Never) to 5 (Almost every day), and must have provided data for at least 75% of items to be included in analyses (96.5% of respondents met this criteria). A total externalizing behavior scale score was constructed by averaging the 24 items (α=.89). Preliminary analyses indicated that externalizing was the only moderately kurtotic variable (kurtosis=6.25), and no corrective transformation was performed.

ResultsConfirmatory factor analyses

Confirmatory factor analyses were conducted in order to evaluate the ASRI. Model fit was evaluated through several fit indices. The minimal requirements for good model fit were a non-significant X 2-fit statistic, a X 2-fit/df ratio of less than 3, a Comparative Fit Index (CFI) greater than .90, a Root Mean Squared Error of Approximation (RMSEA) ranging from .05 to .10, and a Standardized Root Mean Square Residual (SRMR) close to 0 (Klein, 1998). Items were considered to “load” on the factor if their coefficient was greater than .3 or .4.

Single factor model

A single factor model was evaluated using confirmatory factor analysis. All items from the long-term and short-term scales were constrained to load on one factor. This model’s fit was less than adequate, X 2 fit (324)=557.53, p < .001, X 2 fit/df ratio=1.72, CFI=.74, RMSEA=.07, SRMR=.08. All but five items loaded on the single factor. The five exceptions were items 6, 8, 12, 15, and 21 (please see Appendix for item text).

Two factor model

A two factor solution in which the long-term and short-term self-regulation factors were allowed to correlate was also evaluated. The adolescent self-report self-regulation model fit slightly better than the single-factor model, X 2 fit (323)=542.35, p < .001, X 2 fit/df ratio=1.68, CFI=.76, RMSEA=.06, SRMR=.08. The long-term and short-term factors were strongly correlated (r=.83). All long-term items except for two (items 12 and 15; see Appendix) loaded on the long-term factor. All short-term items except for three (items 2, 6, and 8) loaded on the short-term factor. The model modification indices indicated that the model fit could be improved by allowing items 18 and 21 to load on the long-term factor, and by permitting items 15, 23, and 26 to load on the short-term factor. The model fit could also be improved by allowing several items to correlate.

Revised two-factor model

These suggestions to improve the model were incorporated in a revised two-factor model. In this model, items 18 and 26 were revealed as Heywood cases (Heywood, 1931) when allowed to load on both factors. Thus, a third model in which items 15, 21, and 23 were permitted to load on both factors was evaluated (Figure 1). Items 12 and 15 persisted in not loading on the long-term factor, and items 2, 6, and 8 did not load on the short-term factor. However, these changes collectively improved model

fit: X 2 fit (314)=423.32, p < .001, X 2 fit/df ratio=1.35, CFI=.88, RMSEA=.05, SRMR=.07.

Fig. 1. Final two-factor confirmatory factor analysis model

The ASRI scale scores were adjusted to reflect this better-fitting model. Specifically, items 15, 21 and 23 were included on both scale scores. Items that did not load on their specified scales were not removed from the averaged indices. Thus, 15 items were present in both the long-term and short-term ASRI scores. This slightly improved the coefficients of the short-term scales (adolescent self-report α=.75; parent self-report α=.76; parent report about child α=.90), while the coefficients of the long-term scales did not substantially change (adolescent self-report α=.80; parent self-report α=.73; parent report about child α=.91). These revised scale scores were used for all of the following analyses, and the descriptive statistics are included in Table 1.

Construct validityThe construct validity of the ASRI was evaluated through an examination of the correlations between the ASRI scales and the comparison measure of self-regulation

(Novak and Clayton, 2001). Correlations for all reporters are presented in Table 2. For all three reports (adolescent self-report, parent self-report, and parent report about their child), both ASRI factors were positively and strongly correlated with the comparison measure (r range=.68–.92). ASRI factors were also correlated with each other (r range=.74–.91). Parent reports about their children were moderately correlated with youths’ self-reports (r range=.35–.47) on both the ASRI and the comparison measure. The majority of parent and child self-reports were not correlated; however, parents’ self-reported comparison measure scores were modestly correlated with their reports on their children’s long-term self-regulation and comparison measure scores (r range=.24–.25). Table 2. Correlations between forms of the ASRI and the comparison self-regulation scale

1 2 3 4 5 6 7 8

Adolescent Self-Report

1 Short-Term (ASRI)

2 Long-Term (ASRI) .74**

3 Comparison Scale .74** .68**

Parent Report about Child

4 Short-Term (ASRI) .35** .44** .44**

5 Long-Term (ASRI) .35** .47** .46** .91**

6 Comparison Scale .35** .44** .48** .92** .91**

Parent Self-Report

7 Short-Term (ASRI) .01 −.05 .05 .12 .16 .14

8 Long-Term (ASRI) −.06 −.12 .01 .06 .20 .16 .80**

9 Comparison Scale −.01 −.01 .10 .17 .24* .25* .72** .73** * p<.05. ** p<.01 (2-tailed).

Concurrent validityA second set of correlations were examined in order to confirm that adolescents’ responses on the ASRI were statistically associated with constructs theoretically and empirically linked to self-regulation. These bivariate associations are presented in Table 3. Table 3 Correlations between the ASRI and relational and behavioral constructs associated with self-regulation

1 2 3 4 5 6 7 8 9

1 Short-Term Self-Regulation

2 Long-Term Self-Regulation .74**

3 Comparison Self-Regulation Measure .74** .68**

4 Acceptance .23** .40** .34**

5 Psychological Control −.30** −.30** −.33** −.54**

6 Behavioral Control −.13 −.03 −.09 −.31** −.37**

7 Academic Performance .41** .41** .49** .15 −.24** .02

8 Prosocial Behavior .20* .31** .15 .23** .05 .19* .02

9 Internalizing −.35** .27** −.48** −.23** .34** .16* −.25** .10

10 Externalizing −.34** −.46** −.33** −.21** .21** .02 −.34** −.17* .15* p < .05. ** p < .01 (2-tailed).

Parenting

Both short-term and long-term ASRI factors were positively correlated with warmth and negatively correlated with parental psychological control. Neither form was associated with behavioral control. The same pattern was evident for the comparison self-regulation measure.

Adjustment

Short and long-term self-regulation were associated with all indicators of adjustment. Higher levels of both forms of self-regulation were associated with better school performance, higher levels of prosocial behavior, and lower levels of internalizing and externalizing difficulties. With one exception, the same pattern was evident for the comparison measure, which was not correlated with reported prosocial behavior.

Incremental validitySeven hierarchical regressions were conducted to evaluate the incremental validity of the ASRI. For all models, the comparison self-regulation index was entered on the first step, followed by the addition of the short- and long-term self-regulation factors on the second step. Incremental validity is evidenced by significant changes in the variance explained by the second step of the model. When model results suggested mediation was possible, Sobel tests were computed using an internet-based calculator (Preacher and Leonardelli, 2001). Results of these analyses are presented in Table 4. Table 4 Results of regression analyses predicting relational and behavioral constructs associated with self-regulation

Parental warmth

Psychological control

Behavioral control

Academic performance

Model 1 2 1 2 1 2 1 2

Comparison SR .34*** .25* −.33*** −.23* −.09 −.03 .49*** .37**

Short-Term SR

−.27* −.04 −.22+ .03

Long-Term SR

.43*** −.12 .15 .13

Model R 2 .11 .19 .11 .12 .01 .03 .24 .25

Prosocial behavior Internalizing Externalizing

Model 1 2 1 2 1 2

Parental warmth

Psychological control

Behavioral control

Academic performance

Model 1 2 1 2 1 2 1 2

Comparison SR .15+ −.11 −.48*** −.52*** −.33*** .06

Short-Term SR

.00 −.04 .05

Long-Term SR

.39*** .10 −.46***

Model R 2 .02 .10 .23 .23 .11 .22

All regression weights are standardized (βs).+p < .10. * p < .05. ** p < .01. *** p < .001.

At the first step of the parental warmth model, the comparison scale predicted a significant portion of the variance, F (1, 161)=20.42, p < .001, R 2=.11. The addition of the two ASRI scales on the second step significantly increased the portion of variance accounted for by the model, ΔF (2, 159)=7.67, p < .001, ΔR 2=.08. Long-term self-regulation was positively associated and short-term self-regulation was negatively associated with parental warmth. Additionally, the association between the comparison scale and parental warmth was reduced at this model step. Sobel tests indicated that the long-term ASRI factor mediated the effect of the comparison scale on parental warmth, Sobel ratio=4.95, p < .001.

The comparison scale also successfully predicted parental psychological control, F (1, 161)=20.18, p < .001, R 2=.11. Youths reporting lower levels of self-regulation as measured by the comparison scale reported higher levels of parental psychological control. Although the addition of the ASRI scales did not significantly increase the portion of variance accounted for by the model, the effect of the comparison scale was reduced on the second step, ΔF (2, 159)=.93, p > .05, ΔR 2=.01.

The comparison scale did not explain a significant portion of the variance in behavioral control, F (1, 161)=1.27, p > .05, R 2=.01. The addition of the ASRI scales

on the second step did not improve or contribute to the model, ΔF (2, 159)=1.49, p > .05, ΔR 2=.02.

For academic performance, the comparison scale predicted a significant portion of the variance in school grades, F (1, 161)=49.72, p < .001, R 2=.25. The addition of long- and short-term self-regulation on the second step reduced the contribution of the comparison measure but did not significantly increase the portion of variance accounted for by the model, ΔF (2, 159)=1.25, p > .05, ΔR 2=.01.

For prosocial behavior, the comparison scale did not predict a significant portion of the variance, F (1, 163)=3.64, p > .05, R 2=.02. The addition of the ASRI factors on the second step increased the portion of variance accounted for by the model, ΔF (2, 161)=7.37, p < .001, ΔR 2=.08. Long-term self-regulation was the only significant predictor of prosocial behavior on the second model step. A Sobel test indicated that the indirect effect of the comparison measure via the long-term ASRI factor was significantly different from zero, Sobel ratio=3.96, p < .001.

A significant portion of the variance in internalizing behavior was explained by the comparison self-regulation scale, F (1, 158)=46.73, p < .001, R 2=.23. The addition of short-term and long-term self-regulation on the second step did not explain any additional portion of the variance in internalizing, ΔF (2, 156)=.43, p > .05, ΔR 2=.00. However, the association between the comparison measure and internalizing increased on the second step, suggesting a suppressor effect. Follow-up analyses suggested that the presence of the ASRI long-term factor in the model is necessary for this suppressor effect to occur.

The comparison scale predicted a significant portion of the variance in externalizing on the first step, F (1, 160)=19.93, p < .001, R 2=.11. The addition of the ASRI factors on the second step increased the portion of variance accounted for by the model, ΔF (2, 158)=10.65, p < .001, ΔR 2=.11, and fully attenuated the effect of the comparison measure. Long-term self-regulation remained the sole significant predictor of externalizing difficulties, suggesting that long-term self-regulation fully mediates the effect of the comparison measure on externalizing. A Sobel test revealed that the indirect effect of the composite measure on externalizing via long-term self-regulation was significantly different from zero, Sobel ratio=−5.76, p < .001.

In sum, the long-term ASRI factor significantly increased the portion of variance in 3 models, and fully or partially mediated the effect of the composite measure on two parenting variables and three indicators of adjustment. The ASRI long-term factor also revealed a suppressor effect for internalizing. No variables predicted behavioral control in the hierarchical models.

DiscussionFactor structure of the ASRI

The Adolescent Self-Regulatory Inventory (ASRI) was developed specifically to correspond with theories of self-regulatory development in adolescence. Items representing long- and short-term contexts were included, in accordance with theory that adolescents are able to self-regulate for longer durations of time than children and in accordance with their own long-term goals (Demetriou, 2000). This is the first known study to examine the temporal distinction in self-regulation. Most previous studies have used items exclusively tapping short-term regulation, or have not made the distinction between the types. The ASRI’s items were also designed to represent five self-regulatory functions from Barkley’s (1997) hybrid model of self-regulation, including monitoring, activating, adapting, persevering, and inhibiting. Finally, study goals necessitated the creation of three versions of the ASRI, including forms for adolescent self-reports, parent (adult) self-reports, and one for parental report on their teenager’s self-regulation.

The CFA and reliability analyses demonstrated that, for the most part, the internal consistency of each factor was satisfactory. A two-factor model evidenced better fit than a single factor model, and in turn, the revised two-factor model had the best fit. The degree to which individual items from separate factors correlated was unanticipated but not undesirable. Items from two related constructs should be correlated, and this is consistent with other studies in this area (e.g., Raffaelli et al., 2005). Furthermore, several items that were judged by expert raters to be consistent with either short-term or long-term self-regulation loaded strongly on the incorrect factor. Upon further consideration, some of the terms used in these items may be perceived as vague. Additional rewording of these inconsistent items so that their temporal context is explicit may reduce the incidence of this problem. For example, item 15 (“If I really want something, I have to have it right away”) may be clearer if recast as “If I really want a special item, I must purchase it immediately.” These results suggest that the inclusion of additional items and the revision or omission of some of the current items is one area in which the ASRI may be further improved.

The small sample size of the parent self-report sample and the relatively few parent reports on their child’s self-regulation prohibited performing confirmatory factor analyses with either of the parental reports. It remains unknown whether the items that performed poorly in the adolescent self-report version of the ASRI performed better or worse in the parental report versions. Presumably, the parent report on child version would perform better, evidenced by the greater consistency in their responses. Parents may respond to the ASRI and other similar questionnaires with global conceptualizations of their children’s self-regulatory competencies in mind. Conversely, parents and adolescents may recall more detail or perceive “shades of

gray” in their self-regulatory self-schema, evidenced by scales that have comparatively poorer internal consistency.

Validity evaluations

The construct and concurrent validity of the ASRI is suggested through the correlations: specifically, the ASRI was statistically associated with other indicators was as expected. For all three forms of the ASRI, both ASRI factors were strongly correlated with the comparison questionnaire. Furthermore, parental reports on their children and children’s self-reports were also statistically linked. These results support the notion that the ASRI has sufficient construct validity. The modest correlations between the comparison self-regulation scores of parent reports of their own and their child’s self-regulation were unanticipated and are most likely due to reporter bias.

Correlations between the adolescent self-reported ASRI factors, parenting and adjustment were also evaluated. Associations were consistent with hypotheses. Both the short-term and long-term ASRI factors were positively correlated with parental warmth and negatively correlated with psychological control. Neither ASRI factor was statistically associated with parental behavioral control. This pattern is consistent with other studies using different measures of self-regulation (Finkenauer et al., 2005). Likewise, the four adjustment indicators were also statistically linked to both ASRI factors. Specifically, youth reporting higher levels of long- and/or short-term self-regulation also tended to report better school grades, more prosocial behavior, and less internalizing and externalizing behavior. The association between the self-regulation factors and prosocial behavior was unique to the ASRI, as there was no link from the comparison measure to prosocial behavior.

Finally, the incremental validity of the ASRI factors was scrutinized. Regression analyses revealed that the long-term ASRI factor increased the portion of variance accounted for by the comparison index of self-regulation in 3 of the 7 models (parental warmth, prosocial and externalizing behavior). The short-term ASRI factor also explained a unique portion of variance in parental warmth. These findings extend previous studies and support the incremental validity of the ASRI, suggesting that regulation over the long term contributes to momentary control processes that promote adjustment. Likewise, these results suggest that long-term self-regulation is another potential product of warm parenting behavior. Neither ASRI factor increased the portion of variance explained in the remaining 4 models. Although just less than half of the study models were improved by the ASRI, these results are not at all discouraging. It remains possible that the two aspects of self-regulation are implicated in different forms of adjustment or are the result of distinct parenting behaviors. For example, good short-term regulation may prevent youths’ involvement in impulsive acts in the face of momentary temptations that may lead to decreased

competence. Likewise, depressive symptoms may undermine youth’s momentary control processes. Alternately, good long-term self-regulation may keep teenagers on goal-oriented pathways which reduce the number of temptations they encounter. For example, teenagers who value academic achievement may choose to avoid large parties in which alcohol will be present. Regardless, much more work is also needed to elucidate whether the short- and long-term aspects of self-regulation have different antecedents or are linked to different adjustment outcomes.

Study limitations and future directions

Like any other study, the current study has several limitations. First, study data was primarily obtained through adolescent self-report questionnaires. Parental reports were sought for parts of the study, but an insufficient number responded for additional analyses to be conducted. For future studies, multiple reporters should be recruited to provide their perspective on additional study variables. Common method variance is another relevant limitation of the current study, as all measures were questionnaire-based (Podsakoff and Organ, 1986). Correlations between measures may be spurious because respondents provided all of their own questionnaire data. These limitations can be addressed by seeking complete reports from multiple respondents and by including objective measurement techniques whenever possible.

Second, study data were collected at only one time point, thus the cross-sectional nature of this study precludes any firm directional or causal interpretations. The temporal stability of the ASRI factors could not be examined in the current cross-sectional study. Thus, short-term longitudinal work is needed to determine the test-retest reliability of the ASRI.

Sample characteristics may have affected the results in unknown ways: participants were volunteers, and thus were not representative of all youth in the United States or all youth in this particular school district. Study participants reported relatively low levels of maladjustment and overall good parenting (high warmth and behavioral control, low psychological control). Problems related to study self-selection and study generalizability are well known (e.g., Campbell and Stanley, 1963). This is attributable to two issues. Youth were relied upon to return the parental consent form in order to participate in the study, and teens with particularly poor self-regulation may have found this task to be more difficult than their peers with better self-regulation (a problem mentioned by teachers at the study site). The size of the adolescent sample was relatively small, which also may be attributed to recruitment and parental consent procedures. Finally, youth or parents with particularly poor relationships may have actively or passively opted not to participate. Future studies with more representative samples will be invaluable in order to replicate these findings.

Despite these limitations, the current study makes a substantial contribution to the literature by evaluating a new theoretically-based questionnaire of adolescent self-regulation. The measure scrutinized in this study is the first to demonstrate the value of separating long- and short-term self-regulation in empirical work with adolescents. The ASRI provides a new direction for theory-driven research on self-regulation in adolescence, and makes direct tests of these theoretical assumptions possible. This new measure provides a basis for future longitudinal research on the development of long-term and short-term self-regulation, and how these dimensions may be differentially associated with parenting behaviors and adjustment. In time, this line of research may help improve youth risk-prevention programs, as well as increase the effectiveness of parenting interventions.

Acknowledgments This research was conducted while the author was a doctoral candidate in the Department of Psychology at the University of Nebraska - Lincoln, and was funded by a grant to M. Raffaelli and L. J. Crockett from the National Institutes of Mental Health (R01-MH62977). Research assistance was provided by Matthew Maguire, Megan Milligan, and Lisa Palmer. Many thanks to Daniel S. Shaw and Emily Trentacosta for their comments on this manuscript, and to Scott Novak for sharing his self-regulation questionnaire.

Appendix

Short Term Self-Regulation Items

2. When I’m sad, I can usually start doing something that will make me feel better.

6. When I’m bored I fidget or can’t sit still.

8. I can usually act normal around everybody if I’m upset with someone.

9. I am good at keeping track of lots of things going on around me, even when I’m feeling stressed.

11. I can start a new task even if I’m already tired.

13. Little problems detract me from my long-term plans.

14. I forget about whatever else I need to do when I’m doing something really fun.

16. During a dull class, I have trouble forcing myself to start paying attention.

17. After I’m interrupted or distracted, I can easily continue working where I left off.

18. If there are other things going on around me, I find it hard to keep my attention focused on whatever I’m doing.

19. I never know how much more work I have to do.

21. It’s hard to start making plans to deal with a big project or problem, especially when I’m feeling stressed.

22. I can calm myself down when I’m excited or all wound up.

Long-Term Self-Regulation Items

3. If something isn’t going according to my plans, I change my actions to try and reach my goal.

4. I can find ways to make myself study even when my friends want to go out.

12. I lose control whenever I don’t get my way.

15. If I really want something, I have to have it right away.

20. When I have a serious disagreement with someone, I can talk calmly about it without losing control.

23. I can stay focused on my work even when it’s dull.

25. I can stop myself from doing things like throwing objects when I’m mad.

26. I work carefully when I know something will be tricky.

28. I am usually aware of my feelings before I let them out.

29. In class, I can concentrate on my work even if my friends are talking.

30. When I’m excited about reaching a goal (e.g., getting my drivers license, going to college), it’s easy to start working toward it.

31. I can find a way to stick with my plans and goals, even when it’s tough.

32. When I have a big project, I can keep working on it.

37. I can resist doing something when I know I shouldn’t do it.

Items Not Included in Analyses

1. It’s hard for me to notice when I’ve “had enough” (sweets, food, etc.).

5. I lose track of the time when I’m doing something fun.

7. It’s hard for me to get started on big projects that require planning in advance.

10. When I’m having a tough day, I stop myself from whining about it to my family or friends.

24. I usually know when I’m going to start crying.

33. I can usually tell when I’m getting tired or frustrated.

34. I get carried away emotionally when I get excited about something.

35. I have trouble getting excited about something that’s really special when I’m tired.

36. It’s hard for me to keep focused on something I find unpleasant or upsetting.

References

Bandura A, Caprara GV, Barbaranelli C, Gerbino M, Pastorelli C (2003) Role of affective self-regulatory efficacy in diverse spheres of psychosocial functioning. Child Dev 74:769–782

Barber BK, Harmon EL (2001) Violating the self: Parental psychological control of children and adolescents. In: Barber BK (ed) Intrusive parenting: How psychological control affects children and adolescents. American Psychological Association,

Washington, DC, pp 15–52 Barkley RA (1997) ADHD and the nature of self-control. Guildford, New York Barkley RA (2004) Attention-deficity/hyperactivity disorder and self-regulation. In: Baumeister RF, Vohs KD (eds) Handbook of self-regulation: Research, theory, and applications. Guilford, New York, pp 301–323 Baumrind D (1991) The influence of parenting style on adolescent competence and substance use. J Early Adolesce 11:56–95

Brody GH, Ge X (2001) Linking parenting processes and self-regulation to psychological functioning and alcohol use during adolescence. J Fam Psychol 15:82–94

Brown BB, Clasen DR, Eicher SA (1986) Perceptions of peer pressure, peer conformity dispositions, and self-reported behavior among adolescents. Dev Psychol 22:521–530

Brown JM, Miller WR, Lawendowski LA (1999) The self-regulation questionnaire. In: VandeCreek L, Jackson TL (eds) Innovations in clinical practice: A sourcebook, vol. 17. Professional Resource Press/Professional Resource Exchange, Inc, Sarasota, FL, pp 281–292 Campbell DT, Stanley JC (1963) Experimental and quasiexperimental designs for research. Rand McNally, Chicago Capaldi DM, Rothbart MK (1992) Development and validation of an early adolescent temperament measure. J Early Adolesce 12:153–173

Carlo G, Randall BA (2002) The development of a measure of prosocial behaviors for late adolescents. J Youth Adolesce 31:31–44

Center for Human Resource Research (2002) NLSY79 child & young adult data users guide: A guide to the 1986–2000 child data and 1994–2000 young adult data. The Ohio State University, Columbus, OH

Colder CR, Chassin L (1997) Affectivity and impulsivity: Temperament risk for adolescent alcohol involvement. Psychol Addict Behav 11:83–97

Committee on Integrating the Science of Early Childhood Development, Shonkoff JP, Phillips DA (2000) From neurons to neighborhoods: The science of early childhood development. Na