Student Self-Regulated Learning in an Urban High School

Definition and Assessment of Self-Regulation

Over the past couple of decades, there has been extensive research linking SRL processes to various academic skills, such as mathematics (Pape, Bell, & Yetkin, 2003), writing (Graham & Harris, 2009), and content area subjects such as science (Cleary et al., 2008). In general, SRL is often conceptualized as a multi-dimensional process that involves purposeful efforts to behaviorally, cognitively, and metacognitively control and optimize learning in a particular context (Pintrich, 2000; Zimmerman, 2000). Two core components of most SRL models include students’ motivational beliefs (e.g., self-efficacy, interest) and their use of self-regulation strategies to enhance their achievement. From a social-cognitive perspective, motivation beliefs are hypothesized to be the primary sources underlying student efforts to reach personal goals (Pintrich, 2000). Thus, students who are most likely to engage and persist on school tasks typically possess strong beliefs in their personal capabilities (i.e., self-efficacy), exhibit interest and enjoyment in completing academic tasks (i.e., mathematics interest), and perceive academic tasks to be valuable or personally meaningful (Eccles & Wigfield, 2002; Usher & Pajares, 2008).

Although these motivational beliefs underlie the “will” to learn, it is the nature of students’ strategic skills that often determines their success in school. For example, research has shown that high-achieving students frequently use cognitive strategies to learn, such as elaboration and rehearsal, and routinely use regulatory strategies to manage their behaviors and learning environments, including help seeking and time management (Karabenick & Berger, 2013; Zimmerman & Martinez-Pons, 1988). In addition to using adaptive strategies, sophisticated self-regulated learners also seek to minimize personal displays of negative or maladaptive behaviors, such as avoiding help, self-handicapping, or procrastinating (Cleary & Chen, 2009; Flowers, Bridges, & Moore, 2012). In this study, we were interested in examining students’ motivation beliefs and their adaptive and maladaptive regulatory behaviors using self-report questionnaires and a teacher rating scale.

Assessment of SRL

Self-report questionnaires represent the most common type of SRL assessment. These measures typically involve having students use a Likert scale to retrospectively rate their regulatory beliefs, behaviors, and metacognition. Although desirable in many respects (e.g., ease of administration), self-report questionnaires have been criticized in the SRL literature because students are often unreliable reporters of their behaviors and because many of these measures lack adequate situational referents to guide students’ responses (Winne & Jamieson-Noel, 2002). To counter these potential limitations, SRL researchers have developed a variety of dynamic, contextualized assessment approaches in recent years, including think alouds in hypermedia (Azevedo, Johnson, Chauncey, & Graesser, 2011), behavioral traces (Winne & Perry, 2000), and contextualized structured interviews (Cleary, 2011). Although these innovations have been useful, there is a need to devote more attention to using teacher ratings as sources of student SRL behaviors and processes.

Teacher ratings of student SRL are important for several reasons. First, consistent with a multi-dimensional assessment approach, gathering SRL data from multiple sources (e.g., student reports, observations, teacher ratings) enables clinicians to converge assessment data and to facilitate hypothesis development (Sattler, 2008). In addition, because teachers spend much time with students and thus have many opportunities to directly observe student actions in the classroom (e.g., help seeking), they have the potential to provide unique information about student SRL processes. Finally, there is evidence showing that parents and teachers tend to be more accurate sources of students’ externalizing behaviors than students (Gould & Shaffer, 1985; Kamphaus & Frick, 2002).

Primary Objectives

Given the potential value yet paucity of SRL teacher rating scales, we developed the Self-Regulation Strategy Inventory–Teacher Rating Scale (SRSI-TRS) and examined its concurrent and predictive validity. Although a couple of SRL teacher rating scales are available (Zimmerman & Martinez-Pons, 1988), to our knowledge no scales have been developed to parallel a student SRL questionnaire (Cleary, 2006). Furthermore, we were interested in developing a measure that targeted student help-seeking behaviors across different situations, their use of self-motivation tactics to optimize effort in class, and their use of organization-related behaviors. We selected mathematics as the target domain in this study because of the strong link between student SRL and mathematics achievement (Pape et al., 2003) and because student motivation in mathematics remains an important issue in education circles (Ashcroft, Krause, & Hopko, 2007).

In this study, concurrent validity was examined by correlating the SRSI-TRS with a set of SRL self-report questionnaires. Given that the SRSI-TRS and all self-report measures were linked to the same academic class (i.e., mathematics) and due to the established link between students’ motivation beliefs (e.g., self-efficacy, interest) and their regulatory behaviors (Schunk & Zimmerman, 2008), we expected to observe high correlations between the SRSI-TRS and all self-report measures. In terms of predictive validity, we were primarily interested in examining the amount of variance in algebra test scores accounted for by the SRSI-TRS after controlling for prior math achievement and student responses on SRL questionnaires. We anticipated that the SRSI-TRS and the self-report measures would contribute unique variance to the prediction of students’ mathematics test grades after controlling for prior mathematics achievement.

Method

Results

Pearson correlations and hierarchical linear regression analyses were used to address the research questions. Prior to statistical analysis, missing data and the kurtosis and skewness of all variables were examined and found to be within acceptable limits (Kline, 1998). There was no missing data for the survey items. Furthermore, given that there were no significant group differences across gender, ethnicity, and SES status across the core measure used in this study, all data were collapsed. However, 10 students were removed prior to the regression analysis due to unavailability of MAP scores or classroom test scores. One-tailed tests using a significance level of p < .05 were used in all analyses unless noted otherwise.

Concurrent Validity of SRSI-TRS

The first objective of this study was to examine whether the SRSI-TRS correlated with student ratings of their motivation beliefs and regulatory behaviors. We calculated bivariate correlations between the SRSI-TRS and four student self-report measures (see Table 1). Statistically significant correlations were observed between the SRSI-TRS and student reported mathematics interest (r =.32), SRSI-Maladaptive Regulatory Behaviors (MRB; r = −.41), and LASSI–Test Taking (r = −.42). All of these correlations were in the medium range (Cohen, 1988). In general, these results indicated that students who were rated by their teachers as being strategic during classroom activities also tended to self-report high levels of interest or enjoyment in mathematic class and low levels of maladaptive regulatory behaviors. Contrary to expectations, the SRSI-TRS did not relate to student mathematics self-efficacy.

OPEN IN VIEWER

Table 1.

Descriptive Statistics and Correlation Coefficients Between SRL Measures and Math Achievement.

Measure	M	SD	1	2	3	4	5	6
1. MAP	226.17	12.87	—
2. Mathematics test average	65.90	16.34	.37*	—
3. Self-efficacy	7.35	1.77	.42*	.26*	—
4. Mathematics interest	3.36	0.62	.03	.29*	.38*	—
5. LASSI–Test-taking	2.44	0.84	−.41*	−.34*	−.52*	−.37*	—
6. SRSI-SR Maladaptive	2.25	0.65	−.09	−.44*	−.35*	−.50*	.50*	—
7. SRSI-TRS	3.27	1.26	.32*	.55*	.15	.32*	−.42*	−.41*

Note. The self-efficacy scale used an 11-point Likert scale (0-10), while all other SRL measures utilized a 5-point Likert scale (1-5). High scores on the LASSI-Test-Taking and SRSI-SR Maladaptive subscales represent greater maladaptive skills, whereas high scores on all other measures represents adaptive functioning. SRL = self-regulated learning; MAP = Measure of Academic Progress; LASSI = Learning and Study Strategies Inventory; SRSI-SR = Self-Regulation Strategy Inventory-Self-Report; SRSI-TRS = Self-Regulation Strategy Inventory–Teacher Rating Scale.

*

Denotes statistically significant finding using one-tailed test at p < .05.

Predictive Validity of SRSI-TRS

The key objective of this study was to examine whether teacher ratings emerged as a significant predictor of students’ mathematics achievement after controlling for four self-report questionnaires and prior mathematics achievement. Given that we were also interested in evaluating whether student reports accounted for unique variance, we conducted a hierarchical regression analysis and reported changes in R² across three models. We also reported semipartial correlations to convey the precise percentage of unique variance in mathematics achievement accounted for by each predictor after controlling for all other predictors (see Table 2).

OPEN IN VIEWER

Table 2.

Hierarchical Regression Analysis for Teacher and Student Predictors of Student Mathematics Classroom Test Scores.

Variable	Zero-order correlation	Semipartial correlations (sr2)	β	t	R 2
Step 1					.136
MAP	.37	.369 (13.6%)	.37	3.4*
Step 2					.284
MAP	.37	.288 (8.3%)	.32	2.9*
Self-efficacy	.24	−.058 (0.3%)	−.08	−0.58
Mathematics interest	.28	.091 (0.8%)	.12	0.90
LASSI–Test-taking	−.34	−.007 (0.0%)	−.01	−0.07
SRSI-SR Maladaptive	−.43	−.268 (7.2%)	−.35	−2.67*
Step 3					.379
MAP	.37	.187 (3.5%)	.22	1.99*
Self-efficacy	.24	.021 (0.0%)	.03	0.23
Mathematics interest	.28	.025 (0.1%)	.03	0.27
LASSI–Test-taking	−.34	.044 (0.2%)	.06	0.45
SRSI-SR Maladaptive	−.43	−.154 (2.4%)	−.21	−1.64 a
SRSI-TRS	.55	.307 (9.4%)	.39	3.26*

Note. Step 1: Adjusted R² = .125; ∆R² = .136 Step 2: Adjusted R² = .234; ∆R² = .148 Step 3: Adjusted R² = .325; ∆R² = .094. sr² = semipartial squared represents the proportion of unique variance in mathematics test scores accounted for by a specific predictor after controlling for all other variables. MAP = Measure of Academic Progress; LASSI = Learning and Study Strategies Inventory; SRSI-SR = Self-Regulation Strategy Inventory–Self-Report; SRSI-TRS = Self-Regulation Strategy Inventory–Teacher Rating Scale.

a

p = .05.

*

Denotes statistically significant finding using one-tailed test at p < .05.

In general, all predictors accounted for approximately 38% of the variance in mathematics achievement. Of greater importance was that each block of variables accounted for a medium increase in R². For example, in the first block, students’ prior mathematical achievement accounted for approximately 14% of the variation in classroom test scores, F(1, 75) = 11.82, p < .01, ΔR² = .136. In the second block, a significant and robust improvement in the prediction of mathematics achievement was observed, F(1, 71) = 3.67, p < .01, ΔR² = .148, although the MAP and SRSI-MRB emerged as the only significant predictors. Adding the SRSI-TRS in the third block accounted for an additional 9% of the variance in mathematics test grades, F(1, 70) = 10.64, p < .01, ΔR² = .094. In the final model, three variables emerged as significant predictors; the SRSI- TRS, SRSI-MRB, and prior achievement, with the TRS emerging as the strongest predictor (see semipartial correlations in Table 2).

Discussion

In this study, we hypothesized that a broad teacher measure of student SRL would correlate with student self-reports of their motivation processes and regulatory behaviors (e.g., seeking out help), and would also emerge as a significant predictor of mathematics achievement. Our hypotheses were largely confirmed, supporting a key general premise that different sources of student SRL can contribute unique variance to students’ academic achievement.

The SRSI-TRS accounted for a medium level of variation (9.4 %) in mathematics classroom test scores, after controlling for approximately 28% of the variance attributed to prior mathematics skills and student-reported motivation beliefs and regulatory behaviors. In a broad sense, these findings are consistent with prior research showing that teacher ratings can accurately predict a variety of academic outcomes (Perry & Meisels, 1996; Whitebread et al., 2009). Furthermore, the fact that a self-report measure of regulatory behaviors (i.e., SRSI-MRB) emerged as a significant predictor in the final regression model (see Table 2) is particularly noteworthy because many researchers have questioned the utility of self-report questionnaires relative to more objective or dynamic, event-based measures of SRL (Winne & Jamieson-Noel, 2002).

The second key finding in this study was that students who were identified by their teachers as being highly strategic and motivated during class activities were likely to self-report high levels of interest in mathematics activities and fewer maladaptive regulatory behaviors, such as avoiding help or procrastinating. A few important points pertaining to this latter finding are warranted. First, we showed that the SRSI-TRS correlated positively with students’ adaptive self-reported processes (i.e., mathematics interest) and negatively with their maladaptive self-reported regulatory behaviors. The fact that the SRSI-TRS correlated in the expected negative direction with two self-report measures targeting maladaptive regulatory behaviors further buttresses claims that the SRSI-TRS is in fact a measure of adaptive regulatory behaviors, such as seeking out help when confused, motivating oneself when struggling to learn, and being organized and prepared for class.

If one perceives teacher ratings of student behaviors as a more objective type of measure than self-reports, our results also appear to contradict SRL research showing that student reports of their behaviors do not correspond to more objective indicators of those behaviors. For example, Winne and Jamieson-Noel (2002) showed that student self-reports about their strategic behaviors during a studying activity exhibited very poor relations with real-time data about their behavioral traces during the studying activity. Furthermore, there is evidence in clinical contexts showing that student ratings of their behaviors often exhibit low correlations with teacher or parent reports of those same behaviors (Achenbach, McConaughy, & Howell, 1987; Loeber, Green, & Lahey, 1990).

Although we cannot make any claims that SRSI-TRS responses were “correct” or more accurate than student self-reports, there are a couple of explanations that support the observed relations in the current study. First, there was much overlap in the format and structure of the SRSI-TRS and the student measures. That is, both measures were contextualized to the algebra class and relied on retrospective, aggregated accounts of student SRL. Although speculative, a second explanation is that the nature of items in the SRL self-report questionnaires (i.e., focus on maladaptive regulatory behaviors) impacted the observed correlations. A few research studies have demonstrated that negatively worded self-report measures are sometimes more strongly correlated with teacher ratings of achievement and work habits (Bush, 1999; Malone, 1998). In our study, we surveyed students’ adaptive motivation beliefs but not their adaptive regulatory behaviors. As a result, we cannot draw any definitive conclusions regarding how the nature of the SRL questionnaires impacted the observed correlations. However, in light of prior research and the moderate concurrent validity coefficients observed between the teacher and student measures of SRL behaviors, future research should further explore this issue.

The lone unexpected finding of this study involved the non-significant relation between student self-efficacy and the SRSI-TRS. The interpretation of this finding is somewhat challenging because the self-efficacy measure was reliable and exhibited medium to large correlations with all other self-report scales and with both measures of mathematics achievement. Furthermore, our findings contradict the general finding in prior research that highly efficacious students in a particular domain exhibit more strategic behaviors in that context (Cleary & Zimmerman, 2001; Pajares & Urdan, 2006). One potential explanation, however, pertains to whether mathematics self-efficacy perceptions should be expected to relate to all types of classroom behaviors that are observable to a teacher. For example, six of the items on the SRSI-TRS targeted help-seeking which is an important regulatory strategy (Karabenick & Berger, 2013). Based on prior research, it is possible that some students in our study who exhibited high levels of confidence did not perceive the need to ask questions about tests or to seek out help when confused, whereas other confident students frequently sought out support from the teacher because of their inherent interest in learning (e.g., mastery goal orientation) or perhaps their high performance expectations (Karabenick & Berger, 2013). Although some studies have examined the relation between self-efficacy and help-seeking behaviors (Bong, 2008; Ryan & Shin, 2011), the precise nature of the relationship is inconsistent and needs to be studied in greater depth.

Conclusion

In this study, we found that student self-report SRL measures and a SRL-TRS collectively accounted for 24% of the variation in classroom-based mathematics tests after controlling for prior mathematics achievement. In the final model, the SRSI-TRS and the SRSI-MRB (maladaptive regulation) emerged as the key SRL predictors of student achievement. These results support the general premise that a multi-dimensional assessment approach is an important way to not only understand how students might regulate in specific contexts, but to also develop more robust explanations for student performance in schools. From our perspective, when practitioners encounter discrepancies between teacher ratings and student questionnaires, it is important for them to resolve such discrepancies by gathering additional data with other types of SRL assessment tools, such as direct observations, think-aloud protocols, behavioral traces, and SRL microanalytic interviews (Winne & Perry, 2000; Zimmerman, 2008).