The Effectiveness of Current Interventions to Reverse the Underachievement of Gifted Students: Findings of a Meta-Analysis and Systematic Review

Theories of Motivation for Learning

Motivation for learning is an important and complex topic. Most frameworks recognize the importance of multiple factors including personal characteristics of the individual, such as self-efficacy, interests, goals, beliefs, values, and environmental influences (e.g., school and family environments and cultural and societal attitudes; Dweck, 1999; Dweck & Leggett, 1988; Pintrich, 2000; Urdan, 2010; Wigfield & Cambria, 2010). Major theoretical frameworks recognize the complexity of motivation and typically include multiple, interacting components. There are several, however, that undergird much of the educational research on motivation (Linnenbrink-Garcia & Patall, 2015). A notable trend in research on gifted underachievers is the increasing integration of major motivational and self-regulated learning theories as a framework for these studies. These frameworks include intrinsic and extrinsic motivation, expectancy-value theory, self-determination theory, achievement goal theory, and self-regulation. Readers can find additional overviews of these theories and related key constructs in the works of Eccles and Wigfiled (2002) and Wigfield and Cambria (2010) and discussions of these as they relate to gifted students in Clinkenbeard (2012) and Patrick et al. (2015).

Intrinsic and Extrinsic Motivation

Most frameworks of motivation for learning include and contrast the constructs of intrinsic and extrinsic motivation. Intrinsic motivation is learning that is internally driven by interest, curiosity, and desire to know or improve—“learning for learning’s sake.” Extrinsic motivation is learning that is driven by external rewards such as good grades, high test scores, positive affirmation from adults, or prizes. Csikszentmihalyi (1991) emphasized intrinsic motivation in his concept of flow, and Deci and Ryan’s (1985) self-determination theory, which emphasizes a learner’s need for autonomy, relatedness, and competency, is built on the concept of intrinsic motivation.

Achievement Goal Theory

Of these major theoretical frameworks, achievement goal theory is the most prominent in research on gifted underachievement (Linnenbrink-Garcia & Patall, 2015; Siegle & McCoach, 2005). Five out of the 10 quantitative intervention studies included in the current meta-analysis were based on achievement goal theory. Central to this framework is goal orientation, which pertains to students’ purposeful engagement in learning-related tasks. A widely cited theoretical framework of goal orientation is that of Elliot and colleagues (Elliot, 2005; Elliot & Church, 1997). This model posits “mastery” and “performance goal” orientations. A mastery goal orientation manifests when a learner undertakes a task primarily to master content, increase competence, or enjoy a sense of learning something new. Evidence supports the positive link between mastery goal orientation and academic achievement (Hulleman et al., 2010). A performance orientation is evident when a learner’s engagement and effort on a task are primarily motivated by their intention to outperform their peers or show off their superior competence. Performance goal orientations have been classified as “performance-approach” (i.e., wanting to achieve higher than peers) and “performance-avoidance” (i.e., wanting to avoid failing or looking bad) orientations. Some research shows that a performance approach orientation can enhance motivation and achievement (Elliot & Harackiewicz, 1996).

Expectancy-value frameworks place a primary role on individual’s beliefs in effecting motivation for learning. Eccles et al. (1998) developed a comprehensive expectancy-value model and have conducted considerable research on its various components (see Eccles & Wigfield, 2002, and Wigfield & Cambria, 2010). In their model, individual’s beliefs about how well they might do in a future learning task or learning in a particular domain is related to their willingness to expend effort with higher expectations resulting in greater effort and engagement. Another critical component is the importance of the task or domain to some future goal of the individual, with higher value associated with greater motivation. These beliefs interact such that the value of a task may be diminished if the individual believes success will involve too much work, investment, or effort.

Self-Regulated Learning

Self-regulated learning is another motivational framework related to achievement goal theory. It is used as the underlying theory for at least two underachievement interventions reviewed in the current meta-analysis. Self-regulated learning “. . . is an active, constructive process whereby learners set goals for their learning and then attempt to monitor, regulate, and control their cognition, motivation, and behavior, guided and constrained by their goals and the contextual features in the environment” (Pintrich, 2000, p. 453). It involves the engagement and interplay of the learner, task, and/or context and consists of four different areas for regulation: cognition, motivation and affect, behavior, and context (Pintrich, 2000). Research shows that the processes of self-regulation involve the interplay of varied motivational factors, suggesting that motivational and cognitive processes are likely inseparable in self-regulated learning (Pintrich et al., 1993).

Self-regulated learning generally proceeds in four sequential phases involving (1) planning, goal setting, and activating perceptions and knowledge; (2) monitoring; (3) controlling or regulating; and (4) reacting and reflecting on various activities and processes. Ample research shows that self-regulatory activities are significantly related to academic achievement, often accounting for the mediational relationships between the learners’ characteristics and the learning contexts (Zimmerman, 1990). Some research supports that poor self-regulation skills are a significant factor for underachievement among gifted secondary students (Baum et al., 1995; McCoach & Siegle, 2003b).

The Achievement Orientation Model of Interventions

Scholars have developed several theoretical models in pursuit of interventions to address gifted underachievement. These include Rimm’s trifocal model, emphasizing the synergetic involvement of the student, the home, and the school (Rimm, 2008a, b) and Renzulli and Reis’s schoolwide enrichment model, which emphasizes individual interests and authentic enrichment (Reis & Renzulli, 2009). The most prominent one is the achievement orientation model (AOM), which focuses on how students’ beliefs and values can regulate their motivation to engage and achieve (Siegle & McCoach, 2005). The AOM was built upon an incorporation of multiple motivation theories including self-efficacy theory (Bandura, 1986), attribution theory (Weiner, 1986), expectancy-value theory (Eccles & Wigfield, 1995), theory of locus of control (Rotter, 1966), person–environment fit theory (French et al., 1974), and self-regulated learning (Pintrich, 2000).

The AOM hypothesizes that students’ self-perceptions in four key areas collectively effect an individual’s learning motivation. They are self-efficacy (e.g., belief that one can succeed), goal valuation (e.g., belief that learning is meaningful), perceptions about the environment (e.g., feeling supported by teachers, high expectations of others), and self-regulation. Furthermore, the AOM posits that positive self-perceptions and beliefs in these areas help improve learning behaviors, such as setting appropriate goals and utilizing self-regulation skills, which subsequently increase students’ learning motivation and engagement and improves their academic performances (Siegle, 2013). Findings from an empirical validation study of the AOM, based on a sample of 156 mathematically gifted sixth and seventh graders, confirmed the existence of most of the conceptual relationships that the AOM hypothesizes (Ritchotte et al., 2014). Key components in the AOM, as measured by five equivalent factors of the School Attitude Assessment Survey–Revised (academic self-perceptions, attitudes toward teachers and classes, attitudes toward school, goal valuation and motivation, and self-regulation; SAAS-R; McCoach & Siegle, 2003a), can be used to differentiate high and low achievers of differing developmental levels and cultural and socioeconomic backgrounds (Dedrick et al., 2015). The AOM has served as a framework for research on underachievement among gifted students and the design of interventions (e.g., Barbier et al., 2019; Brigandi et al., 2018; Rubenstein et al., 2012).

Understanding Characteristics of Gifted Underachievers

This area of research focuses on understanding the characteristics of gifted underachievers, primarily through comparing them with high-achieving peers on factors such as motivation, self-efficacy, attitudes toward teacher and school, self-regulation, study skills, goal valuation, attributions, learning strategies, school environments, or family circumstances (Baker et al., 1998; Clemons, 2008; Figg et al., 2012; Landis & Reschly, 2013; McCoach & Siegle, 2003b; Mofield et al., 2016; Ruban & Reis, 2006; Siegle & McCoach, 2018; White et al., 2018).

For example, Zimmerman and Martinez-Pons (1990) found that high- and low-achieving students differed in their academic motivation and self-regulation, with the former demonstrating some advantages in actively creating, adjusting, and modifying strategies in concert with new circumstances. As another example, Peterson (2002) examined 14 gifted adolescents identified in high school as at-risk for poor educational outcomes through a longitudinal lens. Peterson found that, 4 years after their high school graduation, most of these youths had improved on some measures, such as believing that they were better able to resolve conflict with their parents, gaining autonomy, and maintaining good emotional health. However, the majority still lacked career direction and had difficulty developing mature relationships.

Pathways to Underachievement

A substantial body of research on underachieving gifted students focuses on uncovering factors, causes, and probable pathways or trajectories of underachievement. A few studies of the developmental path of underachievement are particularly notable. For instance, Snyder and Linnenbrink-Garcia (2013) examined the developmental path of underachievement from early childhood through young adulthood using a longitudinal, person-focused approach.

Building upon their previous work, Snyder et al. (Snyder, Carrig, et al., 2019) studied how teacher-reported underachievement developed in 932 elementary students from the first to sixth grade, with a focus on students’ self-reported changes on four motivational beliefs: reading self-concept, task importance for reading, psychological cost value for reading (e.g., worrying little or very much about doing badly in reading), and global self-worth. The results of latent class growth analyses revealed two patterns of underachievement between the first and sixth grade. One was “sustained underachievement,” featuring chronic or constant underachievement over time. The other pattern was “growing underachievement,” or underachievement that increased with higher levels of academic challenge. For the sustained underachievement pattern, there was a significant relationship between moderately lower levels of self-perceived psychological cost value, self-worth, and achievement in middle school. In contrast, the growing underachievement pattern was found to be significantly related to lower levels of student self-concept and student-perceived task importance in the early stage of elementary education which further declined over time. These findings raised concerns over the degree of heterogeneity in the developmental pathways of gifted underachievement.

Underachievement Interventions

The third key research area explores interventions aimed at stopping or reversing underachievement. Interventions studied generally involved instructional or counseling interventions or some combination of these (Dowdall & Colangelo, 1982; McCoach & Siegle, 2014). Instructional interventions included part-time or full-time special classes for gifted underachievers (Fehrenbach, 1993; Whitmore, 1980). Generally, these involved modifications to the learning environment (e.g., ability grouping) to provide greater academic support for students, curriculum modifications (e.g., curriculum compacting and choice-based differentiation), and activities designed to teach students learning strategies and improve their self-worth and self-concept (Dowdall & Colangelo, 1982; Siegle & McCoach, 2018).

Counseling interventions primarily focused on helping students understand and respond to a range of psychological and/or social–emotional challenges they may experience, such as low motivation, low self-efficacy, bullying by peers, anxiety, or poor relationship skills (Pfeiffer & Prado, 2018). Counseling interventions typically involved individual, group, and/or family therapy or cognitive behavioral training sessions. Some included providing psychological and social–emotional support through close interaction and connections with mentors or providing peer support groups.

Reviews of Underachievement Interventions

Studies of effects of interventions targeting gifted underachievers report mixed results. Some reviews concluded that underachievement interventions were largely ineffective in helping gifted underachievers improve academic performance. For example, Dowdall and Colangelo’s (1982) review of research between 1960 and 1982 found that neither instructional nor counseling interventions showed much success, citing nine studies of counseling interventions deemed ineffective in improving achievement and only one study showing limited positive impact. They concluded that “The programs we reviewed were rather consistent in reporting the ineffectiveness of interventions. . . . Whatever spurious improvements can be documented, there is no evidence that these changes are reliable or long-lasting” (Dowdall & Colangelo, 1982, p. 182).

Findings of some recent reviews are generally in line with the conclusion above (McCoach & Siegle, 2014; Siegle & McCoach, 2018). Other reviewers have seen promise in some recent interventions and/or theoretical models, such as Rimm’s trifocal model (Rimm, 2008a, 2008b), Morisano et al.’s (2010) goal-setting intervention, and Rubenstein et al.’s (2012) motivational and self-regulation interventions. Reviewers have also noted some earlier programs, such as Whitmore’s (1980) full-time and Supplee’s (1990) part-time programs for gifted elementary underachievers, also seemed to be successful (see McCoach & Siegle, 2014; Siegle & McCoach, 2018). McCoach and Siegle (2014) concluded that research on the effectiveness of underachievement interventions is challenged by methodological problems that also plague much of the research in gifted education including small sample sizes, weakness of research designs, and the lack of use of advanced statistical analysis techniques. White et al. (2018) echoed many of McCoach and Siegle’s concerns, stating that the research designs of studies lacked the capacity to effectively distinguish between gifted achievers and underachievers. McCoach and Siegle also stressed the necessity to do more research on the effectiveness of different kinds of interventions for different types of gifted underachievers, for example, those with low self-efficacy versus those who selectively underachieve in certain areas only.

Overall, empirical evidence regarding the effects of underachievement interventions on gifted underachievers still remains fragmentary, mixed, and inconclusive. Therefore, meta-analyses may be especially useful for clarifying the effects of underachievement intervention. Our literature survey revealed an ongoing meta-analysis involving 53 studies of underachievement intervention (i.e., Snyder, Fong, et al., 2019). This meta-analysis found that interventions were moderately effective in improving achievement and psychosocial outcomes. However, the vast majority of study samples were typical students. No prior meta-analysis has reviewed research on the effects of underachievement interventions exclusively focusing on samples of gifted underachievers as in our review.

Achievement Outcomes

Determining and categorizing outcomes of interest is a crucial process in study coding. The current meta-analysis focused on two types of outcomes. The first was achievement outcomes. Nine of the 10 quantitative studies assessed the effects of underachievement interventions on 16 specific outcomes under eight different names. These outcomes included grades in English, math, science, history, and social studies, course grades (in general) and overall grade point average. They also included one outcome named “retention rates” which was based on the number of college students in the intervention and comparison groups who maintained a “full-time” course load of nine credits or more in the postintervention semester (Morisano et al., 2010). These outcomes were then further grouped into three relatively broad categories: course grades (n = 13), GPA (n = 2), and retention rates (n = 1).

Psychosocial Outcomes

Regarding psychosocial outcomes, a total of 44 specific outcomes under 18 different names appeared in eight of the 10 quantitative studies. The 44 specific outcomes were further coded into five relatively broad categories: (1) academic self-efficacy, referring to students’ self-perceived confidence to tackle new learning tasks, improve skills, or grasp course materials (Bandura, 1997); (2) goal valuation, regarding whether students value a learning task and find it meaningful enough to do the work (Siegle, 2013); (3) environmental perceptions, concerning how students feel about the atmosphere of their environments, such as finding teachers to be supportive or feeling insecure around other classmates (Siegle et al., 2017); (4) self-regulation/motivation as a psychosocial outcome category encompassing activities, beliefs, or behaviors relevant to self-regulated learning or achievement motivation (McCoach & Siegle, 2003a); and (5) psychosocial functioning, referring to a variety of mental, social, or emotional states or experiences, such as anxiety, depression, self-consciousness, distress, loneliness, feeling of belonging, or learned helplessness (Puhr et al., 2019; Ziolkowski et al., 2019).

Below are examples of specific outcomes for each of the five psychosocial outcome categories as shown in the respective parentheses. A complete list of all 44 psychosocial outcomes is reported in a table later:

The decision to assign each specific outcome to a category was based on careful considerations of several key aspects of the studies involved, especially regarding the types of interventions, the targeted outcomes of the intervention, how the outcomes were assessed, and references to specific theories of achievement motivation and self-regulated learning. For example, Niederdeppe’s (2009) study of a theatre-arts intervention reported an outcome called “affective disposition.” At face value, it seemed reasonable to place this outcome in the category of psychosocial functioning. However, it was found that “affective disposition” was used as an indicator of school engagement, measured by five additional questions added to the SAAS-R. Therefore, this outcome was ultimately included in the category of self-regulation/motivation. Another similar case was the outcome regarding the number of students who “Did not lose credit” in Johnson’s (2002) study, which was placed in the category of “Self-regulation/motivation.” This was because the outcome “Did not lose credit,” as well as “Completing in-class assignment” and “Completing homework”—two other psychosocial outcomes in the study—were all assessed to indicate students’ learning behaviors relevant to the constructs of self-regulation and motivation.

Also, the five-category outcome structure largely paralleled the AOM framework. Four of the five outcome categories (excepting psychosocial functioning) corresponded to the four key components of the AOM, that is, students’ beliefs in three key motivational areas including self-efficacy, goal valuation, environmental perceptions, as well as a self-regulation constituent. Moreover, the five outcome categories closely resembled the five factors of the SAAS-R, which is an instrumental presentation of the AOM. Relatedly, the SAAS-R was used to assess psychosocial outcomes in four of the 10 eligible quantitative studies reviewed.

Computing Mean Effect Sizes

Meta-analytic analyses were conducted for achievement and psychosocial outcomes separately. At the core of the analyses was computing mean effect sizes by integrating comparable outcomes across respective group of studies. The analysis was conducted with the Comprehensive Meta-Analysis (Borenstein et al., 2006) software, with effect size as the unit of analysis and a random-effects model (Cooper, 2017; Cooper et al., 2009). There were two reasons for using effect size instead of study as the unit of analysis. First, the total number of studies was rather small (n = 10), and the number was even smaller for studies with achievement outcomes (n = 9) and for those with psychosocial outcomes (n = 8). In contrast, the number of effect sizes for achievement and psychosocial outcomes was 16 and 44, respectively. Second, effect size independence was achieved first because meta-analysis was conducted separately on achievement and psychosocial outcomes; additionally, independence of effect sizes was made possible because of the wide variety of specific outcomes and the approach used to create relevant outcome categories (Lipsey & Wilson, 2001).

One may argue that a fixed-effects rather than a random-effects model might be a reasonable choice given that the current meta-analysis was based on only 10 quantitative studies. As it is advised that the choice of models should be primarily in line with the kind of inferences meta-analysts desire to make (Borenstein et al., 2009; Konstantopoulos & Hedges, 2009), a random-effects model was judged to be appropriate for the current meta-analysis, which aimed to make generalizations to a broad population regarding the effectiveness of underachievement interventions. Under a random-effects model, a mean effect size was calculated with each contributing study being weighted by the inverse of variance (i.e., dividing the value of one by the variance) with the Hedges-Vevea (Hegdes & Vevea, 1998) method. Therefore, a random-effects model has the additional advantage of balancing the weights of effect sizes between studies and thus reducing the risk that the mean effect size being overly influenced by any single study (Borenstein et al., 2010). We also conducted analyses with a fixed-effect model. The results appeared to be consistent under the two models. Therefore, we proceeded with only reporting the results from a random-effects model.

Heterogeneity Analysis and Testing for Moderators

Heterogeneity analysis was conducted to statistically assess whether there was significant variation in average effect sizes between those of relevant studies reviewed. In the current meta-analysis, for instance, there were nine studies and therefore nine average effect sizes for achievement outcomes. If average effect sizes varied significantly across studies, testing for moderator analysis (also called subgroup analysis) was then conducted with the method analogous to the analysis of variance (ANOVA). The goal of testing for moderator analysis was to explore possible sources of heterogeneity in the studies’ effect sizes (Borenstein et al., 2009; Hedges & Pigott, 2004). Unlike heterogeneity analysis, which evaluates the variations among average effect sizes of studies, testing for moderator analysis uncovers whether certain confounding variables moderated the effectiveness of an intervention on study outcomes.

The presence of such moderating influence of a confounding variable was indicated by whether the average effect sizes of subcategories of a variable varied significantly. For example, if the average effect size of counseling interventions and that of instructional interventions differed significantly as shown by the statistically significant heterogeneity between the two subcategories of interventions (e.g., the p value of the test for null hypotheses of Q_b was less than .05, .01, or .001), then, by implication, intervention type would be a moderator of underachievement interventions’ effects on achievement outcomes.

From approximately 40 variables containing extracted data regarding major characteristics of the interventions and studies, seven were selected for testing for moderator analysis because they were most relevant to the questions that this meta-analysis intended to address (Littell et al., 2008). They included intervention type (i.e., a counseling or instructional intervention) and intervention maturity (i.e., replicating or piloting), both having two subcategories. The other five variables were intervention dosage (i.e., 90-150 minutes “low,” 300-900 minutes “moderate,” or 1,200-1,760 minutes “high”), development level (i.e., elementary, middle, or high school, or college), data source (i.e., self-reporting, teacher-reporting, or school record), report type (i.e., journal article or a dissertation), and country (i.e., in which country was the study conducted: United States, Germany, or Canada).

Assessing Publication Bias

Publication bias may arise when a meta-analysis was based on a disproportionally large number of small studies with significant or larger effect sizes. Presence of severe publication bias may lead to biased (usually overestimated) results that weaken the validity of a meta-analysis (Rothstein et al., 2005). To evaluate whether mean effect sizes varied due to possible publication bias, the effect sizes were first visually assessed for the symmetry or lack thereof of a funnel plot (Peters et al., 2008). The trim and fill procedures (Duval & Tweedie, 2000) were subsequently carried out, which involved iteratively removing or filling in effect sizes until publication bias no longer appeared to be an issue of concern, as shown by a symmetrical distribution of the funnel plot. Last, the recalculated (adjusted) mean effect sizes (including 95% confidence intervals and p values) through the trim and fill procedures were presented along with the original (observed) mean effect sizes and other relevant statistics.

Evaluating Quality of the Evidence

Problems with the design and implementation of intervention studies raise concerns about quality of the evidence and ultimately the validity of their findings. The current meta-analysis conducted basic evaluation of quality of the evidence of the 10 quantitative studies by assessing risk of bias often responsible for either overestimation or underestimation of the true intervention following a method as described in the Cochrane Handbook for Systematic Reviews of Interventions (see Higgins et al., 2017). The evaluation of quality of the evidence primarily focused on risk of selection bias and detection bias because these two sources of biases were especially relevant and important.

In the context of this meta-analysis, for example, the risk of selection bias might be high if the underachieving gifted students in the intervention condition happened to have better self-regulation skills at baseline than gifted underachievers in the comparison group because of the preintervention differences in self-regulation skills, the intervention itself, as well as a range of factors that may collectively affect students’ postintervention performance. Consequently, the quality of the research evidence would be questionable, and so would the validity of the study findings. Similarly, risk of detection bias may arise when study outcomes are affected when the persons who assess outcomes are not blind to whether students are in the intervention or comparison condition. The risk of detection bias might be high especially on psychosocial outcomes in this meta-analysis because the vast majority of such data were assessed through various student self-reported or teacher-reported rating scales. Furthermore, these intervention studies were also evaluated in light of whether they addressed issues of the fidelity and quality of implementation.

Key Features of Quantitative Studies

Table 1 presents the key features of 10 quantitative studies eligible to be included in the current meta-analysis. Four studies appeared in published journal articles and six studies in dissertations. Four studies were experimental studies that involved using random assignment of participants (i.e., Morisano et al., 2010; Sivaraman, 2012; Stoeger & Ziegler, 2005; Sullivan, 2016). All six dissertations employed a mixed-method research design, but they largely leaned on quantitative data and supplemented them with some limited narrative information. Mathematics, reading, and English language arts were the most frequently involved subject areas across the studies. The intervention maturity for seven studies fell in the category of replicating and the remaining three were cases of piloting. The number of sample sizes ranged from 28 to 85 across the 10 studies, with 398 (83.3%) of the total 478 gifted participants being underachievers.

OPEN IN VIEWER

Table 1.

Key Features of Quantitative Studies of Underachievement Interventions.

Study a	Report	Research design	Subject	Intervention maturity	Underachievers N	Total N b
Castro (2008)	Dissertation	Quasi-experimental		Piloting	43	43
Cortés-Cabello (2013)	Dissertation	Quasi-experimental	Mathematics, science, ELA, and social studies	Replicating	45	45
Johnson (2002)	Dissertation	Mixed-method		Piloting	32	32
Morisano et al. (2010)	Journal article	Experimental (with random assignment)		Replicating	85	85
Niederdeppe (2009)	Dissertation	Mixed-method	Creative arts	Replicating	34	34
Obergriesser and Stoeger (2015)	Journal article	Quasi-experimental	Reading and writing	Replicating	24	85
Rubenstein et al. (2012) Study 1	Journal article	Quasi-experimental	Mathematics or reading/ELA	Piloting	46	46
Sivaraman (2012)	Dissertation	Experimental (with random assignment)		Replicating	44	44
Stoeger and Ziegler (2005)	Journal article	Experimental (with random assignment)	Math	Replicating	36	36
Sullivan (2016)	Dissertation	Experimental (with random assignment)	Math	Replicating	9	28

Note. ELA = English language arts.

a

Of the 10 quantitative studies, 2 were conducted in Germany (Obergriesser & Stoeger, 2015; Stoeger & Ziegler, 2005), 1 conducted in Canada (Morisano et al., 2010), and the remaining 7 were conducted in the United States. ^bThe total number of gifted students equaled to the number of gifted underachievers in eight studies because these studies only included gifted underachieving students as study participants. In Obergriesser and Stoeger’s (2015) and Sullivan’s (2016) studies, the percentage of gifted underachievers relative to total participants was 28.2% (24/85 = 28.2%) and 32.1% (9/28 = 32.1%), respectively.

Characteristics of Underachievement Interventions

As shown in Table 2, interventions investigated in these studies included self-regulated learning (Obergriesser & Stoeger, 2015; Stoeger & Ziegler, 2005), motivation and self-regulation (Rubenstein et al., 2012, Study 1), goal setting (Morisano et al., 2010; Sivaraman, 2012; Sullivan, 2016), theatre arts (Niederdeppe, 2009), strength-based academic (Cortés-Cabello, 2013) intervention, and support groups or networks involving peers, parents, teachers, and/or mentors (Castro, 2008; Johnson, 2002).

OPEN IN VIEWER

Table 2.

Key Characteristics of Underachievement Interventions: Quantitative Studies.

Study	Intervention	Implementation setting	Grade	Duration	Frequency	Total minutes
Castro, 2008	Peer and mentoring support network	Small group during the 7th period of the school day	9th-11th	13 weeks (11 sessions)	Weekly, (60 minutes/session)	660
Cortés-Cabello, 2013	Strength-based academic intervention	Remedial class for gifted underachievers	6th-8th	9 weeks	40 sessions (30 minutes/session)	1,200
Johnson, 2002	Student, parent, and teacher support groups	The district’s gifted education program	5th-8th	32 weeks	Student groups 16 biweekly sessions (50 minutes/session)	800
					Parent groups 8 monthly sessions (1 hour/session)	480
					Teacher group 8 weekly sessions (1 hour/session)	480
					Total	1,760
Morisano et al., 2010	Goal-setting intervention	Regular instruction	College	2.5 hours	One-time	150
Niederdeppe, 2009	Theatre-arts intervention	After school program	6th-8th	10 weeks	Weekly, 90 minutes/week	900
Obergriesser & Stoeger, 2015	Self-regulated learning	Regular instruction	Primary school	7 weeks	Daily, 40-60 minutes each day	1,750
Rubenstein et al., 2012, Study 1	Motivational and self-regulation interventions	Regular instruction	Middle school (5th)	6-9 weeks	Weekly, 30-60 minutes each week	338
Sivaraman, 2012	Goal-setting intervention	After school small-group	10th-12th	2 weeks (2 sessions)	1.25 hours/session	150
Stoeger & Ziegler, 2005	Self-regulated training	Regular instruction	4th	6 weeks	Weekly	~300
Sullivan, 2016	Goal-setting intervention	Cross-grade math grouping class a	10th-12th	9 weeks	10-15 minutes/week	90

a

The class was called regular math analysis in the school. It was a high-level math course, consisting of students of 10th to 12th grades. Therefore, it was indeed a cross-grade math grouping class.

Four interventions were implemented in regular instruction classrooms, two in after-school time (Niederdeppe, 2009; Sivaraman, 2012), one in a cross-grade math ability-grouping class (Sullivan, 2016), and one in a remedial class for gifted underachievers (Cortés-Cabello, 2013). Two interventions were carried out with students in elementary school (Obergriesser & Stoeger, 2015; Stoeger & Ziegler, 2005), four in middle school (Cortés-Cabello, 2013; Johnson, 2002; Niederdeppe, 2009; Rubenstein et al., 2012, Study 1), three in high school (Castro, 2008; Sivaraman, 2012; Sullivan, 2016), and one intervention was with college students (Morisano et al., 2010). The interventions lasted as short as 90 minutes (Sullivan, 2016) and as long as approximately 30 hours (1,760 minutes) across dozens of sessions over a 32-week period (Johnson, 2002). Most of the intervention activities took place on a weekly basis.

Table 3 presents key activities and underlying theoretical frameworks of the interventions. For example, key interventional activities in the Sivaraman’s (2012) study involved teachers working with students to enhance time management skills, self-efficacy, and self-reflective learning behaviors. Although the names of the interventions varied across studies, the goals of the interventions and the kinds of activities involved suggest that they were either instructional (n = 7) or counseling interventions (n = 3) at their cores. The majority of interventions were designed on the basis of underlying frameworks of achievement motivation and/or self-regulated learning including theories pertaining to a variety of constructs such as self-efficacy, attribution, goal valuation, attitudes toward self, teachers and school, social support, and mentoring.

OPEN IN VIEWER

Table 3.

Underachievement Interventions, Activities, and Theoretical Frameworks: Quantitative Studies.

Study	Intervention and activities	Theoretical framework
Castro, 2008	Peer and mentoring support network	Social support and mentoring
	7 mentors, each working with 2-3 gifted underachievers, supported 23 gifted underachieving Latino high school students to develop skills related to:	The intervention was specially designed for the study based on theories supported by gifted educational research such as McCluskey et al. (2005) and Olszewski-Kubilius & Lee (2004). “. . . if gifted underachieving students were monitored, given guidance and support from an adult in a forum where they could receive the support and encouragement of their peers, the work habits and achievement of participating underachieving gifted students would improve, . . .” (Castro, 2008, p. ix)
	• Improved work habits
	• Self-regulation
	• Collective accountability
Cortés-Cabello, 2013	Strength-based academic intervention	Theories of achievement motivation and strengths development
	Students assessed their strengths online with the Clifton Youth Strengths Explorer. They then received training on how to develop and take advantage of their strengths	The Clifton Youth Strengths Explorer (Gallup Organization, 2007) posits that paying attention to one’s strengths and learning how to develop them further can increase students’ motivation. Subsequently, raised motivation is linked to personal growth and improved achievement (Clifton et al., 2006)
Johnson, 2002	Student, parent, and teacher support groups	Beneficial impact of student, parent, and teacher support groups
	• 16 student support sessions held during the course of the school year	The interventions were specially designed for the study based on the theory that providing support groups for gifted students can help them improve coping skills, self-awareness, self-discipline; Parent support groups may help address issues on academics, social skills, motivation, discipline, peer relations, and coping strategies; Teacher support groups can help improve understanding and cooperation among regular and gifted education teachers
	• 8 monthly sessions held to improve parental social–emotional support systems for gifted students
	• 8 weekly sessions held with the teachers
Morisano et al., 2010	Goal-setting intervention	Goal-setting theory
	College students received online training on setting specific personal goals and on strategies for improving achievement	Originally developed by Peterson and Mar (2004), the intervention was adapted for use with university students based on the theory that explicitly setting appropriate and meaningful goals for a given task can improve performance on the task because clear goals can lead to a chain of positive changes such as increased self-regulation, attention, effort, enthusiasm, persistence, and use of more efficient strategies (Locke & Latham, 2002)
Niederdeppe, 2009	Theatre/creative-arts intervention	Opportunities to experience social support and cultural acceptance
	Students participated in a 10-week after school theatre-arts program to produce a modified play of Romeo and Juliet	Students participated in a modified play of Romeo and Juliet based on the theory that a creative arts intervention provides gifted underachievers with opportunities to experience new, challenging and creative endeavors, to engage in in creative and critical thinking, and to interact with peer groups (cf. Catterall, 2007)
Obergriesser & Stoeger, 2015	Self-regulated learning	Self-regulated learning theory
	Students learn and practice using:	This intervention is founded upon a simplified social–cognitive theory–based model of self-regulated learning intended to promote learning through the use of cognitive and meta-cognitive strategies. Research shows that elementary school students benefit from earlier interventions on this area (Stoeger et al., 2014; Zimmerman et al., 1996)
	• Cognitive strategies (i.e., text-reduction)
	• Meta-cognitive strategies (i.e., self-assessment, goal setting, and monitoring)
Rubenstein et al., 2012, Study 1	Motivational and self-regulation interventions	The achievement orientation model
	Teachers led students to practice strategies addressing:	This version of the AOM intervention was designed specifically for this study based on four motivational theories:
	• Self-efficacy	• Self-efficacy theory
	• Meaningfulness	• Attribution theory
	• Environmental perceptions	• Expectancy-value theory
		• Person-environmental fit theory
	Additionally, students experienced interest-based projects and curriculum compacting intended to promote self-regulation	Students’ self-perceptions in self-efficacy, goal valuation, and environmental perception impact their motivation and academic achievement (Siegle & McCoach, 2005)
Sivaraman, 2012	Goal-setting intervention	Goal-setting theory
	High school students undertook two-step goal setting activities wherein:	The intervention was originally developed by Peterson and Mar (2004) and adapted for use with high school students based on the theory that explicitly setting appropriate and meaningful goals for a given task can improve performance on the task because clear goals can lead to a chain of positive changes such as increased self-regulation, attention, effort, enthusiasm, persistence, and use of more efficient strategies (Locke & Latham, 2002).
	• They envisioned what their desired future would look like.
	• They then set two specific goals to reach their ideal future.
Stoeger & Ziegler, 2005	Self-regulated training	Self-regulated learning theory
	Teachers led students in activities intended to develop skills on:	Developed on the basis of Zimmerman et al.’s (1996) model of self-regulated learning, the self-regulated training program was designed to address several causes of underachievement through
	• Time management	• Improving inadequate learning and work behavior
	• Self-efficacy	• Assessing one’s own strengths and weaknesses
	• Self-reflective learning behaviors	• Improving self-concept, motivation, and self-control
		• Acting as managers of one’s own learning
Sullivan, 2016	Goal-setting intervention	Goal valuation/self-regulation
	Students met with the gifted resource teacher for 10-15 minutes a week to complete different goal-setting exercises	The training modules developed by McCoach (2002) were based on Hidi & Harackiewicz’s (2000) theory that children can enhance their self-regulation and motivation by setting clear achievement goals which in turn affect how they handle academic tasks

Note. AOM = achievement orientation model.

Identification of Gifted Underachievers

Table 4 provides summaries regarding how gifted achievers and underachievers were identified in each study. Three out of the 10 studies identified gifted students as those who scored the top percentiles (e.g., IQ = 120, 95%, or 96% percentiles or above) on some type of cognitive ability test and/or demonstrated superior performances on state or local standardized achievement tests (Johnson, 2002; Rubenstein et al., 2012, Study 1; Sivaraman, 2012). Three studies provided no detailed identification criteria other than mentioning that gifted students were identified by a local school or school district (Castro, 2008; Cortés-Cabello, 2013; Sullivan, 2016). Two studies considered students to be gifted if they scored in the top range on Raven’s Standard Progressive Matrices Test (German version; 90% percentile, Obergriesser & Stoeger, 2015; scored 130 or above, Stoeger & Ziegler, 2005). One study (Morisano et al., 2010) assumed that undergraduates attending a highly selective university were gifted or high achievers.

OPEN IN VIEWER

Table 4.

Identification of Gifted Achievers and Underachievers: Quantitative Studies.

Study	Identification criteria a
	Gifted achievers	Gifted underachievers
Castro, 2008	102 Latino students were identified as gifted by the high school	43 (44%) of the 102 gifted students were identified as “at-risk” by school counselors and determined to be moderately to severely underachieving based on the researcher’s criteria (not specified)
Cortés-Cabello, 2013	Participants were drawn from a pool of students identified as gifted by the school district	Gifted students who earned core course grades of “C” or lower in one or more core courses during the Fall 2011 school semester
Johnson, 2002	Students in a pullout gifted program (the top 2% of students in the school district) called Quest. They also met the following eligibility criteria for this study:	Quest program students were considered to be underachieving based on: 1. Insufficient participation in Quest activities 2. Insufficient completion rate of in-class assignments 3. Insufficient completion rate of homework assignments 4. Loss of credit on special projects due to failure to complete them prior to established deadlines 5. Exhibited low self-discipline and motivation, were disorganized, and wasted excessive amounts of class timeOnly 23 of 36 students in the program were achieving at expected levels in Quest class
	1. IQ scores at or above the 96th percentile on an approved cognitive assessment instrument, or
	2. Superior performance on a standardized measure of academic achievement
Morisano et al., 2010	Not given	Undergraduates from McGill University met three criteria:
		1. Taking a full-time course load (nine credits) each semester while in the study
		2. On official academic probation or had a cumulative GPA under 3.00
		3. Experiencing academic difficulty
Niederdeppe, 2009	410 (top 12th percentile) Latino students of a large urban middle school were identified as gifted and enrolled in honors sections of their four academic courses (English, math, history, and science) using three criteria:	Gifted students were first identified as underachievers by having lower than a B (3.0) GPA on the previous semester’s grades in their four academic coursesz-standardized GPA in main subjects (i.e., math, German, and basic science) at least one standard deviation below their z-standardized IQ scoreMet one of the two criteria: 1. Grades in the bottom half of the class in reading/language arts and/or mathematics, or a C average or below in reading/language arts and/or mathematics and 2. Recommendation by a classroom teacher, gifted specialist, and/or counselor as a bright underachiever
	1. “B” grade average or better
	2. 392+ language arts score on the California Standards test
	3. 401+ math score
Obergriesser & Stoeger, 2015	Scored at or above the 90th percentile on Raven’s Standard Progressive Matrices Test (German)
Rubenstein et al., 2012, Study 1	Teachers nominated students as gifted if they met one of the two criteria:
	1. Individual IQ test score (Stanford-Binet LM or Wechsler Intelligence Scale for Children, 3rd ed.) of at least 120 (93.3 percentile) from no earlier than age 6
	2. Composite standardized achievement test scores in the 90th percentile locally from within the past 3 years
Sivaraman, 2012	School district used three criteria to define giftedness:	127 (52%) of 362 academically gifted students were identified as underachievers because they had a GPA of 3.0 or below, and therefore would not be qualified to enter the University of California system
	1. Giftedness of intellectual ability:
	Scoring at or above the 95th percentile on a reasoning test
	2. Giftedness in the form of high achievement:
	• Obtaining an advanced score of 450 or above on the California Standards Test in both English and mathematics in two consecutive years and
	• Obtaining As in these subjects for the second and third semesters
	3. Giftedness in a specific academic ability:
	• Advanced scores of 450 or above on the California Standards Test in three consecutive years and
	• Grades of As or 4s in either English language arts or mathematics in both the second and third trimesters
	4. Giftedness of intellectual ability:
	Scoring at or above the 95th percentile on a reasoning test
	5. Giftedness in the form of high achievement:
	• Obtaining an advanced score of 450 or above on the California Standards Test in both English and mathematics in two consecutive years and
	• Obtaining As in these subjects for the second and third semesters
	6. Giftedness in a specific academic ability:
	• Advanced scores of 450 or above on the California Standards Test in three consecutive years and
	• Grades of As or 4s in either English language arts or mathematics in both the second and third trimesters
Stoeger & Ziegler, 2005	Students scored 130 (97.8 percentile) or above on Raven’s Standard Progressive Matrices (SPMs) were identified as gifted	Underachievers were identified as those whose z-standardized GPA in main subjects was at least one standard deviation below their z-standardized IQ score
Sullivan, 2016	264 Students, representing the top 13.3 percentile of students, were identified as gifted by a public high school which had a total of 1,989 students	Students were identified as underachieving if:
		1. Received a C, D, or an E at the first quarter progress report in their math analysis class and
		2. Received a C, D, or E on composite scores at the end of the first quarter

Note. GPA = grade point average.

a

Boldfaced contents denote the key quantified standards employed.

Only two studies (Obergriesser & Stoeger, 2015; Stoeger & Ziegler, 2005) quantified the discrepancy between students’ cognitive ability (potential) and actual achievement performance. They did so by identifying giftedness as scoring at the 90% percentile or above on the German version of Raven’s Standard Progressive Matrices Test, and determining underachievers as those who met the above criteria but whose z-standardized GPAs in math, German/native language, or basic science was at least one standard deviation below their z-standardized IQ scores. Three studies identified gifted underachievers as those with Grade C or lower either in a single or in multiple subject areas (Cortés-Cabello, 2013; Rubenstein et al., 2012, Study 1; Sullivan, 2016). Three studies identified gifted underachievers as those with a GPA of 3.0 or below (Morisano et al., 2010; Niederdeppe, 2009; Sivaraman, 2012). Castro (2008) provided no specific benchmarks but reported that 43 (44%) of the 102 gifted Latino students were determined to be moderately to severely underachieving by school counselors based on the researchers’ criteria.

In sum, gifted underachievers were identified with relatively clear criteria in most cases. Across the 10 studies, the highest IQ benchmark for qualifying as gifted was a score of 130 (top 2.2% percentile; Stoeger & Ziegler, 2005) and the lowest threshold was the top 13.3% percentile of 1,989 students in a public high school (Sullivan, 2016). Gifted underachievers were mostly identified following Reis and McCoach’s (2000) definition, which emphasizes discrepancies between expected achievement based on one’s intellectual potential and demonstrated performance. Among the samples of these 10 studies, as high as 52% of academically gifted students were identified as underachievers (Sivaraman, 2012).

Outcomes and Effect Sizes

Table 5 lists specific outcomes, outcome categories, outcome assessments, and effect sizes (Hedges’s gs) along with corresponding variances by each of the 10 studies. Nine of the 10 studies reported achievement outcomes while Obergriesser and Stoeger’s (2015) study only provided psychosocial outcomes. Eight of the 10 studies assessed psychosocial outcomes, and the remaining two only studied achievement outcomes (Cortés-Cabello, 2013; Rubenstein et al., 2012, Study 1). Six studies assessed both achievement and psychosocial outcomes.

OPEN IN VIEWER

Table 5.

Specific Outcomes, Outcome Categories, and Outcome Assessments, and 60 Effect Sizes Extracted.

Study	Intervention	Specific outcome	Outcome category	Assessment	Hedges’s g	Variance
Castro, 2008	Peer and mentor network	Science grades	Course grades	Course grades	−0.14	0.21
		Mathematical grade	Course grades	Course grades	0.32	0.20
		English grades	Course grades	Course grades	−0.02	0.18
		Attitudes toward teachers	Environmental perceptions	Rating scale	0.27	0.08
		Attitudes toward school	Environmental perceptions	Rating scale	0.22	0.08
		Goal valuation	Goal valuation	Rating scale	0.13	0.08
		Motivation and self-regulation	Self-regulation/motivation	Rating scale	0.45	0.09
		Academic self-perception	Self-efficacy	Rating scale	0.14	0.08
Cortés-Cabello, 2013	Strength-based remediation intervention	Course grade (English)	Course grades	Course grades	−0.67	0.09
		Course grade (social studies)	Course grades	Course grades	0.49	0.09
		Course grade (science)	Course grades	Course grades	−0.49	0.09
		Course grade (math)	Course grades	Course grades	0.49	0.09
Johnson, 2002	Student, parent, and teacher support groups	Course grades	Course grades	Course grades	−0.06	0.06
		Did not lose credit	Self-regulation/motivation	Interview records	0.35	0.10
		Completing in-class assignment	Self-regulation/motivation	Rating scale	0.94	0.07
		Completing homework	Self-regulation/motivation	Rating scale	1.49	0.08
Morisano et al., 2010	Goal-setting intervention	Retention rates	Retention rates	Retention rates	1.73	0.65
		Reduced negative affect	Psychosocial functioning	Rating scale	0.46	0.05
		Enthusiasm for the study	Self-regulation/Motivation	Rating scale	0.16	0.05
		College GPA	GPA	College GPA	0.50	0.05
Niederdeppe, 2009	Theatre-arts intervention	Affective disposition	Self-regulation/motivation	Rating scale	0.39	0.23
		Science grades	Course grades	Course grades	−0.68	0.24
		History grades	Course grades	Course grades	−0.41	0.23
		Attitudes toward teachers	Environmental perceptions	Rating scale	0.64	0.23
		Attitudes toward school	Environmental perceptions	Rating scale	0.65	0.23
		Goal valuation	Goal valuation	Rating scale	−0.05	0.23
		Science work habits	Self-regulation/motivation	Rating scale	−0.42	0.24
		Science cooperation marks	Self-regulation/motivation	Rating scale	−0.62	0.23
		Motivation and self-regulation	Self-regulation/motivation	Rating scale	0.48	0.23
		History work habits	Self-regulation/motivation	Rating scale	−0.78	0.23
		History cooperation marks	Self-regulation/motivation	Rating scale	−0.44	0.23
		Academic self-perception	Self-efficacy	Rating scale	0.64	0.23
Obergriesser & Stoeger, 2015	Self-regulated learning	Anxiety	Psychosocial functioning	Rating scale	0.03	0.16
		Text-reduction strategies	Self-regulation/motivation	Rating scale	0.30	0.16
		Self-efficacy	Self-efficacy	Rating scale	0.12	0.16
Rubenstein et al., 2012, Study 1	Self-regulation/ motivation	Course grade	Course grades	Course grades	0.38	0.04
Sivaraman, 2012	Goal-setting intervention	Attitudes toward teachers	Environmental perceptions	Rating scale	0.15	0.18
		Attitudes toward school	Environmental perceptions	Rating scale	0.16	0.18
		Time perspective	Goal valuation	Rating scale	0.34	0.18
		Planning ahead	Goal valuation	Rating scale	−0.05	0.18
		Goal valuation	Goal valuation	Rating scale	0.10	0.18
		Future orientation	Goal valuation	Rating scale	0.22	0.18
		Anticipation of future consequences	Goal valuation	Rating scale	0.00	0.18
		Semester GPA	GPA	Course grades	0.58	0.18
		Motivation and self-regulation	Self-regulation/motivation	Rating scale	−0.07	0.18
		Theory of Intelligence	Self-efficacy	Rating scale	0.12	0.98
		Academic self-perception	Self-efficacy	Rating scale	0.21	0.18
Stoeger & Ziegler, 2005	Self-regulated learning	Helplessness	Psychosocial functioning	Rating scale	−0.39	0.22
		Math exam scores	Course grades	Course grades	−0.51	0.23
		Time management	Self-regulation/motivation	Rating scale	0.53	0.22
		Self-reflection of own learning	Self-regulation/motivation	Rating scale	0.48	0.22
		Persistence/willingness to exert effort	Self-regulation/motivation	Rating scale	0.19	0.22
		Self-efficacy	Self-efficacy	Rating scale	0.32	0.23
		Aspiration level for the subject of mathematics	Self-efficacy	Rating scale	−0.28	0.22
Sullivan, 2016	Goal-setting intervention	Motivation and self-regulation	Self-regulation/motivation	Rating scale	−0.14	0.79
		Goal valuation	Goal valuation	Rating scale	0.22	0.78
		Course grades	Course grades	Course grades	0.39	0.40
		Attitudes toward teachers	Environmental perceptions	Rating scale	0.37	0.82
		Attitudes toward school	Environmental perceptions	Rating scale	0.86	0.80
		Academic self-perception	Self-efficacy	Rating scale	0.67	1.10

Note. GPA = grade point average.

A total of 60 specific outcomes of interest were extracted, including 58 continuous and two dichotomous outcomes (i.e., retention rates and the number of students who did not lose credits). Odds ratios were obtained for each of the two dichotomous outcomes and then converted to Hedges’s gs of 1.73 and 0.35, respectively. Of the 60 effect sizes corresponding to the total number of specific outcomes, 16 (26.7%) were associated with achievement outcomes, ranging from −0.68 to 1.73, eight of them being negative and eight being positive. Forty-four effect sizes (73.3%) were associated with psychosocial outcomes, ranging from −0.78 to 1.49, with 11 of them being negative and 33 being positive.

Nearly all (n = 43) of the psychosocial outcomes were measured with some type of rating scale or subscale. However, one outcome in Johnson’s (2002) study (“Did not lose credit”) used as an indicator of students’ self-regulation /motivation was based on teachers’ recollections during interviews. Four out of the 10 studies used SAAS-R (McCoach & Siegle, 2003a), a psychometrically validated instrument designed to measure adolescents’ perceptions on five factors (described in the “Introduction” section earlier) central to address the underachievement of gifted secondary students.

Mean Effect Sizes

The mean effect sizes of the nine studies (and respective interventions) ranged from −0.54 to 0.87, four of them being negative (Cortés-Cabello, 2013; Johnson, 2002; Niederdeppe, 2009; Stoeger & Ziegler, 2005) and five being positive (Castro, 2008; Morisano et al., 2010; Rubenstein et al., 2012, Study 1; Sivaraman, 2012; Sullivan, 2016), but none of them was statistically and significantly different from zero, p values ranging from .071 to .891. Overall, there was no evidence that the interventions significantly affected the academic achievement of gifted underachievers, g = 0.09, k = 16 (effect sizes), 95% CI = [−0.12, 0.30], p = .387. Interventions’ effects on achievement outcomes varied significantly, Q_b = 30.80, df = 15, p = .009, I² = 51.30. Additionally, there was no evidence of publication bias affecting the effect sizes, suggesting that the observed overall mean effect size of 0.09 was neither an overestimation nor underestimation of the true effectiveness of the interventions.

Mean Effect Sizes by Specific Achievement Outcomes

The mean effect size for each of the eight specific achievement outcomes ranged from −0.45 to 1.73, three were negative (English, history, and science grades) and five were positive (course grades, GPA, math grades, retention rates, and social studies grades). On average, gifted underachievers receiving the interventions had significantly higher GPAs than their comparison peers, g = 0.51, k = 2 (effect sizes), 95% CI = [0.13, 0.90], p = .008. Gifted underachievers’ retention rates, an indicator of the number of college students whose course load remained at full-time status (nine credits or more) in the postintervention semester, was significantly higher than those of their nonintervention comparison peers, g = 1.73, k = 1 (effect size), 95% CI = [0.15, 3.31], p = .031. However, gifted underachievers receiving the interventions scored significantly lower in science grades than their comparison peers, g = −0.45, k = 3 (effect sizes), 95% CI = [−0.89, −0.01], p = .044. Interventions’ mean effects on academic performance varied significantly across the eight outcomes, Q_b = 30.80, df = 15, p = .009, I² = 51.30.

Mean Effect Sizes by Achievement Outcome Categories

The mean effect sizes for the outcome category of course grades, GPA, and retention rates were −0.04, 0.51, and 1.73, respectively. On average, gifted underachievers receiving the interventions performed no different on course grades from their comparison peers, g = −0.04, k = 13 (effect sizes), 95% CI = [−0.29, 0.22], p = .786. Not surprisingly, the interventions’ effectiveness on GPA and retention rates remained the same as above because the number of relevant effect sizes was the same regardless of whether they were treated as a specific outcome as above or as an outcome category here. Again, intervention mean effects on gifted underachievers’ academic performance varied significantly across the three outcome categories, Q_b = 30.80, df = 15, p = .009, I² = 51.30.

Testing for Moderators

Table 7 shows the results of analyses regarding whether underachievement interventions’ effectiveness on students’ academic achievement was moderated by any confounding factors. Intervention dosage appeared to be the only significant moderator out of the seven variables tested, Q_b = 6.88, df = 2, p = .032. The three levels of intervention dosage were based on the total number of minutes that the interventions lasted (see Table 2). Counterintuitively, low intervention dosage (90-150 minutes; g = 0.57, p = .002) appeared to be much more effective in improving the academic performance of gifted underachievers than moderate (300-900 minutes; g = −0.02, p = .914) or high (1,200-1,760 minutes; g = −0.05, p = .843) levels of dosages.

OPEN IN VIEWER

Table 7.

Testing for Moderators of the Effectiveness of Underachievement Interventions on Achievement Outcomes.

Variable	k a	Hedges’s g b	Variance	95% CI		p value	Heterogeneity statistics
				Lower	Upper		Q b	df	p value
Intervention type							2.08	1	.149
Counseling intervention	6	−0.12	0.02	−0.43	0.190	.549
Instructional intervention	10	0.20	0.03	−0.15	0.548	.136
Intervention dosage							6.88	2	.032
Low (90-150 minutes)	4	0.57	0.03	0.21	0.92	.002**
Moderate (300-900 minutes)	7	−0.02	0.03	−0.35	0.31	.914
High (1,200-1,760 minutes)	5	−0.05	0.05	−0.50	0.41	.843
Intervention maturity							0.15	1	.701
Piloting	5	0.16	0.02	−0.11	0.421	.250
Replicating	11	0.07	0.04	−0.31	0.441	.730
Developmental level							4.45	3	.217
Elementary school	1	−0.51	0.23	−1.45	0.43	.291
Middle school	8	−0.06	0.03	−0.41	0.29	.739
High school	5	0.22	0.04	−0.19	0.62	.296
College	2	0.87	0.32	−0.24	1.99	.126
Data source							0.06	1	.808
Teacher-reporting	12	0.06	0.02	−0.20	0.32	.643
School record	4	0.17	0.17	−0.65	0.98	.687
Report type							1.95	1	.162
Journal article	4	0.37	0.06	−0.11	0.86	.129
Dissertation	12	−0.02	0.02	−0.30	0.25	.873
Country							3.44	2	.179
Canada	2	0.87	0.32	−0.24	1.99	.126
Germany	1	−0.51	0.23	−1.45	0.43	.291
United States	13	0.03	0.02	−0.23	0.29	.831

Note. CI = confidence interval; df = degrees of freedom.

a

The sum of k for each variable was 16, the total number of effect sizes for achievement outcomes. ^bThe Hedges’s g denoted the mean effect size for each subcategory of variables. **p < .001.

However, a further examination revealed that it is highly possible that this finding was merely a result of chance. This was because the two studies in which the dosages of the interventions were low (150 minutes in both Morisano et al., 2010, and Sivaraman, 2012) happened to have almost the largest effect sizes. Specifically, effect size was 1.73 for retention rates and 0.50 for college GPA in Morisano et al.’s (2010) study; similarly, the effect size was 0.58 for semester GPA in Sivaraman’s (2012) study. Also, the extremely small number of effect sizes in each category weakens confidence in the robustness of findings. Therefore, it seems reasonable to conclude that there was no evidence that interventions’ effects on academic achievement were significantly moderated by any confounding factors.

Overall Mean Effect Sizes for Psychosocial Outcomes

Table 8 presents eight underachievement interventions’ effects on 44 psychosocial outcomes. Overall, gifted underachievers receiving interventions significantly outperformed their comparison peers on psychosocial outcomes, g = 0.22, k = 44 (effect sizes), 95% CI = [0.09, 0.34], p = .001. This observed mean effect size of .22 appeared to a reasonable estimate of the true effectiveness since there was no evidence of influence of publication bias. Interventions’ effects on psychosocial outcomes appeared to be relatively homogenous unlike the case in achievement ones, Q_b = 49.74, df = 43, p = .223, I² = 13.54. The average effect sizes by study ranged from 0.05 to 0.38.

OPEN IN VIEWER

Table 8.

Effectiveness of Underachievement Interventions on Nonachievement Outcomes: Mean ES, ESs by Study and Outcome Category.

Nonachievement outcomes		k	Hedges’s g	Variance	95% CI		p value
					Lower	Upper
Mean ES across studies		44 a	0.22	0.004	0.09	0.34	.001
Study	Intervention
Castro, 2008	Peer and mentoring support network	5	0.24	0.02	−0.01	0.50	.064
Johnson, 2002	Student, parent, and teacher support groups	3	0.94	0.10	0.33	1.56	.002
Morisano et al., 2010	Goal-setting intervention	2	0.31	0.02	0.005	0.61	.046
Niederdeppe, 2009	Theatre-arts intervention	10	0.05	0.03	−0.30	0.41	.774
Obergriesser & Stoeger, 2015	Self-regulated learning	3	0.15	0.05	−0.31	0.60	.524
Sivaraman, 2012	Goal-setting intervention	10	0.12	0.02	−0.15	0.39	.395
Stoeger & Ziegler, 2005	Self-regulated training	6	0.14	0.04	−0.24	0.52	.467
Sullivan, 2016	Goal-setting intervention	5	0.38	0.17	−0.43	1.18	.357
			Heterogeneity statistics
			Q b	df	p value	I 2
			49.74	43	.223	13.54
		k	Hedges’s g	Variance	95% CI		p value
Nonachievement outcomes					Lower	Upper
Outcome category b
Self-efficacy		8	0.19	0.03	−0.13	0.50	.246
• Aspiration level for mathematics
• Academic self-perception
• Theory of Intelligence
Goal valuation		8	0.11	0.02	−0.18	0.40	.456
• Goal setting
• Future orientation
• Planning ahead
• Anticipation of future consequences
Environmental perceptions		8	0.32	0.02	0.03	0.61	.030
• Attitudes toward teachers and/or school
Self-regulation/motivation		17	0.27	0.02	−0.02	0.56	.070
• Using cognitive strategies
• Persistence/willingness to exert effort
• Time management
• Completing in-class assignment and homework
• Managing not to lose course credit
Psychosocial functioning		3	0.16	0.06	−0.32	0.65	.504
• Anxiety
• Negative affect
• Helplessness
			Heterogeneity statistics
			Qb	df	p value	I 2
			49.74	43	.223	13.54

Note. df = degrees of freedom; CI = confidence interval; ES = effect size.

a

Forty-four was the total number of effect sizes for nonachievement outcomes extracted from eight studies reporting nonachievement outcomes in this meta-analysis. ^bA brief description of the five outcome categories is presented in the text. Table 5 lists all 44 specific nonachievement and 16 achievement outcomes.

Gifted underachievers receiving interventions scored significantly higher than their comparison peers in the studies of Johnson (2002): (g = 0.94, k = 3 effect sizes, 95% CI = [0.33, 1.56], p = .002) and Morisano et al. (2010): (g = 0.31, k = 2 effect sizes, 95% CI = [0.005, 0.61], p = .046). However, this finding seemed to be either overestimated or mostly spurious. All three psychosocial outcomes in the Johnson (2002) study were based on data reported by teachers who were also involved in implementing the intervention, thus, the risk of detection bias might have led to overestimations of the intervention’s effect. In fact, the respective effect sizes for the three psychosocial outcomes were 0.35, 0.94, and 1.49. Also, of the 44 psychosocial outcomes, only these 3 were teacher-reported, and the other 41 were self-reported by students. Last, when interpreting the effects of the goal-setting intervention in Morisano et al.’s (2010) study, it is necessary to bear in mind that this was the only study based on college students.

Mean Effect Sizes by Psychosocial Outcome Categories

Also shown in Table 8, the mean effect size was 0.19 for self-efficacy (k = 8 effect sizes, p = .246), 0.11 for goal valuation (k = 8 effect sizes, p = .456), 0.32 for environmental perceptions (k = 8 effect sizes, p = .030), 0.27 for self-regulation/motivation (k = 17 effect sizes, p = .070), and 0.16 for psychosocial functioning (k = 3 effect sizes, p = .504). On average, gifted underachievers receiving interventions outscored their comparison peers on self-reported environmental perceptions but not on the other four outcome categories, g = 0.32, k = 8 (effect sizes), 95% CI = [0.03, 0.61], p = .030. Underachievement interventions’ effects on students’ environmental perceptions appeared to relatively consistent and robust, considering that these effects were manifested across all four studies (Castro, 2008; Niederdeppe, 2009; Sivaraman, 2012; Sullivan, 2016) that assessed students’ attitudes toward teachers and school using the SAAR-S (McCoach & Siegle, 2003a). Again, interventions’ effects on psychosocial outcomes appeared to be relatively homogenous, Q_b = 49.74, df = 43, p = .223, I² = 13.54. Therefore, testing for moderators was not conducted. Table 9 provides interested readers an opportunity to view the distributions of the mean effect sizes for the subcategories of seven selected variables at face value.

OPEN IN VIEWER

Table 9.

Mean Effect Sizes for Subcategories of Key Variables: Nonachievement Outcomes.

Variable	k	Hedges’s g	Variance	95% CI		p value
				Lower	Upper
Intervention type
Counseling intervention	18	0.30	0.02	0.04	0.57	.023
Instructional intervention	26	0.19	0.01	0.03	0.35	.021
Intervention dosage
Low (90-150 minutes)	17	0.21	0.01	0.02	0.41	.033
Moderate (300-900 minutes)	21	0.16	0.01	−0.02	0.33	.074
High (1,200-1,760 minutes)	6	0.59	0.06	0.11	1.07	.016
Intervention maturity
Piloting	8	0.51	0.03	0.17	0.85	.003
Replicating	36	0.16	0.01	0.02	0.30	.028
Developmental level
Elementary school	9	0.14	0.02	−0.15	0.43	.334
Middle school	13	0.31	0.04	−0.09	0.70	.132
High school	20	0.19	0.01	0.01	0.38	.036
College	2	0.31	0.02	0.00	0.61	.046
Data source
Self-reporting	37	0.22	0.004	0.09	0.35	.001
Teacher-reporting	7	0.27	0.11	−0.37	0.91	.406
Report
Journal article	11	0.22	0.01	0.01	0.43	.038
Dissertation	33	0.27	0.01	0.10	0.45	.002
Country
Canada	2	0.31	0.02	0.00	0.61	.046
Germany	9	0.14	0.02	−0.15	0.43	.334
United States	33	0.27	0.01	0.10	0.45	.002

Note. df = degrees of freedom; CI = confidence interval.

The Interventional System: Theories of Change and Beyond

The AOM is one of the most researched and utilized theoretical frameworks on interventions for underachieving gifted students. As elaborated in the “Introduction” section, the key premise of the AOM is that students’ academic self-efficacy (e.g., believing that they can succeed), goal valuation (e.g., believing that learning is important and meaningful), and perceptions about the environment (e.g., feeling that teachers are supportive) collectively effect their motivation (including self-regulated learning) to engage, learn, and perform in school. An AOM-based intervention aims to help gifted underachievers improve their academic performance by boosting their learning motivation and engagement, which serves as a bridge to the end goal of improving school performance.

Results of the current meta-analysis support that gifted underachievers seemed to benefit significantly from interventions in terms of targeted psychosocial outcomes, as noted above, in comparison with underachieving gifted peers not receiving interventions. However, they were not better off than their comparison peers in terms of traditional academic outcome measures, such as grades in the K-12 levels. There are many possible reasons for this, but a very likely one is that substantial change in academic performance requires sustained intervention to translate improved self-beliefs into actual academic performance. Students who have been underachieving may need time to “catch up” academically. Additionally, a successful intervention may need to include a concomitant focus on academic behaviors, such as learning strategies and organizational and study habits. Finally, it is possible that there is still a psychosocial “missing link” needed to move students to actualized improved academic performance.

The Role of Student Engagement

Interventions in nine out of the 10 quantitative studies (with the exception of the study by Cortés-Cabello in 2013) and in all four qualitative inquiries reviewed in the current meta-analysis, included some types of activities for boosting student engagement. This emphasis is in line with current empirical research on the importance of student engagement for school completion and dropout prevention (Reschly & Christenson, 2012). Student engagement is a broad, encompassing construct, covering active learning across the lifespan, including involvement in learning activities in the community, at home, as well as in school. Engagement is viewed not only as a process or mechanism through which external and internal factors work together to facilitate learning but also as an outcome of learning; furthermore, it acts as a protective factor that promotes resilience, coping behaviors, and social support (Wang et al., 2019). Christenson et al. (2012) defined it as including academic (e.g., time on-task, completion of assignments, accruing necessary credit hours), behavioral (e.g., attendance, compliance, discipline referrals), affective (e.g., a sense of belonging, believing that teachers are supportive), and cognitive components (e.g., valuing school, finding learning meaningful, feeling bored in class, embracing cognitive challenges).

The four dimensions of student engagement provide an array of potential levers for intervention because the dimensions may be influenced by a wide variety of activities and/or services that are facilitated by multiple parties, including parents, teachers, peers, and others and within diverse contexts such as home, school, and social networks. The dimensions of student engagement have important implications for both identifying gifted students at risk for underachievement as well as designing interventions and understanding their efficacy (Landis & Reschly, 2013). Future research may consider how each dimension of engagement influences achievement and the types of interventions that are most impactful on each dimension. By doing so, researchers may be able to provide practitioners with suggestions regarding how to provide more targeted and effective interventions for individual gifted underachievers.

Identifying Underachievement Among the Gifted

How to define and measure underachievement in a gifted population remains one of the fundamental and challenging issues facing researchers, no doubt rooted in the lack of consensus on the definition of giftedness. For instance, McBee and Makel (2019) found that the proportion of students categorized as gifted varied from less than 0.5% to approximately 90% depending on the definition of giftedness and measures used in the identification protocol. Many underachievers had been achievers at one time, and these children are relatively easy to identify by parents and teachers. More difficult to identify are children whose level of achievement relative to their academic potential is unknown. Universal screening in the early years may be useful in identifying some of these children. However, this may not be enough to identify underachievement among students with learning disabilities or children whose potential is unseen due to lack of opportunity. When students’ ability itself is underestimated, students attaining basic levels of achievement may not be recognized as gifted underachievers who may benefit from interventions. More extensive evaluation by school psychologists and/or early preparatory programs that expand opportunities for hidden potential to be revealed (for example, see Project SPARK, Little et al., 2018, and STEM Starters+, Robinson et al., 2018) may be required to find and support such students.

It is also worth noting the obstacles to studying underachievement among gifted students to support future research designs. First, to be identified as a gifted underachiever entails, as a prerequisite, being identified as gifted. In most schools, achievement is a significant factor in being identified as gifted. This is obviously true where giftedness is identified on the basis of achievement measures and/or grades. However, it is likely also true where access to the identification process depends on teacher referral or recommendation. This presents a high risk of subject selection issues in studies of gifted underachievement, as “underachievement” is disproportionately likely to be a temporary state for students in both treatment and control groups, given that today’s “gifted underachievers” in many schools must (by definition) have been high achievers at some point in the past. Regression toward the mean in achievement measures in both intervention and comparison groups of gifted students is potentially a significant barrier to detecting effects of interventions that may be overcome by increasing the number of observations and timeframe of studies.

Future researchers should also be cognizant of potential restriction of range issues when using many common achievement measures and take into consideration issues such as limited degrees of freedom in course grades or GPA, the influence of subjectivity on grading, and test ceiling effects that may limit the usefulness of grade-level achievement test scores to reliably measure real academic growth for gifted students and subsequently detect statistically significant group differences in achievement. Above-grade–level assessments (which raise test ceilings to better distribute scores of advanced learners) as measures of achievement change may be useful, especially when it is not feasible to create very large intervention and control groups.

Broaden the Definition of Intervention Success

Our findings on student engagement also highlight the need to look beyond the current ideology and practice concerning educational outcomes of underachievement interventions. Research on the efficacy of underachievement interventions for gifted students has largely been limited to a few traditional measures of school performance, such as course grades or GPA. For example, course grades emerged to be an overwhelmingly dominant indicator of achievement outcomes for students in the K-12 levels in the 10 quantitative studies reviewed in the current meta-analysis. To be sure, it is important to include such outcomes when examining the efficacy of underachievement interventions. However, evaluating the success of interventions solely on the basis of such outcomes may result in undervaluing interventions that yield other important benefits for student well-being and long-term talent development, such as confidence in one’s ability to excel in school, heightened motivation, increased participation in school-related activities, decreased anxiety, perceptions of greater social support, evidence of future educational goals, or even more independent learning at home. These effects may be an important prelude to changes in student achievement, which may begin with small steps before they are evident as substantial improvements.

Furthermore, it is worth noting the possibilities within the AOM framework for underachievement to be caused by factors extrinsic to the student that student-focused interventions cannot address. For example, with respect to the dimension of goal valuation, an intrinsically motivated student’s perception that there is little to be gained by exertion of effort when curriculum lacks opportunities for new learning may be accurate when viewed through the lens of “learning for learning’s sake.” Similarly, underachieving students’ perceptions that their current academic environments are not supportive may sometimes be accurate in settings where their abilities are not valued or where students experience issues such as frequent bullying or harassment. Thus, potential to change achievement outcomes by changing student perceptions may be limited when actual affordances for learning and development are not available in a student’s current setting. This study does not preclude the possibility that interventions that positively influenced student self-efficacy and motivation but did not appear to positively influence traditional achievement measures could yield academic benefits if combined with changes to academic curricula or modifications of the learning context.

Merits of Mixed Methods on Understanding Gifted Underachievement

The current study suggests that mixed methods would be especially useful in investigating gifted underachievement. Research has highlighted the heterogeneous and individualized nature of underachievement among gifted students (Siegle & McCoach, 2018; Snyder, Carrig, et al., 2019; Snyder & Linnenbrink-Garcia, 2013) and fluctuations in achievement levels from one setting to another (Baker et al., 1998; Matthews & McBee, 2007). Relatedly, successful interventions might require more individualized approaches as suggested by the four qualitative studies reviewed in the current meta-analysis (for a recent review on counseling gifted students particularly using a model of evidence-based practice, see Pfeiffer & Preado, 2018). Future research should focus on studying the match between student characteristics and environments and targeted interventions.

Implications for Practice

Although the present study does not point to a transformational one-intervention-fits-all approach to addressing gifted underachievement as measured by traditional academic measures, we see promise in interventions provided on a targeted basis. Practitioners should consider the likely reality that underachievement is a symptom with many potential causes. Therefore, effective approaches likely involve first diagnosing these underlying causes, as the diagnosis of “underachievement” itself may not be specific enough to be actionable. Further, practitioners evaluating potential intervention strategies for application should remember that an underlying assumption of interventions targeting learner self-efficacy, motivation, and engagement to influence academic outcomes is that learning opportunities that would be seized by a student with an optimal mind-set for academic achievement are actually available in the student’s learning environment. High levels of student effort and engagement invested in low-level curriculum may not yield achievement gains for gifted students. If interventions yield measurable improvements in self-efficacy and attitudes toward learning but do not ultimately move the needle on achievement, it may be worth considering the possibility that there are curriculum and/or setting issues to address. Future researchers exploring the issue of gifted underachievement are encouraged to capture and report data on curriculum content and academic setting so that potential interactions between student mind-set and learning context can be better understood to guide development of more precise diagnostic tools and the design of interventions that significantly impact underachievement among gifted students.