Date of Award

Summer 8-21-2015

Document Type


Degree Name

Doctor of Philosophy in Education (PhD)



First Advisor/Committee Member

Arthur K. Ellis

Second Advisor/Committee Member

William Nagy

Third Advisor/Committee Member

John B. Bond


meta-analysis, problem-based learning, project-based learning, middle school, junior high school, high school


Researchers and proponents of problem- and project-based learning (PBL) indicate that PBL as a curriculum and instruction approach (Savery, 2006; Schmidt, Loyens, Van Gog, & Paas, 2007) provides an effective way for teachers to respond to students’ needs, provides opportunities for students to actively engage in and take responsibility for learning by engaging in meaningful and relevant work, and provides students opportunities to directly apply their knowledge and skills (Hmelo-Silver & DeSimone, 2013; McCombs, 2010; Parker et al., 2011). Although primary research within secondary (6-12) contexts indicated that problem-and project based learning (PBL) is often superior to traditional, lecture-based instruction (Mergendoller, Maxwell, & Bellisimo, 2006; Wirkala & Kuhn, 2011) and meta-analyses at the post-secondary level indicated that PBL is at par with or superior to traditional, lecture-based instruction (Dochy, Segers, Van den Bossche, & Gijbels, 2003; Vernon & Blake, 1993; Walker & Leary, 2009), a synthesized and quantified exploration of the strength of relationship between PBL and academic achievement within middle high school student populations (Grades 6-12) was needed. The results in this meta-analysis indicate that overall, PBL students outperformed traditionally instructed students, g = 0.54, on content and skills exams across academic subject types and grade levels. Analysis of the funnel plot suggests publication bias; however, an adjustment of the mean effect using Duval and Tweedie’s (2000) Trim and Fill rendered a similar summary effect of g = 0.50. Although the mean summary effect is relatively robust, effect sizes varied depending on subject area and specific types of outcome measures. The test of homogeneity indicated that 90.49% of the variance between studies was unexplained. An insufficient number of studies rendered meta-regression unfeasible, hindering exploration of possible explanations for this variance.

Copyright Status

Additional Rights Information

Copyright held by author.


Copyright Status