Active learning changes classroom time from predominantly listening to answering, discussing, explaining, practicing, or applying ideas. The available evidence spans undergraduate STEM, statistics and biostatistics courses, and elementary physically active lessons, so the results should be compared with care: these populations and implementations are not interchangeable.
Contents
- Overall STEM performance and failure outcomes
- Achievement gaps and equity outcomes
- Student perceptions and cognitive effort
- Peer instruction and collaborative learning
- Statistics and biostatistics classroom experiments
- Physically active lessons, attention, and retention
Overall STEM performance and failure outcomes
A major undergraduate STEM meta-analysis combined 158 studies published through 2010. Across those studies, active learning increased examination or concept-inventory performance by 0.47 standard deviations compared with traditional lecturing. Examination-score gains were statistically significant in every STEM discipline represented by more than 10 experiments. Gains were also statistically significant in small classes of 50 or fewer students, medium classes of 51–110, and large classes above 110.
The largest performance effects occurred in undergraduate STEM classes with 50 or fewer students. However, the analysis found 0 statistically significant differences in active-learning effect sizes between courses for majors and nonmajors or between introductory and upper-division courses. That suggests the measured advantage was not confined to one course level or enrollment group in this evidence base. Source: Active learning increases student performance in science, engineering, and mathematics, studies published through 2010.
Failure outcomes showed a similarly important pattern. Among 67 undergraduate STEM studies with failure data, the odds of failing were 1.95 times higher under traditional lecturing than under active learning. Average failure was 21.8% in active-learning courses and 33.8% in traditional-lecture courses. The meta-analysis described the traditional-lecture failure rate as 55% higher and the traditional-lecture risk of failing as 1.5 times the active-learning risk.
| Outcome in undergraduate STEM studies | Active learning | Traditional lecturing |
|---|---|---|
| Average failure rate | 21.8% | 33.8% |
| Relative failure comparison | Reference | 1.95 times higher odds of failure |
| Performance comparison | Reference | 0.47 standard deviations lower on average |
These are pooled findings from studies published through 2010, not a guarantee for every active-learning lesson. The source measured examination or concept-inventory performance and reported course-level failure outcomes; it did not establish that every activity or teaching method produces the same result.
Achievement gaps and equity outcomes
A later equity-gap meta-analysis examined studies published through 2019, with the analysis published in 2020. Its examination-score evidence included 9,238 students from 15 studies. Its course-passing evidence included 44,606 students from 26 studies. In this analysis, “minoritized” and “non-minoritized” are the source’s comparison groups.
Active learning reduced the average STEM examination-score gap between minoritized and non-minoritized students by 33%. The average gap was −0.62 standard deviations under traditional lecturing and −0.42 standard deviations with active learning. For course passing, active learning reduced the average gap by 45%: the average passing-rate gap was 7.1 percentage points under traditional lecturing and 3.9 percentage points with active learning.
The study-level pattern was not universal, but it was substantial. Ten of 15 examination-score studies found disproportionate active-learning gains for minoritized students. Fifteen of 26 passing-rate studies found a disproportionate reduction in failure rates for minoritized students under active learning.
Intensity also mattered in the reported comparisons. With low-intensity active learning, the examination-score gap averaged −0.48 standard deviations; with high-intensity active learning, it averaged −0.36. The corresponding passing-rate gaps averaged −9.6 percentage points with low-intensity active learning and −2.0 points with high-intensity active learning.
Source: Active learning narrows achievement gaps for underrepresented students in undergraduate science, technology, engineering, and math, studies published through 2019; analysis published 2020. These results describe pooled STEM evidence and should not be treated as an identical effect in every institution or student population.
Student perceptions and cognitive effort
Active learning can improve measured learning while feeling less comfortable or less fluent to students. A randomized crossover study of 149 introductory physics students used course experiments in spring and fall 2018 and was published in 2019. Students in active classes scored 0.46 standard deviations higher on tests of learning than students in passive-lecture classes. At the same time, active-class students reported feeling 0.56 standard deviations less learning than passive-lecture students.
The study’s tests contained 12 multiple-choice questions for statics and 12 for fluids. Students rated the passive physics class as involving them as listeners at a mean of 3.9 on a 5-point scale, compared with 2.9 for the active class. The difference illustrates why perceived ease or listening time is not the same measure as demonstrated learning.
Interviews helped explain the uncomfortable perception. In follow-up interviews, 15 of 17 physics students described active instruction as disjointed or lacking flow. Fourteen of 17 cited interruptions between group activities and instructor feedback as a concern, and 14 of 17 reported frustration or confusion about the active classes. Yet 14 of 17 said that seeing the study results would change how they studied.
A semester-long intervention also reported a shift in attitudes. After a 20-minute explanation of active learning and learning perceptions, more than 65% of students reported improved feelings about active learning, and 75% said the intervention helped them feel more favorable toward it. Source: Measuring actual learning versus feeling of learning in response to being actively engaged in the classroom, spring and fall 2018 course experiments; published 2019.
Peer instruction and collaborative learning
Peer instruction gives students a structured opportunity to answer a question, discuss reasoning, and reconsider an answer. In a chiropractic immunology and endocrinology study conducted from April 2013 through March 2014, 102 students received peer instruction and 96 received traditional lectures. The peer-instruction course used 42 active quiz activities across 20 lectures.
After peer discussion, 97.5% of changed answers in the 102-student peer-instruction group were corrected answers. Mean comprehension-quiz scores rose 30.9 points after peer instruction, compared with a 17.9-point rise for recall quizzes. The peer-instruction group’s cumulative final-exam score was 3.7% higher than the traditional-lecture group’s score, and peer instruction explained 4.6% of cumulative final-exam performance.
The advantage was larger for comprehension than recall. The peer-instruction advantage was 8.4 percentage points on comprehension questions and 4.7 points on recall questions. The final exam contained 60 questions: 47 recall questions and 13 comprehension questions. Prior academic preparation was similar but not identical: the peer-instruction group had a mean prior GPA of 3.29 versus 3.23 in the control group.
Source: Brief, cooperative peer-instruction sessions during lectures enhance student recall and comprehension, April 2013–March 2014. The findings come from this chiropractic course and its particular quiz and exam design.
Statistics and biostatistics classroom experiments
Evidence from statistics education is more mixed than the broad STEM meta-analysis. A randomized introductory-biostatistics trial in a 2006 course assigned 69 students to eight cooperative-learning sessions, 97 to internet-learning sessions, and 96 to no intervention. In intent-to-treat analysis, there were 0 statistically significant examination-score differences among the cooperative-learning, internet-learning, and control groups.
Among participating biostatistics students, each cooperative-learning session was associated with an adjusted 1.7-point examination improvement, with a 95% confidence interval from −1.8 to 5.2. Each internet-learning session was associated with an adjusted 2.1-point improvement, with a 95% confidence interval from −1.4 to 5.5. After four examinations, four cooperative-learning sessions were associated with an adjusted 5.3-point improvement per examination, with a 95% confidence interval from 0.4 to 10.3. Four internet-learning sessions were associated with an adjusted 8.1-point improvement per examination, with a 95% confidence interval from 3.0 to 13.2.
A randomized first-year statistics-class trial reported in 2022 tested real-time interactive polling. After covariate adjustment, polling increased tutorial attendance by about 22%; the unadjusted estimate was about 24%. The estimated treatment effect on final marks was 0 within statistical resolution, and the estimated treatment effect on the probability of failing was also 0 within statistical resolution.
| Statistics-education result | Finding |
|---|---|
| Cooperative-learning sessions in 2006 biostatistics trial | 1.7-point adjusted improvement per session; 95% CI −1.8 to 5.2 |
| Internet-learning sessions in same trial | 2.1-point adjusted improvement per session; 95% CI −1.4 to 5.5 |
| Interactive polling and tutorial attendance | About 22% adjusted increase; about 24% unadjusted |
| Interactive polling and final marks | 0 within statistical resolution |
Source: METHODS OF LEARNING IN STATISTICAL EDUCATION: A RANDOMIZED TRIAL OF PUBLIC HEALTH GRADUATE STUDENTS, 2006 introductory biostatistics course; and ACTIVE LEARNING AND ACADEMIC PERFORMANCE: THE CASE OF REAL-TIME INTERACTIVE STUDENT POLLING, study period reported 2022. The course experiments measured different outcomes, so attendance gains should not be read as final-mark gains.
Physically active lessons, attention, and retention
Physical movement is one form of active academic lesson, but it belongs to a different age group and implementation from undergraduate STEM active learning. A cluster-randomized fourth-grade study included 2,716 children from 28 Central Texas elementary schools. It assigned 10 schools to active math lessons, 9 to active spelling lessons, and 9 to sedentary academic lessons. The study was conducted before publication in 2018.
Time on task exceeded 85% after physically active academic lessons. Economically disadvantaged children represented 34% of the sample. The sample was 46% White, 32% Hispanic, and 9% African American. Class-level differences explained 14% of the variance in time on task after active academic lessons, while school-level differences explained 9%.
Race, gender, and socioeconomic status produced 0 statistically significant moderation effects on the active-learning effect on fourth-grade time on task. The study also found 2 opposite step-dose associations: more steps predicted higher time on task in active lessons but lower time on task in control lessons.
Source: Active learning improves on-task behaviors in 4th grade children, study conducted before publication in 2018. These attention findings concern fourth-grade children in Central Texas and should not be generalized directly to university statistics courses.