Leading Edge

Commentary

Blended Learning Improves Science Education

Brent R. Stockwell,1,2,3,* Melissa S. Stockwell,4,5,6 Michael Cennamo,7 and Elise Jiang1
1Howard Hughes Medical Institute, Department of Biological Sciences, Columbia University, Northwest Corner Building, MC 4846, 550 West

120th Street, New York, NY 10027, USA

2Department of Chemistry, Columbia University, New York, NY 10027, USA
3Department of Systems Biology, Columbia University, New York, NY 10032, USA
4Department of Pediatrics, Columbia University, New York, NY 10032, USA
5Department of Population and Family Health, Mailman School of Public Health, Columbia University, New York, NY 10032, USA
6NewYork-Presbyterian Hospital, New York, NY 10032, USA
7Columbia Center for New Media Teaching and Learning, Columbia University, New York, NY 10027, USA

*Correspondence: bstockwell@columbia.edu
http://dx.doi.org/10.1016/j.cell.2015.08.009

Blended learning is an emerging paradigm for science education but has not been rigorously as-
sessed. We performed a randomized controlled trial of blended learning. We found that in-class
problem solving improved exam performance, and video assignments increased attendance and
satisfaction. This validates a new model for science communication and education.

Blended Learning Is an Emerging traditional lecture-and-textbook-based (Hake, 1998). Finally, low-stakes forma-
Instructional Pedagogy approach to teaching and learning (Glazer, tive assessments improved exam perfor-
At the undergraduate level, science is 2012). First, pre-class online video assign- mance when compared across students
most commonly taught using a lecture ments may offer advantages over textbook in different class sections with different
and textbook format. In this pedagogical assignments, especially for introductory teachers (Roediger et al., 2011). Thus,
approach, students are assigned a text- science courses with complex and dense active, problem-based learning may im-
book to read at home before class and material that is unfamiliar to beginning prove student performance on exams
listen to an instructor lecture on the as- students (Kagohara, 2010). Reading a (Che et al., 1998; Knight and Wood, 2005).
signed material during class. Although textbook engages mainly visual, language We wondered if pre-class video assign-
this approach to learning is effective for comprehension, and cognitive neural path- ments and in-class problem solving
some students, it is suboptimal for many ways and requires that the reader is able would result in increased engagement,
undergraduate students and may con- to select the most relevant material for satisfaction and more effective learning.
tribute to students leaving the sciences application to the course (Wandell, 2011). While it has been suggested that these
(Handelsman, 2007). Video instruction, on the other hand, adds techniques in some contexts and in isola-
Recognition of the limitations of the auditory engagement to visual, language tion improve learning (Fitzgerald and Li,
traditional instructional approach to sci- comprehension and cognitive processes, 2015; Freeman et al., 2014; Glazer, 2012;
ence using a textbook and lecture for- and allows for more varied emphasis of Handelsman, 2007), studies involving
mat has led to suggestions for examining the importance of content. Video assign- undergraduate science education have
alternative methodologies (Handelsman, ments are typically more engaging for a generally not been conducted as random-
2007). In recent years, online learning large introductory science course and ized controlled trials under otherwise iden-
options, including massive open online may stimulate greater engagement with tical conditions with the same material and
courses (MOOCs), have become increas- the course material. the same instructor. Rather, they have
ingly available as a means to produce In addition, listening to a lecture in class been conducted using different instructors
learning in students who cannot attend involves mostly recording and recalling or different course materials, making it
classes in person and have been proposed information, which are lower levels of difficult to assess if changes seen were
as alternative learning paradigms (Reich, Bloom’s taxonomy of learning (Bloom, due specifically to the teaching methods.
2015). However, the low completion 1956). In contrast, solving problems in We endeavored to rigorously examine
rates of online courses, such as MOOCs, real time during class forces students to whether video-based preparation and in-
and the importance of instructor-student synthesize and apply knowledge as they class problem solving increase student
and student-student interactions in class- process it (Amador et al., 2006). A histori- engagement, as indicated by class atten-
rooms have suggested that online learning cal comparison has shown that a struc- dance, and ultimately exam performance,
alone is unlikely to be the most effective tured course with in class problem solving for undergraduate science students.
strategy for teaching and learning (Glazer, improves performance and reduces the
2012; Reich, 2015). achievement gap (Haak et al., 2011). A Randomized Controlled Trial to
Nonetheless, the online learning para- Moreover, a survey of pre/post-test data Evaluate Blended Learning
digm offers valuable tools that could in physics courses revealed that interac- A total of 172 students enrolled in Bio-
supplement or replace aspects of the tive teaching improved student learning chemistry I: Structure and Metabolism,

Cell 162, August 27, 2015 ª2015 Elsevier Inc. 933

 In advance of class, students were dents with the preparation material. There
provided either a link to a video or a link was significant improvement in satisfac-
to a textbook reading, covering the tion with the preparation material among
same material, as per their randomization the students who received the video
assignment; students were instructed not assignment (4.3/5.0) compared to stu-
to review other materials and had little dents who received a textbook assign-
incentive to do so, since the study results ment (2.9/5.0) (p < 0.0001, Mann-Whitney
were not included in their grade, as partic- test) (Figure 2). This is consistent with the
ipation was voluntary, and since there fact that fewer students who received the
was no assigned textbook for the course textbook assignment chose to attend
for them to otherwise consult. During class and suggests that textbook prepa-
class, students either listened to an ration for a science class is less satisfac-
instructor-delivered lecture or listened to tory and engaging for students compared
the same lecture material interspersed to assignments that involve watching a
with instructions to solve problems re- video.
lated to the presented material, again as At the end of each class, students took
per their randomization arm. To ensure a 20 min online, multiple-choice exam,
that the content was the same in each while still present in class, to test their un-
class, the students who attended the lec- derstanding of the material. First, in order
ture were also provided with the same to test whether the exam measured the
problems and their solutions, but these same performance characteristics found
were explained by the instructor rather in the rest of the course, we compared
than being solved by the students. Thus, exam performance for students with a
the difference between the lecture and higher prior exam performance versus a
problem-solving class formats was not in lower one. Indeed, we found that prior
the content but rather whether the stu- exam performance (high versus low)
dents actively solved the presented prob- correlated with the study exam perfor-
lems or were simply told the answers as mance: the median exam scores for stu-
part of the lecture. All lectures were pro- dents in the high prior exam group (73/
vided by the same instructor. 100) was significantly (p = 0.006, Mann
The first outcome measured was class Whitney test) greater than the median
attendance. Since participation was vol- exam score for students in the low prior
Figure 1. Blended Learning Combines untary, students were able to discontinue exam group (60/100), suggesting that
Different Means of Content Delivery participation at any point in the study. the students exerted their typical effort
We analyzed whether the rate of class and exhibited similar performance char-
an undergraduate biochemistry course attendance after receiving the preparation acteristics, despite the fact that participa-
at Columbia University, were invited to material was different between the stu- tion was voluntary.
participate in a randomized controlled trial dents who received a video versus text- We then examined the performance of
in the fall of 2014. A total of 111 students book pre-class assignment. Indeed, we the four study arms on the end-of-class
enrolled in the study. We used a two-by- found that more students randomized to exam to determine the effects of teaching
two study design, in which we compared the pre-class video assignment attended pedagogy on student performance. We
the effects of both video versus textbook class (84%, 47/56) compared with those found that the median exam score
pre-class assignments and lecturing with randomized to the textbook assignment increased within each arm, from arm 1
instructor-demonstrated problems versus (67%, 37/55) (p = 0.04, Pearson’s chi- (textbook preparation, in-class lecture;
lecturing with student problem-solving in square test) (Figure 2). In other words, 61/100) to arm 2 (video preparation, in-
class. Students were randomized to one twice as many students chose not to class lecture; 67/100) to arm 3 (textbook
of four arms: (1) textbook preparation for attend class after receiving a textbook preparation, in-class problem solving;
lecture, (2) video preparation for lecture, assignment (18/55; 32.7%) compared to 73/100) to arm 4 (video preparation, in-
(3) textbook preparation for problem- those who received a video assignment class problem solving; 80/100) (Figure 2).
solving class, or (4) video preparation for (9/56; 16.1%). This result was consistent Indeed, the students in the arm that ex-
problem-solving class (Figure 2). The stu- with the hypothesis that a video is a perienced fully blended learning, with
dents were stratified for randomization by more engaging way to present new video preparation and in class problem
gender and prior exam performance (low: and complex material to students and solving, had the highest median exam
lower third versus high: upper two-thirds) stimulates students to be interested score (80/100).
into each of the four arms to ensure equal in learning more about the topic by We found that the most significant
representation of these students in each attending class. intervention was the implementation of
study arm; 54% of students in the study To test this hypothesis, we examined student-centered problem solving during
(60/111) were male (Figure 2). the level of satisfaction among the stu- class, as the median score on the exam

934 Cell 162, August 27, 2015 ª2015 Elsevier Inc.

Figure 2. Randomized Controlled Trial to Evaluate Teaching and Learning Strategies for Undergraduate Biochemistry
(A) The study followed the indicated scheme, in which students were invited to participate voluntarily in a study and asked to indicate their informed consent. They
were randomized into one of four groups, receiving either a textbook or video assignment before class and either a lecture or a lecture with problem solving during
class. The impact on attendance and exam score was then measured.
(B) Equivalent numbers of male and female students and students with high and low prior exam scores were assigned to each group.
(C) Students who received a video assignment were more likely to attend class than students who received a textbook assignment. The number of students who
attended or did not attend class after receiving a video or textbook assignment is indicated.
(D) Students who solved problems in class performed better than students who only listened to a lecture. The median score of each group of students on the end
of class exam is indicated (out of 100 possible points). The comparison of scores is shown for the students in the lecture and problem-solving groups. Video
assignments were more satisfying to students than textbook assignments; students were asked to rate how satisfied they were with their pre-class assignment on
a scale of 1 (low satisfaction) to 5 (high satisfaction).

was higher for those who were random- quality and that the difference in exam increase the diversity of individuals who
ized to the lecture-plus-problem-solving performance was not due to unconscious choose science careers and to broaden
class (74/100) compared to those ran- instructor bias in the delivery of the lec- the pool of scientifically literate citizens
domized to the lecture-only class, irre- ture material. Moreover, it suggests that globally. The traditional textbook-and-
spective of preparation material (63/100) student satisfaction alone is not a re- lecture approach to teaching under-
(p = 0.03, Mann Whitney test) (Figure 2). liable indicator of learning gains and graduate science, while effective for a
We also compared the exam scores of that institutions and instructors should subset of individuals, is not the most
students in the textbook versus video be cautious in evaluating pedagogical effective means of stimulating learning in
preparation groups but found no statisti- methods using only student satisfaction the broadest group of students. We and
cally significant difference in this relatively measures, which are the typical metrics others have sought to examine the effec-
modest sample size, despite the trend used in course evaluations. Indeed, tiveness of an alternative pedagogy that
toward higher scores in the group that objective measures of learning gains replaces textbooks with video assign-
received the video assignment. may ultimately be the most accurate ments and traditional lecturing with
We recognized that the instructor in a means of assessing courses. active, student-centered problem solving.
randomized controlled trial may exhibit Overall, we conclude that providing stu- One result of this trial was that students
unconscious bias. We sought to measure dents with problems to solve during class who received a video assignment were
this objectively by evaluating student results in significantly improved exam both more likely to attend class and to
satisfaction with the lecture and prob- performance, compared to simply having rate their assignment as providing a
lem-solving classes. Student satisfaction the instructor describe the same prob- higher degree of satisfaction, compared
did not differ significantly between stu- lems and their solutions during the course to students who received a textbook
dents who attended the problem-solving of the lecture. assignment. Thus, in this study, video as-
versus lecture classes (4.1 versus 4.1, signments were more effective at stimu-
out of 5.0, p = 0.99, Mann-Whitney test) Implications for Science Education lating student interest and engagement
(Figure 2). This suggests that the tradi- In recent years, there has been a growing with the course material. Therefore,
tional lecture was engaging and of high interest in improving science education to providing supplemental videos for science

Cell 162, August 27, 2015 ª2015 Elsevier Inc. 935

courses, either in place of, or in addition to, the same content, the same instructor, way that they can then translate to
textbook assignments, may enhance stu- the same point of time, and the same novel situations they will encounter in
dent engagement and motivation. pool of students randomized to different their academic and professional ca-
A second finding was that students who conditions, we were able to minimize the reers, which is the hallmark of effective
listened to a traditional instructor-focused impact of variables beyond the ones we learning. Institutions and instructors may
lecture underperformed students who sought to measure. In addition, the trial wish to consider how to support blended
actively solved problems during the class was conducted in a pragmatic fashion learning paradigms in their science
period. This was despite the fact that the with students within their usual course curricula.
students in these two groups rated their with their usual peers and instructor and
classes as equally satisfying. Thus, while is therefore reflective of results in a real- ACKNOWLEDGMENTS
students enjoyed passively listening to a world setting. It also demonstrates the
lecture as much as being asked to solve feasibility of conducting such pragmatic We thank the Columbia Center for New Media
Technology and Learning (CCNMTL) for assis-
problems during class, they learned the education trials.
tance in preparing and disseminating videos to
material better when they actively worked Unfortunately, we were only able to students, Daniel Rabinowitz for advice regarding
on problems, rather than simply being examine the effects of teaching style in statistical analyses, and the students who partici-
given the problems and answers as part one class. Additionally, we did not pated in this study. Brent R. Stockwell is an Early
of the lecture. Therefore, instructors may examine the long-term effect of these in- Career Scientist of the Howard Hughes Medical
wish to provide frequent opportunities terventions on learning; it would be valu- Institute. Financial support for this study was
provided by Columbia University’s Office of the
during class for students to apply the con- able to have students take another exam
Provost. The authors do not have any financial
tent presented to specific problems. This several months or even years after partici- conflicts of interest to report.
provides ongoing formative assessments pating in the study to examine how well
for students to test their learning and refo- they retain the information. We propose REFERENCES
cuses student attention during an other- that such longitudinal assessments of
wise extended period of lecturing. In addi- learning gains could be organized at the Amador, J.A., Miles, L., and Peters, C.B. (2006).
tion, asking a student to apply knowledge institutional level by regularly adminis- The Practice of Problem-Based Learning: A Guide
to Implementing PBL in the College Classroom
to new contexts may assist them in tering follow-up exams to students in
(Anker Pub. Co.).
learning the material in a more effective specific majors who have completed the
Bloom B.S., ed. (1956). Taxonomy of Educational
way that allows them to extrapolate the same required courses early in their
Objectives. Handbook 1: Cognitive Domain (David
knowledge to new situations in the future; program. Perhaps by coupling such McKay Company, Inc.).
a related case-based pedagogy has been programs with a compensation system Che, S., Wu, W., Nelman-Gonzalez, M., Stuken-
used successfully in business, medical, for participation, the long-term impact berg, T., Clark, R., and Kuang, J. (1998). FEBS
and law schools. of course structures and pedagogical Lett. 424, 225–233.
One of the difficulties faced when trying methods could be assessed. Drits-Esser, D., Bass, K.M., and Stark, L.A. (2014).
to rigorously assess the effectiveness of Overall, the results from this trial sug- CBE Life Sci. Educ. 13, 593–601.
teaching and learning strategies is the gest that a blended teaching approach, Fitzgerald, N., and Li, L. (2015). J. Chem. Educ.
challenge of delivering identical content which uses video assignments in advance Published online March 6, 2015. http://dx.doi.org/
via the same instructor and same pool of of each class to stimulate interest in 10.1021/ed500667.
students such that it is the teaching the topic and provide foundational Freeman, S., Eddy, S.L., McDonough, M., Smith,
method that is being assessed and not knowledge, coupled with lectures having M.K., Okoroafor, N., Jordt, H., and Wenderoth,
M.P. (2014). Proc. Natl. Acad. Sci. USA 111,
other factors. A variety of strategies have in-class problem solving, is a more effec-
8410–8415.
attempted with limited success to deal tive strategy for science education
Glazer, F.S. (2012). Blended Learning: Across
with this challenge, such as historical compared with traditional approaches. It
the Disciplines, across the Academy (Stylus
comparisons among the students who is worth noting that video assignments Publishing).
take a particular class that has altered its did not improve student exam perfor-
Haak, D.C., HilleRisLambers, J., Pitre, E., and
methodology, comparisons among in- mance on their own but did increase Freeman, S. (2011). Science 332, 1213–1216.
structors who use different teaching attendance and satisfaction. By providing Hake, R.R. (1998). Am. J. Phys. 66, 64–74.
methodologies, or comparisons among foundational information pre-class, video
Handelsman, J. (2007). Scientific Teaching (W.H.
the same instructor using different teach- assignments can also create time in Freeman and Co.).
ing methodologies within one course but class for active learning, such as student Kagohara, D.M. (2010). Dev. Neurorehabil. 13,
for different topics and at different points problem solving. Thus, video pre-class 129–140.
of time in the semester. We suggest that assignments can serve a critical role in Knight, J.K., and Wood, W.B. (2005). Cell Biol.
it is most effective to use the rigor and the blended learning paradigm for edu- Educ. 4, 298–310.
methodology of randomized controlled cation. Reich, J. (2015). Science 347, 34–35.
trials, which are able to measure the These results also illustrate the feasi- Roediger, H.L., Agarwal, P.K., McDaniel, M.A., and
impact of patient interventions in health bility of using the clinical trial methodol- McDermott, K.B. (2011). J. Exp. Psychol. Appl. 17,
and biomedicine, to examine the impact ogy in educational intervention evalua- 382–395.
of teaching approaches on student tions. Blended learning approaches may Wandell, B.A. (2011). Ann. N Y Acad. Sci. 1224,
learning (Drits-Esser et al., 2014). By using help students learn information in a 63–80.

936 Cell 162, August 27, 2015 ª2015 Elsevier Inc.