School & District Management Opinion

Short Videos are Better for Learning, Right? Maybe Not.

By Justin Reich — March 05, 2014 4 min read
  • Save to favorites
  • Print

I got this question from a HarvardX faculty member the other day:

Do you have a repository of HarvardX-germane literature or case studies that speak to the effectiveness of full lectures versus short segments (possibly interspersed with Q&A), ideally that predates MOOCs? Is there any other canonical literature you’d recommend we circulate within our team in the interests of bringing folks up to speed on existing research and results along those lines?

This was an interesting question. We all know that shorter videos are better for learning. Right? Right?

Much of the research on lecture-style teaching and lecture length (some of which predates video lectures) has been driven by pyschologists interested in issues of attention. These are studies that look for evidence of when student attention is maintained or declined during a lecture performance. Attention is a great thing to study. It is almost assuredly one of the correlates of learning outcomes (hard to learn if your mind is elsewhere), but attention is a necessary but not sufficient condition for learning.

So on the terms of this research own terms, as an evaluation of attention span, some researchers argue that the findings are inconclusive. In a 2007 review of the literature, two pyschologists argued that while there were widespread accounts in the literature that lectures shouldn’t be more than 10 or 15 minutes, they couldn’t actually find robust evidence of a particular attention cut-off at that time.

More importantly, few of the studies they cite make the connection between lecture attention, lecture length, and learning outcomes, and in those that did, the results were largely null. So at least in this one literature review, there is little evidence to support short lectures from an attention perspective and less from a learning outcome perspective.

So while this notion of the optimal length of lecture as 5-15 minutes has a lot of traction, it perhaps is more received wisdom than well supported in the literature. And from other perspectives, these “short” videos are terribly long!

For instance, Richard Mayer is a multimedia learning researcher and pyschologist who argues that brains have limited pathways for taking in information and multimedia can overwhlem those channels quickly. He and colleagues conducted a study where information presented in a multimedia animation was broken down into 16 segements that were 8 to 10 seconds long. Students clicked continue to advance each section. Perhaps the real problem with 5 minute lectures is that they are still 30 times too long! In the specific context that Mayer looked at in this study, these segements produced better learning outcomes on a post-test than watching all of the segments at once. Of course, whether students would click 8 seconds at a time through hours and hours of course material is an open question.

As new data comes out from platforms like edX, I’d encourage folks to be careful about making the leap between studies of video watching that look at engagement and those that actually look at learning. For instance, Philip Gao got early access to edX data from multiple institutions, and he has a study coming out advocating for videos of less than six minutes.

He comes to this conclusion by sorting videos into five bins of length ( 0-3 minutes, 3-6, 6-9, 9-12, and 12-40 minutes) and then arguing that viewing completion rates drop as videos get longer. This is a useful early entry into making sense of findings from edX, but it’s not clear that these findings lead clearly to the six minute video recommendation. It’s not clear why grouping into three minute bins is useful, as opposed to developing a continuous measure. Given that the normalized engagement time of 6-9 minute videos is about the same as 3-6 minute videos, it’s not clear why the recommendation is six minutes. The study doesn’t appear to account for the purposes or function of particular videos. For instance, most of the longer videos in one of the studied courses, PH207x: Health in Numbers, are from guest lecturers whose material is more supplementary to the course than essential to the mid-terms and final exams. It’s possible that if those videos were shorter, they’d still be watched less.

And ultimately, the study uses engagement as the outcome rather than learning. If the goal is to get people to watch videos to completion, then it stands to reason that shorter videos will be conducive to that end. But it is one more step to argue that these shorter videos will lead to better learning outcomes. After all, we could get completion rates very, very high on a per video basis if we broke the videos down into 8 second segements!

There are compelling examples of long form video lectures that also have high levels of engagement. Michael Sandel’s hour long lectures (with Socratic discussion) on Justice have been watched millions of times on YouTube (though I don’t know how often they are watched to completion). Would the series be as popular if it had been broken into 6 minute segements? Do we have any way of knowing whether or not that would have improved the learning experience of viewers?

We have much to learn about how people experience online learning through different forms of media. As big as our big data is, it remains observational data. We’ll need more deliberate experimentation if we hope to understand how features of course design like lecture length impact both engagement and the learning outcomes that we care most about (at least I hope we care most about!).

For regular updates, follow me on Twitter at @bjfr and for my papers, presentations and so forth, visit EdTechResearcher.

Related Tags:

The opinions expressed in EdTech Researcher are strictly those of the author(s) and do not reflect the opinions or endorsement of Editorial Projects in Education, or any of its publications.