This is a summary of the key points I made in the webinar on ‘What do we know about effective educational videos’ I delivered on 27 Feb as part of the iSpring series. It is also a record of a work in progress on investigating the current landscape of use of video in education.
Introduction
We are living in the middle of a video revolution. From being a curiosity to a nice-to-have, video has become ubiquitous as a mode of learning. Learners find themselves in a rich video environment, a complex ecosystem of video platforms, creators and knowledge producers. When I ask audiences at talks what was the last time they learned something from watching a video, they always have a response. These range from learning how to use a piece of software to learning how to fix a washing machine or cook a recipe. Video learning resources are not just readily available, they are being used. This is backed up by the 2018 Pew Research report showing that 87% of its users rate YouTube as an important source of learning how to do new things. Education YouTubers with millions of subscribers and tens of millions of views are not rare. There are also an increasing number of paid and free platforms that centre exclusively around video learning.
Yet, when we ask the question “what makes an effective instructional video” we get a cacophony of answers with very little agreement. What research evidence there is does not always match what actual learners are drawn to in practice and what content creators want to make. This complaint is not new, in 2006, Jack Koumi, who was involved in video learning content production at the Open University going back to the 1970s, compared the recommendations derived from research to practical decisions made by education video creators. This showed that the relevance of research was limited because its recommendations were at a high level of generality.
What are our sources of evidence?
What then, are the sources of evidence to which we could turn to for information about effective videos? I would suggest that we need to combine four sources which may or may not always align:
- Experimental laboratory research that contrasts the impact of various closely controlled aspects of a video and investigates their impact on short-term learning.
- Observational research that surveys real life behaviour or longer-term impact of embedding video into the learning process.
- Information about the video ecosystem which tells us what users will spend their time and money to watch.
- Recommendations by video creators and reports by video users (including ourselves).
The bulk of experimental research into the effectiveness of videos works has focused on investigating the effects of cognitive load viewers deal with while watching videos. This work is associated with the work of Richard Mayer and is not focused exclusively on video but on multimedia learning in general. Its recommendations were summarised in a book by Clark and Mayer (2016) under the heading of 12 principles of multimedia learning. However, this 5-point summary by an NIH report (Brame 2016), gives us a good picture of their overall thrust:
- Keep videos brief and targeted on learning goals.
- Use audio and visual elements to convey appropriate parts of an explanation; consider how to make these elements complementary rather than redundant.
- Use signaling to highlight important ideas or concepts.
- Use a conversational, enthusiastic style to enhance engagement.
- Embed videos in a context of active learning by using guiding questions, interactive elements, or associated homework assignments.
Unfortunately, this research has significant limitations acknowledged by Mayer. 1. It has been done almost exclusively on STEM content and 2. To investigate effects of learning, all this research focuses on learning by complete novices. In fact, there is some evidence that many of these principles lose much of their relevance when applied to more advanced learners. As an illustration, people who want to learn more about these principles can watch a video recording of Mayer himself (Mayer 2014) which breaks all of these recommendation (except 4) but can still be a very effective way of learning about them.
The other strand of research relies on observational data and much of it has been conducted on analysing datasets from MOOCs. It mostly confirms the principles above but also shows that learners use video in a variety of ways that cannot be captured in the lab. Most recently a study looked at whether learners use computers or tablets to view videos and found great range of uses.
The great limitation of these studies is that they often lag behind developments in the quickly evolving video learning landscape. For example, a study (Kizilcec, Bailenson, and Gomez 2015) that looked at the effect having the presenter’s face in the video, used videos with the face in a square in the corner of the video which is much less common. This applies to the experimental research, as well. If we look at the illustrations in Clark and Mayer, they are all obviously significantly behind what a viewer of a modern video will see.
What does the video ecosystem look like?
What then does the video ecosystem look like? Here, we obviously need to start with YouTube which houses countless sources of instructional videos. These videos are presented in a whole range of styles and genres from simple screencasts to elaborate animations on the one hand and from recordings of live lectures to mini-video essays on the other. What these videos have in common is that they are personal. They are made by somebody who wants to share knowledge with the rest of the world. They may or may not be tightly scripted but they always have a personal message.
But there’s also a language of video that has developed on YouTube which we might call the vlogger style. Apart from personality which corresponds to the research-derived principles, this style is characterised by humour, quick transitions and jump cuts, and visuals that interact with what is being said. All of these characteristics go against the research recommendations, yet, we see millions of people drawn to them. A good illustration of this language is the popular Crash Course channel. Many aspects of YouTube videos are evolving and changing as YouTube creators learn new skills and acquire production resources. We can see this evolution when we compare early and recent videos on channels like Marginal Revolution University and Up and Atom.
Another principle most YouTube videos break is video length. The research recommendation is to make videos under 6 minutes, yet most education videos on YouTube are longer than that. This, too, as evolved as the platform first allowed clips over 10 minutes and then actively encouraged longer videos by rewarding total time viewed. While we still see channels focused on very short videos, most instructional videos are over 10 minutes long with videos of 20 minutes being increasingly common. We also see an increase in videos approach an hour in length. Again, this can be seen by comparing old and recent videos on many education-focused channels. The popular CGP Grey’ explainer videos illustrate this trend nicely.
However, it would be a mistake to think that only videos that track the up-to-date developments in style and approach can be successful. Khan Academy has seen a steady growth in popularity from its simple beginnings on YouTube, yet its style has changed very little. Another example of this is the Insights into Mathematics YouTube channel run by Norman Wildberger which maintains which produces videos that usually exceed 40 minutes and consist of Wildberger talking in front of a whiteboard.
All of the above principles are mirrored by the commercial ecosystem. Here, we also see variety of approaches from LinkedIn Learning (formerly Lynda.com) and its competitors (Skillshare, Udemy) to Creative Live or Great Courses Plus. We see an evolution of style and production values and wide range of formats. Again, very rarely do we see a strict adherence to the 6-minute length recommendation. Unfortunately, there’s been very little research into how and how well people learn from these commercial platforms. Yet, we have to assume that their commercial viability indicates that hey provide value to their customers. Which is also supported anecdotally by people sharing personal experiences of learning.
What do video creators tell us
Finally, we need to turn to creators themselves to explain their process. I have already mentioned Jack Koumi’s practice-inspired guide to making videos but while much of this is still relevant, it does not reflect the whole range of video approaches.
But there are others as well. Sal Khan (founder of the Khan Academy) explains (Khan 2012) his philosophy and approach to making videos (in a video using his style). His recommendations both echo the research results and contradict them.
Many YouTube creators also talk about their process in podcasts which have become a growing supplement to the video platform. Unfortunately, this is rarely systematic, but I can recommend the podcast Hello Internet by two education YouTubers which often addresses questions of YouTube video creation. Relevant episodes can be found via the show notes.
How are videos watched
The question that is asked much less often is how to present videos to their audience after they have been created. Yet, this should not be neglected. The research here is ambiguous as to what is the most effective but shows that people have a variety of preferences for watching videos. Some of the options people are making:
- Watching at increased speed
- Watching on a mobile device (either at home or while travelling)
- Listening to videos as a podcast
- Watching videos without sound and simply reading captions
- Navigating videos via bookmarks or playlists
- Sharing videos and their parts
Yet, the most common approach by learning management systems is to simply embed a video in a page or even force people to download a video file. But we see from both YouTube and the commercial platforms that people require more. Here, LinkedIn Learning is a good example of a feature set that should be increasingly expected of all learning platforms.
Conclusion
If there is one thing that all the research and practice agree on, it is that videos that are made as a personal message that addresses the audience are better than videos that are impersonal. But where they diverge is the importance of cognitive load and length. Here, research needs to take into account the whole depth and breadth of the video learning ecosystem. This is quite difficult within the existing research approach, but it will be necessary if research wants to stay relevant to the needs of video producers.
Editor’s note: a full list of references is available from the author
Author
Dominik Lukes
Digital Learning Technologist, Saïd Business School, University of Oxford, UK
