June 3rd 2025 in Learning
MIT study reveals reading improves learning over video
A field study by MIT found that reading outperforms video for long-term recall, suggesting that enterprise training should pair short films with text and focus on retrieval practice rather than watch time.
Oli Huggins
CEO and Founder
Enterprises still lean on video
Walk into an enterprise classroom today and the first thing you see is a play bar. Covid 19 forced learning teams to put almost every programme online, and managers stayed with the format even after offices reopened. In the first weeks of remote work, time spent on LinkedIn Learning jumped by 130%, an extra 4.8 million hours in only 2 months (source). Budgets followed. 61% of learning leaders now plan to cut traditional classroom spend and two thirds will add more live virtual sessions, while 60% will put funds into on demand content (source). Video feels efficient. It scales, and its dashboards promise tidy proof of effort. A single field experiment from the Massachusetts Institute of Technology asks whether the hours watched actually stay in working memory.
Inside the MIT experiment
The setting was the World Education Congress in Indianapolis. Researchers from the MIT Integrated Learning Initiative borrowed a meeting room and asked delegates to volunteer for a study on learning habits. After a short briefing the delegates drew slips that placed them in one of two sessions. One group sat in front of a screen and watched senior lecturer Peter Senge explain the Four Fields of Listening. The second group received the same words on paper, along with each diagram that appeared in the film (source). Everyone knew a short quiz would follow the next morning, long enough after the lesson to test genuine recall rather than a temporary echo. With the formalities done, the researchers let the participants return to the conference party, trusting the mix of early nights and late drinks to mirror real working life.
Unpacking the results
Breakfast brought the assessment. The average scores were close, 81.3% for the viewers and 82.5% for the readers (source). A deeper look told a sharper story. The median reading score reached 90% while the median for video settled at 80%, and (source) readers also showed a narrower spread, which hints at steadier comprehension across individuals. The survey data added more. Roughly 30% of the delegates said they usually preferred reading, while only 20% named video as their first choice (source). Preference mattered. Participants who learned in their favoured medium scored about 10 points better than the peers asked to switch (source). Sleep added an unexpected twist. More than 60% had 6 hours of rest or less. Among weary readers the median slipped by 12, but weary viewers held steady, a pattern the researchers noted but could not explain (source).
Why reading retains an edge
The numbers invite a question. Why did text outperform video? Part of the answer is control. A reader adjusts tempo without thinking about it, pausing after a dense paragraph and skimming familiar ground. Stopping a film takes clicks and a timeline search, friction that discourages reflection at the moment it matters most. The page also invites marks in the margin. Each underline or quick note becomes a cue for later retrieval, and generating that cue is itself a memory workout. Video rarely prompts the same effort. The experimenters even saw this in language. Non-native speakers who read the transcript outscored native speakers by 20 points, evidence that patient engagement with text can close fluency gaps (source).
Guidance for learning leaders
The Indianapolis study covered one concept on one day, but its message carries weight for enterprise training. When accuracy and longevity count, offer the reading path first. Release a full transcript the moment you publish a clip, and encourage learners to skim the script before they press play. Build interactive checkpoints that ask for written answers rather than simple clicks, so every session includes retrieval practice. Keep the video short and direct. If a demonstration is essential, let it live in the lesson, but do not let it replace the sentence where a critical figure or legal clause will later be searched. Measure delayed quiz results and in-document search queries rather than minutes viewed. The metrics will feel slower, and they reveal the learning that survives once the dashboard is closed.
For more on the strategic context here, see our complete guide on book-to-course transformation. On the related territory of multimodal learning content for engineering teams, our pillar guide is here.