long-form video – PJFP.com

A new study published in Communications Psychology, a Nature Portfolio journal, tests a question most of us have quietly wondered about while scrolling: can you actually learn anything from a feed of short videos? Meiting Wei, Yandan Li, Guang-Heng Dong and colleagues at Yunnan Normal University ran three experiments, including an fMRI scan, comparing people who learned from a 10-minute stitched-together sequence of TikTok-style clips against people who learned the same information from one continuous 10-minute video. The answer was not close. You can read the full paper, “Learning via short videos impairs memory accuracy and reduces brain synchrony,” here.

TLDR

Across three experiments with college students in China, people who learned from social-media-style short videos (5 to 7 clips of 30 seconds to 2.5 minutes, spliced into 10 minutes) remembered significantly less than people who watched a single 10-minute long video, even though both carried almost exactly the same spoken narration (about 3,000 words each). In Experiment 1, where viewers were told it was just a relaxation break, short-video viewers still scored lower on an immediate memory test. In Experiment 2, where viewers were told to learn the content, the short-video group scored 43.5% versus 65.8% for the long-video group and forgot 46% of what they knew by the next day versus 20% for the long-video group. In Experiment 3, fMRI inter-subject correlation (ISC) analysis showed that short videos reduced neural synchrony in the superior parietal lobule, precuneus and middle occipital gyrus (regions tied to visuospatial attention, episodic memory and top-down control) and increased synchrony in temporal and frontal regions tied to bottom-up, stimulus-driven attention. Functional connectivity between visual, attentional and cognitive control regions was weaker in the short-video group, and frontal synchrony tracked self-reported short video dependency and self-control failure. The authors conclude that the fragmented, rapidly switching format trades deep encoding for attention capture, while cautioning that well-designed, segmented instructional videos are a different thing entirely.

Thoughts

The most useful move in this paper happens in the introduction, before any data. The authors draw a hard line between two things people lump together as “short video learning.” One is instructional segmentation, where a teacher deliberately cuts a coherent lesson into logical chunks, often paired with retrieval practice, which the research says works well. The other is the feed: incidental, passive, algorithmically sequenced clips with no scaffolding. This study is only about the second one. That distinction matters because the popular defense of TikTok as an educational tool usually borrows credibility from the first category while describing the second. Chopping a lesson into pieces is fine. Having the pieces arrive in random order between unrelated content, with no reason to connect them, is the problem.

The most interesting result is the difference between Experiments 1 and 2. When nobody was trying to remember anything, short videos lowered immediate accuracy, but the forgetting rate the next day was essentially identical across groups (about 36% in both). Once people were told to learn, the long-video group’s forgetting rate dropped to 20% while the short-video group’s rose to 46%. In other words, intention to learn paid off enormously for long-video viewers and barely at all for short-video viewers. Effort did help inside the short-video group (people who reported trying harder scored better and forgot less), but the whole group stayed far below the long-video group anyway. The authors call this an “implicit cognitive cost” of the format that effort alone cannot buy back. That is the practical takeaway for anyone who tells themselves they are scrolling educational content on purpose: the format caps what trying can get you.

The neural picture is not simply “less brain activity.” Short videos produced more synchrony in the superior and middle temporal gyri, the middle frontal gyrus and the superior frontal gyrus, and pulled in the ventral attention network and cerebellum, which long videos did not. The regions that lost synchrony, the superior parietal lobule, precuneus and middle occipital gyrus, are the ones that hold a scene together over time and integrate it into episodic memory. So the brain is working hard on short videos. It is just doing the work of detecting and reorienting to each new salient thing, rather than building one coherent model of what it is watching. The fact that middle frontal gyrus synchrony in the short-video group rose with people’s short video dependency and self-control failure scores hints that heavy users may be trained into exactly that reactive mode, though with 28 people in that group, that correlation deserves caution.

The limitations section, which comes near the end, is more candid than most, and it is worth reading closely. The short-video condition used 2 to 3 non-informative filler clips (aerial landscape shots) to match word counts, and the clips were different travelogue segments rather than one story cut up, so the authors admit that format and content coherence cannot be fully separated. The ISC analysis used only a 60-second window (120 to 180 seconds into the video) of a 10-minute stimulus. The fMRI viewers could not scroll or choose clips, which is a big part of real short-video behavior. The samples were college students, and the study was not preregistered. None of this reverses the finding, and the behavioral effect sizes are large (Cohen’s d around 1.7 for accuracy in Experiment 2). But it does mean the precise claim is “a fragmented, incoherent sequence of clips is worse for memory than a coherent continuous video,” which is somewhat narrower than “short length itself damages memory.”

The closing argument ties this paper to the team’s earlier work on memory retrieval. That work found deficits when people tried to recall short-video content, and this study suggests the problem starts earlier, at encoding: the information never gets properly bound together in the first place. For anyone trying to learn, that points to a simple change in how to spend attention. If you want to keep something, give it a continuous block of time, watch or read the long version, and try to remember it on purpose. If you are scrolling, be honest that it is entertainment. The same ten minutes can leave you with two-thirds of the material or less than half of it, depending on whether the information arrives as one story or as fragments.

Key Takeaways

The study was published in Communications Psychology (2026, volume 4, article 120), a Nature Portfolio journal, by Meiting Wei, Yandan Li, Haosen Ni, Zhenglong Li, Jiang Liu and Guang-Heng Dong.
The research asks whether social-media-style short videos are better or worse than long videos as tools for learning and memory.
The paper cites survey figures of 58.4 minutes per day on TikTok for American adults and 151 minutes per day for Chinese users.
Short videos are defined as user-generated clips from a few seconds to five minutes long, with high sensory salience, fragmentation and algorithm-driven personalization.
The authors separate pedagogically designed instructional segmentation, which the literature shows helps learning, from the incidental, passive short videos found on social media, which are the only thing this study tests.
Prior research shows short videos can increase motivation, engagement and interest, and have been used for language and skill learning, which makes a direct test of memory outcomes important.
Theory predicts trouble: the Atkinson-Shiffrin model says information needs rehearsal in working memory to reach long-term storage, and cognitive load theory says overload impairs learning.
The time-based resource-sharing model of working memory suggests frequent attention shifts cause working memory representations to decay, and short videos force exactly those shifts.
Both video conditions ran exactly 10 minutes. The long video was one continuous excerpt from a 30-minute source. The short-video condition spliced 5 to 7 independent clips of 30 seconds to 2.5 minutes.
All material was neutral travelogue content about lesser-known overseas destinations, chosen to limit prior knowledge and emotional arousal.
Spoken narration was matched almost exactly: 3,016 words for the short-video set used and 3,012 words for the long video.
To match information density, the short-video sequence included 2 to 3 filler clips with no narration, such as aerial landscape shots.
A pilot study with 72 participants confirmed the videos did not differ in positive emotion, negative emotion, pleasure, arousal or familiarity.
Experiment 1 (180 college students) told participants the video was a “relaxation session” with no mention of memory, to test incidental learning.
In Experiment 1, short-video viewers scored significantly lower on the immediate test, with a large effect of video type even after controlling for short video dependency.
In Experiment 1, the next-day forgetting rate did not differ between groups (about 36% in each).
Experiment 2 (185 college students) explicitly told participants to remember the content and that they would be tested.
In Experiment 2, immediate accuracy was 43.5% for short videos versus 65.8% for the long video, a very large effect (Cohen’s d of about 1.76).
In Experiment 2, the short-video group forgot 46% of what they initially knew by the next day, compared with 20% for the long-video group.
Delayed tests were given 24 hours later without warning, using a different but equivalent question set, to prevent rehearsal.
Reported memory effort was similar across groups, so the gap was not explained by short-video viewers simply trying less.
Within the short-video group in Experiment 2, more effort correlated with better accuracy and lower forgetting, but it did not close the gap with the long-video group.
Experiment 3 scanned 59 participants with fMRI (28 short video, 31 long video) while they watched the videos under instructions to remember.
Behaviorally, Experiment 3 replicated the result: short-video viewers had much lower recall accuracy (Cohen’s d of about 1.65).
Inter-subject correlation measures how similarly different people’s brains respond to the same natural stimulus, and higher ISC has been linked to real classroom engagement.
Both groups showed synchrony in visual, dorsal attention, default mode and frontoparietal networks, as is typical when watching movies or narratives.
Long videos produced higher synchrony in the superior parietal lobule, precuneus and middle occipital gyrus, regions tied to spatial attention, episodic memory, contextual integration and event segmentation.
Short videos produced higher synchrony in the superior temporal gyrus, middle temporal gyrus, middle frontal gyrus and superior frontal gyrus, plus the ventral attention network and cerebellum.
The authors read the short-video pattern as neural resources shifting toward detecting and responding to rapidly changing salient stimuli rather than integrating a global narrative.
Functional connectivity was weaker for short videos in six pairs: SPG-MOG, SPG-MFG, SPG-STG, MOG-calcarine, MOG-SFG and STG-MFG, linking visual, attentional and cognitive control regions.
Middle frontal gyrus synchrony in the short-video group correlated positively with short video dependency and self-control failure scores.
In the long-video group, weaker SPG-STG and MOG-calcarine connectivity correlated with higher short video addiction and self-control failure scores.
Combined with the team’s earlier retrieval-focused study, the results suggest short-video memory problems start at encoding, not only at recall.
The authors explicitly say the results are not evidence against well-designed instructional short videos in structured educational settings.
Limitations include a college-only sample, no direct measurement of cognitive load or attention, and no intervention testing.
The design could not fully separate presentation format from content coherence, and the scanner prevented natural scrolling and self-paced switching.
The study was not preregistered, but fMRI and behavioral data and analysis code are publicly available on OSF.

Detailed Summary

Why short videos look like good learning tools, and why they might not be

Short video platforms such as TikTok and Douyin are among the most used apps on earth, and “educational” short videos have exploded on them. On paper, the format has a lot going for it: vivid audio and visuals, brevity that fits busy schedules, and recommendation algorithms that serve people what they want. Studies have shown short videos can raise motivation, engagement and interest, and the reward circuitry involved (the ventral tegmental area and amygdala) helps explain why they are so compelling. But the authors argue that the underlying cognitive architecture tells a different story. Learning depends on moving information from sensory memory into limited-capacity working memory and then, through rehearsal, into long-term memory. Fast pacing and dense information threaten working memory overload, and constant attention switching keeps resetting processing so stable knowledge representations never form. Earlier studies linked short video use to memory loss, weaker short-term and prospective memory, and poorer academic performance, but mostly through correlation. This study set out to test the format directly.

Building a fair comparison between short and long videos

The researchers built tightly matched materials. Both conditions lasted exactly 10 minutes. The long video was one continuous slice of a 30-minute travelogue. The short-video condition was 5 to 7 independent travelogue clips of 30 seconds to 2.5 minutes each, reflecting the rhythm of popular short-form platforms. All content featured lesser-known overseas scenic destinations with a neutral narrative style. Every word of narration was transcribed, and word counts were matched almost exactly (3,016 versus 3,012 for the set used), with 2 to 3 silent filler clips inserted into the short-video sequence to equalize the verbal information load. All videos used original Mandarin narration with Chinese subtitles. A 72-person pilot confirmed the videos did not differ in emotion, pleasure, arousal or familiarity. Memory was tested with multiple-choice questions drawn from the narration, split into two counterbalanced sets so the immediate and next-day tests used different questions.

Experiment 1: incidental learning during a “relaxation session”

In the first experiment, 180 college students were told the video was a relaxation session to settle in before the real experiment. Nothing was said about memory. Immediately afterward they took a surprise 10-question test, and 24 hours later they got an unannounced follow-up test online. Participants who reported high memory effort were excluded, to keep the test genuinely incidental. The groups differed in short video dependency scores, so that was controlled statistically. Even after that adjustment, video type had a significant, large effect on immediate accuracy: short-video viewers remembered less. Forgetting rates, however, were nearly identical (about 36% for both groups), meaning that when nobody was trying to learn, both formats lost information at similar rates after the initial gap.

Experiment 2: intentional learning widens the gap

The second experiment, with 185 college students, was identical except that participants were told to remember the content and that they would be tested. Both groups reported high effort, and effort did not differ between them. The gap grew dramatically. Short-video viewers answered 43.5% correctly versus 65.8% for long-video viewers, an effect size (Cohen’s d of about 1.76) that is very large by psychology standards. By the next day, the short-video group had lost 46% of what they initially remembered, while the long-video group lost just 20%. Inside the short-video group, people who tried harder did better and forgot less, which shows effort still matters. But the format ceiling held. The authors describe this as a format-related cognitive cost that increased subjective effort cannot fully offset.

Experiment 3: what the brain does during short videos

For the imaging experiment, 59 screened participants (right-handed, healthy, not addicted to short videos, not anxious or depressed) watched the same videos in a Siemens 3T scanner after being told they would be tested. They then answered 20 questions, and the short-video group again performed much worse. The researchers analyzed inter-subject correlation, a data-driven method that measures how similarly different viewers’ brains respond to the same naturalistic stimulus. Using a leave-one-out approach on a window from 120 to 180 seconds into the videos, chosen to capture short-video content switching, they compared synchrony maps between groups. Both groups engaged visual, dorsal attention, default mode and frontoparietal networks. Long videos drove stronger synchrony in the superior parietal lobule, precuneus and middle occipital gyrus. Short videos drove stronger synchrony in the superior and middle temporal gyri and the middle and superior frontal gyri, and uniquely recruited the ventral attention network and cerebellum.

Bottom-up capture versus top-down integration

The authors interpret the two patterns as two different cognitive states. The superior parietal lobule supports spatial attention and task-oriented control and is linked to episodic memory and contextual integration. The precuneus, a hub of the default mode network, supports self-referential processing and memory retrieval and exerts top-down control over visual processing and event segmentation. The middle occipital gyrus handles higher-level visual analysis. Together these form a network for sustained attention and integrating a coherent narrative, and they synchronized more during long videos. The regions that synchronized more during short videos are associated with semantic and phonetic processing, novelty response and attentional reorienting. That fits a brain preoccupied with catching the next salient thing. The finding that middle frontal gyrus synchrony tracked short video dependency and self-control failure suggests a link between this reactive pattern and reduced cognitive control among heavier users.

Weaker connections between visual, attention and control regions

Functional connectivity analysis showed six connections weaker in the short-video group: between the superior parietal lobule and the middle occipital gyrus, middle frontal gyrus and superior temporal gyrus; between the middle occipital gyrus and the calcarine cortex and superior frontal gyrus; and between the middle frontal gyrus and superior temporal gyrus. The parietal and occipital links form a visuospatial processing network, so their weakening fits rapid visual transitions that never demand deep integration. The weaker link between primary (calcarine) and higher visual cortex suggests fragmentation disrupts dialogue across the visual hierarchy. Reduced coupling with the superior frontal gyrus, a frontoparietal control hub, is consistent with limited attentional resources theory, where high information flow pushes resources toward external stimuli and away from control. Overall, short videos appear to reorganize processing into a less integrated mode than long videos.

From encoding to retrieval, and the limits of the evidence

The fMRI data came from the same scanning session as the team’s earlier study, which looked at brain activity during memory retrieval. Putting the two together, the authors argue that retrieval deficits after short videos likely originate in poor encoding, visible here as lower synchrony and weaker network integration while watching. They list clear limitations: only college students, no direct measurement of attention or cognitive load (so those interpretations are theory-driven), no intervention testing, and a trade-off between control and realism. Because short-video clips were not one story chopped up, format and content coherence could not be fully separated, and participants could not scroll or choose clips in the scanner. The authors call for designs that manipulate format and coherence independently, and for more naturalistic paradigms. Their conclusion is carefully bounded: passive short-video consumption in low-scaffolding environments like social media carries measurable cognitive and neural costs, but this is not a verdict against well-designed instructional short videos.

Notable Quotes

“SVs not only fail to enhance learning outcomes but also accelerate memory loss and increase the likelihood of recall failure.”
Wei, Li, Dong et al., summarizing the behavioral results of Experiments 1 and 2

“The fragmented and rapidly switching nature of typical social media short videos enhances bottom-up attentional capture at the expense of top-down cognitive processes critical for deep learning and long-term memory consolidation.”
The authors, in the paper’s abstract, on the core mechanism

“SVs, through their rapid and frequent attention-switching, constantly reset the cognitive processing, making it difficult for working memory to form stable knowledge representations.”
The authors, on why the format undermines working memory

“Consumption is typically incidental, passive, and largely devoid of instructional scaffolding that supports deep cognitive processing.”
The authors, distinguishing social media short videos from designed instructional segments

“Such a format-related constraint may constitute an implicit cognitive cost, which may be difficult to fully compensated for by increased subjective effort alone.”
The authors, on why trying harder did not close the gap

“Higher neural synchrony may reflect more stable and shared cognitive states across viewers when processing coherent narratives.”
The authors, interpreting stronger parietal and precuneus synchrony during long videos

“Neural resources are preferentially allocated toward the immediate detection of and response to rapidly changing, highly salient stimuli, rather than toward the deep integration of a global narrative.”
The authors, on the distinct brain pattern evoked by short videos

“Deficits observed during retrieval may, at least in part, originate from suboptimal encoding dynamics.”
The authors, connecting this study to their earlier retrieval research

“These findings should not be interpreted as evidence against the effectiveness of well-designed instructional short videos embedded within structured educational contexts.”
The authors, in the conclusion, on the scope of their claim

The full study, including figures, tables and links to the open data and analysis code, is published in Communications Psychology. Read the full paper here.

Tag: long-form video

Short Videos Impair Memory and Reduce Brain Synchrony, fMRI Study Finds: Why TikTok-Style Learning Leads to Faster Forgetting Than Long Videos