Event Segmentation Theory: Why Some Training Feels Clear and Some Feels Like One Continuous Mistake
You’re watching a colleague demonstrate a process in a new piece of software. She moves through it at a reasonable pace, talking as she goes, clicking between screens, adjusting settings, entering data, and checking outputs. It takes about twelve minutes. By the end, you feel like you followed along, but you couldn’t confidently repeat any of it. The whole thing blurred together into one continuous flow of mouse movements and narration.
A week later, a different colleague demonstrates a different process in the same software. This one takes roughly the same amount of time, but something about it is different. Each time the goal shifts, from setting up the record to entering the data to running the verification, she pauses briefly, names what’s about to change, and carries on. By the end, you can recall the distinct phases and roughly what happened in each one. You could probably have a reasonable go at it yourself.
The difference between those two experiences isn’t about presentation skill, or pacing, or even the complexity of the task. It’s about whether the structure of the demonstration aligned with the way your brain was trying to organise the information. Event Segmentation Theory, developed primarily by Jeffrey Zacks and colleagues at Washington University in St. Louis, explains why.
What your brain does with continuous experience
Your brain doesn’t process the world as one unbroken stream. It automatically divides ongoing experience into discrete chunks, which researchers call “events,” and does so continuously, without you deciding to do so or being aware that it’s happening. When you watch someone make coffee, your brain silently registers separate episodes: getting a mug, adding grounds, pouring water. When you drive to work, your brain segments the journey into getting out of the driveway, joining the main road, navigating the roundabout, and pulling into the car park. Different people watching the same activity will place those boundaries in remarkably similar places (Zacks and Tversky, 2001).
The mechanism driving this is prediction. Your brain maintains a running model of what’s happening right now, and it uses that model to predict what will happen next. Most of the time, those predictions are roughly correct, and the current model holds steady. But when the predictions fail because a goal has shifted, the location has changed, a new person has entered the situation, or the causal chain has broken, the brain registers what researchers call an “event boundary.” It closes the file on the current episode, clears the working model, and starts building a new one from fresh sensory input (Zacks et al., 2007).
Note: This is a relatively simplified version of what is actually happening in the brain because, of course, as L&D, HR, and performance enablement professionals, we don’t need to become neuroscientists. We just need a working understanding.
If you are interested, I would suggest reading the source material behind all of this, as it goes into much greater detail.
This process is automatic, and it’s hierarchical. Your brain segments at multiple levels simultaneously: fine-grained events (reaching for the mug) sit within coarse-grained events (making breakfast), which sit within even larger episodes (the morning routine). The theory was first outlined by Zacks and Tversky in 2001, building on earlier work by Darren Newtson (1973), who had shown that people spontaneously parse continuous behaviour into meaningful units when asked to mark transitions. The full theoretical architecture, including the role of prediction error and neural substrates, was published by Zacks, Speer, Swallow, Braver, and Reynolds in 2007.
Neuroimaging studies have confirmed that this isn’t just a metaphor. Brain activity, particularly in the hippocampus, posterior temporal cortex, and lateral frontal regions (collectively known as the posterior medial memory network which represents a contextual system that binds specific details, events, and spatial information with the hippocampus acting as the core node for encoding and retrieval and the connected regions providing context), transiently increases at the moments when people perceive event boundaries, even during passive viewing with no explicit instructions to segment anything (Zacks et al., 2001b). All of this tells us that the brain treats these moments as transition points rather than arbitrary pauses.
Why boundaries matter for memory
Information present at a boundary, the moment when one event ends and another begins, gets encoded more strongly than information in the middle of an event. The boundary acts like an attentional gate: it opens briefly to let new information in, and that information gets a better foothold in long-term memory as a result (Kurby and Zacks, 2008).
People who segment events in ways that are consistent with how most other people segment them, sometimes called “good segmenters” in the literature, remember significantly more about what they watched, even after controlling for working memory capacity, processing speed, and general knowledge. Segmentation ability predicts event memory independently of those other cognitive factors (Sargent et al., 2013), and the benefit persists for at least a month (Eisenberg, Zambrano, Aue, Flores and Zacks, 2017).
There’s a trade-off, though. While boundaries improve memory for what happens at the transition point, they impair memory for temporal order across the boundary. Items that span a boundary are harder to sequence correctly and are remembered as being further apart in time than they were (Ezzyat and Davachi, 2014). The most vivid illustration of this is the “doorway effect”: walking through a doorway into a new room causes forgetting of recently acquired information, because the spatial transition triggers an event boundary and the brain files away the previous event model (Radvansky, Krawietz and Tamplin, 2011). In other words, when you walk into the kitchen and can’t remember why you came in, you’re feeling the doorway effect. The doorway created a boundary, and the information from the previous room became less accessible.
For training design, the key finding comes from Pettijohn, Thompson, Tamplin, Krawietz, and Radvansky (2016), who showed that memory was better when information was distributed across two events, separated by a boundary, than when it was combined into a single event. Introducing more boundaries increased the benefit further. Deliberately placing meaningful transitions into training content can improve recall, provided those transitions represent shifts in meaning rather than arbitrary breaks.
To up your learning science game in 2026, consider attending the Evidence-Informed Practice Conference as we prepare to explore practical insights from the worlds of cognitive neuroscience, psychology, behavioural science and much more.
What counts as a meaningful boundary
Six features of a situation reliably trigger event boundaries: spatial or location changes, character entrances or exits, new object interactions, goal shifts, changes in causal structure, and temporal discontinuities (Speer, Zacks, and Reynolds, 2007). In practical terms, the most reliable triggers for workplace training are changes in what you’re trying to achieve (the goal), changes in where you are or what you’re looking at (the environment), and changes in why the current action matters (the causal structure).
This is where the distinction between meaningful segmentation and arbitrary segmentation becomes important. Splitting a software training video at five-minute intervals provides some benefit simply by giving the user processing time. But it misses the structural mechanism that EST describes. If the break falls in the middle of a coherent action sequence, say halfway through configuring a particular setting, it can disrupt the situation model the learner is building rather than reinforcing it (Spanjers, Van Gog and Van Merriënboer, 2010).
Boundaries placed where goals change (”now we’ve finished setting up the case, we’re going to enter the clinical data”), where the screen environment changes (moving from one module to another), or where the causal logic shifts (”we configured the alert; now we need to verify it’s triggering correctly”) align with the brain’s natural segmentation process. They help users perceive the underlying structure of the task, which is what produces the learning benefit.
The evidence for segmentation in instruction
A meta-analysis of 56 studies covering 88 comparisons, conducted by Rey, Beege, Nebel, Wirzberger, Schmitt, and Schneider (2019), found that segmenting instructional content produced small-to-medium improvements in both retention and transfer, and that segmentation also reduced overall cognitive load. The effect size was reported as a Cohen’s d of around 0.4, which in practical terms means that if you took two groups and gave one segmented instruction and the other continuous instruction, about 66% of the segmented group would outperform the average of the continuous group. That’s a meaningful difference, especially given that segmentation costs nothing to implement.
For readers less familiar with effect sizes: a Cohen’s d measures how large the difference is between two groups, expressed in standard deviations. A d of 0.2 is considered small, 0.5 medium, and 0.8 large. Most instructional interventions produce effects in the small-to-medium range, so a d of 0.4 from segmentation alone is a reasonable return for a design decision that doesn’t require additional resources.
Both system-paced segmentation (where the system controls the pace, inserting pauses or breaks at set points) and user-paced segmentation (where the person clicks to advance) were effective. Some earlier studies by Mayer found effect sizes as large as d = 1.0 for problem-solving transfer tasks, though these used very specific experimental conditions that may not generalise straightforwardly to typical workplace training (Mayer, Dow and Mayer, 2003). The benefit is strongest when the material is complex and fast-paced, and the user is relatively new to the topic.
Spanjers, Van Gog, and Van Merriënboer (2010; 2012) identified two distinct mechanisms behind the segmenting benefit. The first is a pausing mechanism: giving learners processing time between segments. The second is a temporal cueing mechanism: making natural boundaries visually or auditorily salient so the learner can perceive the structure of the content. Their 2012 study showed that temporal cueing, briefly darkening the screen at meaningful transition points, contributed to learning independently of the pausing effect. This means that a “continue” button alone isn’t sufficient; the transition itself needs to communicate that something has changed.
Practical application for training design
The most direct way to apply EST is to structure process training and standard operating procedures around the natural event structure of the task. Rather than organising steps by convenience or by how they appear in a system, map them to the hierarchical structure of the activity: major phases first (the coarse events), then detailed steps within each phase (the fine events). If you’re documenting a procedure, try recording it on video and asking experienced practitioners to mark where they perceive natural transition points. The boundaries they identify will correspond closely to the boundaries learners’ brains will try to impose, and those are the structural skeleton your documentation should follow.
For software training videos and screen recordings, insert clear transitions at the points where the goal changes or the screen environment shifts. Name the transition explicitly: “We’ve finished configuring the alerts; now we’re going to test whether they fire correctly.” This verbal signal serves as both a temporal cue and a micro-summary, helping the learner close off the previous event model and prepare for the next one. Adding a brief pause at this point gives them processing time without requiring them to click anything.
For scenario-based learning, structure your decision points around event boundaries. A scenario that moves from “gather information” to “form a hypothesis” to “take action” represents three natural events with clear boundary triggers: the goal shifts at each transition. These are the cognitively optimal moments for reflection prompts, because the learner’s brain is already in the process of updating its situation model. A question like “what’s different about the situation now?” or “what information do you still need?” prompts explicit updating of the mental representation, which strengthens encoding.
Even simple auditory or visual cues at event boundaries improve memory for both younger and older adults. Gold, Zacks, and Flores (2017) found that cues aligned with event boundaries were significantly more effective than cues at mid-event points, though even suboptimally timed cues provided some benefit. For live demonstrations and facilitated training, this translates to pausing, changing position, or verbally signalling when a phase is complete before moving to the next one.
Three mistakes worth avoiding
Segmenting by time rather than by meaning
Splitting a twenty-minute video into four five-minute chunks is better than no segmentation at all, because it provides processing time. But the breaks are arbitrary, and if they fall mid-event, they can disrupt the user’s developing mental model rather than support it. The evidence is clear that boundary-aligned breaks produce better outcomes than random breaks (Gold, Zacks and Flores, 2017).
Adding visual breaks or headings that don’t communicate a shift in goal, context, or causal structure
A new slide or a section divider that doesn’t tell the user what has changed is segmentation in appearance only. It provides a pause, which has some value, but it doesn’t trigger the situation model updating that produces the deeper learning benefit. Meaningful labels name what has changed: “Now that we’ve identified the root cause, we need to choose an intervention,” not just “Part 3.”
Over-segmenting
This can prevent users from building the higher-order event model that gives a procedure its coherence. If you fragment a process into dozens of tiny micro-modules, each covering a single action, users may end up with a collection of isolated fragments rather than an integrated understanding of how the whole thing fits together. The microlearning trend carries this risk: granularity is not the same as clarity. The goal is to align your instructional structure with the task’s event structure, not to make everything as small as possible.
To up your learning science game in 2026, consider attending the Evidence-Informed Practice Conference as we prepare to explore practical insights from the worlds of cognitive neuroscience, psychology, behavioural science and much more.
One more thing worth knowing
The segmenting effect interacts with user expertise. Spanjers, Wouters, Van Gog, and Van Merriënboer (2011) found that segmented animations were more efficient for students with lower prior knowledge, but students with higher prior knowledge learned equally well from non-segmented versions. Experienced performers already possess domain knowledge and event schemata that allow them to mentally segment content. For them, explicit segmentation is redundant and may even be counterproductive if it forces them into a pace or structure that doesn’t match their existing mental model. This is consistent with the expertise reversal effect observed elsewhere: scaffolding that helps novices can become a hindrance for experts.
The practical implication is straightforward. Segment heavily for people who are new to a task or domain, making boundaries visible, naming transitions, adding processing time and reflection prompts. For experienced performers, consider self-paced navigation or less structured formats that allow them to impose their own mental organisation.
Where to start
If you’re designing process training, screen recordings, scenario-based learning, or standard operating procedures, the starting point is to identify the natural event structure of the task. Ask where goals change, where the environment shifts, and where the causal logic moves from one phase to another. Build your instructional structure around those transitions. At each boundary, name what’s changed, give the user a moment to process, and consider adding a brief prompt that asks them to update their understanding before moving on.
Event Segmentation Theory has been accumulating evidence for over twenty years; the core findings are robust, and the practical applications don’t require special tools or budgets. One of the most common mistakes in training design is segmenting by time or by convenience rather than by meaning, and it’s one of the simplest mistakes to fix.