Chunking and simplifying
You lose them at minute forty of a hard technical topic
You can see the room go at about minute forty, and adding a break in the middle did not fix it.
What it is
Breaking a session into segments that end where the argument ends, alternating exposition with something students do. The fellowship’s session on this came out of memory research, and its wider content on spacing and retrieval belongs here too.
The fellowship supplied a number, a thirty-minute lecture limit with ten-minute segments. It never supplied a procedure, and the number has no source. Cut on the argument instead.
Run it
- Cut at the seams of the argument, not on a clock. A segment ends where a student could state in one sentence what was just established. If that takes eleven minutes, the segment is eleven minutes.
- Sequence the worked examples deliberately. The rule that discriminates between two confusable concepts belongs before the second worked case, not after it.
- Hold one number constant across both worked cases, so the only thing that changes is the thing you are teaching.
- Open with a roadmap slide naming the day’s four or five beats in order, so the seams are visible before students hit them.
- Post the literal return time on the break slide. “We resume at 2:40,” not “ten-minute break.” A posted duration starts when the last person leaves the room. A posted clock time starts now.
Say this
Seven-beat template for any paired-concept block
- Criteria for concept A, attached to one named case.
- Concept A applied to that case, with the numbers on the slide.
- Concept B, by definition only.
- What B shares with A, then the discriminating rule. Before the second case, not after it.
- Concept B applied, with only one variable changed from step 2. Hold the anchor number constant.
- The decision algorithm, withheld until both cases exist.
- One if-then line per possible pattern, including the pattern you did not work.
This is my abstraction of my own EPI 501 deck, not a fellowship artifact.
Techniques
Retrieval practice
The same session’s material on memory supports a second, cheaper move. Open a session with two or three questions from two or three weeks back, notes closed, ungraded, answers given immediately. Recent material feels productive and does almost nothing; the requirement to produce an answer from memory is the whole mechanism, and it is the step that gets quietly dropped.
Spoken at the start of class
“Three questions from week four, notes closed. This is not graded and I am not collecting it. You will get them wrong, and getting them wrong now is the point, because that is what makes them stick until March.”
In my courses
Planned for EPI 601, every week.
I will break each class session into segments of no more than 20-30 minutes, alternating between mini-lecture and active application. Complex topics (e.g., interaction on additive vs. multiplicative scales) will be introduced incrementally: concept first, then notation, then worked example, then student practice.
601 meets Mondays for three hours, which makes this the practice with the widest reach and the one checked for all fourteen weeks. The plan is to pre-identify the two or three hardest concepts per session and chunk specifically around those, with confounding, interaction, and precision named as the hardest. New for 601; not used before the fellowship.
Evidence
Segmenting improves retention (d = 0.32) and transfer (d = 0.36) across 88 comparisons (Rey et al., 2019). Studying worked examples beats unguided problem solving for novices on novel technical material (Sweller, 1988). Retrieval practice is better supported than either: practice testing beats restudy across meta-analyses (Adesope et al., 2017; Rowland, 2014) and the effect survives in real classrooms (Roediger et al., 2011; Yang et al., 2021).
Two numbers to stop repeating. The fifteen-minute attention ceiling has no source; reviews of the studies behind it found they provide little support for the estimate (Bradbury, 2016; Wilson & Korn, 2007). In live classrooms, lapses of under one minute are the most common kind and attention cycles shorten as a lecture proceeds, with no cliff (Bunce et al., 2010). The finding worth carrying is that the greatest variability in attention comes from differences between teachers rather than from lecture format.
The segmenting meta-analysis is also not about live lectures. Segments averaged 75.5 seconds inside materials running about nineteen minutes, learning time came out at d = −0.92 so segmenting costs time, and learner-paced segmenting showed no reliable retention effect at all.
One boundary on retrieval: when spacing is controlled, restudy becomes effective too (Soderstrom et al., 2016). Retrieval’s advantage is partly a spacing advantage.
Finally, watch uniform scaffolding. Support that helps novices actively degrades performance for students who already have the schema (Kalyuga et al., 2003), which is the most transferable result here for a seminar with mixed preparation.