Interleaving Against the Timetable: Why Blocked Curricula and Narrow Rubrics Reward the Same Missing Skill

ThisThis page sits between two others on this site that, read separately, leave a gap worth naming directly. Memorising vs Learning makes the cognitive-science case for interleaved, spaced practice over blocked, massed practice, but stops at the level of individual study habits. Rubrics in HE is explicit that its own headline evidence — rubrics' well-documented benefit to marks — is not the same claim as a benefit to learning retained, and names Craik and Lockhart's levels-of-processing framework as the reason the gap matters, without developing it further¹. This page connects the two: institutional curriculum structure and rubric design are not separate problems sitting next to the cognitive-science one — they are frequently the same mistake, made at two different levels of the same system, and the fix turns out to be structurally identical at both levels.

Why institutions default to blocked structures

TheThe tension is real, and it reflects a genuine gap between administrative convenience and cognitive efficiency, not a simple case of institutions ignoring evidence they should know better than to ignore. Three separate, individually reasonable institutional pressures push toward blocked, topic-by-topic structure:

• Cognitive ease and lower initial friction. Blocked practice creates immediate fluency — solving ten problems of the same type in a row feels like fast, confident progress, which shows up directly in module satisfaction scores and reduces visible student anxiety, even when (as the sections below cover) it's teaching less than it appears to.
• Administrative simplicity. Scheduling, modular credit structures, and standard timetabling map far more easily onto discrete, sequential topic blocks than onto a curriculum that deliberately, continuously revisits earlier material.
• Alignment with assessment and audit. Writing an assessment titled "Week 4: Dynamic Programming" is straightforward to write, straightforward to map to a learning outcome, and straightforward to defend to an external examiner or accreditation panel. An assessment that deliberately mixes current and much earlier material is harder to file cleanly against a single module outcome, even when it's the more honest test of what a student can actually do.

This last point connects directly to something Rubrics in HE already establishes from the assessment-design side: a rubric, once built, is expensive to write well and cheap to reuse only if the assignment it's built for keeps recurring in a stable, predictable shape². A curriculum that mixes topics unpredictably doesn't just complicate scheduling — it removes the very stability that makes rubric reuse (and therefore rubric quality, since a reused rubric gets refined against real student work over successive cohorts) affordable in the first place. The administrative preference for blocked structure and the practical economics of good rubric design are pulling in the same direction, for genuinely defensible reasons — which is exactly why the fix below works with that constraint rather than by demanding institutions abandon it.

The cognitive cost, named precisely

MemorisingMemorising vs Learning covers the mechanism in depth: blocked practice quietly gives away the hardest part of problem-solving for free³. When a student is working inside a "dynamic programming" block, they never have to work out which algorithmic approach a problem calls for — the block structure has already told them, before they've read a single problem statement. What gets practised is execution once the approach is already known, not the harder, more transferable skill of recognising which approach a novel, unlabelled problem actually needs.

This is precisely Torrance's "criteria compliance" mechanism from Rubrics in HE, operating through a different channel⁴. A rubric criterion phrased as "correctly implements the dynamic programming solution" makes the discrimination step just as invisible as the block structure does: a student can satisfy that criterion perfectly well having only ever practised execution inside a labelled context, and the rubric — like the block — never once required them to notice which technique the problem in front of them actually called for. Blocked curriculum structure and task-specific rubric wording are not two separate risks that happen to point the same direction. They are the same missing skill, protected from ever being tested, at two different levels of the same course.

Two symptoms, one fix: Popham's rubric distinction, applied to curriculum structure

RubricsRubrics in HE already contains the fix, developed for a different reason. Popham's response to rubric-driven instrumentalism is not to abandon rubrics, but to distinguish task-specific rubrics (which describe how to complete one assignment) from rubrics built around the general skill the assignment is meant to be evidence of⁵. A task-specific rubric can be satisfied by a student who has only ever practised the one labelled context it was written for; a general-skill rubric cannot, because it has to specify what "doing this well" looks like across contexts the rubric author didn't write it around.

Read against the interleaving research, this turns out to be exactly the same design axis. A criterion written as "correctly applies [named technique from this week's block]" is task-specific in Popham's sense and rewards blocked execution in Rohrer and Taylor's sense — the same rubric wording is vulnerable to both critiques simultaneously, because it's the same underlying gap. A criterion written as "selects and justifies an appropriate approach among plausible alternatives, and executes it correctly" is general-skill-shaped in Popham's sense and can only be satisfied by a student who has practised genuine discrimination — the interleaving research's central finding, stated as an assessment criterion instead of a study strategy.

Rubric wording	What it can be satisfied by	What it's actually testing
"Correctly implements a hash table for this problem."	A student who only ever practises hash-table problems in a hash-table-labelled context	Execution once the tool is already named
"Selects and justifies an appropriate data structure for this problem, and implements it correctly."	Requires genuine discrimination among plausible alternatives	Recognition and execution — the actual, transferable skill
"Correctly derives the recurrence relation for this dynamic programming problem."	A student inside a labelled DP block, never required to notice DP was the right paradigm at all	Execution once the paradigm is already given
"Identifies whether this problem is best solved by recursion, dynamic programming, or a greedy approach, and defends the choice."	Requires the same recognition step blocked practice removes	The judgement a real, unlabelled problem actually demands

Bridging the gap without dismantling the timetable

NoneNone of this requires institutions to abandon modular scheduling — the practical fixes work inside a blocked timetable, adding deliberate interleaving and spacing without touching the administrative structure sitting above it.

Cumulative retrieval warm-ups. Spending the first five to ten minutes of a session on a low-stakes retrieval quiz mixing material from three weeks ago, two months ago, and today forces recognition practice rather than only current-topic execution, at negligible cost to the timetable. This is a smaller-scale, session-level version of what Bruner's spiral curriculum argued for at the level of a whole programme: deliberately revisiting earlier material with increasing depth, rather than treating a topic as closed once its block ends⁶.

Mixed-problem sets. Rohrer, Dedrick and Stershic's classroom trial gives this a real, ecologically valid evidence base rather than only a laboratory one: 126 seventh-grade maths students received either mostly blocked or mostly interleaved practice, using the same total set of problems, over a three-month period, with an unannounced test given either one day or thirty days later. Interleaved practice produced higher scores at both delays, with the advantage growing at the longer delay (Cohen's d = 0.42 at one day, 0.79 at thirty days) — precisely the durable-retention pattern this page and Memorising vs Learning both argue for, demonstrated in a real classroom rather than a lab⁷. Structuring a lab or problem set so a majority of items cover the current topic and a genuine minority are deliberately drawn from earlier, contrasting topics costs no additional timetable time — it only changes which problems go in the set already being written.

Explicit structural contrast. When introducing a new pattern, showing it directly alongside a previously-learned one it's easily confused with (recursion next to iteration; a stack next to a queue) gives students the side-by-side comparison Gick and Holyoak found is what actually triggers schema induction — without deliberate comparison, structurally similar knowledge tends to stay inert and fails to transfer even when it's technically been taught⁸. This costs a few minutes of a single lecture, not a redesigned syllabus.

Redesigning the rubric to match

TheThe curriculum-level fixes above only pay off if assessment actually rewards the discrimination skill they're building — a course that interleaves practice but keeps assessing with task-specific, block-shaped rubric criteria is training a skill it then declines to test. Rubrics in HE's own scaffolding argument extends naturally here: just as rubric detail should fade deliberately across a programme as students gain independent judgement, rubric content should shift from execution-focused criteria early on toward discrimination-and-justification criteria later, mirroring the same widening scope a spiral curriculum applies to content⁹. A first-year criterion reasonably rewards correct execution of a named technique; a final-year criterion, assessing the same underlying skill area, should increasingly require the student to have chosen that technique themselves, from among real alternatives, and to justify the choice — the thing a real, unlabelled problem will actually demand of them the moment they graduate.

The honest limits of this argument

SimplySimply put: most of the sharpest interleaving effect sizes come from tightly controlled studies. Firth, Rivers and Boyle's systematic review of interleaving specifically as a concept-learning strategy — synthesising 26 studies — found real benefits for both memory (Hedges' g up to 0.65) and transfer to novel items (up to 0.66), but also confirmed that the benefit is largest when the concepts being interleaved are subtly similar to each other, and smaller or inconsistent when they're already easy to tell apart¹⁰. Interleaving is not a universal multiplier on all content — it does the most work precisely where blocked curricula do the most harm: paradigms and techniques students are likely to confuse, not ones that were never going to be confused in the first place.

The same review also independently replicated the specific metacognitive illusion this page and Memorising vs Learning both describe: participants who learned via interleaving outperformed those who learned via blocking, and still judged blocking to have been the more effective method¹⁰. So expect module-satisfaction scores to dip, at least initially, even where the actual learning outcome improves — the same fluency-versus-effectiveness mismatch that makes this whole area worth teaching in the first place will show up on a student feedback form the week you introduce it. That's not a reason to skip the change; it's a reason to know in advance which metric is measuring what.

Where this connects

• Memorising vs Learning — the underlying cognitive-science case for interleaving and spacing, at the level of individual study.
• Rubrics in HE — the full evidence base on rubric costs, benefits, and the task-specific/general-skill design distinction this page extends.
• Scaffolding, the ZPD, and Using GenAI Well — the same "support that must fade deliberately, not stay constant" argument, applied to AI-assisted learning rather than curriculum or rubric design.

References