The week before exams: sleep, retrieval, and load — what the evidence actually says

1. Memory consolidation and why sleep is not optional decoration

Sleep supports the offline reorganisation of memory traces — including stabilisation and, in some paradigms, enhancement of declarative learning. Reviews in cognitive neuroscience consistently frame sleep as part of the memory system, not a passive break (e.g. Diekelmann & Born, 2010; Walker & Stickgold, 2006). Sacrificing sleep for extra re-reading often trades away the very process that makes yesterday’s study stick for tomorrow’s paper. The practical implication is blunt: protect a minimum sleep window the way you protect an exam slot — not because discipline is fun, but because the architecture of consolidation is time-locked to sleep in normal learners.

2. Retrieval under time pressure beats passive re-exposure

The testing effect — that retrieving information strengthens retention relative to restudying the same material — is one of the most replicated findings in the learning sciences (Roediger & Karpicke, 2006; Karpicke & Roediger, 2008). In the last week, the highest-yield sessions often look like short, clocked attempts to produce answers, then check against a key or mark scheme, then repeat the same skill gap — not another passively smooth read-through that feels fluent. Fluency after re-reading is a poor proxy for what you can generate in the hall.

3. Cognitive load: your plan has to fit in working memory

Sweller’s work on cognitive load distinguishes demands that serve learning from those that consume limited working-memory capacity without adding understanding (Sweller, 1988). When you are overwhelmed, “revise everything” is extraneous load: it is vague, so your brain spends capacity negotiating the task instead of retrieving and repairing specific knowledge. Shrinking the next step to one line (“ten questions, topic X, timed 20 minutes, then mark”) reduces extraneous load and makes the next action computable. That is not aesthetic minimalism — it matches how attention is budgeted under stress.

4. A seven-day structure you can defend to yourself

• → List every assessment with a real calendar date. Rank by consequence if ignored, not by comfort.
• → Each day: at least one timed retrieval block (questions, past-paper fragment, closed-book prompts) and explicit marking of where marks died.
• → Repeat the wound (question type, definition set, worked step) until the failure mode changes — not another full paper for comfort.
• → End each day with three written lines: finished, hurt, first move tomorrow — externalise the plan so working memory is not the only storage.

What Offload aims to do (without giving away how we build it)

Offload exists because the science is clear that retrieval, spacing, sleep, and load-aware planning matter — but the administrative work of holding deadlines, review timing, and calendar truth in your head is itself a cognitive-load problem. We are building a student-facing system oriented around one thesis: reduce the executive overhead of “what next, when, and what if my week blows up,” while keeping human agency and honest feedback in the loop. We do not claim a single proprietary formula replaces decades of research; we aim to operationalise what the literature already recommends so that, in a real term, the loop survives contact with shifts, labs, jobs, and care responsibilities. The implementation details stay in the product — the obligation to align with evidence does not.