What Is the Operation Span Task?
Operation span — and the broader family of "complex span" tasks it belongs to — is the other standard way researchers measure working memory. Where the N-back asks you to continuously match a stream of items, complex span asks you to do two things at once: a processing task and a storage task, interleaved. In the classic version (Turner & Engle 1989; automated by Unsworth and colleagues in 2005) you see a short math equation, judge whether it's true or false, then see a word or letter to hold in memory. After several of these equation–item pairs, you recall all the stored items in order. The math isn't decoration — it exists to block rehearsal, so the items have to be maintained under genuine dual-task load.
The paradigm earned its place because it works as an instrument. Complex span scores correlate reliably with fluid intelligence and reading comprehension — more strongly than simple storage tasks do — which is why the operation span became a fixture of working-memory research alongside the N-back (Engle 2002). The honest caveat is the same one that applies to the whole field: scores on the task predict performance on other measures, but evidence that practicing the task transfers to anything beyond it is thin. The task measures a capacity; whether training on it changes that capacity is an open question.
What separates a real operation span from a lookalike
A few details separate an honest complex span from a memory checklist with math bolted on:
- Interleaved processing and storage. Each episode must pair a genuine processing judgment (verify an equation) with a to-be-remembered item — not math first, then a separate memory list. The interleaving is what prevents rehearsal, and rehearsal-prevention is the entire point.
- Ordered recall. Complex span tests serial recall: you must reproduce the items in presentation order, typically by selecting from a pool that includes distractors. Recognition alone ("was this word shown?") is a weaker, different task.
- A bounded span. Published complex-span work tops out around 5–7 items per trial, consistent with known working-memory limits (Daneman & Carpenter; Cowan 2001). Apps that scale the list forever are measuring mnemonic strategies, not span.
- A real processing demand. The equations need a time window tight enough to actually interrupt maintenance, and hard enough to occupy attention. If the math is trivial or untimed, it's not loading working memory — it's a delay interval you can rehearse through.
- Recall-time pressure. Conway and colleagues' scoring conventions treat the recall window as part of the task: unlimited retrieval time invites reconstruction strategies that inflate scores.
Operation span in Avoken
Avoken ships an operation-span workout called OpSpan that follows the Unsworth-style automated structure. Each repetition runs a sequence of episodes — a true/false equation to verify within a processing window, followed by a word to hold — and then presents a shuffled recall pool of the target words plus same-category distractors. You tap the words in the order they appeared; any wrong word or order fails the rep, and equation answers give immediate green/red feedback but don't count toward the score — the recall is the measure, matching the published convention.
Across its eight levels, the workout moves the axes the literature uses: span grows from 2 to a hard cap of 5 items (the published complex-span ceiling), the processing window tightens from 4000 to 2500ms and the storage window from 2000 to 1200ms, equations progress from single-digit addition to multi-step operations, and the distractor count grows from 2 to 6. A dimmed decoy word list can flash before recall — a published interference wrinkle the instructions warn you to ignore. Session presets run fewer, longer reps than most Avoken workouts, since each rep is a full encode–recall cycle.
Worth saying plainly: Avoken is a training app, not a treatment. Practicing this task makes you practiced at this task — we don't claim broader transfer. If you like the paradigm, ours is a careful version of it.