Verbal short-term memory — forward and backward
Staircase: starts at length 3. Two trials at each length. Move up if at least one is correct; stop after both trials at a length are failed. Span = the longest length with at least one correct reproduction.
Timing: one item per second — 900 ms on, 300 ms off. Slower presentation inflates span because it leaves room for rehearsal and grouping. Worth demonstrating deliberately: have one student chunk the digits aloud in pairs and watch their span jump.
The discussion prompt that always works: ask who scored near 7, then ask what Miller actually claimed. He proposed roughly seven chunks, not seven items. A student scoring 9 has almost certainly chunked — ask them how, and you've taught chunking without a slide.
Do not present this as Digit Span from a Wechsler scale. The paradigm is public domain; the Wechsler item sets, norms and scaled scores are protected and are not reproduced here.
Forward span is a storage measure: how much material the speech-based store can hold while you rehearse it. Backward span asks you to hold the material and operate on it at the same time, which recruits executive control on top of storage. Backward is normally the shorter of the two, and the size of that drop is more informative than either number by itself.
Digits are held phonologically — which is why span shrinks for longer words and for items that sound alike, and why saying something irrelevant out loud while you rehearse wrecks it. For the spatial counterpart of this task, where silent rehearsal doesn't help at all, run the Corsi board.
Companion demo Corsi Block-Tapping — the spatial half of the same working-memory system →Who devised it. The digit-span method was introduced by Joseph Jacobs (1887), who had schoolchildren repeat back lists of increasing length and called the resulting measure "span of prehension." George A. Miller (1956) gave the finding its famous framing, and Nelson Cowan (2001) later argued the pure capacity limit is closer to four once rehearsal and chunking are blocked. The forward/backward distinction and its interpretation as storage vs. executive load follow the working-memory model of Alan Baddeley and Graham Hitch (1974).
Reproduced faithfully:
Adapted, and why: