Visuospatial short-term memory — forward and backward
Staircase: starts at length 2. 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. This is the two-trials-per-length, discontinue-after-two-errors procedure described by Kessels et al. (2000).
Timing: roughly one block per second — 900 ms lit, 300 ms dark. Slower tapping inflates span by leaving room to verbalise the path.
The demonstration that lands: have a student run this while counting backwards from 100 out loud. Their digit span will collapse; their block span will barely move. That double dissociation is the entire argument for separate phonological and visuospatial stores, and students remember doing it far better than they remember a diagram of it.
A caution worth voicing: sequences of the same length are not equally hard. A path that crosses itself or doubles back is harder than one that sweeps across the board. Standardised versions fix their sequences for this reason; this demo generates them fresh, so span will wobble a little between runs.
The blocks are placed irregularly on purpose. If they sat in a neat grid you could name the positions — "top left, middle, bottom right" — and convert a spatial task into a verbal one. Scattered placement blocks that shortcut and forces you to hold the sequence as a path through space, which is exactly the store the task was designed to reach.
That is why this task earns a place beside digit span rather than duplicating it. Verbal material is held in a speech-based store you can rehearse silently; spatial paths are held somewhere else, and talking to yourself doesn't help. Patients with damage to one system can show a badly reduced span on one task and a normal span on the other — the dissociation that put two separate slave systems into the Baddeley and Hitch model in the first place.
Backward block span asks you to hold the path and reverse it, adding an executive load on top of storage. Interestingly the forward-to-backward drop is usually smaller for blocks than for digits, which is itself worth arguing about in a seminar.
Companion demo Digit Span — the verbal half of the same working-memory system →Who devised it. The block-tapping task was developed by Philip M. Corsi in his 1972 doctoral work at McGill University, supervised by Brenda Milner, as a spatial counterpart to digit span for studying patients with temporal-lobe lesions. Corsi's original apparatus was a board of nine identical wooden cubes fixed in an irregular arrangement, which the examiner tapped by hand. The administration and scoring procedure used here follows the standardisation published by Roy P. C. Kessels and colleagues (2000). The interpretation of the task as tapping a separate visuospatial store follows the working-memory model of Alan Baddeley and Graham Hitch (1974).
Reproduced faithfully:
Adapted, and why: