Quad-stream visual working-memory trainer
Four visual streams at once: position, shape, color and rotation. Press a stream's pad when it matches the item N trials back. Ignore the lures that match N−1 or N+1.
Streams in play
Intention for this set
Each trial shows one item for a fraction of a second. It carries four features at once.
Press every pad whose feature matches the item N trials ago. Several pads can match on one trial.
Features that match N−1 or N+1 back feel familiar but are wrong. Holding back scores points.
The tier rises above 85% and eases below 65%, so you train near your breaking point.
Telemetry
Peak N reached in each set
Set accuracy with tier-up and recalibration lines
Lifetime accuracy per stream · lure blocks
Minutes per day, last 5 weeks
| When | Tier path | Accuracy | Flow time | Lures blocked | Peak sync | Score |
|---|
Configuration
Design rationale & sources
Every mechanic below maps to a finding about dopamine, reward or working memory. The first six come from the original brief. The last six are added mechanics drawn from further research. Each entry says what the game does and what the evidence actually shows.
In the game: position, shape, color and rotation are tracked at once, with no audio. All load lands on visual and spatial working memory.
Evidence: multi-stream n-back training improves n-back performance. Transfer to unrelated abilities is weak and debated.
Jaeggi et al. 2008, PNAS · Melby-Lervåg, Redick & Hulme 2016, Perspect. Psychol. Sci.
In the game: the stimulus stays fixed (350 ms by default). The gap before the next one is the base interval plus a fresh random 300–600 ms on every trial.
Evidence: the brain builds precise timing expectations for rhythmic events. Random timing blocks that shortcut, so attention has to stay up.
Nobre & van Ede 2018, Nat. Rev. Neurosci.
In the game: each stream is generated at 40% true matches, 35% lures (the N−1 or N+1 item, never the target) and 25% noise.
Evidence: lure trials produce most n-back false alarms because familiarity has to be overridden by control.
Kane, Conway, Miura & Colflesh 2007, J. Exp. Psychol. LMC
In the game: before many hard frames, a 0.5×–2.5× multiplier rolls in the HUD and locks just before the item appears. It applies to that trial only.
Evidence: ventral striatum activity rises while people anticipate an uncertain reward, before it arrives.
Knutson, Adams, Fong & Hommer 2001, J. Neurosci.
In the game: a sliding window of recent trials drives the HUD breakpoint gauge. Above 85% the tier rises; below 65% it drops at once. The target band is 75–82%.
Evidence: prefrontal working-memory neurons respond to D1 stimulation along an inverted U. Separately, learning models are fastest at about 85% accuracy.
Vijayraghavan et al. 2007, Nat. Neurosci. · Wilson et al. 2019, Nat. Commun.
In the game: blocking a lure during a sync streak can fire a canvas particle burst. Whether it fires is probabilistic, and its speed scales with streak, tier and multiplier, so it stays surprising.
Evidence: dopamine neurons fire for rewards that beat expectation and adapt to rewards that are fully predicted.
Schultz, Dayan & Montague 1997, Science
In the game: about 1 in 11 trials is marked with a “?”. A perfect answer on that trial opens a bonus exactly half the time.
Evidence: sustained dopamine-neuron activity during anticipation peaks when reward probability is 50%.
Fiorillo, Tobler & Schultz 2003, Science
In the game: the progress bar brightens as the set nears its end, and the last 20% of trials score ×1.25.
Evidence: striatal dopamine ramps up as an animal gets closer to a distant goal.
Howe, Tierney, Sandberg, Phillips & Graybiel 2013, Nature
In the game: each set gets a new accent for the grid frame and HUD. Stimuli never change, so learning carries over.
Evidence: novel stimuli activate the substantia nigra and VTA more than familiar ones.
Bunzeck & Düzel 2006, Neuron
In the game: points grow with N1.4, so harder tiers pay far more for the same response.
Evidence: dopamine shapes how much effort, including mental effort, people will spend for a reward.
Salamone & Correa 2012, Neuron · Westbrook & Braver 2016, Neuron
In the game: you pick a goal before each set: flow time, sync streak, lure blocking or a tier climb. The summary checks it.
Evidence: the chance to choose is itself rewarding and engages the striatum.
Leotti & Delgado 2011, Psychol. Sci.
In the game: the summary flags late-night sessions and days past about 25 minutes, and the sleep-first advice comes with the reason.
Evidence: one night without sleep was linked to lower D2/D3 receptor availability in the ventral striatum.
Volkow et al. 2012, J. Neurosci.
Two small PET studies link working-memory training to dopamine changes. McNab et al. (2009, Science) found that about 14 hours of training changed cortical D1 receptor binding in 13 young men; binding went down as working memory improved. Bäckman et al. (2011, Science) found more striatal dopamine release during an updating task after five weeks of training.
No study has shown that this app, or any game, raises D1 or D2 receptor density or baseline dopamine. Expect clear gains on n-back-style tasks and uncertain transfer beyond them. This is a training game, not a medical treatment. Sleep, exercise and regular practice matter more than any single session.