Recent clean blocks
No V0.310 blocks yet.
DUAL N-BACK · POSITION + SOUND · LOCAL-FIRST
Dual N-Back adaptive blocks, weak review & clean local progress
Train one clean block at a time: N warm-up cues, then scored visual-position and sound decisions. Adaptive training uses the weaker channel to set the next N, while Weak Review, Daily 3 and focused single-stream blocks stay local to this browser.
Live level2-back
V0.310 · TRAINING FIRST
Adaptive Dual N-Back
N warm-up cues → 20 scored trials → separate position and sound evidence → next-level recommendation.
Recommended2-backToday 0 / 3 blocks
Session setup
Adaptive/Weak/Daily use your recommended N. Learn 1-back locks N=1 and shows the comparison rule. Settings are stored only in this browser.
BlockReady
Trial0 / 20
Position—
Sound—
Streak0
Local best—
Ready. No score starts until the warm-up cues finish.
How adaptation works
Each active stream uses balanced accuracy: average target hit-rate and non-target correct-rejection rate. If every active stream is at least 80%, next N rises; if either falls below 65%, next N falls; otherwise it stays. The recommendation is task-local, not an IQ or clinical score.
Weak Review appears after enough local evidence and trains the lower-performing stream directly.
Advanced analytics, fixed benchmarks & legacy modes
V0.208 signal-detection analytics, Clean-5 benchmarks, reproducible setup/backup, legacy Visual/Audio/Dual/Adaptive 30-trial modes and trial review remain available below. They are secondary to the training loop above.
Dual 30Watch the square and listen to the tone. Respond separately when either channel matches N trials back.
Trial 0 / 30Cue Ready
N trials backPosition: A · Sound: L
Ready. Start when you can stay on this tab for the whole session.
Response accuracy—
Hits0
Misses0
False alarms0
Sessions0
Recommended start2-back
Total match hits0
Daily streak0
PB this mode + N—
Recent sessions
No completed Practice sessions on this device yet.
What the score means
Every active channel on every trial is classified as a hit, miss, false alarm or correct rejection. Response accuracy is the share of those decisions that were correct. A false alarm means marking a match when the current cue did not actually equal the cue N trials earlier.
The platform also stores a local 0–1,000 Practice score so My Mix and Progress can compare your own sessions. That internal scale is not a percentile, not an IQ estimate, not a medical or attention assessment and not a population norm.
Signal Detection & Repeatability Lab
Keep the base game fast, then inspect exactly what happened. V0.208 stores browser-local trial outcomes, response times, channel-specific d′ with loglinear correction, n±1 lure errors and repeatability evidence. d′ here is only a task-local signal-detection metric — not an IQ, attention, clinical or population score.
No reviewed Dual session yet.
Finish a base Dual N-Back session to retain trial-by-trial task evidence locally.
Visual d′—
Audio d′—
Visual hit RT—
Audio hit RT—
Trial review
No trial selected.
Complete a Dual session first.
0/0
Recent reviewed sessions
Clean-5 repeatability
Clean 0/5Complete Fixed Dual 24 seeds A→E without leaving the tab.
The five benchmark sequences are deterministic but different, so Clean-5 does not repeat one memorized stream. It summarizes your own accuracy, d′ and timing consistency only.
FIXED / ADAPTIVE PLAYER
Next: Fixed Dual 24 · A
ReadyFixed runs are Dual 2-back. Adaptive drills target the weakest local pattern from your recent evidence and never enter Clean-5.
Reproducible setup & local backup
Copy the current Practice seed/mode/N or Daily date. JSON backup includes base N-back stats plus V0.208 review and repeatability data; nothing requires an account.
Display & evidence settings
V0.208 lab ready. Finish a base Dual session or start the fixed benchmark.
Visual 30
Ignore sound and mark Position match only when the current square is the same square shown N trials earlier. This isolates the spatial stream and is useful for learning the rule before combining channels.
Audio 30
The grid is visually muted while browser-local synthesized tones play. Mark Sound match when the current tone frequency equals the tone N trials earlier. Use Sound check before starting if your device is muted.
Dual 30
Track position and sound independently. The two channels can match together, separately or neither. A single trial can therefore require zero, one or two responses.
Adaptive 3×18
Three dual-channel blocks use your actual block accuracy and false alarms to choose the next N. Strong controlled performance can move the next block up; repeated misses or false alarms can move it down.
Deterministic Practice
Practice uses a deterministic URL seed. Reopening the same mode, N and seed recreates the same cue stream, including the planned match locations, without a server puzzle database or external data corpus.
Daily N-Back
Each date deterministically selects a 2-back or 3-back Dual 30 sequence. Daily N-Back history and streaks are specialist records and stay separate from the frozen Today’s 6 lineup.
N-back is a working-memory task, but a browser session cannot diagnose memory, attention, ADHD, intelligence or neurological status. Research on whether repeated N-back practice transfers broadly to unrelated abilities is mixed. Use the numbers as task-specific practice feedback and compare your own sessions under similar audio and device conditions.