v · elvárás — the expectation engine

meglepetéses · surprise — the sound that does not belong

Half a minute of a morning, told entirely in sound: a door, stairs, a tap, a kettle, a radio being tuned, a chair, the door again. It repeats. Your job is to press before the next thing happens — not on it, before it. Get it right and the loop is rewarded with less: it thins out and it speeds up. Get it wrong and it thickens. And once you know the loop well enough, it begins to violate itself.

Every sound on this page is synthesised, and that is a compromise. The built version of this strand runs on location recordings, concatenatively assembled so that no two runs are identical and no event lands where memory expects it. A browser cannot carry that corpus, so the door, the stairs, the kettle and the radio here are made out of filtered noise and swept oscillators. They are recognisable enough to be learned, which is what this particular game needs — but the sister game M1, which turns on whether a sound belongs, would be measuring the synthesis rather than the listener, and so it is described below rather than built.

what this strand isolates

Expectation and its violation — the mechanism that all music, and a good deal of comedy, actually runs on. David Huron’s account of it in Sweet Anticipation (2006) is a five-stage machine: imagination, tension, prediction, reaction, appraisal. This piece takes that machine apart and plays with the pieces. You build a prediction; the loop rewards a correct one by giving you less to hear, which is the opposite of how a game normally pays; and then it breaks the thing you learned, which is where the pleasure is. That the pleasure is in the violation is not a design flourish, it is the thesis of the field the piece comes from.

The source’s own construction is unusually precise, and worth reading slowly. It is not “detect the odd sound”. It is: into a multitude of sounds that fit a given life-situation and carry narrative content, we scatter sounds that do not belong. Three requirements, all load-bearing — the scene must be plausible, it must have narrative shape, and the intruders must be scattered rather than presented. That is a signal-detection task wearing a short story.

What makes it more than a lab paradigm is that “does not belong” is culturally constructed. A muezzin’s call belongs in one morning and is an intruder in another. The intruders can therefore be graded along a semantic distance rather than a physical one, and the player’s false alarms become a portrait. What a person believes does not belong in a kitchen, a tram, a schoolyard is one of the more revealing things you can learn about them without asking a question. That is where the strand stops being a test and becomes an artwork — and it is why the panel above counts false alarms without ever calling them mistakes.

This piece also has a second life: paired with the biofeedback rig it is the project’s cleanest cross-strand experiment. A measured expectation, a measured violation, and a measured body — skin conductance answering one to five seconds later — all on one timeline.

the rest of the strand

1. M1 Betolakodó · The Intruder
   One button, eight to twelve minutes, and a continuous, entirely plausible scene: a kitchen in the morning, a night tram, a school corridor at break. Events accumulate. Press when something does not belong. Intruders are graded by semantic distance rather than by loudness — a kettle in a kitchen (belongs) → a kettle on a tram (wrong place) → a kettle reversed (wrong time) → a kettle whose spectrum is faintly human-voiced (wrong category) — and are mixed at the same level as everything else, with nothing signposted. It is a properly constructed signal-detection experiment with an art-directed stimulus set, and the by-product is more interesting than the result: exhibited, the aggregated false alarms across a city’s visitors — the sounds people here believe do not belong in a kitchen — are the piece. measures — d′ and criterion per intruder class · reaction time by class · the false-alarm profile
2. M3 Hamis emlék · False Memory
   Two buttons — heard it, didn’t — and six minutes. Twelve sounds are presented as a scene; later, sounds are played one at a time and you answer whether each was there. Some probes are strong semantic associates that never occurred: a lighter’s click when only an inhale was heard, a kettle’s click when only the boil was there. The question is never asked in words — the probe is the sound itself — so the game is fully language-independent and runs identically in any country with no translation. It delivers a real, established cognitive measure as a two-button game with no instructions and no reading, which makes it usable with young children, with people who have limited literacy, and across languages. That combination is rarer than it sounds. measures — false-recognition rate for associates against unrelated lures · confidence, from press latency