Why Game Music Is So Memorable: The Neuroscience
Updated August 5, 2026
The most repeated claim about game music, that it shows “40% higher spontaneous recall than film music” in a 2024 University of Sheffield study, has no source we can find. We searched for the paper, the preprint, the press release and the dataset, and came back with nothing. The real research is messier: in 657 British adults, memories cued by music showed greater episodic reliving than memories cued by television, while a diary study from one of the same researchers found music and food produce memories indistinguishable on vividness, valence and specificity. A music-evoked autobiographical memory is a personal episode that surfaces while a piece of music plays, and the best measured rate comes from Janata, Tomic and Rakowski, where 30% of song presentations evoked one.
Where did the “40% higher recall” number come from?
We cannot tell you, and that is the point. The claim carries a specific institution and year, which is the shape a number takes after enough blog posts have laundered it into looking load-bearing. Searching by institution, year, phrasing and comparison returns music degree listings at Sheffield and unrelated memory papers.
A real recent paper on game music and memory does exist, and it claims something far smaller. Tuuri, Koskela, Tissari and Vahlo collected 183 written “game music memories” and analysed the language people used, in Entertainment Computing (2025, volume 52, article 100883). Remembrances involving awareness of the self predicted stronger affective language. That is a finding about how people write about game music, not a recall percentage, and it makes no comparison with film.
If you see the 40% figure again, ask for the DOI. We could not produce one, and we tried.
What does the research actually show about music and memory?
The papers this article leans on, and what each one measured:
| Study | Published | Sample | Comparison | Finding |
|---|---|---|---|---|
| Janata, Tomic and Rakowski | Memory, 2007 | popular music corpus | songs against autobiographical recall | 30% of song presentations evoked a memory |
| Murty, DuBrow and Davachi | Journal of Neuroscience, 2015 | choice and fixed trials | chosen screens against assigned ones | better recognition 24 hours later after choice |
| Belfi, Karlan and Tranel | Memory, 2016 | songs and famous faces | music against faces | music-evoked memories carried more perceptual detail |
| Jakubowski, Finkel, Stewart and Müllensiefen | Psychology of Aesthetics, Creativity, and the Arts, 2017 | 3,000 survey participants | 100 earworm songs against 100 never named | common contours, unusual gradients, faster tempos |
| Simchy-Gross and Margulis | Music & Science, 2018 | 20 environmental sounds | repeated against unrepeated | repetition alone pushed ratings toward music |
| Jakubowski, Belfi and Eerola | Music Perception, 2021 | 657 British adults | music against television | music ahead on episodic reliving and personal significance |
| Murre | Royal Society Open Science, 2021 | 16 divers | matched against mismatched context | no context effect, the 1975 result did not replicate |
| Jakubowski and colleagues | British Journal of Psychology, 2023 | 78 participants | music against food | no difference in vividness, valence or specificity |
The foundational work is Janata, Tomic and Rakowski in Memory (2007, volume 15, pages 845 to 860). They played participants songs drawn from a large popular music corpus and found that, on average, 30% of song presentations evoked an autobiographical memory. Note how modest that is. Seven times out of ten, a song from your own era plays and no personal memory arrives.
Music does beat some cues: Belfi, Karlan and Tranel compared songs against famous faces in Memory (2016) and found music-evoked memories carried more perceptual detail. The closest thing to a game-versus-film comparison is Jakubowski, Belfi and Eerola in Music Perception (2021, volume 38, pages 435 to 455). Working with 657 participants representative of the British adult population on age, gender, income and education, they compared music-evoked to television-evoked memories. Music won on episodic reliving, personal significance and social content, and produced more positive and more intense emotions. Most of those differences survived in a subset of responses where the music and television cues were matched on familiarity, so familiarity alone does not explain it.
Is music actually special as a memory cue?
Here the popular understanding falls apart, and the counterevidence comes from the same lab. Jakubowski and colleagues ran a diary study with 78 participants in the British Journal of Psychology (2023), comparing music-evoked memories to food-evoked memories in everyday life.
Music triggered memories more often, produced a greater proportion of involuntary memories, and produced memories rated as more personally important. But on vividness, valence, specificity and age of the memory, music and food did not differ. The authors state it plainly: this “provides counterevidence to the idea that music is broadly ‘special’ as a retrieval cue.”
The summary is narrower than the internet version. Music is unusually good at ambushing you with a memory you did not go looking for. Once it arrives, a smell from a kitchen delivers it about as richly.
Why might playing encode music differently from listening?
This is where games have a real claim. The support for it comes from outside music research entirely, from work on what agency does to encoding.
Murty, DuBrow and Davachi published The Simple Act of Choosing Influences Declarative Memory in the Journal of Neuroscience (2015, volume 35, pages 6255 to 6264). Participants studied objects hidden behind occluder screens. In the choice condition they picked which screen to lift; in the fixed condition they were told which button to press. Recognition tested 24 hours later was significantly better for objects from choice trials.
The critical detail is what participants were not choosing. They chose screens, not objects, so their choices had no bearing on what they were remembering. The authors conclude that “the perceived sense of agency, independent of actual control over the content of memoranda, contributes to the mnemonic benefits of active learning.” Striatal activation during the choice cue, interacting trial by trial with right hippocampal activity, predicted which items were later remembered.
Apply that to a game. You do not choose the soundtrack. You choose the route, the fight, the door, the tenant. The music is the occluder screen, arriving during a long stretch of voluntary choosing. Nobody has run this with game soundtracks, so we are reasoning across a gap. But it predicts something testable: the effect should be weaker for cutscene music you watch than for level music you act under.
Do loops make music stick, or just wear it out?
Repetition is the biggest structural difference between a game score and a film score. Simchy-Gross and Margulis, in Music & Science (2018), took 20 environmental sounds, from bee buzzing to machine noise, and played each ten times. Listeners rated them from purely environmental to purely musical, and repetition alone pushed the ratings toward music. As Margulis put it: “The sounds themselves didn’t change, but something changed in the minds of the listeners to make them seem like music.”
The surprising part is the contrast with speech. In earlier speech-to-song work the illusion required exact repetition. For environmental sound it worked whether the clips ran in order or jumbled, because scrambling speech destroys meaning while scrambling a machine noise does not. Repetition does different things to different sounds, depending on what the sound was for in the first place.
On earworms, Jakubowski, Finkel, Stewart and Müllensiefen analysed 3,000 survey participants, comparing 100 frequently named earworm songs against 100 never named across 83 melodic features, in Psychology of Aesthetics, Creativity, and the Arts (2017, volume 11, pages 122 to 135). Earworm tunes had more common global melodic contours but less common average gradients between turning points, plus faster tempos. In plain terms: an ordinary overall shape with unusual leaps inside it. That describes a lot of chiptune-era writing, where a line had to survive three channels of square wave.
One assumption worth correcting: most earworms are not unpleasant. Williamson, Liikkanen, Jakubowski and Stewart surveyed 11,833 Finnish and 5,989 English respondents for PLoS ONE (2014) and reported that most involuntary musical imagery episodes are not bothersome.
How do composers write music that reacts to the player?
The techniques are older than most players assume. LucasArts filed on 25 November 1991 and was granted US Patent 5,315,057 on 24 May 1994, inventors Michael Z. Land and Peter N. McConnell. This is iMUSE. The patent describes a “composing decision tree”: the composer embeds conditional hook and marker messages at musically graceful points, so the sequencer holds until a suitable moment and then jumps to a transitional phrase. The music never cuts on the game event itself.
A small verified irony: Monkey Island 2 Special Edition, the Steam re-release of the first game to ship iMUSE, never mentions the system in its store description. Its only music line is “Re-mastered and re-recorded musical score using live instruments.”
Koji Kondo laid out the other tradition at GDC on 8 March 2007, in a talk covered by Game Developer and archived at the Internet Archive. He named three things he keeps in mind, rhythm, balance and interactivity, and said “Interactivity is the most important part of the interactive game sound track.” His example was vertical layering before the term existed: percussion tracks in Super Mario World that play or stay silent depending on whether the player is riding Yoshi, reused for Super Mario Sunshine more than a decade later. His description of the failure mode is the best short argument for adaptive scoring we have read: “If the music doesn’t reflect the rhythm of the game, it becomes background music that might as well be piped in from any other room.”
Does game music really bring back the room you played in?
People say this constantly: a track returns and with it the carpet, the CRT, the specific winter. The research on context-dependent memory is weaker than its reputation.
The canonical demonstration is Godden and Baddeley’s 1975 study in which divers learned word lists on land or underwater and recalled better when the environments matched. Jan Murre attempted a direct replication with 16 divers in Royal Society Open Science (2021, article 200724). It failed. He reports that “contrary to the original study, we did not find that recall in the same context where the words had been learned was better than recall in the other context.” The original effect was so large, a Cohen’s d of 2.708, that the replication had over 99.9% power. It still found nothing.
The room may well come back when the music plays. What wobbles is the tidy laboratory story people cite to explain it. What survives is the autobiographical literature: Janata’s 30%, the reminiscence bump, and those 183 game music memories, where the strongest emotional language appeared in memories involving awareness of the self. Calling that a controlled context-matching effect would be overclaiming.
What this means for Grandma Left Me a Building
Most of the adaptive toolkit above assumes threat or urgency somewhere in the design. The iMUSE patent’s own worked example is a fight resolving into victory or defeat hooks. Grandma Left Me a Building has no time pressure and is built to be planned at your own pace, so the standard toolkit of musical escalation has nothing to escalate toward. Nothing chases you down a corridor.
What the game does have is the other lever the research points at: continuous small choices. Which of 30+ tenants to drag into which room, when a trait-based interaction system decides whether two neighbours settle or clash. Which of the 20+ building upgrades to fund out of happiness-driven income. Whether the philosopher who answers every question with “but why?” belongs on your quietest floor. If the Murty result generalises, that stream of voluntary choosing is the condition under which incidental material encodes well. We have written about the choice architecture in how we designed Grandma Left Me a Building and the cast in designing tenants that feel alive.
We have nothing announced about this game’s audio, so nothing in this article describes it, and we are not going to invent a feature to fit an article we wrote. The game is unreleased. We have no data on whether anyone remembers a note of it, and will not until people have played it and enough time has passed for a memory to become a memory. Anyone promising that their unreleased soundtrack will lodge in your head is selling something.
Frequently asked questions
Is video game music objectively more memorable than film music?
No published study we could find makes that comparison directly. The nearest verified result compares music to television, not games to films, and found music produced greater episodic reliving and personal significance among 657 British adults. Any percentage claiming to compare game music recall to film music recall should arrive with a DOI attached.
Why do old chiptune melodies stick harder than modern orchestral scores?
Part of it is the reminiscence bump, documented in popular music by Krumhansl and Zupnick in Psychological Science (2013, volume 24, pages 2057 to 2068), in which the songs people respond to most strongly cluster around adolescence and early adulthood, so whichever era you were fourteen in holds a structural advantage. Part of it is melodic: sticky tunes combine common contours with unusual leaps between turning points, and hardware limits pushed composers toward exactly that. The two explanations are not in competition, and neither of them requires retro styling, a point we take further in why nostalgia works so well in video games.
Does looping music help memory, or is it just cheaper?
Repetition demonstrably changes perception. Playing an environmental sound ten times shifted listener ratings from “sound” toward “music,” and the effect held even when clips were jumbled. Whether looping improves long-term recall of a track has not been tested in a game context we could find, so the perceptual effect is established and the memory effect is inferred.
What is the difference between adaptive music and vertical layering?
Adaptive music is the general category: any score that changes in response to game state. Vertical layering is one method, muting and unmuting parallel tracks over the same underlying music, as Kondo did with the Yoshi percussion. Horizontal approaches jump between musical segments, which is what iMUSE’s hook and marker system was patented to do gracefully.
Key takeaways
- The Sheffield 40% figure has no traceable source. We searched by institution, year, phrasing and comparison and found no paper, preprint or press release. Treat it as unsourced.
- Music beats television as a memory cue, and ties with food. The 657-participant study put music ahead on episodic reliving; a 78-participant diary study found music and food indistinguishable on vividness, valence and specificity.
- Agency improves encoding even when you do not choose the content. Participants who picked which screen to lift remembered the objects behind them better 24 hours later.
- Earworm melodies have ordinary shapes and unusual leaps. Across 83 melodic features, earworm tunes had more common global contours, less common gradients between turning points, and faster tempos.
- The classic context-dependent memory study did not replicate. A 2021 attempt with 16 divers and over 99.9% power found no effect, so “the music brings back the room” rests on autobiographical research instead.
Grandma Left Me a Building is a cozy apartment management sim about restoring the building your grandmother left you, with 30+ odd tenants and no time pressure, and it is on Steam.
