The Prediction

You are not moved by the chord. You are moved by the conversation between what you expected and what arrived.

humanitysociety

You have heard a piece of music you had not heard in years. Maybe in a supermarket, or from an open window, or from a car that passed too quickly. The first three notes arrived and something happened before the fourth. Your body knew what came next. Not your conscious mind — your body. A tightening somewhere between the chest and the throat. By the time the melody reached its first phrase boundary, the tightening had become something else. You were moved. You would say the music moved you.

That is not what happened.

In 2009, Petr Janata published a study in Cerebral Cortex that upended the common model of how music and memory interact. The prevailing assumption — Proust’s version, the one repeated in every explanation of why songs can make us cry without warning — is sequential: the music arrives, the music triggers a memory, the memory produces the emotion. Trigger, retrieval, response. Three stages.

Janata found something different. Using functional imaging while subjects listened to music from their past, he showed that the dorsal medial prefrontal cortex responds simultaneously to the degree of autobiographical salience experienced over the course of a musical excerpt AND to the real-time tonal movement of the music itself. These are not separate responses happening in sequence. They are the same neural population tracking both at once — integrating the music and its associations as a single moving event. Not trigger then retrieval. The music and the memory arriving together, continuously, as one thing in motion.

There is no moment where the music finishes arriving and the memory begins. They are the same process. The melody you haven’t heard in twelve years and the afternoon you first heard it are not two things, one calling the other. They are one neural event, unfolding in time, tracked by the same tissue.

In 2013, Valorie Salimpoor and Robert Zatorre published in Science what may be the most important finding about why music feels like anything at all.

They had subjects listen to novel music — pieces they had never heard before — while measuring neural activity. The question was simple: what predicts whether a piece of music will be experienced as rewarding? The answer was not activity in the auditory cortex. It was not activity in the nucleus accumbens alone. It was the functional connectivity between them. The degree to which the reward-evaluation system and the pattern-recognition system were communicating with each other predicted how much subjects would pay for the music they were hearing.

The implication is precise: musical reward is not a property of the sound. It is a property of the relationship between what your auditory cortex predicts and what arrives. The nucleus accumbens does not evaluate beauty. It evaluates whether you were right. Whether your prediction — generated from everything you have ever heard in a structure like this one — was confirmed, productively violated, or met in some way that your pattern-recognition system finds satisfying.

Pure novelty does not reward. Music with no learned pattern to predict from produces no prediction to satisfy. Pure repetition does not reward either — the prediction is trivially correct, the system has nothing to evaluate. Reward lives in the space between: a prediction system working well enough to have expectations, encountering a signal complex enough to test them.

In 2022, Krzysztof Basinski and David Quiroga-Martinez published a review in Neuroscience and Biobehavioral Reviews that added the architecture this mechanism operates through. The brain does not predict music at a single timescale. It monitors multiple temporal levels simultaneously: rapid local predictions at the note-to-note level, phrase-level predictions at an intermediate timescale, and form-level predictions that operate over minutes. These are not independent parallel tracks. They are hierarchical. The longer timescales actively shape expectations at the shorter ones, top-down.

This means that when you are four minutes into a symphony and the harmonic tension has been accumulating since the opening theme, your prediction about which note comes next is being constrained — below conscious awareness — by your prediction about when and how the form resolves. The macro shapes the micro. Your auditory cortex does not process each note in isolation. It processes each note as positioned within a phrase, within a section, within a form, within a history of forms like it that you have absorbed across a lifetime.

When the resolution arrives and you are moved, the reward is simultaneous at multiple levels. The correct note. The correct phrase close. The correct formal arrival. A multi-scale satisfaction, happening all at once. This is what the tightening in your chest actually is. Not one prediction confirmed but many, nested inside each other, all arriving at their destination in the same moment.

This means you are not one listener. You are the accumulation of every listening that shaped what you now predict. The person moved by the final chord of a piece has been calibrating prediction systems across thousands of formal resolutions, thousands of productive violations, thousands of moments where the long expectation was met or exceeded or redirected in a way that the system recognized as meaningful rather than random. The emotion is not a response to the sound. It is the record of a history encountering its present.

A child hears music differently than an adult — not because the child’s ears are different but because the child’s prediction system has less to predict from. Fewer patterns absorbed. Fewer violations catalogued. The architecture is the same; the accumulated material is thinner. An experienced listener is not someone with better taste. An experienced listener is someone whose prediction system operates at more levels simultaneously, with more context, generating more specific expectations that can be more precisely satisfied or more productively surprised.

This is also why a piece of music can move you at forty in a way it did not at twenty. Not because you understand it better. Because your prediction system is deeper. It has more to bring to the encounter. The same notes arrive at a richer apparatus. More levels of expectation are engaged. More simultaneous satisfactions become possible. You have not learned to appreciate the music. You have become a more complex instrument for it to play.

The music you heard in the supermarket. The three notes and then the fourth. The tightening.

What happened was this: a signal arrived at an auditory cortex that had been shaped by decades of listening into a prediction-generating system of extraordinary specificity. The system recognized the pattern. It generated expectations at multiple timescales simultaneously — what note next, what phrase shape, what formal trajectory. The nucleus accumbens began evaluating: were the predictions right? Were they productively wrong? Meanwhile, the dorsal medial prefrontal cortex integrated the incoming signal with its autobiographical associations — not after the music, not because of the music, but as the music, as part of the same moving event. The afternoon you first heard it was not triggered by the melody. It arrived with the melody, inseparable, a single neural event unfolding in time.

You were not moved by the music. You were moved by the conversation between what you expected and what arrived — at every temporal scale from the next note to the shape of the whole form — conducted by apparatus that has been learning what to expect since before you could speak.

Not what the music does to you. What your history of listening does with the music.

Sources

- Solen