I recently read a neuroscience study on imagined music. Ten epilepsy patients who already had electrodes placed on the surface of the brain for clinical monitoring listened to the beginning of familiar songs, silently imagined how the melodies continued, and then hummed what they had imagined. From the brain activity recorded during the silent phase, the researchers were able to identify relative pitch relationships and partially reconstruct the rising and falling contour of the melodies.
This does not mean that scientists can play back a complete piece of music from someone’s mind. The study mainly decoded relative pitch categories such as do, re and mi, not exact timbre, dynamics or a complete sound waveform. What it does show is that silently imagined music is more than an abstract sense of knowing a tune. It has measurable pitch relationships and temporal structure.
Inner music is not an external sound wave, but it is a real internal event. It is private as an experience, yet its neural organisation can be observed and partly decoded. Before any physical sound is produced, the direction of a melody, the relationships between pitches and their order in time are already present in the mind.
The study also helped me make sense of what has changed in my own piano practice. When I first learnt pieces, I was not playing without inner music. The music in my mind and the movements of my hands were broadly synchronised. The real issue was not whether inner music was present, but which relationships carried the memory.
My earlier playing relied heavily on horizontal links between movements. One finger movement triggered the next, one hand shape led into another, and repetition turned the sequence into a fluent motor chain. The music accompanied that chain, but if I stopped in the middle, the later movements often lost their point of entry. I could not simply resume where I had stopped. I had to return to the beginning and let the earlier movements lead me back through the piece.
I later realised that secure playing requires both horizontal and vertical connections. Inner music unfolds horizontally through time, and the movements of the hands also need continuity. At every moment, however, the imagined sound must also have a direct vertical connection with a position and movement on the keyboard.
Inner sound → physical position and movement
Each note needs this connection. In chords and polyphonic music, it must extend to harmonic structure, independent voices, hand shapes and coordinated movement. As these vertical sound-to-movement mappings unfold through time, they naturally produce a fluent horizontal sequence. The movements still connect with one another, but those connections are no longer the sole basis of memory.
Both approaches eventually produce what is commonly called muscle memory. The difference lies in how that memory is organised. A memory built mainly from one movement cueing the next resembles a route with only one entrance. If the route is interrupted, the later section becomes difficult to retrieve on its own. When each inner sound can directly evoke its physical location, many points in the piece become possible entrances. If I can hear the music clearly in my mind, my hands can recover their place. This allows me to begin from different bars and to resume quickly after a mistake.
Many music teachers encourage students to sing note names aloud while playing. I understand its usefulness at an early stage. It draws attention to pitch and can prevent the student from merely watching the keyboard and memorising movements. I do not, however, think solfège should remain a permanent link between inner music and the hands.
Solfège is a symbolic system. If every imagined sound must first be converted into do, re or mi, and only then converted into a keyboard position, it adds an intermediate layer that need not remain. Moving between movable-do and fixed-do systems can create another conversion task. Note names can serve as temporary scaffolding, but once pitch and movement can connect directly, dependence on that scaffolding should gradually disappear.
Naming a note is also different from hearing or humming it accurately. Naming requires a symbolic label; humming requires a clear internal pitch. I can sing some familiar melodic pieces in movable-do solfège, but with many others I cannot immediately name every note even though I can hum the melody accurately. The music is still clear. It simply has not been translated into note names.
Much piano music cannot, in any case, be fully represented by one singing voice. The piano can carry melody, bass, harmony and several independent voices at the same time, sometimes without a stable traditional key. A singer can produce only one pitch line at a time. A skilled pianist may be unable to sing every element of a complex work in solfège while still hearing the registers, voices and harmonic relationships with great precision. The essential ability is therefore not to vocalise everything, but to develop a finely differentiated mapping between inner sound and physical execution.
In my experience, memory for familiar music is often more durable than memory for specific movements. A melody can remain clear after decades, and a single wrong note — even one semitone away — can be immediately apparent. Fingering and keyboard positions, by contrast, are more easily lost. It therefore makes sense to let the more durable inner music regenerate the movement, instead of relying on a fading motor chain to recover the music.
This approach has another major benefit. It helps break the habit of looking down at the hands and using vision to navigate the keyboard. If every movement needs visual confirmation, the eyes are carrying part of the navigation and the motor memory is not yet independent.
Mature motor memory involves more than the fingers. Posture, shoulder and upper-arm movement, the direction of the forearm, adjustments of the wrist, the shape of the hand and the action of the fingers form one coordinated system. More precisely, this involves not only muscle memory but also proprioception and touch. Without looking, the body can sense how far the arm has travelled, the angle of the wrist and whether the fingers are contacting black or white keys. The repeating groups of two and three black keys also give the hands a tactile map of the keyboard.
The fuller relationship is therefore:
Inner music → whole-body and hand movement → keyboard position
Vision can still be used to calibrate a large leap or a difficult change of position, but it should not be the hands’ main navigation system. The important distinction is not whether I ever look at my hands, but whether the movement can continue when I do not.
Once inner music can directly evoke coordinated movement, and the body can locate the keyboard through proprioception and touch, playing without looking is no longer a separate trick. Sight-reading also improves because the eyes can remain on the score instead of repeatedly shifting between score and keyboard. The notation evokes inner sound, inner sound evokes movement, and the hands find their position through bodily awareness. Playing by ear becomes more direct as well, because the sound actually produced is continually compared with the sound anticipated internally.
Reliable piano memory is therefore not a choice between inner music and muscle memory, nor does it require rejecting solfège, singing or vision. The task is to put them in the right relationship. Solfège and visual guidance can support learning, but they should not remain indispensable intermediaries. Inner music has its horizontal continuity, bodily movement has its own horizontal continuity, and at every point the two remain connected vertically through precise sound-to-movement mappings.
The horizontal connections give the performance flow. The vertical connections keep it grounded. One produces continuity and speed; the other provides accuracy, stability, recovery and the freedom to re-enter the music at any point. However fluent the hands become, their movement should never lose its note-by-note connection with the music heard within.
Research references
Reconstruction of Imagined Melody with Relative Pitch Decoding in Electrocorticography
https://doi.org/10.1523/ENEURO.0289-25.2026
Songs You Imagine Can Be Decoded Directly From Your Brain
Discover more from Geoffrey Chen
Subscribe to get the latest posts sent to your email.