A soloist performing a lengthy concerto entirely from memory, with no printed score in front of them at all, is drawing on several genuinely distinct kinds of memory working together, not a single unified skill called "memorisation." Musicians and music education researchers generally distinguish at least three separate layers: structural memory, an understanding of the piece's overall form and how its sections relate to each other; muscle memory, the physical, largely automatic sense of where the hands need to move; and aural memory, an internal sense of how the music should actually sound at each point, which a performer can check their own playing against in real time.
Muscle memory alone is fast but genuinely fragile
Muscle memory, the physical, largely automatic sense of the correct hand and finger movements built up through extensive repeated practice, lets a well-rehearsed passage be played quickly and fluently without requiring the performer to consciously think through each individual movement. But muscle memory alone is a notoriously fragile foundation to rely on entirely — a performer relying purely on physical muscle memory, without a genuine structural or aural understanding of the music backing it up, is often unable to recover cleanly from even a small memory slip, since they may have no clear independent sense of exactly where in the piece they actually are once the automatic physical sequence has been interrupted, with nothing else to fall back on to reorient themselves.
Structural and aural memory provide the independent backup muscle memory lacks
Structural memory — a genuine conscious understanding of the piece's overall architecture, its distinct sections, and how they connect and develop into each other — gives a performer an independent way to know where they are within a piece even if physical muscle memory alone momentarily falters, letting them consciously navigate back to a secure, known point rather than being entirely dependent on an uninterrupted physical sequence continuing unbroken. Aural memory, a performer's genuine internal sense of exactly how the piece should sound at any given point, offers yet another independent check, letting a performer immediately notice when what they're actually playing has started to diverge from what it should sound like, well before a listener might otherwise notice anything is wrong. Music educators generally recommend deliberately cultivating all three memory layers together during practice, precisely because relying on any single one of them alone leaves a performance more vulnerable than having multiple independent, mutually reinforcing systems in place.
What we're still unsure about
The general distinction between structural, muscle and aural memory, and the real vulnerability of relying on muscle memory alone, are well supported by both music education research and generations of performers' own documented practical experience. What's less precisely settled is exactly how much deliberate practice time a performer should allocate to each specific memory layer for a given piece, and how that ideal balance might vary between different performers, instruments and musical styles — music education researchers continue to study which specific practice strategies most efficiently build robust structural and aural memory alongside muscle memory, without a single universally agreed formula for exactly how to divide practice time between them.
This sits inside Music Memorisation Strategies, one of seven topics in Performance, one of four domains in Music, one of seventeen subjects the app can quiz you on.