On this page
Here is a hunch that started as one about myself. A lot of people who learned piano by ear, or from YouTube tutorials and falling-note videos, can play far better than they can read. They have real technique, real musical instinct, a repertoire they can perform from memory — and they still freeze in front of a page of sheet music the way a fluent speaker freezes in front of an alphabet they never learned.
If that is you, the useful news is that the research points somewhere more flattering than you expect. Reading is a separate skill you simply never practised, and the ear you spent years building is the harder half of it.
Can you play piano well without reading music?
Yes, and demonstrably so. Sight-reading and playing are separable skills. Lehmann and Ericsson found that solo specialists sight-read worse than pianists who had accumulated accompanying experience, despite comparable training. Zhukov and colleagues (2016) documented underdeveloped sight-reading in advanced conservatory pianists. Meinz and Hambrick put the finding in their title: deliberate practice is necessary but not sufficient to explain differences in piano sight-reading.
So a competent player who cannot read is not a contradiction and not a failure of discipline. It is the expected result of a route that never asked you to decode a staff.
What falling-note tutorials actually teach
Synthesia-style tutorials — coloured bars raining onto a keyboard diagram — are genuinely good at things. They get you playing something recognisable on day one, they show you which physical key to press, and they are patient and infinitely repeatable. Nobody should sneer at the thing that kept them at the instrument long enough to get good.
Synthesia does not claim otherwise, either. Its own description calls it a game that helps you learn to play using falling notes, treats sheet music as something you can switch on should you choose to, and positions itself as a companion to lessons rather than a replacement. The format is honest about what it is. The gap opens when it becomes the only tool.
Because what a falling bar transmits is which key and when, and that is close to the entire message. The clearest symptom is not bad technique — it is a hard ceiling on repertoire. One player on the Piano World forums put it better than any critic could:
“I cannot read sheet music, I learn from watching tutorials online. There is a specific song I’d like to learn but there is no tutorial out there.”
That is the real cost, and it is about access rather than talent. The public-domain repertoire runs to hundreds of thousands of scores, nearly all of it distributed as notation and none of it waiting for someone to film a tutorial first.
What does notation carry that a falling-note video cannot?
Almost everything, and this is worth being precise about. The MIDI file behind a tutorial does hold exact timings and velocities — but that file is not what you are given. What reaches you is a video. From bars sliding past a keyboard you can judge roughly which key and roughly when, by eye, at whatever frame rate you are watching. Exact timing is not readable from it. Dynamics are not rendered at all: nothing in the picture tells you this phrase is piano and that one swells to forte. Articulation, voicing, and how a note is spelled are simply absent.
There is a deeper difference in what kind of document each one is. A recording or a roll is descriptive: a report of how one performance happened to sound. A score is prescriptive: a blueprint for how the music should be made to sound, which deliberately leaves the details of any one performance to you. A falling-note video is neither — it is a picture of a performance, with most of the performance removed.
That blueprint encodes things a stream of bars structurally cannot. Which hand takes what, and which notes belong to which voice. Whether that black key is an A♯ or a B♭ — the same key under your finger, a different meaning in the harmony, and something MIDI cannot express at all, since it only knows key number 61. Where a phrase breathes. Which finger. And structural compression: repeats and da capo marks fold a five-minute piece onto two pages, an operation a piano roll has no equivalent for.
Take one small case. A crescendo is a single wedge under the staff. It names no note; it describes a continuous change across a whole measure — a shape, not an instruction. Here is a real one, from Bach's Air:
A falling-note render of that same measure gives you six rectangles and no indication they belong to one gesture. The wedge is the composer talking about intent, and it survives on paper for three hundred years.
Does playing by ear hurt your sight-reading?
No — the evidence points the other way. Mishra's meta-analysis of ninety-two studies found sight-reading associated most closely with improvisational skill, ear-training ability and technical ability, and concluded it behaves as a developable musical skill rather than mere visual decoding. McPherson's work suggests a student's sight-reading may be influenced by how well they play by ear. Robert Woody has argued for years that ear playing is a foundation for fluent reading rather than a detour from it.
The eye-tracking work explains why. Perra and colleagues (2024) tracked pianists across five conservatory levels and found that as expertise rises, tempo goes up while fixations get shorter and fewer and the eye-hand span widens. But give the same experts an atonal score and their advantage collapses. Fluent reading is not superior eyesight or faster decoding — it is prediction, built on knowing how tonal music tends to go.
Which is exactly the machinery an ear player has already built. The most charming evidence for this is an error. Sloboda documented the proofreader's effect in musicians: skilled readers silently “correct” misprints toward what the music ought to have said. The pianist Boris Goldovsky found a misprinted note in a Brahms score that accomplished readers had been playing past for years — the person who spotted it was the weakest reader in the room, because she was the only one actually reading the symbol rather than her expectations.
Your ear is the thing that will tell you instantly when you have misread a bar. Readers raised on the page frequently lack it.
How long does it take to learn to read sheet music?
Less than you think, if the practice is the right shape. Note recognition is a domain-specific perceptual skill, and it responds to targeted training in hours. Wong and Fang (2025) report that after around eight hours of computer-based practice, intermediate readers recognised short note sequences 44.1% faster, reaching levels comparable to experts. Eight hours is a fortnight of ten-minute sessions.
Two further findings explain why this works. Wong and Gauthier found experts identified four-note sequences in about 465 ms against novices' 1,281 ms — with no comparable advantage on letter strings, so it is trained musical perception, not general visual talent. And Burman and Booth watched perceptual span grow from four or five notes to eleven across twenty rehearsals, with the gap between skilled and less-skilled readers disappearing entirely.
The same review notes why an explainer video rarely gets you there: the psychological mechanisms behind fluent note recognition are largely not verbally explainable. Nobody can tell you how to see a chord as a shape. It has to be trained by graded exposure, which is a job for drilling rather than for a tutorial — and the paper recommends exactly that, on the grounds that it needs no teacher present and adapts its difficulty to the learner.
The honest counter-argument
Plenty of magnificent musicians never learned to read, and whole traditions — blues, much of jazz, folk music across the world — are transmitted ear to ear. When Arthurs and Petrini (2024) asked musicians what reading does for them, the answers were specific: it lets them play more music, more varied music, and play it with other people. The same musicians also said notation can inhibit expressive performance and improvisation. Both halves of that are real.
So reading is not a moral virtue, and it is not a requirement for making music. The claim here is narrower: it opens a door that is currently shut, and the door has most of the written music behind it.
Why the usual advice fails
Most people who decide to fix this try to read the pieces they already play. It is the obvious move and it is the wrong one. Reading at the edge of your technique means your hands are working too hard for your eyes to learn anything; you fall back on memory within two bars and conclude you are bad at reading.
Reading improves under conditions that feel almost insultingly easy: material well below your playing level, short sessions done often, without stopping to correct mistakes, and — the hard one — without looking at your hands. Which runs into a practical problem. Where do you find an endless supply of material that is easy enough, new every time so you cannot memorise it, and aimed at the notes you personally keep missing?
What we built for exactly this
PracticeClef is a browser instrument for pianists. Connect a MIDI keyboard and open a score: the notation follows your playing, notes light as you strike them, and your sustain pedal turns the page. Nothing to install, no account.
A sight-reading trainer that writes its own exercises
The sight-reading trainer generates exercises instead of drawing them from a book, so it never runs out and you cannot memorise it. It starts further back than most people expect. Level one is a single note — middle C, right hand — and looks like this:
Each of its twelve levels adds exactly one new thing: another pitch, the bass clef, the other hand, a shorter note value, a wider reach, two notes at once, an accidental, then both hands together. By level eight you are reading alternating hands in eighth notes:
Underneath is a spaced-repetition scheduler that treats every pitch on every staff as a separate skill, on the forgetting-curve logic flashcard software uses. Read G in the treble reliably but hesitate over the same G in the bass, and it quietly feeds you more bass G until the hesitation goes. New notes arrive a couple at a time, and only once the ones you have are holding up. That is the graded, adaptive, no-teacher- required exposure the perceptual research argues for.
A score that stays anchored to your hands
Wait mode holds the cursor until you play the written notes, so you cannot coast on memory. Find by playing works the other way round: start playing from anywhere in a piece and the score finds your place and follows you — which is the move an ear player makes naturally, turned into a way into the notation. Hover any note to see what it is, which key and clef it sits in, and how long it lasts.
Every piece is parsed from its source notation, checked, its key and mode derived, and graded against a published 1–9 difficulty rubric shown on the piece's own page — which turns “find something genuinely easy to read” from a guess into a query. There are 1,089 public-domain scores from 47 composers, all free to open, plus a theory library for when a score shows you a symbol you have not met.
And one more thing, nearly ready
We have been training a model that predicts hand and finger assignment directly from the score — which hand takes which note, and which finger — for the overwhelming majority of music that ships with no fingering marked. It is harder than it sounds: on the standard research benchmark, two professional pianists agree with each other on only about 71% of notes, so there is no single right answer, only better and worse ones. Ours is nearly finished, and it deserves its own post rather than a paragraph here. That post is coming shortly.
A protocol you can start tomorrow
If you take one thing from this: read music far too easy for you, for ten minutes, most days. Eight hours of that is a fortnight. The rest is detail, but the detail helps.
1. Read below your level, not at it
If you play grade 5 by ear, read grade 1. The embarrassment you feel about that is the exact reason this has not worked before.
2. Never stop to fix a mistake
Sight-reading is the skill of continuing. Stopping to correct trains a different and less useful habit. Wrong note, keep going.
3. Do not look at your hands
Find the keys by feel. Your ear will tell you when you missed — that is your advantage over someone who started on the page.
4. New material every session
The moment you recognise a piece it stops training reading and starts training memory. This is the one thing a generator is strictly better at than a book.
5. Keep playing by ear
Do not trade one skill for the other. The evidence says your ear is an asset here, and the players who read and hear are the ones who end up unstoppable.
Start with the sight-reading trainer if you want the material chosen for you and the difficulty handled automatically. Start in the library if you would rather pick something easy and real and play it badly on purpose. Either way, connect your keyboard — the docs cover MIDI setup — and give it ten minutes tomorrow.
A note on the evidence. That reading is a distinct trainable skill, that note recognition responds quickly to graded perceptual practice, that expert reading is prediction rather than decoding, and that ear-players hold abilities associated with reading well — each is supported by the work cited above. Two things are our reasoning rather than measured results. We found no controlled study showing that readers learn new pieces faster than ear-players, so the cost we describe is framed as access to repertoire, not speed. And the claim that a score is a better thing to learn from follows from what notation encodes, not from a trial comparing it against falling-note video. We think the case is strong. The trainer is, among other things, how we intend to find out.