Of the five sections in the MusIQ test, tonal memory is the one people talk about after they finish, because it's the only one where the source material plays once and never again — whatever you didn't catch in those few seconds is simply gone. I built it that way on purpose: that exact pressure — "I only get to hear this once" — maps onto something musicians do constantly, from learning a part off a record to taking dictation in an ear-training class. This page covers what the section feels like, what it's actually measuring, and — since I wrote the scoring code myself — exactly how your score gets calculated, down to the formula.
You're presented with a "Play Sequence" button and an empty grid. Nothing else. You press play, a short melody sounds, and while it's playing the grid is inert — you can't pre-click anything, you just listen. Then the melody ends, the button goes gray and stays that way, and you're staring at an empty grid with only your memory of what you just heard to go on.
The grid is simple once you understand it: one column per note in the melody, each column holding twelve stacked boxes for twelve chromatic pitch options. You click a box in the first column for your guess at note one, a box in the second column for note two, and so on. Every box you click plays that pitch out loud immediately, which matters more than it sounds like it should — you're auditioning candidates against whatever's still echoing in your head, not answering blind. Once every column has a guess, a Submit button appears.
What's being isolated here is short-term auditory memory for a sequence of pitches — the ability to hear a short run of notes and hold their order and identity active in mind for a few seconds, not just their general up-and-down shape. Psychologists usually call this kind of capacity working memory — the mental scratchpad you use to hold information active while you do something with it. It's not a new thing to try to measure, either: Carl Seashore's 1919 Measures of Musical Talent, one of the earliest attempts to formally test musical ability, included a tonal memory subtest of its own.
For a working musician, this is the ability behind learning a bass line off a recording without sheet music, hearing a horn line once at a session and nailing it on the second take, or taking melodic dictation in a theory class. None of that requires perfect pitch or years of theory — it requires catching a sequence of notes on one listen and not letting it slip before you can act on it. If you want to see how this connects to memory more broadly — why some melodies stick for days while others vanish instantly — that's covered in Music and Memory: Why Melodies Stick.
This is the part nobody else can tell you, because it's our code — no hiding behind vague language about "advanced algorithms."
The section runs five melodies of increasing length: one note, then two, three, four, and five — fifteen notes total. Each melody is generated fresh for your session by drawing notes at random, without repeats, from a fixed pool of twelve chromatic pitches spanning one octave. They aren't pulled from a scale or written to resolve anywhere — no tonic to lean on, no "that note wants to go up" intuition to bail you out. You're memorizing bare pitch positions in a genuinely random order, which is what makes it a memory test and not a music-theory quiz. Each note rings for a full second and they run back-to-back with no gaps, so the five-note melody is about five seconds of continuous sound.
Grading happens one time slot at a time, not as a "did you get the right notes somewhere" comparison. Your guess in column one is checked against the correct pitch for column one, your guess in column two against column two, and so on — there's no credit for placing the right pitch in the wrong slot, and no separate credit for getting the general shape of the melody right if the actual pitches are off. What gets measured is how many steps apart, on that twelve-box pitch grid, each of your guesses was from the correct answer. Those distances get added up across the whole melody into one error total. Ten minus that error total is your raw score for the question, and it can't go below zero — badly miss enough notes and the question is worth nothing, but it never goes negative.
Here's the detail that actually matters for understanding your score: every question is worth the same maximum — four points, whether it's the one-note melody or the five-note one. The one-note question forgives up to nine grid-steps of error and still pays out something, while the five-note question spreads that same budget across five guesses — being off by just two steps on each of five notes zeroes the question out just as fast as one badly-missed guess would. The later questions aren't scored more generously to compensate for being harder to remember, which is why they feel disproportionately punishing. One more quirk: the grid doesn't wrap around, so a note at the very bottom and one at the very top score as maximally far apart even though, on a keyboard, they might sit right next to each other across the octave — a side effect of the grid's layout, not a judgment on how wrong you actually were.
Two more mechanics worth knowing. The original melody plays exactly once — the button that triggers it disables itself permanently the moment you use it, so there's no secret second listen hiding anywhere. And once you've placed your guesses, you can replay your own submitted sequence back to yourself as many times as you want before hitting Submit, a genuinely useful way to check your answer against what you remember before committing. After you submit, the grid shows the right answer in green and your wrong picks in white, right there before the next melody loads.
Each question caps at four points, so the section tops out at twenty — the same cap every section of the test uses. Your final MusIQ score is simply those five section scores added together, capped at a hundred. Tonal memory isn't weighted up or down against the others; it's a flat one-fifth of your total score.
Less than you'd think, and also more than you'd hope. Tonal memory isn't spectrally demanding the way counting voices is — you're hearing one note at a time, not pulling apart several tones fused into a chord, so you don't need golden-ear frequency resolution here. But because the melody plays exactly once, anything that degrades that single playthrough costs you with no way to get it back. A Bluetooth dropout, a stutter, a dog barking in the next room — on any other section you'd shrug it off and catch the next trial. Here, if it happens during your one listen, that information is simply gone. A quiet room and a stable, low-latency connection — wired headphones remove the variable entirely — give the section the fairest shot at measuring your memory, not your Bluetooth chipset.
A rough tonal memory score doesn't mean you're "unmusical," and it isn't a diagnosis of anything — this is general information about one narrow perceptual skill, not a medical or clinical assessment, and MusIQ isn't built to diagnose memory problems of any kind. Auditory sequence memory is also just a different skill than pitch accuracy or rhythm — plenty of excellent instrumentalists read everything off a page and never had to train this particular muscle.
That said, it's trainable, and the training is concrete. Chunking helps enormously — instead of holding five individual notes, hear the melody as two or three small groups, the way you'd remember a phone number in blocks instead of seven separate digits. Singing or humming back what you just heard, even silently, converts a fleeting auditory impression into a motor memory that sticks around longer. And try encoding contour before exact pitch — lock in the up-and-down shape first (up, up, down, way down), then go back and pin down the exact distances, rather than nailing everything in one pass. The immediate green/white feedback MusIQ gives after each question is genuinely useful here — treat a few run-throughs like flashcards and notice which kinds of errors you keep making. For a fuller set of drills, the ear training guide is the practical hub.
No, and this trips people up constantly. Perfect pitch is the ability to name or produce a note without any reference tone — a fixed, absolute sense of pitch that very few people have. Tonal memory is relative and short-term: can you hold a sequence of notes in your head long enough to reproduce their pattern, using the grid as your reference rather than your own internal pitch sense. You don't need perfect pitch to do well here, and having perfect pitch doesn't automatically make you good at it either, since it's really a memory task wearing a pitch-test costume. More on that distinction over on Perfect Pitch: What It Is, Who Has It, and Can You Learn It?
It's deliberate, not a bug — unlimited replays would mostly measure patience, not memory. A teacher playing a phrase once, a bandmate humming a line at rehearsal: real ear-training rarely comes with unlimited replays either.
Not on its own. The five sections run as one continuous test, so retaking the full thing is currently the only way to get another crack at this score.
Since each of the five questions caps at four points, anything in the high teens out of twenty means you caught nearly every note in nearly every melody, including the five-note one at the end. A modest score on your first attempt is normal — this is one of the more unforgiving sections because of the one-play rule, and it rewards practice noticeably faster than most people expect.
They're generated fresh each time, drawn randomly from that twelve-note pool with no repeated pitches inside a single melody, so you won't get the same five melodies twice. Only the underlying skill transfers between attempts, not a memorized answer key.