← Real Sight Reader

Chain the Keys, Not the Tune

By Mayank Sanganeria

"Learn it in all 12 keys" is standard advice for a working jazz repertoire, and the standard way people follow it is the worst way to follow it: take one tune, run it through Db, then D, then Eb, then E, all the way around the circle, before touching anything else. It's exhausting, it front-loads twelve key changes onto material you've barely learned in one key, and most of what you drill that way doesn't stick. There's a better order hiding in plain sight. Don't take one tune through all twelve keys. Chain your keys across different tunes, one new key per tune, and let your key vocabulary build itself gradually, by association.

The grind, and why it doesn't stick

The 12-key drill exists for a real reason: a working player needs to be able to play a tune wherever a singer's voice sits or a horn player calls it, not just in the key on the lead sheet. That part of the advice is correct. What's not correct is doing it as a single marathon against a single tune. Twelve keys back to back means twelve almost-identical passes through material you're still learning, where the only thing changing is the key — so the transposing itself becomes the whole task, and the tune's actual content (the melody's shape, where the changes really move, what the bridge is doing) gets buried under mechanical relabeling. By key seven or eight the drill has stopped being about the tune and started being about endurance.

It's also a bad match for how keys actually get learned. A key isn't really "learned" the first time you play something in it — it's learned by being revisited, under slightly different content, until the shapes in that key stop needing to be worked out and start being recognized. Twelve keys in one sitting gives every key exactly one exposure, then moves on and doesn't come back. There's no second encounter to consolidate anything, and by the time you'd naturally want one — a week later, on a different tune — the drill has already moved past that key for good.

Chain instead of grind

The fix doesn't require doing less transposing. It requires sequencing it differently. Every time you take on a new tune, put it in a key you've already used for a different tune — not a fresh key, and not the tune's "default" key from the lead sheet. That one constraint changes everything about how the key gets practiced. Instead of twelve keys tied to one tune, you get one key showing up across several tunes, each time with different melodic shapes and different harmonic motion sitting on top of the same key signature. The key gets revisited on purpose, spaced out over real time, carrying different content each time — which is exactly the second encounter the grind never gives you.

Do this consistently and something else happens almost for free: your key vocabulary grows tune by tune instead of arriving all at once. You're never asked to hold twelve unfamiliar keys in your head for one piece of material. You're asked to hold one or two keys you already half-know, applied to something new. The chain does the same job the 12-key drill was trying to do — genuine fluency across the keys you actually play in — without the overload that makes the grind version so easy to abandon after two or three tunes.

The worked example

Here's a chain, five tunes deep, starting from nothing:

Read across it and the pattern is the whole method. Row 1 plants Fm under All the Things You Are. Row 2 doesn't start a new tune in a new key — it puts My Favorite Things partly back in that same Fm, so Fm gets a second, different piece of music sitting on it, and it also opens Em, a key with no prior tune attached yet. Row 3 goes back to All the Things You Are — a tune you already know — but now in Em, the key that arrived one row earlier from a different tune entirely. Row 4 takes Em further (into Summertime) and opens Dm. Row 5 closes the loop by bringing My Favorite Things into Dm. Every row after the first reuses a key from a row above it, but never reuses it on the same tune it came from. No tune ever gets a marathon. No key ever gets exactly one exposure and then abandonment. And it keeps going — the next row picks up Dm or Em for a third tune, or opens a new key from whichever tune is currently active, and the chain gets longer and denser the same way every time.

An even better version: anchor everything in C

The chain solves the sequencing problem — how to order the practice so it actually sticks — but it leaves something else unsolved: every transposition in the chain is still a one-off. Moving All the Things You Are from Fm to Em is its own mental move, unrelated to moving Summertime from Dm to Em. Nothing about doing one makes the next one easier, because there's no shared reference point between any two keys in the chain.

There's a fix that costs almost nothing extra: learn every tune in its original key and in C, and treat C as the fixed point you pivot through whenever you move a tune anywhere else. C has no key signature to fight — no sharps, no flats — so it's the cleanest possible view of what a tune's changes and melody actually do, stripped of the specific key it happened to arrive in. Learning a tune in C isn't really learning a "13th key" the way learning it in Db would be; it's closer to learning the tune itself, with the key-signature noise turned off.

Once a tune has a C version, transposing it anywhere stops being a direct, one-off jump and becomes two much smaller moves through a key you already know cold: original key → C, then C → wherever you actually need it. It's the same trick as routing every currency conversion through dollars — you're not memorizing a separate direct route between every pair of keys a tune might ever need, you're memorizing one route in and one route out, both anchored on the same hub. Combine this with chaining and the two ideas reinforce each other: the chain still governs which key a new tune claims from the growing pool, but C sits underneath every tune in it as the place you go to check your own work, or to work out a fresh key you haven't chained to yet.

Why this works better than grinding one tune alone

This is a specific case of a much more general finding in skill-acquisition research: mixing related-but-distinct material together — interleaving it — rather than practicing one thing in a solid, repeated block, produces worse-looking performance in the moment (it feels harder, slower, less fluent while you're doing it) but noticeably better retention and transfer once you test it later, away from the practice session. Blocked practice, one tune through twelve keys in a row, is the comfortable version: pattern-matching against the pass you just did, which feels like progress but is largely riding on short-term memory of the previous repetition. Interleaved practice forces actual retrieval — you have to reconstruct the key each time from a colder start, because something else came in between — and retrieval under those conditions is what tends to stick.

Chaining is that principle applied to exactly this problem. Keys aren't drilled in an isolated block; they're spaced across different tunes and different sessions, and every return to a key forces you to actually reconstruct it rather than coast on the pass from ten seconds ago. The price is the same as any interleaved scheme — it feels less smooth than the grind while you're doing it, especially early on, when your chain is only two or three tunes long and every new row genuinely is a cold start. That discomfort is the method working, not a sign it's failing.

Build your own chain

Start with one tune in one key — whatever key you already know it in is fine. From there, the only rule that matters: a new tune claims a key already used by a different tune. A simple grid keeps the constraint visible while you're planning it, before you ever sit down at the piano:

Fill the "Key(s)" column with one or two keys per row, same as the worked example. Fill "Reused from" honestly as you go — if a row doesn't actually reuse an earlier key, it's a grind entry, not a chain entry, and it's fine, just notice it. Check off "In C too?" once you've also worked the tune through in C — that's the hub version, the one every future transposition of that tune routes through, so it's worth having before you chain the tune's key any further. Two practical rules keep the chain from turning back into a marathon: work a tune in its new key until it's genuinely playable, not perfect, before moving to the next row — the chain depends on returning to keys later, not on finishing each one now. And let old rows resurface. Nothing says a key or a tune only gets used once; the whole point is that Fm from row 1 is fair game again at row 12, on yet another tune, for another real encounter.

What this doesn't replace

Chaining is a sequencing fix, not a substitute for the underlying work. If you can't yet find the ii-V-I in a key at all, no amount of clever ordering makes that key easier to reach for the first time — you still need the harmonic and technical groundwork, in some key, before there's anything to chain. And a chain with only one or two tunes in it doesn't yet have the spacing that makes the method work; the benefit is cumulative, and it shows up after several tunes are in rotation, not on the first row. This also isn't an argument that the full 12-key drill is worthless — there's real value in eventually being able to call a tune in any key on demand, and some players do want that for one or two staple tunes. The claim is narrower: as the default way to build a working key vocabulary across a growing repertoire, chaining beats grinding, because it spaces the practice a grind can't.

The same idea about ordering practice for retention rather than for a smooth feel in the moment is the subject of the hand-independence essay from a completely different angle — timing instead of key vocabulary, but the same lesson that the version that feels hardest in the room is often the one that's actually working.

Real Sight Reader's own library ends up doing a version of this by accident: real repertoire isn't written in one convenient key, so working through pieces as they're actually notated puts you through several different keys without having to engineer the rotation by hand — which is a reasonable way to start a chain if you don't already have one running.

← More long-form pieces