Interval ear training that remembers what you miss

Because a fixed root teaches the wrong thing. Drill enough fifths from middle C and what improves is your recognition of the specific pair C-G, which does not transfer to the same fifth played from E flat — and the improvement feels real right up until the moment somebody plays it in another key. Moving the root through a full octave forces the interval itself to be the thing you learn. It also stops the session from quietly training absolute pitch, which is a different ability with its own page here.

Because they are a different judgment wearing the same name. When two notes play in turn, you compare the second against a memory of the first, and the answer is a distance. When they play together, there is no sequence to compare — you are hearing a single sonority, and what identifies it is its quality: the beating of a mistuned pair, the hollowness of a fifth, the tension of a tritone. Many people who name melodic intervals fluently are near chance on harmonic ones for months, and the fix is not more melodic practice. Tick the ones you miss into the drill and hear them stacked until each has a character rather than a size.

Because they run out exactly where the difficulty starts. The familiar mnemonics — a film theme for a fifth, a folk tune for a fourth — are all ascending and all from the root, so they cover perhaps a third of what you actually need. There is no well-known tune that starts with a descending minor sixth, and none at all for a harmonic anything, which is why the mnemonic method produces students who are quick on rising intervals and helpless the moment a line goes down. Nothing here forbids using them; the page simply refuses to test only the third of the material they cover, which is why direction is drawn at random on every trial.

Two things, both counts. A per-interval running total of correct and attempted, and one line per finished session holding the date, the totals and the widest answer set you reached — twenty sessions, after which the oldest falls off. There is no record of which intervals played in what order, no timing, and nothing that could identify you. It lives in this browser's local storage for this site alone, so it is not on your other devices and not on any server. The delete button under the history clears it immediately, and clearing this site's data does the same.

So that you can see your own odds. If all twelve buttons were on screen while the ladder was only playing fifths and octaves, a guess would look like a one-in-twelve shot when it was really a coin flip, and both you and the summary at the end would be reading the run wrong. Showing exactly the set in play makes the guessing rate visible at every moment, and it is why the summary computes chance per trial from the set that was actually on screen instead of quoting one number for the session.

Shorter and more often beats long and occasional, and the reason is specific to this material rather than general advice. Interval recognition consolidates between sessions rather than within one, so twenty trials on six days does more than a hundred and twenty in one sitting — and the second half of a long session is where the confusion table fills up with mistakes made through fatigue rather than through not knowing, which then pollutes the thing you were going to work from. A useful stopping rule is the moment you notice yourself answering by elimination instead of by hearing.

Because it has no neighbors to be told apart from until the rest of the set exists. A tritone is identified largely by what it is not — not the fourth below it, not the fifth above it — so introducing it into a two-interval set teaches nothing except that the odd-sounding one is the new button. By the time it arrives, both of its neighbors are in play and the discrimination is real. It is also the interval most reliably named as something else, which is why the confusion column beside it is usually the most informative row in the whole table.