Hear a third. Sing it back. Don’t confuse m3 with M3.

What’s happening.Thirds are the bones of every triad — every major chord stacks M3+m3, every minor chord stacks m3+M3. Untrained singers echo single notes well enough but routinely flatten a M3 into a m3 (or vice versa) because the auditory cortex hasn’t differentiated the two qualities yet. This drill trains the discrimination by forcing production: you can’t fake your way through with a glide because the2.0-second hold inside ±50¢ commits you.

Why mixing m3 and M3, why both directions. Pure-m3 drilling lets you build a single motor template; pure-M3 the same. Mixing them randomly forces the ear to choosewhich template fires — that’s the actual transferable skill. Adding ascending+descending randomization rules out direction-as-cue: you can’t infer the quality from which way the line moves. The interval has to live in your hearing, not your habit.

Why this is the capstone for two-note echoes. Once thirds are clean, P4 and P5 (fourth-fifth-echo) come quickly — wider intervals are easier to disambiguate because the gap is bigger. After that, scale fragments (major-pentachord, one-octave-scale-fragment) and arpeggios (tonic-dominant-arpeggio) all assume m3/M3 discrimination is settled. This is the smallest interval pair where quality matters — get this and the rest of interval ear-training accelerates.