Voice Leading

Voice leading is the way each individual line in a piece of music moves from one chord to the next — what it holds on to, what it gives up, and how far it travels — as distinct from which chords are being played at all. (also called part writing, British usage and most classroom teaching, where the four parts are "parts" rather than "voices")

Up to now a chord has been a single object: a stack of thirds with a name. That is the vertical reading, and it is the one a chord symbol gives you. There is a second reading of exactly the same notes. Take a fixed number of lines — four is the usual laboratory — give each one a note of the first chord, and then ask each line separately where it goes next. Nothing about the harmony changes. What changes is that you now have four short tunes instead of one sequence of blocks, and the four tunes are what a listener actually follows.

You can hear the difference in the two clips because the ear is built to track continuity. In the blocked version every note leaps the same distance at the same instant, so there is nothing to follow and the ear hears four unrelated thumps. In the four-part version the top note holds still through the first change, which makes the ear treat the two chords as one connected event rather than two. Continuity is a property of the lines, not of the chords, which is why the same four chords can sound stitched together or dropped on the floor.

The word "voice" is a historical accident that stuck. The rules of this chapter were worked out for singers, and the four standard parts still carry choral names — soprano, alto, tenor, bass, top to bottom. Nothing requires actual voices. A string quartet has four; a guitarist fingering a chord has as many as they have fingers down; a pianist has two hands and about ten. The reasoning is identical in every case.

Smooth voice leading is writing in which each part moves as little as the harmony allows: it holds any note the next chord also contains, and where it must move it moves to the nearest available note rather than to the one that happens to sit in the same fingering.

The rule has two halves and they are worth separating. The first half is hold the common tones: if the chord you are leaving and the chord you are going to share a note, give that note to a part and leave it alone. The second half is move everything else by step: of the notes that do have to move, send each one to the nearest member of the next chord instead of to the note in the same position in the stack. Between chords a third apart you will find two common tones and the rest of the work is trivial; between chords a step apart — FPlay to GPlay, the third change in the first clip — there is nothing to hold and every part has to move, which is exactly where the effort goes.

That is the default, not a law, and the reason for it is practical. Small moves are easier to sing, easier to play, and easier to follow. A line that leaps constantly costs the listener attention, so leaps become expensive and therefore expressive: save them and a single wide jump in an inner part becomes an event. Spend them everywhere and nothing registers.

The four names in the table describe the relationship between any two parts at any one change, and they are the vocabulary the next section needs. Note that they are relative: at a single change one pair of parts can be in contrary motion while another pair is oblique. You do not describe a whole passage as "contrary"; you describe what two named parts did between two named chords.

Parallel fifths are two parts that are a perfect fifth apart and then move to another perfect fifth in the same direction — and parallel octaves are the same thing an octave apart. In four-part common-practice writing both are avoided, deliberately and systematically, and almost nothing else in that tradition is policed as strictly. (also called consecutive fifths, British usage, and the older term in both traditions; consecutive octaves, British usage, for the same thing at the octave)

The reason is not that the sound is ugly. It is that the sound is too well blended. The octave and the perfect fifth are the two most consonant intervals there are: their overtones coincide almost completely, so the ear fuses them into one pitch with a thickened colour rather than hearing two. As long as the distance keeps changing, the ear can pull the parts apart again. Keep the distance fixed across a change and there is nothing left to pull apart — the two parts stop being two parts and collapse into one. In a texture whose entire premise is four independent lines, losing one of them for a bar is a structural failure, which is why the tradition treats it as a mistake rather than a preference.

That premise is not universal, and this is the part that classroom teaching routinely leaves out. Parallel fifths are a rule inside one tradition, not an acoustic fault. The earliest written polyphony in Europe, ninth-century organum, consists of parallel fifths and fourths from beginning to end, and it was the normal sound of church music for two centuries. The power chord — a root and its fifth and nothing else — is the foundation of hard rock, punk and metal precisely because moving it around in parallel gives you weight without harmonic colour. Debussy wrote strings of parallel fifths on purpose to get a sound that was deliberately not tonal. When a modern engraver or arranger uses them, the fusion is usually the point.

So the working question is never "are there parallel fifths here". It is "am I writing in a texture where independent lines are the point". If yes, check for them. If the texture is a guitar riff or a doubled bass line, the check does not apply and never did.

A bass line is the lowest part of a texture read horizontally — the tune the bottom notes make when you follow them across a progression instead of reading each chord downwards. It is the part a listener tracks most closely after the melody, and it is the one that most decides whether a sequence of chords sounds connected.

The lowest note does two jobs at once, and they can conflict. Vertically it decides the chord's inversion: put the root at the bottom and the chord sounds settled and final, put the third at the bottom and it sounds lighter and in transit, put the fifth at the bottom and it sounds unstable and asks to move on. Horizontally it is a line, and it wants what every line wants — to move by step, to have a shape, to arrive somewhere. Playing every chord in root position gives you the strongest vertical reading and, usually, the worst horizontal one: the bottom part becomes a series of leaps between roots.

Inversions are how the conflict is resolved. Because the inversion is decided by which note is at the bottom and nothing else, you can choose the bottom note to suit the line and let the inversion fall out of that choice. That is the whole of the second clip: three of its eight chords are turned over so that the bottom part descends a scale, C3Play down to D2Play, without once repeating a note. The harmony is untouched. Only the bottom of each chord moved, and the passage stopped sounding like eight events and started sounding like one. Both clips are keyboard settings rather than chorale ones, and throughout they double that bottom note one or two octaves above itself — the deliberate octave doubling the previous section listed among the uses this motion is actually put to, rather than the lapse a four-part exercise would call it.

The descending stepwise bass is common enough to have become a genre in itself. It is the shape of a Baroque lament, of Pachelbel's Canon, of "A Whiter Shade of Pale" and of a large fraction of the ballads written since. Its opposite — a bass that climbs by step — pushes rather than settles, which is why it turns up under a rising phrase and almost never under a final one.

A non-chord tone is a note that sounds against a chord it does not belong to: not a wrong note and not a change of harmony, but a note borrowed from the scale, or from outside it, that sits between two notes of the chord and resolves on to one of them. (also called non-harmonic tone, much North American theory writing, for the same thing)

It is defined entirely by what happens on either side of it. The note itself is not special — D5Play over a CPlay chord is only a passing tone because of how it was reached and how it was left. Approach it by step and leave it by step in the same direction and it is a passing tone. Approach it by step and go back where you came from and it is a neighbour. Hold it over from the previous chord so that it becomes a dissonance the moment the harmony changes, then step down out of it, and it is a suspension. The catalogue in the table classifies journeys, not pitches, which is why the same note can be three different things in three different bars.

What they are all for is motion. A line that touches only chord members has to leap, because the members of a triad are thirds apart; threading the scale notes between them turns a series of leaps into a line you can sing. The dissonance is not an unfortunate side effect — it is the engine. A note that does not belong wants to resolve, so the ear leans forward, and the moment of resolution is a small arrival. String enough of those together and you have a melody instead of an arpeggio.

Two of the six carry most of the expressive weight. The suspension is the one you can hear from across a room: because it lands on the beat, the dissonance arrives at the strongest moment in the bar and the resolution a beat later feels like a sigh. The appoggiatura is its more violent cousin — leapt into rather than held over, so the dissonance appears from nowhere — and it is what a songwriter reaches for on the emotional peak of a chorus.