Which Voices Should Be Heard in Polyphonic Music?

Geoffrey Chen

Building a Piano Piece · Season Three: From Score to Interpretation: Building a Performance of Your Own · Article Six

One of the most familiar instructions in contrapuntal practice is “Every voice must be heard.” If that means that all voices should be equally loud, the result is usually not clarity but congestion. Hearing must organise simultaneous tones into distinguishable streams while integrating them into a whole. Performance must provide enough cues for separation without sacrificing harmony, metre and shared direction.

The question is therefore neither “Which voice is always loudest?” nor “Must every subject entry jump into the foreground?” A better question is: what relationship does a listener need to recognise now? Which parts carry material, direction, bass foundation or harmonic completion? When should the focus move?

Audible does not mean equally prominent

Voices serve different functions. A subject entry must be identifiable, yet a countersubject, continuing part and bass may simultaneously provide indispensable direction. If the subject always overwhelms everything by a large margin, successive entries become a sequence of emphases and harmony becomes hollow. If every line has identical onset and weight, the listener struggles to form separate streams.

Auditory research describes polyphonic perception as an interaction of segregation and integration. Pitch proximity, timbre, synchrony, common motion and attention can affect whether sounds are heard as separate streams. Research on auditory streaming in polyphonic music and experimental work on segregation and integration in multipart listening cannot prescribe the balance of a particular fugue. It does show why loudness is only one cue among several.

Learn the grammar of each voice

Playing parts separately is not preparation for eventually combining four soloists. Separate work should answer questions. Where does this voice breathe? Which notes form its motivic skeleton? How does dissonance resolve? How does it cross the metre? If an inner part has direction only when played alone but becomes undifferentiated filling when combined, the performer has not yet preserved its articulatory and temporal identity.

Next practise pairs: top and bass, subject and countersubject, neighbouring inner voices. Ask one relational question each time. Who initiates an imitation? Which voice completes the cadence? Where do the lines cross? Establishing two-part relations before adding a third is more effective than repeatedly making each part louder within the full texture.

Create a voice-focus plan

Make four columns for a passage.

The first lists material events: subject, countersubject, sequence, pedal point, cadence or free part. The second identifies current focus: which line must the listener most readily identify now. The third names a supporting voice that must retain direction without competing. The fourth specifies the differentiating means: loudness, articulation, duration, slight onset difference, pedal, colour image or distribution of weight.

The plan should move. An initial subject may need a clearly established contour. Later entries may be recognised through register and articulation without each becoming louder. Near an important cadence, bass direction may deserve more foreground than another statement of the subject. Counterpoint is neither a democratic vote nor a succession of soloists seizing the stage. Functions continually reorganise.

Differentiate with more than loudness

Loudness contrast is easy to make and easy to overuse. Try four alternatives.

First, use articulatory contrast: perhaps the subject is more connected and another part more clearly separated, or the reverse if text and style support it. Second, use release contrast: let supporting material release more cleanly at a boundary, leaving perceptual room for an entry. Third, use duration and connection to preserve one line’s continuity without markedly increasing its level. Fourth, test limited onset displacement. Natural movement and historical performance can involve small asynchronies, but deliberately delaying every subject entry becomes an effect. It must survive recorded comparison.

Pedal is particularly consequential. It can help a line connect where fingers cannot, but it also combines releases that would otherwise separate streams. Establish voices without pedal, then add only what harmonic resonance and necessary connection require.

A six-step contrapuntal experiment

  1. Mark actual voices in different colours rather than dividing by right and left hand.
  2. Sing and play each part, indicating direction and cadence.
  3. Practise all relevant pairs and record the relationship being trained.
  4. Combine them into a balanced baseline without deliberately spotlighting the subject.
  5. Follow the focus plan to create a loudness-led version and an articulation-led version.
  6. Record at an ordinary listening distance and ask a listener which line can be followed and where it disappears.

Add a focus-removal test. Return the foreground part towards balanced loudness while retaining its articulation and direction. If it remains audible, the former level difference may have been excessive. Use the reverse test temporarily to exaggerate an inner part, learn its relation, and then return it to proportion.

Hands are not voices

Piano textures often assign two lines to one hand or pass one voice across both. Practice organised only by hands encourages the ear to confuse notated voice with motor unit. Extract a cross-hand line and redesign fingering and weight transfer so that it does not break where the responsible hand changes.

Some controlled research examines inter-hand balance and melodic lead in pianists, reporting relationships between proficiency and the control used to project a target melody over accompaniment. That experimental evidence is not a law for Bach performance. It nevertheless supports the practical point that balance is trained differential control, not something guaranteed merely because the player imagines a voice internally.

Attention must be rehearsed as well

A voice plan is not only a distribution of finger dynamics. It is a route for the performer’s attention. If a complete performance watches only the subject, changes of bass, countersubject and cadential preparation lose agency. Make three attention runs. Follow only the bass in the first, an important inner voice in the second, and move at predetermined points in the third. Try to preserve the sounding design and notice whether a change of attention alone transforms it. A large change reveals that control of some parts is not yet stable.

Then practise dual monitoring: one voice is focus and another is reference. Follow the subject contour while checking bass arrival, or hear an inner resolution while retaining the large beat. The aim is not conscious note-by-note supervision of every part. Establish a few decisive checkpoints. Attention available in performance is finite; an interpretation must decide which relations deserve it.

A listener will not always follow the line the pianist designates. After a subject has been established clearly, emphasis can recede and memory can carry recognition. If every entry re-teaches the identity of the subject, rhetoric becomes dense. Genuine clarity permits attention to move and leaves room for the listener’s selection.

Decide whether the result works

Success does not mean that a listener can transcribe every note of every line. Test three outcomes. Can important entries be identified? Do non-focused voices retain syntax rather than becoming noise? Does the whole retain a shared pulse and harmonic direction? If greater separation makes the music sound like unrelated strands, integration has been lost. If the whole is rich but no relationship can be followed, separation is insufficient.

The distinctive challenge of polyphony is that no fixed melody-to-accompaniment proportion can govern the entire work. Attention and sounding responsibility must be redistributed continuously.

Every voice needs life, but not every voice needs the foreground at every moment. Contrapuntal clarity depends on mobile hierarchy, not equal loudness.

Primary sources and further reading

Continue reading: Building a Piano Piece series page

If you would like to bring these ideas about listening, understanding, and practice to the keyboard, you might try ScoreFlow, an app I developed to make score reading and daily practice flow more naturally together.

Discover more from Geoffrey Chen

Subscribe to get the latest posts sent to your email.