JimM: This is a very complex subject. Let me see if I can add to the confusion.

Your piano has a number of sampled sounds hard-wired into its memory. The sample was created by playing a note and recording it. Better samples record it at several velocities, let it ring longer, record more notes without interpolating sounds (up to 88 for a piano) etc. The note is the full sound spectrum.

The piano's software tweaks these samples in a variety of ways to create the instruments you get when you press the appropriate button. Some of the tweaking can use up more than one note of polyphony. Layering is an example of this. The only way I know to tell if an instrument is layered is to read the manual or look at the programming. Other tweaks don't consume polyphony - they add reverb or other effects, change the sound's envelope, etc.

For a basic piano sound, the usual relationship is one note = two voices of polyphony because the sounds are sampled in stereo. So if you play a 4-note chord you'll use up 8 voices. If you hold the sustain pedal down and then play another 4-note chord you're now using 16 voices because the first 8 are still sounding. Get the idea?

So does a choir use lots of voices? Not necessarily. If they sampled, say, a full choir singing in stereo, then each note might only take 2 voices. But if they sampled soprano, alto, tenor and bass voices and then layered them for a "full" choir you might use 4x2=8 stereo voices per note.

Most of this has very little to do with the "advanced" technology of the piano or lack of it. Better instruments typically have bigger and better samples so the sound is richer, lasts longer, whatever, but the basic polyphony issue remains 1 note = two voices. Usually.

Note-stealing happens when the piano runs out of polyphony. Different manufacturers handle this differently. The easiest algorithm is "first in, first off" - that is, the oldest sound is cut off first. There are other ways of doing it that are more complicated and may take more processing power, so I guess that's a place where faster computers help. Modern note-stealing techniques are pretty good.

Bottom line: more polyphony is better; 64 notes is probably good enough for solo piano; complex orchestrations want more than that; note stealing works OK for most things.[/b]

SteveY: JimM addressed this subject well, but I suspect that there are still a few misconceptions. I'll try and shed some additional clarity on the subject:

Let's consider some terminology first. The type of sound that your keyboard creates (a piano for example) is called a "Patch" or an "Instrument". This "instrument" is made up of "tones". Tones are individual "samples" or recordings of acoustic instruments -- in this case a piano. In order to recreate the sound of an acoustic piano with any sense of realism, multiple recordings/samples are used. These samples are typically layered (as many as four deep) in order to create a realistic piano sound. It's even common to trigger a different set of tones depending on how hard a key is played. One "tone" usually equals one note or voice of polyphony. JimM is correct in saying that sometimes keyboards contain stereo samples. In that case, one "tone" would equal two voices. However, while virtually all keyboards made today play in stereo, most samples are actually mono. They are "converted" to a stereo signal as they go through the keyboard's onboard effects processor.

So in the most typical case, a single key played is using anywhere from 1 to 4 notes of polyphony. A keyboard's processor has nothing to do with how many tones make up a given sound. It has everything to do with the sampling technique and the choices made by the sound designer. I believe that this was pointed out earlier, but the use of the sustain pedal uses up polyphony as well. A 6-note chord played with a 4-tone patch/instrument will use 24 notes of polyphony. When the chord changes, if there's any overlap in the chords due to the sustain pedal, that number can double instantly. 64 note polypony should be enough for a solo piano assuming that the player has good sustain pedal technique. But as others have said, when you add MIDI sequencing to the mix, 64 may not be enough.[/b]