Pianists are usually in front of the microphone; but as producer of my own concert albums, I work on the other side, too. Often I'm asked how I approach this double role; and what follows is an attempt to sketch an answer integrating the views from both sides, the technical and the artistic. To me, music is a language of the emotions. I want to record not mere sound but a communication in that language. Communication implies someone to communicate with: an audience. And it implies a reciprocal exchange which is specific to the occasion. We can say that in a performance, artist and audience are in a communicative loop, linked by the hall acoustics. The audience hears the music; the artist hears his sound in the hall and the audience's response. If the room is reverberant, intuitively the artist plays slower for clarity. If it is "dry," a quicker tempo maintains the flow. A bright or dull hall likewise calls for the player to make appropriate changes of tonal balance. If listeners fidget—the artist knows by the quality of their breathing and the million little noises restless people make—their attention may be recaptured by shaping melodic lines more lucidly or making the dynamics more extreme: louds louder and softs softer. If the audience is deeply responsive, the performer may be inspired to reach interpretive heights. In these moments, the artist transcends more than just the notes. He forgets about the music itself, and experiences only the emotion flowing through him. Such an exalted state is rare. But, for me at least, the chances of even an ordinarily good performance without an audience are nil. So the producer's first responsibility, I feel, is to create a favorable occasion by bringing together in a good hall an audience ready to listen and a performer ready to play. Then the goal should be simply to record faithfully the sound of the performance. The recording process needs to be faithful because it occurs outside the communicative loop. That is, the artist does not hear the recorded sound, and therefore cannot correct for inaccuracies in it. If, for example, reverberation is added electronically in the recording, then the tempo may make no sense; for we have seen how the artist adjusts tempo to the hall's reverberation. If the tonal balance is changed by "equalization," then the artist's ear for sonority may seem defective. If "compression" is used to make the louds softer and the softs louder, the recording will be technically easier to make; but one of the prime carriers of musical emotion, namely dynamic range, will have been thrown away. All such changes make the artist sound like a person who cannot hear what he is doing. The first imperative, then, is: Leave It Alone! And leave the continuity alone, too. Every time a wrong note is corrected by splicing in a replacement, the shape of the musical line is at risk. Since most recordings have dozens or even hundreds of splices, it's no wonder that the note-perfect result rarely adds up to a convincing performance. After "leaving it alone," and before attempting to satisfy what we might call "esthetic man" in the listener, we must satisfy "evolutionary man," who, as both predator and prey, wants to know where a noise is coming from, instinctively turning his head toward any sound of interest. Virtually no modern recording allows you to locate the instruments this way; nor do you necessarily get a sense that they had a stable location in the recording hall. These problems come from the use of spaced microphones; this is something like having one ear at the piano and the other in the balcony. A solution has been known for over 60 years, since stereo recording was invented by Alan Dower Blumlein. Use two (bi-directional) microphones in the same spot, one immediately above the other and aimed at right angles. This gives superb stereo, such as most people have never heard. As a bonus, it avoids yet another weakness of spaced mikes, namely the smearing of instrumental attacks, the noises which begin all musical notes. Crucial to identifying each instrument, attacks are also altered for expressive purposes by the artist. When microphones are spaced, attacks arrive at the two (or more) mikes at different times. To the listener's ear, and depending on the exact distance between the microphones, every attack will be either multiple or at least broadened in duration. Even the lesser effect is enough to make a piano sound like a giant guitar; and no matter what the instrument, this approach homogenizes the performer's expressive variety. Coincident mikes don't have this problem for the simple reason that there is no time delay between them. The actual sound of instruments is best caught by "ribbon" type microphones, in my experience. A piano note, for example, has an attack with a certain "adamant" quality. Revealed as this dies away is the sustained body of tone, in which various harmonics swell and decay over time. Within the broadly rhythmic flow of the performance, the artist's precise timing may depend on matching the point of development of the old note to the attack of the new one. (See my article "The Music of Sound".) The almost universal "condenser" microphones (and digital recording, too) seem to muddle the harmonic development of the notes; to that degree, they lose the meaning of these interpretive points in the performance. Ribbon mikes seem to do better with the entire piano sound, and with that of other instruments, too. So far I have described three major differences between my own approach and that of most others: recording with audience, miking with a "coincident pair," and using ribbons. If I were magically granted three wishes for improving the world's recordings, then as a listener as well as a performer and producer, I would wish for everyone to adopt these practices. (Or so I would have said a few years ago. Now that digital recording is so widespread, I would also ask for a fourth wish!) The producer who has followed the path I've indicated might well at this point feel sanguine about his recording. He has the performance going well, he's picking it up beautifully with coincident ribbons, and he can finally retreat to his own den, the control room, where the sound heard through the monitor speakers is truly remarkable. At last, he has a high-quality electrical signal to be recorded on tape, cut on a master lacquer, and pressed in vinyl. From here on, it should be a straightforward process. He now finds to his dismay that the artist can give a good performance more reliably than the producer can make a good recording of it. Tapes, lacquer discs, vinyl pressings—all have their effects on the sound and thus on the musical emotion. One vinyl compound makes a finished record which sounds very close to the master tape, while another gives muddy attacks, compressed dynamics and altered tonal balance. Use the better of the two vinyls but change to another brand of record pressing machine, and again you have an unacceptable record. Dynamic inflection, attacks, articulation—all are vulnerable. Tempo survives best; but if reverberation is lost from the signal anywhere in the recording chain—whether by a poor transformer or badly-designed circuitry—then even a factually correct tempo will be nonsense, just as much as when reverberation is intentionally added. In this morass of equipment and processes, my own choices of analog recording, tube electronics, ribbon mikes and the rest are dictated not by ideology nor a desire to be different, but by my perception, as a musician, of what best serves the music I make. To me digital sound is anti-musical. The digital bandwagon is a popular conveyance; but to borrow an injunction from elsewhere, "Thou shalt not follow a multitude to do evil" [Exodus 23:2]. Bertrand Russell said of a friend that he would not follow a multitude even to do good! One doesn't want to be like that fellow, and I trust that listening to the meaning of the music will keep us on the right path, no matter which side of the microphone we're on.