"This raises the question--does the composer know how his piece is to sound?
The problem is a delicate one, and it lies at the heart of our conception of music. If music is not a mere notation on paper, then its realization in sound is crucial. We assume generally that the ideal performance is the real piece, and that this imagined ideal performance is the real piece, not the notes on paper or the wrong notes of an actual performance. But this assumption is flimsy and fails to stand up under examination. And none of these--not the imagined or the actual performance or the schematic representation on paper--can be simply equated with a work of music.
Let us put this in the simplest possible terms. When a conductor in 1790 conducted from the keyboard, we know from contemporary testimony that he often stopped playing to raise his hands. There is no way of knowing when he did this, but he did not play throughout. When Haydn imagined the sound of one of his symphonies, he must indeed have expected a certain amount of piano or harpsichord sonority as being likely here or there, but there is no place in the music where he implied this as necessary or even desirable except for the little joke in the Symphony no. 98.
This means that a composer's idea of his work is both precise and slightly fuzzy: this is as it should be. There is nothing more exactly defined than a Haydn symphony, its contours well outlined, its details clear and all audible. Yet when Haydn wrote a note for the clarinet, that does not indicate a specific sound--there are lots of clarinets and clarinetists, and they all sound very different--but a large range of sound within very well-defined limits. The act of composing is the act of fixing those limits within which the performer may move freely. But the performer's freedom is bound--or should be--in another way. The limits set by the composer belong to a system which is in many respects like a language: it has an order, a syntax, and a meaning. The performer brings out that meaning, makes its significance not only clear but almost palpable. And there is no reason to assume that the composer or his contemporaries always knew with any certainty how best to make the listener aware of that significance.
New ways of composing precede new ways of playing and singing, and it often takes as long as ten to twenty years for performers to learn how to change their own styles and to adapt themselves. The use of the continuo in the piano concerto was, by 1775, a vestige of the past that was to be completely abolished by the music itself, and we have every reason to believe that the figured bass was already nothing more than a conventional notation which provided the soloists and the conductor with a substitute for a score during performance, or, at most, a way of keeping an orchestra together which had no longer any musical significance. The occasional indignation about its omission either from performance or edition is historically unwarranted and musically unjustifiable.
In 1767, Rousseau complained that the conductor at the Paris opera made so much noise beating a rolled-up sheet of music paper on the desk to keep the orchestra in time that one's pleasure in the music was spoilt. The audible use of a keyboard instrument during a symphony or the orchestral section of a concerto written after 1775 is no doubt less irritating, but its authenticity and its musical value are the same.
-after a discussion of continuo's role in Mozart's piano concerti. From Rosen's 1997 Expanded Edition of The Classical Style: Haydn, Mozart, Beethoven, p. 195-6.