Artur Schnabel said that he liked to play music that was better than it could be performed. Mozart's music would certainly fit this description: no single performance of one of his great works could possibly bring to life all the facets of its structure and design or all its expressive depth and range. I might add that his music is also better than it can be analysed. I have worked on some pieces for 30 years and more, getting closer to them, I hope, but never feeling that I have reached a 'definitive' understanding. In the quixotic attempt to reach such an understanding, whatever it might be, I have perhaps learned more than if I had set my sights on an attainable goal. And I am sure that performers who try to match the structural richness, emotional depth and easy, unpretentious perfection of Mozart's music in their approach to playing and singing become better performers, and not only of Mozart. In many ways, analysis and performance are related activities. Every analysis is a kind of conceptualized performance, and every performance embodies an implicit analysis. One would imagine, therefore, that performers and analysts would want to share and learn from one another's perspectives. Indeed, one would imagine that, to some extent, analysts and performers would be the same people. On the whole, however, performers tend to be resistant to analysis, even to doing it themselves; and in this regard players of period instruments are not so different from their colleagues who play modern instruments. Insofar as resistance to analysis stems from a distrust of abstract and sterile intellectualism, it is an understandable and even healthy attitude. Performers, after all, have to use their ears, muscles and hearts as well as their minds. But insightful analysis can speak to ears and hearts-sometimes even to muscles-as well as to minds; and there are good reasons why Mozartians might investigate what theory and analysis have to offer them. First of all, the performing milieu of Mozart's time included a good deal of consciousness about theoretical issues. The great teaching manuals of the middle and late 18th century-Quantz, C. P. E. Bach and Leopold Mozart, among others-discuss such aspects of playing as articulation, dynamics and ornamentation in close connection with melodic contour and harmonic structure. For the most part, however, they enunciate general principles-valid, important principles, to be sure, but of necessity not always applicable to specific situations, for the principles that deal with the articulation of the musical surface may conflict with those that deal with tonal structure. Both Leopold Mozart and C. P. E. Bach, for example, tell us that the first note under a slur gets an accent; they also tell us that chromatically altered notes are to be emphasized.' But they don't tell us what to do when, say, the second note under a two-note slur is chromatically altered. And, indeed, no set of precepts, no matter how cleverly conceived, can be adequate to the complexities of musical experience. The only way to approach these complexities is through study of the individual piece, and here 20thcentury analysis can often provide helpful insights. No earlier period has produced so large a body of work devoted to the close study of musical compositions; and if many of the concepts of modern analysts were unknown to theorists of Mozart's time, it does not follow that they have no bearing on his music. Such writers as Tovey, Schoenberg, Lowinsky and Rosen, to name a few, have much to teach us. But since my own work has followed the approach of Heinrich Schenker, it is mainly is ideas that I am going to discuss here. I believe that Schenker went beyond any previous theorist or analyst in the depth and scope of his insights, but his point of departure was, more often than not, the theoretical sy tems of the 18th century-thoroughbass, species counterpoint and, despite his attacks on Rameau, Rameauvian fundamental bass. Incidentally, Schenker was far in advance of his time in his ideas about performance practice. Early in the century he was urging performers to revive continuo playing, to learn stylistic lly appropriate ornamentation, to pay attention to Beethoven's metronome markings (he had more faith in them than I do), and to discard the featureless continuous legato suggested by the 'phrasing slurs' of corrupt editions in favour of the varied and lively articulation shown in the composers' autographs. Mozart's music is better than even Schenker can analyse, but his ideas can bring performers closer to it, as I shall try to show. In this paper I shall relate these ideas to aspects of Mozart performance practice, concentrating to a large extent on the implications of Mozart's slur markings. One of Schenker's most important ideas is what he calls the Auskomponierung eines Klanges, the composing