It is sobering to discover after a lifetime of attending the musical theater how little sticks in the memory. From the ballet, for example, I can conjure up only three clear images. One is of the great leaps of Yurek Shabelevsky in the Polovetsian Dances from ''Prince Igor.'' A second is of Leonid Massine's feckless parvenu Peruvian in "Gaite Parisienne." And the third is of the fluttering hands of the lovely Tatiana Riabuchinska, reawakened from death to life by the music of Paganini's guitar(!) in Michael Fokine' s "Paganini."

"Paganini" was, for whatever reason, a highpoint in my admittedly not terribly extensive ballet-going career. I suppose the choreography has been irrevocably lost in the intervening four decades, but I have been left with a perhaps irrational love of the score, which is none other than the Rachmaninoff ''Rhapsody on a Theme of Paganini.'' In checking out this work for this essay, I was delighted to discover that the basic outline of the ballet was the composer's own. Fokine, taken by the music, had proposed such subjects as Perseus, Andersen's "The Shepherdess and the Sheep,'' Peter the Great, and the creation of the world (as envisioned by Ovid). Rachmaninoff's letter of reply said, in effect, "How about the Paganini legend?" and went on to sketch the various episodes that Fokine later developed, including the apotheosis in which the violinist is surrounded by diabolic fiddling caricatures of himself. The London premiere was a splendid success, and the Times noted that the score was "unexpectedly good ... for dancing."

In those lofty circles occupied by the self-appointed tastemakers over the last half-century, Rachmaninoff's music is regarded as only slightly less unforgivable than Puccini's and Tchaikovsky’s. (All three commit the cardinal sins of being melodic and popular!) Moreover, the composer-Rachmaninoff of the later years is said to be self-imitative and burnt-out. Maybe so, but, from personal taste, I find myself returning to the "Rhapsody" more often than I do to, say, the second piano concerto. The work is built on two sure-fire themes. The chief one is that of Paganini's 24th Caprice for solo violin (itself a set of variations), which has attracted similar treatment by Brahms, Liszt, Boris Blacher, Witold Lutoslawski, and others; the other--a favorite of the dour Rachmaninoff -- is the "Dies irae" from the Gregorian mass for the dead-­familiar from Berlioz "Fantastic Symphony" and Liszt's "Totentanz". The variation form was particularly agreeable to the composer, and this set is both tautly structured and imaginatively conceived. For most people the high point is the extended 18th variation, whose haunting melody is nothing more than the Paganini theme turned upside-down!

Unlike his concerted pieces, Rachmaninoff's sonatas are among the least-known and most infrequently played of his major works. There seems to me a tendency to dismiss the two for piano as "interesting only for pianists." The second (B-flat minor) was written in 1913. The composer himself felt that it sprawled too much, and revised and condensed it in 1931 (Rachmaninoff loved inversions). (Horowitz once worked up a synthesis that preserved the best features of the original and the revision.) Considering the generally condescending critical evaluations of the piece, it is nice to find the sonata-authoritative William S. Newman writing about it with real enthusiasm and at some length. His summary: "rich, unusual, advanced, genuine," --and he dismisses the objections of the late Eric Blom as Pecksniffian.