Although the Bohemian composer Leopold Koželuch (1747–1818) did not go as far as Haydn in his recapitulatory revisions, he also had a strong penchant for producing altered reprises. As with Haydn’s practice, Koželuch’s “new twists of the old” (Vogler) cannot satisfactorily be explained by invoking (the most common variant of) the redundancy hypothesis, according to which classical composers sought to delete the secondary theme (or its basic idea) in the recapitulation in the case of a monothematic exposition. It seems to be more likely that, in revising the primary theme in the recapitulation, Koželuch was responding to the theme’s repetitive structure, aiming both to forestall thematic redundancy and to produce a more continuous formal design. In addition, the use of the primary theme in the development section – as a ‘medial return’ or a false (tonic or off-tonic) recapitulation – might be construed as provoking its alteration in the reprise.
Apart from the desire to avoid redundancy, recomposed recapitulations can also be explained by reference to the composer’s intention to address a deficient event located at some point before the recapitulation. (1) A non-tonic opening is assumed to have far-reaching consequences for the recapitulation, in which the allegedly deficient event reappears in either a regularized or a properly contextualized fashion. (2) A weak harmonic preparation of the recapitulatory onset (most commonly by V/vi) is said to be a non-normative strategy that would call for a “compensatory dominant zone” (Russakovsky), necessitating a rewrite of the reprise.
In scrutinizing these hypotheses, I will not focus on Koželuch’s practice in isolation, but rather will discuss it in the broader context of a contemporaneous (mainly Viennese) corpus of roughly 600 movements. In so doing, I shall seek to demonstrate the advantages of using statistics and generalized corpus data for the analysis and interpretation of individual works.