Those fortunate enough to have seen the recent production of Jean-Baptiste Lully's Atys mounted by Les Arts Florissants will remember the sumptuously staged realm of sleep in Act III, during which costumed lute and recorder players appeared alongside the singers and dancers. Although conductor William Christie and director Jean-Marie Villégier made no attempt to reproduce the original seventeenth-century staging, they did adhere to Quinault's instructions for this scene to the extent of making musicians prominently visible. Atys is not exceptional in calling for stage musicians: Lully regularly included instrumentalists among the dramatis personae of his tragédies en musique, the genre on which he lavished most of his creative energies after 1672, and the practice is even more evident in the thirty or so ballets he composed for Louis XIV's court during the preceding two decades. The phenomenon of on-stage instrumentalists – much more extensive than the use of the banda in nineteenth-century Italian opera – has been studied only for the information it affords about the development of Lolly's orchestra or the iconography of French Baroque opera. This article is concerned rather with why instrumentalists appeared on stage at all, what they represented, how they functioned as characters, and the impact they had on the visual spectacle.

Conservative writers in seventeenth-century France tended to agree that the morals of women were degenerating, though they disagreed as to whether the cause was the excessive strictness or the leniency of husbands. We would perhaps characterise the change more as a newly found sense of freedom and self-confidence that allowed women, however tentatively, to contest the double standard of the age. The new morality was seen in fashions that bared progressively more of the neck, shoulders, breasts, ankles and legs, revealing new erotic frontiers; in a relaxation of the taboo against female swearing and coarse language; in an epidemic of female gambling; and in scandalous reports of women's over-indulgences in the sensual pleasures of food, drink, nicotine and sex.

The operatic diva, a singer of strange songs, and too often a turbulent, unkind girl, haunted the nineteenth-century imagination, as evidenced by the musical tales of E. T. A. Hoffmann and numerous retellings of those tales in theatre, ballet and opera. Each adaptation of Hoffmann's ‘Rat Krespel’, ‘Der Sandmann’ and ‘Don Juan’ reflects an ambivalent attitude towards women performers, whose potent voices make them simultaneously desirable and fearsome. How do these stories about female singers contrive to contain and manage the singing woman’s authority? And how does the prima donna's voice repeatedly make itself heard, eluding and overcoming narrative attempts to shape or contain its turbulent noise?

Let me begin with an excerpt from ‘Rat Krespel’ (1818), which might serve as a parable for relationships between female singers and male music lovers in the Romantic imagination. Krespel, a young German musician, travelled in Italy and was fortunate enough to win the heart of a celebrated diva, Angela, whose name seemed only appropriate to her heavenly voice. Unfortunately, her personality was less than heavenly, and when she was not actually singing he found her violent whims and demands for attention very trying. One day, as he stood playing his violin:

[Angela] embraced her husband, overwhelmed him with sweet and languishing glances, and rested her pretty head on his shoulder. But Krespel, carried away into the world of music, continued to play on until the walls echoed again; thus he chanced to touch the Signora somewhat ungently with his arm and the fiddle bow.

‘Quoi du reste’, to paraphrase Derrida on Hegel in Glas, ‘ici, maintenant, d'une Carmen?’ What's left of Carmen here and now? Aside from its intrinsic interest, the question seems worth asking in light of a bias that recent treatments of opera – particularly those influenced by poststrucuralist theory – seem to betray. The most prominent, Catherine Clément's Opera, or The Undoing of Women, Michel Poizat's The Angel's Cry and Jeremy Tambling's Opera, Ideology, and Film, regard opera as an institution rather than as a body of texts. Each of the authors, to my mind at least, allows a prior structure or structures – the systemic presence of male domination and its construction of women in society, a quasi-Lacanian understanding of unconscious fantasy, the bourgeois construction of operatic experience – to constrain what operas, and opera, can mean. They thus produce what are in effect reception studies or analyses of the audience, which is perhaps why they operate at some distance from the detail of the texts, musical or verbal, of the operas they analyse.

At one point in Paul Rudnick's play, Jeffrey, the title character, who claims to have slept with 5000 different men, speaks directly to the audience: ‘I know it's wrong to say that all gay men are obsessed with sex. Because that's not true. All human beings are obsessed with sex. All gay men are obsessed with opera.’