Each year for the past decade, the French vocal and instrumental ensemble Les Arts Florissants has visited the Brooklyn Academy of Music. In that time, three of their productions stand out as particularly significant: Jean-Baptiste Lully's Atys (1989 and 1992), Marc-Antoine Charpentier's Medée (1994) and Hippolyte et Aricie by Jean-Philippe Rameau (1997). This trilogy of landmarks in the history of French opera was presented in fully staged productions conceived by stage director Jean-Marie Villegier and conductor William Christie.

The age of Louis XIV was an age of contrasts. It was fond of spectacle and ostentation, typified in the great fêtes of Versailles, yet it is rightly known as ‘l'époque classique’, characterised by order and decorum. Its greatest playwright was Jean Racine, whose eleven tragedies had only the barest of sets, presented one action in one place during a maximum of twenty-four hours, chose subjects from history or mythology that featured only mortal characters in easily believable situations, relegated violence and other unseemly behaviour to the wings and to descriptions (récits), and used a limited, noble vocabulary to explore the depths of the human condition and to create poetry of extraordinary beauty. However, in 1673, during the peak of Racine's career, Quinault and Lully created French opera (tragédie lyrique), which, while also using a limited, noble vocabulary, featured spectacular sets and costumes for each act, allowed subplots (even comic ones in the first three operas), staged battles, storms and divine interventions, eschewed historical characters, and presented simpler situations and characterisations in order to leave time for music and dance.

The historiography of Italian opera is particularly well suited to illustrate some problems in the general field of music history and musicology. On the one hand, there is little doubt that Italian opera belongs to the canon, not to say the museum, of learned western music; indeed, today's opera houses surpass concert halls in projecting the ‘museum character’ in which musical tradition seems ‘frozen’. On the other hand, it is also true that only in recent years has international musicology accepted Italian opera as unquestionably deserving of attention. The reasons for this delay are clear enough. Some were easily overcome, connected to the very history of our discipline: since the beginning of the nineteenth century, the musical language of Italian operatic composers diverged from the mainstream Austro-German tradition; the dramaturgy of Italian opera was difficult to understand in a cultural context moulded by Wagnerian theory and practice (in part also by Shakespeare, Schiller, etc.). Other factors, however, are more deeply embedded, and continue to have an effect even in intellectual conditions very different from those of traditional musicology. These include: the manner in which extra-artistic factors determine the operatic work; the various creative competencies that take part in operatic production; the considerable importance accorded to performers, particularly singers; the possibility that parts of an opera may be moved from one work to another, or from one author to another; the fact that in die history of Italian operatic conventions, shared codes and repetition of formulas often prevailed over the search for novelty.

Leonardo Vinci's Li zite 'n galera (The Newlyweds on the Galley) has been afforded a moderate degree of attention by musicologists because it is the first Neapolitan commedia per musica to survive with music almost completely intact. It is also Vinci's first extant opera, and has thus been considered in terms of the light it sheds on the style of his later drammi per musica for which he is better known today.

To what degree does late seventeenth-century English opera contain politics? Some recent critics have assumed that political commentary conveyed by allegory is a pervasive feature of ‘Restoration’ masques and operas. Is this true? Quite a few political interpretations of particular works have been published but no one has systematically enquired to what extent allegory and/or ideology was presumed to be built into operas mounted in late seventeenth-century London. Theoretical statements of the time about opera are scant and contradictory, their authors disinclined to take up political issues. Some of the political content is glaringly obvious (the allegory in Dryde'ns and Grabu's Albion and Albanius); some of it is sharply disputed. How should we read a work like Dryden's and Purcell's King Arthur? Is it essentially a muddled adventure story? An expression of British nationalism rising above current politics? A piece of covert Jacobite propaganda?

Donizetti's involvement with Paris has been understood to date from February 1834, when Rossini, acting as the Théâtre Italien's music director, commissioned Marino Faliero for performance the following winter. Though written in Italy, the work was substantially revised in Paris during the two months Donizetti spent there before the première on 12 March 1835, hence William Ashbrook's assertion that Donizetti ‘wrote for the first time for Paris, absorbing the musical tastes of that city, when he presented Marino Faliero at the Théâtre Italien’. While this opera unquestionably remains the one that officially launched the composer's Parisian career, new manuscript and printed musical sources reveal that Donizetti's ties to the city actually date back to 1833, when plans were laid for productions of two of his earlier operas: Gianni di Parigi (1831) and Gianni da Calais (1828). Like Marino Faliero, the first of these works was composed in Italy with a Paris (or a London) première in mind, Donizetti having hoped diat Giovanni Battista Rubini would introduce it at one of his benefit performances soon after rejoining the Italiens (part of the troupe spent spring and summer seasons at the King's Theatre). A close study of the work might shed light on the composer's understanding (or ignorance) of operatic practice in the city at this point in his career; indeed, it was precisely because Gianni di Parigi was so French at least in terms of its libretto (Felice Romani's adaptation of Claude Godard d'Aucour de Saint-Just's libretto for Boieldieu's classic and closely protected Jean de Paris, in the repertory of the Opéra-Comique since 1812) that the comic melodramma was never mounted in Paris. But it was the second work, composed for the Teatro del Fondo in Naples (where it was premièred on 2 August 1828), and performed at the Théâtre Italien twice, on 17 December 1833 and 4 January 1834, that Donizetti actually helped adapt to Parisian taste.

The day following the première of Jean-Philippe Rameau's Hippolyte et Aricie (1733), Voltaire observed in a letter to Pierre-Robert Le Cornier de Cideville that the opera's libretto was by the abbé Simon-Joseph Pellegrin, and he added with characteristic irony that it was ‘worthy of the abbé Pellegrin’. As we know, Voltaire was not alone in his sentiments. Few poets were reviled with the relish one finds in criticism of Pellegrin. Doomed to be a less-than-perfect poet in a literary culture, Pellegrin was forced by circumstance to do something few of his time could understand or accept: he earned a living writing poems on commission. This made him a faiseur. a hack who pursued art for money rather than inspiration, a paid servant who, improbably, sought to move in the very social circles that employed him. It is emblematic of Pellegrin's situation and his reputation that in 1774, twenty-nine years after his death, Voltaire did write noëls when the marquise du Deffand requested them, but complained that it was ‘une rude commission’ and compared his lot to Pellegrin's. Chastened, the marquise apologised. Pellegrin's artistic destiny would seem to have little to do with diat of die urbane and widely admired Voltaire, but I will argue that, in spite of this, the two shared some similar attitudes and experiences regarding the emplotment of tragedy and tragic opera, and further, that these shared views may have had a direct bearing on Rameau's development as a composer.

The libretto for a tragèdie en musique might be regarded as a tabula rasa. Whereas an Italian dramma per musica generally differentiates between recitative poetry and aria poetry — versi sciolti on the one hand, strophic versi misurati on the other — its French counterpart contains page after page of supple vers libres, which lend themselves to a great variety of musical shapes. Indeed, apart from a few highly conventional gestures, one cannot necessarily tell simply by looking at the libretto how a tragédie en musique will have been set. The point might be illustrated by an example, a pair of invocations sung by Phèdre in Jean-Philippe Rameau's Hippolyte et Aricie (1733); one of them is a single quatrain, and the other consists of eight lines divided by their rhyme scheme into two quatrains. The columns at the left show the length of the vers (with + indicating feminine rhyme) and the rhyme scheme:

The protagonist of Thomas Mann's novel Doktor Faustus (1947) is the composer Adrian Leverkühn, whose Faustian pact with the demonic consists of a long, voluntarily untreated syphilitic infection, which his brain craves as an exhilarating if destructive liberation from its icy germanic discipline. The craving is implanted on the very day on which, having renounced the call to theology, he commits his life to music. Arriving in Leipzig late in 1905 to take up his studies, he has a guided tour of the city, among other things visiting Bach's Thomaskirche. But his guide concludes the tour by dumping him in a brothel. Confused, he heads for the piano and tries to work out a harmonic problem: ‘Modulation from B major to C major, … as in the hermit's prayer in the finale of the Freischütz … on the six-four chord on G’. Then he rushes out, but not before a prostitute has brushed him on the cheek and indelibly fascinated him.

When Luigi Bassi entered the stage of the Prague National Theatre in 1787 to create the title role of Mozart and Da Ponte's Don Giovanni, he could have drawn inspiration from a rich tradition of theatrical, pantomimic and marionette representations of the legendary Don Juan, to which this new opera was the latest contribution. Previous incarnations had been shaped by the likes of Tirso de Molina, Molière, Shadwell, Purcell and Gluck; yet it is Mozart and Da Ponte's version that has for us become the definitive: the Don as paradox; an uncomfortable blend of the despicable and the admirable, hero and anti-hero. Lecher, rapist, liar, cheat, murderer, he is the brutal epitome of macho striving for power and domination, yet clothed with a seductive panache, conviction and bravado — the reckless-heroic libertine phallocrat who would rather face the fires of eternal damnation than curb his appetites.

Towards the end of a life spent meditating on music, philosopher Vladimir Jankélévitch wrote that ‘music is the silence of words, just as poetry is the silence of prose’. With this lapidary remark, he bottled up essences of French symbolist musical doctrines, their elevation of music above language, their impulse to urge poetry towards the condition of musical sound, and their sense that music is ineffable. ‘And the ineffable’, Jankélévitch wrote, ‘cannot be expressed because there are infinite and interminable things to be said of it’. Denying that music was inarticulate – that what it invokes is indicible, unable to be said – he saw instead a hall of sonorous mirrors, saturated with implications and suggestions diat draw the mind on to a vanishing point that can never be reached. Music has the perfect apparitional quality that others would deem characteristic of symbols per se, as that which materialises in the distance and cannot be caught when we journey towards its meaning: it will always recede, leaving us with our glowing trail of assumed connotations. Thus symbolist literature, with its rhapsodies to musical mystery, dealt profitably in poetic images of suggestive but unimaginable sound. Such sounds have their most famous incarnation in the melody of Mallarmé's faun's flute, repository for all that cannot be thought, said or remembered. Mallarme created a poetic symbol, verbally couched, which as symbol per se borrows the ineffability of music. It is a mirage, which we approach and lose even as we imagine its meaning. By choosing a musical object (a flute melody), Mallarmé drew doubly on music's apparitional quality — as if to say: even if the impossible could occur and we were to grasp the symbol and its single luminous meaning, hearing what the flute plays, that sound, like all music, would merely send us wandering onwards again.

In a recent issue of this journal, David J. Levin proposed an approach to evaluating the work of directors and producers of opera. The idea that one might be able to theorise the difference between good and bad stagings is appealing, not least because many of us would like to feel able to raise the standard of debate on this subject. Most public discussions of opera productions (or at least of those productions that generate public discussion) can be predicted in advance, more or less verbatim; the persistence of the arguments used on all sides is in itself enough to suggest that little progress is being made. On the other hand, it is not easy to see how academic debate contributes. Theatre is where operas enter public discourse. Performance might seem like a decisive act of interpretation – choices have to be made about how to present the given work – but, paradoxically, it also marks the point at which opera escapes the attentions of the academy in favour of a constituency which is (presumably) less grounded in theory and less committed to consciously interpretative acts. With understandable reservations, Levin suggests the use of commercially available videos to analyse details of a staging, but details of this sort are not likely to contribute significantly to a theatre audience's experience of how a production works and what it has to say. Videotape permits us the mastery of freeze-frame enquiry, and at the same time confines us within the flattened perspective chosen at each moment by the camera's eye. Both its advantages and its drawbacks are incommensurate with theatre, where (especially in opera, with its simultaneous but distinct modes) the stream of information is diverse and continuous, and our eye moves in relative freedom, never capturing the totality of the stage. One might draw an analogous distinction between academic criticism, which works by isolating certain elements of the ‘text’ or its contexts and subjecting them to intense scrutiny, and the more holistic act of sitting in the opera-house watching a ‘work’ unfold. The pause button creates a sequence of discrete images submitted to the critic's intellectual play. In the theatre, a staging is more likely to achieve its effects through what we might call its ‘feel’, its general character and stance. When Peter Sellars set Così fan tutte in a diner, the air of incongruous modernity – conveyed through costume, set, the characters' ways of behaving–must have determined the audience's sense of his interpretation far more powerfully than (for example) the fact that he had Ferrando and Gugliemo sing ‘Secondate, aurette amiche’ in their own characters rather than their assumed ones.

In the original version of the libretto of Hippolyte et Aricie, the librettist Simon-Joseph Pellegrin prefaced Act V with the rubric ‘le théatre ne change qu'à la troisiéme scéne’ [‘the stage changes only at the third scene’]. This is one of the few acts in the entire repertoire of tragédies en musique where the conventional formula ‘le théatre représente …’ [‘the stage represents … ’], followed by a description of what the stage was intended to depict, was not used. The annotation in Hippolyte warned audiences that, instead of the change of scenery that would normally occur during the entr'acte, the first two scenes of Act V retained the setting of Act IV (‘a wood by the sea, consecrated to Diana’), and the new décor (‘a delightful garden comprising the avenues of the Forest of Aricie’) was revealed only in the third scene. The sense of discontinuity at this ‘internal’ scenery change was also heightened by a break of liaison de presence (that is, none of the actors in scene 2 remained on stage for scene 3). A dispute arose over this breach of convention, and resulted in the first two scenes being omitted from performances sometime during the opera's first season in 1733, ostensibly to avoid the breach of unity of place caused by the change of scenery. This was not the only revision made to the opera in that season, but it was arguably the most significant.

At midnight on 31 December 1913 the low string melody of Parsifal's opening rose upwards in the Teatre Liceu, Barcelona. That year-turn marked the end of the copyright period within which performances of Parsifal outside Bayreuth were prohibited, and made Barcelona the first city in the convention-governed world to stage Parsifal legally. Of course, the timing was a marketing ploy, but for many it was also a gesture of outreach, of stretching beyond the mountainous boundary of the Iberian peninsula and of hoping to become a credible cultural and political power in northern Europe. Alas for Barcelona, Europe didn't notice.