A History of Film Music

With our modern obsession with technology, and related tendency to equate technological advancement with improved quality, the introduction of synchronized sound to the motion picture in the late 1920s seems, with hindsight, to have been not only inevitable but downright essential – a sine qua non of the development of a meaningful cinematic art. In spite of this widely held view, film soundtracks were until the last quarter of the twentieth century almost entirely ignored by film scholars and theorists, who continued to devote exclusive attention to the visual image. Rick Altman has argued that this situation arose as a result of two deeply ingrained fallacies: one historical, in which the introduction of sound to the cinema may be regarded as an afterthought, and the other ontological, in which it is believed that sound must by its very nature always be regarded as subservient to the primacy of the image as the chief vehicle of meaning and expression (Weis and Belton 1985, 50). Theorist Rudolf Arnheim went so far as to consider the introduction of sound as directly violating the motion picture’s ontological status (Flinn 1992, 41). At the opposite extreme was Walter Ruttmann’s experimental documentary Week-End, which in 1930 took the bold step of dispensing with an image track altogether and simply running through the projector a suggestive soundtrack made up from dialogue, noise and music (Winter 1941, 154).

It is a curious paradox that the most famous of early film pioneers, Thomas Edison, had established recorded sound first and then sought to add moving images to it – a reversal of most film-makers’ subsequent priorities. Tom Gunning has drawn attention to the enormous and lasting psychological impact of Edison’s innovations, noting that ‘the phonograph had in effect separated the human senses, divorcing ear from eye, and . . . Edison’s original intention in pursuing motion pictures was to bring them back together’. This ideal was not motivated by a ‘need for perfect representation or a bourgeois desire for coherence’, but sprang from ‘a deep anxiety aware of the manner in which technology, while doubling the human, also seems to be splitting it up, transforming the nature of human subjectivity’ (Abel and Altman 2001, 16, 29).

Naturalistic sound in a motion picture may be deemed necessary only in the interests of cinematic realism. Even from the earliest days, however, directors had shown markedly differing attitudes towards the latter concept, and their work had tended to polarize in one of two directions. On the one hand, the Lumiéres and their followers devoted their efforts to a naturalistic, quasi-documentary style of reportage, their films termed actualités. On the other, the illusionist Méliés had developed a style of fantasy narrative laced with symbolism which was important not only for breaking away from realism, but for developing a whole host of technical devices (including time-lapse photography, superimposition and dissolves) in order to achieve freshness and originality of presentation: in short, Méliés ‘broke from the photographic impulses of the primitives to show that the movie camera could lie’ (Parkinson 1995, 18). As with painting and photography, it was quickly shown that, even if such a thing were attainable, a realistic presentation of events was not necessarily desirable.

The introduction of synchronized sound, on the simple realist level a natural extension of the filmic medium, necessitated a marked shift of emphasis in terms of the medium’s visual syntax. As the French film composer Maurice Jaubert commented of the soon-to-be-outmoded silent film:

Driven by the absence of speech to a lengthy method of visual paraphrase in order to make the story clear, the silent film built up for itself, little by little, a special idiom designed chiefly to compensate for the silence of the actors. This convention became familiar to all habitués of the cinema, who believed, legitimately in those days, that it gave occasion for a special art of the screen – an art which in its finest development would be essentially allusive, and so poetic. But as soon as speech came to destroy this early convention, the cinema – although hardly anyone recognised it at first – changed its character. It became, it is, and it remains realistic.

The relationship of the new sound component to the cinema’s musical provision, and the artistic possibilities opened up by the sound element itself, proved to be more complex than might have been the case without this fundamental shift in aesthetic emphasis.

Many influential film-makers and theorists in the 1920s were deeply sceptical about the value of synchronized sound, and it is instructive to examine the principal arguments generated by consideration of whether the sound film represented a significant advance on the potentialities of the silent medium. Documentary maker Paul Rotha complained in 1930 that ‘The attempted combination of speech and pictures is the direct opposition of two separate mediums, which appeal in two utterly different ways’, and feared that an obsession with sound would now take precedence over the visual image (D.Cook 2004, 225). Others continued to draw attention to the redundancy of sounds which merely duplicated or reinforced the actions portrayed on screen. (Considering so much effort had been expended on finding mechanisms by which sound and image could be accurately synchronized, as detailed below, this was somewhat ironic.) Béla Balázs predicted that the sound film would ‘destroy the already highly developed culture of the silent film’, going so far as to deem it ‘a catastrophe, the like of which had never occurred before in the history of any other art’. As late as the early 1950s he found little cause to revise this bleak prophecy, lamenting the preponderance of ‘speaking photographed theatre’ and still expressing the hope that, as the silent film had before it, the sound film would finally reveal ‘new spheres of human experience . . . based on new principles’ (Balázs 1953, 194–5).

In 1928, the Soviet film-makers Sergei Eisenstein, Vsevolod Pudovkin and Grigori Alexandrov published their Statement on Sound in which they stressed the importance of continuing the suggestive montage techniques already developed in the Soviet silent cinema, now threatened by the new technology. If sound were to be used merely realistically, they argued, it

will destroy the culture of montage, because every mere addition of sound to montage fragments increases their inertia as such and their independent significance; this is undoubtedly detrimental to montage which operates above all not with fragments [per se] but through the juxtaposition of fragments.

Only the contrapuntal use of sound vis-á-vis the visual fragment of montage will open up new possibilities for the development and perfection of montage.

The first experiments in sound must aim at a sharp discord with the visual images.

The writers predicted that such tension would result in an ‘orchestral counterpoint’ of vision and sound, and claimed that their ‘contrapuntal method’ of structuring would have the benefit of becoming a truly international cinematic language. (This last concern was, in their case, the result of artistic idealism: in contrast Hollywood moguls at this time were, for more commercial reasons, concerned that the sound film, with its heavy dependency on dialogue, would not prove to be as exportable as had been the case with silent films, which readily permitted the insertion of intertitles in the language of countries to which they were sold.) In a book on film technique published in 1929, Pudovkin asserted that ‘the first function of sound is to augment the potential expressiveness of the film’s content’ (Pudovkin 1958, 184) and advocated the imaginative use of asynchronous sound to this end. Examples of varying degrees of audio-visual counterpoint in Soviet films made in c.1930–34 have been closely analysed by Kristin Thompson (1980).

The term ‘counterpoint’ has persisted in theoretical writings on film music, not without controversy. Jean Mitry felt that only through the use of multiple screens, as in Gance’s Napoléon, could visual images meaningfully reflect the independent simultaneous strands of true musical counterpoint (Mitry 1998, 142), and condemned fashionable and ill-considered uses of the term to describe the combination of music and image: ‘Counterpoint is precisely the sort of impressive word which looks very good in magazine articles; however, that does not mean that it is any the less irritating, since it is generally misused . . . [T]o refer to counterpoint with reference to the opposition of feelings expressed by music and film is utter nonsense. Would it be an oversimplification to use the only really appropriate word to describe the effect, namely contrast?’ (Mitry 1998, 250). The influential French film theorist Michel Chion also drew attention to the inappropriate way in which the term has often been used when divorced from its strictly musical meaning, noting that ‘dissonant harmony’ might be a more accurate musical parallel for ‘momentary discord between the image’s and sound’s figural natures’ (Chion 1994, 37). Kathryn Kalinak asserts that the concepts of ‘parallel’ and ‘contrapuntal’ music in film reinforce a regrettable tendency for music always to seem subservient to the visual image (Kalinak 1992, 24–6), though overstates the case somewhat, since the theorists she cites – Rudolf Arnheim, Balázs and Kracauer – all stress that music can be free to pursue its own course in film in addition to reinforcing what is latent in the image. As far as the early makers of sound films were concerned, the relationship between music and image centred far more on the difference between ‘realistic’ diegetic music, for which the source was immediately obvious, and more ambiguous applications of music and sound in which their possibly diegetic source was not explicit. The two approaches were later dubbed ‘synchronous’ and ‘asynchronous’ respectively (Kracauer 1960, Chapter 7). Although opinion was sharply divided between these two options in the early years of the sound film, as with many other technical innovations in art of all kinds the most fruitful way forwards proved to be various hybrid accommodations between the two approaches. In modern cinema, carefully manipulated contrasts between diegetic and nondiegetic music and sound, and the blurring of the distinction between the two categories, remain powerful tools for creating tension or ambiguity.

One obvious drawback of a fixed soundtrack was its inevitable exclusion of any possibility that a film’s sound provision might differ from one screening to the next. (This rigidity in film exhibition of course applies equally to the fixed visual images, and Polish director Krzysztof Kieślowski dreamt of releasing his La Double vie de Véronique (1991) in numerous slightly different versions so that movie-going would retain an element of chance, the spectator not knowing in advance which of them would be screened at a particular venue: see Stok 1993, 188.) Few musicians would argue against the notion that, in many ways, live performance is preferable to recorded music – for reasons such as spontaneity, immediacy of impact, richness of sound, spatial awareness and so on; another important factor is that live performances of the same piece, as in different performances of the same play, can vary significantly. The importance of these performative variables disappeared from sight as a result of the general misconception that silent-film music was – like the silent cinema in general – essentially ‘primitive’. It was only with the renascent interest in live music for films towards the end of the twentieth century that this vital dimension of the art was recaptured. The principal negative factors in the special case of live cinema music are the difficulties of synchronization with the visual image and effective balance of volume with other elements of the total sound component, though the latter is no more challenging than the task facing the conductor of opera or concerti.

Whatever the artistic merits and demerits of the use of recorded music in the cinema, a direct practical consequence of the introduction of prerecorded soundtracks in the late 1920s was the catastrophic downturn in the employment opportunities available to musicians who had made careers for themselves in movie-theatre bands. Aformerly flourishing livelihood disappeared almost overnight, in spite of vociferous protests by several musicians’ unions, most prominently the American Federation of Musicians. The problem was by no means restricted to the USA: in the UK, for example, as much as 80 per cent of all professional music-making in c.1928 came as a direct consequence of the silent film (Ehrlich 1985, 199). In Germany, musicians were in early 1929 dismissed without notice as movie theatres were adapted for sound: the 200 players who had in 1928 been working for the theatres of Berlin’s powerful conglomerate, UFA (Universum Film Aktiengesellschaft), had dwindled to around 50 by 1930, these having been retained because silent films had not yet fully died out. In late 1930, some 8,000 German musicians had been sacked as a direct consequence of the advent of the sound film, this figure nearly doubling in the following three years (Kater 1992, 26). In France, an article in Ciné-Journal published on 7 December 1928 drew attention to what the editors regarded as ‘perhaps the greatest danger since professional musicians first came into existence’:

Mechanical music is substituting immediate, human performance. Everyone is threatened: already some cinema orchestras are being abolished and replaced with soundtrack equipment. Others will follow. No musician, no conductor should be indifferent to such a threat. If strenuous measures are not taken, an army of unemployed will be created, who will threaten colleagues still in employment . . . But aren’t the musicians, in the circumstances, the architects of their own misfortune? . . . [A soundtrack] repertoire is being created that will one day supplant orchestral performance almost completely . . . Faced with such a danger, and while awaiting new regulations appropriate to the new situation, the union committee [of the Fédération Française de la Musique] has decided to forbid, until further notice, all soundtrack recordings.

As might have been predicted, in the face of industrial action French producers simply set about recording their soundtracks abroad, an act provoking further hostility from the trade press – in which cultural chauvinism reared its ugly head. On 3 January 1930, Ciné-Journal deplored the glut of foreign films with foreign music and declared that ‘a significant part of the musical accompaniment of synchronized films shown in France must be given over to French music . . . the cinema should not become in France the instrument of the infiltration of foreign views into the national culture’ (Lacombe and Procile 1995, 45).

Towards the end of the silent era, various devices had been introduced to aid musicians in their attempts to synchronize their performances with the action on screen. In Berlin, for example, Carl Robert Blum invented a rhythmonome, which he exhibited in 1926. This recorded sound in the form of a rhythmogram, which charted the exact rhythmic course of the music in a diagrammatic form that was played back in accurate synchronization with the image projector so the conductor could match the tempo to the moving diagram; the device found an application in live theatre, being used for the first production of Ernst Křenek’s opera Jonny spielt auf in 1927. Milhaud, who used the device at Hindemith’s suggestion at Baden-Baden in 1929 (see Chapter 7), recalled that Blum’s apparatus made it possible

to run off the film at the same time as a reel of similar size bearing two staves on which the music was written, so that the music could follow the slightest movement of the picture. During the performance, the musical score was thrown on the conductor’s desk at the same time as the images were projected on to the screen. In this way the conductor was able to synchronize his playing exactly with the film.

A similar invention, named cinépupitre, had been developed in Paris by Pierre de laCommune and used by Honegger when conducting his music for Gance’s La Roue in 1922 (London 1936, 66–8). At the Paris Opéra premiére of Gance’s Napoléon in 1927, the unusual priority accorded to the music saw the conductor equipped with a device for altering the speed of the projection to suit the pacing of the score (Abel 1996, 40).

The next important step, which sounded the death knell for live music in the cinema, was the use of phonograph recordings of music linked to the projector and played back with some degree of synchronization – an idea with which inventors had been toying since the first attempts around 1900 by Messter in Germany and Gaumont in France. Use of recorded music was relatively common in early movie theatres, being especially popular in France where Pathé, like Edison, had a vested interest in promoting his phonographs as much as his moving pictures; but, because they were not yet synchronized, phonographic accompaniments to film screenings were generally considered to be inferior to those provided by live ensembles. In the 1920s, appropriate recordings could be sourced from sound libraries that supplemented printed cue sheets for live music, and synchronized playback equipment became more widespread. The sound operators, who controlled the turntables while watching the screen, were sometimes musical directors from the silent era whose jobs were now in peril. According to London,

The individual turn-tables had a sound-overlapping plant, that is, the music on one record could be made to fade into the next . . . If a third turn-table was available, one could venture at certain junctures on an attempt to set some noise, perhaps even words, in counterpoint against the running music . . . These ‘acoustic overlaps’ . . . were at that time necessary to preserve the illusion of continuous line in the musical accompaniment to some extent . . . Individual volume controls, one fitted to each turn-table, which enabled the dynamics to be graded at will, completed this system.

Cinematic disc-jockeys were aided by the availability of ‘illustrative scores’ for specific pictures, which gave correspondences between the duration of each segment of film and the recorded music intended to accompany it. Films which had live sound recorded on disc at the time of shooting, or which had orchestral music specially recorded in synchronization with the image, were screened using automated playback devices:

The projecting machine for the sound-on-disc film consisted of a normal projector, from which a flexible coupling proceeded to revolve the appropriate gramophone record . . . The difficulties of the transition from one record to the next were overcome as follows. Shortly before one record had finished playing, an electromagnetic coupling was switched on through a small metal cell attached at a suitable point in the perforation holes of the film strip running through the projector. This coupling in turn set the second record in motion by means of the flexible driving shaft mentioned above. When the latter had reached the proper speed, the first record stopped, and the operator could then change the records at his leisure and arrange their ‘entrance’ at the exact point indicated.

An example of a silent film which was intended to be screened with a portion of disc-recorded music was Griffith’s Dream Street (1921), the first American film to include a mechanically synchronized song; in the event, the synchronized sequence was quickly abandoned on account of poor synchronization and inferior sound reproduction (Barrios 1995, 15). Prophetically, the large discs used in many movie theatres played at a speed of 3313 revolutions per minute, many years in advance of the recording industry’s adoption of this as the standard speed for the long-playing record.

The most successful method for supplying sound-on-disc was the Vita-phone system developed by the Western Electric Company and purchased by Warner Bros. in 1925. At first sceptical about its potential, both artistically and commercially, Warners restricted the new technology to the recording of musical accompaniments in standard silent-film idioms, not daring to attempt to record and synchronize dialogue; they nevertheless marketed the technology aggressively, both in the USA and Europe, by drawing attention to the fact that even small movie theatres could now, in effect, house their own large (if pre-recorded) orchestras with none of the expense or inconvenience such accompaniment had formerly entailed. The system was launched at a lavish presentation in New York in August 1926, at which several short sound films showcasing the work of prominent musicians were followed by the screening of a major new feature, Don Juan (dir. Alan Crosland), with a special Vitaphone score composed by William Axt and David Mendoza and recorded by the New York Philharmonic. In a specially filmed address, Will H. Hays (the leading light of Motion Picture Producers and Distributors of America Inc., a regulatory body popularly known as the Hays Office) informed the audience:

In the presentation of these pictures, music plays an invaluable part. The motion picture too is the most potent factor in the national appreciation of good music. That service will now be extended as the Vitaphone will carry symphony orchestrations to the town halls of the hamlets. It has been said the art of vocalists and instrumentalists is ephemeral, that he [sic] creates but for the moment. Now neither the artist nor his art will ever wholly die.

The Vitaphone shorts essentially perpetuated the vaudeville-like bill of fare at typical movie theatres: they could be shown as part of a variety programme, only now the minor live performers of the provinces were replaced by the finest and most famous singing and acting stars of the day in recorded form.

2.1 Poster for The Jazz Singer (Warner Bros., 1927), depicting one of the sensational scenes in which Al Jolson’s voice was recorded live and synchronized with the marvel of Vitaphone technology.

Eager to capitalize on the success of Vitaphone, which had impressed far more in the shorts than in the highly conventional Don Juan, Warner Bros. quickly released the first ‘talkie’, The Jazz Singer (dir. Crosland, 1927), starring Al Jolson, which was accompanied by a Vitaphone score conducted by Louis Silvers. Jolson had already appeared in Vitaphone shorts, but The Jazz Singer now made his a household name. The film’s main attractions, which had materialized almost by accident, were the brief sequences of dialogue which Jolson reportedly improvised on set as the recording apparatus was running: unlike the stiff and stilted formal presentations of speech and music in the shorts, ‘here was Jolson not only singing and dancing but speaking informally and spontaneously to other persons in the film as someone might do in reality. The effect was not so much of hearing Jolson speak as of overhearing him speak, and it thrilled audiences bored with the [melodramatic] conventions of silent cinema’ (D. Cook 2004, 210). Jolson’s featured songs were diegetic, and recorded on set live. In all other respects, The Jazz Singer was a thoroughly conventional silent film, with intertitles used to render other dialogue and essential information, to which a disc-recorded synchronized score had been added in an equally conventional silent-film idiom. An old warhorse of the compilation score, the love theme from Tchaikovsky’s fantasy overture Romeo and Juliet, featured prominently – not because it had the remotest relevance to the film’s plot, but simply because it was a popular tune of the day; it appeared alongside the equally well-known Jewish melody Kol Nidre, used to establish the cultural milieu for the plot.

In 1928, Warner Bros. issued Tenderloin (dir. Michael Curtiz) and Glorious Betsy (dir. Crosland), both hybrids that united silent-film procedures with sporadic Vitaphone inserts carrying synchronized dialogue and song; in Glorious Betsy, the Marseillaise was sung by a star of the Metropolitan Opera, André de Segurola. The first film to do away with silent sequences altogether was Warners’ Lights of New York (dir. Bryan Foy), released in the summer of 1928 and significant not only for its use of synchronized dialogue throughout, but also for its uncredited but competent musical score. Warners’ preferred Vitaphone composer remained Silvers, who for their second ‘all-talkie’, The Terror (dir. Roy Del Ruth), supplied an ‘elaborate wraparound musical score’ that threatened to swamp the actors’ voices (Barrios 1995, 48). In September of this same momentous year, Warners issued a second Jolson vehicle, The Singing Fool (dir. Lloyd Bacon), another uncomfortable hybrid of stilted silent-film presentation and live Vitaphone inserts for which Silvers provided more overblown and omnipresent orchestral music; the film contained two spectacularly successful songs, one of which (Buddy DeSylva’s ‘Sonny Boy’) not only sold over a million copies as sheet music but also a million sound recordings (Sanjek 1988, vol. 3, 106–7). Both The Jazz Singer and The Singing Fool, which owed their appeal far more to their star’s charisma and sheer novelty value rather than to any inherent artistic merit, were widely exported – in both Vitaphone and silent versions – and Warner Bros. saw its profits increase six-fold within a year. After the advent of sound-on-film, the company doggedly persisted with outmoded Vitaphone technology and only began to employ soundtracks recorded on celluloid in January 1931; even then, it continued to release films in both formats until the middle of the decade (Barrios 1995, 117).

In an attempt to rival Warners’ initial success with Vitaphone, the Fox Corporation released Sunrise (dir. F. W.Murnau) with a synchronized score by Hugo Riesenfeld in 1927, and re-released in New York the silent films Seventh Heaven (dir. Frank Borzage) and What Price Glory? (dir. Raoul Walsh) with synchronized music by Ernö Rapée replacing the compilation scores prepared by R. H. Bassett for the Los Angeles premiéres of both films (McCarty 2000, 36, 237). More important was Fox’s development of a sound-on-film system called Movietone, launched in a series of newsreels also commencing in 1927. The Fox newsreels were shot with a single sound-and-image camera, which permitted a rapid development of the film (no editorial synchronization being required) ideal for the prompt reporting of newsworthy events; Movietone’s first film, of the West Point parade, demonstrated how diegetic sound could provide its own spatial perspective, since the music of the marching band inevitably crescendoed as it came towards the camera (Barrios 1995, 30). Western Electric was also developing a sound-on-film system at the same time as it was promoting Vitaphone; another rival optical soundtrack system, Photophone, was under development at Radio Corporation of America (RCA). In 1927, a Big Five agreement was signed by five leading studios, in which they undertook to adopt whichever sound-on-film system proved best suited to their needs; this was considered to be the Western Electric system, though RCA’s Photophone remained an important competitor, and was showcased by the company’s newly merged studio interest, Radio-Keith-Orpheum (RKO), which began producing sound films in 1929.

Hundreds of silent films in production at the time of these momentous innovations had optical soundtracks added to them so that the work and investment would not be wasted, while the rewiring of cinemas to accommodate the new technology proceeded at a phenomenal rate. The operation was horrendously expensive – even the Vitaphone system cost some $23,000 per installation (Barrios 1995, 26) – and required substantial backing from investors. Consequently it was the introduction of sound that was in large part responsible for the domination of film production by wealthy businessmen-turned-producers who subsequently came, quite literally, to call the shots in Hollywood.

Sound-on-disc was both cumbersome and limited in its applications, and for many years before its advent inventors had been vying to perfect and patent various methods of recording synchronized sound directly onto film stock. In 1904, Eugéne A. Lauste carried out the first successful attempt to record sound as an optical pattern using a photo-electric process. Similar experiments, in France, Germany and the USA, explored ways of recording sound on a film strip in the form of surface indentations, swollen thicknesses, roughened surfaces and differences of electrical resistance in the material itself (London 1936, 101); there had also been some experimentation with magnetic tape, though this was not to be generally adopted for sound recording until the 1950s. The leading sound-on-film system was developed by Tri-Ergon from c.1918, and their films were screened in Germany in 1922. Developments proceeded apace once important innovations in the field of radio and telephone technology were adapted for use in experimental sound-on-film systems. The American inventor Lee De Forest exhibited his sound-on-film movies in New York from 1923, including some with pre-recorded musical scores composed by Riesenfeld. De Forest made his name primarily as a pioneer of the electrical amplifier; he had attempted to persuade the Hollywood studios to adopt his sound-on-film system, without success, but briefly enjoyed fame in London in 1925–6 when his films – which featured recorded dialogue, sound effects and musical scores – were screened to some acclaim (Huntley [1947], 30). Only in 1935 was De Forest’s innovative film work legally recognized in the USA, when he received substantial compensation for the widespread efforts of his unscrupulous competitors to infringe upon his patent.

Optical-soundtrack technology was, in conception if not execution, very simple. During the recording process, electrical impulses from a microphone were fed via an amplifier to a light source which illuminated the film stock as it ran past a small slit. The resulting soundtrack ran down the extreme edge of the same strip of celluloid that carried the visual images, and was blanked off when run through the projector so that it did not appear on the screen. The projector housed a photo-electric cell that contained selenium, a substance of which the electrical conductivity varies according to the intensity of the light striking it; the cell converted the optical soundtrack patterns back into electrical impulses which were fed to an amplifier and loudspeaker. Although sound and image were directly synchronized by this method, the soundtrack patterns for each photographic frame were recorded nineteen frames in advance of the relevant image, for two reasons: this physical separation was required by the distance between the image projection gate and the audio photo-electric cell (which could not be housed in the same part of the projector), and the gap also meant that a transport mechanism could be devised whereby the film could be made to pass smoothly across the photo-electric cell in contrast to its designedly jerky frame-by-frame motion through the visual projection gate, which would have seriously distorted audio output. Tri-Ergon had invented and patented a mechanism for precisely this purpose, and their monopoly on the device earned them a considerable income from projector manufacturers before William Fox acquired the rights to their invention in 1927 (D. Cook 2004, 207). The nineteen-frame separation was ideal for sound synchronization in large cinemas but the distance was later increased, first to 20 frames and then 21 frames, in order to reflect the demands of differing projection conditions. When films appear to have poor sound synchronization in the modern cinema, this is usually because the operator has threaded the film incorrectly through the projector (Weis and Belton 1985, 417).

There were two ways of recording the soundtrack, according to whether the recording mechanism involved a lamp or a mirror. A variable-density (or variable-intensity) soundtrack took the form of a pattern of parallel lines, not unlike a modern computer barcode, their differing thicknesses reflecting the strength of the emissions from a lamp flashing through the recording slit. A variable-area soundtrack was a zigzag pattern of varying width, created by linking a galvanometer to a small mirror which was deflected by different extents according to the strength of the current it received and cast correspondingly varied patterns of light on the film (London 1936, 108–11; Sabaneev 1935, 4–9). The variable-density method, developed by Western Electric, proved better for dialogue than music, while the opposite was true of the variable-area method, used in RCA’s Photophone system; the latter was therefore modified in 1936 to remove dialogue distortion and the optical variable-area soundtrack – to give it its full name – became the industry standard.

Regardless of whether a single sound-on-film camera or two separate cameras (one recording images, the other sound) were used, it was at first only possible for the soundtrack to be recorded simultaneously with the shooting of the visual image. Severe restrictions were caused by inadequate microphones hidden on set and the need for the cameras – which had become noisy since they were motorized to operate at a new speed of 24 frames per second in order to improve the quality of the soundtrack – to be housed in sound-proof booths; this in turn made the camera apparatus exceptionally unwieldy and limited the range of camera movements at a director’s disposal. It was virtually impossible to ensure silence on set, where the hiss of the arc lights was a constant distraction. Exactly as Eisenstein and his colleagues predicted, imaginative montage was abandoned in favour of static, theatrical blocking of the action in which everything was dependent on the location of the fixed microphones.

Control over the quality of the soundtrack was primitive. Where multiple microphones were used, the signals were mixed on set before entering the sound camera, which ran in synchronization with the image camera. In the single-camera system, the soundtrack was already at nineteen frames’ remove from the image track, as was the combined release print of the film stock from the two-camera system, and this meant that cut-and-paste editing was virtually impossible until the introduction of the sound Moviola (see below). Everything in these early years was therefore filmed and recorded live in single takes. As with the sound-on-disc recordings in The Jazz Singer, a variety of camera shots within a single scene could only be achieved by the expensive business of running multiple cameras simultaneously in different locations – as is done with video cameras in modern television studios – so that the resulting images could be intercut without sacrificing sound synchronization (Fairservice 2001, 229). Background music would either have to be recorded live on set at the time of shooting, or else recorded in advance and mechanically reproduced on set so it could be re-recorded simultaneously with the dialogue. Both methods yielded lamentably poor sound quality, partly because the low-contrast film stock ideal for image photography failed to provide the level of contrast needed for the soundtrack patterns.

2.2 A Vitaphone sound camera in the late 1920s.

Disappointing recording fidelity was only one of several reasons that contributed towards the general avoidance of nondiegetic music in films from the period c.1928–33. The novelty of synchronized diegetic sound temporarily put background scores out of fashion: music that appeared to emanate from the motion picture itself could, in the interests of realism, be better justified if it were strictly diegetic in origin. In Hollywood, many early sound films included music only for opening and closing credits in addition to diegetic needs; as Max Steiner related, a violinist might be gratuitously included in the background of a love scene solely to justify the use of what would otherwise be invisible romantic underscoring (Naumberg 1937, 219). Some theorists and film-makers were dubious of the very nature of nondiegetic music, Paul Rotha commenting that ‘Music in the form of accompaniment to realist story-films – no matter how well composed and written – is an anachronism, a hangover from the silent cinema which few film-makers are brave enough [even in the 1950s] to discard . . . All non-source music is an artificial aid to stimulate the emotions of the audience and not an integral and valid part of the film aesthetic’ (Rotha 1958, 22–3, n.1). When nondiegetic music was reintroduced after c.1933, its function had subtly changed, a shift summarized by Kalinak: ‘the basis of the silent film score, the principle of continuous playing and selective reproduction of diegetic sound, was rejected in the sound era in favor of a model based on the principles of intermittent music and faithful reproduction of diegetic sound’ (Kalinak 1992, 40). Along with intermittency came a reduced need for coherent structuring, as Mitry observed: ‘Good film music can do without musical structure provided that its intrusion into the film at a specific moment should have a precise signification . . . from which it gains all its power once associated with the other elements: images, words, and sounds’; he cited Roland-Mañuel’s dictum that ‘music must deny its own structure if it is to be an ally of the image’ (Mitry 1998, 249).

In 1929–30, King Vidor and Lewis Milestone experimented with sound dubbing in their respective films Hallelujah! and All Quiet on the Western Front, shooting certain sequences with fluid camera movements (as had been possible in silent films) and adding the soundtrack afterwards. In the same period, Ernst Lubitsch made creative use of dubbing in several features, notably Monte Carlo (1930; music by W. Franke Harling et al.), in which the rhythmic clatter of a train’s wheels merges into the film’s theme song. Such postsynchronization had obvious technical and artistic advantages, and by the mid-1930s several independent tracks were used for the separate recording of dialogue, music and sound effects. Sound quality significantly improved from the poor fidelity apparent in, for example, the nondiegetic music mixed into the soundtracks of Laurel and Hardy films of 1931–2 (Weis and Belton 1985, 42). With the development of lighter and more mobile sound-proofed (blimped) cameras, the portable boom microphone and relatively silent incandescent lighting, sound recording on set also became much more versatile.

The Moviola, an editing device introducedin1924 which allowed footage to be stopped on an individual frame without the risk of setting fire to the extremely flammable film stock (as would invariably happen in a conventional projector at this time), was in 1930 adapted for sound. It could run image and sound film stock in parallel, with the playback heads either locked together or independent; only for the final release print was the soundtrack pulled down to run in advance of the relevant images, enormously simplifying the editor’s task. (The Moviola remained the principal editorial tool until it gradually yielded to video monitors and computer software in the 1980s.) The seamless editing of soundtracks was enabled by the invention of blooping, which was the superimposition of a black triangular sticker – later a diamond-shaped blob of black paint – above the thick ridge of film where two pieces of film stock were spliced together, this bump having formerly produced a loud bang when played back through the projector: the black bloops obviated the abrupt changes in electrical signal that had caused the obtrusive noise (Fairservice 2001, 234–5). This technique allowed the editorial process to appear as transparent (i.e. invisible) in the soundtrack as it became in image editing, such transparency becoming a dominant feature of mainstream cinema until the 1950s.

The acoustic limitations of primitive sound-recording technology remained a source of irritation to composers throughout the 1930s. Stringed instruments sounded poor in reproduction so, in Paris, Eric Sarnette and Hanns Eisler took the radical step of trying to do away with them altogether in favour of more ‘microgenic’ instrumental timbres – not entirely for practical reasons, as Sarnette, like Stravinsky in his 1920s concert works, was suspicious of the emotive power of string clichés and felt the instruments represented a regrettable ‘feminist’ dimension in modern music (London 1936, 193). Recording fidelity of other instruments was variable, though the comments of contemporaneous musicians are fairly subjective on this point with, for example, one judging the bassoon a poor instrument for recording purposes (Sabaneev 1935, 60) and another praising its qualities (London 1936, 171). In conjunction with Adolphe Sax, Jr, Sarnette developed special ‘microphone instruments’, such as the saxotromba with its movable bell, the saxotrombone and a bass trumpet with a bell shaped like a saucepan lid (London 1936, 191). These instruments had little impact and now seem quaint, but their use should not be divorced from Sarnette’s devotion to ‘a clean and plastic method of composition’. London declared that ‘[Sarnette’s] orchestra conceals nothing; the almost pitiless clarity of sound-production requires of composers that their every note should have sense and meaning. The days of romantic sound-painting with chords and sound-mixtures of indefinite nature are over’ (London 1936, 194). Similarly, Sabaneev recommended that aspiring film composers should avoid complex harmonies, more than two contrapuntal lines, extremes of register and cluster chords low in the texture (Sabaneev 1935, 37). These strictures were not just the result of limitations in recording fidelity, but part of a growing European feeling that the increasingly overblown romantic idiom of Hollywood film music was inappropriate for the modern age.

An even more radical method for bypassing poor sound reproduction was so-called ‘animated sound’, experiments with which began almost as soon as optical soundtrack technology became widely available. These involved the use of graphic patterns drawn, printed or photographed directly onto the soundtrack in order to generate synthetic sounds without the need for audio recording.

In Germany during the early 1930s, Rudolph Pfenninger and Oskar Fischinger created synthetic musical tones with soundtrack patterns written by hand that gave ‘the impression of a remarkable abstraction of tone, reminiscent of certain electrical instruments, which can be traced to the absence of overtones’ (London 1936, 198). Similar procedures were subsequently developed by director Rouben Mamoulian (see below), and also by the Bauhaus artist Lászlo Moholy-Nagy (who believed that composers could henceforth ‘create music from a counterpoint of unheard of or even non-existent sound values’), by Evelyn Lambert in Canada, Jack Ellit in London, John and Jack Whitney in the USA, and by Arseni Avraamov and N. V. Voinov at the Scientific Experimental Film Institute in Leningrad. Avraamov achieved microtonal effects and won recognition for them as early as 1931, while Voinov developed graphic representations of the notes of a seven-octave chromatic scale and used these to prepare animated soundtracks realizing pre-existing pieces of classical music, as did Pfenninger independently using a similar method (Prendergast 1992, 198–200; R. James 1986, 81–4).

Following on from Pfenninger’s pioneering work, Canadian artist Norman McLaren’s system of animated sound employed a ‘small library of several dozen cards, each containing black and white areas representing sound waves, [which] replaced traditional musical instruments and noisemaking devices’ (McLaren 1953). When Francis Poulenc sawMcLaren’s Blinkity Blank at Cannes in 1955, he singled it out as the most musically revelatory film of that year’s festival and hailed the director’s demonstration of the ongoing need for film-makers to continue striving to find new techniques uniquely suited to the medium of film (Lacombe and Porcile 1995, 121: see Chapter 7 of the present volume for further discussion of McLaren’s work).

Outside the limited realm of experimental visual art, however, it was virtually impossible to secure sustained financial backing for commercial applications of these ventures: when David Fleischer patented an animated-sound system in 1931, it was bought by Paramount but apparently never used (R. James 1986, 86). Occasional experiments with graphic sound were made in narrative features such as Dr Jekyll and Mr Hyde (see below) and The Devil and Daniel Webster (see Chapter 5), but such novel procedures would soon lose out to the ever-more frequent use of far more conveniently achieved and increasingly versatile electronic sonorities prominent in film scores from the 1940s onwards.

Most early commercial sound films seem crude by any standards, and do not stand comparison with the artistic successes achieved by the best examples of the final flowering of the silent film in the late 1920s. Because synchronized and realistic sounds were initially a gimmick, they often drew attention to themselves in a manner that today seems both obsessive and naíve. French director Jean Renoir, annoyed by the trend, revealed of his first sound film, On purge bébé (1931):

to register my bad mood, I decided to record the sound of a toilet flushing. This was a kind of revolution that did more for my reputation than the shooting of a dozen worthwhile scenes. The most noteworthy artists and scientists of the great sound communities declared that this was an ‘audacious innovation’.

Imaginative directors, however, boldly experimented with the illusory and emotional possibilities offered by both sound effects and music in the new synchronized medium.

Former theatrical director Rouben Mamoulian enthusiastically threw himself at the challenge, strongly believing that film sound ‘should not be constantly shackled by naturalism. The magic of sound recording enabled one to achieve effects that would be impossible and unnatural on the stage or in real life, yet meaningful and eloquent on the screen’ (quoted in Lack 1997, 84). It is worth noting that by no means all creative American film-making in this era was monopolized by Hollywood: Mamoulian’s first film Applause (1929) was shot in Paramount’s studios in Astoria, New York. He used experimental techniques such as double-track mixing of pre-recorded whispering and singing, and included many examples of his belief that the source of a diegetic sound need not be visible on screen. (Asynchronicity of this kind was later investigated by Chion, who coined the term ‘acousmatic’ to describe it and ‘acousmêetre’ for specific instances where it involves the powerful effect of a disembodied voice, which he felt to be deeply rooted in traditional notions of the unseen voice of a deity: see Chion 1994, 128–31, and Chion 1999, 17–57.) Applause is also notable for its use of a sound dissolve – the noise of bottles and a glass at the end of one scene merging into diegetic cymbal playing in a restaurant band at the start of the next – and a yearning convent girl’s dream sequence accompanied by an appropriate collage of popular tunes and the ‘Ave Maria’ (Weis and Belton 1985, 241–2); its mobile camera work – and mobile microphones – became a reality only after fierce confrontations between the director and his crew, whose mentality was deeply rooted in static theatrical blocking. Mamoulian’s City Streets (1931; music by Karl Hajos) contains the first example of a voiced-over flashback repeating dialogue heard earlier in the film. In Dr Jekyll and Mr Hyde (1932) Mamoulian used a mélange of avantgarde sonic expressionism, which became dubbed ‘Mamoulian’s stew’, to accompany the transformation from urbane Jekyll to monstrous Hyde: the astonishing sonic collage was created by using a mixture of animated sound, reversed recordings of gong strokes, amplified heartbeats and reverberating bells. With commendable restraint, Mamoulian used this extraordinary effect only once in the film, thereby immeasurably intensifying its emotional impact. Its outlandish and deeply disturbing quality is strikingly prophetic of alien soundscapes that did not become commonplace in film soundtracks until decades later.

Among the first important sound films in Europe were Alfred Hitchcock’s Blackmail (1929), which was initially shot as a silent then partly remade to include a synchronized score, and René Clair’s Sous les toits de Paris (1929), both of which transferred the major conventions of silent-film music to the sound screen but also made significant creative use of diegetic song. Both Hitchcock and Clair were, in different ways, keenly aware of the creative possibilities offered by the soundtrack. On a basic level, both – like Mamoulian – used diegetic sound to promote a sense of three-dimensional space. And the necessity of adding sound to films that had been conceived as silents in the event proved to be more of a liberation than a hindrance, given that large parts of their transitional soundtracks had to be prepared using flexible postproduction techniques.

At this stage in his career Hitchcock, like most other directors in this period, tended to avoid nondiegetic music. In Blackmail the sporadic nondiegetic scoring (by Hubert Bath et al.) is sometimes crude, notably in the cue that accompanies the opening and closing police chases, which is rooted in melodramatic clichés and stolidly sequential in construction. But the score also incorporates impressionistic reworkings of a motif borrowed from a song first heard diegetically and then used to comment on aspects of the drama as it unfolds, a technique anticipating the similar procedure used many years later by Steiner and others. As Royal S. Brown comments: ‘it is not, of course, the whole song that has left the diegesis and made its way to the nondiegetic music track but rather small fragments of it: in essence, the entire song [‘Miss Up-to-Date’, by Billy Mayerl] and its ironies are expressed in the isolation of its first four notes’ (R. S. Brown 1994, 43). Hitchcock’s Murder (1930) is more typical of its time: after conflating the opening of Beethoven’s Fifth Symphony and the rhythmically similar scherzo theme from the same symphony to serve as ‘fate knocking on the door’ main-title music, the film’s music is restricted to diegetic cues – including a long segment from the opening of Wagner’s Tristan und Isolde. Although the introduction of the latter and its justification both seem crudely executed by today’s standards – the music is evidently present solely to provide an atmospheric background accompaniment but Hitchcock over-emphasizes the fact that its source is diegetic, making us listen to a stilted radio announcer telling us precisely what we are about to hear – this does not lessen the originality of the scene in question, which is the first use of stream-of-consciousness voice-over in cinematic history. At the film’s climax, diegetic music is again contrived to serve as suitable dramatic accompaniment: as a trapeze artist sets about hanging himself in full view of the horrified public, the diegetic circus music (implausibly) grows more tense and melodramatic, stopping altogether so that the grisly act is carried out in disturbing silence. At the end of the film, we see hero and heroine united in a fond embrace and assume that the lush accompanying music is nondiegetic – until the camera dollies backwards revealing them to be actors framed by a proscenium arch, and the love music is therefore now presumed to be issuing from the theatre’s orchestral pit. Murder affords further examples of asynchronous sound: the source of the scream at the opening is not seen, court-room dialogue at the moment of pronouncing the verdict is dubbed over shots of the empty room that the jury has just left, and a dialogue flashback gives insight into the accused’s thoughts as she sits alone in her police cell.

In the opening commuter sequence from Hitchcock’s Rich and Strange (1931), Hal Dolphe’s nondiegetic score catches every action – including pointing out successive entries in a ledger, a book slamming shut and umbrellas snapping open – to create a cartoon-like sequence of almost balletic mickey-mousing. Later in his career, however, Hitchcock remained concerned that nondiegetic music might detract from filmic realism, famously requiring none in Lifeboat (1944) on the grounds that one would not expect to encounter an orchestra playing in the middle of the ocean. To this suggestion, David Raksin equally famously replied, ‘ask him where the camera comes from and I’ll tell him where the music comes from!’ (quoted in Kalinak 1992, xiii). Nonetheless, Hitchcock’s extended collaboration with Bernard Herrmann in the 1950s and 1960s elicited some of the most celebrated film scores is cinema history, even if the critical acclaim accorded to Herrmann’s music placed some strain on their relationship (see Chapter 5).

Clair’s use of music in the early sound era was more adventurous than Hitchcock’s, and formed part of the French director’s vision for developing a type of intellectual yet popular sound film in which dialogue was avoided in favour of a sustained application of sound effects and music. In this regard he directly influenced Mamoulian, by whom dialogue was similarly considered as only one of the sonic possibilities available to the director – a view also shared by Jean Renoir. At a conference on the sound film held in Brussels in 1926, Clair’s fear that dialogue would come to dominate all aspects of film production led him to condemn the new technology as a ‘terrible monster, a creation against nature, courtesy of which the screen will become an impoverished theatre – the poor man’s theatre’ (Lacombe and Porcile 1995, 225). Sous les toits de Paris, featuring music by Armand Bernard and a popular song by Raoul Moretti, owes its success in large part to Clair’s fine sense of integration between visual dynamics and aural momentum; as Arthur Knight has pointed out, the relationship between music and movement in Clair’s films often resembles a ballet (Weis and Belton 1985, 217), a comparison useful to an understanding of the function of nondiegetic music in many sound films of the succeeding generation. Clair’s fluid camera work and asynchronous use of sound may have been partly inspired by Harry Beaumont’s seminal musical The Broadway Melody, which he sawi n London in May 1929 (Barrios 1995, 68–9). As Gorbman notes in her analysis of Sous les toits (Gorbman 1987, 140–50), Clair avoids naturalistic sound until the climactic street fight, which suddenly restores it and then displaces the tension by overdubbing the noise of an unseen train in an example of what she terms ‘auditory masking’. Music is not present in this scene, its absence adding to the stark effect, but is elsewhere pervasive and prominent. It too serves as auditory masking when it directly replaces dialogue, and it can stimulate a new sense of spatio-temporal perception when it appears to bind together disparate characters and locations through the agency of pseudo-communal singing – as when the camera pans slowly from one tenement window to another and every character it encounters along the way is singing or playing (in real-time continuity) successive phrases from the title song, first introduced diegetically by a plugger trying to sell the sheet music to the neighbourhood crowd at the beginning of the film. The wit here lies in their irritation that they cannot now rid their minds of the tune: ‘Ça va, ça va! On la connait!’, snaps one character in a bar before the theme is again taken up by others in the succeeding interior shots. Towards the end of the film a recording of Rossini’s William Tell overture, played diegetically on a gramophone in the bar (where it conveniently accompanies a brawl), is the mechanism for more wit characteristic of Clair: the needle sticks in the groove and, as the music jerkily repeats the same bar over and over again, an old punter watching the fight thinks he must surely be drunk.

While working on A nous la liberté (1931), Clair demonstrated the importance of music to his overall conception of the film by involving the composer, Georges Auric, in all aspects of the production from the writing of the script onwards. Before working with Clair, Auric had already begun his collaboration with Jean Cocteau, historically significant for its experimentation with random associations between image and ostensibly unrelated music (see Chapter 8). In A nous, Auric supplied a sparkling comedic score, very much in the brittle manner of Les Six and featuring prominent saxophone, cor anglais and percussion, in its moments of comic capering directly anticipating his later work for Ealing Studios (discussed in Chapter 6). Auric’s music is constantly foregrounded because Clair’s direction remains rooted in the purely visual humour of the silent cinema and for large stretches is not reliant on dialogue. Themusic track sometimes sharply manipulates moods in a rather basic fashion, as when cheerful music for toys seen in close-up turns sombre when the camera reveals them being manufactured by convicts. Literal-minded mickey-mousing occurs when rays of sunshine strike each of a prison cell’s window bars in turn to the individual notes of an arpeggiated vibraphone chord, an idea wittily recapped when the prisoner attempts to hang himself from the same bars, which buckle under the strain and allow him to escape through the window instead. Immediately after this illustrative cue, Clair disconcerts his viewers by playing with their expectations of film sound: a woman’s singing has been issuing enticingly from a window, but the music deteriorates, stops and restarts incongruously once she has emerged into the street, revealing the source of the singing to have been a recording played from another apartment. This trick, exposing the essential artificiality of all film music, later became an endearing cliché of comedy soundtracks. Other felicitous touches include Auric’s delicate accompaniment while convicts eat in diegetic silence, a cue that later returns to accompany a scene of an industrial conveyor-belt production line – thereby underscoring the film’s suggestion that, for some, life outside prison bars is not much more inspiring than life behind them. In contrast, two parallel scenes of workers clocking in at their factory are given different musical accompaniment even though the images are virtually identical: the first receives mechanistic industrial music typical of many concert composers’ experiments in the 1920s, while the second cue is surprisingly dreamy. Diegetic music is used creatively in the passage-of-time montage during which another escaped convict progresses from lowly gramophone salesman to become a magnate of the hi-fi industry (featuring a series of overlapping recordings played on increasingly sophisticated gramophone equipment), and the title song creeps in and out of the diegesis, the two lead characters chipping vocal lines into the nondiegetic score when least expected.

Germany was another centre for musical experimentation in the early years of the sound film. Wolfgang Zeller contributed a substantial through-composed score to Carl Dreyer’s Vampyr in 1932, the year before Steiner put nondiegetic film music firmly on the map with King Kong (see Chapter 3). A creative use of original diegetic music was made by Karol Rathaus, who scored Der Mörder Dimitri Karamasoff (The Brothers Karamazov; dir. Erich Engels and Fyodor Otsep, 1931) in conjunction with Kurt Schröder. Rathaus’ score was much admired by Herrmann, who commented in a lecture in 1973:

This, to me, was the first great realization of the dream of melodrama . . . [Rathaus] treated for the first time the music of a film as an integral part of the whole, not as decoration. Because the film deals with one of the Karamazovs falling in love with a prominent harlot and visiting her in her establishment wherein a gypsy orchestra plays, the music of the picture begins with a gypsy orchestra simply playing Russian gypsy music. But as the picture progresses and the brother becomes more and more involved with the harlot, the music stops being ornamental and becomes an emotional mirror of him. It becomes more and more tragic and more and more hysterical. It reaches its greatest moment, I think, when the brother hysterically drives a troika through a raging blizzard accompanied musically by a great battery of percussion instruments.

Director Josef von Sternberg, like Mamoulian, was an early exponent of the art of using asynchronous diegetic sound to suggest the existence of space outside the limits of the screen. Working in America, he first included an ironic use of diegetic song in his Paramount film Thunderbolt (1929), for which main-titlemusic was composed by Hajos. Like Hitchcock’s, however, Sternberg’s attempts to include atmospheric accompanimental music by using strictly diegetic pretexts were somewhat contrived, as shown by the implausibility of the appearance of music-making prisoners on death row (Kalinak 1992, 69).Made in Germany in 1930 and catapulting MarleneDietrich to international stardom, his The Blue Angel (Der blaue Engel) – which was shot in parallel German, English and French versions – significantly furthered both techniques. The stark manner in which the closing door of the dressing room used by the vamp Lola-Lola (Dietrich) frequently cuts off the sound of the band in the eponymous night-clubmay today seem unduly crude and overly repetitive – it was clearly a novelty at the time – but the film is shot through with other sound events crucial both to the furthering of the narrative (e.g. the chiming of a clock) and characterization, especially that of the protagonist, Professor Immanuel Rath, with his trombone-like nose blowing, whistling and cockerel imitations. As Sternberg commented, ‘the sound film affords me the opportunity to orchestrate an action in such a way that the instrumentation becomes a necessary organic constituent of the entire work’ (quoted in Sudendorf 2001).

Friedrich Holländer’s music for The Blue Angel is almost exclusively diegetic, apart from a brief medley overture that immediately introduces the sound of a tolling bell that becomes so important later in the film, and an evocative orchestral epilogue accompanying Rath’s death in a haze of shimmering Wagnerian harp arpeggios harmonizing the asynchronous tolling of his death knell; this last cue was a memorable afterthought added eleven days after the film’s premiére at Berlin’s Gloria Palast. Dietrich’s interpretations of Hollãnder’s songs are accompanied by the most famous German dance band of the time, Stefan Weintraub’s strongly American-influenced Syncopators (with whom Hollãnder had provided music for Berlin revues since 1927), and these numbers retain their power both to tease and promote dramatic irony: ‘the images were startling, irrevocable, the most overtly sexual yet connected with musical performance’ (Barrios 1995, 320). Recordings of songs from the film sold widely on their tie-in release in 1930, but their independent commercial success obscured the songs’ real power in the film, which lay in throwing the progress of the pathetic and deranged Rath’s downfall into sharp relief. Throughout the film, ironic use is also made of pre-existing melodies: Holländer’s overture pointedly quotes a theme sung by Papageno in Mozart’s Magic Flute, the ever-present clock chimes a song-tune concerned with loyalty and fidelity, and the African musical doll examined by the Papageno-like Rath in Lola-Lola’s bedroom plays him a theme from Schubert’s Die schöne Müllerin.

The Blue Angel grossed a record $60,000 in just one week after it opened at New York’s Rialto in December 1930 and this success significantly boosted the American careers of both Sternberg and Dietrich, who collaborated there on Morocco (1930) and Dishonored (1931), both with music by Hajos, and Shanghai Express (1932) and Blonde Venus (1932), both with music by W. Franke Harling. In 1933, The Blue Angel was proscribed in Germany by the Nazis, Joseph Goebbels considering it to be ‘“offal,” spewed out by the fetid city’; the Jewish Holländer’s music for the film Jungle Princess was also banned (Kater 1992, 23, 45). The Weintraub Syncopators wisely failed to return to Germany after an international recital tour in 1933, most choosing to remain in Australia apart from their pianists Holländer and Franz Wachsmann who, like Dietrich, emigrated to the USA and later became well known in Hollywood under the Anglicized names of Frederick Hollander and Franz Waxman. Other members of the band were less fortunate: trumpeter Adolf Rosner was exiled to Siberia by Stalin and pianist Martin Roman was murdered in Auschwitz.

Undoubtedly, the film musical was the most important genre to emerge during the earliest years of the sound film: it had inspired experimentation in continental Europe, including G. W. Pabst’s version of Brecht’s and Weill’s The Threepenny Opera (see Chapter 4), and a sometimes stunning but more commercially oriented sense of adventure in its Hollywood incarnation. The widespread popularity of featured songs in this period inevitably led to their inclusion in otherwise non-musical dramatic films on both sides of the Atlantic, often as a means of promoting sales of sheet music. Typical American examples from 1930, cited by Richard Barrios (1995, 309–10), are the domestic drama Young Man of Manhattan (dir. Monta Bell) and the Joan Crawford vehicle Our Blushing Brides (dir. Harry Beaumont). In France, the 1931 crime drama Tumultes (dir. Robert Siodmak) similarly contained three gratuitous songs. Partly responsible for this situation was the fact that many early sound-film actors in both countries were seasoned stage-show singers, and were capitalizing on their existing reputations. As film genres steadily became more formulaic in the later 1930s, so the tendency to break into song was rigorously channelled into the unique domain of the film musical, the growth of which in America was partly influenced by the success of Clair’s films stateside. The ground-rules for a style of nondiegetic instrumental music suitable for dramatic features and utterly subservient to the narrative were established in Hollywood; and in the meantime Clair’s imaginative ideal of the ‘musical film’ – in which music would be more important than dialogue in articulating the drama – was consigned to history.