“The past is slowly slipping away from us. It is no longer ours to interpret as we wish, but ours only to reconstruct as faithfully as possible.” — Robert P. Morgan 
“Every performance today is a translation. — Charles Rosen 
The librarian of Europe’s oldest university library, the Archiginnasio in Bologna, Italy, asked me to cancel our live performance. With a ten-piece orchestra, I had been rehearsing Mortimer Wilson’s 1926 score for Douglas Fairbanks’s film The Black Pirate. I was ushered into the librarian’s office where he explained to me that the building was made of a very soft native stone. Three years before, a law had been passed prohibiting musical performances in the library’s courtyard because the vibrations were causing the stone to break off. The librarian had originally given his permission for our screening because he thought a “film muto” would be completely silent. But when he heard our rehearsal, he realized his mistake.
Since cancellation of the performance was not possible at that point, we decided to move the piano away from the walls and ask the percussionist to modify his dynamics. That night during the dress rehearsal with the film, pieces of stone came crashing down, staying clear of the pianist and other musicians thanks to a compromise with the librarian—the projectionist was positioned on the other side of the courtyard on the floor above, on a side that had been destroyed by a bomb during World War II and reconstructed in cement. Apparently this dress rehearsal took care of the loosest fragments because nothing similar happened during the actual performance the next night.
I begin with this story in order to call everyone’s attention to the first and biggest obstacle to my work—the misconception fostered by the term “silent film.” Initially it was used to distinguish films made before 1930 from talking pictures, whose soundtrack resided physically on the film with the images. Movies that did not have a physical soundtrack on the film became “silent” film. The live sound that had accompanied them, that had intensified the images dramatically, surrounding the public so that the whole experience seemed as if in three dimensions, had been performed within the theaters, not on but with the moving images. But with the term “silent film” came the mistaken idea that for the first thirty years of the twentieth century, films were experienced in silence.
A further misconception was fostered by the terms “sound films” and “synchronized sound films” applied erroneously to talking pictures (the “correct” terms should have been “mechanically synchronized sound” and “recorded-sound” films), but the erroneous terms suggested that the moving pictures before 1930 either did not have sound or had music or effects that were not synchronized to the moving pictures. Actually, the difference between mute and talking pictures was the difference between live and mechanically achieved synchronization and between live and recorded sound. Finally, the entire subject of sound was eliminated by cinema studies specialists who assumed that the history of cinema could be written exclusively about the images, no matter that after 1915 music and/or sound effects constituted fifty percent of the experience. Music historians, on the other hand, thought that whatever music had accompanied films before 1930 was a kind of extinct animal that had nothing to do with sound practices after 1930, and anyway, all film music was inferior in quality to highbrow music. If it had anything to do with previous practices, it was associated with the music for melodramas, a lowbrow form of entertainment not worthy of study.
The Passion of Joan of Arc
(Carl Theodor Dreyer, 1928)
That was the situation in 1978 when I was hired by the Library of Congress Music Division to get bibliographic control over their copyright deposits of music for moving pictures. My appointment was somewhat ironic, as I preferred listening to music or reading a book to watching a film and knew very little about moving pictures. However, as I broke the spines of scores for films before 1930 in order to prepare them for microfilming, I came across the original piano conductor score for what I soon learned was one of the most famous films of all time, Dreyer’s The Passion of Joan of Arc. The music was scored for orchestra, chorus, and three vocal soloists. The copyright deposit consisted of the piano conductor score and four of the five string parts. The piano part was sparsely cued to the film, meaning that there were references to actions on the screen over specific measures of music. The musicians were expected to arrive at that measure when the specified action was on the screen. I played through the score and liked what I heard. It aroused my curiosity. I wanted to find out what the full orchestration with chorus and three vocal soloists would sound like, how it would work with the film, what kinds of problems I would run into when trying to put the music back together with its orphaned images, and what audience reactions today might be.
That was the first original film score I ever tried to reconstruct. American audiences were moved by my first performance with only piano, strings, organ, chorus and soloists that took place in St. Mark’s Church behind the Library of Congress in Washington, and later elsewhere in the United States and Canada. Eventually, after the missing orchestral parts were reorchestrated, a performance with the vocalists and orchestra of the Teatro Comunale in Bologna elicited a more distant reaction. After one of my early performances, Jose Manuel Costa of the Cinemateca Portuguesa hypothesized that at the original Paris premiere of the film, the presence of the music might actually have been part of the movie’s censorship. The French censors excised scenes from the film but I, too, now believe that the lovely but grand, romantic pseudo-Gothic French religious score by Léo Pouget and Victor Alix could have been deliberately constructed to overwhelm Dreyer’s intimately photographed, anticlerical motion picture. Without these present-day performances we probably never would have suspected that the original musical accompaniment might have played a role in the censorship at the Paris premiere.
The Passion of Joan of Arc was my first score, but I consider my reconstructions (as distinct from the restorations) to be experiments, the results hypothetical. We cannot know what the scores were really like for The Passion of Joan of Arc, Nosferatu (dir. F. W Murnau, 1922), Häxan (dir. Benjamin Christensen, 1922), Master of the House (dir. Carl Theodor Dreyer, 1925), Stark Love (Brown, 1927), Rosita (Ernst Lubitsch, 1923) or Forbidden Paradise (Lubitsch, 1924). The remains are scarce, but they can be made to function with the information that has survived. Someone else working with the same material might have found a solution different from my own. The advantage of using music and film from the same epoch is that they both evoke the period from which they came.
With due respect to those concerned about elitism, I began with some of the remaining orchestral music because the orchestral tradition represented the deluxe cinema practice of the era. It was an important force in the “gentrification” of the moving picture and continued to be important to expressivity throughout the recorded-sound period, so there is no reason to downplay its significance. My dwelling on the orchestral tradition, however, does not deny the existence of improvised accompaniments, the use of the theatre organ, phonograph record, silence, or poorly executed accompaniments, which I have described elsewhere.  But the professional, closely synchronized orchestral tradition in the mute film era seems to have had more influence in the recorded-sound film era than did all the informal practices.
I have now restored, reconstructed, or had scores created for over fifty mute moving pictures. Some required reorchestration, some merely the refurbishing and synchronization of the orchestral parts, while others required a combination of reorchestration and the location of the original arrangements used for the compilation of the preexisting music. Still others necessitated the creation of a score from a cue sheet, or a hypothesis about the order in which to put the extant suites that had been extracted from a now-lost original.
My overall aim has simply been to perform this music to find out what I could learn, and then to join what I thus learned to the information preserved in other sources. To make this combination has been time consuming and tedious, but not especially difficult. In certain cases, it brought the original work to life and revealed an even greater masterpiece than expected. While avoiding generalization, I propose only the orthodoxy that we learn some important information by attempting to recombine original orchestrations with their orphaned images. The public reaction to my performances indicates that most of the films do indeed work with their original scores. (The score by David Mendoza and William Axt for King Vidor’s 1926 La Bohème is the exception.) I was and still am motivated by simple curiosity about the role the music played. Many people have written histories of mute cinema and movie music without experiencing the images with their original accompaniments. Yet in many cases the music transformed the pictures (and vice versa) and I wanted that firsthand experience in order to understand the dynamic process involved in the transformation. Eventually I would like to see a cinema history that describes both how the music engages the ears and the images the eyes.
National Gallery of Art Auditorium Screenings/Concerts
Early on, I was exceptionally lucky to have as a performance venue the National Gallery of Art Auditorium and ten members of the National Symphony Orchestra with whom to explore the repertory in actual screenings with the moving pictures. For several years on Saturday afternoons in July, we tried one original score after another with its moving picture. The attempts were very well received in the press, and there were long lines to enter the auditorium. Those performances convinced me that the carefully cued scores were meant to be precisely synchronized with the images, a conclusion that was confirmed by frequent mention of “well synchronized scores” in the press of the 1910s and 1920s. Furthermore, that kind of synchronization could be obtained with a lot of practice and by using one’s eyes rather than the modern Hollywood click track. Those summertime performances eventually ended but were followed by other performances intermittently, including a number with newly commissioned scores constructed according to the principles used during the mute film era. 
The Thief of Bagdad
(Raoul Walsh, 1924)
My first attempt at an orchestral restoration was an original score by American composer Mortimer Wilson for the Douglas Fairbanks film Thief of Bagdad (1924).  The piano conductor score and all the orchestral parts existed. The only things required were to correct all the errors and refurbish the parts so that modern instrumentalists would not revolt because the parts looked so different from what they were used to.
A French horn player in my first performance told me that her part was technically the most difficult she had ever seen. There were long, repeated solos and, to accompany the magic flying horse, a sort of “Ride of the Valkyries” for French horn. Wilson clearly had a very talented soloist in mind, a pretty impractical idea when you realize that the score and parts had to tour all over the country after the premiere, and few horn players could have mastered the part. Clearly this score required many rehearsals, one of the disadvantages in using an original score, which Wilson later admitted.  For second-run theaters, the Wilson score was replaced by a cue sheet.  The deluxe and second-run theater practices coexisted but the synchronization in Wilson’s score was very clearly marked, and certainly realized when the composer was conducting (at least eight months in New York and several more in Boston and other cities).
The Thief of Bagdad set a new high-water mark for film accompaniments in the United States. Fairbanks was trying to make the film a prestigious “silent opera” and the picture’s musical accompaniment one of its main attractions.  His every move in the 1920s was calculated to elevate the quality of his motion pictures and to bring a new, genteel audience to their exhibition. However, his producer for the New York premiere, Morris Gest, tried to have the original score replaced by one totally compiled from preexisting compositions. Wilson claimed that after some anxious moments, his score was restored. 
D. W. Griffith’s Way Down East (1920)
and Intolerance (1916)
My next two restorations were the scores for D. W. Griffith’s Way Down East and Intolerance, both known to be a part of the D. W. Griffith Papers at the Museum of Modern Art (MoMA). In both cases the music came complete with minutely well-synchronized piano-conductor scores and all the orchestral parts. Both called for silences, a practice unfamiliar to most mute film experts but considered the sign of a high-class work.  Theoretically, the refurbishing and synchronizing of the parts would have been the only actions required. But the restoration work on Griffith’s Way Down East gave MoMA the idea of using the original score for his Intolerance to reconstruct the premiere version of that radically ahead-of-its-time movie. Griffith had made many changes in the picture after its premiere and the score no longer matched any of the surviving prints, but it did provide a blueprint for something close to the movie’s original organization. Thus was initiated my most controversial work, the MoMA/Library of Congress reconstruction of the moving picture Intolerance using the original score by Joseph Carl Breil as a guide. I will only partially rehash the controversy it caused here but urge readers to consult what I already have published.  Music was as important to Griffith as his images, but without the MoMA reconstruction, the function of Breil’s score would have remained a mystery. Once restored, one could see that the decision to have different music for each of the four stories contributed to the movie’s lack of cohesion.
While many found the reconstructed film revelatory, some viewers in the audience at the New York and Pordenone, Italy, premieres clearly were shocked. They hated the reconstruction, especially the look of the altered still frames (used to designate missing movie footage) that took the audience out of the story and demonstrated all the changes that Griffith had made since the film’s premiere (especially to the famous long tracking shot in the Babylonian sequence). Audience members also hated the long silences, some of which were documented by printed instructions in the score. In fact, after seeing a silent screening of the reconstruction at MoMA, Lillian Gish, the principal actor, said she preferred the shorter edited version. The New York Times called the reconstruction “hard to watch” but “a triumph of scholarship, sleuthing, and dedication.”  At Le Giornate del Cinema Muto, the popular film festival in Pordenone, it received a fifteen-minute standing ovation.
Thereafter, however, dislike was clothed in scholarly vestments that appeared to subvert the reconstruction’s underlying assumptions and implied that MoMA was out to destroy Griffith’s reputation. The “scholarly” documentation disintegrated under close scrutiny and many of the issues raised were countered.  The only criticism that stuck was the fact that the reconstruction may have been projected at too slow a speed. The critics, though, were still left with their value-laden dislike of what was in effect a scholarly edition of a motion picture. As one filmmaker told me, Griffith would have hated it because the disfigured still frames kept the movie from moving. Nevertheless, MoMA’s focus has always been on providing the means for study, and the reconstruction revealed many facts about Griffith’s and Breil’s actions and intentions, the foremost perhaps being the importance of music to Griffith and his tendency to constantly change his motion pictures.  Certainly, the work on Intolerance is a very special case. For its reconstruction, the complete Metropolis (Fritz Lang, 1927) used the original score by Gottfried Huppertz as a guide (Metropolis 2010), indicating once again that such scores do contain a lot of useful information.
Ben Hur (Fred Niblo, 1926)
David Mendoza and William Axt’s score for Ben Hur involved still different challenges for reconstruction.  The original music had to be reorchestrated and the arrangements of preexisting music they utilized had to be found, photocopied, cut, and pasted. I had to construct a detailed map of each section of the score, noting how many measures to reorchestrate or to take from preexisting material (sometimes partially cut or rearranged) that had been identified and located in music libraries, both private and public. The cues were very precisely laid out, implying the same careful synchronization practices used by Griffith. The same was true for the score composed and compiled by J. S. Zamecnik for Wings (William Wellman, 1927),  the Wilson score for The Black Pirate (Fairbanks, 1926),  the Hugo Riesenfeld score for The Covered Wagon (James Cruze, 1923)  and the compiled scores for The Circus (Charles Chaplin, 1928)  and The Gold Rush (Charles Chaplin, 1925). The scores for The Circus, The Gold Rush and Woman of Paris (Chaplin, 1923) are held by the Chaplin family. It was with their kind permission that the Library of Congress was able to present The Circus at Wolf Trap National Park for the Performing Arts and The Gold Rush at the Kennedy Center.
Häxan (Benjamin Christensen, 1922)
By contrast, the only information about an original accompaniment for Häxan came from a list of musical pieces mentioned in a cinema program the week after its Danish premiere at the Palads Teatret in Copenhagen.  James Luke and I started with the assumption that the list was complete and in the order of its use. Indeed, the music seemed to work that way for us but our solution has to be considered merely a hypothesis.
Häxan used complete overtures or a whole movement of a symphony, all preexisting pieces written well before the movie was made. They fit the picture like a glove even so and our hypothesized long sections seemed to be a repeat of the pattern established in the score for The Passion of Joan of Arc. The way the music worked with the images called my attention to musical and visual logic and to the fact that to work well, long musical excerpts have to be based on the same abstract concept as the images. I tested this idea by creating scores for Pandora’s Box (G. W. Pabst, 1928) using music by Dvorak, Tchaikovsky, and others; Peter Pan (Herbert Brenon, 1924) using music by Mendelssohn and Haydn; and El puño de hierro (Gabriel García Moreno, 1927) using Albeniz’s Iberia.
However, underlying concepts and conventions may have changed. Documentation of this process of change is one of the fascinating opportunities that restoration and reconstruction of mute film scores present. Today Schubert’s Rosamunde Overture is associated with a trip out into the countryside but was used in Häxan (hypothetically) to accompany the witches flying through the night. Underscoring the black mass in Häxan with Max Bruch’s Kol Nidre would be considered anti-Semitic today and synchronizing the aerial warfare in Wings with the fairy music in Mendelssohn’s Midsummer Night’s Dream is not what we would expect, either.
Nosferatu (F. W. Murnau, 1922)
For Murnau’s Nosferatu, two sets of Hans Erdmann’s suites, extracted and arranged from the original movie score, were all that remained of the music.  A version of the suites with all the orchestral parts was deposited for copyright at the Library of Congress. At the head of each of the ten numbers was a verbal explanation of the situations for which each number would be appropriate. Erdmann and Giuseppe Becce’s Allgemeines Handbuch der Filmmusik provided additional clues.
Suite IIB referred to a tune for an inn (there was only one in the moving picture) and a footnote referred to an ABA form with coda that happened at the end of the movie, defined as Suites IE, IIC, IE, and IB. The A section was clearly Nosferatu’s theme, the B Renfield’s chase, and the Coda Ellen’s theme. Berndt Heller has used a ten-instrument version of the Erdmann suites in a very different way, the ending clearly not what Erdmann stipulated but the rest indicating how different these interpretations can be.
Rosita (Ernst Lubitsch, 1923)
My reconstructed score and MoMA’s restored print of Ernst Lubitsch’s Rosita starring Mary Pickford received a standing ovation at the opening of the Venice Film Festival in 2017. The score used only the preexisting publications called for by a 1923 cue sheet,  which contained a list of preexisting music publications considered appropriate for each of the scenes in the movie, defined by their initial intertitles or actions (the “cues,” hence the term cue sheet). I was able to locate all but one of the 45 specified orchestrations in some of the enormous music libraries collected by cinemas and theaters that have been preserved and catalogued. Such music libraries were the backbone of all the music making in cinemas around the world.
The musicians who performed the accompaniment told me they thought the score for Rosita was beautiful. It fit the film perfectly, raising the possibility that Lubitsch, himself a good pianist, might have had a hand in its selection. Contrary to those who consider cue sheets and preexisting music to be of lesser value than original scores, Rosita’s soundtrack was proof that preexisting music, when well chosen, functions just as well as a commissioned original score.
Rosita’s moving images had taken years and years to restore. MoMA had gone through several restoration programs (having to start at the beginning each time) and it would be safe to say that the staff had seen the picture hundreds of times, always silent. After the dress rehearsal, Dave Kehr, a curator of moving pictures at MoMA, made my day. “It is a completely different picture with that music.”
In many mute film scores, each cue was printed over the beginning of a limited number of measures. The composer or compiler apparently wanted that music to occur with those images or intertitles and to last until the next cue, which implied some arithmetic: How to get so-and-so many measures at such-and-such a speed to last X number of seconds? Cue sheets were also tied to duration. Synchronization was implied, but was it realized? My research showed stopwatches being used to synchronize mute-film scores in Europe and the United States, and many reviews specifically mentioned and praised “synchronized scores.” 
Precise synchronization was one of the characteristics that made D. W. Griffith’s almost three-hour Birth of a Nation (with Joseph Carl Breil’s original orchestration) such a sensation, since it was a novelty in 1915.  In the very same year, Hugo Riesenfeld also performed a more loosely synchronized score for DeMille’s Carmen, starring internationally famous opera singer Geraldine Farrar.  Most of the new scenes were cued at the music’s head but ended with the instruction “Play until [the next cue].” Griffith established a new deluxe practice with his close synchronization, but both DeMille and Griffith were attempting to attract a genteel public. DeMille used Geraldine Farrar, the Boston Symphony Orchestra, and a synchronized rendition of the Bizet score; Griffith employed a closely synchronized, sumptuously well-orchestrated score that evoked the prestigious pantomime and pantomime ballet.
Griffith claimed to be presenting “a pantomimic screen spectacle” but what was the connection between pantomime, pantomime-ballet music, and the cinema?  Griffith was only one of many who considered the motion picture to be a form of pantomime—pantomimes being dependent on live synchronized music. As in cinema music, pantomime music called attention to the facial expressions, actions, and body movements of the mimes. Pantomime turned out to be an unexpected source for deluxe cinema music practices that until recently were considered more an offshoot of melodrama. 
Through my research, I realized that a lack of attention to the music and lack of experience matching scores to films had led to an error in cinema histories: synchronized sound had existed prior to the arrival of the so-called talking picture. Recorded sound brought into being only a mechanical type of synchronization; the synchronization that preceded it, having been achieved live, could arrive a little before, exactly on, or a little after a cue. In other words, it could “breathe.” But the synchronization was still perceptible and most certainly existed.
As I said at the beginning of this essay, the difference between mute- and recorded-sound films was not the synchronization of the sound but the live or recorded quality of the sound, and the synchronization of the talking with the image of the actors’ lips. Live synchronized music added weight to the two-dimensional images, the added weight helping to create the illusion of their three-dimensionality. The music made them seem more natural but sometimes also, indeed, more epic. The synchronized scores highlighted scenic beginnings and endings, changes of mood and tempo, and changes in action or expression. They even caught the tempo of actions within the frame. They had an explicit or implicit “pointing” function, calling attention to specific actions or expressions. The live quality of the instrumental sound itself, experienced in the three-dimensional space of a deluxe cinema, also helped to create the illusion of three-dimensionality, a predecessor of surround sound. These are functions that the music still performs with moving images today. When you realize this, surprisingly, the difference in function between music for mute films and for recorded-sound films disappears. This realization was indeed what Charles Rosen described. Whatever the drawbacks to creating and performing these translations, “our knowledge has been increased and our musical life enriched.” 
– Gillian B. Anderson