Are Historical Recordings Reliable?

Are historical sound recordings a trustworthy source of performance practice information and insight, or can their limitations be so severe that they are able to misrepresent the music made in their own time? Could the circumstances of the recording process and their impact on the performers have potentially inhibited the music captured in the grooves or does any of that performativity-type consideration really matter? 

Edward Elgar leading a recording session in 1914

Many discredit historical recordings upon initial encounter due to a visceral reaction against their audio fidelity limitations, especially 78s of the acoustical age. Because these listeners are not able to hear the full palette of sound that not only replicates, but supersedes, the live listening experience (as is the case with modern studio-produced recordings), they put up a mental blockade against listening into the grooves for what is actually there. They often will complain that the recorded sound is cold, distant, unrealistic, and artificial. And that automatically translates into untrustworthy.

Others, perhaps less bothered by the lack of audio fidelity, will cite (and even contrive), deficiencies in the recording process to discredit the music being heard. Probably the most common of these is that the music was played or sung by the musician(s) too quickly, so that a work would fit on a time-limited disc side, especially for 10-inch 78s or those, such as “Berliners,” that were even shorter (7 inches). Other complaints include that the musicians were often under-rehearsed, that the recording cone misrepresented vibrato, that singers especially sounded too shrill in a way that could not be at all realistic, that the recording room conditions encouraged bad playing, and so on. (Interestingly, surface noise seems to be less of an issue now than, say, a decade or so ago, due to the resurgence of interest in LP disc recordings, whereby the non-musical pops and crackles actually are welcome as part of the sonic ambience.)

Complaints with these recordings were not invented in the modern era. Musicians and listeners had mixed reactions to recordings back in the late 19th and early 20th centuries as well. Wildly popular among the masses, 78s caused skepticism among many musicians of the time. These musicians considered the gramophone or phonograph a novelty item or toy for dance music and the like, not a serious mechanism to convey classical music (Symes, 2004, p. 35, 38-40). They also often bemoaned the conditions under which they had to record (not unlike bemoaning having to record today). There is no doubt that the process of recording, whether one hundred years ago or at this very moment, alters and affects the mental state of the musicians being recorded. Knowing that an audio document of a performance will exist indefinitely for others to hear in a variety of circumstances, as opposed to the performed sound disappearing into the air, can have impact on the musical decisions being made and how they are rendered, both consciously and unconsciously. But that does not automatically invalidate those musical decisions in either circumstance.

Musicians such as Salvatore Fucito, who was Enrico Caruso’s coach and pianist, took a more recordings-positive approach, however (Caruso, after all, was the first person to sell over a million recordings). In Fucito’s 1922 book entitled, Caruso and The Art of Singing, he writes:

One need only compare [Caruso’s] singing of La donna è mobile or the Racconto di Rudolfo with his singing of Stradella’s Pietà, Signore or of Halévy’s Rachel: quand du Seigneur la grâce tutélaire – to mention only a few of the many superb phonographic records of Caruso’s great art – in order to be convinced not only of the larger attributes of Caruso’s style, which made his art unique, but of its unusual variety in expression, [and] its flexible representation of his moods as they responded to the significance of the music he was singing (p.187-188).

Fucito, in other words, considered Caruso’s recordings (played on 1922 phonographs no less) to be a sufficient means of absorbing and understanding the various crucial dimensions of Caruso’s artistry. 

Achieving a high-fidelity audio experience is therefore not necessary toward gaining performance practice information from the grooves of the historical recording (even the scratchiest, noisiest early cylinder or Berliner will yield some sort of limited information). Having superior audio engagement that electronically surpasses the sound of the concert hall is not the purpose; hearing the performance elements captured on the recording, as many as possible and by the best means available, is. The listener does not need to be enraptured by the sound quality, or even have all of the elements of reality present, to study various stylistic features in the recorded performance that can be compared to the stylistic features of other recordings of the time, as well to music as it is performed today. Of course, the playback conditions should guarantee that the historical recording is heard as truly and completely as possible, which includes being played either by a reliable high-quality transfer, or on a superior audio system with a turntable at the right speed with a good stylus (see Bailey, ”Historical Recordings and Performance Practice Listening). Listening for the sake of performance practice insight is about gathering and assessing audio information however it presents itself, not about prioritizing or insisting on an aesthetic experience while doing so. That is to say, how the listener reacts to the recording overall is secondary to the performance practice information the recording reveals, whatever and however much is there.

The aria “Magische Töne” from Goldmark’s Die Königin von Saba, through various recordings, illustrates the point. In this first example, tenor Nicolai Gedda provides a remarkable recorded performance enhanced by full sonic audio fidelity. His rendition is intimate, personal, and deeply moving; all the elements of sound are in place. This recording understandably leaves many listeners in awe, and his performance practice stylization is clearly evident. 

In contrast, Leo Slezak’s 1905 recording of the same aria comes nowhere near to matching the audio fidelity mastery of Gedda’s (including the fact that an upright piano is used for the accompaniment, rather than full orchestra). But the listener should put that aside and focus on Slezak’s voice and his phrasing and articulation. Really focus. There is abundant performance practice information apparent – phrase shape, musical contrast, diction stylization, manner of declamation etc. – enough to generate pages of notes on Slezak’s manner of singing this aria.

Hermann Jadlowker also recorded this aria around the same time as Slezak. Listening again with razor-like focus to his voice and its expressive details yields abundant performance practice information, especially in comparison to Slezak. His use of portamento, briefly, toward the beginning of the aria is in and of itself a notable detail of contrast (0:48 - 0:56) compared to Slezak at that same spot. 

With the high-fidelity nature of Gedda’s recording, the listener may sit back and take it all in. With the historical renditions of Slezak and Jadlowker, the listener must lean in and listen with uninterrupted focus and attention, and both recordings will provide ample performance practice musical information, achieving magnificence in their own right. [The youtube links are posted out of convenience, but with some reluctance. Ideally one would listen to all three recordings under the circumstances described in the previous blog posting. Nevertheless, these three fairly well-done transfers should sufficiently convey the point being made, while encouraging the listener, whenever possible, to use higher quality transfers.]

What then of the contention that performers played music faster than normal to fit a work (or a segment of that work) onto a side of a disc? There is ample evidence to show that, as a universal rather than anecdotal contention, this notion lacks foundation. The genesis of the comment is a reaction against the tempos heard on historical recordings, which to some seem faster than normal. Yet the vast majority of recordings, including those that are 10 inches and smaller, still have 20-30-40 seconds left at the end, even a minute in some cases, meaning there was more room on the disc (and even 30 seconds more is quite a bit of time in terms of tempo for a short work or excerpt). The engineers overseeing the recording process also were experts in terms of timing and were particularly exacting in those calculations. For instance, many recordings have cuts, large and small, that enabled the piece or except to fit just fine. Furthermore, this contention has been made over transfers, or the 78s themselves, that actually were being played at the wrong speed, as very many recordings need to be played slower than 78 rpms. This even applies to radio transcription discs of the 1930s and 1940s. Note the following transfer of Pablo de Sarasate playing his Zigeunerweisen in 1904, which is a whole step too high. This also means the tempo is too fast (his quicker passages sound almost inhuman). 

Again, especially in 1904, it was a common occurrence that a recording needed to revolve slower than 78 rpms (sometimes faster as well). Here is a transfer of the same recording much closer to the right speed and therefore the tempo Sarasate actually played.

The transfer at the wrong speed plays for 4:46 and the one at pretty much the right speed for 5:37. Listening to the faster transfer at the wrong speed shaves almost a minute off the performance and of course sounds excessively fast, while the pacing of the phrases is rushed and unnatural. This is why phonographs had speed adjustment. Some musicians also tend to slow down in recording sessions for a sense of caution and are urged to go faster to realize their normal tempo. If other musicians were performing faster than they were used to, which certainly could have been the case in specific circumstances, then it is likely to be apparent in their manner of playing. If, on the other hand, they pulled it off, that also says something about performance practice, all of which becomes part of the listener’s assessment process.

But what if musicians, in certain circumstances, had to endure less than ideal, or even sometimes arduous, recording conditions? What if the room was cramped and hot, the sight lines were off, the spacing was awkward, timing was an issue, and the process was fatiguing? The answer to all this is that it doesn’t matter. Again, It doesn’t matter. The only thing that is concrete and stable is the sound in the grooves, however it got there. To try to factor in any extra-musical elements can only amount to conjecture further colored by modern bias, because there is no way actually to determine just how much these factors actually had impact on the sound of a particular recording, if at all. This is not musicology. This is listening closely to, and assessing, the actual sound documented on the disc. And yes, once we reach the age of radio broadcast recordings, then it is instructive to compare those live studio performances (or Metropolitan Opera broadcasts) with the ones released on commercial 78 rpm recordings. But, again, one does not invalidate the other, it only shows the range of possibility. Who is to say a musician was any less so in certain circumstances, especially to validate bias? The sound captured on the recordings is the only actual music under consideration, and there is plenty there to ponder, compare, and contrast. (I especially appreciate that in J. B. Steane’s The Grand Tradition: Seventy Years of Singing on Record and his Voices, Singers & Critics, he never disparages historical recordings nor cites any of their “deficiencies” as factors in his understanding of the vocal sound he assesses on the recordings themselves.)

There are certain factors regarding the recordings themselves that can have impact on the way the music is heard. For instance, occasionally the apparatus in the acoustical age would wind down in process. This caused a gradual reduction in pitch and tempo, yet some of those recordings were released commercially nonetheless. The A pitch was not standard; in various places it was lower than 440 and in other places a bit higher. Soprano Nellie Melba, for instance, preferred to sing and record at A-435. Singers also transposed arias from time to time to suit their voices, which did not always have to do with their age. Caruso’s 1906 recording of “Che gelida manina” from La Bohème, for example, is a half step lower by intention. Vertical cut 78s by Thomas Edision et. al. require special playback settings, and some recordings were made with groove distortion in louder passages. Reputable companies that make transfers of historical recordings will take all of this into account. When listening directly from the turntable, however, these are things to keep in mind.

Romantic-era performance practice and musicology (especially when musicology is concerned with performativity) may intersect in some musical studies, but they are not the same field. And the concepts articulated in this essay are a response to the latter, through numerous comments in numerous settings may serve to inhibit or even discredit the former when discussing historical sound recordings. The goals of these two disciplines are different as well. Performance practice on record is not endeavoring to recreate the world in which the music was written – an endeavor laced with traps that, arguably, the Early Music movement fell into toward the beginning – nor is it proposing simply to copy the performances on those recordings today. It is rather about enlivening the modern ear with a multitude of sounds generated in a different era, and, through close listening and comprehension, attending to the details that carry and convey those particular sounds as music. Nothing in the understanding of the conditions of how that sound came to be will alter the recorded sound itself, and any influence those conditions may or may not have had will be conjecture. When the topic is musical style and performance as heard on disc, as it is here, the grooves are all that matter.

References

Fucito, S, Beyer, B. J. (1922, Dover 1995). Caruso and The Art of Singing.

Symes, C. (2004). Setting the Record Straight: A Material History of Classical Recording.

Historical Recordings and Performance Practice Listening

Listening to historical sound recordings so as to absorb the most musical information imparted in the groves requires a few levels of consideration and setup. Quite often there is more music on the recording than the listener is hearing because of some less than adequate factor in playback. This not only can cause the listener to miss valuable aspects of a performance on record, but even to misinterpret or misunderstand what is being heard. 

Although all romantic music performance practitioners are welcome to hear recordings at a place such as the Yale collection’s studio, the feasibility of that for so many is obviously limited. Most people, rather, listen to historical recordings remotely. In such cases, there are certain things to keep in mind and account for to hear everything on the recording.

Five types of historical recordings

There are five basic groups of historical recordings that are of particular value to the romantic music performance practitioner: acoustical recordings, electrical recordings, transcription discs, test pressings, and private recordings. These categories by and large exist on disc formats (with the exception of cylinders), while more recently, LPs, CDs, and digital files provide for historical reissues (while formats such as reel-to-reel tapes, cassette tapes, and DAT tapes are outdated). Here is each group summarized using an historical lens:

Acoustical recordings appear near the end of the 19th century in two contrasting formats: cylinders, which were associated with Thomas Edison, and flat discs, which were associated with Emil Berliner. In the first decade of the 20th century, Edison experimented with flat discs and eventually yielded to them. Acoustical recordings were produced mechanically and characterized by the use of the cone (or “horn”) to capture sound vibrations in the recording process with a cutting stylus responding to those vibrations, making grooves in the master recording. Generally these discs existed in 7-inch, 10-inch, and 12-inch shellac formats, single and eventually double sided; most were laterally cut (in terms of the grooves), but some were vertically cut as well. They generally played at 78rpms, though often, depending on the record label especially, these recordings could play slower or even faster. Acoustical classical recordings favored singers accompanied by pianos or instrumental ensembles, but also included solo instrumentalists, and eventually works for groups of musicians. The ensembles, both solo and accompanimental, were reconfigured according to the recording circumstances. Much smaller in size than a full orchestra, these particular multi-instrument recordings featured arrangements that put the tuba on the bass line and transferred many melodies to the clarinet, both of which were more easily picked up by the recording horn. Innumerable acoustical recordings were made and sold throughout the world. An artist such as tenor Enrico Caruso, 1873-1921, is heard only on acoustical recordings.

An Edison cylinder and 7-inch Berliner

In the early 1920s, electrical recordings began to take over. Instead of a cone, microphones were used to capture the music, and electricity was applied throughout the recording process. The most noticeable difference was the enhanced audio fidelity that resulted. As well, regular orchestras could be used with no need to transcribe the instrumentation for smaller, more specialized ensembles. At this point, 10-inch and 12-inch shellac double-sided recordings were the format of choice. Several artists, such as pianist Sergei Rachmaninoff, 1873-1943, and soprano Amelita Galli-Curci, 1882-1963, recorded both acoustical and electrical recordings, while many others younger in age only recorded electricals. 78 rpms was the preferred speed, though many recordings still played faster or slower.  

In the late 1920s (and especially expanding into the 1930s), radio broadcasts of live performances and programs could be captured on electrical transcription discs. These recordings, usually 12-inch, 14-inch, or 16-inch, were distributed for rebroadcast, but also could serve as keepsakes for the performing artists. The Metropolitan Opera began recording its broadcasts in the mid 1930s. Special events such as world fair concerts were even recorded outdoors, as in the early 1940s. The discs generally played at 78 rpms early on, but eventually could also play at 33 rpms. It is not uncommon to encounter a transcription disc that plays with the stylus starting at the label and going to the outer rim. Most discs were made with lacquer, which can exist over glass or aluminum in some cases, though there exist fully aluminum transcription discs as well.

Lacquer peeling off a glass base

Test pressings are another beneficiary of the electrical age. These are unpublished takes of a recording session by an artist or group of artists. Before splicing and editing was a possibility, each recorded version of a piece could be distributed to the executives and performers to decide which one to publish. They were produced for temporary use, similar to transcription discs, which is why many have deteriorated beyond playability. Numerous others still exist, however, and reveal a great deal about performance artistry, including consistency and inconsistency from one take to the next, common or uncommon mistakes, and so on. These play at the same speed as the released commercial version, often 78 rpms, though sometimes a bit slower or faster. The Yale collection, for example, has several test pressings in the late 1920s of pianist Moriz Rosenthal, 1862-1946, playing works by Chopin and Liszt, some of which were published and others that were not. Early test pressings, often called “acetates,” were made like transcription discs, and later on (past the point of this discussion) they were made of vinyl for LPs.

Similar to test pressings, private recordings are unpublished. They also can be extremely rare and hard to obtain, the most difficult to find of any of the other recording types mentioned. Whereas they may contain music, as often they include interviews, family messages, discussion at social gatherings, special announcements and greetings, field work, and an array of informational snippets. These are not only historical but still are being made in abundance. Those older recordings that are on disc generally are made of lacquer and were designed for impermanence (even if they have survived to today), mostly playing at 78 rpms or thereabouts. Very likely the contents on these recordings will be one of a kind.

The listening setup

The easiest way to listen analytically to historical recordings is to rely on expertly made digital transfers (in WAV format) by reputable recording engineers and companies, such as Marston Records (Marston audio transfers currently are only available in CD format, but soon will be downloadable). Extraordinary care and research goes into making these transfers. Also, perhaps a step down from WAV file transfers and CDs, are the high quality streaming files provided by reputable sources such as The Great 78s Project (curated by George Blood) and the files posted on a site such as the Discography of American Historical Recordings (DAHR). Yet another option are those institutions such as Yale that will make digital transfers of 78s on request (Yale provides 30-day streaming access, though other arrangements can be made under special circumstances).

Youtube videos, and the like, can be extremely unreliable, depending who made them. At the very least, the audio quality will be reduced (unless heard on a different platform such as Bandcamp, which maintains higher-quality transfers). These publicly posted, largely homemade, transfers can be fraught with problems, such as the wrong listening speed and pitch (which is also a problem on certain LP reissues), excessive noise reduction (which also diminishes the musical sound), the wrong stylus, dubious audio quality, original ensembles replaced by modern orchestras (which happens a lot with Caruso transfers), zero noise reduction where some is merited, and much more. All these things can blur and lessen the sound of articulation, diction, and other stylistic effects. These recordings may be sufficient for basic entertainment, but not for analytical study.

Listening to historical recordings in order to extract the most detailed musical information from them requires either high-fidelity speakers and/or earphones (the latter being the less expensive option). Laptop built-in speakers are insufficient, as are average desktop computer speakers. Earbuds in most cases also are not enough, as opposed to earphones that cover and insulate the ears.

Audioengine, for instance, makes quality laptop/desktop speakers that will serve the digital files listening experience well, including the HD6, A5+, and HD3 models, which support tabletop listening (all three speakers have amplifiers of varying specifications built into the left speaker). These speakers are also ideal for giving presentations in seminar rooms (the HD3s are especially portable). There are other companies that make similarly impressive speakers as well.

Audioengine A5+ left speaker

To upgrade the audio quality beyond computer speakers, a Digital-Analog Converter (DAC) will allow digital files to play on an analog stereo system. This is ideal for close listening, assuming a high-quality audio setup. A variety of DACs are available at a range of prices (the best being just over $1,000 US), and, once installed, they will require nothing more than the turn of a knob or pressing the right button.

Digital-analog converter (DAC)

If the romantic music performance practitioner desires to listen to the source recordings themselves, this laudable and perfectly possible goal becomes more detailed and complex. It also goes beyond the scope of this essay other than a few suggestions and observations. The stereo system will require an amplifier and preamplifier or a high-quality integrated amplifier with turntable possibilities (this should not be an issue). McIntosh is an undisputed leader in audiophile amplification components, though there are less expensive options. Ideally there would be two turntables, one set up for LPs (and 45s), the other for 78s – otherwise, as 78s and LPs use different cartridges and styli, the listener will frequently have to switch one for the other. Also the turntable dedicated to 78s must have notch by notch speed adjustment and be able to accommodate 16” transcription discs for the sake of some radio broadcast recordings. Of course, so as not to place wear on the 78 if frequent play is expected, an audio component that allows for the digitization of the recording will be necessary, either onto CD-R or to a digital file. This also means obtaining the recordings themselves and building the collection, which is an endless task (though it can be done in conjunction with collecting digital files and LP or CD reissues). 

Audio components, including McIntosh pre-amplifier and amplifier, with two turntables

Turntable setup for 78 rpm recordings

Head shell, cartridge, and stylus for 78 rpm recordings

As a final note, it is highly instructive to listen to 78s played on original phonographs or gramophones (though the discs themselves should be copies and not valuable, since repeated playback will cause wear). Hearing the 78s as they did at the beginning of the 20th century creates a strong connection and even bond with that period. It can open further avenues of awareness and understanding, but it does not provide the means for close, analytical listening for the sake of performance practice study. Modern and sophisticated audio components are best for that.

Early 1920s non-electric phonograph

Thoughts on Performing Romantic Music

To perform music of the romantic era in a manner that accounts for the expectations, conditions, practices, and desires of the time is crucial. These are the ingredients that serve to unlock the full vitality, range of nuance, and scope of expressive possibility in a musical work and its performer(s). If not, approaching the music from an unfiltered modern lens paves over numerous stylistic practices that were in place to help enliven preparation and performance, leaving the music only as a partial reflection of the range of artistic and expressive possibility embedded within.

Pioneers such as Clive Brown, through his innumerable contributions to the field, make this self-evident, alongside the rediscovery of period reviews, essays, instruction manuals, journal entries, correspondence, annotated parts and scores, and historical sound recordings.

Clive Brown

George Kennaway, Neal Peres Da Costa, Jonathan Bellman, to name just a few, have also paved the way – and continue to do so – toward discovering romantic music on its own terms with full consideration given to the period. In the field of opera, Will Crutchfield and Ward Marston top the list alongside others. And when it comes to newer generations of musician-scholars pursuing this realm of study, consideration, and performance, the list of participants and contributors is constantly expanding, too numerous to summarize here. Performing romantic music is now under a new light that serves to re-enliven the foundation of thought and creativity that makes this multi-varied repertoire so compelling.

There is an assumption, however, that romantic-era musical performance practice was created under the umbrella of the early music movement and, in many ways, is simply an extension of that monumental effort and result. While many practitioners in the field of romantic music performance also are active in early music (this author included), there are distinctive factors between the two. In fact, many of the fundamental features of romantic-era style, such as frequent or regular tempo fluctuation, upward and downward portamento, unaligned right and left playing for pianists, and so on, are nowhere present in the vast majority of performances and recordings of period early music. Therefore, these factors of the romantic era have tended to be absent when the same musicians perform romantic music, even on instruments of the period. The result is that the music sounds like period baroque or classical repertoire with different harmonies and melodies, and little more than that. While some of the methodologies and incentives for performance practice study are shared between early music and romantic-era endeavors (and are even essential to both), romantic music performance requires a fundamentally different consideration of style, even intent, such that coming into it directly from the early music movement does not initially seem to yield. But this too is starting to change.

Historical sound recordings assure us that applying the many forgotten, reduced, and partially neglected stylistic manners of performing romantic music does not create performance clones, but highly individualistic and self-expressive musical renditions. After all, historical recordings indicate that orchestras under Arturo Toscanini and Willem Mengelberg sounded significantly different; the piano students of Franz Liszt performed their teacher’s compositions in noticeably contrasting ways; the voice students of Mathilde Marchesi embraced highly individualistic manners of singing the same repertoire, and so on. Therefore, romantic music performance practice is guided, but not restricted. And grappling with old possibilities in a new setting is the project at hand.

Willem Mengelberg

Arturo Toscanini

The current status of romantic music performance practice is therefore enveloped in discovery. The field is constantly unearthing new layers of musical information to consider and enable. The thousands upon thousands of historical sound recordings alone, available for consultation, can serve as an invaluable source of performance practice information. And with so much to learn, there is perhaps almost as much to unlearn. When, for instance, a listener hears in an old recording the soloist very slightly rush ahead of the orchestra toward a cadence or significant musical point of arrival, this does not automatically indicate inferior musical coordination, but more likely is part of the expressive intent of the artistry to generate an extra sense of excitement. And how this translates into historically-informed performance today is still a matter of great conjecture and even experimentation. But with each new project, discovery and rediscovery interact with greater vigor and meaning. The excitement this generates is palpable. Admittedly this will take a while, but the results, as they have been so far, promise to yield greater musical possibilities and the ability to hear romantic music, true to itself, in a

new and vibrant way.

Newer generations: Octavie Dostaler-Lalonde (cello)
and Artem Belogurov (piano)