Expressive tempo modifications in early twentieth-century recorded performances of operatic arias

by Massimo Zicari

Music & Practice, Volume 4

Listening to early recordings

Research that has been conducted into responses to early recordings suggests that uninformed listeners and even music students still tend to disregard these and respond with incredulity to the interpretations which they reveal.[1] This is especially true when it comes to discussing such an interpretive device as portamento, which is often viewed with suspicion,[2] or the addition of ornamentation and cadenzas, which has long been viewed by the layperson as an act of abuse committed by the performer at the expense of the composer.[3] Furthermore, uninformed listeners are prone to react critically to early recordings based on the a priori assumption that interpreters nurtured in the late romantic culture abused the composer’s intentions and indulged in tasteless, exaggerated interpretive choices.[4] Regretfully, similar reactions can also be observed among some music practitioners. In exceptional cases, voice students are even forbidden to insert portamentos, to modify tempos or to insert new cadenzas in the arias they are learning on grounds of bad taste and improper style.

On the other hand, it is now common opinion among music practitioners, both professionals and students, that a solid knowledge of the past, the conventions and the performing conditions relevant to the different epochs, is essential for a better understanding of the repertoire. In this regard, Ralph Kirkpatrick was an important pioneer of historically-informed performance who expressed his view thus:

I cannot abandon a conviction that a conscientious interpreter of the music of the past has an obligation to know as much as he possibly can about the way in which that music was conceived by its composer and performed by his contemporaries, however much he may decide to depart from this necessarily insufficient body of knowledge.[5]

Despite this conviction being expressed more than 30 years ago, a gap still exists between many modern performers’ attitudes and the obligation to which Kirkpatrick refers. Outside the early music community, today’s music practitioners tend to conform to a standard that often has little or nothing to do with what we know about the performance practice of past centuries. While imitating the interpretive choices of a musician from the past is often considered neither desirable nor musically valuable, young performers tend to take reference from their favourite virtuosos and develop the interpretation of a piece by listening to their recordings. Although music modelling is quite common among musicians, and imitating original performances by great artists is often considered an engaging and meaningful exercise,[6] when it comes to choosing their reference artists, young musicians are more likely to focus on a rising star than an old master. Something similar happens even in the realm of early music, where a performer specialising in a particular genre or period may come up with an interpretation that appeals to a large audience and therefore makes a big impact in the discographic market on grounds of novelty and originality rather than historical accuracy or verisimilitude, let alone authenticity. In this regard, Richard Taruskin’s idea that today’s notion of authenticity is more commercial propaganda than a description of a historical approach to the past still holds well; as he suggests, often ‘a thin veneer of historicism clothes a performance style that is completely of our time, and is in fact the most modern style around.’[7]

It is not my intention to suggest that we should or could recreate our musical past by imitating what we hear in early recordings, nor am I implying that singers who recorded in the 1910s and 1920s were truer to the intentions of 19th-century composers than any of our modern interpreters. Nor is it necessary to resuscitate the ancient dispute regarding the extent to which operatic composers were willing to endorse a performance practice that, however pervasive, saw their musical intentions constantly defied by the diva of the moment. One might even argue that the operatic vocal style we find in, say, Patti’s and Tetrazzini’s recordings is less a question of historical continuity than personal taste—all the more so if we assume that authenticity is not a notion that early 20th-century interpreters seem to have cherished. Nevertheless, early recordings can help us to reconstruct a moment in the history of a vocal performance practice that has long been misunderstood on the basis of a false, although widespread, notion of earlier generations’ cavalier attitudes to Werktreue (despite the concept itself originating as a nineteenth-century phenomenon). They can help us to shed some light on how  musical taste changes over time and how what used to be considered exceptionally good 100 years ago might sound wrong and tasteless today and yet still embody a sense of faithfulness, albeit one peculiar to its time.

In this contribution, I focus on the manner in which tempo modifications and, to a more limited extent, the use of ornamentation can be investigated in the recordings of Marcella Sembrich (1858–1935), Nellie Melba (1861–1931), and Luisa Tetrazzini (1871–1940). Taking reference from ‘Ah! Fors’è lui’ from the ‘Scena ed Aria [di] Violetta – Finale Atto I’ of Giuseppe Verdi’s La traviata, I make use of some descriptive statistics to draw a comparison among these divas and, when possible, highlight some analogies in order to explore the extent to which they all belonged to a common breed and were part of a still vibrant vocal tradition. Having transcribed and compared their ornamentation and cadenzas, I then suggest that, once the interpretive solution that worked best was chosen, these prima donnas remained consistent with the original choice. Finally, I touch upon the notion of vanity, which, together with that of whim, seems to have been a feature typical of past prima donnas. This notion, if applied uncritically, can lead to dangerously misleading conclusions.

Expressive tempo modifications

As I have already suggested, a number of sources indicate that, throughout the 19th century, opera singers tended to give more expression to their interpretation by changing the tempo.[8] This tendency spread across all musical genres and led to strong reactions from musicians and commentators who remained true to the older aesthetic of classical reserve.[9] Although tempo modifications were generally understood as a legitimate expressive device to render a passage more moving, it was the extent of those changes that made some music critics cringe and complain. In 19th-century Italian opera, several sources emphasize the importance of the connection between the libretto and the expression of its dramatic contents; these had to be conveyed to the audience by way of a proper use of the voice, a clear pronunciation of the words, the necessary changes in the tempo, and the addition of the most suitable coloratura. Tempo modifications were generally considered appropriate when demanded by the expression of a languid or an agitated character. Two tempo-related expressive devices emerge consistently from a vast body of evidence—tempo rubato, which involves small modifications in the melodic line over a steady beat within a single bar, and larger alterations of entire phrases or longer sections, with the voice and the orchestra slowing down and speeding up together. An account of the legitimacy of tempo rubato and the necessity for a singer to use it with a grain of salt comes from Domenico Corri:

Composers seem to have arranged their works in such a manner as to admit of this liberty, without offending the laws of harmony: one caution, however, becomes highly necessary namely, that this grace, or licence, is to be used with moderation and discretion, in order to avoid confusion for too frequent a use of Tempo Rubato, may produce Tempo indiavolato.[10]

Larger tempo modifications, Corri continues, represented an improvement, not an abuse, on the condition that they found their justification in the lyrics:

Another improvement, by deviation from strict time, is to be made by the singer delivering some phrases or passages in quicker or slower time than he began with, in order to give emphasis, energy, or pathos, to particular words.[11]

As noted by Robert Toft, in some cases tempo modifications were suggested when changes in dynamics were called for.[12] Manuel Garcia’s School provides the most extensive description of all the expressive devices a singer should master in order to interpret the relevant repertoire, with a large number of examples taken from composers like Rossini, Donizetti, Bellini, Cimarosa, Mozart, Handel, and Meyerbeer. He also calls attention to the difference between tempo rubato and the use of accelerando and rallentando and describes them in terms consistent with Corri.[13] Garcia clarifies that ‘Time’ (misura in the Italian edition) can be of three different characters—regular, free and, mixed—and that the mixed character is demanded whenever the ‘feelings expressed in a piece exhibit frequent irregularities of movement, as is often the case in tender, melancholy sentiments.’[14] He also explains that rallentando is used to express a decrease of passion or as a preparation for the return of a theme or melody, while accelerando is preferred when it comes to adding greater spirit to the general effect. In either case, a change in the tempo finds its justification in the connection that exists between the text and the music. Similar suggestions can be found in many contemporary singing methods where the idea of expressive flexibility was repeatedly insisted upon.[15]

Things did not undergo major changes during the second half of the nineteenth century, notwithstanding the new dramatized style introduced by Vincenzo Bellini and developed by Giuseppe Verdi. When the Paris premiere of I puritani was in preparation in 1834, Bellini, who had added some metronome marks to the score, wrote to the conductor Francesco Florimo asking him to ignore the metronome and instead quicken and retard the tempos according to the singers’ voices.[16] In his Hints on Singing, Garcia remarks that ‘the works of Donizetti and Bellini contain a great number of passages which, without bearing the sign of the rallentando or accelerando, yet require their use.’[17] The same can be said of the interpretation of Giuseppe Verdi’s operas, as some contemporary reviews suggest. In 1850, Henry Fothergill Chorley pointed out that ‘Verdi’s music, in its solo passages and closes, gives him [the English tenor John Sims Reeves] scope for that slackening of tempo and elongation of favourite notes which are considered by “Young Italy” as the style dramatic.’[18] In 1853, the same critic suggested that Angiolina Bosio’s style suited Verdi’s music, for ‘by him singers are invited, not forbidden, to slacken tempo.’[19]

It is clear that tempo variability was not only a stylistic feature typical of each individual but also an interpretive device to be considered each time a given passage called for special emphasis. The extent of these modifications depended on many factors, both individual- and repertoire-related, and the risk for an interpreter to exaggerate was lurking around every passage.

As I have already suggested, other devices were also typical of the so-called bel canto tradition:

  • the use of portamento;
  • the use of ornamentation;
  • the modification of melodic passages;
  • the insertion of a central semi-cadenza;
  • the insertion of a final cadenza or its modification, should the composer have written one already.

Contrary to our common understanding, these devices were meant to be used, either alone or in combination, to make the music more expressive and not to show off the voice at the expense of the composer.[20] Tempo modifications were chosen in relation to the dramatic text and were also dependent on the kind of vocal ornamentation that a single interpreter added or substituted for. A wonderful example of the strong relation between the extent of vocal ornamentation and that of tempo modifications is provided by Rosina’s Cavatina in Rossini’ Barbiere di Siviglia. As it is possible to hear in Tetrazzini’s and Sembrich’s recordings, for instance, this aria lends itself to a number of insertions that might affect the inter-onset interval (i.e., beat length) in a remarkable manner.

Profiling the interpreters

Three singers were selected for this investigation: Marcella Sembrich (1858–1935), Nellie Melba (1861–1931), and Luisa Tetrazzini (1871–1940). At the outset of the century, each of these three prima donnas was acclaimed as the worthy successor of Adelina Patti (1843–1919); although they had very little in common with regard to vocal training and musical background, each was saluted as living evidence that the art of bel canto had not yet died, and the moment had not yet arrived for its corpse to be buried.

Marcella Sembrich was a Polish soprano who, having studied violin and piano in Vienna, moved to Milan in 1876 to study with Giovanni Battista Lamperti and, subsequently, with his eminent father Francesco Lamperti.[21] It was thanks to Lamperti’s help that Sembrich developed her superb vocal technique and expressive talent. Her debut as Lucia at Dresden Royal Opera House in September 1878 was an immediate success and she was dubbed the ‘Polish Patti.’ Her international career continued in the US, where she made a sensational debut as Lucia in 1883 New York. As Harold Bruder writes in the liner notes that accompany the CD box published by Romophone in 1998, her recordings are among the earliest, and already at the outset of the 20th century her interpretations of short arias and vocal pieces were recorded by two amateurs who were enthusiastic about the new medium: Lieutenant Gianni Bettini and Lionel Mapleson, the Librarian of the Metropolitan Opera Company. The former seems to have recorded Sembrich singing the first part of Johann Strauss’s ‘Voci di Primavera’ in his New York laboratories as early as 1900 while the latter used his cylinder machine placed in the prompter’s box to record a few short selections from live performances during the 1900‒1901 season at the Met.[22] When describing the difficult situation of bel canto at the beginning of the century, in 1920 Herman Klein suggested that Marcella Sembrich (together with Nellie Melba) was one of those few sopranos of the younger generation who could follow in Adelina Patti’s footsteps, even though Patti’s numerous talents remained still unmatched.

And what of her own fin-de-siècle contemporaries? Among these Sembrich and Melba were, perhaps, the only sopranos whom Patti considered capable of upholding the exalted traditions of the fading school. There were still coloratura singers, but very few of them artists of the first rank, and not one whose vocal and histrionic resources would permit her to cover the same wide operatic field that the diva’s versatile genius had enabled her to adorn. Both Sembrich and Melba were endowed with lovely voices, rare charm, and great beauty of style; but their limitations as prime donne stood out clearly when they were compared with the supreme artist [Patti] who had made her European debut a full generation before they came upon the scene.[23]

In his writings on singers and recordings, Klein went into some detail with regard to the difference between Melba’s and Sembrich’s voices: ‘I noted the clarity of Sembrich’s tone, the ease with which she executed her fiorituri, the richer timbre of her voice compared with the silvery quality of Melba’s, yet not excelling it in musical sweetness or flexibility or a clean articulation of every note in the brilliant passages.’[24] He also compared Sembrich and Patti when talking of their recorded interpretations of the ‘Cavatina di Rosina’ from Rossini’s Barbiere:

Her [Sembrich’s] scales (and there are plenty in the ornamental passages) are simply beyond reproach, and her trill is no less perfect. Everything is of crystalline clearness; you hear every note, no matter how intricate the weaving of the vocal arabesque; and the high D in the cadenza, like the C at the end, is taken without effort. These things are largely due to Sembrich’s admirable breathing, in which, as in the unusually rich quality of her “dark” tone she always reminds me of her acknowledged model, the far-famed “Adelina.”[25]

Apparently, Sembrich seems to have found the recording horn intimidating.[26]

Nellie Melba’s earlier teachers in Melbourne were the German-born pianist Julius Buddee, who had been Jenny Lind’s accompanist, and the Italian tenor Pietro Cecchi, who taught at the Presbyterian Ladies College, where young Melba attended school.[27] Most crucially, Melba was to become the favourite student of Mathilde Marchesi (née Graumann),[28] and it is well known that Marchesi dedicated to Melba one of the three cadenzas with obbligato flute she had written for the mad scene in Donizetti’s Lucia di Lammermoor.[29] On the occasion of Melba’s London debut on 24 May 1888 as Lucia, Hermann Klein in the Sunday Times noted ‘the extraordinary beauty of timbre and her exceeding brilliant vocalization’ but found her deficient ‘in that indescribable something which we call charm.’ ‘Her accents lacked the ring of true pathos,’ he continued, and although she possessed admirable intelligence, ‘the gift of spontaneous feeling has been denied her.’[30] According to Henry Pleasants, ‘she learned to go through the motions with professional aplomb, although these motions were said by her detractors not to have gone beyond the raising of one arm in situations of some intensity and two arms for an outburst.’[31] The distance between Melba and her Italian colleagues was also noted by John Pitts Sanborn from The Globe, who heard Tetrazzini and Melba many times throughout their careers.

Melba sang accurately and with dignity of good workmanship. Her singing was stereotyped without the excitement of the unexpected, the suddenly improvised, the inspiration of the heat and joy of song. Sometimes, as Tetrazzini’s harshest critics insist, the soprano injuries the music by the variations she introduces; oftener she lifts it above the clouds. This sort of thing was inherent in the great Italian style as in the Italian temperament. Melba’s style was rather mid-century French, the style of Faust and Roméo and Juliette than that of the older Italian roles, though in many respects she sang those roles so well and so delightfully.[32]

The aplomb these contemporary critics highlighted can still be heard in Melba’s recordings, especially if one focuses on the flawless but slightly inexpressive manner in which she controls the voice and keeps a steady pace.

Luisa Tetrazzini, the Florentine nightingale, was born to a musical family; both her older sisters, Elvira and Eva, were professional singers, and Eva would marry the conductor Cleofonte Campanini, who came to be a prominent figure in Luisa’s career. Having studied in Florence with Ceccherini, Pietro Cesari seems to have played an important role in Luisa’s early training and in the development of her vocal style.[33] Her career was international, and she made regular appearances in London, New York, and other prominent theatres in Europe, North and South America until 1913. Her London debut in 1907 represents a turning point in her career, and on that occasion some critics called her ‘the new Patti.’

Her phenomenal success, which has been heralded the world over, did not come to her, however, until the local autumn season of grand opera at Covent Garden, where she had never sung before. Her first audience went wild over her. She was cheered again and again and the critics exhausted their vocabulary in sounding her praises. She was Patti and Jenny Lind and Melba all in one, they declared. Hers was the voice of the century.[34]

Tetrazzini came to know Adelina Patti personally; the latter was in the audience when the former celebrated her triumphal return to Covent Garden on 30 April 1908, and two letters written by Adelina that year suggest that she held Luisa in high esteem.[35] A few months earlier, when interviewed by the The Sun, Tetrazzini had also declared that she knew Patti’s and Melba’s recorded interpretations: ‘Have I ever heard Patti? Melba? Not until quite recently except through a gramophone, which I listen to frequently.’[36] After 1913, she made her appearance only in concert recitals.

Although personal relationships are sometimes evident, it is not easy to assess whether and to what extent these three artists belonged to a common musical breed, nor is it possible to trace a clear line of continuity that links them. While Sembrich and Tetrazzini studied in Italy and were on good terms with Adelina Patti, Nellie Melba became Marchesi’s favourite student in Paris after taking lessons from Jenny Lind’s piano accompanist in Australia. Patti had sought advice from Rossini, but it is not clear whether and to what extent she handed over to Sembrich and Tetrazzini what she had learned from the composer. Rossini seems to have been behind the vocal training and musical development of Nellie Melba as well, for her teacher, Mathilde Marchesi, had studied with Manuel Garcia Jr., the son of the first interpreter of Count Almaviva in Il barbiere di Siviglia (Figure 1). However, while Melba seems to have been the singer who best incarnates an uninterrupted vocal tradition that connects her to Rossini himself, as soon as we listen to her interpretations, the distance from her colleagues becomes striking, both in terms of singing style and vocal technique. As we have seen, this distance was already noted by some contemporary commentators. The figure below sketches some of the relationships between these singers and with the tradition in which they operated:

Figure 1 shows the links that connect Sembrich, Melba, and Tetrazzini with the tradition.

A preliminary survey of the roles these prima donnas performed over their careers suggests that, although they were all celebrated as the worthiest representatives of the bel canto tradition, only a few operas feature prominently in each of their discographies; among them are Donizetti’s Lucia di Lammermoor and Verdi’s La traviata and Rigoletto.

Marcella Sembrich’s recordings feature mainly those Italian composers who were long inscribed into the 19th-century bel canto tradition, with Lucia di Lammermoor occupying a prominent position throughout her entire career, but her repertoire included also two Wagner roles, Elsa in Lohengrin and Eva in Die Meistersinger, as well as Puccini’s Mimì (Miller).[37]

In 1894, Melba made her appearance at the Met in New York singing Lucia in Lucia di Lammermoor, Nedda in I Pagliacci, Elizabeth in Tannhäuser and Elsa in Lohengrin. She recorded some of the most celebrated arias from Verdi’s Otello and La traviata, Puccini’s La Bohème and Tosca on many occasions. While the mad scene from Lucia appears twice in her discography, none of either Rossini’s or Bellini’s most celebrated arias does.[38]

Donizetti’s Lucia di Lammermoor is the opera that scores highest in Luisa Tetrazzini’s chronology; she sang Lucia in 113 productions over twenty years between 1892 and 1913. Verdi’s Rigoletto follows at some distance with 67 productions, while Rossini’s Barbiere holds the third position with 61 productions. If the bel canto repertoire was the one with which she felt most comfortable, Lucia is probably the opera that gave her the best opportunity to exhibit her talents and show off her voice. Years after her debut in this role, Tetrazzini remembered that ‘no opera could have been selected which gave me a greater opportunity, for Lucia’s arias have more possibilities for the prima donna than any of the other operas.’[39] In her How to Sing, she raised the question of stylistic appropriateness and suggested that every singer ‘must be familiar also with the varying needs of the different schools of music, with the historical traditions associated with them.’[40] However, since every repertoire has its own different requirements in terms of both vocal technique and singing style, it is unlikely if not impossible, Tetrazzini holds, for an artist to excel in all of them. The consequence is straightforward: ‘For this reason the average artist will, I think, usually be well advised to confine himself to the class of work more particularly suited to his talent.’[41] Tetrazzini herself admitted that, at some point in her career, she had to make a choice and steer away from works that did not fit her vocal skills and artistic personality. ‘People blame me sometimes, for instance, for confining myself mainly to music of a certain school. But I think I know best as to this, and that I am exercising sound judgment in adopting this course. There is much music which I admire and love, but I do not always try to sing it.’[42] If flexibility was a quality that any singer should cultivate, it was not advisable to push oneself into foreign lands and embrace a repertoire that did not fit one’s voice. Tetrazzini’s discographic legacy includes repeated recordings of select arias from Rossini, Donizetti, Bellini, and Verdi but features no Italian operatic composer of the younger generation, not to mention Wagner.

‘Ah! Fors’è lui’

To investigate tempo variability and verify the extent to which these three singers used it as a stylistic and interpretive feature, I considered ‘Ah! Fors’è lui” from the “Scena ed Aria [di] Violetta – Finale Atto I’ of Giuseppe Verdi’s La traviata.[43] In part, the scene adheres to the formal segmentation conventionally defined as ‘solita forma’ (the usual form), as Abramo Basevi described it in 1859 when talking of vocal duets in Verdi’s works—Tempo d’attacco, Adagio, Tempo di mezzo, Cabaletta.[44] In Violetta’s first aria, a Cavatina, the Tempo d’attacco is missing; instead, after a short recitativo (Allegro), we have a cantabile (Andantino), a second recitativo (Allegro), and the final Cabaletta (Allegro brillante). I focussed on the Andantino, which lends itself to larger tempo modifications owing to its dramatic and musical content. As we know well, this is the moment in the first act when Violetta, after the ball, is hesitant and, perhaps, confused; she wonders whether Alfredo, who has kindled in her the burning flame of love, is the man her heart has been longing for and dreaming of. Mysterious and unattainable, love is now the torment and delight of her heart. Francesco Maria Piave’s lyrics revolve around the expression of three emotions, Violetta’s longing for a man to love (first stanza), the sense of unrest that follows Alfredo’s words (second stanza), and the thrilling hesitation that accompanies what seems a turning point in her entire life.[45] These three emotional states are presented again in the second group of stanzas to form a large two-section structure (A-A’). In each section, we find two main melodic ideas, the first in f minor, the second in F major, each being segmented following the so-called lyric form: first stanza (a4+4 a4+4,) in f minor; second stanza (b4+4) featuring a modulation; third stanza (c4+4) in F major.[46] While the first melodic idea in f minor presents a sobbing quality and features a minor sixth interval that provides a strong sense of melancholy, the second opens lyrically toward the high register to express Violetta’s sudden, unrestrained abandon. The same connection between the lyrics and the music characterizes the repeat (A’).

Andantino Ah, fors’è lui che l’anima

Solinga nè tumulti

Godea sovente pingere

Dè suoi colori occulti!…

Lui che modesto e vigile

All’egre soglie ascese,

E nuova febbre accese,

Destandomi all’amor.

A quell’amor ch’è palpito

Dell’universo intero,

Misterioso, altero,

Croce e delizia al cor.

  A A (f min.)


B (F maj.)

a 4+4

a 4+4

b 4+4

c 4+4

d 4+4

A me fanciulla, un candido

E trepido desire

Questi effigiò dolcissimo

Signor dell’avvenire,

Quando nè cieli il raggio

Di sua beltà vedea,

E tutta me pascea

Di quel divino error.

Sentìa che amore è palpito

Dell’universo intero,

Misterioso, altero,

Croce e delizia al cor![47]

  A’ A (f min.)


B (F maj.)

a 4+4

a 4+4

b 4+4

c 4+4

d 4+4

As can be observed, Francesco Maria Piave’s choice of wording offers a broad palette of emotions; a richly nuanced vocabulary leads up to the image of love seen as a mysterious combination of joyful bliss and sorrowful grief. Verdi’s music underpins the expression of these feelings with skilled ability, and the different melodic ideas offer a large choice of interpretive solutions involving changes in vocal colour, tempo, and dynamics.

The recordings

Eleven different renditions were investigated, spanning the years from 1903 to 1911. Two companies played a prominent role in the development of the recording technologies and the relevant market: the Gramophone & Typewriter Company, which became the Gramophone Company in 1907,[48] and the Victor Talking Machine Company of Camden, New Jersey (US). As shown in the following tables, none of the recordings includes a complete rendition of the aria; they all present cuts, in part due to the limited length of cylinders and early discs.

Marcella Sembrich recorded this aria on four occasions (Table 1); of the first recording, made in 1903 with Charles Adam Prince accompanying on the piano and initially unpublished, two takes remain that can now be found included in Marcella Sembrich, The Victor Recordings (1908-19), published by Romophone in 1988.[49] The second, also with piano accompaniment, dates back to 1904 (C1900-1, Victor 85035). Of the third recording session, made in 1906 with the Victor Orchestra, two takes exist; the first was soon published by Victor (C3152-1, Victor 88018) while the second (C3152-2) remained unpublished until Romophone included it in the CD box devoted to Sembrich’s 1904‒08 recordings in 1997;[50] the fourth, again with the Victor Orchestra, was realized in 1908 (C3152-4, Victor 88018).

Table 1 shows the recordings of “Ah! Fors’è lui” from ‘Scena ed Aria [di] Violetta – Finale Atto I’ of Giuseppe Verdi’s La traviata that Sembrich realized in the years 1903‒1908.

Title Date Matrix Catalogue 
‘Ah! Fors’è lui…’ 1903 (take 1) 1366 unpublished
‘Ah! Fors’è lui…’ 1903 (take 2) 1366 unpublished
‘Ah! Fors’è lui… Sempre libera’ 5 Nov. 1904 C 1900-1 Victor 85035
‘Ah! Fors’è lui… Sempre libera’ 1 March 1906 C 3152-1 Victor 88018
‘Ah! Fors’è lui… Sempre libera’ 1 March 1906 C 3152-2 unpublished
‘Ah! Fors’è lui… Sempre libera’ 3 March 1908 C 3152-4 Victor 88018

The 1906 unpublished take was not considered for this investigation on grounds of poorer quality when compared to the first, published one. As can be observed in Figure 2, all the recordings present the same cuts. While either recitativo section is likely to have been excluded for musical and time-related reasons, both the Andantino (‘Ah, fors’è lui’) and the Allegro brillante (‘Sempre libera’) present cuts that are likely to have belonged to a long-established tradition: the repeat (A’) in the Andantino and the first section of the Allegro brillante are not present. The Allegro brillante, a typical cabaletta where the same music material is presented twice, was generally intended for vocal display and singers used to enrich the repeat by inserting new and richly ornamented passages.

Figure 2 shows the cuts present in Sembrich’s recordings of ‘Scena e Aria [di] Violetta.’

Melba first recorded this aria in March 1904 for the Gramophone & Typewriter Company. The first two sessions took place at her residence at 30 Great Cumberland Place in Mayfair, London, and she was accompanied by Landon Ronald (1873‒1938) on the piano. Despite the customary cuts, two takes were necessary to record the entire ‘Scena e aria.’ One more recording of ‘Sempre libera’ was made the same month, this time with Ronald conducting the orchestra. The first and third recordings (Mat. 6, Cat. 03017 and Mat. 23, Cat. 03026) were reissued digitally by Naxos in 2002.[51] The aria was recorded two more times, in 1907 (C 4339-1, Victor 88064) and 1910 (C 4339-3, Victor 88064); both recordings are in a single take (see Table 2). Interestingly, although both the Victor recordings had Walter Rogers conducting the orchestra, the one realized in 1907 lasts 4’15’’ while the one made in 1910 is almost five minutes long, a difference that does not seem to find sufficient justification in the number of recorded bars. This suggests that a longer recording time might have led to a more relaxed interpretation of the music. All the recordings made by Melba for the Victor Talking Machine Company have been included in a three-CD set that was published by Romophone in 1994.[52]

Table 2 shows the recordings of “Ah! Fors’è lui” from ‘Scena ed Aria [di] Violetta – Finale Atto I’ of Giuseppe Verdi’s La traviata that Melba realized in the years 1904‒1910.

Title Date Matrix Catalogue
“Ah fors’è lui…” March 1904 6 G&T 03017
“Follie!… Sempre libera” March 1904 6 Unissued
“Sempre libera” March 1904 23 G&T 03026
“Ah! Fors’è lui… Sempre libera” 27 March 1907 C 4339-1 Victor 88064
“Ah! Fors’è lui… Sempre libera” 23 August 1910 C 4339-3 Victor 88064

In Melba’s case, every recording presents cuts; the one in the “Andantino” corresponds to that in Sembrich’s recordings. Instead, the later recordings show differences that could have depended on the more advanced recording technologies available at the time (Figure 3).

Figure 3 shows the cuts present in Melba’s recordings of ‘Scena e Aria [di] Violetta.’

Tetrazzini realized a first partial recording of Verdi’s ‘Scena e aria’ in London in December 1907 for the Gramophone Company, on which occasion the Tempo di mezzo and the cut version of the Cabaletta ‘Sempre libera’ were recorded (2179f/053147). The following year, it was the turn of the opening recitativo and the cut version of the Andantino (‘Ah, fors’è lui’) to be recorded (2573f/053196). These two different takes are now to be found merged in one single track in the first CD of the set Luisa Tetrazzini, the London recordings, which EMI Classics published in 1992. In the same set, again merged in a single track, we find two recordings made in July 1911: ‘Ah! Fors’è lui’ that includes ‘Follie!…’ (ac 5164f/2-053059) and ‘Sempre libera’ (ac 5169f/2-053062). These two takes together make an almost complete recording of the ‘Scena e aria’ despite some of the customary cuts.[53] On 16 March 1911, Tetrazzini recorded for Victor the complete ‘Scena e aria’ in a single take; the recording presents the usual cuts and includes neither recitativo (C 10065-1, Victor 88293)[54] (Table 3).

Table 3 shows the recordings of “Ah! Fors’è lui” from ‘Scena ed Aria [di] Violetta – Finale Atto I’ of Giuseppe Verdi’s La traviata that Tetrazzini realized in the years 1907‒1911.

Title Date Matrix Catalogue
“Sempre libera” December 1907 2179f G&T 053147
“È strano!… Ah, fors’è lui” August 1908 2573f G&T 053196
“Ah! Fors’è lui… Sempre libera” 16 march 1911 C10065-1 Victor 88293
“Ah, fors’è lui… Follie!…” 11 July 1911 ac 5164f/2 G&T 053059
“Sempre libera” 11 July 1911 ac 5169f/2 G&T 053062

Cuts similar to those present in Sembrich’s and Melba’s recordings can also be found in Tetrazzini’s, especially if one considers the Andantino (Figure 4).

Figure 4 shows the cuts presents in Tetrazzini’s recordings of ‘Scena ed Aria [di] Violetta – Finale Atto I.’

Behind the constraints imposed by the still rudimentary recording technologies, a performance tradition lay in the background that already featured cuts of single sections, especially if a repeat presented itself. This seems to suggest that despite a long-lasting tradition, at the beginning of the 20th century, repeats no longer offered themselves as an opportunity to add new coloratura and substitute for the written bravura passages. Instead, they were generally avoided, perhaps because they were considered redundant and musically uninteresting.


Each audio track was investigated using Sonic Visualizer, a software package that allows users to perform several measurements and extract the relevant data.[55] To quantify tempo modifications, the inter-onset intervals were measured by tracking each beat-onset (quaver/eighth note) with a marker in the waveform. When the voice and the piano were not perfectly synchronized, the onset in the voice was considered. To improve accuracy, the procedure was performed at reduced playback speed (down to 50%). Basic descriptive statistics of the mean, mode, standard deviation, and coefficient of variation of the beat duration were calculated.

A first glance at the measured coefficient of variation (CV) values shows three different approaches to tempo variability and, if we consider the mean values, three different profiles (intra-individual consistency). The coefficient of variation is the ratio of the standard deviation (SD) to the mean and shows the extent of variability in relation to the population. In our case, it shows the extent of variability in relation to the mean value of the beat duration. Put in simpler words, it shows how variably each interpreter treated tempo at both intra- and inter-individual levels. As can be observed in Table 4, while Melba presents the lowest degree of variability (mean CV = 0.24), Sembrich (only three years her senior) features the highest variability of all (mean CV = 0.41), with Tetrazzini (the youngest) also scoring a relatively high degree of agogic freedom (mean CV = 0.33).

Table 4

Interpreter Year CV Mean CV Inter-onset interval mean value Mean bpm Mean bpm value


Melba 1904 0.26 0.24 0.93 0.93 64.5 e = 64.5
Melba 1907 0.22 0.90 66.7
Melba 1910 0.26 0.96 62.5
Tetrazzini 1908 0.35 0.33 0.99 0.99 60.6 e = 60.7
Tetrazzini Mar 1911 0.27 0.95 63.2
Tetrazzini July 1911 0.39 1.03 58.3
Sembrich 1903 0.40 0.41 0.93 0.92 64.5 e = 65.1
Sembrich 1903 0.37 0.86 69.8
Sembrich 1904 0.40 0.96 62.5
Sembrich 1906 0.45 0.97 61.9
Sembrich 1908 0.42 0.90 66.7

These differences do not seem to be related to corresponding differences in average speed; although Melba shows the lowest tempo variability, her average tempo value (beats per minute) is higher than Tetrazzini’s but slightly lower than Sembrich’s.

Pearson correlation analysis was used to determine long-term trends, but due to the limited sample size, generalizations were difficult to draw. Although the calculated correlation coefficient between CV and year of recording shows a moderate inverse correlation (r = -.26), which tentatively suggests that the later the recording, the less variable the tempo, the measured values appear to be clustered in three individual-dependent groups (see Table 5). This goes in the direction of a more personal and individual style, rather than a widely shared notion of tempo variability.

Figure 5 shows a slight inverse correlation (r = -.26) between year of recording and Coefficient of Variation for Melba, Tetrazzini, and Sembrich.

By contrast, a stronger and more general tendency toward a slower tempo may be observed (r = -.53), which might have depended in part on the gradual development of the recording technologies and the consequently longer recording time (Table 6).

Figure 6 shows an inverse correlation (r = -.53) between year of recording and metronome values (bpm) for Melba, Tetrazzini, and Sembrich.

Whether the recording technologies played a crucial role or not, the tendency towards a more relaxed tempo is clear, and what the graph shows is consistent with what we already know from several contemporary accounts, suggesting how taxing the recording experience could be.

If we now look at each individual’s tempo profiles, we notice that, despite some differences, they tend to remain consistent over time and are strongly dependent on melodic segmentation. In particular, the following graphs (Figures 7‒13) show that, while the recorded interpretations of the f minor section present larger differences, possibly on account of the more nuanced dramatic content, in the F major section all the singers tend to adhere more strictly to the melodic contour, its climaxes and points of rest.

Figure 7 shows Sembrich’s tempo profile in the f minor section.

Figure 8 shows Sembrich’s tempo profile in the F major section.

Figure 9 shows Melba’s tempo profile in the f minor section.

Figure 10 shows Melba’s tempo profile in the F major section.

Figure 11 shows Tetrazzini’s tempo profile in the f minor section.

Figure 12 shows Tetrazzini’s tempo profile in the F major section.

In Tetrazzini’s case, it is interesting to observe that, in the F Major section, the tempo profile of the March 1911 recording is smoother than that of the others. This could be due to the different orchestra conductor; on that occasion, it was Walter Rogers conducting the Victor Orchestra in Camden, while we do not know for sure who conducted the others—perhaps Percy Pitt, who used to work for the Gramophone Company in London.

These results are consistent with what was noted by Rebecca Plack when she suggested, ‘while it would be accurate to say that some early recordings display greater rhythmic flexibility than any recordings that have followed since, it would not be accurate to conclude that rhythmic style in early recordings was on the whole characterized by greater flexibility—unless we are referring to flexibility of choices’.[56]

Ornamentation and Cadenzas

If we now turn our attention to each interpreter’s use of ornamentation and their insertion of new cadenzas, we also notice that, once the solution that best fit into one’s voice compass and dramatic characterization had been chosen, each interpreter tended to remain consistent with the original choice. Sembrich’s recordings show how little her cadenza to the cantabile changed over time; while the second part did not change at all, except for the final trill, the first was slightly richer in 1903, simplified in 1904, and remained the same in 1906 and 1908 (Figure 13).

Figure 13 shows Sembrich’s cadenza to the cantabile.

In Melba’s case, the three identical cadenzas show not only how consistent she remained over the years but also how distant she was from the idea that a singer should use these moments as an opportunity for quasi-spontaneous vocal display (Figure 14).

Figure 14 shows Melba’s cadenza to the cantabile.

Similarly, Tetrazzini’s discs show no changes in the recorded cadenzas (Figure 15).

Figure 15 shows Tetrazzini’s cadenza to the cantabile.

Most crucially, the interpretative choices present in these recordings are consistent with what was reported by some contemporary reviews of live performances of La traviata. For instance, when Tetrazzini was Violetta in New York in 1908, on 16 January The Evening World wrote that ‘at the end of the air “Ah, fors’è lui,” she took a high C and swelled and diminished it with evenness and precision and at the end of the succeeding “Sempre libera” she sustained a strong and clear E flat.’[57] Both passages can be found in her recordings (Figures 15 and 16).

Figure 16 shows the high E flat with which she concluded the aria.

Similarly, on 16 January 1908, The Sun wrote that ‘her transformation of the plain scale in the opening cadenza of “Sempre libera” into a chromatic scale, though a departure from the letter of the score, was not at all out of taste and its execution fully obtained its right to existence.’[58] The same departure from the letter of the score can be found in her recordings (Figure 17).

Figure 17 shows the chromaticism Tetrazzini used to sing the opening cadenza of ‘Sempre libera.’


The findings presented so far support the notion that singers like Melba, Sembrich, and Tetrazzini tended to be consistent with their interpretive choices over time.[59] In Tetrazzini’s case, the image of a capricious, although phenomenally gifted, singer was nurtured by those who knew her professionally and may have been part of a constructed ‘persona’ designed to be consistent with the public’s view of the proverbial artistic temperament. On 6 November 1907, Kenneth Muir, manager of the Milanese branch of The Gramophone Company, described her as follows:

In character she is capricious and wayward and if you wish to succeed in obtaining her you must pamper her like a spoiled child by sending her gifts, boxes for theatres, paying her compliments and little personal attentions; in a word you must appeal to the woman in her nature. She is extremely dissolute in her private life and much affected by flattery and champagne.[60]

Nevertheless, even if opera divas were sometimes described in terms similar to those used by Muir to portray Tetrazzini’s capricious personality, the roles they interpreted were carefully prepared and consistently performed. We could also doubt that prima donnas like Tetrazzini, Melba, and Sembrich could sing a new cadenza each evening and improvise a new coloratura in each performance. Tetrazzini herself, when interviewed by The Sun in 1908, revealed that she did not invest much time in practising during the season, except when she was going over a new role. When describing her rehearsals, she claimed that ‘when I go on the stage, the orchestra says “la-la” and I respond “la-la,” and the conductor nods that it is all right and I sing away. That is all the rehearsing I have.’[61] We should also remind ourselves that a diva’s professional career involved a very busy routine and a restless travelling schedule that left very little time for practising and rehearsing. As we have already seen, the notion that a stable repertoire that fit one’s voice compass and vocal technique should be preferred was already shared among these singers, who specialized in, and focussed on, a limited number of individual composers and works. In this regard, prima donnas were neither unpredictable nor capricious. Furthermore, any modern singer knows how difficult it is to improvise ex tempore a new embellishment or to add a different coloratura passage with no adequate preparation, especially when performing in a fully-packed theatre in front of a qualified audience eager to listen to an old favourite. No professional singer, I believe, would take such a risk. Given the evidence amassed here, I would maintain that we should no longer cherish the notion that the divas and prima donnas of earlier eras followed the whim of the moment, nor should we assume that they constantly sacrificed the music and the composers’ intentions to their vanity.

In general, early recordings represent a valuable source of information, and the analysis of tempo variability by means of simple descriptive statistics can complement the investigation of traditional text-based historical sources. As anticipated at the outset of this contribution, although generalizations are difficult to draw, early recordings can help us shed some light on how dramatically musical taste changes over time and how what sounds wrong and tasteless today was considered not only appropriate but exceptionally good 100 years ago.

[1]     Daniel Leech-Wilkinson, ‘Listening and Responding to the Evidence of Early 20th-Century Performance,’ Journal of the Royal Musical Association, 135.1 (2010a), 45-62.

[2]     Deborah Kauffman, ‘Portamento in Romantic Opera,’ Performance Practice Review 5.2 (1992), 139-158.

[3]     Philip Gossett, Divas and Scholars. Performing Italian Opera (Chicago: Chicago University Press, 2006), p. 264.

[4]     Nicholas E. Limansky, ‘Luisa Tetrazzini Coloratura Secrets,’ The Opera Quarterly 20.4 (2004), 540-569.

[5]     Ralph Kirkpatrick, Interpreting Bach’s well-tempered clavier. A performer’s discourse on method (New Haven: Yale University Press, 1984), p. 29.

[6]     Tânia Lisboa and others, ‘Mastery through imitation: A preliminary study,’ Musicae Scientiae 9 (2005), 75-109 (p. 78).

[7]     Richard Taruskin, ‘The Pastness of the Present,’ in Authenticity and Early Music, ed. by Nicholas Kenyon (Oxford: Oxford University Press, 1988), pp. 137-206 (p. 152).

[8]     Massimo Zicari, ‘Expressive Tempo Modifications in Adelina Patti’s Recordings: An Integrated Approach,’ Empirical Musicology Review 12.1-2 (2017), 42-56.

[9]     Clive Brown, Classical & Romantic Performance Practice 1750-1900 (Oxford: Oxford University Press, 1999) pp. 375-383.

[10]   Domenico Corri, The Singer’s Preceptor (London: Chappell & C., 1810) p. 6.

[11]   Ibid., p. 6.

[12]   Robert Toft, Bel Canto: a Performer’s Guide (Oxford: Oxford University Press, 2013) pp. 79-84.

[13]   Manuel Garcia, Scuola di Garcia. Trattato completo dell’arte del Canto, diviso in 2 parti. Traduzione dall’originale francese del maestro Alberto Mazzucato : Parte II (Milano: Ricordi, [1842]) p. 24. For the English translation see Garcia’s New Treatise on the Art of Singing (Boston: Ditson, 1857) p. 51.

[14]   Ibid., p. 25; Garcia’s New Treatise, p. 52.

[15]   References to the importance for a singer to approach tempo in a flexible and expressive manner can be found, among others, in Gioacchino Rossini, Metodo pratico di canto moderno (Berlin: Schlesinger, n.d), Luigi Lablache, Complete method of singing; or, A rational analysis of the principles according to which the studies should be directed for developing the voice (Boston: Ditson & Co., [1860, French ed. 1840]), Enrico Delle Sedie, A Complete Method of Singing (New York: Schirmer, 1894), Francesco Lamperti, Guida teorico-pratica-elementare per lo studio del canto (Milano: Ricordi, 1864).

[16]   As related in Luisa Cambi, Vincenzo Bellini. Epistolario (Milano: Mondadori, 1943) p. 490.

[17]   Manuel Garcia, Hints on Singing (New York: Edward Schuberth, 1894) p. 62.

[18]   The Athenaeum, 23 March 1850, p. 320.

[19]   The Athenaeum, 21 May 1853, p. 625.

[20]   For an extensive account of this discussion see again Gosset, Divas and Scholars.

[21]   No extensive monograph exists that deals with Sembrich’s career. However, the New York Public Library – Music Division owns a collection of documents that cover the entirety of Sembrich’s professional career from her childhood training to her death, and provides details of parts of her personal and family life (particularly the early years) through correspondence, legal and financial documents, newspaper clippings, and concert programs:

[22]   Harold Bruder, liner notes to Marcella Sembrich, the Victor Recordings (1908-19) (Romophone, 81027-2, 1998).

[23]   Herman Klein, The Reign of Patti (London: T. Fisher Unwin, 1920) p. 352.

[24]   Herman Klein, Herman Klein and the Gramophone, Being a Series of Essays on the Bel Canto (1923), the Gramophone and the Singer (1924-1934), and Reviews of New Classical Vocal Recordings (1925-1934), and Other Writings from the Gramophone. Ed. by William R Moran (Portland: Amadeus Press, 1990), p. 63.

[25]   Herman Klein and the Gramophone, p. 67.

[26]   Philip Miller, L. ‘Sembrich, Marcella,’ Grove Music Online (2001).

<> [Accessed 01 July 2019].

[27]   Thérèse Radic, Melba. The Voice of Australia (Melbourne: The Macmillan Company of Australia, 1986) p. 24.

[28]   Ann Blainey, I am Melba (Melbourne: Black Inc, 2009) p. 59.

[29]   Mathilde Marchesi, Variantes et Points d’Orgue (Paris: Huguel, 1900) p. 51.

[30]   See Henry Pleasants, The Great Singers from the Dawn of Opera to our own Time (New York: Simon and Schuster, 1966) pp. 271-272.

[31]   Ibid., p. 272.

[32]   Gattey, Charles Neilson. Luisa Tetrazzini. The Florentine Nightingale (Portland: Amadeus Press, 1995) p. 105.

[33]   Ibid., pp. 6-8.

[34]   ‘New Patti here, ready for debut at Hammerstein’s,’ The Evening World, night edition, 11 January 1908, front page.

[35]   Gattey, pp. 101-105.

[36]   The Sun, 16 January 1908, p. 7.

[37]   Miller, Ibid.

[38]   Radic, pp. 194-198.

[39]   Luisa Tetrazzini, My Life of Song (London: Cassell, 1921) p. 74.

[40]   Luisa Tetrazzini, How to Sing (New York: Doran, 1923) p. 95.

[41]   Ibid., p. 95.

[42]   Ibid. p. 95.

[43]   For the musical examples I refer to the edition originally published by G. Ricordi and republished by Dover: Giuseppe Verdi, La traviata (New York: Dover, 1990, M1500.V48T5, 89-755862) pp. 80-107. Of course, one can only speculate about the edition Sembrich, Melba and Tetrazzini may have used, given the number of reprints circulating at the end of the 19th century.

[44]   Abramo Basevi, Studio sulle opere di Giuseppe Verdi (Firenze: Tofani, 1859) p. 191. For a discussion on the use of this conventional formal segmentation see also Harold S. Powers, ‘La solita forma and The uses of Conventions.’ Acta Musicologica 59.1 (1987) 65-90.

[45]   Tutti i libretti di Verdi. Introduzione e note di Luigi Baldacci con una postfazione di Gino Negri, ed. by Luigi Baldacci (Torino: UTET, 1996), p. 276.

[46]   Steven Huebner, ‘Lyric Form in “Ottocento” Opera’ Journal of the Royal Musical Association 117.1 (1992) 123-147.

[47]   Baldacci, p. 276.

[48]   Gattey, Luisa Tetrazzini, p. 333.

[49]   Marcella Sembrich, the Victor Recordings (1908-19) (Romophone, 81027-2, 1998).

[50]   Marcella Sembrich the Victor Recordings (1904-08) (Romophone, 81026-2, 1997).

[51]   Nellie Melba, the 1904 Recordings (Naxos Historical, 8.110737, 2002).

[52]   Nellie Melba the complete Victor recordings (1907-1916) (Romophone, 81011-2, 1994).

[53]   Luisa Tetrazzini, the London recordings (EMI Classics, CHS 7 63802 2, 1992).

[54]   Tetrazzini (Nimbus Records, NI 7808, 1990).

[55]   Chris Cannam, Christian Landone, Mark Sandler, ‘Sonic Visualiser: An Open Source Application for Viewing, Analysing, and Annotating Music Audio Files,’ Proceedings of the ACM Multimedia 2010 International Conference, MM’10, October 25–29, 2010, Firenze, Italy.

[56]   Rebecca Plack, ‘The Substance of Style. How Singing Creates Sound in Lieder Recordings, 1902-1939’ (A Dissertation Presented to the Faculty of the Graduate School of Cornell University, May 2008), p. 188.

[57]   ‘Two critics and Tetrazzini’s high E flat,’ The Evening World, 16 January 1908, p. 3.

[58]   ‘Mme. Tetrazzini welcomed’ The Sun, 16 January 1908, p. 7.

[59]   Gosset, p. 330.

[60]   Letter preserved at the EMI Group Archive Trust in London, already partially reproduced in Charles Neilson Gattey, Luisa Tetrazzini, The Florentine Nightingale (Portland, US: Amadeus Press, 1995) p. 77.

[61]   ‘Making of a Great Singer’, The Sun, 8 March 1908, p. 6.