With the series of five performances as Verdi’s Lady Macbeth in December of 1952 (the only ones she ever gave in this role), Callas reached her zenith as a dramatic soprano—”a performance in a thousand,” I called it when surveying the rival LP pirate editions in 1973, and that is probably an understatement. All aspects of her assoluta self are enlisted (though only the D-flat of the Sleepwalking Scene takes her into the high extension); none of the troublesome bits intrude; and although for me de Sabata often pushes at the tempi, she takes his pulse and runs with it. Arguments can be advanced for some of her 1954-55 performances (the fabulous Scala/Karajan Lucia and Scala/Bernstein Sonnambula—I have not heard them all) as The Lady’s equal, and I won’t dispute the point. By that time, however, she was headed exclusively on a new course. Certainly it was set in part by two men, Tullio Serafin, who now supervised almost all her role preparation, and Walter Legge, who supervised almost all her recordings. But I think it’s clear that she sought an entire transformation, as woman and artist. She wanted an entirely different self-presentation, and was in a position to pursue that, onstage and off. Vocally, the remake involved a tradeoff of power for refinement. Perhaps she imagined that she would retain all the power she would need, but the repertory choices surely indicate a willingness to forego some of it in favor of becoming an Audrey Hepburn of the voice, as she worked with Biki to become an Audrey Hepburn of the figure. And she might fairly have concluded that while she’d had triumphs with her big roles, she was driving the voice too much to achieve them, and putting her high extension and agilità at risk.

But here’s the thing, as I see it: the alignment of the voice, fundamentally determined by that little turnover down there at the passaggio and the necessary compensation an octave higher, was not quite right no matter what repertory is being sung. And once the full dramatic calibration has been engaged, with every appearance of legitimately belonging to the voice, it’s extremely problematic to attempt a strategic withdrawal and use only part of it. A clear distress signal is run up during a RAI/San Remo concert on Dec. 27, 1954. The selections are all from the high/light side of Callas’ repertoire, and three of them sound fine. “Tutte le torture,” from Mozart’s Seraglio, is sung with all possible brilliance, the coloratura zipping along and the high Ds on the button. The Shadow Song from Meyerbeer’s Dinorah, though not vouchsafed quite the ingenuous timbre we might feel belongs to the part, is tossed off blithely. And we can say the same of one of Callas’ merriest showpieces, “Al dolce impero” from Rossini’s Armida. A big tract of Callas’ vocal territory, only recently in full view, is being left out, and that includes anything of any emotional depth, but we don’t really care for this particular occasion. And then, I believe for the first time anywhere, there’s something quite different, “Depuis le jour,” from Charpentier’s Louise—lyrical, sensuous, sustained, and, unlike the Shadow Song or the Lakmé Bell Song, sung in French. I don’t at all agree with Scott that Callas was not temperamentally suited to this example of “modern opera.” The piece is exquisitely felt, it already captures the character and the moment. But it cannot be solved with the dashing energy that works for everything else on the program, or with sheer heft of tone. And the voice cannot decide which parts of itself to hang onto, which to let go of. All the suspended notes above midvoice that cannot be sung full out (and they tend to occur, again, on varieties of “e” (destinée, rêver, grisée, baiser) are unsteady or threatening to be so, and are capped by a long high B on “heureuse” that wavers more widely than anything yet heard. The apparently simpler demands of moderate tempo, sustained legato, and gentle pressure are not simpler at all. They expose an underlying weakness in the “structure.”