Mfcttrf

Messiaen's Prelude for Organ (no opus number, discovered posthumously in 1997, composed probably circa 1929) is rife with ambiguous accidentals. Some can be figured out by examining similar passages, but these two cases seem particularly egregious:

In bar 59, alto, are the unmarked G noteheads flat or natural?

In bar 61, alto, is the last G sharp or natural?

Can we even tell, without conducting a séance?

Asked another way: at those chromatically altered unisons in soprano and alto, which accidental applies to which voice? Is there a rule for that? If we knew the alto's accidental, we could then let it continue for the later alto notes.

(Were the note heads (or the staves!) separate, with an accidental immediately preceding each, the confusion would vanish.)

Explanations from the musical text, from a facsimile edition if such exists, or from rules of notation outweigh reports that so-and-so played it in such-and-such a way.

If you make the assumption that the voices don't cross, and that the accidentals apply to the voices individually, things seem to fall into place.

In the second half of bar 59, if the lower voice is G flat and the upper is G natural, the lower voice has the pattern

Eb F Eb F Gb

Eb Gb Eb Gb G

which has some logic to it.

Similarly in bar 61, if the upper voice is G sharp and the lower is G natural, the final G should be a natural, which repeats the D#-G tritone at the start of the bar.

Ideally, these details should be interpreted with reference to all the rest of the score, but we don't have that available. Messiaen's music is usually highly structured, so these passages are unlikely to be just "free atonal improvisation."

In bar 59, there is no point in putting the natural sign - it's not that the G note was double flatted. So it's Gnat initially, then G♭ for the subsequent two.

In bar 61, the G♯ from the key sig. is made natural on the 1st G, then sharpened with the accidental on the 2nd, and that makes the 3rd G G♯ too.

Straightforward, messed up by bad writing (publishing?). Not played it, but that's what it says to play.

1. 
   

   Here's the argument for bar 59's altered unison's soprano being G natural (and thus alto G flat).

   
   
   

   • In the whole piece, voices never cross. At least not more than a
     semitone in cases like these, so it would be statistically odd to
     have only those as voice crossings. Simpler to say that none were
     intended at all.

   
   

   • In bars 58-60, soprano traces out G natural rising to D, filled in
     variously.

   
   

   • Were bar 59 alto's last note G natural, then it would have been tied
     to the G natural immediately across the barline. Every other
     repeated notehead in the whole piece, even across voices and across
     staves, is scrupulously tied, e.g. bar 61's A-A. This happens at least
     a hundred times. So alto must be G flat, and hence soprano must be G natural.

   
   



2. Beyond no voice crossings, the only argument for bar 61's altered unison's alto being G natural is that it avoids
   making the bar's final chord a prominent open fifth, which would be
   a purposelessly startling consonance in this densely harmonic texture. There's no parallel passage to compare this bar to.

Gardner Read's "Music Notation," 2nd ed., pp. 73-74, covers how to stem and mark altered unisons.
But it says nothing about
the case at hand. All the examples use multiple note heads, to prevent
ambiguity about which accidental applies to which voice. By forbidding a single
notehead (even two-stemmed) with multiple accidentals, this avoids inventing
a morass of rules such as "the first accidental applies to the upper voice."

So the published notation is rubbish. Even if the manuscript used this notation and the editor was too reverential to improve it, at least a footnote to that effect would have been justified!