Literature and score-analysis cutoff: 24 October 2026.

I first encountered Messiaen’s second mode — the octatonic scale — in an analysis seminar during my physics studies, played by a colleague on an upright piano in a rehearsal room with terrible acoustics. She demonstrated something that stopped me: no matter how many times she transposed the scale, she could never find more than three versions of it. After three transpositions up by a semitone she was back where she started. She called it the charm of impossibilities. It took me years to understand why it is impossible, and longer still to see that the answer is not musical but algebraic.

This post is a companion to Spiral Out: Tool’s Lateralus, the Fibonacci Sequence, and the Mathematics of Musical Structure, which found number theory in a prog-rock song. Here we find abstract algebra in twentieth-century sacred music.


Pitch Classes and the Chromatic Clock

Western music divides the octave into twelve equal semitones. For purposes of harmony and counterpoint, the absolute pitch is often less important than the pitch class — the equivalence class of all pitches related by octave transposition. Middle C and the C two octaves above belong to the same pitch class.

We label the twelve pitch classes $0, 1, 2, \ldots, 11$, with $0 = \mathrm{C}$, $1 = \mathrm{C}\sharp/\mathrm{D}\flat$, $2 = \mathrm{D}$, and so on up to $11 = \mathrm{B}$. Addition is taken modulo 12 — the integers wrap around like a clock face, with $11 + 2 = 1$ (two semitones above B is C$\sharp$).

The set of pitch classes with this operation is a group:

$$\mathbb{Z}_{12} = \{0, 1, 2, \ldots, 11\}, \qquad x \oplus y = (x + y) \bmod 12.$$

This is the cyclic group of order 12. It has an identity element ($0$, “no transposition”), every element has an inverse ($-n \bmod 12$), and the operation is associative. If you are used to thinking about the chromatic scale as a linear sequence ending at the octave, $\mathbb{Z}_{12}$ is the insistence that it is actually a circle.


Musical Operations as Group Elements

Two operations are fundamental in tonal and post-tonal music theory.

Transposition by $n$ semitones maps every pitch class up by $n$:

$$T_n \colon x \mapsto x + n \pmod{12}.$$

The twelve transpositions $T_0, T_1, \ldots, T_{11}$ are exactly the elements of $\mathbb{Z}_{12}$, with $T_n$ corresponding to the integer $n$. Composing two transpositions gives a transposition: $T_m \circ T_n = T_{m+n}$.

Inversion reflects the pitch-class circle:

$$I \colon x \mapsto -x \pmod{12}.$$

Inversion maps C to C, D to B$\flat$, E to A$\flat$, and so on — it is the mirror symmetry of the chromatic circle about the C/F$\sharp$ axis. Combining inversion with transposition gives the inversional transpositions:

$$I_n \colon x \mapsto n - x \pmod{12}.$$

The transpositions and inversional transpositions together generate a group of order 24:

$$D_{12} = \langle T_1, I \rangle.$$

This is the dihedral group $D_{12}$ — the same abstract group that describes the symmetries of a regular 12-gon (twelve rotations and twelve reflections). The identification is not coincidental: the twelve pitch classes arranged in a circle are the vertices of a regular 12-gon, and the musical operations are geometrically the symmetries of that polygon.

Twelve-tone rows also use transposition, inversion, retrograde, and their combinations, but retrograde acts on order rather than pitch class alone. A 24-element pitch-class dihedral group does not by itself contain all four row forms or justify a simple coset identification.


Orbits and Stabilisers

Let $S \subseteq \mathbb{Z}_{12}$ be a pitch-class set — a chord, a scale, a collection of any size.

The orbit of $S$ under $\mathbb{Z}_{12}$ is the collection of all distinct transpositions of $S$:

$$\mathrm{Orb}(S) = \{ T_n(S) : n \in \mathbb{Z}_{12} \}.$$

For most sets, all twelve transpositions produce a different set, so $|\mathrm{Orb}(S)| = 12$. The C major scale, for example, has twelve distinct transpositions, one for each key.

But some sets are symmetric under certain transpositions: there exists $n \neq 0$ such that $T_n(S) = S$. The collection of all symmetry transpositions of $S$ is the stabiliser:

$$\mathrm{Stab}(S) = \{ T_n \in \mathbb{Z}_{12} : T_n(S) = S \}.$$

Because composing two symmetry transpositions yields another, $\mathrm{Stab}(S)$ is a subgroup of $\mathbb{Z}_{12}$.

The orbit–stabiliser theorem gives the fundamental count:

$$|\mathrm{Orb}(S)| \cdot |\mathrm{Stab}(S)| = |\mathbb{Z}_{12}| = 12. \tag{1}$$

The number of distinct transpositions of $S$ equals $12$ divided by the number of transpositions that leave $S$ unchanged. The more internally symmetric $S$ is, the fewer new versions you can produce by transposing it.

A set with $|\mathrm{Stab}(S)| > 1$ has limited transpositional orbit in this algebraic sense.


Mode 1: The Whole-Tone Scale

The whole-tone scale contains the six even pitch classes:

$$\mathrm{Mode\ 1} = \{0, 2, 4, 6, 8, 10\}.$$

Transposing by $T_2$:

$$T_2(\{0, 2, 4, 6, 8, 10\}) = \{2, 4, 6, 8, 10, 0\} = \{0, 2, 4, 6, 8, 10\}. \checkmark$$

The set is unchanged. The same holds for $T_4, T_6, T_8, T_{10}$. The stabiliser is the full subgroup of even transpositions:

$$\mathrm{Stab}(\mathrm{Mode\ 1}) = \{T_0, T_2, T_4, T_6, T_8, T_{10}\} \cong \mathbb{Z}_6.$$

By the orbit–stabiliser theorem:

$$|\mathrm{Orb}(\mathrm{Mode\ 1})| = \frac{12}{6} = 2.$$

There are exactly two distinct whole-tone scales. Every pianist learns this: the one on C and the one on C$\sharp$. Composing with whole-tone harmony means working from a stock of only two harmonic pools with no way to modulate into a genuinely new version of the scale. This is Messiaen’s first charm of impossibility.


Mode 2: The Octatonic Scale

The octatonic (diminished) scale alternates half-step and whole-step intervals. Starting on C:

$$\mathrm{Mode\ 2} = \{0, 1, 3, 4, 6, 7, 9, 10\}.$$

Does $T_3$ leave this set invariant?

$$T_3(\{0, 1, 3, 4, 6, 7, 9, 10\}) = \{3, 4, 6, 7, 9, 10, 0, 1\} = \{0, 1, 3, 4, 6, 7, 9, 10\}. \checkmark$$

Also $T_6$ and $T_9$. The stabiliser is the subgroup generated by transposition by a minor third:

$$\mathrm{Stab}(\mathrm{Mode\ 2}) = \{T_0, T_3, T_6, T_9\} \cong \mathbb{Z}_4.$$

The orbit size:

$$|\mathrm{Orb}(\mathrm{Mode\ 2})| = \frac{12}{4} = 3.$$

There are exactly three distinct octatonic collections under transposition. The cosets of the subgroup $\langle T_3 \rangle$ show how they fit together. Write

$$C_k = \{k,\ k+3,\ k+6,\ k+9\}, \qquad k = 0, 1, 2,$$

for the three cosets, the diminished seventh chords on C, C$\sharp$ and D. Mode 2 on C is $C_0 \cup C_1$, and its transpositions are

$$\mathrm{Oct}_k = T_k(\mathrm{Mode\ 2}) = C_k \cup C_{k+1}, \qquad k = 0, 1, 2,$$

with the index read modulo 3. Each collection is two cosets and omits the third, so it is the complement of a diminished seventh chord, and any two collections share exactly one coset, four pitch classes (Fig. 1). $T_3$ maps every coset onto itself, so it fixes the scale; $T_1$ moves $C_0$ to $C_1$, $C_1$ to $C_2$ and $C_2$ back to $C_0$, so three semitone steps return the scale to where it started.

Figure 1. A whole-tone scale is one coset; an octatonic collection is two. Pitch classes run clockwise from 0 at the top; filled points are the scale. Mode 1 on C is the subgroup $\langle T_2 \rangle$ (solid), and its other coset (dashed) is the other whole-tone scale. The octatonic collections are unions of two of the cosets $C_0$ (solid), $C_1$ (dashed) and $C_2$ (dotted) of $\langle T_3 \rangle$; the faint square is the coset each omits.

The Subgroup Lattice and All Seven Modes

The orbit–stabiliser theorem constrains which stabiliser sizes are algebraically possible. Since $\mathrm{Stab}(S)$ is a subgroup of $\mathbb{Z}_{12}$, its order must divide 12. The proper non-trivial subgroups of $\mathbb{Z}_{12}$ — those with order strictly between 1 and 12 — are precisely:

SubgroupGeneratorOrderOrbit size
$\langle T_2 \rangle = \{T_0, T_2, T_4, T_6, T_8, T_{10}\}$$T_2$62
$\langle T_3 \rangle = \{T_0, T_3, T_6, T_9\}$$T_3$43
$\langle T_4 \rangle = \{T_0, T_4, T_8\}$$T_4$34
$\langle T_6 \rangle = \{T_0, T_6\}$$T_6$26

These four subgroups exist because the proper divisors of 12 that are greater than 1 are exactly $\{2, 3, 4, 6\}$. The subgroups of $\mathbb{Z}_n$ are in bijection with the divisors of $n$ — a consequence of the fundamental theorem of cyclic groups. Since $12 = 2^2 \times 3$, the proper divisors are $1, 2, 3, 4, 6$. Twelve Is Not an Accident: The Group Theory of Musical Tuning draws the same subgroups as orbits on the pitch-class circle.

Messiaen’s seven modes occupy all four rows. The table writes each from C, with the number of transpositions he gives [[1]], its steps in semitones, and its stabiliser (Papadopoulos writes Mode 7 a semitone higher, which is the same class):

ModePitch classesStepsStabiliserTranspositions
1$\{0, 2, 4, 6, 8, 10\}$(2) × 6$\langle T_2 \rangle$2
2$\{0, 1, 3, 4, 6, 7, 9, 10\}$(1, 2) × 4$\langle T_3 \rangle$3
3$\{0, 2, 3, 4, 6, 7, 8, 10, 11\}$(2, 1, 1) × 3$\langle T_4 \rangle$4
4$\{0, 1, 2, 5, 6, 7, 8, 11\}$(1, 1, 3, 1) × 2$\langle T_6 \rangle$6
5$\{0, 1, 5, 6, 7, 11\}$(1, 4, 1) × 2$\langle T_6 \rangle$6
6$\{0, 2, 4, 5, 6, 8, 10, 11\}$(2, 2, 1, 1) × 2$\langle T_6 \rangle$6
7$\{0, 1, 2, 3, 5, 6, 7, 8, 9, 11\}$(1, 1, 1, 2, 1) × 2$\langle T_6 \rangle$6

The step column puts into numbers Messiaen’s description of the modes as “divisible in small symmetric groups” (Papadopoulos’s translation [[1]]): one block of steps repeated around the octave. If the shortest repeating block adds up to $t$ semitones, the stabiliser is $\langle T_t \rangle$, of order $12/t$, and by Eq. (1) the mode has exactly $t$ distinct transpositions. Mode 3’s block, a whole tone and two semitones, adds up to four: four transpositions.

The containment relations are: $\langle T_2 \rangle \supset \langle T_4 \rangle$ and $\langle T_2 \rangle \supset \langle T_6 \rangle$ and $\langle T_3 \rangle \supset \langle T_6 \rangle$. Correspondingly, Mode 1 (stabiliser $\langle T_2 \rangle$, order 6) is “more limited” than Mode 3 (stabiliser $\langle T_4 \rangle$, order 3), in the sense that $\langle T_4 \rangle \subset \langle T_2 \rangle$: every transposition that fixes Mode 3 also fixes Mode 1.


Why Exactly Seven Modes?

After presenting his seven, Messiaen declared the list complete: “The series is closed. It is mathematically impossible to find others, at least in our tempered system of 12 semi-tones” ([[2]], vol. 1, p. 51, as translated in [[1]]). Counting checks the claim. Every non-trivial subgroup of $\mathbb{Z}_{12}$ contains $\langle T_4 \rangle$ or $\langle T_6 \rangle$, so a set has limited transposition exactly when $T_4$ or $T_6$ fixes it. A set fixed by $T_t$ is a union of some of the $t$ cosets of $\langle T_t \rangle$, so there are $2^t$ of them: $T_6$ fixes the 64 unions of tritone pairs $\{x, x+6\}$, $T_4$ the 16 unions of augmented triads, and both together only the 4 sets that $T_2$ fixes. Of the $2^{12} = 4{,}096$ pitch-class sets,

$$64 + 16 - 4 = 76$$

have a non-trivial stabiliser. Two of them, the empty set and the full chromatic aggregate, are fixed by every transposition. Sorting the other 74 by their exact stabiliser (each row subtracts the sets that a larger subgroup also fixes) and dividing by the orbit sizes from Eq. (1) gives the transposition classes:

StabiliserSetsDistinct transpositionsClassesMessiaen’s modes
$\langle T_2 \rangle$$4 - 2 = 2$211
$\langle T_3 \rangle$$8 - 2 = 6$322
$\langle T_4 \rangle$$16 - 4 = 12$433
$\langle T_6 \rangle$$64 - 4 - 8 + 2 = 54$694, 5, 6, 7

That is fifteen classes, eight more than Messiaen lists; a brute-force enumeration of all 4,096 sets gives the same numbers. His seven follow size: every class with more than six notes (Modes 2, 3, 4, 6 and 7), two of the five six-note classes (Modes 1 and 5), and none of the five smaller ones, chords such as the augmented triad and the diminished seventh. The six-note classes left out are the hexatonic collection $\{0, 1, 4, 5, 8, 9\}$ and the mutually inverse pair $\{0, 1, 3, 6, 7, 9\}$ and $\{0, 1, 4, 6, 7, 10\}$ (Fig. 2).

Figure 2. Messiaen’s seven include every class with more than six notes and none with fewer than six. Each point is one of the 15 transposition classes with a non-trivial stabiliser besides the empty set and the aggregate, placed by its distinct transpositions, $12/|\mathrm{Stab}(S)|$, and its number of notes; coincident points are spread sideways. From an enumeration of all 4,096 subsets of $\mathbb{Z}_{12}$.

Why these seven is not settled. Papadopoulos offers two readings: the sets left over are truncations of the seven, or Messiaen simply missed them [[1]]. The first has support in the later Traité, where, according to Papadopoulos, Messiaen calls further six-times-transposable collections truncated modes rather than modes [[3]]. On that reading the closed-series claim is true: each omitted class lies inside a transposition of Mode 3 or Mode 7. But truncation proves too much. Modes 1, 2, 4, 5 and 6 lie inside Mode 3 or Mode 7 as well, so by that criterion two modes would have been enough. My own reading is that he kept the scales he could use as raw material for his harmonic language, which would explain why the seven are the large ones. Inclusion was not use, though: he lists Mode 1 but says he will avoid it unless it is hidden in a superimposition of modes, because Debussy and Dukas left nothing to add ([[2]], vol. 1, p. 52, as translated in [[1]]).

The group theory explains the constraint, not the selection. What it determines is the number of levels (four: orbit sizes 2, 3, 4, 6), the fifteen classes that occupy them, and the impossibility of any mode with, say, five distinct transpositions (since 5 does not divide 12).


What Messiaen Knew — and Did Not Know

Messiaen described his modes in Technique de mon langage musical (1944) [[2]]. His account is musical and phenomenological, not algebraic: a mode is “two times transposable” or “three times transposable” (Papadopoulos’s translation [[1]]), and the “charm of impossibilities”, at once voluptuous and contemplative, lies “particularly in certain mathematical impossibilities of the modal and rhythmic domains”: the modes cannot be transposed further “because one always falls back on the same notes” ([[2]], vol. 1, p. 5; my translation of the French quoted in [[3]]). The listener is promised “a certain effect of tonal ubiquity”, leading towards a “theological rainbow” (p. 13, quoted in [[3]]). I read this as harmonic stasis that a devout Catholic heard as a figure of divine eternity: a mode that cannot depart is a glimpse of the unchanging. That is my gloss, not his wording.

He was not doing group theory, if that means its formal apparatus: he did not write pitch classes as integers modulo 12, had little mathematical background and used his own words for the objects [[1]]. Papadopoulos reads his awareness more strongly, as a conscious and systematic use of “mathematical concepts derived from symmetry and groups” [[1]]. The difference is between deliberate symmetry, which I do not doubt, and the theory that names it. The orbit–stabiliser theorem (in its abstract form) postdates Lagrange (1771), Cauchy (early 19th century), and Galois (1832), and as far as the sources I checked show, its language reached music theory after his treatise: some serialists knew in the 1950s that they were dealing with groups [[1]], Milton Babbitt was among the first to analyse music with group theory, around 1960 [[4]] [[5]], David Lewin’s first article (1959) tied interval content to the discrete Fourier transform in a two-line remark [[6]], and the pitch-class set framework used here is set out in Allen Forte’s 1973 book [[7]] and generalised in Lewin’s 1987 book [[8]].

Messiaen enumerated the transpositions without presenting this group-theoretic formalism. The algebra retrospectively proves the orbit counts; it does not tell us whether he discovered the modes by ear or by another procedure.


From Messiaen to Lewin

Lewin’s transformational theory (1987) generalises the $\mathbb{Z}_{12}$ framework to arbitrary musical spaces [[8]]. A Generalized Interval System is a triple $(S, G, \mathrm{int})$: a set $S$ of musical objects, a group $G$ and a map $\mathrm{int} : S \times S \to G$ with two axioms [[9]]. Intervals compose, $\mathrm{int}(r, s)\,\mathrm{int}(s, t) = \mathrm{int}(r, t)$, and for every $s \in S$ and $g \in G$ exactly one $t \in S$ has $\mathrm{int}(s, t) = g$. Lewin proved this equivalent to a group acting simply transitively on $S$ [[9]].

Pitch classes under transposition are the model case, with $\mathrm{int}(x, y) = y - x \bmod 12$, and nothing in the axioms requires pitch: time points and durations satisfy them too. The transpositions of a Messiaen mode do not. $T_1$, $T_4$, $T_7$ and $T_{10}$ all carry $\mathrm{Oct}_0$ to $\mathrm{Oct}_1$, and only the quotient $\mathbb{Z}_{12}/\langle T_3 \rangle \cong \mathbb{Z}_3$ acts simply transitively on the three octatonic collections.

Neo-Riemannian theory, which grew out of Lewin’s work in the 1980s [[5]] and was surveyed by Cohn (1998) [[10]], applies this framework to triads. P, L and R exchange a major or minor triad with its parallel, its leading-tone exchange and its relative respectively, and the group they generate on the 24 major and minor triads is dihedral of order 24, isomorphic to $D_{12}$ above [[5]]: the same abstract group acting on a different musical space. Messiaen’s modes needed only its transposition subgroup $\mathbb{Z}_{12}$; inversion never entered.

Emmanuel Amiot’s work applies the discrete Fourier transform to pitch-class sets [[11]]. With the convention of Two Chords, One Interval Fingerprint, the coefficients of a set $S$ are

$$\widehat{S}(k) = \sum_{s \in S} e^{-2\pi i k s/12}, \qquad k = 0, 1, \ldots, 11.$$

Limited transposition is a pattern of exact zeros. If $T_t(S) = S$, relabelling $s \mapsto s + t$ gives $\widehat{S}(k) = e^{-2\pi i k t/12}\,\widehat{S}(k)$, so $\widehat{S}(k) = 0$ unless 12 divides $kt$; conversely, if those coefficients vanish, $T_t(S)$ and $S$ have the same coefficients and are the same set. The octatonic collection keeps only $k = 0, 4, 8$, the whole-tone scale only $k = 0, 6$. The magnitudes measure in a graded way how close a set comes to a regular polygon [[6]]; only the zeros identify a stabiliser.

Group theory is one important mathematical framework in music theory, not its sole or dominant framework; Tymoczko’s geometry of chords is another [[12]] [[13]]. The limited transposition of Messiaen’s modes is an algebraic fact; their musical significance is not determined by algebra alone.

The limitation is a theorem of the transposition action. The charm is Messiaen’s.


References

[1] Papadopoulos, A. (2014). Mathematics and group theory in music. Preprint, written for the Handbook of Group Actions, vol. II (L. Ji, A. Papadopoulos & S.-T. Yau, Eds.). arXiv:1407.5757

[2] Messiaen, O. (1944). Technique de mon langage musical (2 vols.). Alphonse Leduc. English translation by J. Satterfield, The Technique of My Musical Language, Alphonse Leduc, 1956.

[3] Papadopoulos, A. (2019). Messiaen et les mathématiques. In R. Illiano (Ed.), Twentieth-Century Music and Mathematics. Brepols. Preprint: arXiv:2008.11936

[4] Babbitt, M. (1960). Twelve-tone invariants as compositional determinants. The Musical Quarterly, 46(2), 246–259. DOI: 10.1093/mq/XLVI.2.246

[5] Crans, A. S., Fiore, T. M., & Satyendra, R. (2009). Musical actions of dihedral groups. The American Mathematical Monthly, 116(6), 479–495. DOI: 10.1080/00029890.2009.11920965

[6] Amiot, E. (2006). Gammes bien réparties et transformée de Fourier discrète. Preprint. arXiv:math/0604210

[7] Forte, A. (1973). The Structure of Atonal Music. Yale University Press.

[8] Lewin, D. (1987). Generalized Musical Intervals and Transformations. Yale University Press. (Reissued Oxford University Press, 2007.)

[9] Fiore, T. M., Noll, T., & Satyendra, R. (2013). Morphisms of generalized interval systems and PR-groups. Journal of Mathematics and Music, 7(1), 3–27. DOI: 10.1080/17459737.2013.785724

[10] Cohn, R. (1998). Introduction to neo-Riemannian theory: A survey and a historical perspective. Journal of Music Theory, 42(2), 167–180. DOI: 10.2307/843871

[11] Amiot, E. (2016). Music Through Fourier Space: Discrete Fourier Transform in Music Theory. Springer (Computational Music Science). DOI: 10.1007/978-3-319-45581-5

[12] Tymoczko, D. (2006). The geometry of musical chords. Science, 313(5783), 72–74. DOI: 10.1126/science.1126287

[13] Tymoczko, D. (2011). A Geometry of Music: Harmony and Counterpoint in the Extended Common Practice. Oxford University Press.


Changelog

  • 2026-01-14: Changed “cosets of $D_{12}$” to “cosets of $\mathbb{Z}_{12}$ (the transposition subgroup)” in the twelve-tone composition paragraph. $D_{12}$ (order 24) already includes both transpositions and inversions, yielding only 2 cosets in the full serial group. The four row forms {P, I, R, RI} correspond to 4 cosets of the transposition-only subgroup $\mathbb{Z}_{12}$ (order 12) in the full group of order 48.
  • 2026-07-11: Separated Messiaen’s seven selected modes from the subgroup lattice, corrected the serial/coset and Fourier overstatements, and removed unsupported discovery and influence claims.
  • 2026-10-03: Corrected “one semitone above B is C♯” to “two semitones” (11 + 2 ≡ 1; one semitone above B is C), and corrected the opening anecdote: a minor-third transposition of the octatonic scale returns the same scale at once, whereas the three distinct versions appear under semitone transposition (12/4 = 3). Orbit sizes, subgroup table and stabilisers rechecked unchanged.
  • 2026-10-24: Added the count behind “Why seven?”: 76 of 4,096 sets have a non-trivial stabiliser, forming 15 transposition classes besides the empty set and the aggregate, of which Messiaen names seven (Fig. 2); his “the series is closed” holds only for truncations. Replaced “scales of moderate cardinality” with every class above six notes plus two six-note classes. Worked out Modes 3–7 in a table and the octatonic cosets exactly (Fig. 1). Corrected “the same dihedral group that governs Messiaen’s modes” (they use only $\mathbb{Z}_{12}$), the order of L, P and R (P parallel, L leading-tone exchange, R relative), and “every symmetry of Mode 3 is also a symmetry of Mode 1’s stabiliser” to “every transposition that fixes Mode 3 also fixes Mode 1”. Narrowed “emerged from Lewin’s work in the 1980s and 1990s and was systematised by Cohn” to Lewin’s 1980s work, surveyed by Cohn, and “not part of music-theoretic discourse until Milton Babbitt’s work in the 1950s” to the evidence: serialists knew of groups in the 1950s, and Lewin’s 1959 article used the Fourier transform. Added the second GIS axiom and the Fourier-zero criterion, marked the divine-eternity reading as mine, and numbered the references (five added).