Abstract
Scales are a fundamental concept of musical practice around the world. They commonly exhibit symmetry properties that are formally studied using cyclic groups in the field of mathematical scale theory. This paper proposes an axiomatic framework for mathematical scale theory, embeds previous research, and presents the theory of maximally even scales and well-formed scales in a uniform and compact manner. All theorems and lemmata are completely proven in a modern and consistent notation. In particular, new simplified proofs of existing theorems such as the equivalence of non-degenerate well-formedness and Myhill's property are presented. This model of musical scales explicitly formalizes and utilizes the cyclic order relation of pitch classes.
1. Introduction
The concept of a musical scale as a collection of pitch classes is universal throughout the world (CitationSavage et al. 2015), even though these scales exhibit great variety and differences. While the notion of the octave and octave equivalence are found universal, differences involve the amount of the “tone material” within the octave, the scale's interval patterns, and the tuning systems. On the other hand, commonalities include the amount of tones in a scale often ranging between five and sevenFootnote1, intervals between adjacent scale tones being between one and three semitones in size, and asymmetric scales being more common than symmetric ones (CitationPatel 2008).
Mathematical scale theory studies the inner symmetry structures of a scale, or, respectively, a single chord, key, or rhythm. It originally emerged from the study of European and North-American classical and popular music (CitationClough and Myerson 1985; CitationCarey and Clampitt 1989; CitationClough and Douthett 1991; CitationTymoczko 2008) and was generalized to study rhythmical structures (CitationWooldridge 1992) and musical chords inside scales (CitationDouthett 2008). In particular, classical diatonic scale theory studies and generalizes the properties of major scales such as the C major scale consisting of the pitch classes C, D, E, F, G, A, and B. Two important outcomes of mathematical scale theory are the discovery of general scale properties that can be found in keys as well as in chords and rhythms and the development of a concise vocabulary of music theoretical concepts.
This paper addresses the research communities of mathematical music theory, computational musicology, and music cognition. Its main contribution is to reformulate mathematical scale theory within an algebraic framework based on the concept of cyclic order, presenting the theory in a compact manner and at a more abstract level than in previous work. We describe core results and proofs of the field in a modern mathematical language, revealing cross-relations between previously unrelated proofs. The second most important contribution is the self-contained presentation of the arguments used in the proofs of this paper. Music research is inherently interdisciplinary. We therefore give detailed descriptions of the arguments used in the proofs to open up scale theory to a broad audience of music researchers with sufficient mathematical background. We believe that it is crucial to facilitate interdisciplinary research on music wherever possible.
Cross-cultural commonalities of scales have been argued to arise from psychological constraints. The typical number of scale tones relates to the amount of distinct pitches that a human can reliably keep track of (CitationMiller 1956; CitationClough, Douthett, and Ramanathan 1993). The prevalence of moderately small intervals between consecutive scale degrees may result from constraints that large intervals are more difficult to sing and too small intervals are hard to distinguish by the human ear Citation(Burns and Ward 1978). The most remarkable commonality is the asymmetry that enables the creation of complex tonal structures by making the positions in the scale easy to perceive and to remember (CitationBalzano 1980; CitationTrehub, Schellenberg, and Kamenetsky 1999). It has been found from a mathematical perspective, however, that there is only one – up to isomorphism unique – “maximally asymmetric” scale for each tone material and given scale size (CitationClough and Douthett 1991). More sophisticated concepts and quantifications of asymmetry are therefore necessary to account for subtle characteristics of scales. Both the diminished-seventh chord and the octatonic scale (also known as the second Messiaen mode or the half-step/whole step diminished scale (CitationLevine 1995) can for example be considered symmetric. Closer inspection reveals, however, that the former is totally symmetric – that is every tone has the same relations to all other tones – while the latter is not.
This paper argues that the transpositional symmetry of a scale can be formalized using automorphism groups that model different degrees of symmetry. Central results from previous research are the formulation of scale properties such as maximal evenness (CitationClough and Douthett 1991), well-formedness (CitationCarey and Clampitt 1989), and Myhill's property (CitationClough and Myerson 1985) and theorems about the relations between those properties to distinguish between different kinds of transpositionally non-symmetric scales. A taxonomy of scale properties can be found in CitationClough, Engebretsen, and Kochavi (1999), a formal concept lattice is presented in CitationNoll (2015), and an overview of different scale definitions used in the literature can be found in CitationHarasim, Schmidt, and Rohrmeier (2016).
More generally, both pitch-class collections and rhythms are cyclic structures, the former in the pitch class dimension and the latter with respect to bars in western music or talas in Indian music. In this analogy, a rhythmic pattern is a sequence of note onsets drawn from a metrical grid just as a western scale is a sequence of pitch classes drawn from a chromatic universe (CitationPressing 1983; CitationWooldridge 1992). Because of the similar formal nature of chords, keys, and rhythms, this paper therefore uses the word scale to refer to any of these concepts. In particular, chords such as triads can be understood as a collection of pitch classes chosen from a diatonic scale analogous to that a western scale is a collection of pitch classes chosen from the chromatic scale. The framework proposed here therefore particularly includes the description of the building blocks of sequential structures of music (CitationConklin and Witten 1995; CitationCohn 1996, Citation1997; CitationTymoczko 2006; CitationRohrmeier 2011; CitationHarasim, Rohrmeier, and O'Donnell 2018; CitationSears et al. 2018; CitationMoss et al. 2019).
2. Scales as cyclic embeddings
All scales in the enharmonically identified chromatic universe using western 12-tone equal temperament can be visualized by a vertex-colored directed circle of length 12 where vertices contained in the scale are colored black and all remaining vertices are colored white. A scale is called symmetric if this graph has a non-trivial automorphism group. Figure shows three scales. The first one is a diatonic scale. It contains seven tones and is not symmetric. The second scale is a hexatonic scale. It is a symmetric scale containing six tones that is isomorphic to its complement. The third scale of Figure is an octatonic scale. It distributes its eight scale tones as evenly as possible in the chromatic scale.
Musical scales are equipped with a cyclic order relation. For example, the pitch class (pc) D is not generally higher than the pc C, since the pitch of D can be realized in a lower octave than the pitch of C. If the pc cycle is, however, cut at any pc, it defines a linear order with respect to the cut point as the lowest pitch. If the pc cycle is for example cut at B, then the result inherits a linear order in which C is lower than D. If the cycle is in contrast cut at C♯, then D is lower than C in the implied linear order. Although previous research has described formal properties of musical scales and proposed mathematical proofs of their relations, to the best of our knowledge so far no study has explicitly modeled this cyclic order relation nor given an axiomatic description of musical scales. Mathematical scale theorists have, however, probably always thought about cyclic orderings of, for example, the chromatic scale, but never used this relational concept explicitly, see for example CitationCarey (1998), CitationNoll (2007), and CitationTymoczko (2008). Cyclic ordering is a standard concept of modern algebra formally defined as follows (CitationHuntington 1916; CitationMegiddo 1976; CitationGalil and Megiddo 1977).
Definition 2.1
A partially cyclic ordered set consists of a set S and a ternary relation such that
(strictness)
(cyclicality),
(transitivity), and
(antisymmetry)
for . It is called (totally) cyclic ordered if additionally it is total. That is, either or for distinct .
Note that the statement implies that s, t, and u are mutually different. Cyclic order relations are therefore most naturally compared to strict total order relations. For any set S and ternary relation , the statement that is cyclic ordered is equivalent to the statement that for all , the binary relation defined by is a strict total order relation on S. For distinct , the intuition behind is that t lies on the path from s to u.
Definition 2.2
Let S and T be cyclic ordered sets. A cyclic morphism is a mapping that reflects the cyclic order. This is for all . It is a cyclic embedding if additionally it is injective and preserves the cyclic order, for . A cyclic isomorphism is a surjective cyclic embedding.
Definition 2.3
A scale is defined as a cyclic embedding of a finite cyclic ordered set D of diatonic tones into a finite cyclic ordered set C of chromatic tones. The cardinalities of the domain and codomain of σ are called diatonic size and chromatic size, respectively. They are denoted by d and c.
Figure illustrates the definitions using the example of the C major scale. The sets of diatonic and chromatic tones D and C are modeled by and , the rings of integers modulo 7 and 12. Throughout the whole paper, we use the symbol for any positive natural number n to denote the set of non-negative numbers less than n together with the addition modulo n. We particularly use instead of just + in situations where we want to stress calculations modulo n. We also use the subscript n to denote other operations modulo n, such as . We identify the pitch class of C with 0. Then the diatonic scale is visualized in Figure . The diatonic and chromatic sizes of, for example, the interval between the tone D, identified with 1 mod 7 and 2 mod 12, and the tone A, identified with 5 mod 7 and 9 mod 12, are and , respectively.
A characteristic feature of a scale is the possibility to calculate the size of musical intervals between tones (paths along the circle in clockwise direction) in two different ways (CitationClough and Myerson 1985, Citation1986). In the circle visualization of Figure , the specific or chromatic size of an interval between two (black) scale tones is one plus the number of white and black vertices strictly in between them in clockwise direction. The generic or diatonic size of an interval between two scale tones is one plus the number of only black tones strictly in between them in clockwise direction.
3. Cyclic ordered sets, generalized interval systems, and regular group actions
This section provides the preliminary definitions and notation of generalized interval systems and regular group actions as subclasses of them are in 1-to-1 relation to finite cyclic ordered sets. Technical proofs and lemmata are presented in the appendix in detail. The core result of this section is Theorem 3.5. It shows how the algebraic concepts,
finite generalized interval systems with cyclic interval group,
regular group actions of a finite cyclic group, and
finite cyclic ordered sets,
are three different viewpoints of a pitch class system. Generalized interval systems formalize how intervals between tones are calculated. Regular cyclic group actions model the concept of musical transposition. A transposition of a tone t up by an interval i is denoted by (read as “t transposed by i”). Note that other authors also use the notation (CitationLewin 1987; CitationClough and Douthett 1991; CitationTymoczko 2008). Cyclic orderings show how to formalize the intuition that, for example, the pitch class of F is higher than the pitch class of E and lower than the pitch class of G. Because of the cyclic nature of the pitch class system, a binary order relation cannot be used here to model that intuition.
Definition 3.1
A generalized interval system (GIS) is a triple where S is a set of tones, is a group of (generalized) intervals, and maps pairs of tones to intervals such that the restriction is bijective and for all .
We use the definition by CitationLewin (1987) here with the notation by CitationNeumaier and Wille (1990) who referred to GISs as tone systems. The group does in general not need to be commutative. In this paper, however, only cyclic interval groups are considered that are commutative. Note that the equations and follow directly from the definition for all .
Definition 3.2
A (right) group action is a mapping that satisfies
(identity) and
(compatibility)
for and where S is a set and is a (in general not necessarily commutative) group. We jump back and forth between prefix and infix notation as needed. In the case of for some , the second argument of is always understood modulo n. A group action is called regular (also known as simply transitive) if the restriction on the second argument is bijective.
Lemma 3.3
(CitationLewin 1987)
Generalized interval systems and regular group actions are in 1-to-1 relation by for and .
Proof.
See Appendix A.1.
A cyclic ordering on a set S can be used to define respective open, semi-open, and closed intervals , , and . The semi-open intervals can be used to define a generalized interval system on a cyclic ordered set. The (musical) interval mapping then returns the cardinality of the respective semi-open (mathematical) interval, . The following results show that the so defined δ is indeed an interval mapping of the generalized interval system .
Lemma 3.4
Let be a cyclic ordered set. Then with is a generalized interval system.
Proof.
See Appendix A.4.
Theorem 3.5
Finite cyclic ordered sets generalized interval systems and regular group actions are in one-to-one relation by for and where is the cardinality of S and denotes the additive cyclic group of cardinality n.
Proof.
See Appendix A.5.
Note that in particular, and for all and . To illustrate the theorem, we characterize the canonical cyclic order on . Let n be a positive natural number. The additive group acts on itself via the action for . The respective interval mapping is thus defined by for . Following theorem 3.5, the corresponding cyclic order is then characterized by for .
The fundamental correspondences between finite cyclic ordered sets, finite cyclic generalized interval systems and finite regular cyclic group actions described in Theorem 3.5 are used throughout the whole paper. All cyclic ordered sets are thus assumed to be finite. The following corollaries show the interactions between these structures.
Corollary 3.6
The interval mapping is invariant under musical transposition. That is for all tones s, t and all intervals i.
Proof.
See Appendix A.6
Lemma 3.7
Let S and T be cyclic ordered sets with . For a mapping the following are equivalent:
f preserves the cyclic order,
f is an injective cyclic morphism, and
f is a cyclic embedding.
Proof.
(i)⇒(ii) Let f preserve the cyclic order, and . Then and Therefore, f is an injective cyclic morphism.
(ii)⇒(iii) Let f be an injective cyclic morphisms. Then f is a cyclic embedding since for all .
(iii)⇒(i) All cyclic embeddings preserve the cyclic order by definition.
Corollary 3.8
A mapping with equally sized domain and codomain is a cyclic isomorphism if and only if the interval mapping is invariant under it.
Proof.
Let S and T be finite cyclic ordered sets and let be a cyclic isomorphism. Then for all .
Conversely, let be a mapping such that for all . Then for all . Therefore, f is a cyclic embedding by Lemma 3.7. It is an isomorphism, because its domain and codomain are of equal cardinality.
Corollary 3.9
Musical transposition is a cyclic automorphism for all .
Proof.
The interval mapping is invariant under musical transposition by Corollary 3.6. Musical transposition is thus a cyclic automorphism by Corollary 3.8.
Corollary 3.10
The mapping is a cyclic isomorphism between the cyclic ordered set S and the cyclic ordered interval group for all .
Proof.
For all , Therefore, is a cyclic isomorphism by Corollary 3.8.
Corollary 3.11
The automorphism group of a finite cyclic ordered set S with is isomorphic to .
Proof.
Fix some . We show that is a group isomorphism. It is a group homomorphism since for all and . It is injective since for all . For showing the surjectivity of Φ, let be an automorphism of S, fix any , and set . We show that . For all , Therefore, for all .
4. Scale morphisms
Definition 4.1
A scale morphism from a scale to a scale is a pair of cyclic morphisms and such that the following diagram commutes.
That is . A scale morphism is called a scale isomorphism if both φ and ψ are cyclic isomorphisms. A scale is called transpositionally symmetric if its automorphism group is not trivial.
The concept of a transpositionally symmetric scale corresponds nicely to the musical notion. Consider for example the diatonic, hexatonic, and octatonic scales defined in Figure . The cyclic automorphism group of the C major scale is trivial whereas the cyclic automorphism groups of hexatonic and octatonic scales are isomorphic to and , respectively.
Since the automorphism group of a cyclic ordered set is isomorphic to the additive cyclic group of its cardinality (see Corollary 3.11), two scales that share their diatonic and chromatic tones are isomorphic if and only if there exist and such that for all . We denote the fact that σ and are isomorphic by . In sum, we have
Theorem 4.2
For every scale there exists an isomorphic scale .
Proof.
Fix any . By Corollary 3.10, the mapping and the mapping are cyclic isomorphisms. Thus, is a scale from to such that the following diagram commutes.
Therefore, is a scale isomorphism.
The next theorem shows how our scale definition sharpens the formalization of scales used by theorists such as CitationClough and Douthett (1991) in their article “Maximally even sets.”. In the following quoted definition, the set corresponds to a scale .
Definition 4.3
To indicate a subset of d pcs selected from the chromatic universe of c pcs we write . Thus is a subset of . We generally assume . We say that d is the diatonic cardinality of . We write to represent the set of all sets.
Theorem 4.4
For each subset of a cyclic ordered set T, there exists exactly one cyclic ordering of S such that the identity is a scale.
Proof.
The cyclic ordering of S is uniquely determined by the cyclic ordering of T, since must hold for all if is a scale.
5. Maximal evenness, Myhill's property, well-formedness, and the fundamental theorem of diatonic scale theory
This section presents the established scale properties maximal evenness, Myhill's property, and well-formedness, and shows their fundamental connection. Maximal evenness and Myhill's property presuppose the concept of the spectrum of a scale that maps the diatonic scale intervals to the set of their possible chromatic realizations. In a major scale for example, (any third is either a minor third spanning three semitones or a major third spanning four semitones) and (any fifth is either a diminished fifth spanning six semitones or a perfect fifth spanning seven semitones). The notion of a spectrum of an interval was introduced by CitationClough and Myerson (1985) as follows:
Definition 5.1
The spectrum of a scale is defined by
The following lemma is helpful for arguments that include the spectrum mapping. It shows that in all scales, the sum of all specific intervals of any generic size equals the product of i and the scale's chromatic size.
Lemma 5.2
(CitationClough and Myerson 1985)
For all scales and diatonic intervals
Proof.
For all diatonic intervals , since , is bijective for all by Lemma 3.9.
A maximally even scale can intuitively be understood as a scale whose tones are as evenly distributed as possible in the underlying scale, for example in the chromatic scale. While six tones can be totally evenly distributed in the chromatic scale, seven tones cannot. The result of distributing seven tones in the chromatic scale as evenly as possible gives the diatonic scale. Asymmetric maximally even scales were first studied by CitationClough and Myerson (1985). The general notion of maximally evenness was in introduced by CitationClough and Douthett (1991).
Definition 5.3
(CitationClough and Douthett 1991)
A scale is maximally even if the spectrum of any generic interval consists either of a single integer or two consecutive integers. It is totally even if additionally where denotes the greatest common divisor of c and d.
Definition 5.4
A scale is non-degenerate maximally even if it is maximally even and
For example, the diatonic, hexatonic, and octatonic scale shown in Figure are respectively non-degenerate maximally even, not maximally even, and degenerate maximally even. The whole-tone scale is an example of a totally even scale. More recently, CitationQuinn (2004) and CitationAmiot (2007) presented an equivalent more intuitive definition of maximal evenness using the magnitude of Fourier coefficients. For a third alternative but equivalent definition using interacting weighting functions see CitationDouthett and Krantz (2007).
We choose the term non-degenerate maximally even in analogy to non-degenerate well-formed proposed by CitationCarey and Clampitt (1989), since both describe scales which satisfy . Note that under this definition, degenerate maximally even scales (like the octatonic scale) are “not as degenerate” as degenerate well-formed scales (like the whole-tone scale). Alternatively, degenerate and non-degenerate maximally even scales could be called transpositional symmetric and transpositional asymmetric maximally even scales, respectively. Since non-degenerate maximally even scales are the scales which were studied by CitationClough and Myerson (1985), CitationHarasim, Noll, and Rohrmeier (2019) also call them Clough–Myerson scales.
Non-degenerate maximal evenness is a strong condition and Myhill's property can be seen as a generalization of it as the next lemma shows.
Definition 5.5
(CitationClough and Myerson 1985)
A scale has Myhill's property if the spectrum of any non-zero diatonic interval consists of exactly two distinct numbers.
Lemma 5.6
(CitationClough and Myerson 1985)
All non-degenerate maximally even scales have Myhill's property.
Proof.
The lemma is shown by contradiction. Assume there is a diatonic interval i and a chromatic interval a such that is a singleton set. Then and thus . Since , d therefore must divide i which contradicts 0<i<d.
The concepts of generated scales and well-formed scales are additional means to characterize scales. They are – intuitively speaking – complementary or orthogonal to the spectrum mapping. The C major scale is for example generated by the chromatic interval of seven semitones beginning by F (pitch class 5), It is also well-formed since there is a diatonic interval class (diatonic interval class 4 in this example) that corresponds to the 7 semitones such that the generation process can also be formulated in the diatonic domain of the scale,
Definition 5.7
(CitationCarey and Clampitt 1989)
A scale is generated if there exists a and a specific interval such that . The interval g is then called a generating interval for σ. A scale is well-formed if additionally there is a such that for all . A well-formed scale is called non-degenerate if .
Note that for a well-formed scale σ, q is invertible in since implies .
The next theorem is what we call the fundamental theorem of diatonic scale theory. It connects the properties maximal evenness, well-formedness, and Myhill's property and will further be used for proving their characterizations in the next sections. The theorem was first stated and proven by CitationClough and Myerson (1985), although they did not yet use the terms maximally even and well-formed. For the example of the C major scale, we have c = 12, d = 7, s = 5 (pitch class of F), q = 4, and g = 7 as described above.
Theorem 5.8
(CitationClough and Myerson 1985)
Non-degenerate maximally even scales are non-degenerate well-formed. In particular, all non-degenerate maximally even scales with equal diatonic and chromatic sizes are isomorphic.
Proof.
Let σ be a non-degenerate maximally even scale. We show the existence of and such that for all . For all there is a such that , because σ fulfills Myhill's property (Lemma 5.6). Let . Then by Lemma 5.2. Since for all The quantity λ is thus the number of how often the generic interval q is realized as specific interval g−1. By setting , a g that generates σ and its corresponding diatonic size q can be obtained from the definition of λ. q is given by since and g is given by .
In particular, all non-degenerate scales with diatonic size d and chromatic size c are generated by the same interval g as calculated here and are thus isomorphic.
6. Maximal evenness and the J-function
This section shows the 1-to-1 relation between maximally even scales and the so called J-functions, introduced by CitationClough and Douthett (1991). This correspondence is particularly useful to show that for all diatonic and chromatic cardinalities d and c there is up to isomorphism exactly one maximally even scale. For natural numbers c, d, and m with 0<d<c and , the J-function is defined by where denotes the floor function. The number m is called the mode index of the scale. In this definition, k iterates over the diatonic tones of the scale and assigns k to the “nearest” chromatic tone that it finds, where the mode index m shifts the pattern. For example, the C major scale is the J-function .
The first theorem in this section shows that all J-functions are scales. The rest of the section then shows that a scale is maximally even if and only if it is isomorphic to a J-function, formulating the argumentation by CitationClough and Douthett (1991) in our axiomatic framework.
Lemma 6.1
As a function every J-function preserves the strict order of . In particular, all J-functions are injective.
Proof.
Let such that and . Then there is a positive natural number k such that i + k = j. Therefore, which implies where denotes the fractional part and denotes the integer part of x. Thus, .
Theorem 6.2
Every J-function is a scale.
Proof.
Let and such that 0<d<c, , and . We use the abbreviation and show . Then J is a scale by Lemma 3.7. Without loss of generality, let . Then and by the previous lemma. Therefore, which is .
Lemma 6.3
For each maximally even scale σ and all divisors e of its diatonic and chromatic sizes, for all .
Proof.
We show for by contradiction. Assume there is an such that (the case is analogous). Since by Lemma 5.2, there is then an such that , contradicting the maximal evenness of σ.
Lemma 6.4
Let be a maximally even scale and a divisor of its diatonic and chromatic sizes d and c. Then the restriction is a maximally even scale.
Proof.
Lemma 6.3 implies for all since Denote by . Then modulo for all . Therefore, is maximally even if σ is maximally even.
The previous lemmas imply that every maximally even scale consists of copies of the smaller maximally even scale . For non-degenerate maximally even scales such as the C major scale , this is trivial. For the octatonic scale (induced by ; see Figure , right) which has transpositional symmetries, e = 4 and . This octatonic scale therefore consists of 4 copies of the scale which is induced by .
Theorem 6.5
(CitationClough and Douthett 1991)
A scale is maximally even if and only if it is isomorphic to a J-function.
Proof.
All J-functions are maximally even since modulo c for all with 0<d<c, , and .
We now show that all maximally even scales that share the same diatonic and chromatic sizes are isomorphic. It then follows that every maximally even scale is isomorphic to a J-function since every J-function is maximally even.
Let σ be a maximally even scale and . By Lemma 6.4, is maximally even and thus unique up to isomorphism according to its diatonic and chromatic sizes by Theorem 5.8. Therefore σ is also unique up to isomorphism according to its diatonic and chromatic sizes, since Lemma 6.3 implies that σ consists of e copies of .
To sum up, this section showed how maximally even scales can be understood. On the one hand, there is up to isomorphism one unique maximally even scale for each diatonic and chromatic size and this scale is given in closed form by a J-function. On the other hand, there are two distinguished kinds of maximally even scales: non-degenerate (without proper transpositional symmetries) and degenerate (with proper transpositional symmetries) where each degenerate maximally even scale can be obtained from a non-degenerate maximally even scale by copying it multiple times.
7. The equivalence of Myhill's property and non-degenerate well-formedness
Lemma 7.1
A scale that has Myhill's property fulfills .
Proof.
Assume . Then by Lemma 5.2. Therefore, Myhill's property implies the existence of a positive natural number and a partition of D into two not empty disjunct sets such that for all and for all . Thus, either in case and alternate, or otherwise, contradicting Myhill's property.
Definition 7.2
A scale that has Myhill's property is reduced if .
Lemma 7.3
Reduced scales that have Myhill's property are non-degenerate maximally even.
Proof.
Let be a reduced scale that has Myhill's property. Then for all and . We show that the spectra of all non-zero intervals consist of consecutive integers. Let . Then modulo c. Therefore, consists of consecutive integers.
We prove the equivalence of Myhill's property and non-degenerate well-formedness using the concept of interval words of scales. The connection between interval words and well-formed scales was first described by CitationClampitt, Domìnguez, and Noll (2007) and CitationDomínguez, Clampitt, and Noll (2009). Further studies on this topic include for example CitationCastrillon Lopez and Domìnguez Romero (2016) and a recent special issue of the Journal of Mathematics and Music (CitationBrlek, Chemillier, and Reutenauer 2018).
Definition 7.4
Let be a scale with where . An interval word of σ is a word w such that for all and some where denotes the kth letter of the word w. Two scales and are equivalent under the interval word relation, denoted by , if they have a common interval word.
Note that the letters of interval words are the indices of the intervals in the totally ordered set . Interval words therefore do not contain information about the concrete sizes of the intervals, but only information about the ordering of the interval sizes. Consider for example the pentatonic scale σ induced by . The interval words of this scale are the cyclic permutations of 21121, because . The scale , induced by , has the same class of interval words as σ. Therefore, . Since σ is maximally even but is not, this example shows that maximal evenness is not invariant under the interval word relation.
Lemma 7.5
Myhill's property and well-formedness are invariant under the interval word relation.
Proof.
Let and be equivalent scales and let and be natural numbers such that for all there is an with For all , there then exist non-negative such that In particular, for all which implies that σ has Myhill's property if and only if has.
Furthermore if σ is well-formed, then there exist , , and such that for all Therefore, for all and is thus well-formed.
Theorem 7.6
(CitationCarey and Clampitt 1996)
A scale has Myhill's Property if and only if it is non-degenerate well-formed.
Proof.
() We first show that non-degenerate well-formed scales have Myhill's property. For this, let be a well-formed scale with . Then there exist , , and such that for all . Thus, for all and modulo c. In the following, we calculate for arbitrary and to show . Let . Since , there exist unique such that and modulo d. Therefore, If , then modulo c. If otherwise , then modulo c and modulo c, because .
() We now show that a scale which has Myhill's property is non-degenerate well-formed. Let be a scale with that has Myhill's property. We construct a reduced scale ϱ with by for . Since and ϱ is reduced, by Lemma 7.5, Lemma 7.3, Theorem 5.8, and again Lemma 7.5, respectively. follows additionally from σ having Myhill's property by Lemma 7.1.
The first part of the proof of Theorem 7.6 is elementary and in a similar way already presented by CitationCarey and Clampitt (1989). The second part was first proven by CitationCarey and Clampitt (1996), also in an elementary way. In contrast, our new proof of this established theorem connects it closer 1) to the theory of maximally even sets by factoring out Theorem 5.8 and 2) to combinatory word theory by utilizing the fact that well-formedness and Myhill's property only depend on the interval word of a scale.
To sum up, this section characterized the different kinds of well-formed scales. The first kind, degenerate well-formed scales, are scales that are generated by a generator that is not coprime to the chromatic size of the scale. All these scales are subscales of a totally even scale. The second kind are maximally even scales that do not have transpositional symmetries. The third and last kind are scales that are pumped-up versions of scales of the second kind. They also have Myhill's property, but are not maximally even.
8. Summary and conclusion
This paper proposes an axiomatic algebraic framework to model musical scales. It utilizes cyclic order relations to define scales as embeddings of cyclic ordered sets and to express their properties and relations. This is the first paper that brings together the core results and arguments of diatonic scale theory into a uniform system to study symmetry properties of scales. We did not state new propositions but reformulated fundamental theorems and proofs in a modern mathematical language. This improves existing frameworks by providing a firmer foundation for future research. We particularly hope that it also will help to establish new connections to related fields such as music cognition and computational musicology.
Additionally, we provided a new more sophisticated proof for the equivalence of Myhill's property and non-degenerate well-formedness. In particular, we showed that the fundamental theorem of diatonic scale theory (Theorem 5.8) can be used to prove both, the characterization of maximally even scales through the J-function and the characterization of non-degenerate well-formedness by Myhill's property.
There are two main directions on which future work can proceed from this paper. The first one is to incorporate other scale properties into the unified framework and to study mathematically the commonalities and differences of the symmetry properties of scales and rhythms from different cultures. The second direction is to use the presented scale-theoretic basis to study voice-leading induced geometries as proposed by CitationHarasim, Schmidt, and Rohrmeier (2016).
Acknowledgments
The authors would like to thank Thomas Fiore, Fabian C. Moss, Christoph Finkensiep, the editors Jason Yust and Emmanuel Amiot, and the anonymous reviewer for their constructive and helpful comments on previous versions of this paper.
Disclosure statement
No potential conflict of interest was reported by the authors.
Additional information
Funding
Notes
1 with the notable exceptions of scales used in Jazz and extended tonality of western music of the nineteenth century.
References
- Amiot, Emmanuel. 2007. “David Lewin and Maximally Even Sets.” Journal of Mathematics and Music 1 (3): 157–172. https://doi.org/10.1080/17459730701654990.
- Balzano, G. J. 1980. “The Group-Theoretic Description of 12-Fold and Microtonal Pitch Systems.” Computer Music Journal 4 (4): 66–84. doi: 10.2307/3679467
- Brlek, Srečko, Marc Chemillier, and Christophe Reutenauer. 2018. “Music and Combinatorics on Words: A Historical Survey.” Journal of Mathematics and Music 12 (3): 125–133. doi: 10.1080/17459737.2018.1542055
- Burns, Edward M., and W. Dixon Ward. 1978. “Categorical Perception-phenomenon or Epiphenomenon: Evidence from Experiments in the Perception of Melodic Musical Intervals.” The Journal of the Acoustical Society of America 63 (2): 456–468. http://asa.scitation.org/doi/10.1121/1.381737.
- Carey, Norman. 1998. “Distribution Modulo 1 and Musical Scales.” Ph.D. thesis. http://www.researchgate.net/profile/Norman_Carey/publication/235712212_Distribution_modulo_1_and_musical_scales/links/004635153159b6e710000000.pdf.
- Carey, Norman, and David Clampitt. 1989. “Aspects of Well-Formed Scales.” Music Theory Spectrum 11 (2): 187–206. doi: 10.2307/745935
- Carey, N., and D. Clampitt. 1996. “Self-Similar Pitch Structures, Their Duals, and Rhythmic Analogues.” Perspectives of New Music 34 (2): 62–87.
- Castrillon Lopez, Marco, and Manuel Domìnguez Romero. 2016. “LR Property of Non-Well-Formed Scales.” Journal of Mathematics and Music 10 (1): 18–35. doi: 10.1080/17459737.2016.1164907
- Clampitt, David. 2009. “Mathematical and Musical Properties of Pairwise Well-Formed Scales.” Communications in Computer and Information Science CCIS 37: 464–468.
- Clampitt, David, Manuel Domìnguez, and Thomas Noll. 2007. “Well-Formed Scales, Maximally Even Sets and Christoffel Words.” Proceedings of the MCM.
- Clough, John, and Jack Douthett. 1991. “Maximally Even Sets.” Journal of Music Theory 35 (1/2): 93–173. doi: 10.2307/843811
- Clough, John, Jack Douthett, and Lewis Rowell N. Ramanathan. 1993. “Early Indian Heptatonic Scales and Recent Diatonic Theory.” Music Theory Spectrum 15 (1): 36–58. doi: 10.2307/745908
- Clough, John, Nora Engebretsen, and Jonathan Kochavi. 1999. “Scales, Sets and Interval Cycles: A Taxonomy.” Music Theory Spectrum 21 (1): 74–104. https://doi.org/10.2307/745921
- Clough, John, and Gerald Myerson. 1985. “Variety and Multiplicity in Diatonic Systems.” Journal of Music Theory 29 (2): 249–270. doi: 10.2307/843615
- Clough, John, and Gerald Myerson. 1986. “Musical Scales and the Generalized Circle of Fifths.” The American Mathematical Monthly 93 (9): 695–701. doi: 10.1080/00029890.1986.11971924
- Cohn, Richard. 1996. “Maximally Smooth Cycles, Hexatonic Systems, and the Analysis of Late-Romantic Triadic Progressions.” Music Analysis 15 (1): 9–40. doi: 10.2307/854168
- Cohn, Richard. 1997. “Neo-Riemannian Operations, Parsimonious Trichords, and Their Tonnetz Representations.” Journal of Music Theory 41 (1): 1–66. doi: 10.2307/843761
- Conklin, Darrell, and Ian H. Witten. 1995. “Multiple Viewpoint Systems for Music Prediction.” Journal of New Music Research 24 (1): 51–73. https://doi.org/10.1080/09298219508570672.
- Domínguez, Manuel, David Clampitt, and Thomas Noll. 2009. “WF Scales, ME Sets, and Christoffel Words.” Communications in Computer and Information Science 37 CCIS: 477–488.
- Douthett, Jack. 2008. “Filtered Point-Symmetry and Dynamical Voice-Leading.” In: Music Theory and Mathematics: Chords, Collections, and Transformations, edited by Jack Douthett, Martha M. Hyde, and Charles J. Smith, 72–106.Rochester: University Rochester Press.
- Douthett, Jack, and Richard Krantz. 2007. “Maximally Even Sets and Configurations: Common Threads in Mathematics, Physics, and Music.” Journal of Combinatorial Optimization 14 (4): 385–410. doi: 10.1007/s10878-006-9041-5
- Galil, Zvi, and Nimrod Megiddo. 1977. “Cyclic Ordering Is NP-Complete.” Theoretical Computer Science 5 (2): 179–182. doi: 10.1016/0304-3975(77)90005-6
- Harasim, Daniel, Thomas Noll, and Martin Rohrmeier. 2019. “Distant Neighbors and Interscalar Contiguities.” In Mathematics and Computation in Music, Proceedings of the 7th International Conference (MCM 2019), Madrid, Spain, June 18-21, 2019 Vol. 11502 of Lecture Notes in Computer Science, edited by Mariana Montiel, Francisco Gomez-Martin, and Octavio A. Agustín-Aquino, 172–184. Cham: Springer International.
- Harasim, Daniel, Martin Rohrmeier, and Timothy J. O'Donnell. 2018. “A Generalized Parsing Framework for Generative Models of Harmonic Syntax.” 19th International Society for Music Information Retrieval Conference.
- Harasim, Daniel, Stefan E. Schmidt, and Martin Rohrmeier. 2016. “Bridging Scale Theory and Geometrical Approaches to Harmony: The Voice-leading Duality between Complementary Chords.” Journal of Mathematics and Music 10 (3): 193–209. https://doi.org/10.1080/17459737.2016.1216186.
- Huntington, E. V. 1916. “A Set of Independent Postulates for Cyclic Order.” Proceedings of the National Academy of Sciences 2 (11): 630–631. http://www.pnas.org/cgi/doi/10.1073/pnas.2.11.630.
- Levine, Mark. 1995. The Jazz Theory Book. Petaluma, CA: Sher Music.
- Lewin, David. 1987. Generalized Musical lntervals and Transformations. New Haven: Yale University Press.
- Megiddo, Nimrod. 1976. “Partial and Complete Cyclic Orders.” Bulletin of the American Mathematical Society 82 (2): 274–277. doi: 10.1090/S0002-9904-1976-14020-7
- Miller, G. A. 1956. “The Magical Number Seven, Plus or Minus Two: Some Limits on Processing Information.” The Psychological Review 63 (2): 81–97. doi: 10.1037/h0043158
- Moss, Fabian C., Markus Neuwirth, Daniel Harasim, and Martin Rohrmeier. 2019. “Statistical Characteristics of Tonal Harmony: A Corpus Study of Beethoven's String Quartets.” PLOS ONE 14 (6): e0217242. http://dx.plos.org/10.1371/journal.pone.0217242.
- Neumaier, Wilfried, and Rudolf Wille. 1990. “Extensionale Standardsprache der Musiktheorie: eine Schnittstelle zwischen Musik und Informatik”.
- Noll, Thomas. 2007. “Musical Intervals and Special Linear Transformations.” Journal of Mathematics and Music 1 (2): 121–137. https://doi.org/10.1080/17459730701375026.
- Noll, Thomas. 2015. “Triads as Modes within Scales as Modes.” In: Collins T., Meredith D., Volk A. (eds) Mathematics and Computation in Music. 5th International Conference (MCM 2015), London, 22-25 June 2015, vol. 9110 of Lecture Notes in Computer Science, edited by Tom Collins, David Meredith, and Anya Volk, 373-384. Cham: Springer.
- Patel, Aniruddh D. 2008. Music, Language, and the Brain. New York: Oxford University Press.
- Pressing, Jeff. 1983. “Cognitive Isomorphisms Between Pitch and Rhythm in World Musics: West Africa, the Balkans and Western Tonality.” Studies in Music 17, 38–61.
- Quinn, Ian. 2004. “A Unified Theory of Chord Quality in Equal Temperaments.” Ph.D. Thesis, Eastman School of Music.
- Rohrmeier, Martin. 2011. “Towards a Generative Syntax of Tonal Harmony.” Journal of Mathematics and Music 5 (1): 35–53. https://doi.org/10.1080/17459737.2011.573676.
- Savage, Patrick E., Steven Brown, Emi Sakai, and Thomas E. Currie. 2015. “Statistical Universals Reveal the Structures and Functions of Human Music.” Proceedings of the National Academy of Sciences 112 (29): 8987–8992. doi: 10.1073/pnas.1414495112
- Sears, David R. W., Marcus T. Pearce, William E. Caplin, and Stephen McAdams. 2018. “Simulating Melodic and Harmonic Expectations for Tonal Cadences Using Probabilistic Models.” Journal of New Music Research 47 (1): 29–52. https://doi.org/10.1080/09298215.2017.1367010.
- Trehub, Sandra E., E. Glenn Schellenberg, and Stuart B. Kamenetsky. 1999. “Infants' and Adult' Perception of Scale Structure.” Journal of Experimental Psychology 25 (4): 965–975.
- Tymoczko, Dmitri. 2006. “The Geometry of Musical Chords.” Science 313 (5783): 72–74. doi: 10.1126/science.1126287
- Tymoczko, Dmitri. 2008. “Scale Theory, Serial Theory and Voice Leading.” Music Analysis 27 (1): 1–49. doi: 10.1111/j.1468-2249.2008.00257.x
- Wooldridge, Marc Charles. 1992. “Rhythmic Implications of Diatonic Theory: A Study of Scott Joplin's Ragtime Piano Works”. Ph. D. diss., State University of New York at Buffalo.
Appendix 1.
Proofs of Section 3
Proposition A.1
(CitationLewin 1987)
Generalized interval systems and regular group actions are in 1-to-1 relation by for and .
Proof.
Let fist be a generalized interval system. Then , defined by for and is a regular group action. It is a group action since 1) implies and 2) for with and . The regularity of follows from the bijectivity of δ in the second argument. Let in converse be a regular group action. Then δ, defined by for and is the interval mapping of the generalized interval system since for follows from with the injectivity of . The bijectivity of δ in the second argument follows from the regularity of . It is moreover obvious that the presented constructions which translate between generalized interval systems and regular group actions are inverse to each other.
Lemma A.2
Let be a cyclic ordered set. Then for all with where denotes the complement .
Proof.
For all with , Therefore, and thus also .
Lemma A.3
Let be a cyclic ordered set. Then for all distinct the following are equivalent.
Proof.
We show the equivalences in two cycles, proving , and , and , and . In any step, we prove only one of the three equalities. The other two ones follow always analogously.
. We show and . Let . Then either or v = t or or v = u. Therefore, or v = u, since . Thus, . Let in converse be . If v = u then . Otherwise either , implying , or . The latter and imply and thus . In any case, .
. We show . Since , by Lemma A.2.
. Since , by the argumentation of .
Let . Then or v = t, and or v = u. Therefore, Hence, which contradicts . There thus does not exist a , implying .
. We show . Since , by Lemma A.2.
. Since , by the argumentation of .
Proposition A.4
Let be a cyclic ordered set. Then with is a generalized interval system.
Proof.
Denote the cardinality of S by n. We first show mod n for any pairwise distinct considering two cases. If s, t, and u are not pairwise distinct, one of , , and is zero and the equality holds trivially.
Case 1: . Then by Lemma A.3. Therefore, Case 2: . Then by Lemma A.3. Therefore, mod n by Lemma A.3.
We show now by induction over that is bijective for any . The base case follows from . For the inductive step, let be the unique element of S with . A with can be found using the following procedure. At first, choose randomly some that is distinct from u. If holds for all , then mod n. If otherwise there is a with , then set t: = v and iterate. The algorithm terminates since S is finite. For showing the uniqueness of the found t, assume there are distinct with . Then and . Since either or , either or , implying which contradicts the assumption.
Theorem A.5
Finite cyclic ordered sets finite generalized interval systems and regular group actions are in one-to-one relation by for and where is the cardinality of S and denotes the additive cyclic group of cardinality n.
Proof.
The one-to-one relation between generalized interval systems and regular group actions is shown in Theorem 3.3. Theorem 3.4 shows further that with is a generalized interval system. We first here show that for all generalized interval systems , the relation is a cyclic ordering on S.
Since < is a strict total order on , the binary relation defined by is a strict total order on S. Hence, is a cyclic order. For the cyclicality, let with and . Then, and therefore where and denote the addition and subtraction modulo n, and + and − denote the usual addition and subtraction of integers. Thus, which is by definition, proving the cyclicality.
We now show that the presented constructions that translate cyclic ordered sets into generalized interval systems and vice versa are inverse. Let be a cyclic ordered set and . We show that . (⇒) Let be pairwise distinct elements with . For all with , by transitivity. Therefore, implying which is by definition.
(⇐) Let be pairwise distinct elements with . Since is anitsymmetric, either or must hold. Assume that . Then, we have analogously to the if part, contradicting and the cyclicality of . Therefore, .
Let now be a generalized interval system. We show that for . Since for any , Together with the bijectivity of , this implies .
Corollary A.6
The interval mapping is invariant under musical transposition.
Proof.
Let S be cyclic ordered and , , and . Then Therefore, .