A Graph Representation of

Tonal Modulation

By
Kayla Mari Wankowski
Advisor: Dr. Jaromy Kuhl

An Undergraduate Proseminar
In Partial Fulfillment of the Degree of
Bachelor of Science in Mathematical Sciences
The University of West Florida
July 2014

Approval Page
The Proseminar of Kayla Mari Wankowski is approved:

Jaromy Kuhl, Ph. D., Proseminar Advisor

Date

Josaphat Uvah, Ph. D., Proseminar Committee Chair

Date

Accepted for the Department:

Jaromy Kuhl, Ph. D., Chair

Date

Contents
Approval Page

Abstract

Introduction

2.1

Literature Review . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

2.2

Musical Definitions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Main Body

3.1

Graphs . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.2

Connectivity . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.3

Chromatic Number . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.4

Adjacency Matrix . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.5

Tutte Polynomial . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

10

Conclusion

11

ii

Abstract

Tonal Modulation, the musical process of changing the key signature, can be represented by a simple graph on the
musical keys. This graph is called the tonal modulation graph. We discuss graph theory invariants such as connectivity
and the chromatic number to investigate modulation. We also make some observations concerning the adjacency
matrix and show that certain features are unique under the Tutte polynomial. The Tonal Modulation Graph can be
used by a composer wishing to explore the different possibilities of modulating while writing music. In Classical
music, to change the key of a work, the composer must write a transition period to smooth the change. The graph
discussed in this maps out all of the possible paths from one key to another.

2
2.1

Introduction
Literature Review

For this research, we used an article by Adrian Walton titled A Graph Theoretic Approach to Tonal Modulation from
The Journal of Mathematics and Music, Douglas Wests Introduction to Graph Theory and The Tutte Polynomial and
the Generalized Petersen Graph an article by Jaromy Kuhl from the Australian Journal of Combinatorics. Waltons
A Graph Theoretic Approach to Tonal Modulation was the main resource for the application of graph theory to
tonal modulation. Walton creates the graphs used to map tonal modulation and relates the properties of the graphs
created to the properties of tonal modulation in music. While Walton discusses several interesting qualities of the
tonal modulation graph, there are still several useful properties of the graph that could be useful to a composer. To
address these topics we also used Introduction to Graph Theory and The Tutte Polynomial and the Generalized
Petersen graph These two other resources allow us to discuss other topics in graph theory such as the adjacency
matrix and Tutte polynomial.
Other literature for further study are Musical Actions of Dihedral Groups in the American Mathematical Monthly
by Alissa S. Crans, Thomas M. Fiore, and Ramon Satyendra, The Topos of Music by Guerino Mazzola and, Modulation
by Max Reger. Musical Actions of Dihedral Groups discusses how music may be interpreted in terms of the group
structure of the dihedral group of order 24 and its centralizer. Musical group structure is discussed by Walton in his
article as well. Mazzola has been working on mathematical music theory since the 1980s. His text The Topos of Music
contains many intriguing applications of mathematics in music theory. Modulation further discusses techniques of
tonal modulation.

2.2

Musical Definitions

To understand the topics discussed in this paper, one must first understand some basic music theory. Figure 1 demonstrates how to read music in treble clef. The treble clef is the most common musical key, and the only one we will
need to focus on.

Figure 1: Reading music in treble clef

The key signature dictates which scale the music follows, or the notes that will be used most often throughout
the work. In Classical music, when a composer wishes to change from one key to another, they must first write a
transitional period in which the chords move smoothly from one key to the next by using triads that are common to
the previous key and the new key. A chord occurs whenever two or more notes, or pitches, are played simultaneously.
A triad is a chord with three notes, which are stacked on top of each other in thirds, meaning the distance between
one note to the one above it is either a major or miner third. A third is referring to the distance between the bottom
note and the note on top of it. For example, the C major triad, the notes from the bottom up are C, E, and G. To get
from C to E we can count starting on C; C, D, E. That was three notes, so it is a valid bottom interval of a triad. Now
from E we have, E, F, G, again this is a three note interval, so we have formed the C manor triad. A musical scale has
a triad for each note. A triad can be major, minor, diminished, or augmented. For the understanding of this paper, it
is not necessary to know how to distinguish between different the types of triads, only that the difference exist. An
accidental is a sharp, ], flat, [, or natural, \ that occurs before a note to change the quality of the note. These are
used when a composer wishes to create a harmony not found in the key signature. Accidentals are used frequently in
modulations since the composer must use chords from different keys.
We can now define tonal modulation. Tonal modulation is the process in music used to change the key signature
of a work of music. A simple modulation is possible if the starting key and the destination key share a triad as shown
in Figure 2. The pivot chord is the chord in a modulation that can be interpreted as in both the initial key and the
destination key. Figure 3 demonstrates a simple modulation. The pivot chord is circled, and is marked with two roman
numerals, one for the starting key, and one for the destination key. Chords are identified by a roman numeral that

corresponds to the the position of the root of the chord in the scale of that key signature, as shown in Figure 2. Major
chords are given a capital numeral and minor are given a lowercase numeral.

Figure 2: How scales relate to allow for a simple modulation

Figure 3: A simple modulation using a pivot cord

3
3.1

Main Body
Graphs

In this paper we will construct a graph that demonstrates the relationship between musical keys with respect to tonal
modulation. A graph, in this case, is a set of vertices connected by lines, called edges. In the graph we use, the vertices
represent the 12 major keys. It is also important to note that a single vertex represents the major key and its relative
minor, since the relative minor is just the Aeolian mode, which can also be thought of as starting the scale on the
6th note. If there is an edge between two vertices, then there is a common triad between the two keys. In terms of
modulation, this means that a simple modulation is possible. Figure 4 below uses this relationship between keys.
3

Figure 4: The Tonal Modulation Graph

A modulation between two keys that can be constructed using a simple modulation should sound less complex than
a modulation between keys that do not have a chord in common. One can think of a simple modulation only using one
chord that would sound slightly out of place but then resolved immediately. A modulation between two keys without
a common chord will have a longer transition period that would sound more disconnected and out of place. Figure
5 demonstrates both a simple modulation and one that is more complex. The double bar in the middle of the staff
separates the two modulations. In the first modulation you notice that there is only one accidental. Since modulation
from C to G is accomplished with a single pivot chord, this is a simple modulation. The two key signatures also have
more notes in common, which negates the need for many accidentals. The modulation on the right is from C to G[.
As we see in the graph C and G[ do not share an edge, so a modulation from C to G[ will not be simple. Notice the
amount of accidentals required to modulate from C to G[. Since these keys are not neighbors in the graph they have
very few notes in common. All of the added accidentals is also what makes this modulation sound more complex then
the modulation form C to G.
Examining the graph in Figure 4, we see that the graph is planar, meaning that it can be drawn in two-dimensions
without any of the edges crossing. Also, the graph is Hamiltonian, which means that the graph contains a is a cycle
that goes through each of its vertices. Musically, this concept can easily be examined, since tracing the circle of fifths

Figure 5: Comparison of modulation techniques

through the graph will result in a Hamiltonian cycle. Figure 6 is the circle of fifths. Following the circle of fifths clock
wise or counter-clock wise will create a Hamiltonian cycle. Notice in the image how the key signatures are related
from one to the next by adding a sharp or flat, and that the neighbors of a vertex in the graph are the two keys next
to it in either direction. This also means that there is a series of simple modulations starting in one key and using the
remaining 11 keys to end on the starting key, as mentioned by Adrian Walton.

Figure 6: The Circle of Fifths

3.2

Connectivity

Another useful result of the graph comes from the application of Mengers Theorem. This theorem deals with the
connectivity of the graph. A graph is said to be k-connected if the minim number of vertices that must be deleted to
disconnect the graph is k.

Theorem 1 (Mengers Theorem) If x and y are vertices of a graph G and xy

/ E(G), then the minimum size of an
xy-cut equals the maximum number of pairwise internally disjoint paths.
Walton applies this theorem to the graph to create the following proposition.
Proposition 1 (Waltons Proposition 2.2) Between any two relative major/minor pairs there exist four sequences of
simple modulations such that no two sequences share any keys.
P ROOF :[The Tonal Modulation Graph is 4-connected] The graph contains two cycles, if we were to attempt to
disconnect the graph by separating the two cycles, we would delete a cycle. With out loss of generality, we will cut
the outer cycle in two pieces. To do this we must disconnect two vertices from it. Since each vertex in the outer cycle
has two neighbors in the inner cycle, to disconnect a piece of the outer cycle from the graph requires that two more
vertices be removed from the inner cycle. Thus the total number of vertices removed to disconnect the graph is 4, so
2

by Mengers theorem the result follows.

The full proof can be found in [1].

From the key signature of C to G[, the four independent paths can be examined much more easily than from C to
G. But it is still possible to find four independent paths from C to G. We have C,G; C, D, G; C,F,G; C, B[, A[, D[, B,
A, G. It is important to note that these paths are not unique; one can find another set of 4 independent paths between C
and G. This information, as Walton states, is useful to a composer wishing to explore all possible options to modulate
between two keys without repeating a transition chord.

3.3

Chromatic Number

While most of the information discussed is referring to simple modulations, the majority of the time a composer cannot
use a simple modulation, since the departure key and the destination key may not be neighbors. In romantic music it
is common to modulate directly to a key a major third apart. This modulation can not be a simple modulation because
keys a major third apart share no common triads. In Waltons paper, he proposes this as the motivation to partition the
vertices in the graph so that no two vertices in a partition share an edge. In graph theoretic terms, this can be solved
by coloring the vertices to obtain a proper coloring. The chromatic number of a graph is the least number of colors
needed to produce a proper coloring. A proper coloring is a coloring in which adjacent vertices do not share a color.
By the Four-Color theorem stated in [2], the chromatic number is at most four, since the Tonal Modulation Graph is
planar.
Proposition 2 The chromatic number of the Tonal Modulation Graph is three.

P ROOF : The Tonal Modulation Graph contains a triangle so the chromatic number must be greater than two. We
begin by coloring one of the keys in the outer circle with color 1. Now its four neighbors can not be colored with
color 1. Notice that these vertices create three triangles all with the vertices colored with 1. We color the 4 vertices
by alternating between color 2, and color 3. Next we will look at one of the neighbors of the vertex on the outer ring.
With out loss of generality, assume this vertex is colored with color 2. It is already in a triangle that is fully colored.
Thus its neighbors are colored alternating between 1 and 3. Repeat this on the remaining outer vertices. At the second
to last vertex, the graph has a complete 3-coloring. Thus the tonal modulation graph is 3-colorable. Therefore it has a
2

chromatic number of 3.

Figure 7: A proper coloring of the Tonal modulation graph

Figure 7 is a 3-coloring of the Tonal Modulation Graph using shapes in place of colors. The chromatic number
would be useful to a composer who wishes to modulate between all the keys in such a way that each section would
sound distant and significantly different from the last. The result tells the composer how many sections this work
would need. This result can also be used by a composer writing a smaller piece of music who wants to modulate to a
key that would sound distinctly different from the starting key.

3.4

Adjacency Matrix

The Adjacency Matrix of a graph, A(G), is an n n matrix, where n is the number of vertices in the graph, in which
entry aij is the number of edges in G with endpoints at vertex i and vertex j. The graph in Figure 8 has been labeled to
correlate with the adjacency matrices discussed in this section. The adjacency matrix for the Tonal modulation graph is
shown below. The following theorem is well known [4] and shows the information contained in the adjacency matrix
of a graph G.

Figure 8:

A(G) =

(k)

(k)

Theorem 2 If A is an n n adjacency matrix of a graph and aij represents the (i, j) entry of Ak , then aij is equal
to the number of walks of length k from Vi to Vj .
This means that by raising the adjacency matrix to a power k, we can find how many walks of length k to get from
key i to key j. A composer wishing to modulate back and forth between two of the same keys can use this information
to know how many times they can make different modulations of a desired length. Shown below are the squared and
cubed adjacency matrices for the Tonal Modulation Graph. A composer may use the squared adjacency matrix to
determine the number of possible ways to modulate from C to F going through two key signatures. The composer
would find the entry in column C row F, then observe that there are two options. Notice that going from F to C will
yield the same result.
8

The adjacency matrix can be raised to any power to find walks of modulation between keys of that length. Notice,
at the 3rd power the matrix does not have any zeros. This means that a composer can modulate between any two keys
using three key signatures. For example, assume a composer wants to modulate from the key of C to the key of B in
the least amount of chords required. Either with a quick inspection of the graph, or the adjacency matrix, they can see
that this would not be a simple modulation since the entry in column 1, row 10 is 0. There is also a 0 in A(G)2 but
in A(G)3 there are 3 possible walks of length 3. So the composer must go through at least 3 key signatures before
coming to the final key.

A(G)2 =

A(G)3 =

3.5

Tutte Polynomial

The Tutte Polynomial, introduced by Tutte in 1954, contains much of a graphs information, as listed in Theorem 3
from [3]. Next we will show that the Tonal Modulation Graph is T -unique. This means that any graph with the same
Tutte Polynomial is isomorphic to the Tonal Modulation Graph, or any graph with the same information from Theorem
4 must be the Tonal Modulation Graph.
Theorem 3 Let G be a 2-connected graph. Then the following parameters of G are determined by its Tutte polynomial:
1. The number of vertices and the number of edges.
2. For every k, the number of edges with multiplicity k.
3. The number of cycles of shortest length.
4. The edge connectivity of G. In particular, a lower bound for the minimum degree of G.
5. If G is simple, the number of cliques of each size.
6. If G is simple, the number of cycles of length three, four, and five. For cycles of length four, it is also possible to
know how many of them have exactly one chord.
We use the following theorem to show that any graph with the same Tutte polynomial as the Tonal Modulation
graph must be 2-connected, since the Tonal Modulation graph is 2-connected.
Theorem 4 If G is a 2-connected graph and H is T -equivalent to G, then H is a 2-connected graph.
Theorem 5 The Tonal Modulation Graph is T -Unique.
P ROOF : Let H be a graph T -equivalent to the Tonal Modulation Graph. From Theorem 3, H has 12 vertices and
24 edges. Furthermore, H has 12 triangles and 12 4-cycles with a chord, which we denote as C4+ . It also follows that
each 4-cycle in H contains a chord.
Claim 1: H is 4-regular.
From Theorem 3, (H) 4. If (H) > 4, then

vV

d(v) > 48, a contradiction. Thus H is 4-regular.

Claim 2: Each triangle in H is a triangle shared by two C4+ s.

By Theorom 3, H contains 12 5-cycles. In one 5-cycle there must be 2 chords. If we have vertices 1, 2, 3, 4, and
5 in a cycle in that order, the two chords can be from any pair of two already non-adjacent vertices. With out loss
10

of generality, we place the first chord between vertices 2 and 5. Now we have a triangle and a 4- cycle. Since each
4-cycle must have a chord, there cannot be a chord from 1 to 4 or 2 to 3 . This leaves a chord between either 3 and 5
or 2 and 4. With out loss of generality, we choose the chord to be between 3 and 5. This creates the four cycle 1-2-3-5
with a chord between 2 and 5, and the four cycle 2-3-4-5 with a chord between 3 and 5. These two C4+ s share the
triangle 5-2-3. Since there are 12 5-cycles with two chords, each of the 12 triangles are shared by two C4+ s.
Let T1 , . . . , T12 denote the triangles of H and let C1 , . . . , C12 denote the C4+ s of H. To construct H we begin
with the C1 = T1 T2 . Using claim 2, we add a vertex to form T3 and thus construct C2 = T2 T3 . We continue
adding vertices to form Ti for 4 i 10, and thus construct C3 , . . . , Ci1 .
We now have a graph with 12 vertices. In T10 one vertex has degree 4, one has degree 3, and another has degree
2. Since T1 has only been used in one C4+ , the edges in T1 will have the same degree distribution as T10 . By Claim 1,
the two vertices in T1 with degree at most 3 must be adjacent to the vertex in T10 with degree 2, creating T11 . Now we
have a vertex in T1 with degree 3 and a vertex in T10 with degree 3. Since these are the only vertices in H with degree
at most 3, they must necessarily be adjacent. Connecting these two vertices with an edge creates T12 . Since Claims 1
and 2 hold, H is isomorphic to the Tonal Modulation Graph making the Tonal Modulation Graph T -unique.

Conclusion

Even without knowledge of graph theory, the Tonal Modulation Graph can still be useful to a composer. A composer
can use the graph to find the chord sequences needed to modulate to the destination key, by simply tracing a path from
the initial key to the destination key. With the added graph theoretic information, a composer is now able to modulate
between keys using the options discussed in this paper. In [1], Walton goes on to create a graph for tonal modulation
that separates the major keys and their relative minor keys into separate vertices. This graph has twice as many vertices
and each vertex has twice as many neighbors. One can image that this graph is quite complex. It is also possible to
construct a graph to include the melodic and harmonic minor scales.

11

References
[1] Adrian Walton. A Graph Theoretic Approach to Tonal Modulation. Journal of Mathematics and Music 4 (2010),
45-56.
[2] Douglas B. West. Introduction to Graph Theory. Pearson. 2000
[3] A. de Mier and M. Noy. On Graphs Determined by Their Tutte polynomial. Graphs Combin. 20 (2004), 105-119.
[4] Steven J. Leon. Linear Algebra with Applications. Pearson Education, Inc. 2006.
[5] Jaromy Scott Kuhl. The Tutte Polynomial and the Generalized Petersen Graph. Australasian Journal of Combinatorics 40 (2008), 87-97.

12