See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/252818080

The normal modes of vocal fold tissues

Article  in  The Journal of the Acoustical Society of America · October 1996

DOI: 10.1121/1.417435

CITATIONS READS
4 122

1 author:

David A Berry
University of California, Los Angeles
84 PUBLICATIONS   2,639 CITATIONS   

SEE PROFILE

All content following this page was uploaded by David A Berry on 14 October 2014.

The user has requested enhancement of the downloaded file.

Normal modes in a continuum model of vocal fold tissues
David A. Berry and Ingo R. Titze
Department of Speech Pathology and Audiology, National Center for Voice and Speech, The University
of Iowa, Iowa City, Iowa 52242-1012

Received 15 April 1996; accepted for publication 12 July 1996

The Ritz method is used to calculate eigenmodes and eigenfrequencies in a continuum model of the
vocal folds. The investigation represents a rectification and extension of previous studies,
emphasizing the indispensability of utilizing natural boundary conditions when computing the
characteristic modes of a system. Concurring with previous assertions, two of the lower-order
eigenmodes are theorized to play a major role in facilitating self-oscillation of the folds during
phonation. One mode, related to vertical phasing, is shown to have a more direct control over glottal
convergence/divergence than indicated in previous calculations. Unlike lumped element models, the
continuum model predicts that the eigenfrequencies of the two modes are closely spaced over an
extensive range of tissue sizes and stiffnesses. This finding may help explain why the two modes
entrain so naturally over a wide range of phonatory adjustments in human phonation. © 1996
Acoustical Society of America.
PACS numbers: 43.70.Aj AL

INTRODUCTION vibration patterns related to normal human phonation
Moore

Vocal fold vibration during sustained phonation is gen-
erally understood to be a self-excited dynamical system
Ish-
izaka and Flanagan, 1972; Titze and Talkin, 1979; Alipour
and Titze, 1985. Almost universally, glottal aerodynamics
are modeled as an intrinsic part of the laryngeal vibrations

e.g., aerodynamic pressures are usually expressed as func-
tions of vocal fold tissue displacements. Thus a proper
analysis of system dynamics would normally account for this
aerodynamic component. On the other hand, over the past
two decades eigenstudies of the vocal folds, conducted in the
absence of aerodynamic forces, have demonstrated remark-
able predictive power in describing aerodynamically induced
vibrations
Titze and Strong, 1975; Titze, 1976; Kaneko
et al., 1981; Kaneko et al., 1983; Kaneko et al., 1986; Ish-
izaka, 1988.
Of particular note are the empirical studies by Kaneko
et al.
1986. In these investigations, the vocal folds of hu-
man subjects were pulse excited with a mechanical tap to the
thyroid cartilage. Ultrasonic techniques were used to extract
the natural frequencies of the folds from the resultant oscil-
lations. Following this brief procedure which lasted only mil-
liseconds, the subjects were instructed to immediately begin
phonation. Across a wide range of phonatory adjustments,
two dominant eigenfrequencies consistently emerged in the
investigations, the lower eigenfrequency corresponding well
to the phonation frequency that followed.
This correspondence between the fundamental fre-
quency of oscillation and the lowest eigenfrequency has also
been noted in linearized versions of the two-mass model

Titze, 1976; Ishizaka, 1988. Furthermore, the eigenmodes
of the two-mass model, one with the two masses in phase
and the other with the two masses 180° out-of-phase, have
been argued to facilitate energy transfer from the airflow to
the tissue, enabling self-oscillation of the folds
Stevens,
1977; Broad, 1979; Titze, 1988.
At least in a qualitative sense, superpositions of these
two eigenmodes provide a reasonable description of most
and Von Leden, 1958; Hirano, 1975. Moreover, empirical
eigenfunctions were recently extracted from a finite element
simulation of vocal fold vibration which incorporated aero-
dynamic forces and many other nonlinearities
Berry et al.,
1994. In spite of these influences, the empirical eigenfunc-
tions showed a significant correspondence to the linear
eigenmodes, and a superposition of the two dominant empiri-
cal eigenfunctions yielded a precise description of the overall
vibration pattern of the folds.
In view of the foregoing empirical and theoretical stud-
ies, the significance of eigentheory to a study of vocal fold
dynamics cannot be easily dismissed. Although aerodynam-
ics is undoubtedly an intrinsic component of vocal fold dy-
namics, the role played by aerodynamics may be secondary
in comparison to the primary role played by the linear eigen-
modes. Indeed, this is what seems to be suggested by the
foregoing studies.
The present investigation considers the predictions of
eigentheory for a continuum model of vocal fold vibration.
Previously, Titze and Strong
1975 obtained an analytic so-
lution for the eigenmodes of a continuum model of com-
pletely compressible vocal tissues. However, because free
boundary conditions were not utilized in determining the so-
lutions, they did not correspond to the true eigenmodes and
eigenfrequencies of the system. Specifically, because shear
stresses were not required to vanish on the free surfaces, the
calculated modes and frequencies were not necessarily the
characteristic ones of the system.
Although the continuum model was generalized in Titze

1976 to incorporate incompressibility and viscosity of vo-
cal fold tissues, calculation of eigenmodes and eigenfrequen-
cies was not attempted in this investigation. Thus while
eigentheory has shown promise for interpreting many aspects
of vocal fold dynamics for several decades, the specific con-
sequences of eigentheory analysis for continuum models of
vocal fold vibration remain relatively unexplored. The pur-
pose of the present investigation is to help remedy this situ-
ation.

3345 J. Acoust. Soc. Am. 100 (5), November 1996 0001-4966/96/100(5)/3345/10/$10.00 © 1996 Acoustical Society of America 3345
 and X̄, ¯
Y , and ¯
Z are body forces, i.e., forces which act over
the entire volume, such as gravity. In this development, Eq.

1b is not relevant because of the constraint on y displace-
ments for small vibrations. Consequently, the general tissue
displacement vector for small oscillations about the equilib-
rium position may be written as

2

Note that Eq.
1 is a mixed equation, i.e., the equation

contains both stresses and displacements. In the solution of
these equations, one usually expresses the equations in terms
of stresses only, or displacements only. In this investigation,
it is more appropriate to work with displacements, since
eigenmode shapes are expressed in terms of displacements.
Stresses may be removed from the equations of motion using

冉冊 冉 冊
FIG. 1. The rectangular parallelepiped and associated boundary conditions
to approximate a vocal fold like continuum. Boundary conditions at each Hooke’s law:
surface are centered within the appropriate face of the parallelepiped.
After
Titze and Strong, 1975. x  /  x
y 
/y
z  /  z
I. THE EQUATIONS OF MOTION ⫽C  , where  ⫽ ,
3
 xy 
/  x⫹  /  y
As in former treatments
Titze and Strong, 1975; Titze,  yz  /  y⫹ 
/  z
1976, several assumptions are made in the derivation and  zx  /  z⫹  /  x
solution of the equations of motion for the 3-D continuum.
First of all, vocal fold tissues are assumed to be elastic and
transversely isotropic. Because tissues are generally stiffer in
the direction of fibers, transverse isotropy is probably the
simplest realistic assumption that can be made for vocal fold
tissues. To a first approximation, the direction of the tissue
fibers is assumed to lie along the y axis. Isotropy, then, exists
where C is the stiffness matrix and  is the strain vector, as
defined above. In general, C is a 6⫻6 symmetric matrix with
21 independent constants. For the case of transverse isotropy
about the y axis, the inverse of C may be expressed in terms
of just five independent mechanical constants
as adapted
from Lekhnitskii, 1963:

冢 冣
only in the x-z plane, i.e., the plane transverse to the fibers. 1 ⬘
The folds are also assumed to take the shape of a rect- ⫺ ⫺ 0 0 0
E E⬘ E
angular parallelepiped. In general, simple geometrical shapes
are required in order to obtain analytic solutions for normal ⬘ 1 ⬘
⫺ ⫺ 0 0 0
modes. Fixed boundary conditions are imposed at the lateral, E⬘ E⬘ E⬘
anterior and posterior surfaces, while free boundary condi- ⬘ 1
tions are dictated at the medial, superior, and inferior sur- ⫺ ⫺ 0 0 0
E E⬘ E
faces
see Fig. 1. For small oscillations about this equilib- C ⫺1
⫽ ,
4
rium configuration, two additional assumptions are made:
1 1
0 0 0 0 0
the stress–strain relation is assumed to be linear, and
2 all ⬘
y displacements are assumed to be negligible. The latter as- 1
sumption is based on observations of trajectories of vocal 0 0 0 0 0
⬘
fold fleshpoints during self-oscillation
Baer, 1981; Saito
et al., 1985. 1
0 0 0 0 0
For infinitesimal displacements in an elastic continuum, 
Newton’s second law of motion may be expressed as
where E is the Young’s modulus in the transverse plane, E ⬘
 x   xy   zx  2 is the Young’s modulus along the y axis, ⬘ is the shear
⫹ ⫹ ⫹X̄⫽  2 ,
1a modulus along the y axis, is the Poisson’s ratio in the
x y z t
transverse plane, and ⬘ is the Poisson’s ratio along the y
  xy  y   yz ¯  2
axis. Although  is another common engineering constant
⫹ ⫹ ⫹Y ⫽  2 ,
1b
x y z t
the shear modulus in the transverse plane, it is not indepen-
dent of the other constants, but may be expressed as
  zx   yz  z ¯  2 
⫹ ⫹ ⫹Z ⫽  2 ,
1c E
x y z t
⫽ .
5
2
1⫹ 
where ,
, and  are displacements in the x, y, and z direc-
tions; the i ’s are normal stresses; the i j ’s are shear stresses; The corresponding stiffness matrix, C, is

3346 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3346
 冢 冣

⫺E ⬘ ⫹ ⬘ 2 E  E EE ⬘ ⬘
⬘ 2 E⫹ E ⬘  E
⫺ ⫺ 0 0 0

1⫹  k k
1⫹  k
EE ⬘ ⬘ E ⬘ 2
⫺1  EE ⬘ ⬘
⫺ ⫺ 0 0 0
k k k
C⫽
⬘ 2 E⫹ E ⬘  E EE ⬘ ⬘
⫺E ⬘ ⫹ ⬘ 2 E  E ,
6
⫺ ⫺ 0 0 0

1⫹  k k
1⫹  k
0 0 0 ⬘ 0 0
0 0 0 0 ⬘ 0
0 0 0 0 0 

where k⫽⫺E ⬘ (1⫺ )⫹2 ⬘ 2 E. pletely incompressible material, a given lateral strain must
induce an equal and opposite vertical strain
e.g., lateral
compression must result in vertical elongation. This can be
substantiated with the well-known kinematic expression for
II. VALUES OF THE MECHANICAL CONSTANTS
incompressibility, which states that the divergence of the dis-
A brief rationale for the mechanical constants used in placement vector equals zero:
study will now be given. First, consider the role that the
Poisson’s ratio ⬘ plays in scaling the longitudinal strain,

/ y, induced by the lateral strain,  / x:  
“–⌿⫽ ⫹ ⫽0.
10
x z



⫽⫺ ⬘ .
7
y x Through comparison of Eqs.
9 and
10, one sees that
must equal one for the case of incompressibility. Although
Because of the constraint on y displacements throughout the many texts on elastic theory state that the upper bound on the
entire parallelepiped
for small oscillations about the equilib- Poisson’s ratio is 0.5, such a bound is relevant only for iso-
rium position,
and all of its spatial derivatives must be tropic materials As demonstrated by Lempriere
1968, the
zero, including the longitudinal strain in Eq.
7. Because the upper bound on for transversely isotropic materials is given
longitudinal strain remains zero for any arbitrary lateral by
strain, the constant ⬘ must be identically zero. This result
considerably simplifies the stiffness matrix:
1⫺2 ⬘ 2 E/E ⬘ .
11

冢 冣
E E
0 0 0 0

1⫹ 
1⫺ 
1⫹ 
1⫺  Because ⬘ equals zero in this problem, the upper bound on
0 E⬘ 0 0 0 0
8 is one, which corresponds to the condition of incompress-
E E ibility.
C⫽ 0 0 0 0 . A series of experiments have been conducted to quantify

1⫹ 
1⫺ 
1⫹ 
1⫺ 
0 0 0 ⬘ 0 0 E ⬘ for the body, cover, and ligament of the folds
Alipour
and Titze, 1991; Min et al., 1995. Unfortunately, the results
0 0 0 0 ⬘ 0
of those investigations do not directly impact the present
0 0 0 0 0 
study, because E ⬘ does not enter the equations of motion for
the condition of no y displacements. However, a useful ob-
Analogous to Eq.
7, the function of the transverse servation that may still be gleaned from these studies is that
Poisson’s ratio is shown in the following expression, illus- for low strains, the values of E ⬘ for the body, cover, and
trating that a vertical strain / z is induced by a lateral ligament are all on the order of 105 dyn/cm2. In the absence
strain, / x: of measurements for mechanical constants in the transverse
plane, we make the assumption that for low strains, both E
and ⬘ equal approximately 105 dyn/cm2 throughout all the
  layers of vocal fold tissues. Because of the uncertainty sur-
⫽⫺ .
9
z x rounding the values of E and ⬘, one of the topics investi-
gated in this paper will be the influence of these parameters
In the case of a completely compressible material
equals on the resultant eigenfrequencies. This will be accomplished
zero, Eq.
9 predicts that a given lateral strain will induce through an examination of isofrequency contours generated
no vertical strain. On the other hand, in the case of a com- in the E-⬘ plane.

3347 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3347
III. SOLUTION OF THE EQUATIONS OF MOTION  1  1

“ⴛ⌿1  •j⫽ ⫺ ⫽0.
18
By employing Hooke’s law Eq.
3 and the stiffness x z
matrix given in Eq.
8, the equations of motion may be Similarly, the second component is an incompressible wave:
expressed as
 2  2
䊐 2 ⫹
1⫹  ξ gy 
1⫺  x  x
⫹ 冉
z
⫽0, 冊
12a
“–⌿2 ⫽
x
⫹
z
⫽0.

In the limit of incompressibility, the irrotational component

19

䊐 2 ⫹
1⫹   
⫹冉
1⫺  z  x  z
⫽0, 冊
12b
vanishes. However, in the general case, eigenmodes consist
of both irrotational and incompressible components.

where IV. THE FREE BOUNDARY CONDITIONS

2 2 2 2
  ⬘  1  As already mentioned, fixed boundaries are imposed at
䊐 i⫽ ⫹ ⫹ ⫺ , i⫽1,2,
12c
 x 2  z 2  c 2i  y 2 c 2i  t 2 the lateral, anterior, and posterior surfaces; and free bound-
aries at the medial, superior, and inferior surfaces. But what
c 1⫽ 冑 1 E
1⫺ 2 
, c 2⫽ 冑 1 E
2
1⫹  
.
12d exactly is a free boundary? In the literature, free or natural
boundary conditions correspond to those conditions which
require the first variation in the potential energy to vanish
Following a development similar to that given by Teodor-

McFarland et al., 1972; Saada, 1974. For the second-order
escu
1972, an effective method for decoupling the lateral
equations in question, this condition is obtained by requiring
and vertical displacements, and , may be realized by ex-
the normal and shear stresses to vanish on the free surfaces
pressing Eq.
12a in the following form:

Teodorescu, 1972; Leissa and Zhang, 1983. However, be-
D 1 ⫽D 2  ,
13 cause the y displacements,
, have already been forced to
vanish, it is not possible to simultaneously demand that the
where D 1 and D 2 are differential operators. Assuming that
shear stresses in the y direction also vanish. Consequently,
the differential operators are interchangeable, and  can be
only the shear stresses normal to the y axis are required to
expressed in terms of a common displacement function, F:
vanish on the free surfaces.
⫽D 2 F,  ⫽D 1 F.
14 By imposing the fixed boundary conditions at the ante-
rior and posterior surfaces, the method of separation of vari-
Substitution of Eq.
14 into Eq.
12b yields a fourth-order,
ables may be used to obtain an expression for the y depen-
biharmonic wave equation:
dence of and , characterized through means of their
䊐 1 䊐 2 F⫽0.
15 common displacement function F:
Without loss of generality, the function F may be expressed
in terms of two components satisfying distinct second-order F n ⫽F 1n ⫹F 2n ⫽sin 冉 冊
ny
L
 f 1
x,z  ⫹ f 2
x,z  .
20
wave equations:
From the equations of motion Eqs.
16, it appears that
䊐 1 F 1 ⫽0, 䊐 2 F 2 ⫽0, where F⫽F 1 ⫹F 2
16 the method of separation of variables may also be used to
decouple the x and z dependence of and . However, the
and and  may be obtained from the displacement functions
complex boundary conditions do not permit such a decou-
using the following relations:
pling. In fact, the boundary conditions may only be satisfied
⫽ 1⫹ 2 ,
17a exactly through means of an infinite series solution. How-
ever, through appropriate truncation of the series, approxi-
 ⫽  1⫹  2 ,
17b mations may be obtained to any desired degree of accuracy.
1⫹  2 F 1 For example, the method of superposition has been used to
1 ⫽⫺ ,
17c compute eigenmodes and eigenfrequencies of similar prob-
1⫺  x  z
lems utilizing a Fourier series approximation
Gorman,
1⫹  2 F 2 1982.
2 ⫽⫺ ,
17d
1⫺  x  z
V. THE POTENTIAL ENERGY AND THE RITZ METHOD
1⫹  2 F 1
 1 ⫽⫺ ,
17e A simpler approach for calculating the eigenmodes and
1⫺  z 2 eigenfrequencies involves an implicit, rather than explicit,
1⫹  2 F 2 application of the boundary conditions. As previously men-
 2⫽ .
17f tioned, the free boundary conditions are deduced by requir-
1⫺  x 2
ing the first variation of the potential energy to vanish
Mc-
At this point, it is possible to assign a physical interpre- Farland et al., 1972; Saada, 1974. Based on this principle,
tation to the two distinct components of the eigenmode so- the Ritz method is implemented by assuming a general form
lutions. For example, using Eqs.
17, one can verify that the of the solution—usually a polynomial with unknown coeffi-
first component is an irrotational wave about the y axis: cients
Ritz, 1908. Upon inserting the solution into the po-

3348 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3348
tential energy function, the coefficients may be determined TABLE I. Parameter values used in computing eigenmodes and eigenfre-
quencies.
by requiring the first variation in the potential energy to van-
ish. Leissa and Zhang
1983 have shown this to be an effi- Lateral depth, D 1.0 cm
cient method for determining the eigenfrequencies of an iso- Longitudinal
anterior–posterior length, L 1.2 cm
tropic cantilevered rectangular parallelepiped. A cantilevered Vertical thickness, T 0.7 cm
parallelepiped differs from the present problem in that it has Tissue density,  1.03 g/cm3
Transverse Young’s modulus, E 105 dyn/cm2
only one fixed surface, rather than three. The isotropic as-
Longitudinal shear modulus, ⬘ 105 dyn/cm2
sumption also represents a simplification of the present Transverse Poisson’s ratio, 0/0.9999
study.

冉 冊
Longitudinal Poisson’s ratio, ⬘ 0
Washizu
1968 has shown that the total potential en-
ergy, , for an elastic continuum in free vibration may be
written as I1 I1 I1
•••
⫽I 1 ⫺  2 I 2 ,
21a
 v 21 v1 v2 v1 v3
I1 I1 I1
•••
where Q⫽ v2 v1  v 22 v2 v3 ,
25c
I1 I1 I1
1
冕 冕 冕
z⫽T y⫽L x⫽D •••

冉 冊
I 1⫽  T C  dx dy dz,
21b v3 v1 v3 v2  v 23
2 z⫽0 y⫽0 x⫽0
⯗ ⯗ ⯗ ⯗

I 2⫽

2
冕 冕 冕
z⫽T

z⫽0
y⫽L

y⫽0 x⫽0
x⫽D

2 ⫹  2  dx dy dz.
21c
I2
 v 21
I2
v1 v2
I2
v1 v3
•••

I2 I2 I2

As in Eq.
3,  refers to the strain vector. •••
Using the results of Eq.
20, a general solution for a few R⫽ v2 v1  v 22 v2 v3 .
25d
of the lowest-order eigenmodes may be written as I2 I2 I2
•••
v3 v1 v3 v2  v 23
I J
y

x,y,z,t  ⫽sin  t sin
L 
i⫽1 j⫽0
A i jx iz j,
22a ⯗ ⯗ ⯗ ⯗
Utilizing MAPLE V, exact differentiations and integra-
M N tions were performed on all terms in the equations, yielding
y analytic expressions for each element in the Q and R matri-

x,y,z,t  ⫽sin  t sin
L  
m⫽1 n⫽0
B mn x m z n .
22b
ces. Only in the final step of the analysis were numerical
values utilized to compute the eigenmodes and eigenfrequen-
Notice that the general solution automatically incorporates cies. Results were also verified using ABAQUS, a finite el-
all the fixed boundary conditions. In order to simultaneously ement software package.
satisfy the free boundary conditions, the first variation in the
potential energy must vanish, as expressed in the following
VI. RESULTS AND DISCUSSION
equations:
Table I shows the parameter values used in computing
 the eigenmodes and eigenfrequencies. In order to compare
⫽0
i⫽1,...,I; j⫽0,...,J  ,
23a our findings with earlier studies
Titze and Strong, 1975;
Aij
Titze, 1976, the two following cases are examined:
1 the
 case of complete compressibility
⫽0, and
2 the case of
⫽0
m⫽1,...,M ;n⫽0,...,N  .
23b near incompressibility
⫽0.9999. Figure 2 shows how the
 B mn eigenfrequencies converge as the order n of the Q and R
matrices increases. For both cases, three significant digit ac-
The general form of the potential energy allows the forego- curacy is obtained as the order of the matrices approaches
ing equations to be expressed in a standard matrix formula- 60. Also, in both instances, the lowest eigenfrequency is ap-
tion of the eigenvalue problem: proximately 133 Hz. The second and third eigenfrequencies
range between 150 and 160 Hz, and closely approximate
Qv⫽  2 Rv,
24 each other for the case of near incompressibility.
Figure 3 illustrates the dependence of the eigenfrequen-
where cies on the degree of incompressibility. While the first eigen-
frequency appears relatively insensitive to variations in in-
vT ⫽ v 1 , v 2 , v 3 ,...  ,
25a compressibility, the influence of incompressibility on the
next two eigenfrequencies is clear: The second eigenfre-
⫽  A 10 ,...,A IJ ,B 10 ,...,B M N  ,
25b quency increases with incompressibility, while the third

3349 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3349
 FIG. 3. The three lowest eigenfrequencies of the vocal folds plotted as a
function of tissue incompressibility.

this eigenmode is identical to the x-10 of Titze and Strong

1975. It is the only lower-order eigenmode which can be
expressed in terms of a simple trigonometric function and
still meet the free boundary conditions. Next, eigenmode 3
see Fig. 4
c exhibits strong lateral and vertical compo-

FIG. 2. Convergence of the three lowest eigenfrequencies as a function of

the order n of the Q and R matrices for
a completely compressible tissue,
and
b nearly incompressible tissue.

eigenfrequency decreases. This effect is sufficiently pro-

nounced that the ordering of the eigenfrequencies changes
for 0.8.
Figure 4 depicts a few of the lower-order eigenmodes
for the case of completely compressible tissues. Because the
y dependence of the eigenmodes is already known see Eq.

20, the eigenmodes are illustrated as 2-D coronal cross
sections in the x-z plane. While eigenmode 1 see Fig. 4
a
is similar to the z-10 mode presented in Titze and Strong

1975, this eigenmode contradicts their assertion that eigen-
modes for completely compressible tissues are either purely
lateral or purely vertical, with no x-z coupling. Eigenmode 1
clearly contains both lateral and vertical components. How- FIG. 4. The three lowest-order eigenmodes for completely compressible
ever, because it is predominantly a vertical mode of vibra- tissue:
a eigenmode 1, analogous to the z-10 mode;
b eigenmode 2,
tion, it probably does not play a major role in phonatory identical to the x-10 mode;
c eigenmode 3, analogous to the x-11 mode.
Left and right folds shown to give indication of glottal geometry produced
dynamics. In contrast, eigenmode 2 see Fig. 4
b is a purely
by corresponding mode. Upper and lower frames indicate extreme positions
lateral mode of vibration, with no vertical components. With of eigenmodes, spaced 180° apart in a vibratory cycle. Dotted lines indicate
all lateral displacements moving in-phase with each other, equilibrium positions.

3350 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3350
 TABLE III. Irrotational and incompressible components of eigenmodes
in
%.

irrot
i 
incompr
i 

compr
1 , 17.3% 82.7%


compr
2 , 100.0% 0.0%


compr
3 , 39.6% 60.4%


inc
1 , 0.0% 100.0%


inc
2 , 0.0% 100.0%


inc
3 , 0.0% 100.0%

eigenmodes are very similar, although not identical. Further-

FIG. 5. The three lowest-order eigenmodes for nearly incompressible tissue:

a eigenmode 1, analogous to the z-10 mode;
b eigenmode 2, analogous to
the x-11 mode;
c eigenmode 3, analogous to the x-10 mode.
nents. Nevertheless, it is qualitatively similar to the x-11
mode of Titze and Strong
1975 in the role that it plays in
glottal shaping. During oscillation, this eigenmode causes the
glottal airway to alternate between convergent and divergent
shapes, as portrayed in the upper
divergent and lower
con-
vergent frames of Fig. 4
c. In fact, because the medial edge
of this eigenmode is characterized by a straight line, it has a
simpler, more direct control over glottal convergence/
divergence than the x-11 mode, which is characterized by a
sine function along the medial edge.
In contrast to the assumption of compressible tissue, in-
compressibility is believed to better approximate the condi-
tion of the true folds. This is because vocal fold tissues are
composed mostly of water, and are thus essentially incom-
pressible. A few of the lower-order eigenmodes for incom-
pressible tissues are shown in Fig. 5. With one possible ex-
ception, these eigenmodes have not changed remarkably
from the previous case. For example, it is difficult to detect
any difference between the eigenmodes in Figs. 4
a and
5
a, or in Figs. 4
c and 5
b. Dot products of the eigen-
modes, presented in Table II, are a further indication that the
TABLE II. Dot product of corresponding compressible and incompressible
eigenmodes.


compr ,
inc
1  0.996
1
compr inc


2 ,
3  0.987


compr
3 ,
inc
2  0.940
more, although visual differences are apparent in the eigen-
modes shown in Figs. 4
b and 5
c qualitative similarities
are still present. For example, along the medial edge of the
folds, the tissue movement is predominantly lateral and in-
phase, just as in Fig. 4
b. Even numerical comparisons dem-
onstrate the correspondence between these two eigenmodes,
their dot product yielding 0.94.
Earlier in the paper, it was noted that all eigenmodes
may consist of both irrotational and incompressible compo-
nents. Table III indicates the relative dominance of these
components for each of the eigenmodes. Figure 4
b is an
example of an eigenmode that is purely irrotational, i.e., it
has no curvature. The corresponding eigenmode in Fig. 5
c,
however, is purely incompressible.
Note that the other two ‘‘compressible’’ eigenmodes in
Fig. 4
a and
c are predominantly ‘‘incompressible.’’ As
substantiated by Eqs.
18 and
19, even completely com-
pressible solutions may have both irrotational and incom-
pressible components
however, irrotational components do
vanish in the limit of incompressibility. Thus it is not sur-
prising that these two eigenmodes which are largely incom-
pressible, even in the case of complete compressibility,
change very little as the incompressibility condition is man-
dated.
As already stated, the z-10-like mode probably does not
play an important role in vocal fold dynamics. However, it
has long been asserted that the other two eigenmodes are
essential for self-oscillation
Titze, 1988. The x-11-like
eigenmode has already been shown to produce alternating
divergent and convergent shapes in the glottis during oscil-
lation. Because of this correspondence with glottal shaping,
the eigenmode exerts considerable influence over the intra-
glottal pressure
Titze, 1988. The x-10-like eigenmode, on
the other hand, yields a rough description of the net lateral
movement of the folds.
The basic condition for self-oscillation is that the glottal
airflow must impart sufficient energy to the folds to over-
come the energy losses associated with vocal fold vibration

i.e., viscous damping, collision, etc.. In order to satisfy this
condition and maximize the energy imparted to the folds, the
tissue velocity
controlled predominantly by the x-10 mode
must be in-phase with the intraglottal pressure
controlled
predominantly by the x-11 mode. A unique phase relation-
ship between the two eigenmodes ensures that this condition
is satisfied. This relationship has been observed and carefully
investigated in finite element simulations of vocal fold
movement
Berry et al., 1994.

3351 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3351
 The two-mass model of Ishizaka and Flanagan
1972,
upon linearization, also has two eigenmodes which are con-
ceptually equivalent to the x-10 and x-11 modes. For ex-
ample, the mode with both masses in-phase corresponds to
the x-10 mode and the mode with both masses out-of-phase
corresponds to the x-11 mode. Because of this correspon-
dence, throughout the remainder of this discussion, we refer
to the eigenmodes of the two-mass model as x-10 and x-11
modes. Much of the success of the two-mass model might be
attributed to its ability to capture these two eigenmodes
which facilitate self-oscillation.
It has been observed that during self-oscillation the fun-
damental frequency of the two-mass model corresponds
closely to the eigenfrequency of the x-10 mode, which is the
lowest-order eigenmode of the model. In the present investi-
gation of eigenfrequencies in a continuum, it has been ob-
served that the x-10 mode has a lower eigenfrequency than
x-11 mode only for the case of compressible tissues. For the
more realistic case of near incompressibility, it is the x-11
mode, the mode most closely related to intraglottal air pres-
sure, which has the lower eigenfrequency. However, because
these two eigenfrequencies are so closely spaced in the con-
tinuum model, it may not be meaningful to unduly empha-
size this issue.
Indeed, a major discrepancy between the two-mass
model and the continuum study concerns the relative spacing
of the eigenfrequencies of the two dominant eigenmodes. For
typical Ishizaka and Flanagan
1972 parameters, the eigen-
frequencies of the x-10 and x-11 modes are approximately
120 and 201 Hz, respectively. When system nonlinearities
and aerodynamic driving forces are included, the two modes
entrain at a frequency very close to the lowest eigenfre-
quency. In contrast, for the parameters used in this study of
incompressible tissues, the corresponding eigenfrequencies
are 153 and 151 Hz—a relative close spacing. Such a close
spacing of eigenfrequencies would facilitate a 1:1 entrain-
ment of the modes during self-oscillation. Although nonlin-
earities are necessarily neglected in calculating the linear
eigenfrequencies, it is nevertheless understood that the actual
physical system does exhibit some nonlinearities, enabling
an entrainment of the eigenmodes
the 1:1 entrainment being
the most common.
Because of the relative uncertainty concerning the pa-
rameter values used for E and ⬘ in the continuum model,
the eigenfrequencies for the three lowest-order eigenmodes
were computed across the entire E-⬘ plane for the case of
near incompressibility, as shown in Fig. 6. Across the entire
plane, the x-11 mode always had a lower eigenfrequency
than the x-10 mode. However, the spacing of these two
eigenfrequencies Fig. 6
b and
c never differed by more
than several Hz. Furthermore, this close spacing of the eigen-
frequencies was observed in the model over a wide variety of
glottal depths, lengths, and thicknesses. Thus over an exten-
sive range of tissue sizes and stiffnesses, the model suggests
that that these two modes could easily be entrained.
This finding is reminiscent of observations in normal
human phonation. For example, most individuals, male and
FIG. 6. Isofrequency contours in the E-⬘ parameter plane associated with
female, are able to produce normal, periodic phonation over the eigenfrequencies for
a eigenmode 1,
b eigenmode 2, and
c eigen-
a wide range of phonatory adjustments. Why is this so? For mode 3. Computed for the condition of nearly incompressible tissue.

3352 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3352
nearly all cases of periodic phonation, vocal fold vibrations
may be appropriately described as a superposition of two
dominant modes, one with a vertical-phasing component and
one without. The proper combination of these modes is
known to facilitate self-oscillation
Stevens, 1977; Broad,
1979; Titze, 1988. The present investigation suggests a rea-
son for the easy entrainment of these two eigenmodes over a
wide variety of vocal conditions, i.e., the close spacing of
their eigenfrequencies.
In regard to Fig. 6, the influence of the parameters E and
⬘ on the eigenfrequencies also merits discussion. For low
values of E, the eigenfrequencies of all three eigenmodes are
nearly identical. Increasing the value of ⬘ boosts all the
eigenfrequencies. Increasing the value of E also boosts the
eigenfrequencies of eigenmodes 2 and 3; however, it has
little influence on eigenmode 1. Thus the eigenfrequencies of
the lowest-order eigenmode are mostly a function of ⬘,
while those of the two higher-order modes have a strong
dependence on both E and ⬘. One implication of this ob-
servation is that for large values of E, a greater spacing ex-
ists between the lowest eigenfrequency and those of the two
higher-order modes. As already mentioned, the two higher
eigenfrequencies retain their close spacing throughout the
entire E-⬘ plane.
VII. SUMMARY
The eigenmodes and eigenfrequencies of a continuum
model of the vocal folds have been computed using standard
analytic and numerical approximation techniques from elas-
tic theory. In particular, the Ritz method has proven useful in
calculating the eigenfunctions. The new calculations indicate
that the eigenmode related to vertical phasing has a more
direct control over glottal convergence/divergence than pre-
viously supposed
Titze and Strong, 1975, strengthening
previous claims that this mode is intrinsically related to the
intraglottal pressure
Titze, 1988. Another novel outcome of
this investigation was that the eigenfrequencies of the x-10
and x-11 eigenmodes of nearly incompressible tissues were
closely spaced throughout an extensive parameter region of
the model. This finding may help explain why the two eigen-
modes entrain so naturally over a wide range of phonatory
adjustments in human phonation.
ACKNOWLEDGMENTS
This work was supported by Grant No. P60 DC00976
from the National Institute on Deafness and Other Commu-
nication Disorders. The ICAEN computing center at the Uni-
versity of Iowa generously allocated computer time and soft-
ware to allow the results of this investigation to be verified
with the finite element method. The authors also thank Dr. B.
Kent Harrison of Brigham Young University for advice con-
cerning the analytic solution of this problem.
Alipour-Haghighi, F., and Titze, I. R.
1985. ‘‘Simulation of particle tra-
jectories of vocal fold tissue during phonation,’’ in Vocal Fold Physiol-
ogy: Biomechanics, Acoustics, and Phonatory Control, edited by I. R.
Titze and R. C. Scherer
Denver Center for the Performing Arts, Denver,
CO.
3353 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996
Alipour-Haghighi, F., and Titze, I. R.
1991. ‘‘Elastic models of vocal fold
tissues,’’ J. Acoust. Soc. Am. 90, 1326–1331.
Baer, T.
1981. Investigation of the phonatory mechanism. ASHA Reports

11, 38–46.
Berry, D. A., Herzel, H., Titze, I. R., and Krischer, K.
1994. ‘‘Interpreta-
tion of biomechanical simulations of normal and chaotic vocal fold oscil-
lations with empirical eigenfunctions,’’ J. Acoust. Soc. Am. 95, 3595–
3604.
Broad, D.
1979. The new theories of vocal fold vibration, in Speech and
Language: Advances in Basic Research and Practice, edited by N. Lass

Academic, New York.
Gorman, D. J.
1982. Free Vibration Analysis of Rectangular Plates

Elsevier, New York.
Hirano, M.
1975. Phonosurgery: basic and clinical investigations, Official
Rep. 76th Annual Convention on Oto-Rhino-Laryngo. Soc., Kurume, Ja-
pan.
Ishizaka, K.
1988. Significance of Kaneko’s measurement of natural fre-
quencies of the vocal folds, in Voice Physiology: Voice Production,
Mechanisms and Functions edited by O. Fujimara
Raven, New York, pp.
181–190.
Ishizaka, K., and Flanagan, J. L.
1972. Synthesis of voiced sounds from a
two-mass model of the vocal cords, Bell Syst. Tech. J. 51, 1233–1267.
Kaneko, T., Uchida, K., Suzuki, H., Komatsu, K., Kanesaka, T., and Koba-
yashi, N., and Naito, J.
1981. ‘‘Ultrasonic observations of vocal fold
vibration,’’ in Vocal Fold Physiology, edited by K. N. Stevens and M.
Hirano
University of Tokyo, Japan, pp. 107–118.
Kaneko, T., Komatsu, K., Suzuki, H., Kanesaka, T., Masuda, T., Numata,
T., and Naito, J.
1983. ‘‘Mechanical properties of the human vocal
fold—Resonance characteristics in living humans and in excised larynes,’’
in Vocal Fold Physiology: Biomechnics, Acoustics, and Phonatory Con-
trol, edited by I. R. Titze, and R. C. Scherer
The Denver Center for the
Performing Arts, Denver, CO, pp. 304–317.
Kaneko, T., Masuda, T., Shimada, A., Suzuki, H., Hayasaki, K., and Ko-
matsu, K.
1986. ‘‘Resonance characteristics of the human vocal folds in
vivo and in vitro by an impulse excitation,’’ in Laryngeal Function in
Phonation and Respiration, edited by T. Baer, C. Sasaki, and K. Harris

Little, Brown, Boston, pp. 349–377.
Leissa, A., and Zhang, Z. D.
1983. ‘‘On the three-dimensional vibrations
of the cantilevered rectangular parallelepiped,’’ J. Acoust. Soc. Am. 73,
2013–2021.
Lekhnitskii, S. G.
1963. Theory of Elasticity of an Anisotropic Body

Holden-Day, San Francisco, p. 25.
Lempriere, B. M.
1968. ‘‘Poisson’s Ratio in Orthotopic Materials,’’ Am.
Inst. Aeronaut. Astronaut. J. 6, 2226–2227.
McFarland, D., Smith, B. L., and Bernhart, W. D.
1972. Analysis of Plates

Spartan Books, New York, p. 22.
Min, Y. B., Titze, I. R., and Alipour-Haghighi, F.
1995. ‘‘Stress-strain
response of the human vocal ligament,’’ Ann. Otol. Rhinol. Laryngol. 104,
563–569.
Moore, G., and Von Leden, H.
1958. ‘‘Dynamic variations of the vibratory
pattern in the normal larynx,’’ Folia Phoniatr. 10, 205–238.
Ritz, W.
1908. ‘‘Über eine neue Methode zuer Lösung gewisser Variation-
sprobleme der mathematischen Physik,’’ Z. Reine Angew. Math. 135,
1–61.
Saada, A. S.
1974. Elasticity: Theory and Applications
Pergamon, New
York, p. 452.
Saito, S., Fukuda, H., Kitahira, S., Isogai, Y., Tsuzuki, T., Muta, H.,
Takyama, E., Fujika, T., Kokawa, N., and Makino, K.
1985. ‘‘Pellet
tracking in the vocal fold while phonating: experimental study using ca-
nine larynges with muscle activity,’’ in Vocal Fold Physiology, edited by
I. R. Titze and R. C. Scherer
Denver Center for the Performing Arts,
Denver, CO.
Stevens, K. N.
1977. ‘‘Physics of laryngeal behaviour and larynx modes,’’
Phonetica 34, 264–279.
Teodorescu, P. P.
1972. Dinamica Corpurilor Liniar Elastice
Editura
Academiei Republicii Socialiste Romnia. English translation: Dynamics
of Linear Elastic Bodies
Abacus, England, 1975, Chap. 3.
Titze, I. R.
1976. ‘‘On the mechanics of vocal-fold vibration,’’ J. Acoust.
Soc. Am. 60, 1366–1380.
Titze, I. R.
1988. ‘‘The physics of small-amplitude oscillation of the vocal
D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3353
 folds,’’ J. Acoust. Soc. Am. 83, 1536–1552. various laryngeal configurations on the acoustics of phonation,’’ J. Acoust.
Titze, I. R., and Strong, W. J.
1975. ‘‘Normal modes in vocal cord tis- Soc. Am. 66, 60–74.
sues,’’ J. Acoust. Soc. Am. 57, 736–744. Washizu, K.
1968. Variational Methods in Elasticity and Plasticity
Per-
Titze, I. R., and Talkin, D. T.
1979. ‘‘A theoretical study of the effects of gamon, New York, pp. 43–44.

3354 J. Acoust. Soc. Am., Vol. 100, No. 5, November 1996 D. A. Berry and I. R. Titze: Normal modes in a vocal fold model 3354

View publication stats