HOWI}IRDSSTI\G

The mechanismof bird songhas traditionallybeen comparedto either

a wind musicalinstrument or the human vocal apparatus.The analysis
of the bird songsthemselvespoints toward an entirely differentsystem
by Crawford H. Greenewalt

ome years ago I read a book in

which, among other things, the author summarized the theories that
had been advanced to explain the physiological processesemployed by a singing
bird. These theories, it seemed to me,
were totally unacceptable, and I determined, perhaps with more rashnessthan
wisdom, to see if I could ftnd an explanation that did nc violerce to either the
aatomical findings or the laws of physics.
The theories to which I took exception compared the singing bird either to
a musical instrument-the clarinet, e
oboe or the trumpet-or alternatively to
the human voice (that is, a bird lvas believed to employ the same devices we
do rvhen we speak or sing). In a clarinet,
for example, pitch and timbre are controlled by the effectve length of the barrel of the instrtrment; the vibrating reed
is in effect driven by and so forced to
conform to the harmonic spectrum of the
resonator.If a bird sang in this fashion,
piich could be varied only by extension
and retraction of the neck. For a song
sparrow, with its two*octave range, the
neck rvould have to be extended bv a
factor of four-clearly a physical impossibility. Furthermore, birds sing many
phrases in which the wave form is sinusoidal (a train of pure sine waves), without an associated harmonic stectrum.
Resoratorssuch as the barrel of a clarinet must by definition produce harmonics and cannot generate a sinusoidal
lvave form.
As for e human-voice theory, when
we speak we produce at the glottis a
series of pufls of air mathematically
equivalent to a harmonic spectrum with
an infinity of components, and with a
fundamental frequency corresponding
to the interval between puffs. The resulting acoustical disturbance is then

modulated in our oral passagesto produce the spoken word. This mechanism
does not (and cannot) produce a purely
sinusoidal wave form; every speech
sound is associated with a harmonic
spectrum of considerablecomplexity.
Negatives such as these are neither
satisfying nor sporting, and to remedy
both potential criticisms I undertook to
develop a physiological and acoustical
model that would describe a singing
bird. The approach was to study in detail the bird songs themselves, and to develop from their constituent parts some
noon of the associatedacousticalDroc-

esses.It seemed useless to employ an

anatomical approach because the many
excellent anatomical studiesin the scientific literature have produced no useful
conclusions,and because the acoustical
analysis of hunran speech has been so
strikingly successful in elucidating our
own vocal performance.
I might pausehere to note that I have
written a book-Bird Song: Acoustics
and Phy siology (Smithsonian Institution
Press, 1968). In it much detail is given
on the state of the literature, the fiistrumentation my associatesand I have developed for the analysisof bird song, and

AIR SAC

TRACHEA

SYRINX

AIR SACS
C L A V I C U L AS
RA C

VOCAL APPARATUS OF A SONGBIRD is shown in schematic form. The vocal organ it'
self is the syrinx, which is located at the point where two bronchi join to form the trachea.

the proofs, ntathematical and otber',,vise,

for the findings I am rbotttto clescribe.
T'his alticle is in efirectIsumnrrry of that
book, and if the reader should rvish more
detail he u,ill find it there.
1.1",* v'oc'alorgan of l>irds is ctlledtlie
r syrinr. It is located deep in the thoracic cavity, at the point r,vherethe trvo
bronchi join to form the trrcrhea
l,seeilIustrations belowl. lt is difficult to decide from the detailed anatonry of the
syrirrx exact)y rvhrt pirrts are important
in vocalization, and precisely horv souird
is produced. At one trne or another alrrost every anatomicll element in the
syrinx has been assigneda ole in vocalization; unhappily there is an overabundance of possibilities.
I have proposed the relatively sirnple
functional structure depicted in the top
iliustration on page B. The elements
are the tympanic mernbranes and their
assocatedmusculature, the external labia and the systern of internal air sacs
(not shown in the illustration) that force
air tlirough the bronchial passage and
l:ulge the membrane into tlle bronclal
lumen, or passageway. The iilustration
depicts the liighly evolved syrinx of a
songbird. At successivelyearlier stages
of avian evolution the syrinx loses the
external labia, then the syrirrgeal rnus*
cles and ends up as a simple tube with

a membrane on its peliphery and contained rvithin inair sac,

It shoulcl be kept in mind that the
functional elements of the syrinx are
doubled, that is, there is a set for each
bronchus. Since each broricl-rialpassage
lras its orvn mernbralle, musculaiure and
nervous s1'stem,it is evident that birds
can control each passageindepenciently
of the other; they can sing what might
be calieclan internal duet.
The systemoperatesas follorvs: When
a bircl undertakes to sing, it in effect
closesrvalve between the lung and the
svrinx. Then it compresses(rvith its chest
muscles)the air in a systemof sacs.Fressure in the ciavicular sac, which surrounds the syrinx, forces the exceedingly
thin tympanic membrane into the bronchial passage, closing it momentarily.
Tension is then applied to the syringeal
rnuscles,which, acting in opposition to
the sac pressure, r.r'ithdraw the bulged
membrane from the opposite bronchial
wall, thus cleating a passage tlirough
the bronchial tube. Air streaming
through the passage past the tensed
mernbrane stimulates it to viblate, and
song is produced. If oniy one of the two
voices is to be used, no tension is applied
to the other membrane, and the corresponding bronchial passage remains
closed.When a duet is sung, both membranes are under tension; there ale

tw-o airstrerrnsand hence two vibratancl tr.vo simultaneous

;flr1;*nranes
The illustrationson the next two pages
shorv three complex songs analyzed in
terms of arnplitude and frequency. The
songs embrace almost every vocal gym.
nastic of rvhich bids are caparble,Note,
for exa.mple,the extremely rapid amplitude modulations, the rvide range in the
shape and extent of the amplitude envelopes and the relatively large frequency intervals: from just over two to just
undel' seven kilocycles per second. The
arnplitude displays are precise and require no qualification. In the frequency
displays there are a tirne delay and a.n
integration of frequencies over srnall
time intervals that introduce ambiguities
for precise analysis.
The question of rvhether or not a bird
can use its two acoustical sources independently is vital to the elucidation of
the mechanics of bird song. If, for instilnce, the bird's trachea behaved like
the barrel of a clarinet, both sources,that
is, both vibrating memlranes,u'ould be
forced to conform to the resonancesof
the trachea.The two sourcesrvould then
merely reinforce each other, increasing
the amplitude, or loudness,of tlre sound.
If, on the other hand, the bird's trachea
behaved like human oral cavities, and
the vibrating membranes were analo-

_-- f'4USCULATURI
TRACH[AL
RINGS
li
t:
l:

\,
\;,
\,,

BRONCHIAL
RINGS

' JF
E X lE N S I C I O
CLAV]CIJtAR
A]Ii SPACI

A I { A T O M Y O I ' T H E S Y R I N X s h e d sl i t t l e l i s h t o n i t s f u n c t i o n . A t
left is tle skeleton of tle syrinx in the maglie; at right, a section

through

the syrinx of the European

based on studies puLlished

blackbircl. Both drau'ings are

by the anatomist

V. Hr:ker in 1900.

gor.rsto the human glottis, the resuiting

wave form should be equivalent to a
harmonic spectrum of great complexity
involving three fundamental frequencies, one for each source and one for the
first resonanceof the trachea.
fact that the tu.,o soulces cutproTatr"
r duce two harmonically unrelated siuusoidal \\,ave forms was ffrst shorvn by
Ralph K. Potter, George A. Kopp and
H. C. Green of the Bell Telephone Lab-

oratories, for a fragment of the song of

a brown thrrsher.Since the phenomenon
is an important one, I have seal'chedfor
and found many similar examples in species representative of most farnilies of
birds. One exampleis provided by a relatively long phrtrse from the song of a
mockingbird lsee illustration on page
6]. In the displays the two voices were
separated with sharply discriminating
acoustical filters; frequency was determined by measuring time intervals for

10 successivesine waves. The tr.vovoices

are hnnnonically unrelated and overlap
on the time axis. We must conclucie th.it
neither the musical-instl'ument theory
nor the human-voice theory is operative.
Other evidence may be found in
phlases that cover a Iarge frequency
range. If such a "glissando" traverses
a tracheal resonance or antiresonance.
there should be a marked increaseor decrease irr amplitude at the appropriate
frequencies. An example is a glissando

8r

>-a6
Ju, z

=
, I
= >-v

6.1

9u)
#
2s"
ii \. r!

4-

J t ) u

.vLL

'l

lr
.
2i*," - ----*.^^.:.-.^"...

"'..,..,": .,. f..;-'','t*'.'

.1...........:4,..''

,,,'.

l:'.'.:'.,.."r*:...,:'.:j).::".,.
1

> - . n6 8 1

/ \ r ' r
v w
7 ) ^

,q.

-, c-t Y 6-i

J.

=>-Y

"6
l r

-*

##

4 -

t l l v
M J f Y

t'*

21
o

'". -'-l

,:::'tt. :.:::,, |;,.t;;;,tt,,;t::,,:,:.,

:,

6B- o
tlJ

()

z;-iOai
v

l r r

) t \ q

r->Ya

-*

u
4 : ^

-,.,
|!
v

^ :

-*

i5 stb.+...

i **
&

!r

"..'l

0
SONGS OF THREE SONGBIRDS are presented in rraces thar show
amplitude and frequency as a function of time. The top record for

each bird is an oscillogram that displays amplitude; the bottom

record is a sound spectrogram that displays frequency. At a is the

'*t*.

't

sung by a yellow warbler that embraces

a full octave [see illustrution on page
7 ]. There are changesin amplitude in
the glissando, but in an acoustical sense
they are very small (lessthan three decibels), and it is most unlikely that they
are associated with tracheal resonances.
The two minima at 90 and 130 milliseconds appear at frequencies of 4.8 and
6.5 kilocycles per second. One of these
could correspond to a resonance; both
of them could not.

Although the evidence seems quite

conclusive that the prin-rary vibrations
produced in the syrinx traverse tlle
trachea without attenuation, ampliffcation or change of any sort, lve must offer
a valid explanation for this rather surprising acoustical inertness of the trachea. The trachea is a tube. A tube, if its
acoustical lossesare small, exhibits resonances, and one would expect it to be
an effective modulator, even lvith its
relatively soft walls. A possible explana-

*l*

'-' ^'

t.

+s

*tT-

T I M E( S E C O N D S )

s--e\

ll

t.c

T t M E( S E C O N D S )

tion-indeed, the only likely explanation

-is that the irnpedance of the
source (the
vibrating membrane in its constrictecl
passage)closely matches the impeclance
of the trachea. Calculations bised on
reasonableassumptions indicate that this
conditon will obtain where the mean
cross-sectional area of the trachea is
about 10 times that of the syrinx. The
curve relating this area ratio io attenuation is fairly flat and shows that the ratio
can be substantially higher or lorver than

ltran**
I

.c

.r$Srffitta*r"
frr&',
2

tl

2.5

-R"** * "
{

rrttll
1

T t M E( S E C O N D S )
song of a rvhite.erowhed

spamo\,r'; at b, the song of a song sparrolv;

longspur. These song recordings illus.

at c, the song of a Lapland

2.5
trate the extraordinary virtuosity of these singing birds. No human
performance can match the complexity of the individual phrasee.

l0 before tracheal attenuation becomes

important.
There is another way of expressing
the effect of variation in the relatve
cross sections of source and resonator
that may be easier to comprehend. If we
take a tube closecl at one end and open
at the other, the resonances will occur
at multiples 1, 3, 5, 7 and so on of the
fundamental frequency. If the tube is

open at both ends, the resonances will

occur at multiples 0, 2, 4, 6 and so on
of the fundamental frequency, If now
we begin with a tube open at both ends
and gradually close off one end, we will
in due course arrive at a twilight zone
within which the tube shows no resotlant
effects. This presumably is the zone in
which birds sing.
Perhaps this is another examPle of

Nature's ingenuity in dealing with the

needs of her diverse creatures.The necks
of birds (which contain and limit the
trachea) have many functions, for instance feeding, preening and nest-building. It would be odd if such functions
operated to restrict freedom in an activity as important as song.
So far I have undertaken to show that
bird song has its origin in the syrinx,

z
(J
lLl f
a

a 3.5
LIJ

LU
J
(J
(J

o
J

Y
>l)

otr
L'

ul
l
a>
LU

.2
LL

160
(MiLLISECONDS)

"INTERNAL DUET' is revealed when a phrase of the song of a

m o c k i n g b i r d ( a ) i s d i s s e c t e di n t o t w o p h r a s e e ( b , c ) w i t h a p p r o '
priate filters. The plots at bottom show the frequencies in b and c.

tlrat tire syrinx contrinstwo tcoustical

sources u'hich can be independently
controlled to produc'e iur interrrirlduet,
and tliit the sounds rvhich originate in
the syrinx pass through the trchen to
the ear of the listener rvithout further
I must norv show 1>recisely
rnclclulrtiorr.
how tl-re sources in the syririx operate
and horv the extraordinarily rapicl arrd
complexmodu]ationsso cornmonin bird
sorlg are produced.
I have noted that pressurein the clavicular sac forces the internal tvmpanic
rnernltranes into tlie bronchial lumerr
against tension appliecl to the nembrule
by the musclesof the s),linx. Air streaming past the resulting constriction in the
bronchial passage stinruiates tlre urembrane to vibrate. If this postulat.eis correct, it follows that in any rapid modulation amplitu<Ie and frequency must be
coupled; an increasein frequency can be
produced only by increasing the tension
in the syringeal muscles.At a given pressure in the clavicular sac this tension will
increasethe cosssectionof the bronchial
passage, which, as we shall see, must
produce either an increaseor a decrease
in the acousticalamplitu<le.
The direct coupling of frequency and
amplitude, that is, amplitude increasing
with irrcreasing frecluency, appears in
curves for Townsend's solitaire in which
frecluency and amplitude are sirnultaneously plotted against tirne Isce u]]per
illwstration on pl.ge 9]. The reverse process, r,vith amplitude falling as frequency
rises, appears in similar curves for the
red-winged blackbir d lsee louserillustration on page 91. In curves for the song
r sporrow one can see both types of coupling within the modulating period: amplitude rises with increasing frequency
up to 6.8 kilocycles and falls above that
frequency.
These phenomeni:Lcan best be explained by referring again to the simplified model of the syrinx shorvn on the
rrext page. If r.vestart rvith just sufficient
tension (T) on the tympanic membranes
to balance the pressure P, the distance
across the passage(D) will be zero and
no rirw'ill floiv. A small increasein tension will open the passageand allow the
membrane to vibrate at a frequency corresponding to the tension,but the amplitude of vibration will be restricted to low
values by the small distance acrr:ssthe
passage. As tlie tension increases, frequency ar-rdamplitude will increase together, as a larger distance across the
passagepermits greater vibratior"ralamplitude in the memlranes. With a further increasein memlrane tension,and a
correspondingly larger distance across

l,ti--"*--*"..*l

.1:

iittl
1
r
il'

it

':

ti . :i i1l i l

o
(J

t:

i : i

(/) ii ,' ,i l

lt3rls"l

LIJ

CI
trl
L

l;;

C)
U

o
Y
5

LL

i
I

'iiifi
ii^l

i
,

Il ; ; 1

_l

-"---j
i

LL]
:f

*-fi;

I
l

!1i,#r
,1,-.
rii-ttl

ir:i1
iijP-,

O
Z

*--ir.

--**i

^^^R
^Hddt

^Yd'"

:-*

i',#i

lti

t.-;ri
tff

t;l

i:
iiiil
'r ii t l

i1

it';l

3 ti
0

50

r00
N f l Lt | S E C O N D S )
T t N , , t( E

i50

2AO

IN THE GLISSANDOof the yellow warbler

freruencyrises smoothly (plot at bottom)
but amplitude (top) rises only moderately.

the passage,we arrive at a point rvhere

the airflorv through the syringeal constriction can no longer stimulate the
rrtembr:lneto vibrate tlirotrgh the allorvable distance D, and the amplitude will
decrease. At the limit-a tension sufficiently large to open the pissagefullythe amplitude will be zero, and the airflow from the ]rird's lungs will escape
through the trachea without sound production. It should be understood thirt
airflow is used here in an acoustical
sense; it is r time-varying flow that is
controllecl by the vibration of the tympanic nreml>rane.Air coulci r,vell leak
past the syringeal constriction, but this

airflow would be continuous and would

produce no sound.
coupling always
'-\nrplitucle-frecluency
accompaniesrapid modulntion, but
it is not necessarilypresent when the pericd of the modulation is long. Consider
a song sparrow plirase about 50 millisecondslong lsee toyt illustration on tage
l0]. The modulating frequency remains
c<lnstantat about 300 cycles per second,
as does the frequency excursion. There
is, however, a gradual rise and fall in
arnplitude from the beginning of the
phrase to the end. Such comparatively
long-term changesin amplitude without

an accompanying change in frequency are relatively common. We find the

necessarycontrol meclirnismin the external labium l"'EL" in the bottom illustration on page l0], whose movement
into and out of the bronchial lumen provides a simple device for changing amplitude without affecting muscle tension and hence without changing the
frequencies produced in the tympanic
membrane.
I must confessthat this role for the external labium rests on a somewhat shaky
anatomicalbase.Many anatomistsreport
SIMPLIFIED MODEL OF THE SYRINX indicates how it functions. Air flows from the
the presence of this member, particular(P)
the
forces
sac
the
clavicular
pressure
in
the
of
air
trachea at right. The
lungs at left into
ly in the songbird syrinx; it is a pillowtympanic membrane (M) into the bronchial lumen, or passage.Tension (T) is produced in
like structure elected on the bronchial
the membrnne by the syringeal muscles (nfs). The resultant o{ the two forces P and T de'
termines the distance across the lumen (D) and also the vibrating frequency. EL is external
half-ring directly opposite the tympanic
labium, which can regulate distance across bronchial lumen without affecting frequency.
membrane. Only one anatomist, however, describes muscular attachments
that could move the labium into and out
of e bronchial lumen. I trust that some
interested anatomist will note this opportunity for a definitive publication.
The bird could, of course, change the
air pressure in its sac system, but this
would have two effects whose resultant
is at best ambiguous. Increasing the sac
pressure would drive the membrane farther into the bronchial lumen, thereby
decreasingits cross section; at the same
time it would increase muscle tension.
The increased pressure drop across the
constricted bronchial passage would increaseflow and hence would compensate
for the reduction in cross section. I am
not rash enough to predict the resultant
of these phenomena, but it seems clear
that they could not produce the overall
rise and fall in amplitude exhibited in
the song sparrow phrase on page 134.
For that we need the external labium or
some other anatomical feature that affects amplitude alone.
In any event, these rapid and highly
variable modulations are so common as
to be a prime clraracteristicin the songs
of most birds. They can be repetitive,
that is, there is a modulating frequency
that continuesfor 50 to 500 milliseconds,
or they can be nonrepetitive, that is, they
may vary rapidly and randomly in amplitude, in flequency or in both.
Many phrases sung by the song spar"..'lll,.rt''"'
AIRTLCW
row comprise modulations that are repetitive. The modulating frequency averages 300 cycles per second, and amplitude ranges much more widely within a
AMpLITUDE AND FREQUENCY ARE COUPLBD when changesin either are rapid. fn a
modulating period than frequency, In
a tone of a given amplitude is produced when the airflow causesthe tensed tympanic mem'
the songsof other speciesthe modulating
brane (M) to vibrate. In b the tension (T) is increased, causing the membrane to vibrate
frequency ranges from about 100 cycles
mem'
The
passage'
faster. Increased tension also retracts membrane, further opening the
per second to something over 400. We
is
brane can therefore vibrate through a larger distance, increasing amplitude of song. This
are dealing here with two oscillating systhreshold;
a
termed positive coupling. In c tension on the membrane has increased beyond
passage. tems superimposed on each other. The
entire
the
across
vibrate
longer
no
can
great
the
membrane
that
is
so
tension
the
first is the tympanic membrane vibrating
Beyond this thresholil amplitucle decreaseswith frequency, that is, the coupling is negative.

#c' ,

.;

fr*n

u32
LLJ

ca

tLi

LIJ
J
(j

"

all

:)

:- 11
U-

r"j:
."_J

o
-l

II

5
C
z

l<

i1

-2-R

1
I

L!

rr- 2.6 t.-

---*.,-& -,.--.,.......
.

i2
18
T i i l E i f ' , 1L1L I S E C o N
DS)

POSITIVE COUPLING of amplitude and frequency is shorvn in

t h i s a n u l y s i so f * p h r a s e s u n g b y T o r v n s e n d ' ss o l i t a i r c . T l e c o l o r e t l

i
o
-** -L*i
{u o i*--ft
r
l\
i

r\

l\

r*****--**^1

r
+--

i
:r- * -*.r.i **{+
- -"."^)
3
rr
I

ji

L-\-

F"\
_ i"rl
v
: i\l
r)i)fl-

O{

O ,
O-

t*-.
(
I

| lv

24

curve traces the amplitude of the phrase; the black curve traces
the
frertrcnr,y. The curvcs risc ard fall together over the time
reriod.

1
t

I
.._*,." .

ul

E.

rr
a

{
I

l-!l

^\

"\4, i

l l " i .J
) f

*--l"il\f1 \ / I
:
i'',i v j

2 ttt
C)

:)

Lil ; l
J
I-L

j l\{'i

10
T i M E( i l r L t s E C o N D S )
NEGATIVE cOUPLING is represented in this analysis of a phrase
sung by a red-rvinged blar:hbird. I{ere rvherthe amplitude (colorerl

at its llatural, or "carrier," frequency.

Tlie second includes the syringr:rlrnuscles, rvhose vibration rvoulcl produce a
periodic change in tensi<nand hence r
change in both frequency arid amplitude. T'lle rnassurdelasticity of the rruscle systern plesurnably lmits its natural
frequency of vibration to lelatir.elv lon'
vnlties-up to, si]-v,500 c.vcle"^
pe. rec,r,rcl.
"Ihe mr"lch krlver mass of the tvrnl>anic
membraneallorvsvibration up to t0 kilocyclesper second.

curue) decreases, the frequen cy ( black) increases. curves also show

the characteristic lnoclulating period of this phra.se ir the aone.

quency range in the tw-ophrasesis, bv a age of 130 cycles per wave. The nodulg:mai'gn,*'ider than it is in rnvotll- iating frerltrencyof the high voice in the
er bird song.It eriencisfrom .75 to 10.7 second subphraseof the "ghg" is about
kilocvcles per srcolld-nearl-1'four oc- 700 cycles per second,higher by a Iarge
tavesl The nrrxinrumfrequencv at 10,7 margin than any other. These perforrnkilocvclesper secondis higher tllan n.hat ances are truly remarkable. lVhat purs'e have found for inv other bird, just pose is served by a "glug" .o-priii,lg
nosingout the 10.5 kilocvclesper second five rvidely different subphrass, to r tt l l e t o p o f t h c l > l a c k p o l l - u ' a r l > l c r . s o n ggc[lrcr
.
w,itlr a "glceee" irrcluding t note
Ilotir I'oicer-are used in the second sub- of negligible duration, two ropid gliuphrase of the "glug,' rnclthe frecluency sancliand a peak frequency of tb.Z t<llospread betu'een the trvo voices (trvo full
cycles per second, only lvladame Cowoctaves) is exceededonly by that of the bird will knorv.
There is not much point in comparing
\fonreretitive modulations can be seen Arnelican bittern.
r \ in the coultship song of the brownThe first note in the "gleeee" is the the impressionsbird songs make n huheaded corvbird lsee illustration on Tsage shortestI hrr.eencountere.It lasts a bit
rran and on avian ears.The differenceis
-l-l]. This bird of unprcpossessingap- lessthan tu,o rnillisecondsand comtrises moot; bird.sdo not sing to us
but to their
pearance and habits is the undisputed a packet of 12 sine u,aves at 6.4 kilo- orvn kind-to seducea wiilng female, for
winner in the decathlon of avian vocali- cycles per second.The glissancloat 50 example, or to warn off a potential
male
ztion,Roger Tory Peterson chrrac)ter- milliseconcis
irr the "gleeee"is onc of the competitor'. We are inclined to put evizes the first phrase in the illustration as most rapid, covering the rar-rgefrom five erything \r,e see or hear into clirr
or,vn
"glng" and the secondphraseas"gleeee." to eight kilocycles per .secor)din four
standard of reference,and s'e wax lyriConsider the following feiltures: The fremiilisecondsnd23 sine \\tves,
iln aver.- cal over the song of a niglttingale or a

O
z ^^
o :.o
O
LIJ

>* u)
()
^- rt
oa
lt

tl

nn
.L

trU

()
J
*
1.8

REPETITIYE MODULATION is shorvn in

a phrase sung by a song sparrow. The fre'
quency of the modulation (curue at hottom)
is fairly constant at some 300 cycles per sec'
ond. The amplitude of the rnodulation' how'
ever (oscillogram at top)t rises and falls
f r o m t h e b e g i n n i n g o f t h e p h r a s et o t h e e n d '

$
EXTERNAL
membrane

EL

(EL), which is located in the bronchial rvall opposite the tympanic

LABIUM
(M), may be the structure that enables the song sparrow and other birds to mod-

ulate amplitude

without

increasing

or decreasing

frequency.

Normally

a change

in one

alters the other, when both are controlled by tension on the tympanic membrane. If, however, the bird can diminish the distance across the bronchial lurnen by extending the labium
ipto it, it u'ould le able to lou'er amrlitude without alfecting freruency, since this means of
changing distance across the passage rvould not afrect tension on the tympanic mernbrane.

t l r r r r s l rs i r n p l y l x , t ' l u s c t l r c y l u t l l c n t o
sing within the ambit of clur own musical
oxperience.The fact is that the rapid
modulations we lrrve been discussing
cannot be perceiveclas suclt by human
eu's.They lre smearecl,as it rvere, to
produce what to us seemslike a note of a
rliflcrcnt rlrrality,or tirnlrlr:,urld onc that
on the whole we find unpleasant. It is
easy for us to resolve a trill or a tremolo
if its frequency is 30 cyclesper secondor
less,but rvhen the frequcncy risesto 100
cyclesper secondor more we hear a note
of a ratl-rer unpleasant buzzy quality.
Hence the br:autiful complexity of the
Lapland longspur song is completely lost
in our ears, whereas it seemsmore than
likely that Madame Longspur finds it delightful and enticing.
I have examined experimentally frequency perception and time perception
for the avirnear. I conclude thrtits frequency discrimination, as expressed in
the relation Af /f X 10-3 (l is of course
frequency), lies between 2 and 5. Time
discrimination appears to be no greater
than .5 millisecond. For human ears frequency discrimination is about the same,
but time discrimination is pelhaps 50 to
100 times worse for humans than it is for
birds. There is then the strong presumption that birds hear as sucl'r the rapid
moclulations so characteristic of tlleir
songs, and that the information content
even in relatively simple songs must be
enorrnous.One readily understandshow
birds of the same species can recognize individuals from subtleties in their
songsthat are imperceptible to a human
]isterrer,
Qo far I have discussed only those
u phlases in bird song for which the
basic wrve form is sinusoidal rvithout
significantharmonic content. As we shall
now see, harmonics do occur in bird
song, but there is no broad generalization relating to the presenceor absence
of harmonics in the songs of the several
bird families. One might say with some
conffdence that songs of the Passeriformes in rvhich hannonics appear are
relatively rare, whereas in the songsand
calls of birds in other families phrases
rvith substantial harmonic content are
comparatively common. Each statement
will, however, have numerous exceptions,
It can nonethelessbe said with considerable assurance that for any given
species there is a threshold frequency
below which harmonics occur and above
which one herrsa phrase rvithout sigrtificant harmonic content. This tlireshold
vrrieswidely for clifl:erentspecies,from
a value near 4,000 cycles per second for

the blue-gray gnatcatcher to below 500

cycles per second for the barrecl owl.
Such threshold frequencies can be determined only rvhen a bird sings over a
ranIe of frequencies that embrace the
threshold, This circumstance is uncommon, becausethe n-rajorityof the Passeriformes sing only in the frequency range
givirig rise to phra.sesfree of harmonics,
ancl birds of other families sing only
in the range giving rise to harmonic
phrases. Indeed, whether or not a bird
passes through the harmonic threshold
in its songs or calls appears to be a matter of clioic; there is no physiological
reason that would prevent a crow from
singing a harmonic-free phrase, or a
r,vood rvarbler, by way of contrast, from
siuging a phrase with substantial harmonic centent.
As an example of the developrirent.of
harmonic spectra as the fundamental descends below the threshold frequency, I
offer a glissando sung by a smooth-billed
ani that embraces nearly three octaves
(frorn 485 to 3,500 cycles per second).
As the frequency rises dominance shifts
from the fourth harmonic through the
thrcl and secondharmonics to the fundamental frequency. The transition frequencies (the frequency at rvhich the
relitiveamplitudes of adjacent hlrmonics reequ*l) are 1,600 cycles per sec*
ond for the fundamental and the second
harmonic, 950 cycles per second for the
second and third harmonics and 560 cycles per second for the third rndfourth
harrnonics. In all the many cases I
have examined in r,vhichthe fundamental plsses through thc threslrokl, curves
of relative amplitude plotted against frequency shorv a small frequency interval embracing each transition betrveen

4
lf
Z

3.5

f )

tlJ

U)
t

tl-l

0_
U)
LU
() 25

o
J

5
L)
Z

15

LLJ

:l

o
tlJ
t
LL

,5

TIME

6
z
U

Ll-'

U)

.
tU

o_

(/)
)
Lll

O
O
a)
J

5
O
Z

LU

NONREPETITIYE MODULATION is seen

in the courtship song of the brown-headed
cowbird. In the oscillogram and frequency
analysis at top is the phrase described as
" g l u g o ' ; i n t h e c o r r e s p o n d i n gr e c o r d s a t b o t tom is the phrase "gleeee.ooBoth of the
phrases cover a large range of frequencies.

:)
o
.
u
Llt

T t M E( N / r L L | S I C O N D S )

harmonics and a larger interval during

which a particular harmonic coutains a
relatively large fraction of the acoustical
energy. I have found no glissandi for
which the dominant harmonic is higher
than the fourth, but there are many calls
in which the fundamental is constant,
and in which e associated harmonic

spectra show similar characteristics, that

is, a clomirant hnrmonic with adjacent
harmonics falling off rapidly in relative
amplitude.
To understand how harmonic spectra
with these characteristics can be generatecl in the syrinx, tret us return once
again to my simplified model of the organ. Imagine the tension in the trmpanic
membrane gradually being reduced,
with the membrane at ts vibrational
peak approaching the opposing bronchial wall (or the external labium) more
and more closely. The point will come
when the bronchial wall will constrain
the membrane, forcing it to depart from
a pure sinusoidal vibration. At this point
the second harmonic will become evident, increasing in amplitude as memblane tension falls and the constraint of
the opposing bronchial wall influences

an increasing percentage of the period of

vibration (tlie period of the fundamental). As the process continues, the amplitude of the fundamental will fall as the
amplitude of the second harmonic rises.
As rnembrane tension decreasesstill further, the second harmonic will become
constrained and the third hamonic wiil
become dominant. At this point the
membrane can be visualized as being in
a state of rippling vibration, with a fundamental fixed by membrane tension
and the associated harmonic spectrurp
dictated by the constraints imposed by
the passagewithin which the membrane
is vibrating.
Among the ducks and geeseharmonic
spectra are found with many terms and
with amplitudes showing no particular
pattern. Such spectramust be associated
with a form of membrane vibration re-

0510152C25
HARMONiC
*AH'as in'oCharlie." At left is the
INDIAN HILL MYNA SAYS
harmonic spectrum of the sound. At right is the wave form of the

sound, rvhich does not decay in amplitude as it would if it lvere

generated by a resonant system such as the human vocal apparatus.

051015202530
HARMONIC
'AH"as in "Charlie."The waveform shorvsa de'
AUTHOR SAYS
o'ahs"

cay in amplitude.

The two

by a nonresonant

mechanism

t2

sound alike, but one is generated

and the orher by a resonant one. This

difierence between the vocal apparatus of birds

and the vocal apparatus of men. The vertical lines in the harmonic
spectra represent the relative intensity of sound at each harmonic.
is a fundamental

sembling the pulses produced in the

huntan glottis, that i.s,a higli-anrplitude
pulse followed by a ser-iesof less violent
ripples of variable period.
Througliout the harmonic domain,
whatever the species and whatever the
wave form or harmonic spectrum may
be, the evidence shows quite clearly that
there is no tracheal modulation: the call
is produced in the syrinx and passesunthrough the trachea.
.changed
come now to those birds that pro\[/:
v v duce a more or less
convincing imitation of the human voice. The qution
is whether such birds bring into play
some completely nerv physiological or
acou,sticalmechanism, or whether they
achieve tlieir imitatiorrs using the same
mechanisms that produce th"eir normal
sorlgs. I have selected the Indian hill
myna as the standard of reference becauseits imitation of the human voice is
excellent; it might well deceive the lis_
tene-rinto thinking a person is speaking.
A phrase was selected from tlie myna
repertoire and was examined in detail in
comparisonwith the samephrase spoken
by me.
The reader will recall that the human
voice originates as a series of puffs of air
ernanating from the glottis, the period
between puffs corresponding to the fundamental frequency (80 to 1-80pufis per
secondfor the male voice). This acoustical disturbance is rnodulated in passing
through our oral cavities so as to reinforce certain frequencies and attenuate
others. Areas of reinforcement are callecl
"formants" and are characteristic of particular vowel sounds. The r.vaveformind
the corresponding harmonic spectrum
for the vowel "a" as enunciated bv the
myna and by me are shown in the illus-

trations below. The sound was unmistakaltly "rh"for both of us, but thc associateclwave forms are entirely diflerent.
When the acoustical disturbance originating at the glottis stimulatesa resonator (the oral cavities) to vibrate at a natural frequency, the genelated wave form
will have the frequency of the resonator,
and the amplitude of the wave will decay
exponentially over a period corresponding to that of the fundamental. One can
leadily isolate the formants for the myna
and me corresponding to the vowel "a"
and see if the resulting wave form fits
thesecriterir.We find that tlie amplitude
for my formants does indeed decay exponentially rvithin the fundamental periocl; for the myna there is no decay and
only random change in arnplitucle. This'
is the most convincing evidence we can
offer that resonators a=renot involved in
the myna "imitation."
How, then, is the imitation producedP
It is important first to realizl that the
myna need not reproduce the human
wave form at all. Since the human ear is
not sensitive to phase, the myna need
produce only an approximation of the
amplitudes of the several harmonics, and
this can be done with literally an infinity
of wave forms. The ability to produce a
separate harmonic spectrum with each
of the two acoustical sources should be
adequate to produce a reasonably good
imitation, particularly since it has been
shown that much of the human voice
spectrum is redundant even if the criterion is recognition of the speech of a particular individual.

modulated by an elastic membrane vibrating in a restricted passage bounded by the walls of the bronhus. This
source-generated acoustical disturbance
appears not to be modiffed in its passage
through the trachea. The syrinx contains
two independently contrcllablesources,
one in each bronchus, enabling the bird
to produce two notes or phrases simultaneously. Harmonics ariie below a
threshold frequency by mechanical constraints on the vibrating membrane, forcing a departure from a purely sinusoid.ai
wave form. The source-generatedsounds
can be modulated in frequency or in amplitude or (more usually) in both with
extraordinarv rapidity, so rapidly that
human ears cannot perceive-the mod_
ulators as such, receiving instead im_
pressions of notes of varying quality or
timbre.
I end this account by pointing out that
it has been in effect a scientiffc etective
story, with conclusions reached by ana_
lyzing the evidence in the bird songs
themselves.The criminal did not confeis
in the last chapter, and the evidence
must remain circumstantial, without di_
rect proof. Had I the deductive powers
of an Albert Campion, a Gideon Fell or
any of the other erudite detectives of ffc_
tion, coupled with the persuasivenessof
Perry N4ason,I might have done better.
any event I have developed a model,
fn
highly convincing to me,
I shall patiently await experimental""d
evidence tirat
will raise my spirits if the answer is yes
but will not be too devastating if it
should be no.

ej me summarize. The physiology and

f
acouscs of bird vocalization are
"
unique in the animal kingdom. Sound is
produced at e syrinx in an air stream

t3