Proc. Natl. Acad. Sci. USA
Vol. 82, pp. 5997-6000, September 1985
Neurobiology

Auditory representation of autogenous song in the song system of
white-crowned sparrows

(auditory neurophysiology/learning and memory/birdsong/sensorimotor integration/auditory feedback)

DANIEL MARGOLIASH* AND MASAKAZU KONISHI
Division of Biology, California Institute of Technology, Pasadena, CA 91125

Contributed by Masakazu Konishi, May 20, 1985

ABSTRACT The HVc (hyperstriatum ventrale, pars
caudale) is a forebrain nucleus in the motor pathway for the
control of song. Neurons in the HVc also exhibit auditory
responses. A subset of these auditory neurons in the white-
crowned sparrow (Zonotrichia leucophrys) have been shown to
be highly selective for the individual bird's own (autogenous)
song. By using multiunit recording techniques to sample from
a large population, we demonstrate that the entire population
of auditory neurons in the HVc is selective for autogenous song.
The selectivity of these neurons must reflect the song-learning
process, for the acoustic parameters of a sparrow's song are
acquired by learning. By testing with laboratory-reared birds,
we show that HVc auditory neurons prefer autogenous song
over the tutor model to which the birds were exposed early in
life. Thus, these neurons must be specified at or after the time
song crystallizes. Since song is learned by reference to auditory
feedback, HVc auditory neurons may guide the development of
the motor program for song. The maintenance of a precise
auditory representation of autogenous song into adulthood can
contribute to the ability to distinguish the fine differences
among conspecific songs.

The song of birds is learned. Young birds memorize a song
model during an impressionable phase early in life (1, 2). The
development of the motor program for song is dependent on
comparison of the individual's vocal output with the previ-
ously established song memory. For birds such as the
white-crowned sparrow (Zonotrichia leucophrys), the "crys-
tallized" adult song is highly stereotyped throughout life, and
the maintenance of adult song does not require vocal feed-
back (3).
The song of the adult white-crowned sparrow contributes

to his reproductive success by influencing mate attraction
and defense of the breeding territory. In the wild, a white-
crowned sparrow typically competes with conspecifics that
sing similar songs from the same dialect (4). Although the
songs of neighbors are similar, white-crowned sparrows can
distinguish among neighbors and between neighbors and
strangers solely on the basis of song (5, 6).

Recently, a projection between the avian auditory telen-
cephalon and the telencephalic nucleus HVc (hyperstriatum
ventrale, pars caudale) has been described (7). The HVc is
part of the motor pathway controlling song production (8, 9),
and neurons in the HVc also exhibit responses to auditory
stimuli (9-12). We report here that auditory neurons in the
HVc are optimally stimulated by the bird's own adult (autog-
enous) song. Therefore, these auditory response properties
reflect specific modification by vocal feedback during devel-
opment.

MATERIALS AND METHODS
Experiments were conducted with four awake birds carrying
chronically implanted electrodes (see below) and with four
birds anesthetized with 20% urethane (Sigma). These and
other aspects of the experimental procedures have been
described (12). Briefly, birds were collected from Bodega
Bay, CA, as nestlings or adults, and housed in individual
sound-attenuation chambers (Industrial Acoustics, Bronx,
NY). Songs were recorded on analog tape, digitized, and
analyzed by zero-crossing techniques (13). The lack of
harmonic structure and short-time limited frequency modu-
lation of the white-crowned sparrow's song permits accurate
recovery of the song from its zero-crossings. All songs in this
report were presented at peak values adjusted to 70 decibels
(dB) (relative to 20 ,uPa).

Clusters of neurons were recorded, to efficiently sample
from a large number of cells (e.g., Fig. 1A). A simple tech-
nique was developed to quantify the strength of multineu-
ronal responses. Beginning with the onset of a song stimulus,
values representing the strength of response of neuronal
activity were calculated for 500 consecutive 10-msec inter-
vals ("bins"). Thus, sampling was continued beyond the end
of a stimulus until the neuronal activity recovered to spon-
taneous levels. For each 10-msec bin the absolute value ofthe
signal (i.e., signal area) was summed across 50 samples
acquired at a sampling rate of 5 kHz. An average for the
response strength for each bin was maintained across all the
representations of a given stimulus. The final baseline value
representing the spontaneous activity was subtracted from
the bin values (see Fig. 1B). The sum of all bin values
throughout the duration of a song was taken as a measure of
overall response strength.
The duration of a song stimulus will affect this and all

measures of neuronal response in a complex way depending
on the relative duration of the song components that do and
do not elicit excitation. For the present data, however,
differences between the overall duration or duration of
corresponding phrases of the bird's own song and test songs
are small compared to differences in response strength.
Except where noted, correction of response strength for
overall duration does not alter the statistical significance of
the results.
For multineuronal recordings, this analysis enjoys several

advantages over counting spikes. These include obviating the
need to arbitrarily choose a threshold level for discrimination
of spikes, and increasing the sensitivity as a result of signal
averaging. The primary disadvantage is that neurons closer to
the electrode and those that produce larger spikes are given

Abbreviation: HVc, hyperstriatum ventrale (pars caudale).
*Present address: Department of Biology, Box 1137, Washington
University, Lindell and Skinker Boulevards, St. Louis, MO 63110.

5997

The publication costs of this article were defrayed in part by page charge
payment. This article must therefore be hereby marked "advertisement"
in accordance with 18 U.S.C. §1734 solely to indicate this fact.

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 



5998 Neurobiology: Margoliash and Konishi

greater weighting; nevertheless, at any one recording site this
bias is constant across stimuli.

RESULTS

Many HVc neuronal clusters responded vigorously to song;
at these recording sites the individual bird's own song proved
to be a most effective stimulus (Fig. 1B). As an initial test of
song selectivity, four white-crowned sparrows were prepared
for acute recordings. Two ofthe birds, R75 and Y73 (Fig. 2A),
sang a song of the Bodega Bay dialect. For these birds, the
responses of neuronal clusters to autogenous song were
compared with the responses to three test songs (W73, G88,
W91; Fig. 2C) also of the Bodega Bay dialect. The test songs
were chosen as a small sample representative of the range of
intradialect variation. The other two birds, Y8 and Y14, had
been reared in the laboratory and exposed during the im-
pressionable phase to a computer-synthesized song that

A Ace r A

w---~~~~~--I-

A-"-AM.-in I

R71
--ar

T T T~~~~,Ijjjjj:~

B

comprised unnatural elements, in particular a trill with
upward-sweeping frequency modulation (14). This tutoring
paradigm results in individual variation in learning (unpub-
lished results), and indeed Y14 formed a good copy of the
song, whereas Y8 failed to learn. For these birds, the
responses to autogenous song were compared with the
responses to the tutor song.
At 47 recording sites throughout the HVc that responded

robustly to song, a random sequence of songs was chosen;
each song was then presented for 20 repetitions. Y73 and Y14
each received a single row of closely spaced (100 Am)
penetrations that covered the rostrocaudal extent ofthe HVc,
whereas R75 and Y8 received more coarsely spaced pene-
trations in several rows that encompassed the entire HVc. At
each of the recording sites, the bird's own song consistently
elicited stronger responses than the test songs (Table 1). No
systematic differences in song selectivity were observed
throughout the HVc. Abnormal components of song never
observed in the wild-i.e., reverse frequency modulation for
Y14 and a series of eight whistles of descending frequency for
Y8-elicited selective responses from HVc neurons in those
birds.
Although the individual bird's song elicited the maximal

response, the test songs also typically elicited excitatory
responses. At 46 of the 47 recording sites, the overall
response to the bird's own song was excitatory; of the 95
presentations of the test songs, 6 resulted in overall inhibi-
tion, while for 4 there was essentially no response. Further-
more, the response profile to the bird's own song (e.g.,
response to whistle and trill but not to buzz) typically was
similar to the response profile to the other songs (Fig. 1B).
Thus, throughout the HVc, conspecific song is an effective
stimulus, and almost always the optimal song is the individ-
ual's own.

Several experiments were conducted to verify that the
selectivity for autogenous song was not capriciously gener-
ated by sampling bias. As one control, the degree of
intradialect song selectivity of HVc neuronal clusters was
investigated further. At two recording sites in each of three
birds (R71, R77, Y85; Fig. 2B), a more stringent test of
selectivity was applied. At these sites, the responses to 10
different songs, all from the Bodega Bay dialect, were
compared with the response to each bird's own song. These
10 songs were quite similar in general morphology both to
each other and to each bird's song; typically these songs

W91

I I . .. A.
I y,

" 'I

W73
-a -Lh A&-.A LA d d_~w*.~.i~AA." A. I .L-LL

FIG. 1. (A) Neuronal activity elicited by five repetitions of the
bird's own song (BOS), for bird R71. The multiunit activity is strong
during the whistle and trill (first and third phases) of song and weak
for the intervening buzz. (B) Response of the neuronal cluster
represented in A to four songs, including BOS (R71). The baseline
represents spontaneous activity. Arrows mark the offset of each
phrase (see Fig. 2 for sonograms). Note that response to BOS is
strongest, whereas the response to other songs, if any, is at the same
phrases as response to BOS. Each response was summed over 50
repetitions of the song (1 song per 12 sec).

Table 1. Comparison between the bird's own song (SOS) and
three intradialect songs for birds R75 and Y73 or between the
BOS and the tutor song for birds Y8 and Y14

No. of recording
sites with stronger

response to

Bird Test song n BOS Test song P*

R75 W73 11 8 (10) 3 (1) 0.113 (0.006)
G88 11 9 2 0,033
W91 11 6 (11) 5 (0) 0.274 (<0.001)

Y73 W73 13 13 0 <0.001
G88 13 13 0 <0.001
W91 13 13 (12) 0 (1) <0.001 (0.002)

Y8 Tutor 14 14 0 <0.001
Y14 Tutor 9 7 2 0.09
Total 95 83 (89) 12 (6) <0.001
For values that changed when corrected for song duration (see

text), corrected numbers are shown in parentheses. Note that R75
sang an unusually short song, hence the effect of correction for song
duration is significant. Each song was repeated 20 times; repetition
rate was 1 per 12 sec.
*Based on one-tailed sign test.

i AIL _ . -,M.lw_F-ly-

.1 RERN AV4%W*"~s-qwrOrr-pr---w

Proc. Nad Acad Sd USA 82 (1985)

I - T" - . - VwwIffl,

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 



Proc. NatL Acad Scd USA 82 (1985) 5999

R75

K

R71

-mimpm \P\\ \ \

TI .._

R77

4 ______

8't.T-

8 ,
Y85

4 -

4 _*__rmmmm \ \
. 4

\

500 1000 1500 2000

Y73

C G88

W73

W91

i t |4-*\mom!

MSEC

FIG. 2. Sonograms (frequency vs. time representations) of songs used to test HVc song-selectivity. (A) Birds R75 and Y73 were tested both
with autogenous song and the songs in C, at multiple sites throughout the HVc (see Table 1). (B) Birds R72, R77, and Y85 were tested with
autogenous song and 10 other songs, including those in C (see Table 2). (C) Three of the test songs. For A, B, and C all songs are derived from
Bodega Bay dialect birds; note that all songs are similar yet vary in slight detail.

comprised a two-part introductory whistle, a buzz of higher
center frequency, and a two-part trill, with an occasional
terminal buzz. The center frequencies and detailed morphol-
ogy for the different phrases in the test songs, however,
spanned the range of variation found within the Bodega Bay
dialect. Nevertheless, the bird's own song was commonly the
most effective song (Table 2). Thus, the neuronal song
selectivity observed in the HVc was not capriciously gener-
ated by use of a limited set of test songs.

As a final control for sampling bias, in four white-crowned
sparrows (R71, R77, Y48, Y62), 11 pairs of glass-coated
platinum/iridium electrodes were chronically implanted so as
to locate the tips (100 ,um) within the HVc. The possibility of
bias in choosing recording sites on the basis of selectivity for
the bird's own song was excluded by implanting the elec-
trodes before the birds were induced to sing and thus before
their songs were known. Locations of recording sites were
selected when responses to the three test songs verified the

Table 2. Strength of response to test songs relative to bird's own song (BOS)
Relative response to

Bird Site G77 G83 G88 044 W73 W76 W78 W91 Y75 Y80 P*

R71 1 0.82 0.26 0.64 0.86 0.59 1.01 0.72 0.61 0.56 0.49 0.011
2 0.75 0.29 0.78 0.28 0.28 1.06 0.27 0.21 0.17 0.16 0.011

R77 1 0.93 0.61 0.81 0.89 1.05 0.45 0.50 0.62 0.53 1.03 0.055
2 1.20 0.24 0.53 1.07 0.67 1.55 0.76 0.36 0.51 0.52 0.172

Y85 1 1.59 0.42 0.80 0.74 0.62 0.84 0.57 0.48 0.68 0.88 0.011
2 1.26 0.49 0.69 0.85 0.91 0.96 0.51 0.27 0.57 1.32 0.055

For six recording sites in three birds, pair-wise comparisons of the efficacy of each BOS with 10
intradialect test songs. Values <1.0 indicate that the test song elicited a weaker response than the BOS.
For R71 andR77, 25 repetitions per song; for Y85, 20 repetitions per song. One song per 12 sec in all
cases.
*Based on one-tailed sign test.

B

N
I

Neurobiology: Margohash and Konishi

looks 4w mw.
-..- to !I

v N.

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1
 



6000 Neurobiology: Margoliash and Konishi

existence of robust auditory responses to song. Chronic
differential recordings from pairs of electrodes (which reject
noise common to both inputs, such as body-movement-
induced artifactual signals from awake but restrained birds)
were conducted on subsequent days.
The birds came into full song within 1 to 3 weeks after

subcutaneous administration of testosterone. Three of the
birds, two males and a female, developed normal song,
whereas a second female sang an abnormal song, as is
occasionally observed for wild-caught females. While the
birds were developing song, a daily variability in response
strength to the test songs was noted, suggesting fluctuation
either in electrode position or in neuronal responses. Without
exception, however, after song crystallized and testing with
the bird's own song commenced, all three test songs elicited
weaker responses at all of the recording sites in all of the
birds-a total of 60 pair-wise comparisons. Two birds sur-
vived 77 and 97 days after onset of singing, respectively. The
neuronal clusters recorded at all three of the electrode pairs
still functioning were optimally responsive to the individual's
song.
Numerous tests were conducted to ascertain the acoustic

basis for the selectivity for autogenous song. Although a
detailed description is beyond the scope of this report, it was
observed that reversing the temporal sequence of song,
which alters the pattern of time-varying frequency and
amplitude modulation while leaving the overall spectrum
unaltered, systematically reduced the efficacy of the song
stimulus. Other manipulations, including frequency-shifting
and modification of the frequency modulation without affect-
ing amplitude modulation, all reduced the efficacy of
autogenous song (unpublished data). The results show that
HVc auditory neurons are sensitive to the particular acoustic
parameters in autogenous song.

DISCUSSION

The results demonstrate that adult HVc auditory neurons
reflect a developmental plasticity that is molded by the
individual's own song. Auditory neurons in the HVc are
explicitly modified by the vocal-feedback experience and
therefore are probably specified at the time of song crystal-
lization. The location and response properties of these
neurons suggest that during song crystallization they are
likely candidates for selecting those motor patterns in the
HVc that produce vocal feedback that matches the song
template. That is, if HVc auditory neurons in the juvenile
have access both to the memory trace of the tutor song and
to vocal feedback, then during song crystallization the
response properties of these neurons may become fixed as
fortuitous patterns of motor activity produce a match be-
tween the template and auditory feedback. In turn, the output

of HVc auditory neurons might tend to stabilize those motor
patterns.
What advantage does an adult white-crowned sparrow

achieve by maintaining an auditory representation of his own
song? Once song is crystallized, a white-crowned sparrow
can maintain normal song for at least 18 months after being
deafened (3). Since song maintenance does not require
auditory feedback and since auditory responses in HVc are
inhibited during singing (9), maintenance is an unlikely
explanation for HVc auditory-response properties. These
neurons do exhibit behaviorally relevant song selectivity,
however. Thus, they are excellent candidates for mediating
song recognition. It has been suggested that a bird mayjudge
the distance of territorial conspecifics by comparing the
transmission-induced frequency-degradation of the songs of
neighbors with a memorized copy of autogenous song
(15-17). If so, the HVc is a candidate site of such memory,
serving as an "autogenous reference" for those aspects of
conspecific song recognition that may require a reference
component. The hypothesis that the song an individual learns
early in life affects his perception of conspecific songs as an
adult is amenable to further behavioral analysis.
We thank Drs. Nobuo Suga, Dale Purves, and James D. Miller for

reviewing an early draft of the manuscript. This work was supported
by a Del E. Webb fellowship (to D.M.) and a grant from the Pew
Memorial Trust.

1. Thorpe, W. H. (1961) Bird Song (Cambridge Univ. Press,
Cambridge, England).

2. Marler, P. (1970) J. Comp. Physiol. Psychol. 71, 1-25.
3. Konishi, M. (1965) Z. Tierpsychol. 22, 770-783.
4. Marler, P. & Tamura, M. (1962) Condor 64, 368-377.
5. Milligan, M. M. & Verner, J. (1971) Condor 73, 208-213.
6. Baker, M. C., Thompson, D. B. & Sherman, G. L. (1981)

Condor 83, 265-267.
7. Kelley, D. B. & Nottebohm, F. (1979) J. Comp. Neurol. 183,

455-470.
8. Nottebohm, F., Stokes, T. M. & Leonard, C. M. (1976) J.

Comp. Neurol. 165, 457-486.
9. McCasland, J. & Konishi, M. (1981) Proc. Natl. Acad. Sci.

USA 78, 7815-7819.
10. Katz, L. C. & Gurney, M. E. (1981) Brain Res. 211, 192-197.
11. Paton, J. A. & Nottebohm, F. (1984) Science 225, 1046-1048.
12. Margoliash, D. (1983) J. Neurosci. 3, 1039-1057.
13. Margoliash, D. (1983) Dissertation (California Institute of

Technology, Pasadena, CA).
14. Konishi, M. (1978) in Perception and Experience, eds. Walk,

R. D. & Pick, H. L., Jr. (Plenum, New York), pp. 105-118.
15. Richards, D. G. (1981) Auk 98, 127-133.
16. Morton, E. S. (1982) in Acoustic Communication in Birds, eds.

Kroodsma, D. E. & Miller, E. H. (Academic, New York),
Vol. 1, pp. 183-212.

17. McGregor, P. K., Krebs, J. R. & Ratcliffe, L. M. (1983) Auk
100, 898-906.

Proc. Nad Acad Sd USA 82 (1985)

D
o
w

n
lo

a
d
e
d
 a

t 
C

a
rn

e
g
ie

 M
e
llo

n
 U

n
iv

e
rs

ity
 o

n
 A

p
ri
l 5

, 
2
0
2
1