Pattern and Time Specificity in Vocal Responses of Blackbirds Turdus merula L.

Author(s): Jochen Wolffgramm and Dietmar Todt

Source: Behaviour, Vol. 81, No. 2/4 (1982), pp. 264-286
Published by: Brill
Stable URL: http://www.jstor.org/stable/4534207
Accessed: 04-01-2016 08:18 UTC

REFERENCES
Linked references are available on JSTOR for this article:
http://www.jstor.org/stable/4534207?seq=1&cid=pdf-reference#references_tab_contents

You may need to log in to JSTOR to access the linked references.

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact [email protected].

Brill is collaborating with JSTOR to digitize, preserve and extend access to Behaviour.

http://www.jstor.org

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY IN VOCAL RESPONSES
OF BLACKBIRDS TURDUS MERULA L.

by

JOCHEN WOLFFGRAMM and DIETMAR TODT')

(Inst. f. Physiologie, Universitatsklinikum, Gesamthochschule Essen; Inst. f. Allg.
Zoologie, Freie Universitat Berlin; Germany)

(With 9 Figuires)
(Ac<. 6-\V1-1982)

Introduction

Playback experiments provide prospective means for studies on the vocal

behaviour of birds. Widely, they have been used in investigations con-
cerning species recognition and territoriality (MARLER, 1956;
ARMSTRONG, 1963; THOMPSON, 1969, 1972; SCHUBERT, 1971; EMLEN,
1972; CATCHPOLE, 1973; BREMOND, 1977; JENKINS, 1977; FALLS, 1978;
KREBS et al., 1978; KROODSMA & VERNER, 1978; THIELCKE et al., 1978;
HULTSCH, 1980; KREBS & KROODSMA, 1980; SLATER, 1981; PETRINOVIC et
al., 1981). Furthermore this sort of experiments supplements data on the
variability of song organisation (HINDE, 1958; ISAAC & MARLER, 1963;
LEMON, 1968; TODT, 1968), thereby supporting attempts to elucidate
both mechanisms of sensory-motor coordination and rules of vocal com-
munication (TODT, 1971, 1975, 1981; WOLFFGRAMM, 1975, 1979, 1980;
TODT & WOLFFGRAMM, 1975; THIMM, 1976, 1980; WHITNEY, 1981).
Among birds that perform temporally segregated songs, the most wide
spread form of vocal responses between territorial neighbours is vocal
matching (equivalent vocal response). When matching the song
( = Strophe) of a neighbour, a bird utters that particular song of its own
repertoire which is most similar to the heard one (pattern specificity).
European blackbirds (Turdus merula) basicly match the initial part of a
heard song. It has been shown for this species that territorial males also
match playback copies of their own songs this way (TODT, 1970, 1974).
Matching replies occur within distinct latencies with probabilities higher

') We thank H. HULTSCHand M. L. HUNTER for their comments on an earlier

draft of the manuscript; their advice was greatly appreciated. The study was supported by
a grant (To 13/9) of the Deutsche Forschungsgemeinschaft.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 265

than random (time specificity), even when the auditory stimuli are
presented over several hours (TODT, 1981).
In order to investigate relationships between pattern and time specifici-
ty of responses, territorial blackbirds were confronted with playback pro-
grams containing their own songs and modifications of these. Modifica-
tions concerned the timing and the sequence of elements, and were
within the range of variations occurring in blackbird song. By these ex-
periments we wanted to find out how such alterations would affect both
the pattern specificity and the time specificity of the test birds' responses.

Methods
I. Terminology.
Song (in German: Strophe): sequence of elements ( = notes). Pauses within songs &lt; 0.5 s;
mean pause duration between successive songs: 4 s. Blackbirds' songs consist of two sec-
tions: the initial "motif section" (elements between 1-5 kHz, and the terminal "twitter
section" (elements between 1-10 kHz).
Song Class (in German: Strophenklasse): in blackbirds: cluster of all the songs starting
with the same element type. The repertoires of the individuals tested during this study
contained each about 10-25 song classes.
Patternspecificity:Correlation between the pattern types (in blackbirds the song classes)
of stimulus and response. The most frequent kind of pattern specificity is vocal matching.
Time specificity: Correlation between the timing of patterns, e.g. the onset timing of
stimulus and response without regard to their pattern types.

II. Experimental procedure.

The experiments were performed using four free ranging male blackbirds (Turdus merula
L.), two of them with territories in Freiburg (Germany), two of them in Berlin. These
birds were selected after pretesting of their repertoires and their song organisation with
regard to variability in song succession and sufficient relative frequency of defined song
classes.
At the start of the birds' singing seasons (middle of April 1976 and 1977) about 200-500
spontaneous songs of each individual were recorded, sound spectrographed (Kay electric
6061B) and classified in the manner described above (10-25 song classes per individual).
Control recordings taken during the entire singing seasons showed only insignificant
changes in the structure and relative use of song classes. In particular, there were no
changes that could be ascribed to our acoustical stimulations.
The playback-experiments were performed in the field using tapes with experimentally
arranged song composition. The compositions contained both modified and unmodified
songs derived from the particular individuals' own repertoires. For each individual two or
three frequently used song classes were chosen for testing. For each of the selected classes
we took a typical song which was later played back to the individual in modified and un-
modified form. Unmodified songs were played back three times more frequently than any
modification. The pauses between the playback songs simulated fast normal singing
(4 + 1 s). In order to prevent habituation, song succession was varied and songs of other
classes were inserted. The sequence of the songs was determined by randomizing. The
stimulus tapes contained in total from fifty to sixty songs (15-20 of them being songs of
other song classes). The playback was reproduced continuously for hours not only when
the bird was singing but also in the pauses between the song bouts. The loudspeaker was

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
266 JOCHEN WOLFFGRAMM & DIETMAR TODT

positioned within a bird's territory at a distance of several meters from the blackbird's
preferred song post. The following modifications were made (Fig. 1):
1. Modification of the pause between the first and the second element.
The pauses were either shortened or lengthened (shortening up to 0.15 s, elongated up
to 0.6 s).
2. Element erasure.
a. Erasure of the first element.
b. Erasure of the second element (with and without a shortening of the interval
between the first and the remaining third element).
c. Erasure of all elements following the second one.
d. Erasure of all elements following the first one.
3. Modification on the second element.
a. Substitution of the second element by an element selected from another
individual's repertoire.
b. Substitution of the second element by a series of three first elements of other song
classes selected from its own repertoire.
4. Insertion of an additional element.
a. An additional element was inserted at 100 ms or 600 ms before the start of the
original song.
b. Insertion of a short element between the first and the second element.
5. Doubling of initial song sections.
a. Doubling of the first element.
b. Doubling of the first two or first three elements.
All the modificationswere made using tape recorders(Uher stereo report 4200 IC and
Uher stereo report 4400) and recordingat one eighth of the original speed. They were
controlled and measured using a sonagraph.
III. Data analvsis.
A detailed analysis was made on recordings from two birds of three consecutive days. In
this period the males uttered 2125 songs (or 1664 songs, respectively) when exposed to
2704 playback songs (or 2450, respectively). Data of the two remainder birds (400 or 800
songs, respectively) were taken for control analyses. The recordingswere transcribedus-
ing a multi-channel measuring set (Meloscript; WOLF,1977) as continuous temporal
registrations of sound pressure level and frequency. In combination to monitoring the
records, these meloscript diagrams allowed the identificationof song classes. In any case
of uncertainty sonagrams were made. The diagrams permitted labeling of the temporal
start of every song with a precision of+ 10 ms.
The sequential and temporal relationships of stimulus songs and reply songs were
analysed using a computerized program for the examination of behavioural time series
("PROVED"; WOI.FFGRAMM& THIMM, 1976). In order to separate random results from
non-random regularities, expected values were calculated using a suitable null-hypothesis
and then compared to the observed data. The preferencefactor y (THIMM et al., 1974) was
used to quantify the deviation of the observed value (o) from the expected value (e):
o-e

The probability of error p in rejecting the null-hypothesis could be determined using

X2-tests.
In investigatingpatternspecificitythe expected values were generated assuming statistical
independence of song classes in the stimulus and response. The expected relative frequen-
cies ei for the S-R transition i-j were calculated from the transition matrix: ei = fI . fj,

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 267

Original play back song

sound 8 i
frequency 6
in kHz 4-

time ins

Modifications Examples

Erasure of elements i

Substitution of elements ^ t 11 it '

Addition of elements

Doubling of elements

Modifications of pauses J

Fig. 1. Sonagram examples of song modifications broadcasted to one of the experimental

blackbirds (individual B).

where fi and f are the relative frequencies of the classes i and j in stimulus or response,
respectively. if one has reason to assume that several transitions are functionally
equivalent it is possible to group them into a transition cluster group and to calculate a
generalized preference factor using the sum of their observed values and the sum of their
expected values (cluster group preference technique; THIMMet al., 1974). In this paper the
cluster of matching replies was considered. This preference can be differentiated by addi-

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
268 JOCHEN WOLFFGRAMM & DIETMAR TODT

tionally regarding the latency between stimulus and reply. The underlying null-
hypothesis does not include temporal randomness of the vocal response but transitional
independence at any latency.
For the study of temporalspecificitynull-hypothesis is that the latency between stimulus
and response is random. Here, no assumptions are made concerning the classes of either
stimuli or replies. Thus, no overlap or connection between the results of the two pro-
cedures described above can occur. Calculation of expected latency distributionscan be
done using different methods:
1. When the latencies between two subsequent events are random, they describe an ex-
the decrement of which can be estimated statistically (WOLFFGRAMM,
ponentialdistribution
1975).
2. The expected distribution can be estimated by means of computerized simulation.For
this purpose, the series of stimulus songs is displayed by chance against the series of reply
songs (THIMM, 1976).
3. When the null-hypothesisis extended to include all the following responses up to the
next stimulation, the latencies are expected to from a uniformdistribution.If the intervals
between stimulus patterns are known, then it is possible to calculate exactly the expected
frequencies.

The first two procedures have some disadvantages. Procedure (1) can only be used if
there are no internal regularitiesneither within the stimulus nor within the reply pattern
series. Procedure (2) avoids this prerequisite but the resulting distribution is subject to
statisticalvariations. Therefore, we applied procedure(3). In order to attain a higher tem-
poral resolution at short latencies we formed logarithmic time classes.

Results
Pattern specificity of response.
The birds preferentially replied with a song that was of the same class as
the broadcast one (matching response). Thus, the null-hypothesis that
the song classes uttered by the bird were statistically independent from
the preceding stimulus song was rejected. In unmodified playback songs
the preference factor for matching responses was y = + 0.79 (bird A) and
y= + 0.67 (bird B) respectively (Table 1). This meant that matching
song replies were 79 % and 67 % more frequent than statistically expected
(p<0.001). Modified stimulus songs also elicited matching responses
(p< 0.001). In both modified and unmodified stimulations there were no
remarkable differences in the responses among the three selected song
classes. We therefore grouped all matching responses into a "response
cluster group" (see: Methods) and preference factors were calculated for
the sums of observed and expected frequencies. The following sections
summarize the results concerning pattern specific responses for every
type of song modification.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 269

TABLE 1. Response matrix of individual A on unmodified stimulus songs

(song classes A, B and C), modified songs (A', B' and C') and songs
of other classes

Song class Song class A B C Other song E

of the of the classes
stimulus response
A o= 21 10 4 53 88
e= 10.0 13.1 5.2 59.7
y= + 1.10 -0.24 -0.23 -0.11
B o= 5 16 6 38 65
e= 7.4 9.7 3.8 44.1
y = -0.32 + 0.65 + 0.58 -0.14
C o= 4 8 8 57 77
e= 8.8 11.5 4.5 52.3
y = -0.55 -0.30 + 0.78 + 0.09
A' o= 80 43 24 297 444
e= 50.6 66 26.1 301.3
y = + 0.58 -0.35 -0.08 -0.01
B' o= 31 117 11 264 423
e= 48.2 62.9 24.9 287.0
y= -0.36 + 0.86 -0.56 -0.08
C' o= 41 45 36 307 429
e= 48.9 63.8 25.2 291.1
y= -0.16 -0.29 + 0.43 + 0.05
Other song classes o= 60 78 36 425 599
e= 68.2 89.1 35.2 406.5
y= -0.12 -0.12 + 0.02 + 0.05
E 242 317 125 1441 2125

A', B' and C' include all the modifications. o = observed frequency, e = expected frequen-
cy, y = preference factor. Significant preferences are underlined (p<0.05).

I. Alterations of the pause between the first and the second

element.
The relative frequency of matching song responses increased when the
pause between the first two elements was shortened (Fig. 2). Thus, such
a copy could be considered as a "supernormal" one. When the pause
was elongated, a decrease of preference was found which leveled off at
pause lengths of about 500 ms; past this, there was no further decline. It
seems likely that the birds handled such pauses as being ends of the
songs.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
270 JOCHEN WOLFFGRAMM & DIETMAR TODT

original
'i
^Y pause
Duration

+1

20 A^^^ \Lindividual A

0
o | individual B
o0

-0.5-
-0.1 0 0.1 0.2 0.3 0.4 0.5

pause elongation in s
Fig. 2. Preferences of song matching at different pauses between the first and the second
element within the play back copies (individuals A and B).

II. Element erasure.

The first element of a song ("initial element")had primary importance in
the evocation of matching responses. When the initial element was erased
the rest of the stimulus song had no influence on song class choice. Eras-
ing the second element or erasing the rest of the song following the second
element did not diminuish matching responses significantly (Fig. 3). In
one individual erasure of the second element lead to a slight decrease in
the preference factor; in other birds, a small increase was found. This ef-
fect did not depend on the interval between the first and the third remain-
ing element. We think that these differences among the individuals might
be due to different lengths of the initial elements. Following earlier ex-
periments (TODT, 1974), the initial element of a song class was played
back (erasure of the rest). We found a definite decrease in response
preference but matching responses were still statistically significant
(y = + 0.42, p<0.005).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 271

modifications

't 2 element
original song erased

rest of the
+1^
-
1 A song erased

' erased
^o- &lt; |^1 '-l^

* +0.5

h..

0-

A i 1.element
mo erased

-0.5

Fig. 3. Preferences of song matching at different kinds of element erasure within the play-
back copies (individual A).

III. Substitution of the second element.

The modifications described above had shown that the second element of
the stimulus song could be erased without any strong decreases in
matching response. This result might suggest that it would be unimpor-
tant which element type took its place, but this conclusion was wrong.
When we replaced the second element by another one which did not
derive from the individual's repertoire but was taken from another con-
specific preference of matching response decreased significantly to about
0.05; Fig. 4). In other modifications the se-
half of the original value (p &lt;
cond element was replaced by a series of two or four initial elements of
other song classes. These songs were not answered by matching
responses; rather the preference factor was significantly below zero
(p &lt;
0.05; Fig. 4). Hence matchingresponseswere avoidedstatistically.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 272 JOCHEN WOLFFGRAMM & DIETMAR TODT

modifications

7' original song

_l[| A 9i
p inter-element

+1 E '"^ ---

= =- - ^ pre-element

- +0.5-- 2. element
'a \ \ ^| 2* ^ 1 replaced

'!i O_iMR
\;^- &lt;_( pre-element

-0.5,-
Iob
O1
_
- series of
initial elements

Fig. 4. Preferences of song matching at different kinds of element replacement and ele-
ment addition within the play back copies (individual A).

IV. Addition of elements.

We placed pre-elements before the beginning of a stimulus song; these
elements were rather long (200-300 ms) and derived from the bird's own
repertoire. The pause after the pre-element was either short (about 150
ms) or long (600 ms). The effect of this song modification on matching
responses depended on pause duration (Fig. 4). When the pause was
short we found no matching response (y = 0). When the pause was long
stimulus songs were answered like the original ones. Obviously such
pauses were taken as inter-song intervals (see: pause alteration).
Additional elements were also interposed between the first and the
second element. These inter-elements were short (50 ms) and derived
from the individual's repertoire. The interval between the first two
original elements was not lengthened by this procedure. These songs
elicited matching responses with a slightly increased preference factor.
This increase might be due to the shortening of the silent interval.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 273

V. Doubling of elements and song sections.

The effect of doubling within the stimulus song depended decisively on
the number of elements that were multiplied. When the starting element
alone was doubled the preference factor of matching responses sank to
almost -1 (Fig. 5a). This statistical avoidance was similar to the effect of a
starting element series (see: Fig. 4) and was highly significant
(p<0.001). This result concerned only two of the three selected song
classes in individual A. In the third class, the starting element was
originally doubled by the bird itself. These repetitions occurred in spon-
taneous songs with a probability of about 50%. The original playback
song consisted of two such elements. We tested 0, 1, 2 and 3 repetitions of
the initial elements. Unlike results with the other song classes no
matching avoidance was found. Instead, the preference factor increased
with increasing number of repetitions (Fig. 5b).
When the first two or first three elements were doubled the preference
factor increased strongly; i.e. such stimuli were answered better than the
original songs (p <0.01 and p <0.02, resp.) and thus could be regarded
as "supernormal" copies (Fig. 5a).

Temporal relationships between stimulus and response

I. Time specificity of song replies.
Time specificity suggests that there are distinct latencies between
stimulus and response. To distinguish such an effect from pattern
specificity we registered all the songs uttered after the stimulus without
respect to their song classes. The observed latencies were compared to
values calculated on the base of randomness. Earlier results have shown
that in blackbirds the start of a reply was temporally correlated to the
onset of the stimulus (TODT, 1974). Therefore, we first tested the tem-
poral relationship with respect to the beginning of the stimulus song.
Two ranges of preferred latencies were found that were likely to represent
response times: 0.1-0.15 s and 0.25-0.8 s (Fig. 6). These values cor-
responded well to those earlier described for two types of latency intervals
in vocal reaction of blackbirds (TODT, 1970).

II. Temporal dependence of pattern specificity.

For studying temporal changes in pattern specificity the null-hypothesis
assumed that the relative frequency of matching song replies was equal at
any latency between stimulus and response. In this analysis the effect of

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
274 JOCHEN WOLFFGRAMM & DIETMAR TODT

modifications

doubled

original song
+2
/
C

5)

0+1-
0.)

0-

.
i.^ . f_ 1. element
doubled

Y,
3 song B of individual A

o 2
5)

C.

a ~0-
1

/" t It'
original
song

0 1 2 3
number of repetitions

Fig. 5. Above: Preferences of song matching at different kinds of element multiplication

within the play-back copies (individual A, except for song class B). Below: Preferences of
song matching for song class B of individual A at different numbers of repetition of the
initial element within the play-back copies. In this song class, the bird originally doubled
the initial element.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 275

250

0)200 observed
O'
, 150-
100-

50-

2 0.050.1 Q2 0.4 0.8 1.6 32 64128 256 512 512

o expected
A 200-
150-

100-

50-

L.

0.050.1 02 04 0.8 1.6 32 6412 256 512

latency in s

Fig. 6. Time specificity of song responses. Above: histogram of the intervals between the
beginning of a bird's song and the start of the preceding stimulus song. Center:histogram
of expected intervals calculated on the base of temporal randomness (see: Methods).
Below: preferences of the temporal onset of the response, calculated from the histograms
above
above (individual
(individual A).
A).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
276 JOCHEN WOLFFGRAMM & DIETMAR TODT

time specificity was neglected. The results showed a wide range of

positive preference (Fig. 7). Matching responses were significantly more
frequent than expected not only at latencies of 0.25 s-0.8 s, (one range of
temporal specificity) but up to values of about 6 s. The null-hypothesis
could be rejected at p<0.02. We therefore concluded that beside a
"rapid" reaction - 0.25 s to 0.8 s after the start of the stimulus - there
existed another type of response that could be called a "late" one. In late

I matching
responses

X 200-

100-

L
0
0.050.1 02 0.4 0.8 16 3.2 64 12B
o v
_0
3- /

2-

0-01 ,

0.050.1 0.2 04 0.8 1.6 32 6.4 12.8

latency in s
Fig. 7. Pattern specificity of song responses. Above: histogram of latencies between
stimulus and song response discriminating matching and non-matching responses. Below:
preferences of song matching at different latencies, calculated from the relative part of
matching responses (individual A).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 277

replies we found no temporal specificity i.e. latencies did not occur more
often than expected. The timing of these songs did not follow the timing
of the stimulus but comprised normal pauses between the songs. At these
latencies matching responses were preferred, too. In contrast, rapid
response at A t= 0.1 s to 0.15 s (formerly termed unspecific response;
TODT, 1970); showed only temporal specificity but not pattern specificity.
The results did not answer the question whether rapid and late
matching responses pointed to different mechanisms of auditory pattern
processing. For this reason we counted both types of replies separately
and calculated the preference factors for matching song responses by dif-
ferentiating four groups of stimulus songs,(Fig. 8): (a) original songs and
the following modified songs, (b) subnormal copies, (c) normal copies,
(d) supernormal copies.

rapid late
response response
supernormal a ----
3
normal b
2.5

original C -- --- . \2\

1.5

subnormal d --
0.5

'"3. 0-

-05
-1
Fig. 8. Preferences of song matching in stimulations with different classes of modifications
(supernormal, normal, subnormal copies) and original songs within rapid (latency
L<0.8s) and late (L>0.8s) matching responses (individual A).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
278 JOCHEN WOLFFGRAMM & DIETMAR TODT

In general, preference factors were higher in rapid responses than in

late ones but there was a correlation between both types of replies. The
higher the y-factor was in rapid responses, the more also late matching
was preferred.

III. Time specificity of rapid matching responses.

Since matching responses require a preceding recognition of element
types, we examined the time specificity of matching responses in respect
to the ends of the stimulus elements. The procedure was the same as
described in section I except that only matching replies were considered.
A short time (0.1 s-0.3 s) after the end of the first, second or third ele-
ment of the stimulus song, the preference of response was strongly in-
creased (Fig. 9). The height of the preference peak decreased gradually
from the first to the third element but remained statistically significant
(p <0.01). This result seemed unusual because the third element was not
important for the pattern specificity of responses and the second element
had only a subordinate role.
The experimental modifications on the original pattern were per-
formed mainly on the second elements of the stimulus songs. We
therefore examined time specificity of latencies from the end of the second
element to the start of the matching response in different modifications of
the original song (Fig. 9b). All these preference diagrams were rather in-
dependent of the modification. In contrast when the first element was
omitted or when matching was inhibited in another way any kind of tem-
poral coordination with the stimulus was suppressed (Fig. 9b).

Discussion
In our experiments the relationships between a series of modified and un-
modified stimulus songs and the song utterances of the test birds have
been investigated. The modifications concerned only a few selected song
classes but they formed vocal patterns that might have been sung by a
neighbouring conspecific. For instance, the repertoires of blackbirds can
comprise songs that differ only slightly in temporal parameters and com-
binations of elements. Furthermore singing birds can terminate a song
before its normal end. It is still not answered whether a bird being con-
fronted with its own songs or with modified copies of own songs
discriminates these patterns from the songs of neighbouring conspecifics.
Our experiments confirmed a high tendency of the blackbirds to respond
vocally to such stimuli. The different types of responses delivered hints for

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 279

Y-
0 *m M...
3-
I _ .... _. ...
2-
~' 4

"2-' \
1. o
r\ \6\
-1 I I I I I
\
.
y- .I ~JV ...
.1

y a

i i...
2- A iA 2- 1- -
co 1-

L.
Q) -1 . g a,
o,
-1 , . .i I .I I I I

-/'v \ ...

-1 I.. . . . a 0.05 0.1 0.2 0.4 02 1.6 32 6.4 12.8

0.05 0.1 0.2 0.4 08 1.6 32 6.4 12.8

latency in s
Fig. 9. Left: Temporal preferences of the latencies within matching responses. The
latencies were measured from the end of the first (above), second (center), and third
(below) element of the play back song. Right: Temporal preference of latencies (measured
from the end of the second element) within matching responses at different kinds of'
song modifications (above: original copy; center: alteration of the pause and replacement
of the second element, respectively; below: erasure of the first element) (individual A).

processing mechanisms in the experimental birds. However, we are not

able to discriminate details of the processing, since a bird's responses
reflect the complete result of the processing procedure which is addi-
tionally influenced by the bird's intrinsic program of vocalization.

I. Signal processing.
In many thrush species (blackbirds, nightingales, redstarts, wood
thrushes, robins and robin chats) there is evidence that conspecific song
stimulus might modify a bird's vocal program in regard to pattern choice
and pattern performance (TODT, 1970, 1971, 1975, 1981; WOLFFGRAMM,
1973, 1975, 1980; THIMM, 1973, 1976, 1980; HULTSCH, 1980). Modifica-
tions might last for a short moment (some seconds), or for a longer time.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
280 JOCHEN WOLFFGRAMM & DIETMAR TODT

Considering short term modifications, relationships between a stimulus

song and a reply song can be described by two kinds of relationships, in
this paper called pattern specificity and time specificity. Depending on
which of both is present one can distinguish three types of vocal
responses:
Response Time specificity Pattern specificity

type I +
type II + +
type III +

Our analysis showed that in blackbirds being exposed to conspecific

songs all these response types did occur. In contrast to type III, type I
and II responses were described in an earlier study using other methods
(TODT, 1970). Studies concerning other song bird species dealt however,
mainly with type II responses.
Within a response being equivalent to a normal choice reaction the
stimulus should elicit the response and at the same time determine the
response pattern. In contrast to type I and III, type II responses show a
correlation between reply triggering and song class choice. However, the
first two elements contributed to these two processes in different ways.
Although the third element had no measurable effect on the pattern
specificity of the response it could trigger the timing of the reply.
However, this effect did only occur if the first element was present. If it
was lacking or masked by a pre-element, any triggering was suppressed.
Thus, we suggest that recognition of song initials is a basic requirement
for the release of the matching reply (Fig. 9) but that after the ending of
the discriminative process, the timing of the response is influenced by ad-
ditional extrinsic and intrinsic factors not necessarily linked to the first
step.
In blackbirds, the existence of the three types of response in our ex-
periments can be interpreted in the following way:
(I) More frequent than it should be expected on the base of the bird's self-
program it reacted after a latency of only 0.05-0.15 s. Since here existed
no pattern specificity between the stimulus and the response we assume
that the bird uttered that song that had previously been chosen by its self-
program. The effect of the stimulus was only a triggering one.
(II) After a latency of 0.4-0.8 s the birds reacted more frequently than it
should be expected following their self-program by a matching response.
Beside the triggering effect the stimulus here influenced the choice of the

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 281

song class. Since such an effect was not always present but expressed
itself only in a statistical preference, we conclude an interaction of the
auditory input and the intrinsic components of song control (e.g.
periodicities, inhibitions, etc.) thus facilitating a decision in favour of a
matching song utterance.
(III) After latencies of some seconds, when no time specificity was found
the birds still responded by matching more frequently than it could be ex-
plained only by regarding their self-program. Here, the timing of the
response seemed not to be triggered by the stimulus. The individual took
the stimulus pattern into respect when determining the song class of its
utterance but it performed the response when it would have usually
started to sing.

II. Functional aspects.

Rapid responses of type I and especially of type II are usual not only
under experimental conditions but also during normal countersinging.
Since nearly all blackbirds songs are longer than 1 s, the responding in-
dividual, here, jams the song of its partner. It has been shown that the
jammed bird shortens its singing, and avoids song posts being connected
to such replies (TODT, 1981). Responses that can be assigned to type I
and type II have also been found in duetting thrush species where not
only neighbouring males but also paired individuals were com-
municating. In contrast to countersinging, the responding mate does not
match the song of its partner, but selects a particular antiphonal duet
motif. Even in different species, vocal reaction times correspond well to
the latencies measured for type II responses (0.2-0.8 s; TODT, 1970, 1971;
THORPE, 1972). Once the motifs have been selected, temporal coordina-
tion between the dueting mates can further be obtained by a continuous
control following type I responses (Temporal triggering including laten-
cies of about 0.1 s; TODT, 1970; TODT et al., 1981).
There are still few examples describing the communicative use of type
III responses (e.g. in redstart song: THIMM, 1976; THIMM & WOLFF-
GRAMM, 1979). In blackbirds such responses lead to "late matching"which
predominantly can be observed apart from the "dawn chorus" and near
to the end of the singing season. In contrast to "rapid matching" which
has been interpreted as "vocal threatening" late matching may be con-
sidered as "vocal greeting" (TODT, 1981). In general, type III responses
may contribute to alternating singing of neighbours. There is evidence
that in nightingales particular territory holders (obviously, subordinate

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
282 JOCHEN WOLFFGRAMM & DIETMAR TODT

birds) establish alternating singing by referring their replies to the end of a

neighbour's song (inserting response; HULTSCH,1981; HULTSCH&TODT,
1982). Inserting is particularly precise in bird species which - like
nightingales, and differently than e.g. blackbirds - vocalize songs with a
loud and distinct terminal element type (TODT, 1971; HULTSCH,1980).
Provided that jamming the songs of a rival is an appropriate strategy in
song competitions, the key role of a song's initial element seems to be
very efficient, for the following reasons:
1. The reply can be affected much faster than it would be possible if
furrer elements had to be taken into account.
.Due to the dichotomic pattern of element succession within a song,
initial element types occur much more frequently than any other element
types. Thus, the chance to hear a known element in the song of a rival
and to match it is rather high (TODT, 1970, 1981).
3. Comparing the element repertoires of neighbouring blackbirds one
finds much more correspondences in the initial part of the songs than in
the middle or the end (HALL-GRAGGS, 1962; TODT, 1968; WOLFFGRAMM,
1979). Thus, song matching concerning only the initial elements of a
song can widely be used in interactions between conspecifics. There are
indications that by this means other individuals can be discriminated,
whose repertoires correspond more or less to the own one. As a result,
distinct preferred territorial distances between neighbours with a high
amount of repertoire sharing are affected (WOLFFGRAMM, 1979; HULTSCH
& TODT, 1981; WOLFFGRAMM, in prep.).
4. A discriminative mechanism regarding mainly initial elements
should allow to reduce interferences in situations when several
neighbouring conspecifics are singing together, and it should be more ap-
propriate during multi-directional interactions than a mechanism requir-
ing mutual recognition of complete songs.

According to their importance in the' coding of own songs and the

decoding of alien, songs initial elements should be protected in a special
way. We suppose that the inhibition of matching observed when initial
elements follow directly one after the other (Fig. 4, 5) is caused by such a
protection mechanism thus preventing interferences that might affect
song class choice (see: 4 above).
Apart from the succession of element types their pause durations in-
fluence the frequency of song matching. At times of increased vocal ac-
tivity, many species of song birds fasten their singing by shortening the
silent pauses between the vocal patterns (TODT, 1970; WOLFFGRAMM,

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 283

1975; TODT & HULTSCH, 1981). We assume that the test birds responded
to playback songs comprising shortened pauses like to songs of an excited
conspecific. Thus, the response attitude of a blackbird is not only
modified by discrete patterns but also by variable parameters. Further
experiments are planned that should additionally concern changes of
these parameters, including e.g. a varied sound pressure level.
The song functions discussed here consider mainly interactions among
communicating males, but left influences of, and on female conspecifics
out of consideration. Further experiments are necessary to find out
whether females assess a heard song in a different way than males do. In
particular, the element types following the initial elements should be
regarded here.

Summary
Male blackbirds (Turdus merula)were confronted with modified and unmodified playback
copies of their own songs presented within the birds normal territories (Fig. 1). The vocal
responses of the birds were recorded and correlated to the playback songs in respect to pat-
tern specificity (song matching) and time specificity (distinct latencies). The following
results were obtained:
1. The initial elementof a song played a key role. If it was erased or masked by a pre-
element the birds did not match the playback copies. In contrast erasure of the second ele-
ment or the rest of the song following the second element did not affect song matching
(Figs 3, 4).
2. The secondelementhad an additional effect. The matching preference increased with a
shortening of the pause between the first two elements, it decreased when the second ele-
ment was substituted by an alien element (Figs 2, 3, 4).
3. A series of initial elements (identical or different element types) led to an inhibition
of song matching. In contrast, copies including a doubling of the first two or three
elements had a supernormal effect (Figs 4, 5).
4. Three types of song responsescould be discriminated: in type I responses (latency:
L = 0.1 s) the birds did not match the stimulus; in type II responses (L= 0.25-0.8 s)
song matching was preferred; type III responses (L> 1 s) showed no time specificity
towards onset of playback songs, but still a preference of song matching (Figs 6, 7).
5. In general, late matchingresponses (type III) were less preferred than rapidones (type
II) but the effects of different song modifications remained the same.
6. In type II responses, the temporal triggeringof the reply and the recognitionof the
heard song class were not completely linked. In contrast to the neglectable influence of the
third elements they were well able to trigger a reply (Fig. 9).

References

ARMSTRONG, E. A. (1963). A study of bird song. - Oxford Univ. Press.

BREMOND, J. C. (1977). Acoustic competition between the song of the wren (Troglodytes
- Behaviour 65,
troglodytes)and the song of other species. p. 89-97.
CATCHPOLE, C. K. (1973). The functions of advertising song in the sedge warbler and
reed warbler. - Behaviour 46, p. 300-320.
EMLEN, S. T. (1972). An experimental analysis of the parameters of bird song eliciting
species recognition. - Behaviour 41, p. 130-171.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 284 JOCHEN WOLFFGRAMM & DIETMAR TODT

FALLS, J. B. (1978). Bird song and territorial behaviour. - Advances in the study of
communication and affect. - Plenum Press, New York 4, p. 61-89.
HALL-CRAGGS, J. (1962). The development of song in the blackbird Turdus merula. -
Ibis 104, p. 277-300.
HINDE, R. A. (1958). Alternative motor patterns in chaffinch song. - Anim. Behav. 6,
p. 211-218.
HULTSCH, H. (1980). Beziehungen zwischen Struktur, zeitlicher Variabilitat und
sozialem Einsatz des Gesanges der Nachtigall (Luscinia megarhynchosB.). - Disser-
tation, FB Biologie; Freie Universitat Berlin.
--(1981). Acceleration and delay of vocal responses to auditory stimuli, a contribution
in functions of alternating singing. - Verh. Dtsch. Zool. Ges. 1981, p. 343.
-- & TODT, D. (1981). Repertoire sharing and song post distance in nightingales
(Luscinia megarhynchosB.). - Behav. Ecol. Sociobiol. 8, p. 183-188.
-- &-- (1982). Temporal performance roles during vocal interactions in nightingales
(Luscinia megarhynchos).- Behav. Ecol. Sociobiol., in press.
ISAAC,.D. & MARLER,P. (1963). Ordering of sequences of singing behaviour of mistle
thrushes in relationship to timing. - Anim. Behav. 11, p. 179-188.
JENKINS, P. (1977). Cultural transmission of song patterns and dialect development in a
free living bird population. - Anim. Behav. 25, p. 305-399.
KREBS, J. R., ASHCROFT,R. & WEBBER, M. (1978). Song repertoires and territory
defense in the great tit (Parus major L.). - Nature 271, p. 539-542.
-- & KROODSMA, D. E. (1980). Repertoires and geographical variation in bird song. -
Advances in the study of behaviour 11, p. 143-177.
KROODSMA,D. E. & VERNER, J. (1978). Complex singing behaviour among Cistophorus
wrens. - Auk 95, p. 703-716.
--
(1979). Vocal dueling among male marsh wrens: Evidence for ritualized expressions
of dominance/subordinance. - Auk 96, p. 506-515.
LEMON, R. E. (1968). The relation between organization and function of song in car-
dinals. - Behaviour 32, p. 158-178.
MARLER, P. (1956). Behaviour of the chaffinch (Fringilla coelebs). - Behav. Suppl. 5,
p. 1-184.
PETRINOVIC, L., PATTERSON, T. L. & BAPTISTA,F. L. (1981). Song dialects as barriers to
dispersal: A re-evaluation. - Evolution 35, p. 180-188.
SCHUBERT,M. (1971). Untersuchungen fiber die reaktionsauslosenden Signalstrukturen
des Fitis-Gesangs und das Verhalten gegenuber arteigenen Rufen. - Behaviour 38,
p. 250-288.
SLATER,P. J. B. (1981). Chaffinch song repertoires: observations experiments and a
discussion of their significance. - Z. Tierpsychol. 56, p. 1-24.
THIELCKE,G., WLiSTENBERG, K. & BECKER,P. H. (1978). Reaktion von Zilpzalp und
Fitis (Phylloscopus collybita,Ph. throchilus)auf verschiedene Gesangsformen des Zilp-
zalps. -J. Orn. 119, p. 313-326.
THIMM, F. (1973). Sequentielle und zeitliche Beziehungen im Reviergesang des Garten-
rotschwanzes (Phoenicurusphoenicurus).-J. comp. Physiol. 84, p. 311-334.
-- (1976). Zeitliche und qualitative Organisation der Lautmusterfolge im Reviergesang
des Gartenrotschwanzes. - Diss. Univ. Freiburg/Brsg.
- (1978). Steuerung des zeitlichen Ablaufs im Gartenrotschwanzgesang (Phoenicurus
phoenicurus).- Verh. Dtsch. Zool. Ges. 1978, p. 186.
- (1980). The function of feed-back mechanism in bird song. - Acta XVII.
Congr.
Int. Orn., Berlin 1978, p. 677-681.
--, CLAUSEN, A., TODT, D. & WOLFFGRAMM, J. (1974). Zeitabhangigkeit von Verhal-
tensmusterfolgen. -J. comp. Physiol. 93, p. 55-84.
- & WOLFFGRAMM, J. (1979)..Erkennung und vokale Beantwortung von Lautmustern
durch Singvogel. - Verh. Dtsch. Zool. Ges. 1979, p. 220.

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
 PATTERN AND TIME SPECIFICITY OF RESPONSES 285

THOMPSON,W. L. (1969). Song recognition by territorial male buntings (Passerina). -

Anim. Behav. 17, p. 658-663.
-- -
(1972). Singing behaviour of the indigo bunting, Passerinacyanea. Z. Tierpsychol.
31, p. 39-59.
THORPE, W. H. (1972). Duetting and antiphonal song in birds: its extent and sig-
nificance. - Behav. monogr. suppl. 18, p. 1-305.
TODT, D. (1968). Zur Steuerung unregelmiafiger Verhaltensablaufe. - In: Kybernetic
1968 (MITTELSTEADT, H. ed.) p. 465-485.
- (1970). Gesangliche Reaktion der Amsel auf ihren experimentell reproduzierten
- Z.
Eigengesang. vergl. Physiol. 66, p. 294-317.
- (1971). Aquivalente und konvalente gesangliche Reaktionen einer extrem regel-
matiig singenden Nachtigall (Luscinia megarhynchosB.). - Z. vergl. Physiol. 71,
p. 262-285.
- (1974). Zur Bedeutung der "richtigen" Syntax auditiver Muster fiur deren vokale
Beantwortung durch Amseln. - Z. Naturforsch. 29c, p. 157-160.
- (1975). Short term inhibition of outputs occurring in the vocal behaviour of black-
birds. -J. comp. Physiol. 98, p. 289-306.
- (1981). On functions of vocal matching: Effect of counter replies on song post choice
and singing. - Z. Tierpsychol. 57, p. 73-93.
- & HULTSCH,H. (1980). Functional aspects of sequence and hierarchy in bird song.
-Acta XVII. Congr. Int. Orn., Berlin 1978, p. 663-670.
- , &- DUVALL,F. (1981). Behavioural significance and social function of vocal and
non-vocal displays in the monogamous duet-singer Cossypha heuglini. - Zool.
Beitrage (in press).
-- & WOLFFGRAMM, J. (1975). Uberprufungen von Steuerungssystemen zur Strophen-
wahl der Amsel durch digitale Stimulierung. - Biol. Cybernetics 17, p. 109-127.
WHITNEY, C. L. (1981). Patterns of singing in the varied thrush: II a model of control.
Z. Tierpsychol. 57, p. 141-162.
WOLFFGRAMM,J. (1973). Lautmustersequenzen und periodische Beziehungen im Roller-
Kanari-Gesang (Serinus canaria). -J. comp. Physiol. 85, p. 65-88.
- (1975). Aufbau, Organisation und zentrale Steuerung variabler Lautmusterfolgen
im Gesang des Roller-Kanari (Serinus canaria L.). - Univ. Freiburg Dissertation.
- (1978). Vokale Reaktionen der Amsel auf kunstlich abgewandelte Strophen ihres
eigenen Gesangs. - Verh. Dtsch. Zool. Ges. 1978, p. 187.
- (1979). Vergleich der hochvariablen Gesangvortrage benachbarter Amseln.
Verh. Dtsch. Zool. Ges. 1979, p. 237.
- (1980). The role of periodicities in avian vocal communication. - Acta XVII.
Congr. Int. Orn., Berlin 1978, p. 671-676.
- & THIMM, F. (1976). Bearbeitergesteuerte Analyse von Verhaltenszeitreihen am
Digitalrechner. - Biol. Cybernetics 21, p. 61-78.

Zusammenfassung
Amselmannchen (Turdus merula) wurden im Feldexperiment modifizierte und unveran-
derte Strophenattrappen ihres eigenen Gesangs vorgespielt (Fig. 1). Die gesanglichen
Antworten wurden im Bezug auf Musterspezifitat (gleichklassige gesangliche Antworten)
und Zeitspezifitat (Auftreten bevorzugter Latenzen) analysiert. Dabei ergaben sich fol-
gende Resultate:
1. Das Anfangselementeiner Strophe hatte eine Schliisselfunktion. Fehlte es oder war es
durch ein Vor-Element maskiert, so antworteten die Testvogel nicht gleichklassig. Dem-
gegeniiber hatte die Loschung des zweiten Elements oder des Strophenrests nach dem
zweiten Element keinen negativen Einfluss auf die gleichklassige gesangliche Antwort
(Fig. 3, 4).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions
286 JOCHEN WOLFFGRAMM & DIETMAR TODT

2. Das zweite Elementhatte einen zusatzlichen Einfluss. Bei einer Verkirzung der Pause
zwischen den beiden ersten Elementen der Vorspielstrophe stieg die Antwortpraferenz
an, beim Ersatz des zweiten Elements durch ein Fremdelement eines anderen Vogels
sank sie dagegen ab (Fig. 2, 3, 4).
3. Eine direkte Folge von Anfangselementenim Vorspiel verhinderte gleichklassige Ant-
worten. Dagegen bewirkten Vorspielstrophen mit einer Verdopplung der ersten zwei oder
drei Elemente eine erh6hte Antwortpraferenz (Fig. 4, 5).
4. Drei Antworttypenwaren zu unterscheiden: Bei Typ I (Latenz: 0.1 s) waren die Ant-
worten nicht bevorzugt gleichklassig; bei Typ II (L = 0.4-0.8 s) waren gleichklassige Ant-
worten bevorzugt; bei Typ III (L> 1 s) war keine Zeitspezifitat zwischen Reizbeginn und
Einsatz der Antworten mehr nachweisbar, dennoch blieb eine Bevorzugung fur gleich-
klassige gesangliche Antworten erhalten (Fig. 6, 7).
5. Generell waren spate gleichklassigeAntworten (Typ III) weniger stark bevorzugt als
schnelle(Typ II). Die Effekte der verschiedenen Strophenabwandlungen waren allerdings
in beiden Fallen gleichartig (Fig. 8).
6. Bei Antworten nach Typ II waren die Erkennungder gehorten Strophenklasse und
die Auslosungder Antwortstrophe nichtvollstandig miteinander verknupt. Im Gegensatz zu
ihrer vernachlassigbaren Wirkung auf die Gleichklassigkeit der Antworten wirkten die
dritten Elemente an der zeitlichen Ausl6sung mit (Fig. 9).

This content downloaded from 192.122.237.41 on Mon, 04 Jan 2016 08:18:57 UTC
All use subject to JSTOR Terms and Conditions