THE ROLE OF PITCH CUES IN SONG RECOGNITION BY

CAROLINA CHICKADEES (PARUS CAROLINENSIS)

by

BERNARD LOHR 1), RONALD WEISMAN 2) and STEPHEN NOWICKI 3)

(Department of Zoology, Duke University, Durham, NC 27708)

(With 3 Figures) (Acc. 15-VI-1994)

Summary

Carolina chickadees sing songs characterized by obvious frequency changes between adjacent notes. Several songbird species have been suggested to use the pitch interval cues available from frequency ratios in their song for song recognition. We evaluated whether Carolina chickadees use pitch interval cues in song recognition using playback of song in the field to territorial males as a response assay. In contrast to previous studies, we found a similar response to songs with altered frequency ratios as to those with normal ratios, suggesting that pitch interval cues are not important for song recognition in Carolina chickadees. We also demonstrate that Carolina chickadees do not distinguish between recordings of naturally produced songs and songs synthesized using measured parameters of natural songs, validating our approach to experimental manipulation of song parame- ters. Our results indicate that different songbird species may be using different mechanisms for recognizing conspecifics.

Introduction

Pitch and pitch change, the perceptual correlates of frequency and fre-

quency change (ROEDERER, 1979), are known to contribute features

important for the recognition of conspecific song in many species of

songbird (e.g. DABELSTEEN & PEDERSEN, 1985; WEARY et al., 1986; NELSON,

1988). It is less clear which mechanisms of pitch perception songbirds use

to identify these cues. Absolute pitch perception is the capacity to identify the frequency of a note without comparing it to an external standard

1) Correspondence concerning this article may be addressed to Bernard LOHR, Department of Zoology, Duke University, Box 90325, Durham, NC, 27708-0325. 2) Department of Psychology, Queen's University, Kingston, Ontario, K7L 3N6 Canada. 3) This research was funded by PHS grant DC00402 and an Alfred P. Sloan Fellowship to Stephen NOWICKI and an NSERC Canada to Ronald WEISMAN. We thank Kristin CALVERT, Alicia MAYNARD and Jeffrey PODOS for research assistance, and Stewart H. HULSE, Donald E. KROODSMA, Douglas A. NELSON, Laurene RATCLIFFE and an anonymous reviewer for com- ments on earlier drafts.

2

(WARD & BURNS, 1982). Relative pitch refers to the capacity to identify directional or interval changes between the frequencies of two or more

notes, as in a melody (DEUTSCH, 1982). The pitch interval between two

notes corresponds mathematically to the ratio formed by dividing the

frequency of one note by the other (WARD & BURNS, 1982). Initial work on the use of pitch cues in song recognition focused on the

importance of absolute pitch (e.g. FALLS, 1963; BREMOND, 1968). Recent

laboratory work has shown that songbirds may use both relative and

absolute cues in pitch perception, although absolute cues may constrain

the use of relative ones (HULSE & CYNX, 1985). There is now evidence

from field studies suggesting that songbirds may use relative pitch cues

available from the frequency changes in song. In white-throated sparrows

(Zonotrichia albicollis) and veeries (Catharus fuscescens), song playback experi- ments indicate that pitch interval cues are important in species song

recognition (HURLY et al., 1990; WEARY et al., 1991 ). In black-capped chickadees (Parus atricapillus) production data also

provide evidence for the use of relative frequency mechanisms in song. WEISMAN et al. (1990) found the relative relationships among song note

frequencies to be much less variable than absolute cues. Furthermore,

they found that when individuals shift the overall frequency of their song, birds maintain the same frequency ratio between adjacent notes (RAT- CLIFFE & WEISMAN, 1985; HILL & LEIN, 1987; WEISMAN et al., 1990). The

constancy of relative features in production, it was argued, suggests their

potential importance for song recognition in this species. Carolina chickadees and black-capped chickadees are closely-related

congeners (BRAUN & ROBBINS, 1986; GILL et al., 1989) and their species-

typical songs share many features (SMITH, 1972; FICKEN et al., 1978;

HAILMAN, 1989). Both are composed of simple pure-tone notes with little

or no modulation, alternating between high and low frequencies. Because

of their close phylogenetic relationship, it is of interest to compare pro- cesses of song production and recognition in the two species. LOHR et al.

(1991) found that Carolina chickadees produce songs with relative cues

that are equally as variable as absolute cues, that the frequencies of notes

are not highly predictable from one another, and that common frequency ratios are not maintained across different songtypes. Thus, unlike the

data for black-capped chickadees, frequency relationships in Carolina

chickadee songs do not implicate the use of interval perception in song

3

recognition. On the other hand, neither do these production data rule out

the use of either absolute or pitch interval cues for recognition. We here test the importance of frequency change in song recognition

by Carolina chickadees. We first establish whether responses to synthe- sized Carolina chickadee songs are detectably different from responses

given to recordings of natural song (Experiment 1). We next measure the

response of birds to field playback of songs whose frequency relationships have been manipulated either by increasing or decreasing the frequency

ratio, thus altering the pitch interval cues, to values outside the normal

species range (Experiment 2). Results from both experiments also allow us

to evaluate the use of absolute pitch cues in species song recognition in

Carolina chickadees and compare mechanisms of frequency perception to those of black-capped chickadees.

Methods

Subjects and testing conditions.

Playback tests were conducted at study sites within 10 km of Duke University, Durham NC. Subjects were territorial male Carolina chickadees who sang the type 'A' songs described by LOHR et al. (1991). Birds holding adjacent territories were color-marked for identifica- tion ; unmarked subjects had territories at least I km distant from the nearest other test bird. We tested 12 birds in Experiment I during April 1992 and 24 birds in Experiment 2 during April and early May 1991. Each bird was used only once within an experiment. Adjacent neighbors were not tested on the same day.

Tests were conducted between 0700 and 1500 hrs EST. Songs were broadcast near the center of a bird's territory using a SONY TCM 5000EV and an AR 'Powered Partner' speaker-amplifier on a platform about I m off the ground. Song amplitude during playback was normalized to 95 dB RMS at I m.

Song synthesis.

Playback songs (e.g. Fig. 1) were synthesized digitally using 'SIGNAL' digital signal process- ing software (BEEMAN, 1990) operating on a 486/33 microprocessor with a Data Transla- tion DT-2821 I/O board, at a sample rate of 25 kHz. Song notes of desired frequencies, durations and amplitude profiles were produced using the digital voltage-controlled oscilla- tor and amplifier functions, in 'SIGNAL'.

Playback procedures.

We used a one-speaker, multiple-trial playback test procedure. A playback test began with 4 min of silence, followed in sequence by two or three song trials (depending on experi- ment). A trial consisted of 2 min of song, at 10 songs/min, followed by 4 min of silence. Vocal responses of the subject and dictation of behavioral observations were recorded throughout each trial. To minimize possible bias, playback tapes were randomly assigned, and observers were blind to the order and type of playback presentation in each trial. Our observations included (1) the number of songs sung by the test subject, (2) latency to sing, (3) time spent within 20 m of the speaker, (4) closest approach to the speaker, (5) mean

4

Fig. 1. Example of a synthetic playback song (Type A of Lohr et al., 1991), illustrating the simple pure-tonal nature of song notes. Sonagram produced on a Kay Elemetric DSP 5500

Sona-Graph (0-8 kHz analysis range, 59 Hz frequency resolution).

approach to the speaker, and (6) number of perch changes. Data for the last two of these measures were not collected during Experiment 2.

Playback experiments.

Experiment 1. In order to determine whether synthetic songs are adequate for eliciting normal territorial behavioral responses in Carolina chickadees, we compared responses to recordings of natural songs with synthetic conspecific songs and natural heterospecific songs. Six record- ings of song type A, each from a different bird, were used as models for synthetic songs. Two of these (with the lowest frequency low notes) were classified as 'low' frequency songs, two (low notes at mid frequency values) as 'medium' frequency songs, and two (with the highest frequency low notes) as 'high' frequency songs. Note durations, inter-note intervals, and amplitude envelopes of synthetic songs were constructed to match exactly the parame- ters of the corresponding natural song in each playback trial.

Each natural song and its corresponding synthetic version were used in two playback trials. In addition, one song from each of 4 different Carolina wrens (Thyrothorus ludovicianus) was used as a heterospecific control in each of the playbacks. Carolina wrens are common in the study area and have songs that cover approximately the same time and frequency ranges as Carolina chickadee songs. We played one normal, one synthetic, and one heterospecific song to each bird. Birds were assigned, in the order of testing, to a ran- domized counterbalanced list of song frequencies (low, medium, and high) and song orders (natural, synthetic, and heterospecific).

Experiment 2. To examine the role of the pitch interval in song recognition, we contrasted normal playback songs, having frequency ratios close to the population mean, with altered songs, having frequency ratios well outside normal population parameters. We designed these test stimuli so that songs with different ratios shared common frequencies to control for absolute pitch cues. Four synthetic versions of Carolina chickadee song type A were constructed for playback. Note durations, inter-note intervals, and amplitude envelopes for synthetic songs were mean values measured from the A type songs of 19 individuals (LOHR et al., 1991). Two songs had normal frequency ratios and relatively common note frequen- cies : the low (3429 Hz) and high (3789 Hz) frequency normal songs (Table 1, left side). Holding the frequency of Notes 2, 3 & 5 (the low notes) constant across normal and altered songs, we also synthesized two songs with altered ratios (3429 Hz and 3789 Hz altered songs, Table 1, right side) by substituting Notes 1 & 4 (the high notes) from the other

5

TABLE 1. Note frequency and ratio values of the normal and altered

playback songs in Experiment 2

Also shown are SD units from Mean Song Type A, values for each note and ratio. The frequencies shown in bold were held constant in normal and altered songs.

normal song. Our tests examine adjacent intervals in song. While birds may be calculating other frequency ratios, between Notes 1 & 4, or Notes I & 5 for instance, subsequent studies indicate that this is unlikely (LOHR, unpubl. data). All songs were digitally synthesized de novo and differed only in the particular parameters being tested, so concerns about differen- tial responsiveness due to undetected song idiosyncracies (KROODSMA, 1989) are minimized in the present design (MCGR?GOR et al., 1992).

We played one normal and one altered song to each bird. Equal numbers of birds heard songs with 3429 Hz and 3789 Hz low notes. Birds were assigned, in the order of testing, to a randomized counterbalanced list of song frequencies (3429 Hz vs. 3789 Hz low notes) and song orders (normal vs. altered song first).

Statistical analyses.

We used 3-way mixed ANOVAs with 2 between-subjects factors and I within-subjects factor to evaluate the results. Between subjects factors included low-note frequency and order for both experiments. Within-subjects factors included song type (natural, synthetic, and heterospecific) for Experiment 1, and frequency ratio (normal and altered) for Experi- ment 2. We then used principal components factor analysis with varimax rotation, aggre- gating the response measures into factors for further analysis, in order to increase the power and clarity of comparisons among songs (see e.g. YASUKAWA, 1978). This multivariate technique allows us to combine a number of original response measures into a reduced number of composite measures, or factors, that describe the variation in response to test songs in a more comprehensive manner. We first describe the aggregation of responses into

6

factors and then report analyses of variance performed on the resultant aggregate factor scores.

Two methods for calculating factor scores are in general use. In one procedure, salient responses are transformed to standard scores and then weighted equally (i.e. multiplied by +1 or -I depending on the sign of the loading), then simply summed in the factor score; nonsalient responses are excluded. In an alternative procedure, salient and nonsalient responses are transformed to standard scores, then weighted differentially according to their loadings on the factor. A disadvantage of this latter procedure is that nonsalient variables tend to contribute disproportionately to the error variance of the factor score (WACKWITZ & HORN, 1971; JACKSON & CHAN, 1980). Given that none of the nonsalient response loadings appeared at all ambiguous (i.e. greater than 0.40), and that the purpose of obtaining factor scores was to develop more stable and unbiased estimates of song potency, we calculated factor scores according to the first method, which weights salient responses equally.

Inter-observer agreement.

Playback tests can contain differences in frequency that are occasionally audible to observers in the tests. We therefore recorded results from 36 playback trials in Experiments I with two independent observers, one of which was naive regarding the experimental predictions, in order to establish the objectivity of behavioral response measures. We used Pearson correlations to examine the degree of inter-observer reliability in assessing terri- torial responses to playbacks.

Results

Inter-observer agreement.

Pearson correlations between two observers for 36 playback trials in

Experiment I ranged from .95 to .99 for the 6 response measures, with no

significant differences between the observers for any response (ts(35) <_ 1.54, NS).

Analysis of individual response measures.

Experiment 1.

Natural and simulated playback songs elicited singing and approach to

the speaker whereas heterospecific playback song did not 22.63,

p, < .O1). Natural and simulated songs did not elicit significantly different

numbers of perch changes or closer approach to the speaker 3.22,

NS). Also, although natural playback songs elicited more songs, simulated

songs elicited singing more promptly 5.77, p? < .05).

High and medium frequency playback songs elicited more perch

changes, increased time within 20 m of the speaker, closer approach, and

closer mean approach to the speaker than low frequency songs

5.50, p, < .05). Song latency and number of songs failed to vary signifi-

cantly with the frequency of the playback song 2.45, NS).

7

Experiment 2.

The frequency ratio of the playback song did not have a significant effect

on any response measure 1.75, NS). Lower frequency playback

songs elicited song more promptly and frequently than high frequency

songs 5.98, p, < .05) but the birds did not approach significantly more closely or spend significantly more time within 20 m of the speaker

during low than high frequency playback songs (F, < 1, NS).

Aggregation into factor scores.

In Experiments I and 2, a principal components factor analysis was used

to examine playback responses in a more comprehensive manner. Inde-

pendent principal components factor analyses with varimax rotation

aggregated response measures into the same two factors, 'approach' and

'song', each with an eigenvalue greater than I (Tables 2 and 3). The

factors had low inter-factor correlations (rs >_ .12) and together accounted

for 77%-83% of the variance in each experiment. Responses with load-

TABLE 2. Loadings for a varimax factor analysis of the response measures

in Experiment I

The loadings shown in bold constitute the factors.

TABLE 3. Loadings for a varimax factor analysis of the response measures

in Experiment 2

The loadings shown in bold constitute the factors.

8

ings equal to or greater than 0.60 (shown in bold) on a factor were judged salient to that factor, whereas those with loadings less than 0.60 were

judged nonsalient.

Analysis of factor scores.

Experiment 1.

In Experiment 1, there were no significant differences in song factor

scores between natural and simulated songs or among playback song

frequencies (Fig. 2a, F, < l, NS). Since birds remained over 20 m from the

speaker and sang infrequently when presented with heterospecific song,

Fig. 2. Song (a) and approach (b) scores for playback with natural and synthetic conspecific song (filled and open bars, respectively) at 3 different low note frequencies in Experiment 1 (N = 12). Factor scores do not differ significantly between natural and synthetic versions for any case. Approach factor scores for both natural and synthetic low frequency songs are signifi-

cantly lower than those for medium and high frequency songs.

9

Fig. 3. Song (a) and approach (b) scores for playback songs with normal and altered frequency ratios (filled and open bars, respectively) at 2 low note frequencies in Experiment 2 (N = 24). Factor scores do not differ significantly between normal and altered songs for any case. Song factor scores for both normal and altered ratio songs with low notes at 3429 Hz are

significantly higher than those for songs with low notes at 3789 Hz.

we omitted heterospecific songs from the analysis of factor scores. Anal-

ysis of approach factor scores found no significant difference between the

effects of natural and simulated songs (Fig. 2b, F(l,9) = 2.34, NS) but

revealed a significant effect of playback song frequency (F(2,9) = 9.95, p <

.01). Lower frequency playback songs tended to elicit weaker approach scores than medium or higher frequency songs (Fig. 2b).

Experiments 2.

In Experiment 2, analysis of song factor scores found no significant effect

of the frequency ratio (F < 1, NS) and a significant effect for song

frequency (F(1,22) = 10.30, p < .01). Playback songs with altered ratios did

10

not generate lower song scores than songs with normal ratios, but play- back songs with 3429 Hz low notes tended to elicit higher song scores

than those with 3789 Hz low notes (Fig. 3).

Discussion

Experiment I was designed primarily to demonstrate that synthetic and

natural song elicit equivalent behavioral responses. For most of our

measures, there is no detectable difference in amount of response given to

recordings of natural as. synthetic song. In the two of six response mea-

sures in which results from raw data scores differ, more salient responses are given to simulated songs in one case (latency to sing) and natural

songs in the other (number of songs). Aggregate factor scores for natural

and synthetic song do not differ in any of the frequency presentations.

Also, both synthetic and natural songs elicit significantly greater

responses than heterospecific songs. This demonstrates that Carolina

chickadees do not functionally differentiate between natural and syn- thetic versions of their species-typical song on a scale of response mea-

sures where there are clear distinctions between responses to different

species' songs. The use of synthetic songs during playback allows a

greater degree of versatility in producing controlled stimuli for examining the extent to which absolute pitch may play a role in perception of song

frequencies under natural conditions.

As demonstrated in Experiment 2, altering frequency ratios between

adjacent notes in Carolina chickadee songs does not affect the responsive- ness of territorial males to playback, even when these ratios are increased

by more than 2.7 SDs or decreased by more than 2.5 SDs as compared to

the population mean (Fig. 3). Taken together with production data from

LoHR et al. (1991), these results demonstrate that pitch interval cues are

not likely to be important to Carolina chickadees for identifying the songs of conspecifics. This conclusion contrasts with that suggested earlier for

black-capped chickadees. For that species, production data (WEISMAN et

al., 1990) were used to argue that pitch interval cues should play a role in

conspecific song recognition. Laboratory playback data also suggest some

role for the use of pitch interval cues in black-capped chickadee song

(WEISMAN & RATCLIFFE, 1989), but evidence from the field is less compel-

ling. RnTCLiFFE and WEISMAN (1992) found that responses to black-

11

capped song were significantly reduced only when frequency intervals

were lowered to 0.99, and then in only one measure (number of flights). Results from field studies on several other species of songbirds, how-

ever, provide more convincing evidence for the use of pitch interval cues

for song recognition. Both white-throated sparrows and veeries, two

unrelated species, have been shown to rely on pitch interval cues in song

production and recognition, based on recordings and song playback

experiments in natural, territorial contexts (HURLY et al., 1990, 1991; WEARY et al., 1991). It is even more surprising, then, if there is a signifi- cant difference between Carolina and black-capped chickadees in their

use of pitch cues in song perception, given the close relationship between

the two species (BRAUN & ROBBINS, 1986; GILL et al., 1989). Possible

explanations for such a difference may depend in part on the relatively

greater size of the intervals in song of Carolina chickadees. The accuracy of pitch interval perception for melodic sequences in humans, for

instance, decreases with increased interval size (DEUTSCH, 1982). Addi-

tionally, the ability to distinguish between two tones in humans becomes

relatively poor at frequencies greater than 6000 Hz (HANDEL, 1989). Whereas pitch interval is not a significant feature for song recognition

in Carolina chickadees, data from both of the experiments reported here

suggest that absolute pitch cues may be important. In Experiment 2,

songs with low notes (Notes 2, 3, & 5 in song type A) that were at

frequencies below the species mean elicited significantly more song

responses than songs with low notes approximately the same number of

SD units above the mean. This experiment was not designed to explore the influence of absolute pitch systematically, yet it is clear that variation

in absolute pitch affects responsiveness to song in playback. In contrast, in

Experiment 1, songs with lower frequency notes overall elicited signifi-

cantly weaker approach measures than songs with medium or higher notes. Since note frequencies, timing relationships, and amplitude pro- files were matched to specific natural songs in Experiment 1, and

approach factor scores differed rather than song factor scores as in

Experiment 2, direct comparisons of absolute frequency effects cannot be

made between the two experiments. A contrast between Carolina chickadees and other species in their use

of song features in the field does not necessarily reflect differences in the

perceptual capabilities of the different species. All may be equally compe-

12

tent at the task of discriminating either relative or absolute pitch cues, even though they may emphasize different forms of pitch processing in

the context of song recognition. Black-capped chickadees, for example, can distinguish small differences in frequency ratio in a psychoacoustic discrimination task (WEARY & WEisMAN, 1991). Comparable psycho- acoustic tests with Carolina chickadees are required to determine

whether differences in the use of perceptual cues are reflected in the

perceptual competencies of the two species. Available evidence, however,

suggests that basic perceptual capabilities are relatively uniform among

songbirds (DOOLING, 1980), so we might expect that a significant differ-

ence will not be found. Furthermore, birds may process perceptual cues

quite differently in operant laboratory discriminations and field situa-

tions, reflecting substantial differences in underlying motivations in these

different conditions (NELSON & MARLER, 1990). If Carolina chickadees and other species do not differ in their percep-

tual capabilities, then what other factors might predict differences in their

use of pitch cues? One long-standing hypothesis suggests that song fea-

tures with minimal intraspecific variation (invariant features) will be

emphasized for discriminating conspecific songs from those of other

species (MARLER, 1960; FALLS, 1963; EMLEN, 1972). In black-capped

chickadees, for example, an invariant features model predicts that this

species should emphasize relative pitch cues, which are much less variable

in song production than are absolute pitch cues (WEISMAN et al., 1990). The potential reversed importance of pitch interval cues and absolute

pitch cues in Carolina chickadee song would not have been so predicted,

however, because frequency ratios and absolute note frequencies in this

species are equally variable (LOHR et al., 1991). EMLEN (1972) originally

argued that invariant cues are not necessarily used in recognition, and

both DABELSTEEN & PEDERSEN (1985) and NELSON (1988) have demon-

strated that variable as well as invariant cues can be important in species

song recognition. Another hypothesis suggests that species use song features for recogni-

tion that are most distinctive in the context of the attributes of the song

community, or sound environment, to which they belong (BRTMOND,

1978; DABELSTEEN & PEDERSEN, 1985; NELSON, 1988; NELSON & MARLER,

1990). W'e do not have data on the bioacoustic environments of Carolina

or black-capped chickadees, but the habitats and communities both live

13

in are diverse within each species' range and overlap considerably between species (BREWER, 1963; JOHNSTON, 1971), suggesting that this

hypothesis also would not have predicted the differences we observe. In

fact, Carolina and black-capped chickadees share a narrow zone of

sympatry that extends from New Jersey to Kansas (BREWER, 1963; Ris-

irrc, 1968; JOHNSTON, 1971; ROBBINS et al., 1986) in which their sound

environments are probably identical. The frequency ratio values between

the two species are quite different (1.70-1.81 for Carolina chickadees

versus 1.13 for black-capped chickadees, see LOHR et al., 1991 and VVEis-

MaN et al., 1990, respectively). Thus, we can discount the possibility that

the differential use of absolute cues and pitch interval cues is the result of

avoiding acoustic competition between the two species. Carolina and black-capped chickadees are reported to interbreed in

regions of sympatry (BREWER, 1963; RISING, 1968; JOHNSTON, 1971; ROBBINS et al., 1986), raising further questions about song recognition mechanisms. To the extent that song plays a role in coordinating repro- ductive activity, we must suppose that one species can respond in an

appropriate fashion to relevant features of the other's songs. This suggests that the two species would share common perceptual biases for the ways

in which they encode species information in song. We find the opposite,

however, in the case of pitch cues used by Carolina and black-capped chickadees. In a cross-fostering study, D.E. KROODSMA (pers. comm.) found that young of either species may be capable of learning features of

the other species' songs, although the acoustic frequency relationships of

such songs have not been examined. Also unknown is the degree to which

early experience might influence perceptual biases for relative and abso-

lute pitch cues. The point of our comparison between Carolina chick-

adees and other species of songbirds is not to state unequivocally which

cues any particular species may use in song recognition. Instead, we

believe that our results suggest that different songbirds, indeed possibly even congeners, may be using different mechanisms to recognize

conspecifics.

References

BEEMAN, K. (1990). 'SIGNAL' user's guide. - Engineering design, Belmont, MA. BRAUN, M,J. & ROBBINS, M.B. (1986). Extensive protein similarity of the hybridizing

chickadees Parus atricapillus and P. carolinensis. - Auk 103, p. 667-675.

14

BRÉMOND, J.-C. (1968). Recherches sur la semantique et les elementes vecteurs d'informa- tion dans les signaux acoustiques du rouge-gorge (Erithacus rubecula). - Terre et Vie 2, p. 109-220. (1978). Acoustic competition between the song of the wren (Troglodytes troglodytes) and the songs of other species. - Behaviour 65, p. 89-98.

BREWER, R. (1963). Ecological and reproductive relationships of black-capped and Caro- lina chickadees. - Auk 80, p. 9-47.

DABELSTEEN, T. & PEDERSEN, S.B. (1985). Correspondence between messages in the full song of the blackbird Turdus merula and meanings to territorial males, as inferred from responses to computerized modifications of natural song. - Z. f. Tierpsychol. 69, p. 149-165.

DEUTSCH, D. (1982). The processing of pitch combinations. - In: The psychology of music (D. Deutsch, ed.). Academic Press, New York, p. 271-316.

DOOLING, R.J. (1980). Behavior and psychophysics of hearing in birds. - In: Comparative studies of hearing in vertebrates (A.N. POPPER & R.R. FAY, eds). Springer-Verlag, Berlin, p. 261-288.

EMLEN, S.T. (1972). An experimental analysis of the parameters of bird song eliciting species recognition. - Behaviour 41, p. 130-171.

FALLS, J.B. (1963). Properties of bird song eliciting response from territorial males. - Proc. XIIIth Intern. Ornithol. Congress, p. 259-271.

FICKEN, M.S., FICKEN, R.W. & WITKIN, S.R. (1978). Vocal repertoire of the black-capped chickadee. - Auk 95, p. 34-48.

GILL, F.B., FUNK, D.H. & SILVERIN, B. 1989. Protein relationships among titmice (Parus). - Wilson Bull. 101, p. 182-197.

HANDEL, S. (1989). Listening: an introduction to the perception of auditory events. - MIT Press, Cambridge, Massachusetts.

HAILMAN, J.P. (1989). The organization of major vocalizations in the Paridae. - Wilson Bull. 101, p. 305-343.

HILL, B.G. & LEIN, M.R. (1987). Function of frequency-shifted songs of black-capped chickadees. - Condor 89, p. 914-915.

HULSE, S.H. & CYNX, J. (1985). Relative pitch perception is constrained by absolute pitch in songbirds (Mimus, Molothrus, and Sturnus). -J. Comp. Psychol. 99, p. 176-196.

HURLY, T.A., RATCLIFFE, L. & WEISMAN, R.G. (1990). Relative pitch recognition in white- throated sparrows (Zonotrichia albicollis). - Anim. Behav. 40, p. 176-181.

- -, WEISMAN, R.G., RATCLIFFE, L. & JOHNSRUDE, I.S. (1991). Absolute and relative pitch production in the songs of the white-throated sparrow (Zonotrichia albicollis). - Bio- acoustics 3, p. 81-91.

JACKSON, D.N. & CHAN, D.W. (1980). Maximum-likelihood estimation in common factor analysis: A cautionary note. - Psychol. Bull. 88, p. 502-508.

JOHNSTON, D.W. (1971). Ecological aspects of hybridizing chickadees (Parus) in Virginia. - Am. Midl. Natur. 85, p. 124-134.

KROODSMA, D.E. (1989). Suggested experimental designs for song playbacks. - Anim. Behav. 37, p. 600-609.

LOHR, B., NOWICKI, S. & WEISMAN, R. (1991). Pitch production in Carolina chickadee songs. - Condor 93, p. 197-199.

MARLER, P. (1960). Bird songs and mate selection. - In: Animal sounds and communica- tion (W.E. LANYON & W.N. TAVOLGA, eds). American Institute of Biological Sciences, Washington DC, Publ. No. 7, p. 348-367.

MCGREGOR, P.K., CATCHPOLE, C.K., DABELSTEEN, T., FALLS, J.B., FUSANI, L., GERHARDT, H.C., GILBERT, F., HORN, A.G., KLUMP, G.M., KROODSMA, D.E., LAMBRECHTS, M.M., MCCOMB, K.E., NELSON, D.A., PEPPERBERG, I.M., RATCLIFFE, L., SEARCY, W.A. & WEARY, D.M. (1992). Design of playback experiments: The Thornbridge Hall NATO consensus. - In: Playback and studies of animal communication (P.K. MCGREGOR, ed.). Plenum Press, New York, p. 1-9.

15

NELSON, D.A. (1988). Feature weighting in species song recognition by the field sparrow (Spizella pusilla). - Behaviour 106, p. 158-182.

- - & MARLER, P. (1990). The perception of birdsong and an ecological concept of signal space. - In: Comparative perception, Volume II: Complex signals (W.C. STEBBINS & M.A. BERKLEY, eds). John Wiley & Sons, New York, p. 443-478.

RATCLIFFE, L. & WEISMAN, R.G. (1985). Frequency shift in the fee bee song of the black- capped chickadee. - Condor 87, p. 555-556.

- - & -- (1992). Pitch processing strategies in birds: A comparison of laboratory and field studies. - In: Playback and studies of animal communication (P.K. MCGREGOR, ed.). Plenum Press, New York, p. 211-223.

RISING, J.D. (1968). A multivariate assessment of interbreeding between the chickadees, Parus atricapillus and P. carolinensis. - Syst. Zool. 17, p. 160-169.

ROBBINS, M.B., BRAUN, M.J. & TOBEY, E.A. (1986). Morphological and vocal variation across a contact zone between the chickadees Parus atricapillus and P. carolinensis. - Auk 103, p. 655-666.

ROEDERER, J.G. (1979). Introduction to the physics and psychophysics of music, 2nd edition. - Springer Verlag, New York.

SMITH, S.T. (1972). Communication and other social behavior in Parus carolinensis. - Publ. Nuttall Ornithol. Club, Cambridge, MA.

WACKWITZ, J.H. & HORN, J.L. (1971). On obtaining the best estimates of factor scores within an ideal simple structure. - Multivar. Behav. Res. 6, p. 389-408.

WARD, W.D. & BURNS, E.M. (1982). Absolute pitch. - In: The psychology of music (D. DEUTSCH, ed.). Academic Press, New York, p. 431-451.

WEARY, D.M., LEMON, R.E. & DATE, E.M. (1986). Acoustic features used in song discrimi- nation by the veery. - Ethology 72, p. 199-203.

- - & WEISMAN, R.G. (1991). Operant discrimination of frequency and frequency ratio in the black-capped chickadee (Parus atricapillus). -J. Comp. Psychol. 105, p. 253-259.

- -, LEMON, R.E., CHIN, T. & MONGRAIN, J. (1991). Use of the relative frequency of notes by veeries in song recognition and production. - Auk 108, p. 977-981.

WEISMAN, R.G. & RATCLIFFE, L. (1989). Absolute and relative pitch processing in black- capped chickadees (Parus atricapillus). - Anim. Behav. 38, p. 685-692. --, JOHNSRUDE, I. & HURLY, T.A. (1990). Absolute and relative pitch production in

the song of the black-capped chickadee. - Condor 92, p. 118-124. YASUKAWA, K. (1978). Aggressive tendencies and levels of a graded display factor analysis

of response to song playback in the redwinged blackbird (Agelaius phoeniceus). - Behav. Biol. 23, p. 446-459.