Variation in the Bout Structure of Northern Mockingbird (Mimus Polyglottus) Singing

Variation in the Bout Structure of Northern Mockingbird (Mimus polyglottus) Singing

NICHOLAS S. THOMPSON,* EMILY ABBEY,* JESSICA WAPNER,* CHERYL LOGAN,† PETER G. MERRITT,‡ AND ALBERT POOTH§

*Departments of Biology and Psychology, Clark University, Worcester, MA †Department of Psychology, University of North Carolina, Greensboro, NC ‡Treasure Coast Regional Planning Council, Stuart, FL §Department of Biology, University of Miami, Miami, FL

The highly variable singing of the northern mockingbird (Mimus polyglottus) is distinguishable from that of other sympatric mimids by its organization into bouts: the bird’s tendency to repeat an element several times before proceeding to another. To determine the degree to which this bout structure is a common feature of mockingbird song, 10 samples of singing from widely different populations and circumstances were examined including examples from coastal and central New England, central North Carolina, and the Florida peninsula and examples of day and night and spring and summer song. Measures included note parameters (peak frequency, internote interval, and note duration) and bout parameters (songs per bout and mean values of note parameters). All samples were found to be organized in bouts, but the degree of differentiation of the bouts (i.e., the degree to which the boundaries between bouts were emphasized by the contrast between their songs) varied both within and between samples. Bout differentiation was not maxi- mized: songs of bouts sung in close temporal proximity were more similar than average, and the performance overall seemed to consist of runs of high and low values of note and/or bout parameters. Whether these variations in the bout structure reflect changes in the state of the singer or in his circumstances or serve to enhance the overall effectiveness of his performance remains to be determined.

Mimidae Mockingbird Communication Song variation Song structure

Among the primary functions of birds’ singing species, the male’s singing is practically invariant. are the communication of the species and individual Performances vary only subtly across the individuals identity of the singer. If singing is to identify sing- of the species, and each singer repeats the same sound ers both as species members and as individuals, then or, at most, sings one or two variants. The species- at least one property of the performance must be specific properties of such a performance are obvi- consistent across the members of the species, and at ous. In other species, the individual males sing hun- least one property must be variable from species dreds, perhaps thousands, of different sounds, and member to species member but consistent within a performances vary strikingly from individual to indi- species member (Marler, 1960). vidual. For species with such highly variable perfor- Birds differ strikingly in the degree of variation in mances, identification of the species-specific proper- their singing (Catchpole & Slater, 1995). In many ties of the performance may be less straightforward.

93 94 THOMPSON ET AL.

One taxonomic group particularly known for the about 100 ms in duration separated in time from variety of its singing is the mimic thrushes. Within neighboring notes by silence. Notes are gathered into this group, three species breed sympatrically and units of repetition called songs. Songs can be classi- during overlapping time periods in the eastern United fied into readily recognized discrete types called States: the gray catbird (Dumetella carolinensis), the song types. During singing, the mockingbird imme- brown thrasher (Toxostoma rufum), and the north- diately repeats songs of one type several times be- ern mockingbird (Mimus polyglottus). All three are fore switching to songs of a different type. These repertoire-singing species with an extraordinary va- sequences of immediate repetition of song types are riety of song elements. Both the catbird and the called “bouts.” Mockingbirds may also produce mockingbird are known to sing hundreds of repeat- songs in temporal clusters called “groups.” Group able vocal elements, and the brown thrasher is boundaries do not necessarily correspond to bout thought to sing thousands (Boughey & Thompson, boundaries (Figs. 1 and 2). 1976; Kroodsma & Parker, 1977). Furthermore, all Although the bout structure of mockingbird song three species incorporate the vocal elements of other is thought to be a species-identifying property, it is species in their performances. not an invariant one. Bout length varies in a mock- The extraordinary variability of mimic thrush song ingbird performance, and its frequency distribution elements raises questions concerning auditory dis- overlaps with the bout length distributions of the crimination among the three species. In the field, sympatric mimids (Boughey & Thompson, 1976, the three species are traditionally distinguished by 1981) (Fig. 2). In the field, the birds also appear to the number of repetitions of song elements: catbirds vary the salience of the bout structure by varying tend to sing each element only once and brown the distinctiveness of the songs that make up neigh- thrashers once or twice, whereas mockingbirds typi- boring bouts and by varying the duration of pauses cally repeat each song element several times before that occur within and between bouts. going on to the next (Mathews, 1904). Every mock- The purpose of this investigation was to examine ingbird performance has at least three levels of or- variations in the bout structure of northern mock- ganization called notes, songs, and bouts (see ingbird song performances taken from a range of Derrickson & Breitwisch, 1992; Howard, 1974; circumstances and locations to see if patterns of Moody, Kedoux, & Thompson, 1994; Wildenthal, variation therein were consistent with a species iden- 1965; for additional information and song terminol- tification role or if they suggested other communi- ogy). A note is a continuous utterance of sound of cative functions.

Figure 1. A typical sequence of mockingbird song with multiple repetition of songs. VARIATION IN MOCKINGBIRD SINGING 95

Figure 2. A sequence of mockingbird song with very short bouts. In this unusual sequence of mostly one-song bouts, the singer is linking temporal groups by using some part of each previous group in the next.

Methods Macintosh Quadra 700 running Sound Designer II (Digidesign, 1992) and converted into spectrograms Ten samples of mockingbird song were obtained in Canary 1.1 or 1.2 (Bioacoustics Research Pro- that were designed to represent widely different lo- gram, 1995). The breeding season samples from cations and recording circumstances (Table 1). The North Carolina were digitized at 22,050 Hz; all other samples were obtained from Massachusetts, North samples were digitized at 44,100 Hz. After being Carolina, and Florida. Some were recorded in the digitized, each sample was separated into blocks of day, some in the night; some were recorded in the approximately 20 s. breeding season, some outside the breeding season. Once converted into spectrograms, samples were With one briefer exception, all samples were ap- measured, note by note, by student technicians. Four proximately 3 min in duration and contained 700– song features were measured for each note using 1100 notes. All samples were digitized on a Canary’s 1.x’s automatic logging feature: frequency

Table 1. Sources of the Samples of Mockingbird Song Used in this Study and the Characteristics of the Samples

Samplea Region Season Time of Day Mate Status (if known) Sample Length

NC-1 Central North Carolina breeding day mated 232 s NC-2 Central North Carolina breeding day unmated 190 s NC-3 Central North Carolina nonbreeding day 239 s MA-1 Central Massachusetts breeding day 162 s MA-2 Coastal Massachusetts breeding night 376 s FL-2a Florida breeding night unmated 206 s FL-2b Florida breeding day unmated 45 s FL-2c Florida breeding night unmated 198 s FL-1a Florida breeding day 277 s FL-1b Florida nonbreeding day 196 s

aArabic numerals (NC-1, NC-2, etc.) denote different individuals; a, b, c indicate different recordings from the same individual. 96 THOMPSON ET AL. at peak of amplitude (peak frequency) in Hz, dura- song of the next bout, and (4) the first song of the tion in ms, interval in ms, and peak amplitude (Ca- bout five bouts later in the sound record. The tetrads nary 1.1 and 1.2) (Peak amplitude measurements are provided the basis for three sets of sound similarity valid only for comparisons within sample.) The be- correlations differentiated by the distance apart in ginnings and endings of notes were determined by the sound record: correlations between songs within enlarging spectrograms and playing them at 1/8 the the same bout, between songs from neighboring normal speed. The reliability of each technician in bouts, and between songs five bouts apart in the per- determining where a note began and ended was formance. achieved by using an explicit measurement protocol and was confirmed by correlating his or her Results measurements of the notes of a standard selection done by all previous technicians. The 10 samples exhibited the variability and volu- After samples had been measured, the place of bility characteristic of northern mockingbird sing- each note was identified in the bout/song/note struc- ing. All singers added new note and song types as ture. Song boundaries were determined by the be- their performances progressed and bouts of old types ginning and ending of a repeated sequence of notes, were seldom repeated. bout boundaries in a change in the sequence of songs that were repeated. These distinctions were straight- Variation Between Samples forward in most cases. Difficult cases were notes Every variable we examined varied significantly not obviously elements of songs. Single notes that between samples (one-way ANOVA for sample, were repeated were treated as a bout of one-note df =9, p< 0.0001). This generalization applies songs. Notes not appearing in immediately repeated equally to the note features (peak frequency, dura- sequences were treated as one-note, one-song, bouts. tion, and internote interval) as well as bout features Intervals between notes were divided into three (songs/bout, notes/song, notes/bout, bout sound den- types: interbout intervals, intersong intervals (other sity, and average interbout interval). than those that separated bouts), and internote intervals (other than those that separated bouts and songs). Prevalence of the Bout/Song/Note Structure From this master database, a secondary database was created that contained the bout averages for each The bout/song/note structure of the samples was sample of: notes per song (N/S), songs per bout (S/ reflected in comparisons among their notes and their B), notes per bout (N/B), interbout interval (IBI), songs. Note durations, internote intervals, and note intersong interval (ISI), internote interval (INI), and peak frequencies were more similar when taken from sound density (the proportion of time occupied by notes of the same bout than from notes overall notes). (nested ANOVA for bout within sample, df = 852, Variation in the bout structure of the performances p < 0.0001 for all three variables) (the degrees of was examined by various means. Analyses of vari- freedom are large because they relate to the test of ance located sources of the variation in note mea- the hypothesis that the notes of all samples are drawn surements (within and between bouts). Runs tests from the same population of note measurement). were used to detect serial over- or underdistribution Moreover, for all 10 samples, the first song of each in bout parameters (i.e., the tendency for high val- bout was on average more similar to the last song of ues of parameters to cluster together in the perfor- the same bout than it was to the first song of the mance and at a distance from low values). In addi- following bout or the first song of a bout five bouts tion, the spectrogram correlation module in Canary away (two-way ANOVA for distance and sample: 1.2 was used to evaluate the degree to which the distance, df = 2, p < 0.0001: planned comparison: similarity of two songs was dependent on the num- within bout > neighboring bouts, df =1, p< 0.0001). ber of intervening songs in the performance. This In all samples, the bout/song/note structure was sound similarity analysis was performed on sets of emphasized by the temporal structure in that larger 10 song tetrads drawn from each sample. Each tet- temporal breaks tended to occur at boundaries be- rad contained four songs: (1) first song of a multisong tween bouts and songs (nested ANOVA, interval type bout, (2) the last song of the same bout, (3) the first within sample, df = 2, p < 0.0001; planned compari- VARIATION IN MOCKINGBIRD SINGING 97

son: bout boundary intervals > nonboundary inter- of notes in the figure. The first three notes of the vals, df =1, p< .0001). first group are paired with a new element in the second group. The third group consists of the last two Variation in Differentiation of Bouts notes of the second group. Finally, these notes are Within and Between Samples modulated upward in pitch and repeated to form the fourth group. The auditory effect of such a pattern We found ample evidence that the birds varied of group linkage is of a transition from the elements the distinctiveness of bouts. The samples differed in of the first group to those of the last group via the the degree to which sounds of the same bout were elements of the intervening groups. more similar than sounds of different bouts (two- way, ANOVA for distance, distance × sample, Discussion df = 18, p < 0.02). Moreover, instead of dissimilar bouts appearing together in a manner that would On the one hand, the results support the conclu- emphasize the boundaries between them, similar sion that the bout/song/note structure is a character- bouts were clustered together in the performance. A istic feature of northern mockingbird singing. Many Runs Test determines whether values above or be- features of the performances of our subjects served low the median tend to be under- or overdistributed to emphasize their bout structure, and the widely in time (i.e., the degree to which values above the differing circumstances of our samples confirm that median tend to occur in clusters or to be systemati- the bout structure is species specific. On the other cally separated by low values). If the structure of hand, the results also support the conclusion that the the performances tended to emphasize bout distinc- salience of the bout structure is a variable property tiveness, then singers should put dissimilar bouts of mockingbird performances. adjacent to one another in time, and a Runs Test Two hypotheses have occurred to us to resolve should indicate overdistribution of high and low this apparent contradiction. One possibility is that values of bout parameters. In fact, overdistribution the variation in bout salience is the result of a com- was not observed for any variable in any sample and promise between the needs of species identification underdistribution was characteristic of (p < 0.05, two and other urgent communication needs. Birds’ songs tailed): peak frequency (7 samples), single-note-bout have been shown to carry information not only about vs. other (6 samples), and note duration internote the singer’s species, its individual identity, and its interval and songs/bout (2 samples). Sound similar- quality, but also about its behavioral propensities ity analysis confirms that neighboring bouts are more such as readiness to attack an intruder (Catchpole & likely to be similar than bouts further apart in the Slater 1995; Falls, 1969; Smith, 1997). The varia- performance (one way ANOVA, df =1, p< 0.002: tion required for coding such information would, of planned comparison: within bout > neighboring necessity, increase the degree of variation across the bouts, df = 1, p < 0.002 ). species in one or more properties of singing. Ac- These quantitative measures of interbout linkage cording to this hypothesis, the salience of bouts var- may not entirely convey the extent to which bouts ies because it is carrying information concerning are linked in the performance. Detailed examina- instantaneous variations in the singer’s behavioral tion of the spectrograms reveals what seem to be propensities. elaborate interconnections between bouts. A bird Another hypothesis is that variations in bout prop- may utter a series of bouts in which many song fea- erties are driven by demands on the performance as tures are held constant while only a few are varied. a whole. This line of thought suggests an analogy Alternatively, a bird may link a succession of bouts between a repertoire-singing bird such as a mock- by singing a note from each preceding bout as an ingbird and a bower-building species, such as the element in the songs of each immediately following satin bowerbird (Ptilonorhynchus violaceus) one. Singers may also link temporal groups of sounds (Borgia, 1986; Gould & Gould, 1989). A bowerbird in the same way. Figure 2 provides an example of constructs a bower and decorates it with items taken such intergroup linkages from a selection of noctur- from its physical environment, including some sto- nal singing in which the normal bout structure of len from its neighbors. Its reproductive success is the song is greatly reduced. There are four groups determined in large part by the variety of items ob- 98 THOMPSON ET AL. tained and the skill with which it arranges these items havior group at the University of Massachusetts. I in space. Similarly, a mockingbird constructs a am also deeply grateful for the intellectual support “bower” of sound and decorates it with items taken of W. John Smith without whose kind encourage- from its auditory environment, including some “sto- ment this work would never have seen publication. len” from its neighbors. (Mockingbirds also presum- This research was financed in part by grants from ably modify and improvise items, as well.) If the Clark University’s Faculty Research Fund and by a analogy is apt, the mockingbird’s reproductive suc- grant from the National Science Foundation’s Insti- cess is determined in large part by the variety of items tutional Laboratory Instrumentation Program (ILI) obtained and the skill with which it arranges these to Clark University (USE 9152401). The research items in time. described here was conducted as part of the training Taking the bowerbird metaphor seriously has the of its undergraduate authors. effect of focusing attention on higher order proper- Correspondence concerning this article should be ties of the mockingbird’s performance. Nobody sent to Professor Nicholas Thompson, Department would doubt that, if we are ever to fully understand of Psychology, Clark University, 950 Main Street, the function of bower building, researchers must Worcester, MA 01610. Electronic mail may be sent carefully document the arrangement of items in the via the Internet to [email protected] bowers of individual bowerbirds as they construct, maintain, defend, and use their bowers. So, the anal- References ogy suggests, if we are ever to understand the func- Bioacoustics Research Program. (1995). Cornell, NY. tion of singing in mockingbirds, we must document Borgia, G. (1986). Sexual selection in bowerbirds. Scientific the arrangement of sound elements in the perfor- American, 254, 92–100. Boughey, M. J., & Thompson, N. S. (1976). Species specificity mances of individual mockingbirds as the singers and individual variation in the songs of the brown thrasher construct, maintain, and use them. (Toxostoma rufum) and catbird (Dumetella carolinensis). Which hypothesis one holds has important impli- Behaviour, 57, 64–90. Boughey, M., & Thompson, N. S. (1981). Song variety in the cations for the kind of research one does to under- brown thrasher (Toxostoma rufum). Zeitschrift für stand variation in mockingbird song. If one believes Tierpsychologie, 56, 47–58. that the variation arises from moment-to-moment Catchpole, C. K., & Slater, P. B. J. (1995). Bird song: Biologi- cal themes and variations. Cambridge University Press. changes in the singer’s propensities, then research Derrickson K. C., & Breitwisch, R. (1992). Northern mocking- on mockingbird singing should focus on the rela- bird. The Birds of North America, 7, 1–24. tion between variations in bout structure and mo- Falls, J. B. (1969). Functions of territorial song in the white- throated sparrow. In R. A. Hinde (Ed.), Bird vocalizations ment-to-moment variations in the singer’s circum- (pp. 207–232). Cambridge, UK: Cambridge University. stances, research that could be carried out with rela- Gould, J. L., & Gould, C. G. (1989). Sexual selection. New York: tively short samples of singing on relatively large Scientific American Library (HPHLP). Howard, R. D. (1974). The influence of sexual selection and numbers of singers. If, on the other hand, one takes interspecific competition on mockingbird song (Mimus the bower metaphor seriously, a very different re- polyglottos). Evolution, 28, 428–438. search protocol is demanded. Because the Kroodsma, D. E., & Parker, L. D. (1977). Vocal virtuosity in the brown thrasher. Auk, 94, 783–785. mockingbird’s “bowers” are constructed of many Marler, P. (1960). Bird songs and mate selection. In E. E. Lanyon thousands of elements and are assembled over large & W. N. Tavolga (Eds.), Animal sounds and communication spans of time, the research would require much (pp. 348–367). Washington: American Institute of Biologi- cal Science. longer samples and either much more time-consum- Mathews, F. J. (1904). Field book of wild birds and their music. ing research protocols or many fewer subjects than New York: Putnam. has been customary in studies of species with less Moody, K, K., Ledoux, K., & Thompson, N. S. (1994). A sys- tem for describing bird song units. Bioacoustics, 5, 267– elaborate performances. 279. Smith, W. J. (1997). The behavior of communicating after 20 Author Note years. In D. Owings, M. Beecher, & N. S. Thompson (Eds.), Perspectives in ethology, XII: Evolution and communication (pp. 7–53). New York: Plenum Press. We have had a lot of help from colleagues, nota- Wildenthal, J. L. (1965). Structure in primary song of the mock- bly Don Kroodsma and the members of his bird be- ingbird (Mimus polyglottus). Auk, 82, 161–189.