STRUCTURAL CHANGES IN SONG ONTOGENY IN THE SWAMP SPARROW MELOSPIZA GEORGIANA PETER MARLER AND SUSAN PETERS RockefellerUniversity Field Research Center, Tyrrel Road,Millbrook, New York 12545USA ABS?RAC?.--Sixteenmale Swamp Sparrowstrained with recordedsongs in infancywere recordedon a weekly basis throughoutthe processof song development.Analyses of the 15,000 songsrecorded from them provided the basis for a seven-stageclassification of the stepsin songontogeny, with the four major divisions of subsong,subplastic song, plastic song,and full song.The individuallyisolated males all followeda basicallysimilar devel- opmental progressionthrough the seven stagesof song development.Accounts of other speciessuggest that the pattern seen in the Swamp Sparrowis a rather generalone, when one allowsfor speciesdifferences in songstructure. The abrupt reductionof songduration and its variability was a marker for the crystallizationof plasticsong into full song.At this time there was also a reduction in the number of syllabletypes used and the number of parts per song.Swamp Sparrows produced considerably more syllabletypes in plasticsong than necessaryto generatethe normal,species-specific song. Received 10 June 1981, accepted 19 October 1981. A GENERALpicture is availableof the overall Four males,designated as the "prehatchtreated" structural properties of subsong and plastic group,were injected with CI628on 4 successivedays songand the transformationsto which they are between8 daysbefore hatching and two daysafter subjectprior to the emergenceof full song hatching. Three males, the "posthatch treated" (Nice 1943, Marlet 1956, Thorpe and Pilcher group, were injectedon 6 successivedays between 3 and 8 days after hatching.A third group of five 1958, Nottebohm 1972a). Despite the plethora males,the "late treated"group, were injectedon 4 of studies on mature birdsong, there has been to 6 successivedays, betweenthe agesof 9 and 24 little descriptive analysis of song ontogeny. daysof age.Four males were not treated at all. There The phenomenonof subsongis nevertheless were no lossesas a consequenceof the treatments, of greatpotential interest because of its distri- and no differencesin song could be detected be- bution in birds, which seems to coincide with tween treatmentgroups and controls. the incidence of song learning (Nottebohm All birds were trained for 40 daysbetween 24 June 1972a, 1975; Marlet and Peters in press a). No 1977 and 26 August 1977 startingbetween the ages one, however, has yet conducteda longitudi- of 16and 26 daysof ageand endingwhen the birds were between55 and 67 daysof age. The birds were nal study of songdevelopment in individual trained with one of two setsof tape-recordedsongs, birds to determinethe timing and manner of A and B, designedto exploresome of the species emergenceof learnedthemes. As a first step, differencesin song learning in Swamp and Song we presenthere the resultsof an investigation (Melospizamelodia) sparrows. The detailsof this ex- on the structureof subsongand plasticsong in perimentwill be presentedelsewhere, but we de- the Swamp Sparrow(Melospiza georgiana). scribethe training songsbriefly here. Wild male Swamp Sparrowsongs typically consist METHODS of identical repetitions of one distinctive syllable Sixteenmale SwampSparrows hatched in the wild type at a steadytempo. This patternand the rarely- between30 May 1977and 24 June1977 were brought occurring(<5%) two-partedtrill servedas the basis into the laboratorybetween 2 and 10 daysof ageand for our training songs (for examples of natural rearedby hand to independence.Younger nestlings Swamp Sparrowsong, see Marler and Peters1977, were brooded by canariesbetween feedingsuntil Peterset al. 1980). ScheduleA, used for training six about 10 days of age. The projecton songdevelop- males, contained 16 one-parted trills composed of ment was grafted onto anotherstudy of possibleef- either normal Swamp Sparrow syllablesor artificial fectsof an anti-estrogen,CI628, on the development syllablescreated from notesof Swampand Song of songlearning abilities. The experimentalmanip- sparrowsyllables (see Zoloth et al. 1980for proce- ulationsinvolved had no effecton singing behavior dural details). The remaining 10 maleswere trained and will not be consideredfurther in this paper be- with ScheduleB, which contained11 different songs, yond thesedetails. 6 one-parted trills, and 5 two-parted trills. These 446 The Auk 99: 446-458.July 1982 JvL¾1982] SparrowSubsong and Plastic Song 447 TABLE1. Criteria for a seven-stateclassification of stagesof songdevelopment in the Swamp Sparrow. Syllables Songs Stage Morphology Repetitions Morphology Duration Crystallizedsong I Stereotyped Clear trills Stableorder Short II Stereotyped Clear trills Variable order Short Plasticsong III Minor variations Clear trills Stableorder Longer IV Variable Clear trills Variable order Long and V Rudiments Some Variable order variable Subplasticsong VI Rudiments None Variableorder Variable Subsong VII None None None Variable songswere composedof a total of 16 different intact tographed, sound-spectrographicanalysis of songs Songand Swampsparrow syllables. Songs for both with the frequencyspectrum displayed against time. scheduleswere organized into 3-min bouts by re- In early stageswe analyzeda 5-10-min samplefor peatinga singlesong type onceevery 10 s, a normal eachrecording date for eachbird. With later singing songdelivery rate for Swamp Sparrows.These were all edited material was analyzed. played to subjectstwice per day, moming and eve- For detailed analysisof the songs,we adopted a ning, during the periodsalready indicated. definition of a "song unit" suitable for all stages. After songtraining was completed,each male was Songsof adult Swamp Sparrowsin nature are about housed in a sound-proofchamber, with the lights 2.3 s in duration with a typical intersonginterval of kept on a normal photoperiod by a time-clock more than 5 s. The averageintersyllable interval for changedevery week. We first determined the peak full songis 40 ms and the averageinternote interval singing period during the day. A Crown Interna- is 12 ms. The timing of earlysinging behavior is very tional 800 seriestape deck was set up with a time different, however. Songsseparated by intervals of delayand resetsystem. Once per weekfrom 8 Au- more than 5 s often contain within them intervals gust 1977 to 2 September1977, we sampleda subset greater than 80 ms. Some of the note sequences of 3-7 birds from each training schedule, starting at bounded by 5-s intervalsare extremelyshort. After about 50 days of age and selecting1 out of every 5 someexperimentation, we adoptedthe followingcri- min for 24 h. Most singingoccurred just after dawn, teria for song-likeunits. Any sequenceof notes qual- when the lights were switched on in the chambers. ified as a song if it lastedmore than 0.5 s with no This information provided the basis for the recorded intervals of silence longer than 0.25 s between samplingof singing of all 16 males. Beginningat an notes.Thus, a 1.3-s sequenceof notesthat contained averageof 99 daysof age and continuinguntil about an interval of 0.3 s occurringafter the first 0.5 s would 380 days of age, we recordedeach male for 1 h once be analyzedas two songs.Applying thesecriteria to per week. TandbergSeries 15 tape recorderson time the output of all 16 subjects,we defined a total of clockswere used,with a tape,speedof 33/4inches per 15,062songs for all songstages. Song durations were second.From 593 recordingsessions for all birds, 235 measuredon a Summagraphicstablet and a PdP 11/ h of recordingswere obtained that containedsong. 10 DEC computer,programmed to registertime du- All recordingswere monitoredand then editedfor rations to the nearest 12.8 ms. analysis.The editing procedurediffered slightly for early and later stagesof singing.With early singing, we edited out up to 15 min of continuoussinging RESULTS and estimatedthe remaining amount in the 1-h re- PATTERNS OF SONG ONTOGENY cording. In later stages,song occurredin more dis- creteunits, and all songswere edited out, with two As a first step in the interpretationof song exceptions.Occasionally, recording quality was poor development, we createda seven-stageclassi- becauseof microphoneor tape-recordermalfdnction. fication, based primarily on the presenceor In a few of these casesa small sample of song was absenceof syllabicstructure and on the relative edited out to documentthe stageof singing. Late in stability of syllable morphology and the se- songdevelopment we occasionallyedited only alter- nate songs,when it was clear that a large sampleof quencesin which syllableswere delivered (Ta- the male's crystallizedsongs had already been ob- ble 1). Figure 1 representssamples of songpro- tained. duction from one male Swamp Sparrow, The edited tapeswere run on a PrincetonApplied illustrating the sevenstages in the emergence ResearchSpectrum Analyzer, which providesa pho- of one of this male'stwo crystallizedsyllable 448 MARLER AND PETERS [Auk, Vol. 99 Training syllables 2 3 4 Stage Subsong Stage Plastic song ',"h'l'11,t ' ,','1 syll'5 syll'6 syll3 Subplasticsong syll 2 syll 4 ,,!It I Izttt 1t 71/I ,'t Stage Crystallizedsong syll 2 .5 sec Fig. 1. Songdevelopment in a SwampSparrow. Samples of eachof the sevenstages of developingsong in a male SwampSparrow that was trainedwith ScheduleB. The stageis indicatedat the left and major stagesabove. Imitations of thesix training syllables portrayed at thetop of thefigure are identified by number. Trainingsyllable