https://doi.org/10.5061/dryad.0vt4b8h5n

The buzz segment of Savannah sparrow songs is generally consistent within a population, but varies across populations. Variation occurred within the Kent Island (New Brunswick, Canada) population in two ways: between 1994 and 2004 a total of 19 males sang a variant buzz that has not been sung since, and between 2013 and 2018 foreign (Western) songs were broadcast from speakers on the study site and were copied by some birds during song development. The paper associated with this data set assessed responses to local and foreign dialects using four buzz. By comparing the acoustic characteristics of each bird’s buzz to those of the stimuli it heard, we were able to show that the intensity of a bird’s response is not predicted by the acoustic difference between his song and the stimulus. We found that responses to the local dialect depend on a bird’s experience both during development and during the current breeding season, and that only when the dialect features within a song are both familiar and homogeneous (sung by nearly all the birds in a population) do the birds respond more aggressively to the local dialect than to a foreign dialect. When the local dialect is not homogeneous, responses fall to the same level as those given to unfamiliar dialects and to minority dialects present on the study site.

Description of the data and file structure

Acoustic characteristics of different buzz types and stimuli (2 csv files).

1)  ForeignWtnKIBuzzes.csv has the measurements of acoustic characteristics for buzz segments from 362 Savannah sparrow songs. A total of 222 songs were drawn from recordings made on Kent, Grand Manan, and White Head islands in the Bay of Fundy, New Brunswick, Canada. A total of 94 songs were recorded 500 km away in Williamstown,Massachusetts, U.S.A., and 46 foreign songs from Western North America songs were broadcast through speakers on the Kent Island study site from 2013-2018. The bird or recording identity, the buzz type, and the year are given in the first three columns. The next three columns provide the song type, and that type indexed by numbers so that the file can be sorted appropriately for graphing. We had previously found (Williams et al., 2019, https://doi.org/10.1007) that buzz dialects could be characterized by mean frequency and pulse period, so we used those two measurements for analysis. For each buzz, frequency (Hz)and buzz duration (ms) measurements obtained using Sound Analysis Pro (Tchernichovski et al., 2000, 10.1006/anbe.1999.1416) are included. We counted the number of pulses and used that and the duration to derive the pulse period (ms). For 28 of the 363 birds, only these two critical variables were measured; for the remaining birds, additional variables as measured by Sound Analysis Pro are available upon request to the corresponding author. 

2)  PBstimulussubjects_2016-7.csv includes the duration (ms), mean frequency (Hz), peak frequency (Hz), number of pulses (count), and pulse period (ms) for the buzzes of all stimuli and subjects used in 2016-2017 on Kent Island and Grand Manan Island. The first two columns in each row give the bird identity and the location or stimulus type of the buzz. The third column designates the song type so that the file can be sorted appropriately for figures.

Examples of buzz types and stimuli (4 WAVE files)

1)  Kent Island stimuli.wav includes the eight stimuli of this category used in 2016-2017 (none were derived from songs sung by any of the subjects). All have high mean frequencies and short pulse periods (buzzes are made up of a series of closely spaced sound pulses).

2)  variant Kent Island stimuli.wav includes the eight stimuli of this category used in 2016-2017. All were derived from songs recorded between 1994 and 2004 and have relatively low mean frequencies and long pulse periods.

3)  Williamstown stimuli.wav includes the eight stimuli of this category used in 2016-2017. All were derived from songs recorded 500 km from the study site, and have midrange mean frequencies and midrange pulse periods.

4)  foreign Western buzzes.wav includes the eight stimuli of this category used in 2016-2017. All have relatively low mean frequencies (that do not overlap with the range of Kent Island buzzes) and short pulse periods.

Responses to playback stimuli (3 csv files)

There are three files with data describing responses to playback sessions, one file for each period during which playback studies were performed. The variable names and orders for each file vary slightly but not significantly (they were compiled by different authors at different times), and all include the following variables used in analyses: subject identity (ID), location, age, breeding experience, date and time of playback session, how long the primer was played (Primertime), stimulus category (StimCat), which of the eight stimulus sets was used (StimSet), order of presentation (trial), flights and wing flutters, songs, and response duration (= disengagement time; the primary response measurement was expressed in seconds “DisengmtTsec”).

For the 2016-7 and 2022 files, the subject’s closest approach and average distance to the speaker were noted (in meters), and the subject’s songs were counted and categorized as either normal or soft (whisper) songs.

The number of Flutter flights and the number of Wing flutters were summed to give the variable “Flutters”. The number of Whisper songs was added to Flutters to yield the aggressive displays variable (AggrBeh).

The “Female” variable notes whether the female inhabiting the subject’s territory was present or absent and, if present, whether she responded to the playback stimulus.

The “Pbstatus” variable is the research team’s assessment of whether interference from other birds or humans invalidated a trial because the subject was responding to those other stimuli. Only trials with a status of “good” were used in the analyses, which is why some subjects are not listed as having completed all four trials.

Only the disengagement time measure was taken for primer playback period, and so the other variables for that period are given as “NA” for each subject.

The three files are:

1)  2012playbackbuzzdata.csv summarizes responses of trials completed before the foreign Western songs were broadcast through speakers on the study site. Since the stimuli to be broadcast as part of that study had not yet been chosen, foreign (Western) buzz stimuli (StimType = “Stimulus D”, corresponding to StimCat = shortlow) were not used in the playback study and “NA” appears in all corresponding rows.

2)  2016-2017playbackbuzzdata.csv summarizes responses to trials completed in 2016-2017, when birds also heard foreign Western songs broadcast from speakers while they were on the study site. In addition to the data summarized above, there is a column labeled “SongType”, which describes a bird that had copied a foreign Western song from the broadcast speakers as “golden”. Ages were only available for Kent Island birds, and all subjects were either one year old (“FirstYear”) or two years old (“Experienced”). The differences between the stimulus pulse periods (ms) and frequencies (Hz) were determined for each subject’s buzz and the stimuli he heard. These distances were normalized for each measure, and then used to calculate the Euclidean distance between each subject’s buzz and each of the stimulus buzzes he heard. This value is given by the “stimdistance” variable, which was used to analyze the relationship between a bird’s aggressive response and the acoustic distance between his own buzz and the stimulus buzz. Because the primer had no buzz and some birds’ songs were not recorded, some measurements of the stimdistance variable could not be calculated and are given as “NA”.

2022playbackbuzzdata.csv summarizes responses to trials completed in 2022, when some of the older subjects had heard the broadcast foreign Western songs during development (denoted by the “TwiceSpeaker” variable), and another subset of subjects either sang a copied foreign Western song or heard such a song sung by a neighbor on an adjoining territory (denoted by the “NeighborGolden” variable).

R scripts (2 files)

1)  BuzzFamiliarityHomogeneity.R recapitulates all of the data analyses reported in the paper and also reproduces all but one of the figures from the paper. It calls on the three playback data files described above:

            2012buzzplaybackdata.csv

            2016-2017buzzdplaybackdata.csv

            2022buzzplaybackdata.csv

2)  BuzzCharacters.R reproduces Figure 2 from the paper. It calls on two files described above:

            ForeignWtnKIBuzzes.csv

            PBstimulisubjects_2016-7.csv

Sharing/Access information

Data can also be obtained from the corresponding author (hwilliams@williams.edu). All data were generated from original research by the authors, and should be credited with a citation to the published paper in Animal Behaviour and to this database.