Spectrograms of the ten exemplars used for the playbacks (a–j). Each exemplar came from a different male and we attempted to select unique exemplars, although some song types were shared among males. Each exemplar consists of a sequence of introductory notes (P1), an IPU, and a concluding phrase (P2; labeled on exemplar i). The spectrograms were created in PRAAT with a Gaussian Window (window length = 0.1 ms) and a dynamic range of 70 dB.
Example of active space construction for song exemplar c in a non-urban closed habitat. Pictured top left is the power spectrum for the first replicate of song c recorded at 1 m and the bottom left is the spectrum of the noise. The noise was subtracted from the signal to get a SNR. Pictured right is an example of the regression technique used to determine the active space of the signal for several frequencies (indicated by arrows in power spectrums). The active space at each frequency is the distance at which SNR was equal to the critical ratio. The active space for the signal was defined by the frequency with the largest active space.
Spectrograms of a song that were recorded after propagating in each of the four habitats are shown on the left. Exemplars are in order of propagation distance and include recordings at 0.6, 1, and 5 m and from 10 to 100 m in 10 m increments. The spectrograms were created in PRAAT with a Gaussian Window (window length = 0.1 ms) and a dynamic range of 70 dB. Power spectrums for noise in each of the four habitats are shown on the right. Low frequency noise had more power in the urban than non-urban habitats. In most habitats there was a second peak of noise from 5 to 8 kHz that was generated by insects. In non-urban closed habitats there was also a third spectral peak that was generated by insects and the song of other bird species. The power spectra were created in PRAAT from a concatenation of all the recordings of noise in a given habitat.
Active space as a function of frequency for each of the four habitat types averaged across exemplars using the masked threshold method. There are two main peaks in active space—one at lower frequencies representing active space of the introductory phrase (P1) and one at a higher frequency representing the final phrase (P2). Standard errors ranged from 1.01 to 1.67 m.
(a) Active space of each of the ten song exemplars (entire song) averaged across all recording sites using both the cross-correlation method and the masked threshold method. (b) Active space of the entire brown-headed cowbird song in each of the four habitat types averaged across all exemplars. Data shown are lsmeans ± S.E. which were generated by the LSMEANS statement in PROC MIXED for SAS 9.2.
Box and whisker plot of the frequency band that had the maximum active space (highest SNR at threshold) in each of the four habitats. This active space in this frequency band was then used as the active space for the entire song. The plot shows the mean (+), median (—), interquartile range, and potential outliers (0,*). The dotted lines indicate the entire range of frequencies driving active space in each of the four habitats (excluding outliers).
(a) Active space for the introductory phrase (P1) of a brown-headed cowbird song and the final phrase (P2) for each of the exemplars. (b) Active space for the introductory phrase (P1) of a brown-headed cowbird song and the final phrase (P2) in each of the four habitat types averaged across all exemplars estimated using the cross-correlation method. Data shown are lsmeans ± S.E. which were generated by the LSMEANS statement in PROC MIXED for SAS 9.2.
Article metrics loading...
Full text loading...