Birdsong13.10.15

Background

As a New Englander, I’m sure that I’ve heard American Robins’ songs for my entire life. But I only started to pay attention to them recently and, in paying attention, was struck by the apparent regularity of vocalizations within each song, and of the time between songs – particularly those in the morning and evening.

To investigate what I heard, I used a Sony MZ-M10 minidisc recorder to capture about a minute of American Robin song audio. The audio was captured in Cambridge, a rather urban space, which accounts for the substantial low-frequency background noise. The audio was processed to remove the more egregious low and high-frequency sounds, and cleaned, to smooth – but not remove – pops and scratches.

I imported the audio into Audacity, an open-source audio editing platform, and used its spectrogram features to visualize the song data (time vs. frequency vs. intensity). Label tracks were used to measure the duration of each song, and to measure the duration between songs. Given the nature of the tool, I estimate the time accuracy to be within two tenths (.2) of a second (although time measurements are possible at thousandths). Data were analyzed in Google Sheets.

Findings: 

(1) This American Robin did produce a very regular song; four vocalizations per set, with each set lasting an average of 2.38 seconds (min: 1.931, max: 3.352, with 60% of songs falling within 0.2 seconds of the average).

Robin-Regularity-Sample

Time between songs varied more widely (average: 2.52 seconds, min: 1.673 s, max: 3.219 s) with 40% falling within 0.2 seconds of the average. All but two songs contained four notes; one contained five, the other seemingly three. [ View the data and a graph here ]

(2) This American Robin’s frequency ranged from 2.0kHz to 5.0kHz, with the strongest centers at 2.5 and 3.5 kHz respectively. This is where things get complex. Birds sing with their syrinx, an anatomical part at the junction of the two trachea that split to their lungs. Some birds sing with one side of the syrinx or the other, and as a result, produce only a single tone. Others are polyphonous, and can control both sides of their syrinx to create a song with two simultaneous tones [read more here].

Robin-Hetrodyne-Example

Looking carefully at the spectogram, we may be able to see both; examples of song created by the control of a single tone, and a few examples of song created by the combination of two simultaneous tones. Polyphonic properties that lend (or reduce) credence to the possibility that this Robin sang in two tones:

  • Seeming overlap of the 2.0-2.5 and 3.0-3.5 kHz frequencies across time, in some instances of song
  • Hetrodyne frequency/vibration at the sum of the the two fundamentals, e.g., 2.0 and 3.0 kHz (the herodyne frequency appears around 5kHz). 
  • Alternatively, no evidence of the higher – lower (appearing around 1.0 kHz) hetrodyne frequency is evident.

The result? I’m no bird expert, but it does seem plausible that our friendly American Robin is a polyphonic critter!

Future questions:

Clearly, this investigation leaves a lot to be desired! I’m curious to repeat the experiment; this time, I’d like to (a) compare findings across American Robins (b) work to increase the sound quality, clarity and sensitivity (c) explore the possibility of automatically characterizing songs, or vocalizations within songs via a sound processing library – perhaps Python or Perl?

K-12 Education Implications

This piece relates to an essay I’m writing on technological possibilitarianism – stay tuned!

 


Yes: Leave a comment!