Monday, October 26, 2009

Bass Notes

Singing trees, not singing fish

While I was continuing to get skunked at Moss Island, at least on the double knocker front, "Jacob" had been circulating my 3/18 double knock recording a bit amongst other interested people. I did not expect much response, as the big projects seemed to have dismissed Tennessee a year or more before. However, in June one of the Cornell PIs did have comments, and there was a bit of an e-mail conference between him and us. His initial response was two fold: very impressive double knocks, and because of the 10 second spacing almost surely just the simulations. There was some miscommunication between Cornell and Tennessee about the particulars of the simulations and timings, but even after this was sorted out the Cornellian remained convinced that mistiming, miscounting, and anomalous sound propagation were more likely than an actual response to the double knocker. He commented that, "a response pattern of just three double knocks spaced 10 s apart is unlikely for Campephilus." My observation in February of what was essentially this exact thing (three double knocks in rapid series) in response to a gunshot appeared to be given no significance. Worthy of note, we were not operating from a presumption of Ivorybillness and did not dismiss data because it did not match a priori notions of how an Ivorybill should behave. If we have Campephilus-like (actually, Campephilus-identical) double knocks at Moss Island, that fact is of great significance to the Ivorybill quest regardless of whether or not they truly are from an actual Campephilus. I'll have more to say on that in my concluding posts in this series. In the end he concluded that the sounds I recorded were "quite obviously something else than an Ivory-billed Woodpecker." My closing response to this was, "'Obvious' is a judgement. There are many rational, well-educated people who consider it obvious that the human mind was intelligently designed, that the bird in the Luneau video is a normal Pileated Woodpecker, and that double knocks are caused by duck wingtip collisions."

Beyond these immediate discussions, though, we also talked about what sorts of responses to the double knock simulator might be expected based on what has been seen in other Campephilus species. Much remains unanswered, but doesn't it seem likely that there would be wide variability between species, times of year, and local circumstances? Imagine if you tried to predict the behavior of a Lincoln's Sparrow on the wintering grounds based on the behavior of the congeneric Song Sparrow on the nesting grounds. You would get some broad generalities right, but would be wildly in error on many important particulars.

There is one well-publicized example of the behavior of a pair of Campephilus woodpeckers in response to simulated double knocks: David Attenborough's Magellanic Woodpeckers. Bear in mind, though, that this is a nature documentary, not a piece of raw scientific data. What we see in the final video is very likely a greatly edited abstraction from a large amount of footage. Who knows how many attempts were needed before this worked, if Attenborough's little tapping on the tree with two small rocks is the only attractant they actually used, whether they might have been standing 10 meters away from the birds' nest tree, if every shot is even of the same pair of woodpeckers, etc.? This is primarily entertainment, not scientific documentation. Still, though, it is the best footage of double-knocking Campephilus that I have seen anywhere.

I have wondered ever since I first shot the 3/18 video if there is a way to determine whether the double knocks I recorded are real or simulated based purely on the properties of the sounds themselves, without involving matters of timing, location, etc. To the ear the simulator can be a good approximation to the real sound, but it is not identical. Spectrographically there are some notable and consistent differences, which I alluded to in the previous post. I already posted one sonogram of the sound of the simulator at close range; here are two more along with video clips so you can hear also. As always, click the sonogram to see a larger version:

April 9:

August 21:

And here are samples of double knocks from the same simulator recorded from increasing distances, from August 21-22:

Note that to improve legibility, I boosted the gain by 100% on the middle three and by 200% on the last one, hence there is a lot more drop in loudness than it appears. The final sample, from 600m, was recorded at dawn; the others were at midday. It shows very well the greatly improved sound propagation in the dawn stillness; remember though that it is still less than 1/3rd of the distance that I was from Scott on the morning of 3/18. Looking at the whole sequence, you see the loss of the high frequencies, the blurring out of the sharpness, especially in the second knock and at midday, and the persistence of the long trailing gunshot-like echo at lower frequencies, regardless of time of day.

For comparison, here are two examples of real double knocks. First, a Powerful Woodpecker from the Macaulay Library of the Cornell Laboratory of Ornithology, catalog number 84098:

Next, here are sample double knocks from David Attenborough's Magellanic Woodpeckers (I hope the BBC, who seem to be checking my blog regularly, do not object to this use of a tiny sample of their footage. It would seem like fair use to me):

The most obvious differences are twofold:

A. The simulations have more high-frequency sound than the real double knocks.

B. The simulations show a more pronounced trailing echo than the real thing, looking almost like little gunshots with their broad rightward smear.

The real sounds are variable in their frequency distributions, especially on the low range. The examples that show the strongest base notes are the ones by the male Magellanic when he appears to be rapping close to a cavity, presumably near a void in the tree truck. This deep base is what gives some of the knocks their hollow, resonant quality.

Look more closely at the basso profundo on the Magellanic double knocks. Notice how the vertical line marking each knock bends to the right towards the bottom. There is a delay of a few milliseconds in these base tones as compared to the higher frequencies. In the Magellanic knock with the strongest base, it looks like this delay is actually greatest not in the deepest tones, but a bit above them; it's a bowing out, not a simple rightward bend. On the Powerful sonogram, the basso profundo is much less forte. But in the clearly visible faint tails at the bottom of the knocks, we see a distinct rightward bend as well.

I should say here that I am a hack when it comes to bioacoustics. Sure I have a Ph.D., but it is in Ecology, not Physics, Acoustics, or even Music. My formal training in acoustics would have been nothing beyond college physics about 28 years ago. So I am no expert here. But I'm not a total dope, either. It seems to me that this rightward bowing on the base of the sonogram, indicating a delay in the production of the tones in the frequency range of a few hundred Hz, is likely a property of the resonance of tree trunks. I expect it is this brief delay that especially gives the real knocks their sonorous, resonant quality.

Now, look back at the simulations. They do not show this feature. The initiation of the sound is at the same instant at all frequencies, making a crisp straight vertical line on the sonogram. Whatever the resonance properties of a real tree are that produce this effect, the simulator appears to lack them. Again, I'd hypothesize that the lack of this feature is what makes the simulations sound subjectively less resonant to the ear. There is no lack of the basso profundo tones in the simulated double knocks; they just happen at the exact same instant as the higher tones and thus to the ear are absorbed within the one big sharp "whack." I would propose to add a third distinguishing feature that can potentially differentiate simulations from the real thing:

C. The simulations lack the rightward bend of the sonogram at low frequencies; all frequencies of sound are produced at the same time.

As far as utility in the field, it varies between each feature. Item A, the excess of high frequencies, will fade out rapidly in a forest so it is probably of no real value. Item B, the trailing echos, certainly might be useful. In the data I posted both here and in the previous post showing the simulations at different distances, this only becomes more prominent with attenuation, as the whole sonogram dissolves into mush. Unfortunately I do not have any recordings of known real double knocks at great distance for comparison. Item C, the rightward bending base of the sonogram, might well survive over distance. As the speed of sound in air is largely independent of frequency, this feature (straight versus bent) should be preserved for as long as the sound remains crisp enough to give a good spectrographic image. It's also worth noting that this feature might be useful for separating "real" double knocks from other sounds, such as duck wingtip collisions, construction or vehicle noises, etc.

Finally, to the Tennessee sounds from March 18th one more time. Here is the clearest one, number 3 (click to enlarge):

As would be expected in either case, the high frequencies are gone. In the base tones that remain, I see rightward bowing. In fact it looks almost exactly like the shape of the Magellanic double knock at the same frequency range. There is also not a trace of the trailing echo, or any other smearing. So for the two criteria that might be meaningful in the field at a distance, this one scores as "real" on both.

Here are the other two 3/18 double knocks, along with the audio/video again:

The first knock of #1 is fairly clear; it looks bent to me. The final note of this double knock, plus both notes of double knock #2 are very faint. Still, the faint lines do seem to show a bend.

Score for this round:

Simulations: 0
Real thing: 2

Once again, this still only indicates that the sound is not the simulator and is something else rapping on wood making a sound very much like a Campephilus double knock. As always, this does not in itself prove the presence of a real Campephilus woodpecker at the site. But it again supports the idea that the Moss Island Mystery Double Knocker is a real and coherent phenomenon, not just a random assortment of ordinary sounds being misinterpreted.

Other posts in this series:


At 8:16 PM, Blogger Unknown said...

Here's an aside that you may or may not find of interest. I once sang in a choir under the direction of a conductor who had an orchestra and choir. From my position in the choir (right behind the orchestra), it bugged me that the bass violins always attacked chords an instant before the rest of the orchestra. Then I learned that this was a special trick of the conductor, who recognized that, out in the audience, the bass sound seems to arrive an instant late. When I listen to recordings of this orchestra, it sounds right.

At 9:04 PM, Blogger Bill Pulliam said...

Definitely interesting. Actually, considering that many musical instruments (strings, woodwinds, organ pipes, etc.) are based on the resonant properties of hollow pieces of wood, I'd not be at all surprised to find that my "discovery" is actually an extremely well known phenomenon. Considering that at these frequencies the periods of the sound waves are on the order of 3-5 ms, it's probably just that it takes a multiple or two of this time before peak resonance is achieved and the full sound is produced. The simulator, lacking an enclosed resonating chamber, just releases all the sound at once without showing this effect, like a xylophone or woodblock instead of the string bass. You standing right behind the bass section hear (and see) the initial attack of the bow, but the audience doesn't hear it until the molecules in the whole space inside the instrument are vibrating in unison an eyeblink later. Perhaps it's also a bit like using a rattle: you have to flick your wrist a moment before you want to hear the sound.

The ear is definitely sensitive to these small shifts in timing and quality of sound. Even if you can't consciously hear that "hey, the base tones are coming 5 ms later than the treble" you will sense it as a difference in the quality of the sound. All these spectrographic differences are just quantitative representations of distinctions that the brain makes (probably with finer levels of discrimination) in a holistic, qualitative, even emotional way -- dull, bright, sonorous, eerie, lifeless, etc.

At 9:14 PM, Blogger fangsheath said...

I'm afraid I perceive a profound lack of objectivity in some quarters when it comes to evidence gathered in ivory-bill searches, whether imagery, sounds, or sightings. I think this is well illustrated by the "obvious" remark. This is not the first time I have seen an argument proceed from mere questions being raised to a final conclusion, stated with no trace of doubt, that ivory-bills cannot be responsible for the observations. In one such case I know of, an exclamation point was actually used. Such patterns indicate to me a mind determined to arrive at a foregone conclusion, by whatever means.

Scientists of course are human beings, subject to human foibles. But one has to wonder about all of the evidence that we do not have access to. How much of it has been tossed aside due to presumptions about ivory-bill habitat, presumptions about ivory-bill behavior, unwarranted comparisons with tropical Campephilus, unwarranted comparisons with imagery taken from a blind 20 feet away in the Singer Tract, and highly unwarranted extrapolation of Tanner's conclusions drawn from observations of a few birds in La.? Time and again I have encountered the phenomenon of evidence turned aside because it does not match presumptions. A friend once said, "I can't help how ivory-bills are supposed to behave. I can only tell you what I saw and heard."

At 9:31 PM, Blogger Bill Pulliam said...

In the specific case of the March 18th incident, the simulator is indeed an "obvious" alternative explanation and is what I myself immediately assumed -- you can hear me proclaim exactly this conclusion on the recording. However, on closer examination this explanation has turned out to have multiple problems that make its correctness much less "obvious."

I'm coming into my final few posts where I will be summarizing my thoughts on what it all means, so I don't want to steal too much thunder from that. But as a lead in, if you have the same phenomena being observed at locales A and B, and you want to conclude that at site A they are possible evidence of X but at site B they could not possibly be due to X, well, you've painted yourself into a fairly serious logical corner.

At 3:11 AM, Blogger cyberthrush said...

seems pretty clearcut to me Bill... you have a Magellanic Woodpecker roaming around Western Tenn. Forget Lammertink and Cornell, get David Attenborough up there pronto to assist you guys ;-) But seriously, have search plans, if any, been formulated for the forthcoming season?
...And BTW ditto everything fangsheath says.

At 6:01 AM, Blogger Bill Pulliam said...

Cy -- it's as good a theory as any other!

No one seems to take it seriously when I mention it, but if some relocatable birds ever turn up (i.e. probably a nest) I think getting some high-profile documentary filmmakers involved would be very helpful for public relations and promoting conservation. I don't think they'll come knocking (not even double knocking) as long as we're just chasing phantoms and poltergeists, though.

Someone from is a regular reader, though...

A new reality TV show: Pecker Hunters! Schedule it right between Ice Truckers and Deadliest Catch. I'll go start inflating the balloon...

At 1:58 PM, Blogger Unknown said...

I think sonograms can be over-analyzed with sounds like DKs, especially when comparing them to high quality samples like those you have of the Powerful and Magellanic. If you have "line-of-sight" to the bird, you can get clean-sounding knocks followed by much-attenuated (relatively) echos, which is why you don't see much smearing in the sonograms of those two samples. However, it's more likely in the field not to have line-of-sight, and in that case, what you have are two sets of echos instead. Not single echos, but lots of echos with possibly many different paths of propagation that then eventually interfere with each other at your ear or microphone. The nature of the echo smearing and frequency loss you see on a sonogram are the result of too many variables to count.

The best sensor/pattern recognition system we have by far is our ears and brain. When Melinda said that Scott's DK simulations that morning didn't sound like good DKs, that should carry a ton of weight when comparing them to what you heard. Though I can see the Cornell guy's point of view as well, since those three DKs are exactly 10 seconds apart -- like you'd get with some guy holding a watch. Perhaps the bird responded to Scott's simulations, but not immediately, and mimicked the 10 second cadence to boot.

At 11:52 AM, Blogger Cotinis said...

There is one well-publicized example of the behavior of a pair of Campephilus woodpeckers in response to simulated double knocks: David Attenborough's Magellanic Woodpeckers. Bear in mind, though, that this is a nature documentary, not a piece of raw scientific data. What we see in the final video is very likely a greatly edited abstraction from a large amount of footage.
For what it is worth, I met an entomologist at North Carolina State University (an excellent birder as well as an avid hunter) who has been to that site Attenborough visited, and he said that the birds do come in pretty much as shown in the film. That I would like to see.

At 11:59 AM, Blogger Bill Pulliam said...

Coyle --

I agree absolutely that the ear is able to detect more subtle distinctions that the sonogram; I said this in earlier posts in fact. But the sonogram is useful for reductionistically analyzing sounds and differences between sounds. In the digital era, unfortunately, the experiences of skilled humans are gradually being given less and less weight.

As for propagation and attenuation, these are not mysteries. They follow patterns and can be studied, tending to do predictable and consistent things to sounds. My point here is that the spectrographic differences between what I recorded and the simulations are not the sort of differences that would be expected to be caused by attentuation, echoing, refraction, etc, and could be explained easily if what I recorded was actually a much closer sound similar to the Magellanic double knocks.

At 12:28 PM, Blogger Bill Pulliam said...

Cotinis --

Apparently, though, ONLY during the nesting season on territory.


Post a Comment

<< Home

Site Meter