Monday, November 23, 2009

Video Success and Failure

As promised a week or so ago, more comments about the recent Pearl River video and video in general...

To begin, an important question: Why did so many people, including some who had been quite skeptical of previous Ivorybill evidence, initially get such a pro-Ivorybill impression from Mike Collin's 11/5 video? I think several factors contributed to this. First, when the bird initially appears, it is flying towards the camera. This created a predictable foreshortening of the wings, making them appear narrower than they really are. This is straightforward enough; however, the wings also appeared pointed. This almost kite-like impression in these initial frames colored perception of everything that followed. Secondly, two related phenomena interacted to prevent most from even considering a Red-headed Woodpecker when they first watched the video. Though it was not explicitly stated, I think we generally assumed that the bird had been seen, not just videotaped. and was known to be a large bird. I believe we made an assumption that the video would not even have been circulated if this fact was not known; an unjustifiable assumption, of course. I don't think this assumption was made explicitly or likely even consciously; but it lurked there. This was compounded by the fact that the published and distributed video was a deinterlaced 60 fps video which plays at half speed on most media players. This fact was not stated when the video was first made available; I had to ask to get this piece of information through a couple of links of communication. Of course, this was not done with the intent of deceiving or misrepresenting; it was done to make the video clearer and easier to analyze. Nevertheless, most viewers' first viewing of this clip was at half speed, without their being aware of this, which caused the bird to appear larger with slower movements. If you take the clip and play it at double speed (i.e. true speed), the bird suddenly looks much smaller and far less Ivorybillish. These things combined very quickly -- long pointed wings, assumption of large bird consistent with its apparent flight style -- to put most viewers in a mental state where Red-headed Woodpecker did not even come to mind as an option. Once this impression is in the mind, you become much more forgiving of the things that appear later in the video that might point in other directions.

The lesson from this: Never make any judgement about a video such as this until you have full information about the circumstances of the encounter, what was seen in addition to what is on the tape, and all the possibly relevant technical aspects of the clip. This is in fact why I have never had anything to say about Mike's "fly-under video" from 2008; without having access to the full clip, rather than just selected segments, I feel that I have no context to judge what I might be seeing.

My second point is much more general. Birders do not seem to have figured out yet what to do with video. A video tends to be treated just as a big heap of poor-quality still images. There is a very strong tendency to pick out individual frames in isolation and just interpret what is and is not seen in them. This approach makes use of the worst parts of a video (the image resolution, or lack thereof, combined with numerous artifacts) and discards some of its most useful parts: the documentation of movement, structure, and dynamics. I've gone on at length in the past about the astonishing failures of big-name birders in misinterpreting and misunderstanding imaging artifacts in low-quality video frames; now I'd like to talk about this second aspect.

Some birders tend to speak of the "giss" of a bird as though it is a metaphysical, supernatural property; perhaps an aura that can only be sensed, not measured. This is of course ridiculous. True, "giss" is a "gestalt" phenomenon; indeed before the British term was popularized in this country in the mid 1980s, we in America called it "gestalt birding" not "giss birding." But, it is a gestalt that arises from the physical nature and behavior of the bird. The giss of a bird in flight is created by its physical structure and the dynamics and patterns of its movements. There is nothing mystically incomprehensible about it. All of these attributes such as "wingtip elevation," "wrist angulation," "flap rate," "bound duration," etc. are in fact components of gizz, crystalized and quantified. If you see a "gizz" difference between two videos, you should be able to quantify what is creating it and use this for real, scientific, non-mysterious comparisons to other videos. You can also pick up other consistent, taxon-specific attributes that might not be obvious to the naked eye, such as apparent wingtip shapes, the geometry of the wings on the upstroke, etc. This is the additional information that is available in a video that compensates for the generally lower image resolution. Failing to take advantage of this is a very bad idea. Videos should be examined as a whole, with each frame in context of the temporal sequence, and the added dimensions of time and kinetics used to their full advantage.

As a side note here, many people have mentioned that flight style is a "soft" character, subject to variation. Of course, this is true. But it is subject to variation only within limits. Birds don't learn how to flap fro scratch; that behavior is hard wired. Each species has a set range of flight styles that it can vary within. One will never find a Pileated Woodpecker flying like a Ruby-throated Hummingbird no matter how hard one might look. So, of course if you only have two examples to compare, there's little you can say. But if you have a large suite of comparison material you will get a much better sense of what the range of this "soft" character is, and can in fact determine if an unknown bird is within or outside of this range.

Finally, a point that was illustrated well by this recent video. When you do have the species identification correct, and you have suitable comparison material, you will see everything about the video in question fall in line behind this ID in short order. Allowing for minor glitches and transient illusions, every frame and every feature will be seen to be readily explained by and consistent with the hypothesis that the bird in the video is actually of species X. Once I had a suitable Red-headed Woodpecker video, the recent Pearl video immediately lined up with it so well (frame by frame and in its totallity) that it was clear there was little more to discuss.

It is interesting to note that this has never happened with the Luneau video and attempts to line it up with a Pileated. Only by misinterpreting image artifacts, ignoring flight style and wing dynamics, and focusing on select out-of-context frames can one even begin to line them up. Still, five years later, no one has yet produced a video of a Pileated in flight that even approximately matches the bowed-winged downstrokes of the Luneau bird. It took me one trip to the woods and 30 minutes to get a Red-headed Woodpecker video that was an exact match to the recent Pearl video, in plumage, structure, and movement. The contrast between these two experiences is informative.


At 12:47 PM, Blogger cyberthrush said...

One further tidbit that caused many to jump to conclusions:
Mike initially wrote a cryptic entry on his log saying "Got the sucker!" (or something like that) implying to many that he'd gotten the clearcut pic everyone wants; he followed that with an entry saying he'd attained film that was "quantum leaps" above anything previously produced. He took down these entries within around 24-48 hrs. but no doubt they misled many who saw them.

At 12:57 PM, Blogger Bill Pulliam said...

Yes, I think that was a big part of what contributed to the assumption that he had seen the bird well enough to be certain it was a large woodpecker. When I later heard that he was reported to have said that when he first glimpsed it he thought it was a flicker, that changed my perspective dramatically. Flicker is much closer to Red-headed in size than Ivorybill!


Post a Comment

<< Home

Site Meter