[Prev][Next][Index][Thread]

Re: (meteorobs) Opik

> 
>> I think it's even worse: the Opik method breaks down at rates above maybe
>> 1,000
>> per hour (with really good timing), but then ANY human counting method breaks
>> down around 4,000 per hour
> 
> I was curious to hear more about this last point, Chris: are you aware of any
> perception studies which tend to point to this as an upper limit for
> "counting"?

No, this is my own observation based on my experience on the Leonid MAC
mission. But the problem was pretty basic: just about everybody reported
that after about ten minutes of pressing the mouse button every time they
saw a Leonid -- about once per second -- their fingers were getting pretty
sore. I had chosen a very light mouse button to minimize this problem, but
it definitely hit us.
 
> Also, it may or may not be signficant that the observing conditions for you
> and
> your fellow Leonid MAC team members were very, very different from those that
> most observers face in the field - perhaps better, or perhaps more
> distracting?

Yes, the conditions on board the aircraft were quite different from those on
the ground, but the basic problem of muscle fatigue will bother an airborne
finger as readily as a grounded one.

I will also note a problem that I have been unable to nail down, but which I
am certain exists. Whenever a meteor crosses your field of view, your
attention immediately zeroes in on the meteor -- making you blind to
anything else in the field of view. Note that I am not saying that your
eyeball itself moves -- it does move a bit but the real problem is in the
visual processing region of your brain. The behavioral manifestation of this
struck me when I ran a training video for observers for my ill-fated visual
observations of the 2001 storm. I was using videos taken from the 1999
Leonid MAC mission. At one point in the video, a rather bright meteor zips
across the upper right of the screen, followed a fraction of a second later
by a medium-brightness meteor in the opposite quadrant of the screen. I
myself had noted that second meteor only after watching this five-minute
clip dozens of times. Out of five people using the training video, not one
person saw the second meteor.

Now let's assume then that there's a half-second dead time immediately after
seeing a meteor, during which the brain is so busy recognizing the meteor
that it is simply blind to anything else. If the rate were, say, two meteors
per second, then you'd expect to miss about half the meteors. The actual
calculation is far more complicated because they'd be distributed randomly
over time. But the point remains that a great many meteors will be missed.

Especially complicating this is the fact that people often do catch
simultaneous meteors. I believe that this applies only with bright meteors
or meteors that are close together. To put it rigorously, I refer to the
three-dimensional graph showing the probability that an observer will see a
meteor as a function of its brightness and its distance from the center of
his field of view. Obviously, this surface slopes downward from a peak at
the origin. However, I am suggesting that, during the refractory period
immediately after seeing a meteor, the surface drops downward. It's a very
short-lived phenomenon, and would never really show up at normal rates, but
when we start getting rates higher than one per second, it could have a
serious impact on our calculations. It may therefore be necessary to have
very precise timings of the meteors, then statistically correct the reported
rate by looking at how the observed distribution for counts per interval
deviates from the Poisson.

I am tempted to build a device that I considered for 1999: a voice-driven
switch that the observer uses with very short enunciations. You can say
"dit-dit-dit" faster and with less fatigue than you can press a mouse button
three times. Also, because the brain-to-tongue connection is shorter than
the brain-to-finger connection, there's less "backlash" in the data
reporting. Still, I'm concerned about "back-queued" reports messing up the
statistics.

Chris

The archive and Web site for our list is at http://www.meteorobs.org
To stop getting all email from the 'meteorobs' lists, use our Webform:
http://www.meteorobs.org/subscribe.html
References: