Hi--
Sorry it took so long to get back - I wasn't sure
whether my reply would be long or short, and the
"busy's" got me......

> I think the most important area of agreement
> is that "rectalinear" or rectangular perspective
> is not based on fidelity to vision, but to what
> we know about parallelisms in the subject matter
> based not on seeing but on measurement *at* the
> subject. This conflict is essentially not
> resolvable, and is emphasized by wide fields of
> view forced onto flat planes for viewing.

Yes, and over narrow angles the conflict is less
apparent, since all perspective types converge in their
characteristics if the angle of view is made small.

> One interesting experiment I made was to shoot
> an interior with a 14mm, a fisheye adapter on a
> 50mm, and a Widelux, all without moving the
> tripod. I was not concerned with which had the
> greatest horizontal coverage, since I could have
> altered that by using a 13 or a 15mm, or mounting
> the fisheye adapter on a 45 or 58mm lens, or by
> using a full 360 pan camera instead of the short
> rotation Widelux.
>
> I then chose the widest horizontal coverage that
> was included in all negs, based on inclusion of
> the same two objects as markers for the left and
> right field limits. This had to be done at the
> mid-height of the frame because of the fisheye.
> All three negs were enlarged so that these left
> and right markers spanned the same distance in
> all prints. If I had chosen a diagonal distance
> using the upper corner marker of one side and
> lower corner marker of the other, the inclusion
> of the fisheye would have altered the match that
> I established for marker at the mid-height of the
> frames. Since there is no real reason to prefer
> bisecting the frame horizontally or diagonally, I
> fudged the size of the fisheye print to somewhere
> between the two.
>
> I was not really interested in any "scientific"
> study of optics and imaging, so perfect accuracy
> was not important. My efforts were simply to
> minimize variables because the variable I wished
> to observe would hopefully be more evident that
> way. What I wished to see was the visual effect
> of the three methods of perspective. The final
> prints are about 2.25:1 aspect so the fisheye is
> not screaming "fisheye!!" and the "egghead", or
> corner distortion of the 14 is not on the print.
>
> The main differences visually are that the 14 is
> rendering the outmost objects much larger than
> does either of the other optics, while the fish
> boosts the size of central objects more than the
> other systems. The swing-lens fall between the
> other two and offers the most natural look. I
> find the 14 makes the center look miniaturized,
> the fish makes middle 50% look about right but
> puts the squeeze on the other 50% that resides
> toward both sides.
>
> This is all most unfortunate, my prefering the
> swing-lens image, because that type of camera is
> the least flexible: odd format, no use of flash,
> usually fixed focus, miserable viewfinders, and
> limitted slow speeds with no time exposures. At
> this point, the Quaker Oats Pinholer is looking
> like the camera most likely to produce an image
> that is comfortable both to the eye and the mind.
> As I said, I am not really seeking to establish
> a scientifically acceptable standard for imaging
> accuracy, but am trying to find the wide angle
> method whose images look most natural. Natural,
> to me, means I can include a very wide subject
> field with its wonderful sweep of miscelaneous,
> often "mismatched" elements, and not have the
> final produvt scream "Ultrawide Effects" which
> always distracts the viewer from the "crazy-
> quilt" effect of gathering many eclectic subject
> elements into one frame.
>
> David Rosen golem@capital.net

Hmmm, I agree with your technique, and with your
conclusions. Though technically the fisheye can be
shown to be the best approximation of vision, the
conditions under which it is a good approximation
make the swing-lens view a more attractive (and
familiar) approximation. There are two issues:
- What type of image "looks right"?
- Is it valid to use image types that do not
"look right"?
In the first case, we have all noticed the difficulty
of finding an image type that satisfies many of the
requirements for "looking right" - the choice of a
narrow angle of view is about the only way to satisfy
most people. But some of us are very aware of being
able to see over a very wide angle, and the swing-lens
camera view requires a minimum of cropping, subject
selection, etc. to make it acceptably natural-looking
for many people.
In the second case, I think it is reasonable to use
image types that are obviously quite different from
"looking right", and effectively use the
characteristics of those "unusual" image types for
making photographs (which may have little "naturalness"
to them). It is easy to show that virtually no
photograph is an accurate record of any subject.
Therefore, as in other media, the need to stay "natural"
is really an issue related only to the belief (not the
reality) that photographs are good records of subjects.

Shall I post the above?