Symposium on unidentified flying objects: hearings before the Committee on Science and Astronautics, U.S. House of Representatives, Ninetieth Congress, second session, July 29, 1968.

233
The
Use
of
the
Computer
in
the
Design
of
Recognition
Test
-Materials
and
in
the
Analysis
of
Results
Our
sample
of
only
206
actual
UFO
cases
is
really
too
small
and
haphazard
for the
purpose of ensuring that
all
types
of
reported
shapes are
adequately
represented in
any
proposed
recognition
array.
Many
descriptive
terms
that
have
repeatedly
been
used
(
such
as
"doughnut,"
"ring,"
"mushroom,"
flattened
ball,"
"double
-convex
lense," "bullet,"
"blimp,"
and
"submarine")
didn't
happen
to
ap-
pear
more
than
once
in
our
particular
sample).
Ideally,
for
this
work,
one would
like
access to
a
centralized library
of
all
reasonably
documented
cases—suitably
coded
for
retrieval
via
computer. Indeed,
at
present
the
scientific
study
of
the
UFO
problem
is
greatly
hampered by
the circumstance that
the thousands
of
reported
sightings
have
not been
adequately
coded or
systematized
in
any
uniform
way
and
are, in
fact,
still
scattered
among
such
diverse
and
often
mutually
hostile
organizations
as
the
U.S. Air
Force,
NICAP,
APRO,
and
the
University
of
Colorado
Project—not
to mention
a
number
of
more
-or
-less
private
files
assem-
bled
by
individual
investigators
both here
and
abroad.
Recent
developments
in
computer
technology—particularly in
computer
graphics—could be
utilized,
also,
in
the
construction
of
arrays
of
shapes
for
a
recognition
test.
Thus
for
any
specified
shape,
the
computer
(together
with
suita-
ble
graphical
output
equipment)
could automatically generate
alternative
pic-
tures
of
the
same
object
as
viewed
from
any
desired
angle
(e.g.,
Noll,
1965;
Zajac,
1964);
and
could
even
generate other
test
shapes
intermediate
between
that
shape
and
some
other
specified
shape.
(In
fact,
as
on-line
graphical
facilities
become
more
widely
available,
even
a
relatively
unartistic
witness, seated in
front
of
a
sui.able
display
device,
should
be
able to reconstruct
his
own
object
by
tech-
niques of
these general
sorts.)
For
the
present,
however, perhaps
the
most
promising use of the
computer
in
this
connection
would
be
in
finding
an
optimum
arrangement of the
alternativetest
shapes
in
the
recognition
array.
This
is
a
matter of
real
concern
owing
to
the
large
number
of
shapes
that should
be
included.
(Even
the
63
exhibited in
the
above
figure
fall
far
short
of
covering
all
the
varieties
that
have
been
sketched
or
described.)
If
the
alternatives
could
somehow
be arranged
so that
similar
shapes are
close
together,
then
the witness could
quickly
narrow
down
to the
most
relevant region
of the
array
in
order
to
make
his
final,
most
refined discrimi-
nations.
In
order to
do
this
we
would
first
need
to
obtain
some
measure
of the
perceived
similarity
between
any
two
shapes.
One
could
of
course
obtain
a
direct,
subjective
judgment
of
similarity
from
experimental
subjects.
However,
it
might
be
more
convenient
to obtain
a
derived
measure
of
similarity
hased
upon
the frequency
with
which
different
subjects
will
sort
the
two
shapes
into
the
same
pile,
or
upon
the overlap
in
their
application
of
the
same
descriptive
terms
to
the
two
shapes
in
the kind of
task
described in
the
preceding
section
(cf.,
Rosenberg,
Nelson,
&
Vivekananthan,
in press).
Once
we
have
any
such measure
of
similarity
for every
pair,
we
can
apply
powerful
new
computer
-based
methods
for
mapping
the
objects
into
a
two-dimensional
arrangement
in
such
a
way
that
their
similarities
are
preserved, in
so
far
as
possi,ble,
in
the
spatial
proximities
among
them
(Kruskal,
1964,
Shepard.
1962; Shepard
&
Carroll,
1966).
Moreover,
these
same
methods
could
yield
a
quantitative
metric of
similarity
that
would
then enable
us
to specify just
how
similar
an
object
identified
by one
witness
is
to
the
object
identified
by
another
witness.
Indeed, they
could
even
tell
us something about
the
basic
dimensions
along
which
UFO
phenomena
differ
or,
with the help
of
recently perfected
methods
for
"hierarchical clustering"
(Johnson,
1967), they could provide
an
indication
of the
basically different
classes into
which
these
phenomena
undoubtedly
fall.
Possibly,
some
of these
classes
of
unidentified aerial
phenomena
will
turn out
to
be of
purely natural
origin.
I
once even ventured to suggest
this
for
certain
puzzling
types of
cases myself
(
Shepard, 1967b)
—though,
admittedly, attemptsto develop
such
explanations in
terms of
known
principles
of
atmospheric
physics,
generally,
have run
into
competent and
serious
criticism
(McDonald,
1968).
Still.
even
if
some
of the
phenomena
arc of natural
origin,
a more
complete
and
accuratecharacterization
of
their
appearance
and
,behavior
should be of
some
interest to
the physical scientist—indeed,
all
the
more so to the extent that they appear
toconflict
with
known
physical principles.
In any
case,
it
appears that techniques
now
exist that could provide the basis
for
a
psychologically oriented,
but genuinely
scientific
investigation into uniden-
tified
aerial
phenomena, whatever
their nature
may
ultimately prove to be.
97-818
0-68-16