快猫短视频

SENSING what is up and down should be really straightforward. After all, it鈥檚
such a vital thing to know. It guides how we walk, sit, stand, and tells us what
things will fall over and where. But when you enter the old tilted shack on the
side of a hill in Santa Cruz, California, your view of the world changes. Balls
seem to roll uphill, a pendulum hangs off to one side, and people look like
they鈥檙e standing at impossible angles as if suspended by ropes.

The fun-park owners who built the lop-sided shack call it the Mystery Spot.
They have known for decades that the sight of slanted walls and floors is enough
to distort someone鈥檚 sense of the vertical so strongly that even sceptics get
the feeling they can defy gravity. And all around the world similar attractions
testify to the fascination that visual trickery has for us.

But why should what we see influence our sense of what鈥檚 upright? After all,
with our somatosensory system we can feel the pull of gravity on our limbs and
through our feet or seat, and we have fluid in our inner ears that swishes
around as we tilt our head, acting like a sort of spirit level鈥攖he
vestibular system. But when normally upright things such as walls, floors,
doors, windows and furniture appear to lean over, we must somehow be ignoring
this information. Perception researchers have long been trying to work out why
our eyes can let us get something as seemingly obvious as up and down so
wrong.

While tourists stumble about the Mystery Spot, William Prinzmetal, a
perception psychologist at the University of California at Berkeley, is busy
measuring the angles of the walls. He is studying visual illusions such as these
to discover what type of stimuli cause people to misjudge which way is up. And
he鈥檚 come to a rather surprising conclusion.

Multiple perceptions

For years psychologists believed that powerful illusions such as those at the
Mystery Spot depended on our finding tilted objects that our brains are used to
seeing upright. But more recently they have found that slanted lines alone are
enough to skew our perception of 鈥渦p鈥�.

Now Prinzmetal is taking things a step further. He says some of the first
neurons that work on the input from the eyes are are so sensitive to vertical or
near vertical lines that they can actually be working on more than one
approximation of vertical at a time. The resulting variations in this vertical
sense across the visual field, even though entirely unconscious, may cause all
sorts of visual trickery.

Normally there鈥檚 no conflict between the vertical we see and the one we feel,
so it came as a surprise to many when, in 1941, psychologists Herman Witkin and
Solomon Asch of Brooklyn College in New York showed just how easily our eyes can
fool us. They built a small room, about the size of a shipping crate, open at
the front so you could see in. It was furnished with a table and chair, and
propped up at one end so that the whole thing tilted about 22掳 to one side.
Then they asked someone to sit looking into the room, from a level seat on the
floor of the lab.

The experimenters, hiding behind the back wall, rotated a metre-long rod
until it looked perfectly vertical to their observers. Despite instructions to
ignore the room and to align the rod to gravity, the viewers routinely picked an
angle of between 5掳 and 20掳, somewhere in between real vertical and the
angle of the tilted walls.

Witkin and Asch assumed that the familiar objects in the scene were
important to the illusion, that the subjects鈥� extensive experience with upright
walls, tables and chairs made the illusion robust. If that were true, it would
mean our sense of orientation must rely on fairly complex mental processing. The
objects must be perceived and recognised and their normal orientation recalled
for the brain to work out which way 鈥渦p鈥� ought to be in the scene.

That view has persisted for many years. It seems logical that we would learn
all about vertical as we learn all about the objects around us. But experiments
with a different sort of test room in the early 1990s profoundly changed that
view. Leonard Matin and Wenxun Li, psychologists at Columbia University in New
York, discovered that horizontal and vertical lines on the wall of a tilting
room, with no familiar objects in view were sufficient to trigger all the
orientation illusions of the Mystery Spot.

Matin and Li created a powerful illusion with a large upturned box. But
rather than furnish the box like a room, they merely painted the walls with
horizontal and vertical lines like the mortar pattern of a brick wall. The
viewer sits on a stool inside the box facing the rear wall. As the box has no
floor, the viewer remains upright while Matin and Li tilt the box. In this
particular experiment, the researchers pitch the box forwards and backwards,
rather than tilting it to one side like the Witkin and Asch room. With the room
pitched to different angles, the viewers were asked to adjust a point of light
on the back wall to eye level.

When the box room is level, so that the brick pattern is horizontal and
vertical with respect to gravity, it鈥檚 easy to judge eye level accurately. But
when the box is pitched forwards 30掳, so that the ceiling is sloping down,
the subjects seem to think they are looking straight ahead when in fact they are
looking down. Although the degree of distortion varies from person to person,
the average angle of error is 20掳鈥攔oughly the angle between the top
and the bottom of this page if held at arm鈥檚 length.

How to shrink your friends

The box room illusion is the same as one of the most powerful effects at the
Mystery Spot. As well as leaning at impossible angles, people can seem to shrink
and grow. In normal surroundings, your perception of eye level is accurate. But
in the box room with the pitch forwards, or in the yard at the Mystery Spot, you
would stare your partner in the chest and yet feel as if you were looking
straight ahead. Your friend would seem head and shoulders taller. If the room is
pitched the other way, you look right over the top of your now diminutive
friend鈥檚 head.

鈥淲e can shrink people and make them grow in no time,鈥� says Matin. 鈥淲e can
show you balls going uphill too. Almost all of the phenomena at the Mystery Spot
are reproducible here.鈥� So you don鈥檛 have to recognise objects to lose your
vertical orientation. What鈥檚 important is lines.

Just last year, Prinzmetal tested this further, by showing subjects an
oblique line on a computer screen. Although they had instructions to look only
at the screen, they would see the frame of the monitor and the walls of the room
with their peripheral vision, giving them ample clues to true vertical. Yet when
the subjects tried to align two dots horizontally or vertically, they were drawn
off several degrees by the oblique line鈥攔ather like the early
rod-orientation experiments. Prinzmetal speculated that people judge vertical
from nearby visual cues, ignoring more distant ones even if they are more
numerous or more accurate.

One possible implication of this is that the deception caused by slanted
lines results from a local effect in a small portion of your visual field. It
needn鈥檛 mean you change your overall idea of vertical. In other words, you judge
whether a line is vertical by referring to others nearby. If you can see two
nearly vertical lines, they might both get flagged as good vertical references,
even if they are not quite parallel. In evaluating other lines, your visual
system might compare them only to the nearest line flagged as a vertical
reference, leading to very slight discrepancies from one side of the visual
field to the other.

And there could easily be more than one version of vertical in the mind: one
we consciously use when adjusting a rod in a psychology experiment or a picture
frame on the wall, and another unconscious one that lets us walk around without falling over
(see 鈥淚rresistible Illusions鈥�, 快猫短视频, 5 September 1998, p 32).

Prinzmetal reasoned that if we judge lines by comparison with their
neighbours, that could account for several classical geometrical illusions. In
the Ponzo illusion, for example, two converging oblique lines make one
horizontal line in between look longer than another of equal length. In the
1960s, Richard Gregory, a psychologist at Bristol University, proposed that the
converging lines suggest depth, rather like train tracks disappearing into the
distance, and therefore trigger a perceptual mechanism that enlarges the
apparently more distant line.

Prinzmetal sees it differently. Imagine a pair of short horizontal lines of
the same length, one squarely above the other. Place a vertical line next to
them just a short distance away on the left. Now if you tilt that vertical line
slightly so the top gets closer to the horizontal lines, your local impression
of vertical tilts too, and the left end of the top line will appear to protrude
very slightly more to the left than the left end of the bottom line. Since the
right-hand ends still seem more or less vertically aligned, the top line now
appears longer than the bottom. If you add a line tilted the opposite way to the
other side of the figure, the illusion roughly doubles.

Gregory says he agrees that too little attention has been paid to how
vertical orientation affects illusions, but cautions against overestimating its
importance. 鈥淭here are very few experiments on it,鈥� he says. 鈥淚 think it鈥檚 worth
颈苍惫别蝉迟颈驳补迟颈苍驳.鈥�

Whether or not Prinzmetal is right about geometrical illusions such as the
Ponzo, his skewed line experiments suggest where to look for some of the neurons
that keep us oriented. Since lines are the most basic features recognised by the
visual system, Prinzmetal suspects the neurons involved are among the first in
the brain to process visual images. Suppose certain neurons in the visual
cortex, the region at the back of the brain that analyses visual information,
fire only when a vertical or near vertical line falls on the retina. Such
neurons might label a line as a good reference for vertical, even if it鈥檚
slightly askew.

Do such neurons exist? In fact, in the very first stage of visual processing
the image is split up into lines. Neurons in a cortical area called V1 monitor
the visual information hitting specific regions of the retina and fire when they
detect lines of a certain orientation. But these cells aren鈥檛 enough on their
own. Their preferred orientation moves with the retina, and so with the head. If
your head is tilted, a V1 cell that happens to be sensitive to lines running up
and down the retina from the top to the bottom of the eye is not going to be a
reliable indicator of what is vertical. Prinzmetal would expect to find a type
of neuron beyond area V1 that senses the vertical with respect to gravity, no
matter how the head is tilted.

At least one group of neurologists has recently identified neurons that seem
to fit the bill. Xavier Sauvan and Esther Peterhans at the University of
Z眉rich found cells in other regions of the visual cortex of monkeys that
keep firing in response to a single line, even when the head, and so the retina,
is tilted. In an article soon to appear in Visual Cognition, they
describe how they found these cells in cortical areas V2 and V3, just downstream
from area V1 in the visual processing pathway. Peterhans says these cells must
be integrating non-visual information, probably from the inner ear, to keep
track of the line鈥檚 orientation with respect to gravity.

Could cells like these be responsible for picking out the lines that help us
tell which way is up? Quite possibly, says Columbia鈥檚 Matin. The brain must
integrate vestibular information with the visual information at some point, he
says, because the retina by itself doesn鈥檛 collect enough information to orient
an animal to gravity. 鈥淵ou can think of it as film in a camera,鈥� he says. 鈥淲hat
happens when you turn the camera? You鈥檝e got to have information about where the
camera is.鈥� But Prinzmetal wants to go further. He contends that we don鈥檛 just
use vertical lines to judge up and down, but that we use them to estimate the
properties of lines that seem to have nothing to do with orientation, such as
the length of the top line in the Ponzo figure.

Recently he has begun testing this idea. To see whether vertical orientation
and the Ponzo illusion are linked, Prinzmetal adapted one of Witkin and Asch鈥檚
experiments from the 1940s. They had shown that a subject looking at the tilted
room while sitting in a tilted chair would adjust the rod even farther away from
true vertical. Prinzmetal has now found that tilting people in front of a level
computer screen magnified the errors made in aligning dots next to oblique
lines. He and postdoc Diane Beck have also shown people a Ponzo figure on a
computer screen and asked them to adjust the bottom line until they thought it
was the same length as the top line. On average, upright subjects made the
bottom line 6 per cent too long, but tilted subjects made it 11 per cent too
long.

Although no one knows why tilting magnifies orientation illusions, the
experiment suggests that something similar is going on both in the tilted room
experiments and the Ponzo illusion. It may be that with gravity tugging on the
body in an unfamiliar way, the brain relies even more heavily on the visual cues
than normal.

One thing is clear: the way our view of the world can be distorted is giving
researchers some interesting leads as they try to work out the neurological
basis of perception. That鈥檚 what makes illusions so useful, say Prinzmetal and
Beck, they tell you something about how the brain works normally. 鈥淗umans didn鈥檛
evolve the ability to see these illusions just so we could have fun,鈥� says Beck.
But if you visit the Mystery Spot, you鈥檒l be glad we did.

PSYCHOLOGISTS Ian Howard and Heather Jenkin of York University in London,
Ontario, are studying visual orientation in hopes of reducing space sickness for
astronauts during long stints without gravity. Using a simple optical illusion,
they have managed to convince people that they are upside down.

Jenkin packs people into a chair with lots of padding, so they feel equal
pressure all around and so that somatosensory information is virtually
eliminated, as it would be in zero gravity. In front of the chair is a model
room in which everything is nailed down. Jenkin rotates the room slowly so that
the subject sees the wall on the left go down and the floor start to come up on
the right. Since rooms don鈥檛 normally rotate, says Jenkin, most subjects believe
their chair is rotating. Some people are misled more than others. 鈥淲e鈥檝e had
people do 180掳,鈥� says Jenkin.

Space sickness hits when you lose your orientation. It can strike even
experienced astronauts, Jenkin says. 鈥淵ou鈥檙e working away when suddenly someone
floats by in a different orientation and you realise you aren鈥檛 vertical.鈥�
Jenkin is trying to work out which cues should be included in the International
Space Station to keep the astronauts oriented. Particularly helpful, she says,
are vertical lines, a different colour floor and human faces.