31.07.2015 Views

Steven Pinker -- How the Mind Works - Hampshire High Italian ...

Steven Pinker -- How the Mind Works - Hampshire High Italian ...

Steven Pinker -- How the Mind Works - Hampshire High Italian ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

244 | HOW THE MIND WORKSabout projection (how objects appear in perspective} and someprobabilities about <strong>the</strong> world (what kinds of objects it has). Some of<strong>the</strong> probabilities about projection are very good indeed. A penny, <strong>the</strong>oretically,can project to a thin line, but it does so only when il is viewededge-on. If <strong>the</strong>re's a penny in <strong>the</strong> scene, what is <strong>the</strong> probability that youare viewing it edge-on? Unless someone has choreographed <strong>the</strong> two ofyou, not very high. The vast majority of viewpoints will make <strong>the</strong> pennyproject an ellipse instead. The shape analyzer assumes <strong>the</strong> current viewpointis generic—not poised with pinpoint accuracy to line things up,Ames-style—and places its chips accordingly. A matchsiick, on <strong>the</strong> o<strong>the</strong>rhand, will project to a straight line almost all <strong>the</strong> time, so if <strong>the</strong>re is a linein an image, a stick is a better guess than a disk, all else being equal.A collection of lines in an image can narrow <strong>the</strong> odds even fur<strong>the</strong>r, forexample, a set of parallel or near-parallel lines is seldom an accident.Nonparallel lines in <strong>the</strong> world rarely project near-parallel lines in animage: most pairs of sticks strewn on a floor cross at moderate to sharpangles. But lines that are parallel in <strong>the</strong> world, such as <strong>the</strong> edges of atelephone pole, almost always project near-parallel lines. So if <strong>the</strong>re arenear-parallel lines in an image, <strong>the</strong> odds favor parallel edges in <strong>the</strong> world.There are many o<strong>the</strong>r rules of thumb that say what kinds of seulptings of<strong>the</strong> world can be counted on to give off various markings in an image.Little Ts, Ys, angles, arrows, crows' feet, and parallel wiggles are <strong>the</strong> fingerprintsof various straight edges, corners, right angles, and symmetricalshapes. Cartoonists have exploited <strong>the</strong>se rules for millennia, and a wilyshape analyzer can run <strong>the</strong>m backwards when betting on what is in <strong>the</strong>world.But of course running a likelihood backwards—saying that parallelstuff usually projects near-parallel images, <strong>the</strong>refore near-para lie Iimages imply parallel stuff—is unsound. It is like hearing hoofbeats outsideyour window and concluding that <strong>the</strong>y came from a zebra, becausezebras often make hoofbeats. The prior probability that <strong>the</strong> world containssome entity—how many zebras are out <strong>the</strong>re, how many paralleledges are out <strong>the</strong>re—has to be multiplied in. For an odds-playing shapeanalyzer to work, <strong>the</strong> world had better contain lots of <strong>the</strong> straight, regular,symmetrical, compact kinds of objects that it likes to guess. Does it?A romantic might think that <strong>the</strong> natural world is organic and soft, itshard edges bulldozed in by <strong>the</strong> Army Corps of Engineers. As a literatureprofessor recently declared to his class, "Straight lines on <strong>the</strong> landscapeare put <strong>the</strong>re by man." A skeptical student, Gail Jensen Sanford, pub-

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!