Sie sind auf Seite 1von 4

Close Encounters -- An Artist Shows that Size Affects Shape Author(s): Denis G.

Pelli Source: Science, New Series, Vol. 285, No. 5429 (Aug. 6, 1999), pp. 844-846 Published by: American Association for the Advancement of Science Stable URL: http://www.jstor.org/stable/2898587 . Accessed: 10/04/2013 04:26
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at . http://www.jstor.org/page/info/about/policies/terms.jsp

.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship. For more information about JSTOR, please contact support@jstor.org.

American Association for the Advancement of Science is collaborating with JSTOR to digitize, preserve and extend access to Science.

http://www.jstor.org

This content downloaded from 14.139.62.114 on Wed, 10 Apr 2013 04:26:51 AM All use subject to JSTOR Terms and Conditions

YSCIENCE'S
PERSPECTIVES: VISUAL SCIENCE

COMPASS

PERSPECTIVES

ancient mosaics at Delos and Pompeii. However, noneusedcoarse grids to render three-dimensional shape, so there is no duThemosaics ality. andthepointillist grids aretoofine to readily into flat disintegrate marks, andthebenday screens areuniform, Denis G. Pelli so that arealways they flat at all distances. e easilyrecognize of all objects Nearly all of Close's paintings are of Theonly precedent for the ofClose's duality shapes andsizes,yetno onehas headsproduced byphotographing thesub- recent heads may be the 4th long-lost centuanyidea howwe do it.It seems ject,drawing a gridontothephoto and a ry(B.C.E.) skiagraphia paintings byApolobvious that wemust see shape inthesame similar gridontothecanvas, andmeticu- lodorus in which he achieved intermediate wayregardless ofsize,otherwise we would lously copying one squareat a time from colors ofunbyjuxtaposing large patches recognize ourfriend ortheletter "a" differ- phototo canvas(see cover,thisissue). mixedcolorsthatblended whenviewed ently at eachsize or viewing distance (1). Seenfrom a great the is a from distance, portrait a distance (8). Yettheblockportraits bytheartist Chuck visually accurate reproduction ofthephoThe ChuckClose retrospective (2) exClosevividly showthesize to. In his earliest paintings, Close copied hibits scores ofblock incolor portraits, half ofshape dependence percep- details within eachsquare so that theorigi- (likeBill II) andhalfin blackandwhite, tion.Whenviewing anyof nal gridis notvisiblein theresult, but with a widerange ofmarks across theface Close's1987-1997 portraits since 1973 he has usually filled each (5 to21). facesize(2 to200cm), mark size at their actualsize (2), one squareof thegridon his (0.5 to 9 cm), and mark 200 canmove forward andback, canvas with content that is type. Thisvariety allowed x again and again, and the independent oftheoriginal 100us toundertake a parametX face, solidfrom afar, Closerefers toeach always photo. ricinvestigation ofthesize x collapses into flatmarks filled square as a "mark." dependence of shapeperU) Flat whenseen from near.The duality (solid In 1973,thescientists ception.We used a psyx from afar andflat from near) ofthese paint- Harmon and Juleszpubchophysical "nosetest" to ings shows that sizeaffects the perception of lishedblock portraits of measure the transition from )XX thepopular shape,disproving Lincoln assumption Abraham that were flatto solid in 33 Close that shapeperception is size-independent. important in vision science portraits. While at looking x We havereproduced a recent Close block (5). Their"critical-band" ~10the painting, the observer is Bill 1I portrait, atone-third (1991), ofitsac- explanation of theblock- co asked tomove forward and U. tualsize(see figure, opposite page)toallow portrait effect asimplicitly backward tofind theviewSolid readers to experience thedramatic effect of sumessize invariance and ingdistance at which the size on what they perceive. thus predictsthat face Psychophysical nose from emerges the can2 measurements ofobservers viewing Close's recognition requires a cervas. (Theinstructions em7 10 70 block reveal portraits the ofsize, tainnumber importance of marks per phasizethebridge of the Marks across the face that the effect is visual not (perception) opti- face independent of face nosetominimize the effect cal (physics), andthat itinvolves a competi- size (butthey onlytested The nose test. The criticalface ofthe which nostrils, Close plotted usually tionbetween thefaceand itsconstituent the effect atonesmall size). widthof Close's portraits renders with detail the number of marks across blocksto engageourperception of shape Close,too,showed hisfirst against that ofthegrid.) exceeding lines,one from shading. block portrait (dotdrawing)the face.The regression From afar, the nose(illumifor each offive observers (accountAristotle that (3) noted shape perception in 1973(6). Unlike Close's from nated the side) aping for half of the variance),are couldbe independent ofsizeonly for sizes recent the early block plotsofthe results work, pears to stick outfrom the for nose judging that areneither so huge as toexceed ourvi- portraits by Close, and emergence canvas. As the viewer apon each of the 33 gridsualfield at those by Harmon and ded portraits (-135? visualanglesubtended thenosesuddenfrom the Chuck Close proaches, theeye)norso tiny as to exceed ourvisual Julesz, had small(<1 cm) retrospective a becoming (exceptthe Keith/Six ly collapses, acuity(0.10). As visual acuity(thefineness blocksthat viewers never DrawingsSeries) (2). X and 0 are flat ofunevenly colpatch ofvision)is largely determined byoptics, approached closely enough rawdata fortwo observers. skin. Thetransition is Size in- ored theassumption vision among scientists has toexperience thefull duali- dependencepredicts The observeris a vertical line. abrupt. beenthat observers theshape ofthe ty(7). Instead, identify viewers ex- All five lines have log-log slopes asked to find thetransition retinal (blurry) ofits perienced only a weak close to 1 (mean 1.0, SD ? 0.2), point, imageindependently andthecritical dissize.Thisage-old seems to be one-time assumption effect, elegantly showingthat the perceivedshape tance is measured from the common supported byboth and described by Harmon: does dependon size. experience observer's eyetothepaintpriming experiments (in which thespeed "Viewed from close up, ing.Clearly, sizedoesmatandaccuracy oftarget identification areun- these 'block portraits' appear to be merely ter because thefaceis perceived differently affected sizedifference bythe between ofsquares,. . . [but]once a target an assemblage atdifferent sizes(viewing distances). andprime images) (4). notto Whenviewing all possible face is perceivedit becomesdifficult blockporsee it,as ifsomekindofperceptual hystere- traits there is a division into twodomains: sisprevented theimage from once again dis- Thefaceis seenas flat in onedomain and Theauthor is intheDepartment ofPsychology and solving into an abstract pattern of squares." in solid the other. We the mapped boundary the Center forNeuralScience,NewYork UniverThereare well-known antecedents to dividing the two domains toreveal howsize sity, New York, NY 10003, USA. E-mail: blockpaintings suchas thecoarse"benday" affects denis@psych.nyu.edu shape. For a given painting, the num-

Close Encounters An Artist ShowsthatSize Affects Shape

on the GrandeJatte(1884-1886), and the

screens usedinRoyLichtenstein's cartoons ofthe1960s, A Sunday Seurat's pointillist

844

6 AUGUST 1999

VOL 285

SCIENCE

www.sciencemag.org

This content downloaded from 14.139.62.114 on Wed, 10 Apr 2013 04:26:51 AM All use subject to JSTOR Terms and Conditions

berofmarks acrossthefaceis fixed. The critical facewidth in degrees is theangle subtended bythefaceatthe critical viewing distance. We plotted critical face width against the number ofmarks across the face. The fivesolid linesrepresent theresults from theviewing of 33 blockportraits by fiveobservers (see figure, opposite page). There areobvious differences between observers, presumably because eachmust set his or herowninternal criterion fornose If shapeperception emergence. weresizeindependent, theplotwouldbe a vertical line.Instead, themodest positive slopes indicate that peopleneedmore marks to see thenosestick outon larger faces. Thefive lines haveslopescloseto 1,demonstrating that critical facewidth is proportional to number ofmarks across theface.From the slopeofthelinesit is clearthat themark size(facewidth indegrees divided bynumberofmarks across the face)is constant. So, divides theflat from thesoliddomain. Portraits areseenas flat when marks arelarger Thetype ofmark usedmayaccount for muchof theresidual scatter of thepoints about thelines inthegraph. Closehastried many ofmark, types andthey affect the critical distance at which thenosecollapses. Closeportraits areall thesamesize,based on thesamephotograph andgrid, butuse very different marks (watercolor findots, inkstickscribbles, gerprints, and white Contecrayon) and havedifferent critical
The Keith/Six Drawings Series (1979) of
1.59

itis thecritical marksize (roughly 0.3?) that

than0.3?,and solidwhenmarks are smaller than 0.3?,for facesofall sizes.

z z 3 u
9 0

Butcanthesizeeffects ofClose's paintDV O HZNK _ ingsbe explained bysimple Vision optics? 0.39 0.24 0.15 0.09 0.50 0.32 0.21 0.13 scientists, takingsize independence for granted, havesupposed that increasing the Faces are but a galleryof pictures.BillII(1991), a blockportrait byChuck Close,reduced to one-third viewing distance reveals the facesimply be- of itsactualsizeand cropped. Compare itsappearance from nearand far(>5 m) orcompareitwith the causeit increases theblur, outthe tinycopy of Bill11(see oppositepage,top). Ignorethe pupils, wiping nostrils, and the line betweenthe lips, grid. off By taking one'sglasses (orputting whichhavemuchhigher resolution thanthe 1.3-cm grid thatrepresents therestoftheface.Beloweach on someone onecanblur else's), theimage letter in the eye chartis a number indicating itssize (the observer's acuity)as a fraction of a mark(a (remove the grid) andthe face is revealed, as filled squareofthe painting's grid). (Oil on canvas, 92.4 x 76.2 cm.Photograph byBillJacobson.) viewers of Close'spaintings often discover for themselves. But,whereas onecould(at scaleonfigure, this page).Butwhen the ob- cal to,itsblur with the2-mm pupil.(The leastin principle) walkfarenough away server simply increases viewing distance, retinal illuminances areequated byadding a from a blockportrait to achieve thesame without defocus, thegridis stillapparent 25% transmission neutral density filter to bluras thewrong glassesprovide at short whenthenose pops out,at an acuity of the2-mm Ifweperceived pupil.) theretinal in fact, thesize effects distances, that are about a quarter mark. image's shape independently ofitssize,then most salient in theClose exhibition occur To unequivocally concludethatper- thecritical image size should double when over modest distances the ceived (<6 m) at which shapedepends on thesizeperse of we double theblur. In fact, we find that the blur is only a fraction eye's ofa mark. Read- theretinal image, we tested a condition in critical imagesizes areidentical, showing erscantry this for themselves the whichtheretinal byfinding imagechanges onlyin that retinal image size,notblur, determines ofnoseemergence with points andwithout size.We compared critical distances while the perceived shape. their andcomparing theappearance looking glasses, through either a 1- or 2-mm pinObservers must see marks both substanof Bill II and their visualacuity artificial (size of hole;these pupils aresufficiently tially larger andsmaller than 0.3?toexperismallest readable between thetwo smallto maketheeyediffraction-limited letter) enceduality. To collapse reliably, a portrait conditions. Defocus (blur) reveals thenose (thequality oftheimage depends only on composed of 0.4-cm marks (likeHarmon only when itcompletely smears out the grid, pupil size)(9). Theeye's blur with the1-mm andJulesz's must be viewed Lincoln) from at an acuity ofabout onemark (see acuity pupil is twice as bigas,but otherwise identi- less than 40 cm,closer than most viewers
www.sciencemag.org SCIENCE VOL285 6 AUGUST1999 845

mark sizes,ranging from 0.20 to 0.7?.

O VR
K
1.01 0.80

1.27

0.64

This content downloaded from 14.139.62.114 on Wed, 10 Apr 2013 04:26:51 AM All use subject to JSTOR Terms and Conditions

SCIENCE'S or larger, as in will come. For 2-cm marks, Close's recent1987-97 portraits (see cover, this issue) that distance is at least 2 m, whichmostviewers crossas theyapproach. Some formsof camouflage,like a tiger's maybreakup theanimal'sshapeonstripes, ly whenseen from verynear.Most perception textbooksshow a spottedDalmatian, initially lostin a background of spotty shadows,butwhichusuallyappearsquickly and nevergoes away(like Harmon's description of seeingLincolnwithsmallblocks)(10). I find that theDalmatian, likea blockportrait, does breakup reliably intomereflatspots whenenlarged to makethe (or approached) spotspacingexceed0.3?. Testinga wide range of sizes revealed that thedivision between seeinga blockporas flator solid occursat a cnrtical mark trait size of 0.3? (which is independent of the number of marks per face).This refutes the of shapeperception size invariance and Harmonand Julesz's critical-band theory of the effect. It seemsthat theblocks block-portrait (or their edges) (11) competewiththe face to capturethe visual shape-from-shading process.The size and typeof themarks determine their powerof attraction. This comdetermined petitionis bottom-up, by the not top-down, controlled by the stimulus, observer. Close concedes that,paintingat evenhe cannot arm'slength, see thefaceunless he backsaway(2). One mightsuppose that Close was a naive artist, obsessed by grids,who innocently producedthe coarsely gridded paintingsthatwe use hereto revealthe size deof shapeperception. In fact, Close pendence has devoted his careerto studying just that: "Theselfportrait fom 1967-68 is the first head thatI painted.... The idea portrait was to make something thatwas so large thatit could notbe readily seen as a whole andforcetheviewer to scan theimagein a wereGullivBrobdingnagian way,as ifthey overthesurface ers Lilliputians crawling of theface,fallingintoa nostril and tripping over a mustachehair" (2). He was more thanhis scientific thorough colleagues;the insize of the marksin his block portraits creased by 15% per year from1973 (0.4 exhicm) to 1997 (9 cm). He made surethat bitionsof his workwould conveythe idea, a retrospective thatcouldnotprocanceling vide long viewing distances. So credit thissize-deChuckClose withdiscovering breakdown of our ability to extract pendent well within thebounds shapefrom shading, ofourvisualfieldand acuity.
yearsago,in 1. In 1886,Ernst Machwrote, "Someforty I proposed a societyof physicists and physiologists, simifordiscussion the question, whygeometrically I resimilar. lar figures werealso optically [visually] to member quitewellthe attitude takenwithregard thisquestion, whichwas accountednot onlysuper-

COMPASS

5. L.D. Harmon and B.Julesz, Science180, 1194 (1973); L.D. Harmon, Sci.Am.229 (no. 5), 71 (1973); in 0. J. translated tions, by C. M. Williams and S. Waterlow Grusser and R.K. Klinke (Eds.),Pattern Recognition in (RoutledgelThoemes, London, 1996),p. 109]. Biologicaland Technical Systems(Springer-Verlag, 2. ChuckClose exhibition organizedby RobertStorr, NewYork, 1971),pp.196-219. Museumof Modern Art, New York City, 26 February 6. Keith/1,280, exhibitedat the Bykert Gallery, New to 26 May 1998. The exhibition is at the Hayward 20 to 15 November York City, October 1973. 22 July to 19 September, 1999. All Gallery, London, 7. Harmon (1971) includes a face with 0.4-cm squares. in relevant paintings, exceptMaggie, are reproduced The blocksare 0.4 cm in Harmon and Julesz's (1973) the catalog[R.Storr, ChuckClose (Museumof ModLincoln (cover of Science), 0.8 cm in Harmon's (1973) ernArt, New York, 1998)]. The Close quote is from ofScientific George Washington (cover and American), narration at theexhibition. recorded provided 1.0 cm in his MonaLisa.Close's Keith/1,280 (1973) is 3. "Beautydepends on size and order;hence an exmadeup ofdotson a 0.4-cmgrid. could not be beautiful, tremely minutecreature for 8. E. C. Keuls [Plato and GreekPainting (Brill, Leiden, ourvision becomesblurred as itapproaches the point Netherlands, 1978)], on the basis of textsby Plato of imperceptibility, norcouldan utterly hugecreature andAristotle. be beautiful, unableto take it in all at once,the for, 9. F.W.Campbell and R.W.Gubisch,J. Physiol. 186, 558 viewerfinds that its unity and wholenesshave es(1966). caped hisfield ofvision" [Aristotle's Poetics, translat- 10. Dalmatian photoby R.C. James, The in R.L Gregory, ed byJ.Hutton NewYork, 7]. (Norton, 1982),chapter NewYork, Intelligent Eye(McGraw-Hill, 1973),p. 14. 4. E. E. Cooper, I. Biederman, Can.J.PsyJ.E. Hummel, 11. M. C. Morrone, D. C. Burr, J.Ross,Nature305, 226 chol. 46, 191 (1992). Unlikeidentification, judging (1983). whether one has seen a particular object beforeis 12. I thankS. Bernardete forGreekcitationsand D. R. Fordiscussion size-specific. of the size question, see the artificial Williams forsuggesting pupil. Supported www.visionscience.com/mail/cvnet/199810307.html.bythe National EyeInstitute.
fluous, but even ludicrous" [The Analysis of Sensa-

PERSPECTIVES:

GEOPHYSICS

Hawaiian PlumeDynamics
views aboutEarth'sinterigeologists' or. The apparently fixed position of the Hawaiian "hot spot" led to the theory that deep-seated plumes of hot, buoyant mantlewere responsibleforocean island volcanism at Hawaii and manyotherocean islandchains(1). Chemicaland isotopic differences betweenocean island basalts and mid-ocean ridge basaltshavelongbeenused bygeochemists to constrain modelsofmantleconvection and thechemical evolution of Earth(2-4). In thisissue,Blichert-Toft eta!. (5) present evidencefrom hafnium isotopes ancient suggesting that deep ocean (pelagic) in thesourceof some sediments are present in its own right, Hawaiian lavas. Important thisresult also suggests thatcombined geochemical and seismologic study of the Hawaiian"plume"mayhelp resolveone of themostimportant and long-standing quesscience:whether of tionsin earth convection Earth's mantle is layered. Evidence forlayeredmantleconvection comes primarily fromgeochemistry. Mass balance appearsto require thata substantial in of Earth's mantle is less depleted portion in the continental elementsconcentrated crust[suchas large-ion lithophile (LIL) elethanthehighly ments] depleted uppermantle sampledat mid-ocean ridges(2, 4). Rare especiallyforAr gas isotopedistributions, and He, also suggest thatpart of Earth's mantle retains a largefraction of itsprimoras well as a largefraction of dial gas budget, the 40Ar producedby the decay of 40K (3, thisreservoir forthe 4). To have preserved age of the Earth, the reservoirmust remain convectively isolatedfromtheupper where theprocessesof crust formamantle, tion at mid-ocean ridges and island arcs have stripped a large fraction of the initial A reasonable raregas and LIL elements. locationforthisgas- and LIL-richreservoir is therefore the lowermantle.The change in mantle that occursat a depthof mineralogy 660 km,indicated by a seismicdiscontinuity, was long believedto act as a barrier to of cold downtransfer convection, blocking wellingslabs or hotupwelling plumes.The 660-kmdiscontinuity was therefore thought to markthe boundarybetweena depleted lower upper mantleand a more primitive mantle (6, 7). This model is not consistent withrecent seismic tomographicimages, which are as indicating thatmany widelyinterpreted slabs do not stopat the 660-km subducting to descenddeep discontinuity, but continue intoEarth'sinterior (8, 9). If a substantial fraction of subducted slabs have penetrated intothelowermantle formuchofEarthhischemicallayersignificant long-lived tory, to preserve(10). However, ing is difficult seismic tomographycan only provide a of thecurrent thermal structure of snapshot the mantle.Earthhas been slowlycooling and there forthepast4.5 billionyears, is no a priori could reasonwhymantle convection not have been layered formostof the geologic past even if today such layering apdown(4). pearsto havebroken The mounting evidenceforancient recyin cled crust and sediments the Hawaiian Theauthor is at theMax-Planck-Institute for Chemhow seismologists and geoPostOffice Box3060,55020 Mainz, istry, Germany. plumesuggests forces to constrain the chemists can combine E-mail: lassiter@mpch-mainz.mpg.de SCIENCE www.sciencemag.org

The Hawaiian Islands havelongshaped

JohnLassiter

References and Notes

846

6 AUGUST1999 VOL 285

This content downloaded from 14.139.62.114 on Wed, 10 Apr 2013 04:26:51 AM All use subject to JSTOR Terms and Conditions

Das könnte Ihnen auch gefallen