Lecture 2 - Depth: The Problem Foundations of Mind...

Info iconThis preview shows page 1. Sign up to view the full content.

View Full Document Right Arrow Icon
This is the end of the preview. Sign up to access the rest of the document.

Unformatted text preview: Depth: The Problem Foundations of Mind ! 200.336 ! 2/2/2012 ! Lecture #2 ! Depth Perception 1 ! Slide# 2 Space Percep2on is Hard Some Problems of Visual Space Percep2on • We perceive a stable, con2nuous, 3D spa2al layout How can we perceive 3 ­D from 2 ­D input? • Percep2on seems immediate, effortless & nearly error ­free • The mechanisms of percep2on are complex and puzzling 4 4 3 9 2 4 2 3 4 4 8 9 4 4 4 5 4 3 4 4 5 4 4 4 4 8 All senses operate on contact (e.g., surface touches skin, molecules interact with tastebuds) But, Percep2on brings us knowledge of things at a distance… How? Slide# 3 4 4 9 8 9 8 9 4 4 4 4 4 4 4 4 2 4 4 8 Slide# 4 1 Some Problems of Visual Space Percep2on Some Problems of Visual Space Percep2on (1) Depth (2) Size Slide# 5 Some Problems of Visual Space Percep2on 6 The Problem of Depth (3) Posi2on and mo2on Object mo2on We agree: Depth is not directly observable! Observer mo2on Slide# 7 Slide# 8 2 Monocular Cues to Depth Monocular Cues to Depth Interposi2on Linear perspec2ve Rafael, 1509 ­11 Slide# 9 Ocular Cues to Depth Slide# 10 Binocular Cues to Depth (discussed by Descartes and Berkeley) (discussed by Descartes and Berkeley) Accommoda2on Convergence NEAR in focus FAR in focus Slide# 11 12 3 The Ques2on Two Theories of Space (Depth) Percep2on: What makes us able to interpret these cues so as to perceive depth and spa2al layout? The human mind and brain are built to infer space automa2cally. Perceptual processes are like reasoning processes: geometric inferences Two different answers: We are endowed with special ­purpose cogni2ve systems for interpre2ng these cues in the right ways Core knowledge of depth/ space Na2vism/ ra2onalism Percep&on of depth involves innate unconscious mental computa&ons We are endowed with a general ­purpose capacity to learn about our surroundings. This associa2ve learning leads us to construct appropriate interpreta2ons of spa2al cues. No core knowledge of depth/ space… Empiricism Na2vism (Descartes) Slide# 13 Two Theories of Space (Depth) Percep2on: Slide# 14 Descartes’ Theory of Vision: The human mind and brain built to sense only impinging s2muli (e.g., light, muscular effort) We learn through experience to interpret these s2muli in terms of depth by associa2on with touch: no reasoning involved! demo Percep&on of depth involves rote learning, associa&ve pairing of convergence & accommoda&on with the sense of TOUCH Empiricism (Berkeley) Slide# 15 Slide# 16 4 Descartes’ Theory of Vision: Berkeley’s Retort (1709): René Descartes (The Op&cs, 1637) philosopher, mathema2cian, scien2st Analogy: the blind man “I appeal to any one’s experience, whether, upon sight of an object, he compute its distance by the bigness of the angle made by the mee2ng of the two op&c axes? … In vain shall all the mathema&cians in the world tell me, that I perceive certain lines and angles which introduce into my mind the various ideas of distance; so long as I myself am conscious of no such thing.” Descartes’ theory is: •  Mechanis2c (light, refrac2on, nerves, signals to brain) •  Computa2onal (geometric inferences) •  Ra2onalist and na2vist (“natural geometry”) Slide# 17 Berkeley’s Idea of “Effort”: Slide# 18 Berkeley’s Associa2ve Theory: “Less effort” “More effort!” Accommoda2on NEAR in focus: more efforhul FAR in focus: less efforhul FAR fixated: less efforhul NEAR fixated: more efforhul Convergence Slide# 19 Method: Introspec2on Primi2ves: Sensa2ons (e.g., muscular effort) What is Learned: Associa2on between ocular effort, vision, touch Berkeley’s theory is: •  Empiricist •  Associa2onist Slide# 20 5 Two Theories of Space (Depth) Percep2on: Hermann von Helmholtz on the Origins of Space Percep2on (1850’s): The idea shared by Ra2onalism and Empiricism is that humans do not know things directly but grasp only their impressions (sense data). Q: Do we perceive depth by innately structured computa2onal mechanisms or by associa2ve learning? Both require inference to perceive depth: A: To answer, must study space percep2on in infants and children Ra2onalists adopted mathema2cal deduc2on, Empiricists experien2al induc2on. But we cannot do this! Next best thing: mess around with adult percep2on Slide# 21 Helmholtz on the Modifiability of Percep2on Slide# 22 Helmholtz on the Modifiability of Percep2on (1) Natural observa2on: walking in the forest with new glasses Helmholtz’s next ­best strategy: Study the modifiability of space percep2on in adults (2) Experimenta2on: prism adapta2on • Alter the spa2al rela2onship between the perceiver and the layout so that visual space percep2on is systema2cally in error • Observe whether the perceiver adapts to the altered rela2onship & corrects the error Poin2ng with prism If no adapta2on occurs: The mechanisms of space percep2on are fixed in adults…. most reasonable to suppose they are fixed in infants If adapta2on occurs: Adults are able to learn to perceive space correctly… most reasonable to suppose babies learn too Before adapta2on Slide# 23 Aler adapta2on Slide# 24 6 Helmholtz on the Modifiability of Percep2on More on Prism Adapta2on: Harris (1960’s) (1) Where does the perceptual change occur: in vision or in touch? If vision learned by associa2on with touch, as Berkeley claims, then vision should change in adapta2on experiments Perceptual change or conscious decision? Test via alereffects 10 Poin2ng without prism Adaptive Shift In Degrees Poin2ng with prism Before adapta2on Aler adapta2on Before adapta2on 8 6 4 2 0 -2 Aler adapta2on Conclusions: Space percep2on is modifiable in adults. On grounds of plausibility, probably learned in children….? 25 Slide# Prism Adapta2on Conclusions Visual Targets Auditory “Straight Targets Ahead” (2) how does the perceptual change occur? By associa2ve pairing or ac2ve mo2on? if Berkeley’s associa2ve theory is correct, then associa2ve pairing should suffice Slide# 26 Depth Percep2on: Developmental Methods It’s impossible to know what infants think or perceive (b/c they don’t do much). (1)  Mechanisms of space percep2on are modifiable, but touch is more modifiable than vision (2) Modifica2ons depend on ac2ve, self ­produced mo2on If studies of prism adapta2on in adults shed light on visual space percep2on in infants, then ini2ally Helmholtz’s studies show some support for Berkeley. But Harris’ later work shows that the phenomena more accurately offer more support for Descartes’ na2vist theory Von Helmholtz (1850) Robert Fantz’s Answer (1958): 2 things that infants DO do: 1) Look around 2) Prefer to look at interes2ng things to boring things  ­ ­Touch doesn’t teach vision  ­ ­Learning is not by passive associa2on Slide# 27 Slide# 28 7 Depth Percep2on: Developmental Methods 29 Developmental Method #1: Looking Time 30 Developmental Method #1: Looking Time (1) Infants, adults, everyone likes to look at things they find interes2ng Preference for faces? (2) Looking Time Method: Show newborns different displays and measure how long they stare at them (3) If they look longer at one thing than another, it means infants can discriminate the two. Preference for paoerned over homogeneous displays Newborn infants can see How well can they see? Robert Fantz (1958) 31 32 8 Developmental Method #1: Looking Time Developmental Method #2: Preferen2al Looking Tes2ng visual acuity Robert Fantz (1958) The preferen2al looking method: ??? left image right image peephole 33 Developmental Method #2: Preferen2al Looking 34 Developmental Method #2: Preferen2al Looking Ques2on: Can we use this method to measure visual acuity in infants? Oooh..!  ­ ­Vary stripe width  ­ ­Find the smallest width at which infants of a given age prefer stripes over grey 35 36 9 Developmental Method #2: Preferen2al Looking Developmental Method #2: Preferen2al Looking Aaah..! Wow..! 37 Developmental Method #2: Preferen2al Looking 38 Developmental Method #2: Preferen2al Looking No preference when stripes are too narrow Yawn...! Newborns’ limit = 2º of visual angle (as wide as 2 thumbs at arms length) 39 40 10 Developmental Method #2: Preferen2al Looking Developmental Method #2: Preferen2al Looking Fantz’s measure gives a lower bound es2mate of newborn visual acuity Previously: Visual acuity…. Can also be used to measure contrast sensi2vity Real acuity could be higher: babies may see thinner stripes but not prefer to look at them More recent measures using EEG (see Kellman):  ­ ­Narrowest width: 1° Wow..!  ­ ­Striking qualita2ve agreement with Fantz’s findings: Rapid change in acuity over first 6 months… Good acuity & slower change aler that 41 Developmental Method #2: Preferen2al Looking 42 Developmental Method #2: Preferen2al Looking Previously: Visual acuity…. Can also be used to measure contrast sensi2vity Previously: Acuity, contrast… But what about DEPTH? disc sphere Yawn...! Longer looking at sphere at all ages tested (1 ­6 months) 43 44 11 Developmental Method #2: Preferen2al Looking The General Problem Limita2ons of Fantz’s method: What would na2vists/empiricists predict? Na2vists and empiricists agree that infants have some sensory capaci2es (else learning would be impossible). But disagree about what inexperienced infants perceive. Descartes: ini2al state = depth from natural geometry Berkeley: ini2al state = confusion, effort Both predict preference for sphere, for different reasons! “Whoah, confused!” Berkeley’s predic2on The existence of a visual preference does not reveal the basis for that preference. Helmholtz’s skep2cism returns: How can we learn what infants perceive? “Wow, 3D!” Descartes’ predic2on 45 Two Approaches to Solving this Problem 46 Signatures of Depth Percep2on in Infants? (1) Psychophysical signatures of depth percep2on: STEREOPSIS (Held, Birch & Gwiazda, 1980) 1) Psychophysical signatures of depth percep2on in infants Differences between the two views of a scene at the two eyes provide informa2on for depth: binocular disparity 2) Equivalence of different cues to depth 47 48 12 Signatures of Depth Percep2on in Infants? Signatures of Depth Percep2on in Infants? 1) Psychophysics: Put stereo glasses on babies (1) Psychophysical signatures of depth percep2on: STEREOPSIS (Held, Birch & Gwiazda, 1980) Differences between the two views of a scene at the two eyes provide informa2on for depth: binocular disparity Most adults are highly sensi2ve to this cue  ­ ­Threading a needle  ­ ­Detec2ng counterfeit dollars •  One side = no disparity (flat); other = disparity (3 ­D) •  Vary size of disparity on different trials •  Measure preference for side seen as 3D by adults Are infants sensi2ve to binocular disparity informa2on for depth? 49 Signatures of Depth Percep2on in Infants? 50 Signatures of Depth Percep2on in Infants? But are infants preferring DEPTH?? The sphere ­vs. ­disc problem: what is the basis of infants’ preference? Longitudinal results: Stereopsis begins at 3 ­4 months. Acuity rises rapidly therealer. “Ooh, Depth!” “Yikes, Double images!” Held’s strategy: test for the signatures of mature stereopsis: Three signatures of stereoscopic depth percep2on in adults: 1) Hyperacuity (stereo acuity > regular acuity) 2) Limited range (when dispari2es get too big, we see double images) 3) Horizontal disparity only (ver2cal = double images) 51 52 13 Signatures of Depth Percep2on in Infants? Q: Hyperacuity? Test 2 ­D versus 3 ­D acuity in same infants. A: Stereo acuity IS higher than simple acuity for stripes. (Yes) Q: Upper limit? Test with disparity too large for adults to see as fused objects in depth. A: Infants show no preference. (Yes) Q: Horizontal disparity only? Rotate displays by 90º. A: Infants show no preference. (Yes) 53 14 ...
View Full Document

{[ snackBarMessage ]}

Ask a homework question - tutors are online