Popular in Course
Popular in Psychlogy
This 153 page Class Notes was uploaded by Marco Wolf on Monday September 7, 2015. The Class Notes belongs to PSY 323 at University of Texas at Austin taught by Staff in Fall. Since its upload, it has received 39 views. For similar materials see /class/181815/psy-323-university-of-texas-at-austin in Psychlogy at University of Texas at Austin.
Reviews for PERCEPTION
Report this Material
What is Karma?
Karma is the currency of StudySoup.
You can buy or earn more Karma at anytime and redeem it for class notes, study guides, flashcards, and more!
Date Created: 09/07/15
Sense SeeMg Heanng Bamnce Touch Pose Smell amp Taste Major Sensory Systems Source of information Light Sound Gravity and acceleration Temperature and pressure Joint position and muscle stress Chemical structure Visual system Auditory system Vestibular system Tactile system Kinesthetic proprioceptive system Olfactory system Gustatory system These are some of the sensory systems in humans Overarching Principle Perceptual systems evolve to obtain information about the environment that is relevant forthe tasks the organism must perform in orderto survive and reproduce Corollary The design of a perceptual system is constrained by the tasks it performs by the physicalstatistical properties ofthe environment and by various biological factors Important Perceptual Tasks Identification of objects and materials Navigation through the environment Prediction of motion trajectories Estimation of physical dimensions Object manipulation Speech communication Visual communication Vision and hearing are very important for human survival They involve many important and complex tasks Visual pathway of the macaque monkey whose visual system is very similar to the human visual system Note that the visual areas take up a very large percentage of the monkey s cortex In humans it is a smaller percentage but still very substantial The optic nerves contains 23 of all sensory neurons in humans Major neural connections between the Visual areas in the previous map of the macaque monkey s Visual pathway This diagram hints at the complexity of the perceptual systems 53 Twisted cord illusion The twisted cord is not a spiral but a series of concentric circles Even seemingly simple tasks such as judging the curvature or slant of contours involve complex processing Another twisted cord illusion The letters are not tilted Demonstration of active 1 l t 39 grouping 39 39 by JL l 39 One sees a uctuating collection of circles of various diameters Vision involves active complex subconscious processes movements Demonstration of local adaptation effects Note the close connection of the effects to eye A t sage w v Complex natural scenes that have never been seen before are quickly and accurately interpreted by the visual system Famous painting by Bev Dolittle The Forest Has Eyes demonstrating how good humans are a detecting meaningful structure in images Five Dif cult Problems for Vision Systems Illumination problem The illumination of scenes is highly variable and complex Depth problem The images in the eyes are twodimensional projections ofthe three dimensional environment Context problem Objects often appear in a complex and varying context of other objects Viewpoint problem Objects are rarely seen from the same viewpoint Category complexity problem The specific objects that define a category are often quite different Fundamental Biological Constraints Limited neural resources dynamic ranges and physical space Most natural tasks involve dealing with one or more of these difficult general problems Similar problems hold for the other perceptual systems Furthermore the solutions that the perceptual systems can come up with in natural tasks are constrained by various fundamental biological factors Natural example of importance of solving the context problem in order to recognize familiar objects Recurrent Themes Perception is a very complex process Perception generally involves the integration of many sources of information most of which are not very reliable There are many approaches to the study of perceptual systems and each has made important contributions to our understanding Approaches to Understanding Perception Natural tasks Natural scene statistics Anatomy Responses of individual neurons Responses of neural populations Perceptualbehavioral performance Mathematical and computational modeling The eye encounters a enormous variety of complex visual images Yet the brain somehow manages to correctly interpret almost every visual image it receives It is able to correctly identify objects materials and surface shapes as well as shadows and other lighting effects It does this effortlessly for scenes never encountered before Underlying this remarkable ability is a set of sophisticated perceptual grouping mechanisms that try to link together image features that arise from the same physical source To understand why grouping mechanisms are essential it is useful consider context problem the visual brain must overcome Five Dif cult Problems for Vision Systems Illumination problem The illumination of scenes is highly variable and complex Depth problem The images in the eyes are twodimensional projections ofthe three dimensional environment Context problem Objects often appear in a complex and varying context of other objects Viewpoint problem Objects are rarely seen from the same viewpoint Category complexity problem The specific objects that define a category are often quite different Fundamental Biological Constraints Limited neural resources dynamic ranges and physical space Most natural tasks involve dealing with one or more of these difficult general problems Furthermore the solutions that the visual system can come up with in natural tasks are constrained by various fundamental biological factors Five Dif cult Problems for Vision Systems Context problem Objects often appear in a complex and varying context of other objects Most natural tasks involve dealing with one or more of these dif cult general problems Furthermore the solutions that the visual system can come up with in natural tasks are constrained by various fundamental biological factors Context problem Objects often appear in a complex and varying context of other objects making recognition ofobjects difficult Solution First measure attributes of the images in small regions Second combine the small regions into wholes using rules that are related to the physical laws and statistical facts of nature and to past experience This abstract object contains a recognizable object Erasing the bottom half ofthe abstract object reveals the recognizable object Why are we unable to recognize the squareroot of 16 The context of other contours causes us to group the contour elements in a way that prevents the squareroot of 16 from making contact with our stored memory for that object The implication is that proper grouping is essential for object recognition THISISTRUEEVENWHENTHERE ITSACCOMPANIMENTISPLAYED THEVISUALAREAWHENFIGURES DISCONTINUOUSPARTSBECOME QUITEHOMOGENEOUSGROUND Text from Wertheimer 1923 Another demonstration from one of the founders of Gestat psychology of how perceptual grouping is critical for recognition EVENiTHOUGHiALTERNATIVE ORGANIZATIONSiMAYiBE EASIERiHEREiTHANiINiTHE PRECEDINGiCASEsiITiIsiSTILL TRUEiTHATiAiSPONTANEOUS NATURALiNORMALLYiExPECTED COMBINATIONiAND SEGREGATIONiOCCURsiAND 0THERiORGANIZATIONSiCANiBE ACHIEVEDiONLyiRARELYiUNDER PARTICULARiCONDITIONSiAND USUALLYiWITHiSPECIAL EFFORTiANDiSOMEiDIFFICULTY Text from Wenheimer 1923 Any way of marking word boundaries helps the recognition process Gestalt Grouping Principles Proximity Similarity Good continuation Closure Proximity objects that are nearby tend to be grouped together Similarity objects that are similartend to be grouped together Good continuation contour elements that are consistent with a smooth contour tend to be grouped together Closure Contours that are consistent with a closed form tend to be grouped together I I I I I I I I Position I I I I Orientation I I I I Size OOOOOOOOCOIOI 0 0 0 0 O 0 Motion 0 0 e Gestalt grouping principles proximity position and various possible dimensions of similarity Wertheimer 1923 The Gestalt principle of good continuation Contour elements that are consistent with a smooth curve tend to be grouped together Gestalt principle of closure Because of good continuation the two straight line segments in A tend to look like a pair of crossing sticks Because of closure the same two line segments tend to be split at the middle to become parts of two butterfly wingsquot mm exemplal primitivedelectmn gmupmg i grouping ii I l o w a a i wagonquot 39 eego a 53635 i 0 matching Example of using simple feature detection followed by gestalt grouping rules to form groups in a novel texture pattern Le to right input image gt perform featureprimitive detection like V1 gt use goodcontinuation proximity etc to form small groups gt use shape similarity comparing or matching shapes to form larger groups from small groups Yjunction Tjunction J Arrowjunction 70 Ljunction Perceptual grouping also makes use of principles that are based upon the threedimensional properties of the environment For example these line segments are grouped into two boxes and a cylinder Object corners occluding a background object tend to form an L junctionquot or an Arrow junctionquot Object corners that do not occlude a background object tend to form a Yjunction Occluded contours of an object tend to form T junctionsquot with the contours of the occluding object These principles plus the Gestalt principles are used to group features into wholes that are likely to correspond to physically separate objects The result of appropriate typical grouping A result of inappropriate grouping Why do we have these specific grouping principles How does the brain implement these principles Are there other grouping principles the brain uses A good starting point is to examine the statistical relationship between the natural environment and the images formed in the eye To measure statistics of natural contours we analyzed 20 representative natural images closeups distant shots forests mountains ocean sky water elds animals Each image was analyzed separately and the statistics combined Each red pixel on the left shows an edge element detected using synthetic neurons with receptive fields like those in primary visual cortex V1 Kiwi if it Each short line segment in this image shows the position and orientation of an edge element detected using synthetic neurons with receptive elds like those in primary visual cortex V1 Only one tenth of the edge elements are shown reference d I ZgtltS e Three numbers are required to describe the geometrical relationship between two edge elements from an image the distance between the elements d the direction of one element from the other phi and the difference in orientation between the elements theta From all the edge elements in all the images it is possible to determine for each possible geometrical relationship how likely a pair of edge elements are to belong to the same physical contour 21 Contour Grouping Unlikely to come from same physical contour Likely to come from same physical contour For example these statistical measurements tell us that two of these elements are likely to belong to the same contour but the third is not If human contour grouping mechanisms accurately incorporate these statistical facts of the world then we should be able to predict human ability to detect contours from the statistics of natural images 22 response fixation field fixation field 500 ms 222 222 500 222 222 1000 ms An experiment to measure human ability to detect contours where the only information available in the stimulus for performing the task is the geometry of the edge elements 23 Example Target Contours 3 2 15 1 Fractal Exponent 525 125 25 50 RMS amplitude 20 4O 60 80 Length m G O 27 45 63 Jmer deg Examples ofthe different dimensions of contour shape that were tested in the experiment in previous slide In each case the contour was embedded in a dense background of random contour elements The connected contours on the right are the contour groups obtained using the statistics of contours in natural images Notice how they roughly match the contours you see in the left side In the forced choice experiment we predict that humans pick the interval with the longest group of edge elements 25 A Length 80 C Length 40 g 5 E g E g c g g 3 3 a amp u m u m lt t lt Orientation jitter range deg Orientation jiller range deg 8 Length 60 D Length 20 RMS amplitude RMS amplitude 5 3 g E g E a g e g 8 5 8 lt t lt t Orientation jitter range deg Orientation litter range deg Comparison of human performance solid symbols and the predictions open symbols from the statistics of contours in natural images 26 r091 39 90 i N 0 o o 80 39o 00 o 70 i o aquot 60 7 39 0 O O 0 O 50 Predicted Accuracy 50 60 70 80 90 Measured Accuracy There is a high correlation between human ability to detect contours and the model hypothesis based directly upon the statistics of contour geometry in natural images The neural circuits in the brain that perform contour grouping are unknown at this time although there are some hints from neurophysiological data 27 The problems of deafness are deeper and more complex if not more important than the problems of blindness Deafness is a much worse misfortune For it means the loss of the most vital stimulusthe sound of the voice that brings language sets thoughts astir and keeps us in the intellectual company of man Helen Keller The Miracle Worker Letter to Dr J Kerr Love 1910 in Brian Grant ed The Quiet Ear 1987 Being blind cuts you off from things Being deaf cuts you off from people Also attributed to Helen Keller The inability to hear is a nuisance the inability to communicate is the tragedy Lou Ann Walker A Loss for Words The Story of Deafness in a Family 1986 Got scarlet fever or meningitis at about 19 months of age Her teacher was Anne Sullivan Auncle Slapes Vesubmar newe Tym anum Round wmdQW p Midwe meme Eustachian tube akKlItory ear cavxlv General schematic of the ear Max Brodel1939 WW e ear 3 Posts 65 n Lat lt quot windcw wifh s raPeS Middle 88quot cavity Vesti bulsr nerve Facial nerve l Cuchlear nerve Outer ear Eardrum Ossicles Inner ear Middle ear Plnna Auditory canal Pinna and eardrum Directionai microphone a Middle impedance matching ear and overload protection Inner Frequency analysis ear hair cells Major functions of the structures of the ear Schematic of middle ear malleus hammer incus anvil stapes stirrup The purpose of the middle ear is to convert pressure waves in air at the ear drum into pressure waves in the fluid filled cochlea Without the middle ear only a small fraction of the pressure entering your ear would get into the cochlea Two factors help to covert air pressure to fluid pressure 1 the area of tympanic membrane ear drum is 20 times larger than the area of the stapes 2 the lever effect of the ossicles converts long movements of the ear drum into short movements of the stapes The muscle attached to the malleus is the tensor typmpani The muscle attached to the stapes is the tensor stapedius They help hold the ossicles in place and they implement the acoustic reflex which protects the inner ear from damage Max Bradel 19M Secreting epithelium cala vestibuli pevio c space endolic space evnal VeSH bular membrane Tecmrial membrane Spiral I spiral suicus gangnun Basil Spiral organ membrane Com Scala tympani perio c space sheath Hus om39hlvurls or main 1 Dmzrummnllc cross suclion at u uuclmnr c unl Tho In I d n c nlllllm Hm orgnn If L nrli with 5 half cells the Il nmlu mu humus ul ln39nriluz Frum llusmussnu 1943 1 an trminmumu mm mm m 1 n t m N m nmmm W quoter m Wlmrimllpinlhuml m m mm immmiiu mmm MM u mi N i Denequot ml in diam u u mm min K qunnalultunl mm Shows how tops of pillar cells tile part of the roof of the reticular lamina m a m an 2quot 394 quot Z mania a 14 mummmuummm mm W an up mummy a u m m m u m w a um n mm nu gt4 runWU 4 mm H Imam pmlw a w mm W w m u m l m a mum1m numbm m n quotwa m Mum m m m m M u m m a m a quotwarm u m m um mm M m mm n m Iw waunm mm mm m W nu m m quotAm an m Some Facts about Hair cells Outer Hair Cells OHCs rows of stereocilia in a W pattern about 120 150 stereocilia per OHC basally about 40 80 stereocilia per OHC apically lengths of stereocilia increase in height from base to apex and from inner to outer rows of cells at birth approximately 11200 16000 OHC are more susceptible to drugs acoustic trauma etc than lHCs curiously OHC row 1 is often the most vulnerable resting potential is about 70 mV lnner Hair Cells lHCs 24 rows of stereocilia in a straight line about 48 60 per IHC lengths of stereocilia increase in height from base to apex and from inner to outer on the cells at birth approximately 2800 4400 lHCs are less susceptible to drugs acoustic trauma etc than OHCs resting potential is about 45 mV Some Facts about Primary Fibers Primary auditory fibers About 30000 afferent fibers per ear All travel to and synapse in cochlear nucleus About 510 innervate OHCs Small diameter not myelinated Pattern is a combination of divergent one to many and convergent many to one About 9095 innervate lHCs Larger diameters myelinated for faster conduction times Pattern of innervation is divergent one to many Three subtypes high medium and low spontaneous rate Different frequencies cause responses at different places along the basilar membrane However the motion is in the form of traveling waves as demonstrated by Georg von Bekesy 20000 Base Frequency map Hz of the basilar membrane in the cochlea Note that traveling waves move from the base towards the apex Paradoxically the basilar membrane is narrowest at the wide beginning of the cochlea and widest at the narrow end of cochlea 47 mm mmmhlp hIwuz mm mm nuavvnnuH am 53mm v unitIanquot 5 mm mumInn I mm a Nann mlpa m m Wm m M m m w m Sammy Mus mum durlnn ml um kninlnnW mm m 1 mm m Wm m mu m Marm ms av m m an I m alumnus n ma awn nva le rm mmwgy om mom Imamm RealmMu a o my 91mm Archives momInyngmnny Annien MmeInmc1s12i p m rum pulms fur Inclnrm and basilquot Immbra nus i c mm Teclorial mg39mhmne i 7 Outer rm mm m inm han an Samar membrane new poaxtinu Hair cell receptors are stimulated when the basilar membrane moves up and down Speci cally the motion of the basilar membrane and tectorial membrane together create a shearing force that bends the hairs cilia of the hair cells cum haircells passive Outer mix cells active A D 1 1quot f HH l iH Aclion potentials m4 lintquot Action potentials are generated when the basilar membrane moves upward Anterior vertical Ssmimrcular FOSIEHDl canals vertical Horizontal Cochlear nerve Scale media Spiral ganglion Helicalrema Scala vestibuli Haircslls Relssner s membrane Organ of com Spir l Scale qmpam a ganglion Basilar membrane Structure of the cochlea and the vestibular apparatus The vestibular apparatus is composed of the semicircular canals the utricle and the saccule which sense angular acceleration linear acceleration and gravity The auditory and vestibular receptors are located in uid filled holes in the temporal bone 5w mm a cocmea Tecmnzl membuns 5W gangth Omar haw ce l nner m as V saw memhrana Scala wmvm chkmal mumbling Cross section of the cochlea In humans there are approximately 15500 hair cells in each ear 12000 outer and 3500 inner 2O When cilia are bent channels in the cilia membrane are opened mechanically 21 a 439 a 5 a 5 E 1 a 9 3 g I1VL 7 L n 2 2 2 4 5 7e 7 3 m w n z Dummy m Mm m m I r a a 1 a 20 Dmance hum stapes mm 25 H1 50 Hz 100 Hz 200 Hz 400 H1 300 H IBDD H1 Envelope of the Bekesy traveling wave The peak shifts toward the stapes as frequency increases 22 Anteriov vemcai Ssmlmrcular Posterior Vesubuiar nerve canals vertical 39cie Saccuie Mod 0 us Horizoniai Cuchlear nerve Scaia media Scala vestibuii Scaia media Hairceih Relssner s membrane Organofcom Scaia mpam Spirai ganglion It is possible to record from individual neurons in the cochlear 8th nerve 23 mnmy manquot M511 1mm p115 Note articulation of stereocilia Low and mediumspont fibers drop off more axon endings in ON than do Highspont fibers Lowspont fibers receive more efferent endings than Highspont fibers Myelin is acquired after passing thru the habenulum non i 300 Z 200 too 3 ion an u 3 W W m m 70 quot39 m A W F 309 i 3 e0 quot g 2 g 50 w 3 m g 00A In 3 0 r i n zo 5400 m X g 20 m i a I zoo 39 0 too 40 o m to no 40ng M a u moust mm Wquot 5 5m 0 4 Tuning curves and rulekvd lunumns for ute Imus ofsimllnr cr mu dmmu ski For ma mini the micrkvtl function wrc genamted by presuuinx SD ms Ion bursLs m i a my chtl wu incmlsed m 2dE stcps rwm he Iowa 10 um highm lcvnls ma m mu busts were prrsenld a an lml See mu ror I39unherdeimk mnmy manquot M511 19mm p115 Hi Spont gt18 spikessec about 75 of all fibers Low Spont lt 05 spikessec about 1015 of all fibers about 2060 dB less sensitive at CF Medium Spont in between Gain functions are not at CF Some Characteristics of the Three Categories of Primary Fibers Hiqh D Fibers Spontaneous firing rates of gt 18sec About 75 of all afferent fibers Most sensitive subpopulation Most widely tuned subpopulation Large ber diameter Contact Inner Hair Cell on outer side Gradual sloping gain functions Mitochondrion rich Few no contacts from e erent bers Least branching in cochlear nucleus Make contactwith halflhe number of cN cells as low sR ers Low D Fibers Spontaneous firing rates of lt 05sec About 10 15 of all afferent bers Less sensitive by 2060 dB SPL Least widely tuned Small fiber diameter Contact lHC high in inner side Steeply sloping gain functions Few mitochondria Large number orcontacts from efferent bers Most branching in CM ke contactwith twice the number of cN cells as high sR bers Terminate in the small cell cap see region in cm 26 Hum 4 LOW frequency frequermr A B C D E Comma Basiiar membvane ugmluue ldBSPU Tuning curves measured by nding the pure tone amplitude that produces a criterion response in an 8th nerve fiber Tuning curves forfour different neurons are shown 27 How does the basilar membrane move Temporal Telephone Theory Different frequencies are represented by the temporal response of the whole basilar membrane The basilar membrane responds like a telephone or microphone Place Resonance Theory Different frequencies are represented by different places along the basilar membrane The basilar membrane responds like the separate strings of a piano 28 arm mum elMive amplitude Ll 7 7L L n 71 22 w 5 7e 7 23 n w n 3 0mm mm mm m m a m 20 Distance hum stapes mm 25 H1 50 Hz 100 Hz 200 Hz 400 H1 300 Hz IBDD H1 One puzzle is that the tuning curves of 8th nerve neurons are too narrow to be consistent with the von Bekesy traveling waves One possible explanation is that von Bekesy s measurements were on cadavers 29 Outer hair cells To cochlear nucleus afferent From superior olivary complex efferent Another puzzle is that outer hair cells do not send much if any information to the brain Each inner hair cell is innervated by approximately 20 TypeI 8th nerve bers Each Type N 8th nerve ber synapses with about 10 outer hair cells but each outer hair cell synapses with several nerve fibers There are also approximately 900 efferent fibers fibers that come into the cochlea from more central locations One possible explanation for this anatomical mystery is that the outer hair cells are playing a very different role 30 BOr 70 9 I so J I g 50 I m 39u 40 AP threshold m Is kHz l3i34 dB SPL 30 053433 dB SPL I post marrem I I I Illllll l I l0 Frequency kHz Tuning curves at low noise levels high noise levels and after killing the outer hair cells There is increased sensitivity at lower sound levels when the outer hair cells are functional This effect is called the cochlear amplifierquot effect 31 Contraction and expansion of outer hair cells helps to amplify the responses of the basilar membrane in a narrow region This may be a large component of the cochlear ampli erquot 32 Max Brodel1939 WW e ear 3 Posts 65 n Lat lt quot windcw wifh s raPeS Middle 88quot cavity Vesti bulsr nerve Facial nerve l Cuchlear nerve 34 l l u lllvi39 Sound Ieval as A I I 0 mon zaoo aonu 4000 Frequency Hz Spontaneous otoacoustic emissions measured from two different ears another consequence of the cochlear ampli er 35 Click Stimulus IN OUT Id OZO sec gt Middle Ear Bones Cochlea Response from cochlea amplified 500X 36 1 snmums chck vv I F 35 kHz J 26 kHz J 20 kHz I 118 kHz W I l A I 0 1o 20 11mg ms Click evoked emissions a consequence of the cochlear amplifier 37 Some Characteristics of Otoacoustic Emissions OAEs Spontaneous OAEs SOAEs are essentially pure tones single frequency components emitted continuously by normalhearing ears SOAEs are more prevalent in females than in males About 7585 of females have at least one SOAE compared to about 4565 of males ClickEvoked OAEs CEOAEs and DistortionProduct OAEs DPOAEs are present in essentially all normalhearing ears and are stronger in females than in males SOAEs are more prevalent and stronger in right ears than in left ears CEOAEs are stronger in right ears than in left ears The sex and ear differences seen in adults are also seen in infants and children SOAEs appear to be highly constant through life SOAEs are more prevalent in people ofAsian and African extraction than in Caucasians SOAEs are generally quite weak and most people do not hear their SOAEs SOAEs are NOT the basis forthe ringing in the ears tinnitus that commonly accompanies hearing loss People having at least 4 SOAEs in one ear have hearing sens vity that is several decibels etter than people having no SOAEs Some form of OAEs is present in turtles frogs lizards and birds as well as in mammals 38 Some Readings on Otoacoustic Emissions OAEs McFadden D What do sex twins spotted hyenas ADHD and sexual orientation have in mon Perspectives in Psychological Science 2008 3 309323 McFadden D Masculinization effects in the auditory system Archives of Sexual Behavior 2002 31 93105 McFadden D Westhafer JG Pasanen EG Carlson CL and Tucker DM Physiological evidence of hypermasc 39n39za i n in boys with the inattentive type of attention deficitlhyperactivity disorder ADHD Clinical Neuroscience Research 2005 5 233245 McFadden D and Pasanen EG Comparison of the auditory systems of heterosexuals and homosexuals Clickevoked otoacoustic emissions Proceedings of the National Academy of Sciences USA 1998 95 27092713 McFadden D A masculinizing effect on the auditory systems of human females having male co twins Proceedings of the National Academy of Sciences USA 1993 90 1190011904 39 Approaches to Understanding Perception Nature tasks Natura scene statistics Anatomy Responses of individual neurons Responses of meura populations Perceptualbehavioral performance Mathei na39 cal and computational mode mg Behavioral Tasks A B C description objective feedback identification subjective no feedback estimation Different perceptual tasks can be classi ed by picking one attribute from each column 2AFC Task 13 a interval 1 interval 2 warning response feedback warning Example of an objective identi cation task With feedback If there are just two alternatives then such tasks are often called a discrimination tasks if one of the alternative is uniform in some fashion then such tasks are often called a detection tasks The observer must decide Whether a target pattern is in the first temporal interval or the second temporal interval STIMULI FEEDBACK TONE Igt m D m gt WARNING RESPONSE TONE INTERVAL 5 L0 30 E 2 20 39A E 09 k 3 E 15 Q 9 2 8 08 m m IO N n n 1 2 o 7 q 0 U 0 I z z I Q I 05 J 06 1 31 K A 025 g 2 m g l l I l l a o l i 3 4 5 THRESHOLD DIFFERENCE BETWEEN A AND B Illustration of the twointerval twoalternative forced choice task and the concept of the psychometric function Proportion Correct 0 O O I 00 O O m I I I I I 0 01 02 03 04 05 06 Vernier Offset min of arc A real example of a psychometric function Each data point represents the proportion correct for a block of 30 trials Cumulative normal function 1 2 X 1 zl 6 l Fxa q1qlJ2 e aw Logistic function 1 Fxa Q1 LIW Weibull function Fxa ql q l exa Three different equations that have been used to describe psychometric functions Each has three parameters which are typically estimated with maximum likelihood methods The parameter q represents chance performance in the task roughly speaking the parameter alpha controls the horizontal position of the curve roughly speaking the parameter beta controls the steepness of the curve An objective identi cation task with no feedback The illusion in this example is called the MullerLyer illusion Such illusions have also been studied with descriptive methods the phenomenological approach 0 CD I 0 O I Proportion A quotLongerquot o 4 I 0 N I O I I I I I 6 4 2 0 2 4 LengthA Length B mm point of subjective equality PSE Example psychometric function for an objective task with no feedback Comparison patch Example of an objective estimation task with no feedback The gray scale luminance is the same for the two squares in the checkerboard in fact it is exactly the same gray shown at the tails of the arrows One way to estimate the difference in apparent luminance is to adjust the luminance of a comparison patch against a same xed background to match the brightness of the squares in the two regions of the image The luminance of the lighter comparison patch has an apparent luminance more similar to the square in the shadow The difference in the physical luminance of the comparison gives a precise measure of the apparent psychological luminance difference Approaches to Understanding Perception Natura tasks Naturai ace rue statistics matme Respgnsea at individual neumns Pesoonsea of neurai papula om P6maptuaii behaviotai pa ormance Mathematical and computational modeling ComputationallMathematical Approaches Descriptive models Information processing models Physiological models Ideal observer analysis Ideal Bayesian Observer Perception as Rational Inference Possible stimulus categories 0102 c Prior probability 1701 Posterior probability 1701 IS Rational decision rule make response 139 ifpcl lS gtpc IS for all i j Ideal Bayesian Observer Perception as Rational Inference Possible stimulus categories 0102 c Prior probability 170l Stimulus likelihood 17Slcl Rational decision rule pick category 139 ifpScl pcl gtpScpcJ for all i 2 j Attention in Visual Perception Definition of attention Selection of information for specialized processing usually in the context of some goal or task Questions What are the kinds of specialized processing Where does the selection occur along the path from stimulus to behavioral response What kinds of information can be selected How is the information selected Find Waldo Find all the antlers These tasks demonstrate selective attention Change Blindness u u Compare this image to the one after the next slide Find what is missing This demonstrates that some kind of specialized processing must be selected and applied in the rst image and second image in order to notice What is missing in the second Change Blindness Change Blindness TN ne 5 mud Controlled experiments for studying covert selection attention This is the Posner cueing paradigm E1 mem Search Conuncliun search Spaml con guration I I I I I 39 quot39 I I I lt increasing senile Reaction time ms 9 o msl Vb 4w X n t n 1W 39 m5 item Set size Emciem Inefficient Another popular task is the Visual search task where set size number of cued locations is varied There have been two theories of the set size effects serial processing of each cued location and entirely parallel processing of each location Both predict set size effects Easy Search Hard Search I ll I I I II I I I I I A B 1 1000 a Easy Hard o o 2 09 E 800 H I o 08 0 60 I m D 1 quotI o E 07 a 400 Easy m D Hard 0 O 06 a zoo u A 05 o o 10 20 so 40 o 10 20 so 40 Set Size Set Size Predictions of parallel processing model of covert search based on signaldetection theory SDT E CONTRAST ORIENTATION CONJUNCTWN E 0399 5 Conjunction T o Fcawre Cunlnsl 1 03 E mum Oricnlslion a x U E x E SDTModel g 07 a a Serial Mndcl n 06 KG r I n 5 m IS 10 25 set Size SetSize Comparison of the two models for accuracy in a covert search task Eckstein 1998 The parallel SDT model also predicts the performance in the Posner cueing task Questions About Visual Search With Eye Movements How should the eyes be moved when searching with a foveated visual system Do humans use rational eye movement strategies in performing visual search If not what strategies do humans use 15 Example of a fixation search task In this task the subject is looking for a Windowed spatial sine wave target in a background of lf noise Backgrounds of lf noise are relevant because they have the same amplitude spectrum as natural images The task is to find the location of the target as rapidly as possible Without making mistakes This kind of search task is similar to what goes on in many natural search tasks finding a bird in foliage spotting a plane in the sky spotting a life raft in the ocean or finding a dropped object in the grass Naturalistic Search Task Fixate Target Press Button Press Button When Found Begin Search Central Fixation Time line of visual search task Logm amplitude gt 36 0 U Log 0 spatial frequency cyclespicture Field 1987 Amplitude spectra of natural images the same amplitude spectra The two images on the right have Generic Framework for Fixation Search I Start with prior beliefs about target location I Encode visual image I Update prior beliefs to posterior beliefs I MEMORY if max belief exceeds criterion STOP and pick the max location as target VISIBILITY MAP REPRESENTATION I Choose the next fixation location I repeat the cycle It is useful to think about what sort of processing is involved in a xation search task Here is a generic framework that most if not all models of xation search would fall within Three General Fixation Selection Strategies Random andlor Tiling Search Fixate random locations in the display andor tile the display Featu reBased Search Fixate locations with features similar to the target InformationBased Search Fixate locations likely to provide the best information about the location of the target Within this general framework three general classes of fixation search strategies that have been considered in the literature Detection b Search 15 deg 15 deg To evaluate model searchers and compare it to human performance it is necessary to specify or measure the visibility maps for the targets and backgrounds of interest We measured visibility detectability for a 6 cpd target as a function of target contrast and noise background contrast at the 25 locations indicated by the small circles on the left in a two interval ZAFC detection task while monitoring eye position The stimuli were presented for 250 ms intervals which correspond to the approximate duration of fixations in the search task a 1 E 00 o Nolse 0 3 Contrast LC 393 O 00 E g o g 06 O 005 3 A ll V 01 v 02 V 39 04 001 01 05 001 01 05 Target Contrast Target Contrast d c 015 d V bl M 00 45 5 ISI Ilty ap 35 JN I o A 4 3 a 01 weA A 3 g E 2 1 b 25 E quota O 0 0 E 2 E 005 o 0 Lu 15 O 1 0 I I 6 I I I I I 0 001 002 003 004 6 4 2 0 2 4 6 Noise Contrast Power Eccentricity deg Measurements of Visibilitydetectability for a wide range of eccentricites target contrasts and noise contrasts Here is a typical sequence of fixations for a rational searcher Notice the haphazard pattern and variable saccade length however this is the optimal ideal strategy As it turns out humans are very good at this task and are nearly optimal ideal They can find the target on average With the smallest number of fixations possible given the falloff in resolution of their eyes 20 Noise JN WG Contrast g A A 005 g 0 O 020 Q 9 L 32 5 L 9 8 a g 25 2 O 3 4 5 6 7 Foveal Target Visibility d39 The symbols are the median number of fixations that the two human observers required to locate the target as a function of the visibility d of the target in the fovea for two levels of background noise contrast You can think of the bottom aXis as contrast the greater the d the greater the contrast As can be seen search performance improves as the visibility of the target increases and is better in the high noise condition this is because the visibility maps are broader in high noise for the same foveal d The solid curves show the predictions of the ideal searcher with the same visibility maps as the human observers The stopping criterion for the ideal searcher was set so that it has the same error rates as the human observers The results imply that humans are remarkably efficient at visual search at least under these conditions nearly reaching the performance of the ideal searcher 20 15 r o 0 2 10 o 9 2 5 7 LI O l l W l l l l l l E d so d39 60 d39 70 03959 g 15 7 4 Contrast E o 005 3 Z 10 7 7 o 020 5 7 I 7 7 0 l l l l l l W I W I W I W l W l W l 0 2 4 6 2 4 6 2 4 6 Target Eccentricity deg Here is a more detailed analysis of the search times where the median number of fixations to find that target has been plotted for each condition as a function of the eccentricity of the target from the center of the display Symbols are the average of the two subjects curves are the performance of the ideal searcher The data are less reliable at small eccentricities because there are fewer trials where the target is located there We see that humans behave in a nearly optimal fashion for all target eccentricities 20 20 Noise JN WG Contrast g A A 005 g o o 020 U D X 4 L U H CC 0 L 5 g E LIJ g 25 3 4 5 6 7 Foveal Target Visibility d39 The dashed curves show the performance of a searcher that is ideal in every way except that it makes random fixations Humans far outperform this searcher Humans also outperform an enhanced random searcher that has the added feature of inhibition of return The fact that humans outperform these searchers is a powerful result because it conclusively rejects all possible models of Visual search where fixation locations are selected at random with or without replacement 21 20 Noise MAP Ideal JN WG Contrast A A 005 O O 020 Error Rate Number of Fixations 5 6 7 Fbmm mm 4 Foveal Target Visibility d The dashed curves show the performance of a searcher that is ideal in every way except that it always fixates the location with the greatest posterior probability of being the target location Under these conditions this MAP searcher performs almost as well as the ideal searcher Hence the MAP featurebased searcher cannot be rejected on the basis of overall performance To compare models need to look at eye movement statistics 22 00 Combined Distribution of fixation locations in the stimulus display for two human searchers top and three model searchers bottom Dark red indicates locations that were fixated relatively frequently dark blue indicates regions that were fixated relatively rarely Human and ideal searchers tend to have a donut shaped distribution with more fixations in the upper and lower areas of the display The model searcher that always moves the eye to the current best guess of where the target is located the MAP searcher does not have a donut shaped distribution and tends to fixate more to the left and right The model searcher that makes random eye movements fixates all locations equally often on average 23 Human Saccade Length deg Relative Proportion of Saccades Saccade Direction deg Humans more horizontal saccades shorter horizontal saccades secondary peaks in vertical direction Ideal same trends but fewer horizontal saccades MAP more horizontal saccades longer horizontal saccades no secondary peaks 24 Why Make Eye Movements Limited eld of view Variable retinal resolution Relative acuity 0 at 9 h optic disc 20 10 0 10 20 Horizontal eccentricity deg Why Make Eye Movements So why notjust move our heads as raptors do Efficiency The average human head weighs about 5kg vs 8g per eye 7 vs 002 of body weight In contrast raptors eyes account for a significant proportion of the mass in their heads up to 5 of total body weight Extraocular Musculature 3 complementary pairs of muscles for 3 DOF Superiorinferior rectus Lateralmedial rectus Superiorinferior oblique Innervated by oculomotor abducens amp trochlear nerves Supermr 39ecms Supenur WEE Lamar Memer Remus v 2 LEFT 0mm EVE nYenurR39ecms Measuring Eye Movements Afterimages Eyemounted mirrors EIectro oculogram EOG Scleral search coils Purkinje images Videobased limbus and pupil tracking FnrhWN Types of Eye Movements Vestibuloocular reflex Optokinetic reflex nystagmus Fixational eye movements Vergence Smooth pursuit Saccades Involuntary Vestibuloocular Reflex VCR 11 Detection of rotation gt1 w iJisemicircular canals stabilize gaze lt5 2 inhibition of 2 Excitation of extraoculat extraonular Very fast response muscles muscles on on one the other latency 1 Oms side we Smootth compensates for brief head movements 3 Compensating eye movement Optokinetic Nystagmus OKN Complements VOR using visual optical flow information Slow response compensates for prolonged head movements Triggered by largefield motion Influences OMN through VN Two phases 1 Slow phase stabilizes view during locomotion Fast phase nystagmus keeps eye from pinning out Fixational Eye Movements Mm 4u o Microsaccades microtremors ocular drift Functions mechanisms remain open questions microsaccades Prevents retinal stabilization and Troxler fading Troxler Fading Demo 1 Troxler Fading Demo 2 Vergence amp Conjugacy Vergence Moves eyes in opposite directions to point toward common 3D point V Makes stereo vision possible Prevents diplopia Conjugacy Eye movements normally yoked to coordinate gaze 775 h shifts see slide 6 Failure strabismus leads to amblyopia Smooth Pursuit Used to track moving objects stabilizes target position on retina Uses fovealparafoveal retinal slip signal Retinal Slip Vtarget Veye Corollaries No smooth pursuit without moving stimulus Afoveate animals don t have a smooth pursuit system Saccadic Eye Movements Used for visual exploration of environment Directs fovea to objects or regions of interest Position deg D 2 4 E 6 10 l l y 300 350 l l 0 l Quick jerky eye movements Fast up to 800 degs Ballistic openloop no online feedback Highly stereotyped movements main sequence VelociLy degsec 100 2G 10000 2392900 elerat mn degsecsec 10 UUO 0 Ass Saccadic Main Sequence B Duration ms C Peak vclociLydsgs 80 5 O 60 400 40 300 20 200 c0 20 1000 5 10 IS 20 Amplitude deg Saccadic Adaptation PREADAI39T ADAPTATION RECOVERY 1 looxoo E Q quotU 3 E q lt 2 quotU B U U N U 1 1 800 1000 1200 1400 1600 S b l 1 1800 2000 2200 2400 2600 2800 accade Num er 1 I u I I 10 I I I 100 MS Latem mvapaneta Avea UP Superior commas Superior Colliculus Gaze Displacement Map ERACHIUM OF THE mmuoa coLuchLLvs mmuoa coLuchLLvs FACIAL coLuchLLvs DORSAL COCHLEAR VESTIBUchcIILEA cums NERVE c N vxm chssopmamcmmma cmmmz TUBERCLE c N no 1 CLAVA gt DORSOLATERALSULCUS DORSAL VIEW Robinson 1972 Perceptual Aims and Consequences of Saccades Open Questions How do we select saccadic targets Why are we largely unaware of our constant jerky eye movements Why do we perceive a smooth continuous visual environment instead of the jumpy sequence of images actually formed on our retinas 1 Libra examen 2 Evaluerla situation de mags mat rielle de la famille 3 Donner I39Age 4 Pr vuir ce que la lamille 5 M moriser es vetements des personnes avait iait avant l39arriv e port s par les personnes du Visiteur inattendu Consigne Regarder 7 t ches chacune pendant 3 minutes stimer oombien de em 7 E ps Ie Visiteur inanendu a t loin de la famille 6 M moriser Ia posi on des personnes et des objets Yarbus 1967 Perceptual Stability Oculonmor events leaoe Visual cvunls 5914 mum Motion Perception Uses of motion information Detection Grouping Distance Shape Heading Measuring motion information Motion measurement mechanisms Correspondence problem Aperture problem Brain areas Detection The image of an object must move on the retina or it will disappear Examples stabilized images retinal vasculature Grouping lmage features that move together tend to be grouped together Gestalt principle of common fate Motion cues to distance When xating the horizon while translating in a car or train the nearer an object the faster it moves across the retina When xating a closer object while translating the further an object from the xated object the faster it moves across the retina I Demonstration of shape from motion If small lights are attached at each joint in the human body the static pattern of lights is nearly incomprehensible However as soon as the person moves the scene becomes easy interpret For example here are the static patterns for two people dancing A second demonstration in class showed random dot movies ofa rotating objects Optic ow is a key source of information for heading which is the direction of ones own motion This gure illustrates the local motion vectors produced during translation toward a point on the horizon indicated by the vertical line The length of the line segment attached to each dot shows the speed of the motion the direction of the line shows the direction of the motion Each dot represents an arbitrary point on the ground plane Translation Jgt kYllldb ll KKKJJ ANN KKKKJtlxxNNNNN KKK1JJLI rJJJJLLNNNNNN Translation amp Rotation RKKKKKKK39SRKKKKRK KKKKKKKKKKKKK K KKKKKKKKKKK sshgsgssa lt lt 4 O kkkekerrrvtssxa LlKkKKKJLXVNNN KIILKJztix The upper gure shows optic ow similar to the example in the previous slide The lower gure shows the optic ow vectors when the person is xating at a nearby point while continuing to translate toward the point on the horizon Notice that the ow pattern changes a great deal Humans still correctly interpret this ow pattern The optic ow pattern when approaching a nearby by surface eg a wall can be used to judge the time until contact with the surface even if the observer does not know how fast heshe is moving Motion Perception Uses of motion information Detection Grouping Distance Shape Heading Measuring motion information Visual latency Motion measurement mechanisms Correspondence problem Aperture problem Brain areas Visual Latency Contrast Luminance eg Pulfrich effect Spatial frequency Three factors that affect Visual response latency 8 7 O 7 9 7 AA 7 14 7 A 7 A O A 0 l 0 l 9 0 l 0 30 60 0 40 80 0 25 50 24 Time Shift ms 0 3 l Eff l O O l I gt 4 o gt o gt l J o l a l 0 45 90 0 40 80 0 20 40 16 i 16 i 18 l A G H O I 8 e 7 8 7 7 9 7 0 7 O A O O A 0 0 02 0 e 0 30 60 0 35 70 0 35 70 20 K L A 7 1o 7 7 O O OO O l 0 i Q 45 90 0 30 60 Contrast Changes in response latency for 12 different neurons in primary Visual cortex V1 as a function of stimulus contrast Response latency decreases an average of 2030 ms as the contrast is increased Time shift change in response latency Corresponding points Luminance light intensity also affects latency One demonstration of this is the Pulfrich effect A lter which reduces luminance causes a longer latency of neural response in the left eye Thus the moving green bar can appear at location P in the left eye but at another location in the right eye The result is an effective disparity which causes the bar to appear to be at location Q a different depth The bar appears to be at a greater distance than P when motion is left to right 4W u o 1 o o u 30 3 an 5 3 250 m I 5 I 3 5 7 E II SPAM mum lamaDESI Reaction time as a function of target spatial frequency Breitmeyer 1975 The higher the spatial frequency the longer it takes to see the target and hence the longer to push a button Mechanisms for Motion Measurement Delayed summation Motion is extremely important to measure accurately The brain probably uses several different kinds of mechanism to measure motion Perpendicular Schematic of simple delayed summation circuit for creating a direction selective simple cell A stimulus moving right to left will result in signals arriving at the summation site at different times and so no spikes will be generated at the output Perpendicular On the other hand a stimulus moving left to right will result in signals arriving at the summation site simultaneously and so spikes will be generated at the output
Are you sure you want to buy this material for
You're already Subscribed!
Looks like you've already subscribed to StudySoup, you won't need to purchase another subscription to get this material. To access this material simply click 'View Full Document'