Prof. Greg Francis1PSY 310: Sensory and Perceptual ProcessesPurdue UniversityNeural codesPSY 310Greg FrancisLecture 12Is 100 billion neurons enough?Purdue UniversityCOC illusion The COC illusion looks like real squares because the neuralresponses are similarTrue squaresCOC squaresGanglion cell responsesPurdue UniversityVentral stream From visual cortex to the temporal lobe Involved in recognizing or identifying objectsPurdue UniversityCritical features Cells in inferotemporal cortex respond to complex features instimuli Not easy to measure neurophysiologically Why such bizarre patterns?Purdue UniversityReal stimuli Following the ideas about the COC stimulus, ourperceptual experience is determined by the neuralresponses to stimuli Suppose you see this picture of a catPurdue UniversityCell responses This IT cell might respond strongly And the other cells hardly at allProf. Greg Francis2PSY 310: Sensory and Perceptual ProcessesPurdue UniversityCell responses One might be tempted to suggest that this cell“codes” the perceptual experience of the cat But this would not be correct This cell would respond similarly to many differenttypes of stimuli But our perceptual experience is quite different!Purdue UniversityNeurophysiology and perception Does this mean that our perceptual experience isnot determined by neural responses? After all similar neural responses should give rise tosimilar percepts No. Similarity between ganglion responses andperceptual experience works for the ganglion cellsbecause of their anatomical location Everything else in the visual system is based on theirresponsesPurdue UniversityNeurophysiology and perception The rest of visual perception does not depend onthe responses of this IT cellPurdue UniversityNeurophysiology and perception The rest of visual perception does not depend on theresponses of this IT cell In fact, probably thousands of IT cells respond to thisstimulusPurdue UniversityNeurophysiology and perception The rest of visual perception does not depend on theresponses of this IT cell In fact, probably thousands of IT cells respond to thisstimulus Millions of ganglion cellsPurdue UniversityNeurophysiology and perception The rest of visual perception does not depend on theresponses of this IT cell In fact, probably thousands of IT cells respond to thisstimulus Millions of ganglion cells Millions of orientation sensitive cellsProf. Greg Francis3PSY 310: Sensory and Perceptual ProcessesPurdue UniversityNeurophysiology and perception The rest of visual perception does not depend on theresponses of this IT cell In fact, probably thousands of IT cells respond to thisstimulus Millions of ganglion cells Millions of orientation sensitive cells Dorsal streamPurdue UniversityRepresentation of objects Objects are unlikely to be represented by a single neuron Consider facesPurdue UniversityRepresentation of objects Objects are unlikely to be represented by a single neuron Consider facesWhat do you do here?Purdue UniversityRepresentation of objects Objects are unlikely to be represented by a single neuron Consider facesWhat do you do here?What do you do withdifferent expressions?Purdue UniversityRepresentation of objects Objects are unlikely to be represented by a single neuron Consider facesWhat do you do here?Since we perceive andrecognize all of thesefaces as beingdifferent. There mustbe a different neuralrepresentations.Purdue UniversityDistributed code Each cell codes some feature of the imageProf. Greg Francis4PSY 310: Sensory and Perceptual ProcessesPurdue UniversityDistributed code Each cell codes some feature of the imageThe representationof a face is apattern across thefeatures.No two faces havethe same pattern,so not the samepercept.Purdue UniversityDistributed code Each cell codes some feature of the imageChange ofexpression maychange only someof the features.Similar faces havesimilar patterns; sosimilar percepts.Purdue UniversityFeatures So what are the features? A difficult question to answer They need not be things that we would name Nose, eye, mouth, hair Could be Fourier components Or something completely differentPurdue UniversityFourier features Selecting certain Fourier components as features might notcorrespond to anything that we would name in the imagePurdue UniversityDistributed code How does it all get put together?We don’t seefeatures, we seefaces.Purdue UniversityRelated issue We earlier discussed how we can consider the visual system to consistof many retinotopic layers of activities from cells tuned to differentfeaturesImageRetinaGanglion cellsOrientationcellsProf. Greg Francis5PSY 310: Sensory and Perceptual ProcessesPurdue UniversityRelated issue There are actually many layers we never got a chance to talkabout Color, size, motion, depth, texture, many more To an extent, they process things independently - featuresImageRetinaGanglion cellsOrientationcellsColorMotionPurdue UniversityRelated issue If you see a red car go speeding by, you gets lots of responsesfrom different parts of the brain Which part is the percept? How do they get coordinated?ImageRetinaGanglion cellsOrientationcellsColorMotionPurdue UniversityRelated issue For that matter, if you are thinking of something else orlistening intently to something You may not perceive the car Even though the nervous system responds!ImageRetinaGanglion cellsOrientationcellsColorMotionPurdue UniversityAttention It’s not entirely clear what attention is or does But it seems to be involved in “pulling together” neuralresponses from different parts of the brain Perhaps by synchronizing action potentials Perhaps it selects the features that help solve a particular task E.g., recognize a face Attention seems to be necessary to actually perceivesomethingPurdue UniversityAttentional blink Suppose you have to identify rapidly presented (100 ms)letters e.g., detect J and/or K in a stream of lettersMPKRWSPurdue UniversityAttentional blink Turns out that detection of first letter tends to make detectionof the second letter very difficult if it immediately follows the first Attentional blinkMPKRJSProf. Greg Francis6PSY 310: Sensory and Perceptual ProcessesPurdue UniversityAttentional blink Measure frequency of detectionImplies thatdetecting firstletter causes you to
View Full Document