Prof. Greg Francis1PSY 310: Sensory and Perceptual ProcessesPurdue UniversitySignal detectionPSY 310Greg FrancisLecture 12Is this a van, or a van and a car?Purdue UniversityExperimental psychology We spent the first part of the course looking at theneurophysiological properties of the visual system Now we turn to a discussion of the perceptualexperience itself What do we actually perceive? How do we describe it? How do we measure perceptual experiences? It’s not as easy as you might think! (But it’s not as bad as Fourier analysis either)Purdue UniversityMeasuring percepts There is a temptation to trust our intuition aboutwhat we see Introspection Philosophy However, what we report is often biased by more than just the “pure”perceptual experiencePurdue UniversityMeasuring percepts We tend to see objects in the world Our interpretation of those objects is based onmemory and learning It all seems effortless because it works correctlymost of the timePurdue UniversityMeasuring percepts But we know that the neurophysiology iscomplicated And with some stimuli the difficulties become clear What is the object here?Purdue UniversityThe perceptual processAttended stimulusEnvironmental stimulusActionRecognitionPerceptionProcessingTransductionStimulus onreceptorsProf. Greg Francis2PSY 310: Sensory and Perceptual ProcessesPurdue UniversityPerceptual process Notice that there is a difference between perception andrecognition You can see the black and white parts of this image But it is difficult to recognizePurdue UniversityPerceptual process With a bit of help it is obvious what the image is And once you know, you cannot see it any other wayPurdue UniversityReporting Percepts are subjective experiences I cannot know what you perceive unless I ask you to dosomething On the basis of your actions I can deduce whatyou perceive Description “I see something” You chose the correct label You push a certain button What kind of questions get the best deductions?Purdue UniversityMeasuring thresholds Suppose you wanted to identify the faintest spot of light thatsomeone can detect Method of constant stimuli Or method of adjustment, method of limits Present lots of spots of various intensities For each one the subject indicates whether he sees it or notPurdue UniversityMeasuring thresholds Do this for many trials, repeating stimuli For a faint stimulus, sometimes the subject willreport seeing it and sometimes notPurdue UniversityMeasuring thresholds Do this for many trials repeating stimuli For a faint stimulus, sometimes the subject willreport seeing it and sometimes not Make a graph0102030405060708090100Low HighPercent “I saw it” responsesSpot intensityProf. Greg Francis3PSY 310: Sensory and Perceptual ProcessesPurdue UniversityMeasuring thresholds Identify a threshold by finding the intensity atwhich the subject reports seeing the spot 50% ofthe time0102030405060708090100Low HighPercent “I saw it” responsesSpot intensityPurdue UniversityMeasuring thresholds Identify a threshold by finding the intensity atwhich the subject reports seeing the spot 50% ofthe time0102030405060708090100Low HighPercent “I saw it” responsesSpot intensityPurdue UniversityWhy percentages? Why don’t you always see the same thing? For very faint stimuli there is noise from neural processesImageRetinaGanglion cellsOrientationcellsPurdue UniversityWhy percentages? Why don’t you always see the same thing? For very faint stimuli there is noise from neural processesImageRetinaGanglion cellsOrientationcellsPurdue UniversityWhy percentages? The noise intensity varies from trial to trial Sometimes the stimulus signal is much stronger than the noiseImageRetinaGanglion cellsOrientationcellsPurdue UniversityWhy percentages? The noise intensity varies from trial to trial Sometimes the noise is much stronger than the stimulus signalImageRetinaGanglion cellsOrientationcellsProf. Greg Francis4PSY 310: Sensory and Perceptual ProcessesPurdue UniversityWhy percentages? On some trials the noise will produce a pattern that looks like thestimulus Even though the stimulus was not shownImageRetinaGanglion cellsOrientationcellsPurdue UniversityMeasuring thresholds In experiments of this type, when no stimulus wasshown, subjects sometimes report that it wasshown!0102030405060708090100NoneLowHighPercent “I saw it” responsesSpot intensityPurdue UniversityEnvironmental noise It’s not just neural noise that can make a task difficult Suppose you are waiting for a phone call The phone is in another part of your apartment and you arelistening for its ring You may hear other sounds as well (noise) Washing machine Dishwasher Television Radio Other apartments You have to decide if a sound you heard was your phonewithin the noise or just the noise alone Detection is a discrimination taskPurdue UniversityDiscrimination In a discrimination task there are four kinds ofresponses that can be made: 1) Hit: the stimulus was presented, and youreported it was presented 2) Correct Rejection: the stimulus was notpresented, and you reported it was not presented 3) False alarm: the stimulus was not presented,but you reported it was presented 4) Miss: the stimulus was presented, but youreported it was not presentedPurdue UniversityDiscrimination The situation is complicated further because your reports ofwhether the stimulus was present depends on more thanjust the perceptual experience For example, if it is a very important phone call, you will“respond” to many sounds, even if you think they are unlikelyto be the phone Hits and false alarms will be frequent Misses and correct rejections will be rare If the call is not important and you are busy doing somethingelse, you might decide that what you heard was not thephone Hits and false alarms will be rare Misses and correct rejections will be frequentPurdue UniversityBias Likewise, some subjects may be biased to say “I saw thespot of light.” While other subjects may be biased to say “I didn’t see thespot of light.” These biases change the proportions of hits, correctrejections, false alarms and misses These biases may have almost nothing to do with perception Both subjects might describe what they see in exactly thesame way, but come to