PowerPoint PresentationSlide 2Slide 3Slide 4Slide 5Slide 6Slide 7Slide 8Slide 9Slide 10Slide 11Slide 12Slide 13Slide 14Slide 15Slide 16Slide 17Slide 18Slide 19Slide 20Slide 21Slide 22Slide 23Slide 24Slide 25Slide 26Slide 27Slide 28Slide 29Slide 30Slide 31Slide 32Slide 33Slide 34Slide 35Slide 36Image is filtered at different orientations and spatial scales. Gives a filter template for each location.Slide 38Slide 39Slide 40Slide 41Slide 42Slide 43Slide 44Slide 45Slide 46Methods for Investigating Attentional Selection and Visual SearchAttentional limitations central fact of cognition.What is the nature of the selection process?What gets selected, and how?Reaction TimesEye MovementsBayesian PrinciplesClassification ImagesComputational models of search processQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Posner Paradigm: Attentional CueingAttentional selection can operate independently of fixation: called “covert” attention. Fixation point indicates “overt” attentionCentral Cue - voluntary/endogenous attentionPeripheral Cue - automatic/exogenous attentionPosner, 1980QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Reaction Time advantage for valid trials.Peripheral cue: advantage when cue precedes stimulus by 100 msecCentral cue: advantage when cue precedes stimulus by 300 msecConcept of an attentional spotlight: spatially basedselection.Spotlight “enhances” processing.Spotlight takes time to move.More time for an endogenous cue.Note: assumption that faster = betterVisual Search Paradigm: Another approach to investigating the selection process. (Treisman, 1980’s)Task: Find a target among distractors.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Parallel/ automatic processing of simple features.Selective attention required to bind features into objects.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Parallel search for simple featuresSerial search for items requiring feature bindingEffect of set size reveals nature of searchInterpretation of slopes in terms of msec/itemas speed of covert attention - eg 20 msec/item.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Wolfe: search varies in difficulty - no real dichotomyQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Distributions of slopes are not bi-modalQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Wolfe’s Guided Search: selection on the basis of simplefeature, followed by serial analysisQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Difficult to distinguish between serial and parallel mechanismsQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Texture segmentation usually consistent with reaction time data.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Search difficulty depends on relation of target todistractorsSimilarity of target & distractorsHomogeneity of distractorsQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Presumably reflects the signal/noise characteristics of selectionon basis of these properties.Stimulus TaxonomyNote: assumption that the stimulus is the critical variable.Reaction Time is a useful measure, but a single numbercan’t account for much of the complexity of the brainComputations.Bayesian Approach:Cueing effect in Posner paradigm without enhancedprocessing at attended location.Eckstein et al, 2002.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Greater prior liklihood of stimulus at cues location leads tobetter performance (detectability).QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Classification Image Technique:Subject detects a signal in noiseSort out the False Alarm trialsAdd all the images that resulted in false alarmsReveals the information that led to a false alarm.Cf reverse correlation techniqueQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Simulation results using classification imagesQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Classification images from simulations using differentfilters in cued and uncued locations.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.QuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Classification images from subjects in Posner experiment.Supports Bayesian interpretation of performance, not use of different filters in cued and uncued locations.Attention is a hypothetical internal variable withlimited/no explanatory power.Eye Movements: Overt AttentionIs there any advantage to covert attention inabsence of fixation?Findlay: reaction time to peripheral stimulus maybe faster if subject allowed to fixate.Visual Acuity degrades rapidly with eccentricityAdvantage of re-fixation much greater than covert attentionSet size Set sizeReaction Time Number of SaccadesserialserialparallelparallelSimilar performance for reaction time andsacadic eye movements. Williams, ZelinskyQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.RSVP Method to measure speed of attentionRSVP: Rapid Serial Visual PresentationVarious experiments that pit saccades againstother tasks such as letter recognition show interference. (Kowler, Deubel)Implication is that saccadic target selectioninvolves a shift of attention to the target location.target selectionsignals to musclesinhibits SC saccade decisionsaccade commandSaccadic eye movement circuitryLIP:lateral intra-parietalQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.fMRI shows similar areas involved in covert attention andsacadic eye movements. CorbettaEye Movements during Search in Naturalistic Scenes. Zelinsky et alCenter-of-gravity fixationsAnalytical approaches to understanding searchImage is filtered at different orientations and spatial scales. Gives a filter template for each location.+_Rao et alQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Correlate template with image: only one location has a high correlationSearch targetQuickTime™ and aTIFF (LZW) decompressorare needed to see this picture.Multi-scaleImagerepresentationCompare