perceiving, remembering & knowing in scene cognition: where are the divisions? helene intraub...
TRANSCRIPT
Perceiving, Remembering & Knowing Perceiving, Remembering & Knowing
in Scene Cognition: in Scene Cognition:
Where are the Divisions?Where are the Divisions?
Helene IntraubHelene Intraub
University of DelawareUniversity of Delaware
Framework for Scene CognitionFramework for Scene Cognition
Visual construct Visual construct Spatial constructSpatial construct
““Dividing lines” between perception and memoryDividing lines” between perception and memory
Visual Visual Starting point: Starting point:
representation of therepresentation of the
picturepicture Very short-term storesVery short-term stores
• IconIcon• Transsaccadic memoryTranssaccadic memory• CSTMCSTM• VSTMVSTM
What is transferred?What is transferred? Lost?Lost? Added?Added? Locus of effect?Locus of effect?
Very short termstores
STMVisual Working
Mem
LTM
Intraub & Richardson (1989) JEP:LMC
LTM [minutes – 2 days] (Koriat et al., 2000; Annual Review) nonlinguistic version
RSVP trios, 1 sec masked interval-Intraub, Gottesman, Willey & Zuk (1996), -Bertamini, Jones, Spooner & Hecht (2005)
Boundary Extension (BE)Boundary Extension (BE)
Roediger (1996) Roediger (1996) “… the phenomenon falls at “… the phenomenon falls at the boundary between perceiving and the boundary between perceiving and remembering”remembering”
Spatial Spatial Starting point: Starting point:
spatial representation of thespatial representation of the
scenescene the pic only partially the pic only partially
revealsreveals
Neighborhood
Outdoors
Sky
Street surface
Spatial CognitionSpatial Cognition Starting point: Starting point:
spatial representationspatial representation
below
right left
behindbehind
above
In frontIn front
Graded representation Graded representation
Imagine, response to object imagine in Imagine, response to object imagine in front, faster than in back front, faster than in back
““Resolution” better than front than back Resolution” better than front than back
(Franklin, N., Henkel, L., & Zangas, T., 1995; (Franklin, N., Henkel, L., & Zangas, T., 1995; Franklin & Tversky, 1990)Franklin & Tversky, 1990)
Perception vs. Memory?
Visual: BE is the result of this active process
Spatial: scene understanding is a graded 3-D representation
The region that will constitute the BE error is already there.
BE is the result of a decision – which part of the entire scene representation had a sensory source – was seen before?
When does BE occur?When does BE occur?
Visual construct: Visual construct: • No a priori predictionNo a priori prediction• Early no BE, over some pd of time it emergesEarly no BE, over some pd of time it emerges• Which early visual memory system?Which early visual memory system?
Spatial approachSpatial approach• Predicts – when sensory info is gone Predicts – when sensory info is gone • Conscious memory includes surrounding Conscious memory includes surrounding
swath of spaceswath of space
Visual constructVisual construct
Diff possibilitiesDiff possibilities
Either wayEither way
Either wayEither way
May or may notMay or may not
Spatial ConstructSpatial Construct
When sensory When sensory input is goneinput is gone
Weaker record, Weaker record, more BEmore BE
Weaker weaker, Weaker weaker, more BEmore BE
Same spatial Same spatial scene repscene rep
Timing?Timing?
Retention intervalRetention interval
Stimulus durationStimulus duration
Divided attention?Divided attention?
Haptics?Haptics?
Early time-course experiments Early time-course experiments
are in collaboration with: are in collaboration with:
Christopher A. DickinsonChristopher A. Dickinson
Remains on Screen
______
BoundaryRating
Scene (250 ms)
Interrupt42 ms
Exp 136 Ss 36 Trials (4 types)
CC WW
0
+2
-1
-2
+1
Much moreMuch moreWide-angleWide-angle
Slightly moreSlightly moreWide-angleWide-angle
Much moreMuch moreClose-upClose-up
Slightly moreSlightly moreClose-upClose-up
Identical viewIdentical view::
Trial Type:Stimulus and Test Identical
CC
Trial Type:Stimulus and Test Identical
WW
0
+2
-1
-2
+1
Much moreMuch moreWide-angleWide-angle
Slightly moreSlightly moreWide-angleWide-angle
Much moreMuch moreClose-upClose-up
Slightly moreSlightly moreClose-upClose-up
Identical viewIdentical view::
Close-ups Wide-angle
CC WW-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0M
ean
Rat
ing
Test Type
Exp 1_CC_WW_42 ms
42 ms Interruption
Identical View during Presentation and TEST
CC vs. WW: t(35) = 3.29, p = .002
CW WC
0
+2
-1
-2
+1
Much moreMuch moreWide-angleWide-angle
Slightly moreSlightly moreWide-angleWide-angle
Much moreMuch moreClose-upClose-up
Slightly moreSlightly moreClose-upClose-up
Identical viewIdentical view::
Trial Type:Stimulus and Test Differ
CW
Trial Type:Stimulus and Test Differ
WC
0
+2
-1
-2
+1
Much moreMuch moreWide-angleWide-angle
Slightly moreSlightly moreWide-angleWide-angle
Much moreMuch moreClose-upClose-up
Slightly moreSlightly moreClose-upClose-up
Identical viewIdentical view::
Close-Wide Wide-Close
Park, Intraub, Yi, Widders & Chun; fMRI/PPA/RSC
Slightly Different View during Presentation and TEST
CW WC-1.2-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0M
ean
Rat
ing
Test Type
Exp 1_CW_WC_42 ms
42 ms Interruption
CW vs. WC (deviation from same – asymmetry): t(35) = 6.64, p < .001
SummarySummary
BE occurred when memory ought to have been BE occurred when memory ought to have been excellent!excellent!
Memory Load: 1 pictureMemory Load: 1 picture Task is known & repeated across 36 trialsTask is known & repeated across 36 trials Target location was fully predictableTarget location was fully predictable Extremely brief interruption of sensory input: Extremely brief interruption of sensory input:
• 42 ms (commensurate w/saccade)42 ms (commensurate w/saccade)• Eye tracking, also found across actual Eye tracking, also found across actual
saccadesaccade (103 ms) (103 ms)
Graded Representation
Effect of Presentation time?
100 ms
250 ms
500 ms
1.0
0
-1.0
-.50
.50
100 250 500 Duration (ms)
BE occurs following haptic exploration of space. Blindfolded-signted, and anEarly deaf and blind woman (Lebers Syndrome) – Intraub (2004) Cognition
Quinn & Intraub (2007)Quinn & Intraub (2007) Child DevelopmentChild Development
3-4 month old infants3-4 month old infants(and 6-7 month old infants)(and 6-7 month old infants)
Scene Understanding: Spatial ConstructScene Understanding: Spatial Construct
Visual – BE develops (pic rep – then fill out)Visual – BE develops (pic rep – then fill out)• Perception/memory issuePerception/memory issue
Spatial – BE area Spatial – BE area is already thereis already there• Which part of the full spatial representation was Which part of the full spatial representation was
originally sensory information?originally sensory information?• Reality monitoring (Johnson & Raye, 1981)Reality monitoring (Johnson & Raye, 1981)
• Usually LTMUsually LTM
• But may apply to any scene memory decisionBut may apply to any scene memory decision
Perception v. memory dividing line? Not an issuePerception v. memory dividing line? Not an issue Graded representation –prior sensory, amodal, Graded representation –prior sensory, amodal,
knowledgeknowledge
World is continuous… World is continuous… But sensory input is not!But sensory input is not!
Brain’s goal: Represent the scene Brain’s goal: Represent the scene Don’t retain spurious boundariesDon’t retain spurious boundariesSensory input is gone – mental rep remains Sensory input is gone – mental rep remains Regions are not tagged as “high acuity”, Regions are not tagged as “high acuity”,
“low acuity”, amodally “filled in” or “filled out” “low acuity”, amodally “filled in” or “filled out” On-line, layout projection facilitate view On-line, layout projection facilitate view
integration integration
Supporting a coherent representation that is Supporting a coherent representation that is always based on discrete successive always based on discrete successive views.views.
AcknowledgmentsAcknowledgments
Kristin O. Michod (RA)Kristin O. Michod (RA)
Undergrad RAsUndergrad RAs Dan Bensonoff Dan Bensonoff Dave DrowosDave Drowos Jason Rand Jason Rand Kari SchweifelKari Schweifel Mattie WilsonMattie Wilson
Erika FurlongErika Furlong We thank NIMH We thank NIMH David GriecoDavid Grieco
Kristin’s latest research, “Conceptual Masking: Is it really all about the concept or does layout matter?” Submitted to VSS ’07
Special thanks to..Special thanks to..
James E. HoffmanJames E. Hoffman Anna PapafragouAnna Papafragou Neil IntraubNeil Intraub
For generously spending time listening, commenting, questioning…For generously spending time listening, commenting, questioning…
New RM-related research:New RM-related research:
w Anna and Ozge w Anna and Ozge (Turkish evidentials and ReMon)(Turkish evidentials and ReMon)
w Jessica Hughes (BE and ReMon)w Jessica Hughes (BE and ReMon)
Visual CognitionVisual Cognition BE developed over the BE developed over the
interruptinterrupt
Spatial CognitionSpatial Cognition The content that we The content that we
ultimately define as BE ultimately define as BE was part of the was part of the representation.representation.
Once sensory input is Once sensory input is gone…gone…
Cannot distinguish Cannot distinguish peripheral from amodal…peripheral from amodal…
Can distinguish from less Can distinguish from less constrained knowledgeconstrained knowledge
Spatial LayoutSpatial Layout
Boundary Extension:Boundary Extension: Infants as young a 3-4 months Infants as young a 3-4 months
Quinn & Intraub Quinn & Intraub (in press).(in press). Child Child Dev.Dev.
Children 6–80 y.o. AdultsChildren 6–80 y.o. Adults
Seamons et al., (2002)Seamons et al., (2002) Adult haptic exploration Adult haptic exploration
• Deaf & blind adult (Lebers Syndrome)Deaf & blind adult (Lebers Syndrome)
• Blindfolded adultsBlindfolded adults
Intraub (2004). Cognition.Intraub (2004). Cognition.
Labeling Cognitive ProcessesLabeling Cognitive Processes
Heuristic – Perceiving, remembering Heuristic – Perceiving, remembering IssuesIssues““New Look” psychologist 1940’s/1950’sNew Look” psychologist 1940’s/1950’sBeyond the stimulus/or Perception as CategorizationBeyond the stimulus/or Perception as Categorization
Jerome Bruner ( )Jerome Bruner ( )Hochberg (1968, 1978)– or one of the chaptersHochberg (1968, 1978)– or one of the chaptersQuote about schemaQuote about schema““Levels of processing” 1970’sLevels of processing” 1970’sMemory a byproduct of perceptual processesMemory a byproduct of perceptual processes
Craik & Lockhart ( 1972)Craik & Lockhart ( 1972)Situated Perception (or Embodied) 90’sSituated Perception (or Embodied) 90’sBarsalou and othersBarsalou and others
Exp 2: BE across a SaccadeExp 2: BE across a Saccade
EyeLink II (sampling at 500 hz; chinrest) EyeLink II (sampling at 500 hz; chinrest) N = 24 N = 24 36 trials36 trials Stimulus and test on opposite sides of screenStimulus and test on opposite sides of screen Stimulus locations held constant for each SStimulus locations held constant for each S
• Stimulus right; Test on left for ½ SsStimulus right; Test on left for ½ Ss• Stimulus left ; Test on right for ½ SsStimulus left ; Test on right for ½ Ss
Exp 2:Across a Saccade?Exp 2:Across a Saccade?
Launch SaccadeTo this location
250 ms
Boundary Rating
Exp 2:Across a Saccade?Exp 2:Across a Saccade?
LaunchSaccade
250 ms
Boundary Rating
Mean total time = 298 ms Initiate saccade = 195Saccade duration = 103 (approx. 11° amplitude)Full screen is 28° x 21°
CC WW-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0
Mea
n R
atin
g
Test Type
Exp 2_EyeLink_CC_WW_GC
Target onset is GAZE Contingent (EyeLink II, chinrest)
When eyes land:Identical View Different View
CC WW-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0
Mea
n R
atin
g
Test Type
Exp 2_EyeLink_CC_WW_GC
CW WC-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0
Mea
n R
atin
g
Test Type
Exp 2_EyeLink_CW_WC_GC
59% Single Fixation; 41% two fixations
Target onset is GAZE Contingent (EyeLink II, chinrest)
When eyes land:Identical View Different View
CC vs. WW: t(23) = 2.11, p = .046CW vs. WC (absolute values): t(23) = 3.23, p = .004
Filtered DATA – Single saccade only
When eyes land:Identical View Different View
Single Saccade ONLY trials
CC WW-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0
Mea
n R
atin
g
Test Type
Exp 2_EyelinkCC_WW_GC_ Single Saccade only
CW WC-1.0-0.8-0.6-0.4-0.20.00.20.40.60.81.0
Exp 2_EyelinkCW_WC_GC_ Single Saccade only
Mea
n R
atin
gs
Test TypeFiltered DATA – Single saccade only
When eyes land:Identical View Different View
Single Saccade ONLY trials
Johnson & Raye (1981) Reality MonitoringJohnson & Raye (1981) Reality Monitoring Johnson et al. (1993) Source MonitoringJohnson et al. (1993) Source Monitoring
Spatial RepresentationSpatial Representation Enter room blindfoldedEnter room blindfolded Spatial sense (in front, above, left, behind…)Spatial sense (in front, above, left, behind…) Remove blindfold -- Discrete samplesRemove blindfold -- Discrete samples
- 3 Fixations/sec- 3 Fixations/sec
- Saccade 30-50ms (Rayner, 1998)- Saccade 30-50ms (Rayner, 1998)
– – saccadic suppressionsaccadic suppression
Transaccadic memory (Irwin, 1992)Transaccadic memory (Irwin, 1992)
Is Is BEBE available?available?
Perception vs. Memory?
Visual: extrapolation occurs around the edges
BE is the result of this active process
Spatial: scene understanding is a graded 3-D representation
The region that will constitute the BE error is already there.
BE is the result of a decision – which part of the entire scene representation had a sensory source – was seen before?
Representation of surrounding space is present immediately.
So what determines BE?
Job: distinguish what portion of this large, mostly vague/abstract representation, was derived from sensory experience.
Reality Monitoring
(Johnson & Raye, 1981)
Quality of detail/resolution
Starting with stimulus – BE develops some time after an initial representation is available
Starting with the spatial representation – the information necessary for BE is already there, so is part of the representation as soon as there is no sensory input – can’t distinguish.
Which part of the rep initially was delivered via sensory info?Can’t distinguish from nearby highly constrained part of amodal rep.
Don’t confuse with general knowledge, but do confuse with constrained portion.
Like reality monitoring … (Johnson and Raye, 1981).In fact we know that reading can lead to memory confused with sensory input about scenes that were never seen!
But that always associates with LTM – here it applied immediately following stim offset (Skidmore group and amodal completion of objects).