PDP Class January 10 2011 Overview Modular approaches Marr and Fodor A critique of modular approaches in vision The word superiority effect and the interactive activation model Interactivity in the brain ID: 312208
Download Presentation The PPT/PDF document "Interactive Activation: Behavioral and B..." is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
Interactive Activation: Behavioral and Brain Evidence and the Interactive Activation Model
PDP Class
January 10, 2011Slide2
Overview
Modular approaches: Marr and Fodor
A critique of modular approaches in vision
The word superiority effect and the interactive activation model
Interactivity in the brain
Anatomy and neurophysiologySlide3
Marr’s Modular Approach
Divide and conquer approach to vision:
E.g., Marr suggests studying shape from shading, shape from motion, shape from depth as separate computations.
A key motivation is that sometimes just one type of information is enough (crumpled newspaper example).
Forces close attention to just how much can be done with each source of information alone, and a careful consideration of how it might be done.
Similar to Fodor’s modular approach in
Modularity of Mind
.
Fodor assumes that specific input and output systems are encapsulated (insensitive to inputs from other sources) and reflex-like.
The suggestion is that in order for them to be reflex-like they must be very narrow in the range of considerations they take into account.
If in fact the brain were modular this would be nice, for then function would be considerably easier to analyse.Slide4
Critique of Fodor’s appeal to ‘Reflexes’
Reflexes are not encapsulated.
Instead they involve at least a synapse between the sensory and motor neuron or even an interneuron (as shown).
Additional inputs either to the interneuron or the motor neuron can modulate the response.
The cat’s paw reflex makes clear that context can modulate reflexes.
They are still fast and (pretty) direct but still highly context sensitive.
Influence of subsequent context requires more complex mechanisms.
Illustration from
McGraw Hill Online
Lecture on Motor Control.Slide5
Critique of the Marr’s Modular Approachin Vision (Bulthoff and Yuille, 1996)
Provides no insight into what to do when different modules both provide inconclusive (possibly conflicting) evidence.
More generally, it punts on how we successfully integrate multiple sources of information, as we clearly do much of the time.
Ignores the fact that one source of information can change the way we use information from another source (see next slide).Slide6
Gilcrest, A. L. Perceived lightness depends on perceived spatial arrangement.
Science
, 1977,
195
, 185-187.
Experiment shows that subjects assign a surface a ‘color’ based on which other surfaces they see it as co-planar with. Thus color depends on perceived depth, violating modularity.Slide7
Findings Motivating the IA Model
The word superiority effect (Reicher, 1969)
Subjects identify letters in words better than single letters or letters in scrambled strings.
The pseudoword advantage
The advantage over single letters and scrambled strings extends to pronounceable non-words (e.g. LEAT LOAT…)
The contextual enhancement effect
Increasing the duration of the context or of the target letter facilitates correct identification.
Reicher’s experiment:Used pairs of 4-letter words differing by one letter
READ ROADThe ‘critical letter’ is the letter that differs.Critical letters occur in all four positions.Same critical letters occur alone or in scrambled strings _E__ _O__ EADR EODR
W PW
Scr
L
Percent CorrectSlide8
READ
READ
_E__
OSlide9
The Contextual Enhancement EffectSlide10
The Interactive Activation Model
Feature, letter and word units.
Between-layer connections were + or -; only inhibitory connections within.
Activation follows the ‘iac’ function.
Response selected from the letter units in the cued location according to the Luce choice rule:Slide11
How the Model Works:
Words vs. Single LettersSlide12
Word and Letter Level Activations for Words and Pseudowords
Idea of ‘conspiracy effect’ rather than consistency with rules as
a basis of performance on ‘regular’ items.Slide13
Simulation of Contextual Enhancement EffectSlide14
Role of Pronouncability vs. Neighbors
Three kinds of pairs:
Pronounceable:
SLET-SPET
Unpronouncable/good:
SLCT-SPCT
Unpronouncable/bad:
XLQJ-XPQJSlide15
Autonomous vs. Interactive Approaches
visual or auditory
feature level
Letter/phoneme
identificationSlide16
Can the Models be Distinguished?
Attempts to support IA models seek to demonstrate ‘knock-on’ effects influencing the phoneme level inputs to the word level.
Lexically-Triggered Compensation for Coarticulation
Selective Adaptation
Tuning Phoneme Boundaries
Basic logic:
Use context to determine identity of an ambiguous segment
Show that the contextually-determined segment identity triggers a phenomenon that affects phoneme identification (on the way to lexical access)See TiCS paper in readings for detailsSlide17
Tuning Phoneme Boundaries
Present ambiguous (s/f) segment in a context where lexical information determines its identity as ‘f’, while presenting normal ‘s’ segments:
Consider / Con(s/f)use
Later, test for identification of the ambiguous segment, and identification of words where both interpretations are possible.
(f/s)ear
Participants identify ambiguous segment and the word containing it as though they hear it now as an ‘f’.
This can be explained by assuming participants use the top-down signal to adjust the connection weights mapping features onto the ‘f’ sound.Slide18
Interactivity in the Brain
Bidirectional Connectivity
Interactions between V5 (MT) and V1/V2:
Bullier
Subjective Contours in V1:
Lee and Nguyen
MEG Evidence:
Bar et al (2006)Slide19Slide20
Hupe, James, Payne, Lomber, Girard & Bullier (Nature, 1998, 394, 784-787)
Investigated effects of cooling V5 (MT) on neuronal responses in V1, V2, and V3 to a bar on a background grid of lower contrast.
MT cooling typically produces a reversible reduction in firing rate to V1/V2/V3 cells’ optimal stimulus (figure)
Top down effect is greatest for stimuli of low contrast. If the stimulus is easy to see when it is not moving, top-down influences from MT have little effect.
Concept of ‘inverse effectiveness’ arises here and in many other related cases.
*Slide21
Lee & Nguyen (PNAS, 2001, 98, 1907-1911)
They asked the question:
Do V1 neurons participate in the formation of a representation of the illusory contour seen in the upper panel (but not in the lower panel)?
They recorded from neurons in V1 tuned to the illusory line segment, and varied the position of the illusory segment with respect to the most responsive position of the neuron.Slide22
Response to the illusory contour is found at precisely the expected location. Slide23
Temporal Response to Real and Illusory Contours
Neuron’s receptive field falls right
over the middle of the real or illusory
line defining the bottom edge of the squareSlide24
Feedback loop between OFC and Fusiform Gyrus indicates top-down contribution to object recognition
Bar
, M.,
Kassam
, K.,
Ghuman
, A.,
Boshyan
, J., Dale, A.,
Hämäläinen
, M., Marinkovic, K., Schacter
, D.L., Rosen, B., and
Halgren
, E. (2006). Top-down facilitation of visual recognition.
Proceedings of the National Academy of Sciences
, 103(2), 449-54.