Download presentation source

brasscoffeeAI and Robotics

Nov 17, 2013 (3 years and 9 months ago)

75 views

Announcement

MIDTERM


When: 2/23 8
-
10 PM

Where: 128 Dennison

Last Lecture

What/Where pathways in
humans


Attribute specific deficits


Neuroimaging evidence


This Lecture

Examining the what
pathway


The Visual Agnosias


Neurocognitive
architecture of visual
recognition


The
What
Pathway:

How object recognition breaks down

THE VISUAL AGNOSIAS


Deficits in object and form recognition:


modality
-
specific deficit in recognition


not due to naming impairment


not due to memory impairment


not due to deficits in elementary sensation

Classic Distinction:


Apperceptive versus Associative Agnosia: affect
different "stages" of visual processing

Apperceptive Agnosia


impaired shape identification


impaired copying


impaired matching


difficulty judging orientation: (horizontal vs. vertical)


can trace an image / misled by stray lines


field defects sometimes


some cases "helped" by object motion


can reach for objects accurately; negotiate a path


Domain general: words, objects, faces are all
affected.

Apperceptive Agnosia

Impaired Copy


Impaired Matching

Associative Agnosia


Impaired recognition of complex forms and objects.


In contrast to Apperceptives, early visual processing
is much more INTACT.


Simple form recognition is intact


Copying and Matching OK

Traditional View: Visual perception stripped of
meaning


Perception is intact


impairment associating percept with stored
knowledge


Associative Agnosics can copy...

More examples of

spared copying...

Associative Agnosia


Stored Knowledge is (largely) intact


Evidence: verbal descriptions and drawings
from memory


Helped by context: better with real
objects in scenes than isolated drawings.


Lesion Locus: typically bilateral ventral
(occipital
-
temporal pathway).

Integrative Visual Agnosia


Can copy, but in a labored, fragmented
manner.


Has difficulty integrating parts into a whole.


Deficit is intermediate (between apperceptive
& associative).

Semantic

knowledge

image

Shape

coding

Figure/ground

Feature integration

Grouping

Mapping to

Structural
*

description

Stages of visual recognition

Structural description: representations of shape that are composed

of parts and the spatial relations among the parts.

CLOSURE

SIMILARITY

GOOD CONTINUATION

PROXIMITY

GESTALT

GROUPING

PRINCIPLES

Semantic

knowledge

image

Shape

coding

Figure/ground

Feature integration

Grouping

Mapping to

Structural
*

description

Stages of visual recognition

Structural description: representations of shape that are composed

of parts and the spatial relations among the parts.

Semantic

knowledge

image

Shape

coding

Figure/ground

Feature integration

grouping

Mapping to

Structural

description

APPERCEPTIVE AGNOSIA

Locus of recognition deficit?

Semantic

knowledge

image

Shape

coding

Figure/ground

Feature integration

grouping

Mapping to

Structural

description

INTEGRATIVE AGNOSIA

Locus of recognition deficit?

Semantic

knowledge

image

Shape

coding

Figure/ground

Feature integration

grouping

Mapping to

Structural

description

ASSOCIATIVE AGNOSIA

Locus of recognition deficit?

Visual Agnosia Summary



"Percepts" undergo many transformations so
recognition deficits are variable.


Apperceptive agnosia
-

early deficit, affecting all
visual materials (objects, faces, words).


impaired copy and matching


Deficit in shape coding stage


Integrative Agnosia


can be material
-
specific (word or face processing
may be spared)


copying is slavish and labored


deficit in figure/ground segregation, grouping

… Summary continued


Associative agnosia
-

a "later"
deficit of perception &
association.


can be material
-
specific (e.g., word
or

face
processing may be spared)


Does this agnosia category really exist or
are they all integrative agnosics?


Material
-
specificity of visual
agnosias suggests specialized
visual processors.

QuickTime™ and a
Photo - JPEG decompressor
are needed to see this picture.
Prosopagnosia: Evidence for a special
purpose face processor?


Impairment in recognizing familiar faces


Person recognition can occur non
-
visually


Person memory is intact


Reading and object recognition may be
"intact"


Bilateral Ventral path lesions (or RH only?)

How selective is the impairment?

Aspects of face processing may be spared


Face matching


Age and gender ID


Facial Emotion ID

Non
-
face stimuli may be affected


Impaired within class animal recognition


former bird watcher:

"all the birds look the same"


farmer:

unable to distinguish among his cows
(zooagnosia???)


Some pure cases exist: DeRenzi (1986)

How do we process faces?

Yin’s (1970) Inversion effect


Face recognition suffers more from
inversion than other objects.


Specialized configurational processor

0
1
2
3
4
5
6
Faces
Houses
Upright
Inverted
Semantic

knowledge

image

Shape

coding

Mapping to

Structural

description

Mapping to

Structural

description

Semantic

knowledge

QuickTime™ and a
Photo - JPEG decompressor
are needed to see this picture.
Semantic

knowledge

image

Shape

coding

Mapping to

Structural

description

Mapping to

Structural

description

Semantic

knowledge

QuickTime™ and a
Photo - JPEG decompressor
are needed to see this picture.
Special purpose letter processor?

(Pure) Alexia


Pure Alexia
:
acquired

reading
impairment without other salient
language deficits.


patients read letter
-
by
-
letter


word knowledge is intact


Can dissociate from other recognition problems


Co
-
occurs with achromatopsia


Upper RVF quadrantanopia


Unilateral LH lesion occipital
-
temporal region (area 19
/ 37)

Face recognition



requires encoding of relations among parts
(gestalt) without decomposing into parts.



identification does not depend on any
particular part (e.g. the nose)


Not so for Words:



Parts must be specified.


Distinguishing WORD from WORK


Face and Word Deficits Doubly
Dissociate

Farah theory: Two specialized
(lateralized?) processors

I. PART DECOMPOSITION:


used heavily for word identification but also objects


LH dominant

II. WHOLISTIC PROCESSING:


for complex objects without part decomposition
--
>
faces


RH dominant


Object recognition relies on both


to varying degrees
--
>


one may compensate for the other.

Evidence for two processors


Object agnosia co
-
occurs with pure alexia.


Object agnosia co
-
occurs with prosopagnosia


Rarely, if ever, do pure alexia and
prosopagnosia occur without object agnosia.

Holistic processing

Part Decomposition

Semantic

knowledge

image

Shape

coding

Mapping to

Structural

description

Mapping to

Structural

description

Semantic

knowledge

A

M

g

e

h

K

QuickTime™ and a
Photo - JPEG decompressor
are needed to see this picture.