1. Richard Russell <rrussell@MIT.EDU>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1) open research question/project idea:
Apply techniques that have attempted to determine whether representations
are local or distributed in other domains, such as language, to the
question of whether IT representations are local (ie gnostic) or
distributed in character.
2) Short answer:
What kind of evidence would we look for to determine whether individual
objects are represented by individual neurons or by populations of neurons?
What approaches have investigators taken toward gathering such evidence?
3) Multiple choice:
Most evidence suggests that object recognition in the macaque takes place in:
A) secondary visual cortex
B) the parietal lobes
C) inferio-temporal cortex
D) premotor cortex
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
2. Charisse Massay <charisse@MIT.EDU>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1. Open research question: What happens if you remove a primate from its
natural environment and then expose it to certain objects from that
environment? For example, we remove a gorilla from an interactive zoo
environment (for simplicity, the Bronx Zoo has great environments) and then
take certain aspects of those surroundings, like a specific tree or rock
and test how the gorilla reacts to it. If one can find the gnostic cell,
then the experimenter can test other cell in that area through a
simplification task. This would test whether or not the IT complex is
organized in a column like setup.
2. Short answer question: Describe Tanaka's experiments. What is the
process of simplification and what do these experiments state about the IT
cortex?
3. Multiple choice: Match the pathways with their terminations in the brain:
a. ventral - posterior cortex; dorsal - inferotemporal cortex
b. ventral - inferotemporal cortex; dorsal - posterior cortex
c. ventral - inferotemporal cortex; dorsal - frontal lobe
d. ventral - medulla oblongata ; dorsal - posterior cortex
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
3. Liina Pylkkanen <liina@MIT.EDU>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1. Open question/project idea
Does the FFA perform face detection of face recognition?
2. Short answer question
Why are the effects of face inversion important to
distinguishing between different hypotheses about the role of the FFA?
3. Multiple choice:
The FFA is more responsive to
(i) upright faces than inverted faces.
(ii) faces with hair covered than faces with hair covered.
(iii) faces with eyes than faces without eyes.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
4. Jennifer C. Shieh <jcshieh@MIT.EDU>
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1. Design an experiment that would be able to clarify the Tanaka lab
findings, and more concretely describe a simplification protocol.
2. Describe experiments that tested the specificity of the fusiform face
area (FFA) to the face and explain the results.
3. What is NOT a problem with Tanaka’s model of the IT cortex?
a. It does not explain the role of learning on the representations of
patterns.
b. There is the risk of losing specificity in the final representation.
c. Columnar organization is similar to organization of other visual areas.
d. Templates are still involved in recognition, so there does not appear to
be a simplification.
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
5. Keith Thoresz <thorek@ai.mit.edu>
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1. Why was the fusiform face area not found in all subjects?
2. Give two reasons why an object representation based on highly
selective gnostic (“grandmother”) cells might be less tenable than a
distributed representation?
3. Tanaka’s object simplification protocol led to which conclusion:
a. Ocular dominance columns exist in visual cortex
b. Cortical columns facilitate object invariance
c. There is strong evidence in support of most gnostic cells
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
6. Janice Chen <kanile@MIT.EDU>
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
1. Delorme et al suggest that color is only important in object
recognition for images that cannot be accurately categorized on the basis
of the "first wave" of processing. How about the RT for recognizing an
object that is the wrong color (like a blue banana)? If this kind of object
recognition is not dependent on color, then we should expect no RT difference
for recognition of objects of correct and incorrect color.
2. In the Delorme et al study, how did the performance of subjects with
short RT times compare to that of those with longer RT times?
3. According to Delorme et al, complex form categorization tasks can be
performed on the basis of:
a. Feed-forward processing (correct)
b. Iterative processing
c. Retinal image matching
d. High-level semantic identification
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
7. Jodi Davenport <jodi@MIT.EDU>
+++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Research project:
Create a computational model to learn objects and store representations in
a columnar fashion that can later recognize novel objects.
Short answer:
What types of representations are thought to be stored in columnar
fashion?
Multiple choice:
What area of the brain is thought to be specialized for face perception?
a) the funkiform face area
b) V1
c) the fusiform face area
d) MT
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
8. Andrew Yip <ayip@MIT.EDU>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Open research question/project idea:
Tanaka has attempted to test the existence of gnostic cells, but
there are concerns about the methods of image simplification used.
Perhaps lesion studies could be used to address the question of
gnostic cells; if a particular cell is suspected of encoding a
particular face or object, then one could measure the effect of
damaging that neuron on recogntion performance.
Short answer question:
What sort of evidence would contradict the notion of gnostic cells
for object recognition?
Multiple choice question:
What evidence suggests that a region of the fusiform gyrus is
important in face recognition?
a) stronger response to faces than scrambled faces
b) stronger response to viewing of faces rather than houses
c) stronger response to three-quarter-view faces rather than hands
d) all of the above
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
9. Yuri Ostrovsky <yo@MIT.EDU>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Research Question:
What is the relationship between a single neuron firing and the "meaning" of the signal?
Short Answer Question:
What is a "gnostic" or "grandmother" cell?
Multiple Choice Question:
What is the hypercolumn hypothesis for the organization of neurons in IT?
a. Going down a column (perpendicular to the surface of the brain), neuronal populations code a slightly varied version of an object; across columns, different objects are encoded.
b. A column in IT has most of the information associated with an object: its shape, its color, its size, and its name; different columns store information for different objects.
c. Each layer of neurons parallel to the surface of the brain contains population codes for one object; layers going deeper into the brain store information for different objects.
d. Each column stores low-level features which can be common to any object. The representation of one object is the sum total of these columns (and, therefore, these low-level features).
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
10. "Amrys O. Williams" <amrys@mit.edu>
++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Research Question:
Is there a way to improve upon Tanaka's gnostic cells studies that pursues
the same question using a more uniform system of image simplification?
Short Answer Question:
Describe the columnar structure of IT cortex.
Multiple Choice Question:
If a monkey is trained on two tasks, (I) an object discrimination task and
(II) a landmark task, and then given a lesion to the temporal lobe, which
choice below best represents which tasks the money will be able to perform
properly after surgery?
a) I and II
b) I and not II
c) II and not I
d) neither I nor II