Functional Neuroimaging of
Place Learning in a ComputerGenerated Space
Ming Hsu & W. Jake Jacobs
Introduction
 Our
experiment employed the use of a
Computer-Generated (C-G) Arena in
conjunction with fMRI to study the neural
structures involved in human place learning.
 The C-G Arena was originally designed
after the Morris Water Maze (MWZ), an
apparatus instrumental in the development
of the cognitive mapping theory.
Introduction cont.
 We
have previously shown that the C-G
Arena is a good representation of the human
place learning in real space.
 We have also shown that people can learn
locations within C-G space by observation.
 Thus, we took advantage of this close
correspondence to mount an fMRI
examination of observational place learning.
Introduction cont.
 Following
the predictions made by
cognitive mapping theory, we expect to find
activation in the human hippocampus
during observational place learning.
Experiment Design
 Subjects
were shown a recording of a target
being found from various locations in the CG Arena.
 Two experimental conditions were used:


1. Searches in a room that contains a visible
target.
2. Searches in a room that contains an invisible
target (i.e., visible only upon contact).
Experiment Design cont.
invisible
kaleidoscope
visible
Invisible
1
2
46
90
134
Kaleidoscope
13
22
Kaleidoscope
57
66
Kaleidoscope
101
110
Kaleidoscope
145
154
Visible
Kaleidoscope
35
44
Kaleidoscope
79
88
Kaleidoscope
123
132
Kaleidoscope
167
176
177
All trials11can be roughly divided
into thirds.
First 1/3 of
24
33
Visible the target, second 1/3
theInvisible
trial 55
consists of panning 68towards
77
shows
the last 1/3 of the trial
Invisiblemovement to the target, and
Visible
99
112
121
shows
turning
while
on
target.
Invisible
Visible
143
156
165
Activation in Perceptual Model
Perceptual Model
(1) invisible trials - kaleidoscope
(2) visible trials - kaleidoscope
Model Subjects MF & RD
Precentral Gyrus Activation
RD:
MF:vis
inv.v.v.kal
kal
MF: vis. v. kal
RD: inv v. kal
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
MF: invisible v. kaleidoscope
 Because
subject RD did not contain any
significant clusters of activation, only
activation curves from MF will be shown.
Intraparietal Sulcus
MF:
RD:inv
visv.v.kal
kal
MF:
vis v.
RD: inv
v. kal
kal
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
MF: invisible v. kaleidoscope
Notice again the 2 “bumps”
in the activation curves.
RD: invisible v. kaleidoscope
RD: inv v. kal
Therefore, the latter 1/3 of the
trial appears to be crucial for
subsequent performance in the
CG-Arena
Cerebellum Activation
RD:
MF:
visinv_kal
v. kal
MF: vis_kal
RD: inv v. kal
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
MF: invisible v. kaleidoscope
 Again,
only MF activation curves will be
shown.
MF: inv_kal
Cerebellar activity seems to
mirror, albeit roughly, the
activity in the precentral and
parietal areas.
Activation in Learning Model
Learning Models
(1) first 2 invisible trial - last 2 invisible trials
(2) first 2 visible trials - last 2 visible trials
Prefrontal Cortex
MF:
RD: invisible
visible
RD:
MF:
invisible
visible
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
Temporal Lobe: Anterior
MF:
RD: invisible
visible
MF:invisible
visible
RD:
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
Temporal Lobe: Posterior
MF:
RD: invisible
visible
MF:
visible
RD:
invisible
Activity in temporal lobe appears to be at least
an indicator of learning.
Activation
Highest
Neutral
Low
Low
Deactivation
Highest
MF: MT Activity
MF: invisible
Recapitulation
 Activity
in the precentral cortex, and around
the intraparietal sulcus during the last 1/3 of
the invisible trials is associated with
learning.
 Activity in the prefrontal cortex and
temporal cortex in the first 2 invisible trials
is also associated with learning.
Conclusions & Hypotheses
 “What


& Where” System
Ungleider & Mishkin 1982.
Dorsal/Parietal = Where.
 Therefore,
the time when the relationships between
the target and cues are established is the crucial
period that determines spatial learning.


Ventral/Occipital = What.
Both streams end in inferotemporal cortex,
called in monkeys polysensory cortex.
C&H cont.
 Parieto-precentral
Network: From vision to
motion.


Evidence in monkey and imaging literature.
Unanswered questions within the model.
 How
visual information gets from parietal to
precentral cortex, as motor cortex has only access to
“blind” areas of parietal lobe.
C&H cont.
 Role



of temporal lobe
Temporal activity decreases with familiarity in
monkey and imaging studies.
In this task, temporal lobe activity appears to be
associated primarily with knowledge of spatial
relationships among cues and target--difference
between invisible and visible trials.
Possibility of cognitive mapping within MT.
C&H cont.
 Role



of cerebellum
Abundance of cerebellar activity in imaging
studies.
Cognition, or fine motor control, or facilitation
of cerebral functions?
Possibility of cerebellum as pathway between
parietal and precentral areas.
Future directions/questions
 How
to get hippocampal activation that
argues convincingly for (or against)
cognitive mapping?
 What exactly is the role of cerebellum in all
this?
 Further elucidation of the existence and
function of these networks.
End of Presentation