Transcript PPT - Biointelligence Lab.

Ch. 13 A face in the crowd: which groups of
neurons process face stimuli, and how do they
interact?
KARI L. HOFFMANN
2009/1/13 BI, Population Coding Seminar
Ji Hoon Lee
Summary

Where in the brain does face processing occur?
 Temporal lobe

Neural coding about faces is sparse and distributed

The face cells are clustered

The neurons relevant to face-processing are synchronized.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
1/11
Introduction

Neural responses to face stimuli may seem like an unwieldy
subject for investigating population activity.
 The nature of perceiving and learning about facial signals, in
terms of population dynamics, is very under-explored
territory.
 The population interactions of face-processing systems across
space (circuits), and time (oscillations) will be discussed.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
2/11
The many facets of faces: what can be
extracted from a face

Information provided by the face includes the gender and
identity of the monkey, and his or her associated position in
the social hierarchy.
 Head and gaze direction indicate the intended target of facial
expressions.
 Monkeys are sensitive to communicative gestures such as
silent facial expressions
 Another’s gaze direction appears to be used as an indication
of salient objects or locations in the environment.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
3/11
Where in the brain does face processing
occur?

Human resources
 Temporal lobe


Monkey homologues
Spatial clustering
 Recording from temporal lobe neurons with similar responses
where observed to cluster together
 All neurons sampled were selective for faces as a class
compared to various non-face stimuli as a class.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
4/11
Face encoding I: single-unit information



Coding is considered to be through firing rate in most studies
Latency differences have provided information about stimulus type
Subclass of face cells
 Holistic, Head direction, Gaze, Expression( closed or openmouth), Motion, identity

Neural coding about faces is sparse and distributed
 Sparse
 A neuron will fire strongly to only a few stimuli
 Sparsity leads to greater capacity
 Distributed
 A stimulus will elicit activity in many neurons
 Generalization and pattern completion from incomplete or
degraded stimuli, or from biological noise
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
5/11
Face encoding II: How is information
integrated across single units?




One of the ways that faces might be represented in a distributed code
across neurons and even brain regions is through the synchronized activity
of relevant neurons.
Hebb postulated that the neural representation of an event could be formed
by cell assemblies.
The brain regions responding to actual face stimuli are also selectively
active during imagined faces.
Cell assembly offers a convenient mechanism allowing proper grouping of
relevant distributed groups of neurons in time, and their elicitation as a
whole from incomplete or even absent sensory inputs.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
6/11
Multi-site synchronization and evidence
for cell assemblies (in other systems)

What evidence exists for cell assembly representation of faces, particularly
in the necessary temporal lobe regions?
 Through the synchronized grouping of cells (signs of cell assemblies) in other systems.




Two brain regions, LIP and PFC showed synchronized, or coherent
oscillations (Buschman and Miller, 2007)
Inter-area synchronization during task performance was reported between
areas MT and parietal cortex.(Saalmann et al., 2007)
……
There is growing evidence that oscillations can coordinate activity among
spatially distributed neurons, as in cell assemblies, and that this results in
behavioral benefits (but, limited in the temporal lobe and for face
processing)
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
7/11
Synchronization during
face processing
Rodriguez et al., 1999
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
8/11
Multi-site synchronization and evidence
for cell assemblies



Monkey temporal lobe obtained during slow-wave sleep
Slow oscillations, indicated by vertical lines, can be seen as synchronized
deflections.
The temporal lobe appears to contain multi-site synchronized oscillations,
but the spatiotemporal patterns and their behavioral significance are not
yet understood.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
9/11
Where to go from here




Existing studies suggest that the temporal lobe is concerned
with faces, as part of a network of face-processing regions.
Stimulus coding in neurons appears to be selective for a small
number of stimuli (sparse), but with populations of cells
responding (distributed). The spatial distribution of faceselective cells in the temporal lobe tends to be clustered.
There is very little to inform us about whether cell assemblies
are created and used for face processing in the temporal lobe.
The nature of coding in the temporal lobe for behaviorally
relevant stimuli is an attainable goal.
(C) 2008, SNU Biointelligence Lab, http://bi.snu.ac.kr/
10/11