Transcript (2) Face Recognition

(2) Face Recognition
• These notes are the second part of a twopart lecture roughly corresponding to (1)
object recognition and (2) face recognition
• We'll actually make a smooth transition by
discussing whether face recognition is a
special kind of recognition, or whether it
can be described in terms of more general
abilities
6. Agnosias: Impairments of
recognition
• Agnosia: from Greek, literally, "ignorance";
the inability to recognize basic objects,
usually due to brain damage
Object agnosia
• Patients cannot recognize common objects, but
have no trouble with recognizing faces
• Note that visual acuity in object agnosia is
unharmed
• When object agnosics are shown art with "hidden
faces" in an array of objects, they quickly report
only being able to see the faces…
[See Oliver Sack's The Man Who Mistook His
Wife for A Hat for more case-study details]
Can you find all 13 faces?
Prosopagnosia ("face blindness")
• Patients cannot recognize people's faces,
although basic recognition appears spared
• Again, visual acuity is spared
• Prosopagnosics often learn to identify
people on the basis of their hair, dress,
voice, and other peripheral cues
Bill's Face Blindness Page
(on-line: http://www.choisser.com/faceblind/)
• A prosopagnosic shares some anecdotes from his
life regarding his inability to recognize faces as
well as most; the early age of some of the
anecdotes suggests that he may have been born
with it
• He remembers (age 6) seeing a bank robber movie
and saying that it was silly for bank robbers to
only cover their faces, since you could still see the
rest of them (betraying his reliance on peripheral
cues that most of us ignore)
• Also left the Navy after 4 days-- when everyone
got the same haircut and uniform, he couldn't
recognize anyone anymore!
Double-dissociation?
• It appears that some patients show a deficit
for recognizing objects, while others show a
selective deficit for recognizing faces
• This suggests that objects and faces may be
processed by separate perceptual
mechanisms/brain areas
7. Recognition in the healthy
brain
• View-selective responses in monkey
inferotemporal cortex (IT): When monkeys
are trained in a recognition task like that of
Bulthoff & Edelman (see “Seeing Objects”
lecture), you can find neurons that respond
maximally to one of the trained views of the
object, and responses fall off as the object is
rotated away from this preferred view
Columnar organization in
monkey IT
• Different columns of cortex (see “Visual
Pathways” lecture for summary of how such
columns work) have neurons that are
responsive to different types of basic
shapes, forms, and even faces
Columnar organization in IT
Face-selective neurons in
inferotemporal cortex
• This was a rather controversial result when
first published
• Are neurons really responsive to faces per
se, or just to a complex set of features that
faces happen to have? (And is there a
difference between these 2 criteria?)
• Anyway, neurons in IT give big responses
to faces
Face cells?
Various manipulations make the
argument more convincing:
• Full (monkey) face: large response
• Blank out eyes or mouth: weaker response
• Scramble features randomly: very small
response
• Change angle of face gradually: gradual fall-off
of responses with increasing head-turn
• Human face: large response
Using functional Magnetic
Resonance Imaging (fMRI)
• Subject view faces for a while, then pictures
of objects
• One area becomes more active during faceviewing; another area becomes more active
during object-viewing
Faces vs. Objects (human fMRI)
8. Face recognition is special
• There are several pieces of evidence that suggest
that face recognition might be special:
– Infants show a tendency to track moving faces, at just
30 minutes old
– Face agnosia without object agnosia; object agnosia
without face agnosia
– Human fMRI result of objects vs. faces
– Inversion effect: Normal subjects are better at
recognizing upright rather than inverted faces
– Inverted inversion effect: Prosopagnosics are better at
recognizing inverted rather than upright faces
(!!! Why do you think this happens?)
9. Face recognition isn't special
• However, there are also several pieces of evidence
that suggest that the processing of faces is not
special:
– In monkey IT, face columns are intermingled with other
columns (objects/shapes)
– Actually, prosopagnosics often show impaired
recognition ability for non-face objects within a
category (e.g., birds, cars, etc)
– After training people to recognize non-face objects,
subjects show face-recognition-like performance (e.g.,
inversion effect)…
Greebles
• After this special training, brain activity
shifts from object area to face area
• Specifically, this experiment required
subjects to learn to identify "greebles”,
which are weird little shapes that are
complex (like faces), but don't look much
like them…
Meet the Greebles
Outcome of Greeble study:
• Subjects learned to discriminate between
various types of Greebles, and, as they
became experts, their brain activity shifted
to the putative 'face' area