James DiCarlo - Seeing gender
MIT neuroscientists pinpoint neurons that help primates tell faces apart.
Rob Matheson | MIT News Office
May 8, 2015
do primates, including humans, tell faces apart? Scientists have long
attributed this ability to so-called “face-detector” (FD) neurons,
thought to be responsible for distinguishing faces, among other objects.
But no direct evidence has supported this claim.
Now, using optogenetics, a technique that controls neural
activity with light, MIT researchers have provided the first evidence
that directly links FD neurons to face-discrimination in primates —
specifically, differentiating between males and females.
Working with macaque monkeys trained to correctly identify images of
male or female faces, the researchers used a light-sensitive protein to
suppress subregions of FD neurons in the inferior temporal (IT) cortex, a
visual information-processing region. In suppressing the neurons, the
researchers observed a small yet significant impairment in the animals’
ability to properly identify genders.
“If these face-detector neurons are participating in
face-discriminating behavior — in telling gender of faces apart — then,
if we knock them down, the behavior should take a hit,” says Arash
Afraz, a research scientist at MIT’s McGovern Institute for Brain
Research and lead author of a paper describing the study in the Proceedings of the National Academy of Sciences.
This experiment, Afraz says, marks a step forward in understanding
the links between specific neurons and primate behavior. “You actually
have to perturb the activation of that neuron and see if you can affect
behavior,” he says. “If that happens, it means these neurons are part of
the causal chain for that particular behavior.”
By providing a closer look at primate object-recognition, Afraz adds,
the study could also aid in developing visual prostheses that may
require direct wiring with the IT cortex. More broadly, understanding
the light level needed for optogenetic neural silencing could also aid
in developing implantable treatments for patients with temporal lobe
epilepsy. “We could have devices implanted in the cortex that
automatically turn on when the epilepsy attack starts, and silence the
cortex with light,” Afraz says.
Co-authors of the study are James DiCarlo, a professor of
neuroscience and head of MIT’s Department of Brain and Cognitive
Sciences, and Ed Boyden, an associate professor of biological
engineering and brain and cognitive sciences whose group developed the
optogenetics tools used in the study.
Knocking down neurons
In the 1980s, scientists first hypothesized FD neurons, with studies
that recorded spikes in neural activity in response to images of faces.
“But we [never had] a clear mechanistic connection between the
activation of these neurons and face discrimination, as opposed to face
detection,” Afraz says.
For the PNAS paper, the MIT researchers trained two monkeys
to identify images of gendered faces with about 90 percent accuracy. To
do so, they displayed images of male and female faces with varying
features slightly to the left or right of a middle fixation point of a
screen. Then, they displayed two dots on the top and bottom of the
screen; the monkeys looked at the top dot if the face was female, and at
the bottom dot if it was male.
The researchers then measured neural activity in the IT cortex of the
monkeys, locating a number of subregions where FD neurons were most and
least concentrated. Next, they injected high- and low-FD subregions
with a virally delivered protein engineered by Boyden’s group, called
ArchT, which subdues neural activity in the presence of light.
After a month, the monkeys viewed 1,600 grayscale images of male and
female faces, during 40 separate sessions, while the researchers
delivered random pulses of green light to the ArchT-treated areas.
Suppressing only 1 millimeter of high-FD subregions — not low-FD
subregions — impaired the animals’ ability to correctly identify
gendered faces by, on average, about 2 percent, the researchers found.
Linking tiny clusters of neurons with the perceptual ability to
identify genders suggests those neurons are responsible for processing
gendered faces, Afraz says. “Wherever a signal is encoded more
explicitly in the brain, that part seems to contribute more to the
behavior directly,” Afraz explains. “If we know the information of a
face’s gender is encoded more explicitly in a small bit of cortex,
knocking down that bit of cortex takes a bigger toll on behavior.”
New avenue of discovery
While his lab has researched visual processing for 20 years, DiCarlo
notes that “this collaboration with Boyden — who develops cutting-edge
tools — is what opened the door to this significant advance, and to an
entire new avenue of discovery.”
In particular, as one of the first documented uses of optogenetics to
induce behavioral changes in primates, the study also demonstrates the
potential for using it to study vision and behaviors in primates, Boyden
says. In contrast to traditional neural-suppression methods, for
instance, optogenetics tools can zero in on tiny clusters of neurons for
brief moments, which can better pinpoint specific neurons as
drivers of behavior.
“We’re getting at the actual circuitry of the brain and the exact
neurons that are involved in discriminating [between faces],” Boyden
says. “These tools offer higher temporal and spatial resolution than any
other neural perturbation method.”
William Newsome, a professor of neurobiology at Stanford University,
says the study “addresses a fascinating problem in systems neuroscience …
in a set of very challenging experiments” that utilize both
optogenetics- and pharmaceutical-suppression techniques.
“This,” he adds, “is a powerful demonstration that face-detecting
neurons mediate the perceptual ability to discriminate among faces — a
very cool result.”
The study was funded by the National Institutes of Health.