An important goal of neuroscience research is to understand
the cellular basis of perception and cognition.
To this end, Romo has monkeys carry out a sensory discrimination
task, in which they must make a motor response dependent
on their comparison of two somatosensory stimuli. The
stimuli consist of 10-4OHz vibrations on the monkey's
forefinger, which lead to a perception of "flufter." The
monkey experiences a "cue" stimulus, then after a delay
of one to six seconds must discriminate whether a second,
"response" stimulus has a higher or lower frequency. The
monkey receives a juice reward if it makes the correct
motor response by pulling one of two levers corresponding
to higher or lower response frequency.
This task is valuable, as it requires the monkey to carry
out the most basic of perceptual and cognitive tasks-that
is, to compare two quantities. It was important that the
cue and the response frequencies were varient at random.
Without variation of the cue stimulus, the monkey would
avoid true comparison, and more simply categorize the
response frequency as "high" or "low" based on a previously
learned rule.
During the trials, extracellular recordings of neurons
in somatosensory, prefrontal, and primary motor cortices
demonstrated that neuronal activity correlates with the
task. Romo's group has shown that during the stimulus,
the period of neuronal rhythmic activity in the primary
somatosensory cortex (Sl) contains more information about
stimulus frequency than does the average firing rate (which
increases monotonically with stimulus frequency).
However, the monkey's psychophysical performance in the
task matches the information based on an average rate
code. Further evidence that the monkey's perception of
frequency is based on the average firing rate of neurons
in Sl was provided by showing that neurons in the secondary
somatosensory cortex (S2) do not fire with the periodicity
of the stimulus. Moreover, the monkey was able to perform
the task equally well with an aperiodic stimulus, when
there can be no rhythmic activity of neurons, by comparing
the average frequencies.
A groundbreaking success in these experiments was achieved
by using electrical stimuli of rapidly adapting neurons
in Sl to mimic the vibrotactile stimulus. When electrical
stimuli were used in the place of the mechanical vibrations,
for either the cue or the response in the trial, Romo's
group observed no deterioration in performance. Hence,
they were able to bypass the monkey's sensory system and
inject the information directly into its cortex!
Interestingly, they found neurons in S2 and the prefrontal
cortex, whose firing rates decreased with increasing stimulus
frequency, such that they encoded the stimulus in a negative
monotonic manner. Many neurons in the prefrontal cortex
(PFC), an area known to exhibit the persistent activity
necessary for working memory, maintained task-related
activity throughout the delay period of up to six seconds.
Other PFC neurons, along with neurons in S2, fired at
a stimulus-dependent rate only during the cue, and at
the beginning of the delay, decaying to spontaneous rates
after about one second. Yet other PFC neurons, as well
as neurons in the primary motor cortex ramped up their
activity during the final second of the delay, and fired
at higher rates during the "response" stimulus. The goal
is to unravel the patterns of neuronal spiking, to distinguish
the activities that encode the perception and memory of
a cue stimulus, the comparison of cue and response stimuli,
and the preparation of motor action. They hope to understand
how a monkey makes the simplest of decisions.
