The sensation of flutter is felt when mechanical vibrations
between 5-50 Hz are applied to the skin. Some primary
somatosensory neurons are driven very effectively by periodic
flutter stimuli; their evoked spike trains typically have
a periodic structure, with highly regular time differences
between spikes. It has been strongly argued that these
time intervals may underly a subject's capacity to discriminate
flutter frequencies. Is it true that periodicity in cortical
spikes plays a functional role? Do cortical somatosensory
neurons exploit such a temporal code? We investigated
these hypotheses by analyzing extracellular recordings
from primary (S1) and secondary (S2) somatosensory cortices
of monkeys trained to perform a frequency discrimination
task.
The analysis was based on Shannon's mutual information,
which is a powerful statistic that measures the strength
of association between two variables. We found that the
information about stimulus frequency carried by the periodic
spike intervals was indeed extremely high in S1 but decreased
dramatically in S2, whereas the information provided by
the mean rate was moderate but similar across areas. Many
S2 neurons sustained their responses for a few hundred
milliseconds after stimulus offset, during an inter-stimulus
period in which the monkey needs to remember the stimulus
frequency. The information provided by the firing rate
during this period was still substantial, but the information
from spike periodicity was practically extinguished. Additionally,
the firing rate also conveyed information about average
frequency during discrimination based on periodic vibrations---by
design, information from spike timing is practically eliminated
with these stimuli. Finally, only the firing rate signal
was enhanced by the behavioral significance of the stimuli.
Hence, we conclude that the neural code for flutter frequency
is probably based on firing rate; the exquisite timing
of stimulus-driven spikes in S1 seems unrelated to it.
These results attach a cautionary note to studies in which
functional relationships are inferred exclusively from
observed correlations between neuronal firing and variations
in stimulus or behavior: some neuronal response attributes
may covary greatly and accurately with physical quantities
without necessarily having any meaning in the language
of neuronal interactions.
