Biomedical Engineering Reference
In-Depth Information
measured variable. Some investigators have suggested that such
noise correlations play a crucial role in neural processing
(40)
.
Trial-shuffling plays a key role in discriminating between signal
and noise correlations
(42)
, and is analogous to the within-class
trial shuffling performed in our analysis of predictive informa-
tion to determine if shared predictive relationships were related to
high predictive information (signal correlations) or to shared pre-
dictive relationships (noise correlation). Furthermore, we report
neurons with high noise and low signal correlations (
Table 7.1,
1st
column) as well as neurons with high signal and noise correla-
tions (
Table 7.1, 2nd
column). However, such analyses typically
require robust measurement of signal via a tuning curve; our task
is too simple for such an analysis. Instead, we focus on using mul-
tiple approaches to develop a picture of neural interactions.
In exploring functional correlations between neurons, one
must remember the anatomy of the brain areas of interest. For
example, when recording between distinct cortical areas, one
must understand the probability of direct synaptic connections
and shared inputs. Furthermore, it is helpful to know in which
other areas the neurons might interact (i.e., thalamus), and
whether in other brain areas, projections to the neurons origi-
nate from the same neurons. While this information is difficult to
get and requires painstaking anatomy, it is critical to interpreting
data about how neurons interact. Within a cortical area, columnar
organization, laminar information, receptive field properties, and
recurrent connectivity all will help define and constrain interpre-
tation of functional interactions between simultaneously recorded
neurons
(6, 7)
.
Finally, the goal of studying functional interactions is to
design clear experiments that test hypothesis about how neurons
interact. For instance, we combined JPSTH analyses with tech-
niques that inactivate dmPFC while recording from motor cortex
to make insights about how the two areas might interact
(20)
.
One could imagine future studies targeting microstimulation to
groups with strong functional interactions, and investigate the
role of stimulation of one neuron on some other (distant) neu-
ron. Experiments of this nature will provide novel insights on how
cortical neurons work together to control behavior.
Acknowledgments
We thank Eyal Kimchi and Nicole Horst for critical comments
and helpful discussions. This work was supported by funds from
the Tourette Syndrome Association, Kavli Institute at Yale, and
the John B. Pierce Laboratory for ML and from an NIH training
