| |
Abstract:
A typical neuroimaging study involves the manipulation of
stimulus processing demands to assess brain activity, which then
impacts on behavioural output. Multivariate Partial Least Squares
(PLS) analysis is a powerful tool for the identification of
distributed activity patterns related to stimulus or response. As
implemented in neuroscience, this technique does not examine the
neural linkages between stimulus and response. Given that the brain
is organized in a series of parallel interacting neural systems,
there may exist multiple networks that associate stimulus
processing and response in addition to networks that are unique to
either dimension. Here we introduce a new extension of PLS,
Psychophysiological Convergence, that works simultaneously on
stimulus and response dimensions. Conceptually, the analysis
identifies all dimensions of brain activity that relate different
aspects of stimulus processing to the ensuing behavioural output,
as well as activity patterns that are unique to either dimension.
We illustrate the technique using PET data from two experiments.
The first is a perceptual memory experiment where we demonstrate
that delay related differences in prefrontal lobe activation relate
directly to memory performance. The second is an episodic encoding
and retrieval experiment where we demonstrate that a subset of
regions more active during encoding predict better performance at
retrieval. Among these regions are bilateral temporal poles and
left hippocampus.
|
