| |
Abstract:
One challenge in the analysis of event-related potentials
(ERPs) is identifying which components change as a function of task
manipulations. We present an application of partial least squares
(PLS), that identifies the spatio-temporal distribution of scalp
potentials reflecting experimental effects. Three simulations, 4
task conditions each, examined the efficacy of the PLS analysis.
Scalp recordings reflecting the activity of three midline sources
were generated. Two sources were phasic in nature and had no
temporal overlap. The third was a slow wave, with the same onset as
source 2. Experimental effects were produced by increasing source 1
or source 2 amplitude (task1, task2), or by simultaneously
increasing source 2 and decreasing source 3 amplitude (task3). PLS
identified task-related differences in the simulated ERP waveforms
which reflected the experimental effects when effects did not
overlap in time (Simulation 1), when two effects overlapped in time
(Simulation 2), and when temporally overlapping and non-overlapping
effects were combined (Simulation 3). Spatio-temporal decomposition
of the results correctly identified the original components that
produced the effects. PLS also identified task-related ERP
components in data acquired in a selective attention task, where
discriminations were made on the basis of attention to colour,
shape or the conjunction of colour and shape.
|
