| |
Abstract:
Evoked Brain Potentials (ERPs) provide an important source of
information about the temporal and topological distribution of
language processing events in the brain. However, despite ERP's
high temporal resolution, data bearing upon the crucial early
components of language processing, that is, the first 300
milliseconds, have been difficult to acquire and controversial. We
will show that the application of 'Symbolic Encoding' and the
subsequent calculation of 'Cylinder Entropies' and 'Word
Statistics', reveal a considerable amount of coherent and
contrastive brain behaviour that is obscured by traditional ERP
analysis techniques. The non-linear techniques provide a method of
calculating the Shannon (also Reyni, Kullback and Topological)
Entropies of ERP recording epochs. These measures of system
entropies reflect the information carrying capabilities of the data
stream being analysed. We use the results of a conventional ERP
study, which examined number agreement and Case ambiguities in
German, to demonstrate our point. Analysis of averaged voltages
generated to the experimental conditions revealed a strong P600 in
the number disagreement condition and no significant differences in
the Case disagreement condition. Application of our technique to
the same data set shows contrasts in entropies that are: 1.
statistically significant topologically and temporally during the
first 250 milliseconds 2. statistically significant later
topologically and temporally and which are consistent with
metabolic (fMRI, PET) maps. 3. statistically significant and
correspond to the traditional voltage markers: P600, N400, ELAN,
LAN
|
