Part of Neural Information Processing Systems 0 (NIPS 1987)
D. Lehmann, D. Brandeis, A. Horst, H. Ozaki, I. Pal
The brain works in a state-dependent manner: processin9
strate9ies and access to stored information depends on the momentary functional state which is continuously re-adjusted. The state is manifest as spatial confi9uration of the brain electric field. Spontaneous and information-tri9gered brain electric activity is a series of momentary field maps. Adaptive segmentation of spontaneous series into spatially stable epochs (states) exhibited 210 msec mean segments, discontinuous changes. Different maps imply different active neural populations, hence expectedly different effects on information processing: Reaction time differred between map classes at stimulus arrival. Segments might be units of brain information processin9 (content/mode/step), possibly operationalizin9 consciousness time. Related units (e.9. tri9gered by stimuli durin9 fi9ure perception and voluntary attention) mi9ht specify brain sub(cid:173) mechanisms of information treatment.
BRAIN FUNCTIONAL STATES AND THEIR CHANGES
The momentary functional state of the brain is reflected by the
confi9uration of the brain's electro-ma9netic field. The state manifests the strate9Y, mode, step and content of brain information processing, and the state constrains the choice of strate9ies and modes and the access to memory material available for processin9 of incoming information (1). The constraints include the available range of changes of state in PAVLOV's classical ·orienting reaction" as response to new or important informations. Different states mi9ht be viewed as different functional connectivities between the neural elements.
The orienting reaction (see 1,2) is the result of the first (Mpre-attentiveM) stage of information processing. This stage operates automatically (no involvement of consciousness) and in a parallel mode, and quickly determines whether (a) the information is important or unknown and hence requires increased attention and alertness, i.e. an orienting reaction which means a re-adjustment of functional state in order to deal adequately with the information invokin9 consciousness for further processing, or whether (b) the information is known or unimportant and hence requires no re(cid:173) adjustment of state, i.e. that it can be treated further with well-
© American Institute of Physics 1988
established (·automatic·) strategies. Conscious strategies are slow but flexible (offer wide choice), automatic strategies are fast but rigid.
Examples for functional states on a gross scale are wakefulness, drowsin.ss and sleep in adults, or developmental stages as infancy, childhood and adolesc.nce, or drug states induced by alcohol or other psychoactive agent •• The different states are associated with distinctly different ways of information processing. For example, in normal adults, reality-close, abstracting strategies based on causal relationships predominate during wakefulness, whereas in drowsiness and sleep (dreams), reality-remote, visualizing, associative concatenations of contents are used. Other well-known examples are drug states.
HUMAN BRAIN ELECTRIC FIELD DATA AND STATES
While alive, the brain produces an ever-changing el.ctromagnetic
fi.ld, which very sensitively reflects global and local states as effected by spontaneous activity, incoming information, metabolism, drugs, and diseases. The .lectric component of the brain~s electro(cid:173) magnetic field as non-invasively measured from the intact human scalp shows voltages between 0.1 and 250 microVolts, temporal fr.quencies between 0.1 and 30, 100 or 3000 Hz depending on the examined function, and spatial frequencies up to 0.2 cycles/em.
Brain electric field data are traditionally viewed as time series
of potential differences betwe.n two scalp locations (the electroencephalogram or EE6). Time series analysis has offered an effective way to class different gross brain functional states, typically using EE6 power spectral values. Differences between power spectra during different gross states typically are greater than between different locations. States of lesser functional complexity such as childhood vs adult states, sleep vs wakefulness, and many drug-state. vs non-drug states tend to increased power in slower frequencies (e.g. 1,4).
Time series analyses of epochs of intermediate durations between
30 and 10 seconds have demonstrated (e.g. 1,5,6) that there are significant and reliable relations between spectral power or coh.rency values of EE6 and characteristics of human mentation (reality-close thoughts vs free associations, visual vs non-visual thoughts, po.itive vs negative ~otions).
Viewing brain electric field data as series of momentary field
maps (7,8) opens the possibility to investigate the temporal microstructure of brain functional states in the sub-second range. The rationale is that the momentary configuration of activated neural elements represents a given brain functional state, and that the spatial pattern of activation is reflected by the momentary brain electric field which is recordable on the scalp as a momentary field map. Different configurations of activation (different field maps) are expected to be associated with different modes, strategies, steps and contents of information processing.
SE(J1ENTATI~ OF BRAIN ELECTRIC HAP SERIES INTO STABLE SE(J1ENTS
When Viewing brain electric activity as series of maps of
momentary potential distributions, changes of functional state are recognizable as changes of the ·electric landscapes· of these maps. Typically, several successive maps show similar landscapes, then quickly change to a new configuration which again tends to persist for a number of successive maps, suggestive of stable states concatenated by non-linear transitions (9,10). Stable map landscapes might be hypothesized to indicate the basic building blocks of information processing in the brain, the -atoms of thoughts·. Thus, the task at hand is the recognition of the landscape configurations; this leads to the adaptive segmentation of time series of momentary maps into segments of stable landscapes during varying durations.
We have proposed and used a method which describes the
configuration of a momentary map by the locations of its maximal and minimal potential values, thus invoking a dipole model. The goal here is the phenomenological recognition of different momentary functional states using a very limited number of major map features as classifiers, and we suggest conservative interpretion of the data as to real brain locations of the generating processes which always involve millions of neural elements.
We have studied (11) map series recorded from 16 scalp locations
over posterior skull areas from normal subjects during relaxation with closed eyes. For adaptive segmentation, the maps at the times of maximal map relief were selected for optimal signal/nOise conditions. The locations of the maximal and minimal (extrema) potentials were extracted in each map as descriptors of the landscape; taking into account the basically periodic nature of spontaneous brain electric activity (Fig. 1), extrema locations were treated disregarding polarity information. If over time an extreme left its pre-set spatial window (say, one electrode distance), the segment was terminated. The map series showed stable map configurations for varying durations (Fig. 2), and discontinuous, step-wise changes. Over 6 subjects, resting alpha-type EEG showed 210 msec mean segment duration; segments longer than 323 msec covered 50% of total time; the most prominent segment class (1.5% of all classes) covered 20% of total time (prominence varied strongly over classes; not all possible classes occurred). Spectral power and phase of averages of adaptive and pre-determined segments demonstrated the adequacy of the strategy and the homogeneity of adaptive segment classes by their reduced within-class variance. Segmentation using global map dissimilarity (sum of Euklidian difference vs average reference at all measured points) emulates the results of the extracted-characteristics-strategy.
FUNCTIONAL SIGNIFICANCE OF MOMENTARY MICRO STATES
Since different maps of momentary EEG fields imply activity of
different neural populations, different segment classes must manifest different brain functional states with expectedly different