International Organization for Psychophysiology Congress 2008

St. Petersburg, Russia



September, 2008

Scott Makeig , Swartz Center for Computational Neuroscience, Institute for Neural Computation, University of California San Diego, La Jolla, CA

Linking brain, mind, and behavior

The basis for the science of psychophysiology is the belief that the overall structure and detailed nature of our behavior and experience are tightly linked to and supported by the physiological structure and activity of our body and brain. The concept that coordinated observations of physiology, experience, and behavior can be used to model the linkage between these three domains is at least as old as the nineteenth century investigations of Helmholtz into kinesiology, vision, and musical experience.

The first actual observations of human brain activity associated with (bottom-up) sensation and its (top-down) interpretation employed electroencephalography or EEG. EEG was thus the first brain imaging modality, though its early explorers and, in fact, most EEG researchers today still do not approach their data from this point of view. Yet, modern high-density EEG recordings capture nothing more or less than a moving image or movie of cortical field dynamics, projected onto the scalp surface by volume conduction and further mixed with various non-brain signals or 'artifacts (depending on the interest of the researcher). The continuing computer revolution now allows us to record this moving image from up to 256 or more sites with a resolution of a million or more bits per second, giving us an unprecedented opportunity to study how brain, experience, and behavior are linked.

However, the main obstacle to progress in this direction is the lack of attention to concurrent recording of behavior, which in many EEG and other brain imaging experiments is limited to noting the moments of infrequent small finger presses on a response 'microswitch.' It is clear, however, that our rate of progress in understanding how human brain activity, experience, and behavior are linked must be slow -- if our recordings of brain and behavioral data streams continue to have a information mismatch approaching a million to one! Here, the obvious remedy to this problem is, first, to conduct brain imaging experiments that record more of our behavior that human brains have evolved to organize and whose main function is to control.

Nearly all of the current brain imaging modalities use very heavy, rigidly supported sensors (fMRI, PET, MEG, etc.). Thus, to produce useable data participants in brain imaging experiments must keep their head rigidly fixed in place near the sensors during recordings. EEG electrodes, on the other hand, can be quite light, and in the near future may become nearly weightless and wireless, thus allowing subjects in EEG brain imaging experiments the freedom to make natural head and body movements. Yet traditional EEG experiments have not taken advantage of this freedom. Why not?

First, passive electrodes pass low-level signals back to the signal amplifiers, through electrode cables whose every movement may introduce large, uncontrolled artifacts into the data. This problem may be addressed by using active electrode chips and wireless telemetry. Second, until recently adequate methods and software for separating brain source signals from scalp muscle activities, and eye movements, cardiac artifacts, and other non-brain signals were not available. Here, a marked advance in the last decade or so has been the introduction of information-based signal processing methods, in particular independent component analysis (ICA), that in favorable circumstances can learn spatial filters that separate EEG brain imaging data into functionally distinct signal sources, both brain and non-brain, without starting with a specific model of how or where each source contributes to the recorded signals.

However, the difficulty of extracting clean EEG brain signals from the recorded data is not the only obstacle to better understanding how brain activity, experience, and behavior are interlinked. Another powerful methodological obstacle is the traditional reliance on reducing the recorded EEG data to averaged responses to classes of events that investigators assume in advance are associated with stereotyped patterns of brain activity. Life, however, does not allow the brain to wait for results of response averaging to organize motivated behavioral responses to the continued stream of novel challenges we face moment by moment! Nor have human brains evolved to evoke only a limited number of stereotyped responses to these challenges! Considering a recorded, ever-varying EEG 'scalp movie' to be composed of a limited repertoire of stereotyped stimulus responses, as computed by response averaging, plus non-brain artifacts and ongoing but wholly irrelevant EEG 'noise' is itself a powerful obstacle to achieving better understanding of how our brains respond to the challenge of the moment!

A more promising conception of a more adequate EEG-based psychophysiology begins with simultaneous recording of (1) high-density EEG scalp movies, (2) detailed behavioral records, including eye and body movements, and (3) other psychophysiological measures, as motivated participants deal with a stream of varying challenges in 3-D environments. After adequate data preprocessing to extract relevant data dimensions, new information-based and machine learning methods, applied to the extracted data, may reveal much more about how our continually varying EEG brain activity is linked to our behavior and experience. I will present some first results in this direction.

Return to Current Abstracts