A growing number of anatomic and physiologic studies have shown that parallel sensory and motor information processing occurs in multiple cortical areas. These findings challenge the traditional model of brain processing, which states that the brain is a collection of physically discrete processing modules that pass information to each other by neuronal impulses in a stepwise manner. New concepts based on neural network models suggest that the brain is a dynamically shifting collection of interpenetrating, distributed, and transient neural networks. Neither of these models is necessarily mutually exclusive, but each gives different perspectives on the brain that might be complementary. Each model has its own research methodology, with functional magnetic resonance imaging supporting notions of modular processing, and electrophysiology (eg, electroencephalography) emphasizing the network model. These two technologies might be combined fruitfully in the near future to provide us with a better understanding of the brain. However, this common enterprise can succeed only when the inherent limitations and advantages of both models and technologies are known. After a general introduction about electrophysiology as a research tool and its relation to the network model, several practical examples are given on the generation of pathophysiologic models and disease classification, intermediate phenotyping for genetic investigations, and pharmacodynamic modeling. Finally, proposals are made about how to integrate electrophysiology and neuroimaging methods.

There are few clinical or biologic predictors of response to treatments for depression. This article reviews growing evidence that electrophysiologic and neurocognitive measures of brain function may be of value as predictors of therapeutic response to antidepressants. Initial studies using dichotic listening, quantitative electroencephalography, or event-related brain potential measures have found differences between treatment responsive and nonresponsive subgroups of depressed patients. The neurophysiologic basis for these differences and the potential clinical utility of electrophysiologic and dichotic predictors of treatment out come remain to be determined in future studies.

Magnetoencephalography (MEG) is an electrophysiologic brain imaging technology that has been applied to the study of mental illness, particularly schizophrenia. Like electroencephalography, it provides excellent temporal resolution, and in combination with magnetic resonance imaging, can also provide good spatial resolution. Studies of the auditory system in schizophrenia using MEG have demonstrated an abnormality in functional cerebral asymmetry, in which persons with schizophrenia typically show reduced, or reversed, cerebral asymmetry compared with normal subjects. This abnormality is sex-specific; it is more pronounced in males with schizophrenia. These findings have not been demonstrated using other neuroimaging strategies. Thus, MEG appears to offer a unique and valuable contribution to psychiatric neuroimaging. Current research and clinical applications of MEG are limited, however, by the high cost of instrumentation. The cost of MEG systems should improve as more applications are developed, in schizophrenia as well as other neuropsychiatric conditions, and hospitals begin to invest in the technology.

The use of different imaging modalities provides the clinician and researcher with different views of anatomy and physiology at unprecedented levels of detail. Multimodal imaging allows for noninvasive measurement of structure and function in humans during complex behavior, and thus provides information about the inner workings of the brain previously unavailable. The present paper examines the various imaging techniques available, and describes their application to the clinic—in the case of epilepsy—and to research—in the case of schizophrenia. Because the electroen-cephalogram has a dynamic response in milliseconds, it provides the best temporal sensitivity of functional measures of brain activity. When coupled with high-resolution magnetic resonance imaging measures of brain structure, this multimodal approach provides a powerful tool for understanding brain activity. Clinically, the use of multimodal imaging has provided greater precision in localization of the epileptogenic focus. For researchers attempting to determine the underlying causes of schizophrenia, the use of multimodal imaging has helped lead the field away from a specific lesion view to a more distributed system abnormality view of this disorder.

In this article we discuss the procedures that should be followed in order to develop diagnostic tests based on electro-physiologic parameters that would be useful to practicing psychiatrists in their efforts to diagnose mental disorders. To do this, we start by giving an overview of receiver operating characteristic (ROC) analysis, with which a candidate diagnostic test can be characterized mathematically. This methodology allows for an estimate of the overall discriminating power of the diagnostic test and thus can be used to compare the performance of one diagnostic test with that of another. However, ROC analysis cannot be used for optimizing a given diagnostic test (ie, determining the “best” cut-off) and, because it does not incorporate Bayes' theorem, it cannot be used to compare the performances of diagnostic tests when it is given to different risk groups for the target disorder. Both of these problems can be solved by using the concept of information theory. The equation for information gain automatically takes Bayes' theorem into consideration and also provides an intrinsic criterion for finding the cutoff that best discriminates between individuals who do and those who do not have the target disorder. Neither ROC analysis nor information theory are model dependent; however, if the equation for the frequency distribution of the electrophysiologic variable is known for the disease-positive and disease-negative populations, the calculations are greatly simplified. Therefore, the assumption is often made that both of these distributions (or transformations of the distributions) are Gaussian. However, studies have shown that the results of ROC analysis are quite robust to deviations from normality of the underlying distributions.