Click for next page ( 44


The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement



Below are the first 10 and last 10 pages of uncorrected machine-read text (when available) of this chapter, followed by the top 30 algorithmically extracted key phrases from the chapter as a whole.
Intended to provide our own search engines and external engines with highly rich, chapter-representative searchable text on the opening pages of each chapter. Because it is UNCORRECTED material, please consider the following text as a useful but insufficient proxy for the authoritative book pages.

Do not use for reproduction, copying, pasting, or reading; exclusively for search engines.

OCR for page 43
4 Four New Technologies: Critical Problems In this chapter we discuss some critical problems that must be re- solved if the technologies are to contribute to the further development of cognitive science and its applications. Following some additional background on each technology, key problems with each technology are discussed. Among the six problems highlighted with respect to event-related brain potentials are definition and measurement issues, a lack of integration with the broader field of neuroscience, and un- developed methodologies for field or clinical research. With regard to neuromagnetism, such problems as secondary sources, magneti- cally silent sources, and the inverse problem are discussed. Positron emission tomography presents such problems as the precise location of physiological processes, spatial and temporal discriminations, ap- propriate statistical analysis of the data, and risks to participants producecl by the imaging technique. And studies of the effects of brain damage have been limited by problems of extrapolating from animal to human work. Moreover, in human studies, there has been a lack of understanding of the implications of effects of lesions for information processing. EVENT-RE[ATED BRAIN POTENTIALS Psychophysiology is a discipline that capitalizes on the fact that it is possible to measure the activity of bodily systems while peo- ple engage in a variety of cognitive and affect-producing tasks. The psychophysiologist's goal may be the understanding of the ways of the mind or the understanding of the ways of the brain. However, 43

OCR for page 43
44 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES in both cases the psychophysiologist is building on the variation of the bodily systems measured and the circumstances of the measure- ment. The latter is the crucial point. Whether recording heart rate, evoked potentials, or metabolic activity in the brain, one is subject to the sane basic constraint: the measurements are interpretable only within the context of the control that has been exercised over the subject's information-processing activities. One must guard against seduction by technological marvels, a seduction that often leads one to violate some of the basic tenets of psychophysiological research. The precision ant] resolution with which the bodily systems are measured do not guarantee the quality of the study. One must devote an equal degree of care to the control of the experimental situation. Thus, if an investigator reports a relationship of Preparatory pro- cesses" to the amplitude of ERPs, or to the uptake of radioactive material by brain tissue, it is imperative that the preparatory pro- cesses be defined and measured with the same precision and care that is invested in the physiological measurements. The variation in the psychophysiological measurements (dependent variables) is of interest only if it can be related to the variations in task demands controlled by the experimenter (the independent variables). Often, neither type of variable is wed defined in studies that search for rela- tionships between them. Even if the dependent psychophysiological variables are measured with great precision, a careless definition of the independent task variables renders the experiment inadequate. An investigator may, for example, employ a PET scan and pub fish results of examining the metabolism in different parts of the brain when people are asked to "think" compared with the metabolism when they are asked not to ~think." The results appear exciting, es- peciaITy as they are presented in color pictures that have anatomical reference. However, the results are largely uninterpretable because the putative independent variable, thinking versus nonthinking, is so poorly defined. There is so much that can differ between the two experimental conditions that it is quite inappropriate to claim that the study pertains to the relationship between thinking and metabolic rate. This error is not unique, of course, to PET studies. It has been committed by the practitioners of virtually all types of psychophysiology. Problems arise, of course, in the definition and measurement of the dependent psychophysiological variables as well. We list here five of these difficulties, illustrated within the context of the study of event-related activities.

OCR for page 43
CRITICAL PROBLEMS 45 Definition of Components The definition of the dependent variables presents a challenge to ah psychophysiologists. In virtually every case the data are multidi- mensional and the measurement requires some method of selecting the features of the data to which the measurement vnI} be applied. This Is true whether the data take the form of two-dimensional maps or as multiple time series. There is, for example, a continuing con- troversy in the literature regarding the proper methods for defining an ERP component. The msue is clearly central. If two investigators purport to study the N200, it is imperative that they both, in fact, study the same component. That is, they must both use the same set of operations for extracting the amplitude of the N200 from their data. If one investigation uses the statistical technique known as principal component analysis and the other records the area under the curve in the interval between 150 and 250 msec after the stimulus, they may find themselves In conflict that Is not justified by the data they were recording. There is at this tune no consensual terminology and no systematic attempt to standardize measurement procedures to ensure that components are equivalently defined in all paradigms by ad subjects (Fabian) et al., in press). OverIappmg Components The problem of specified components ~ aggravated by the fact that the measures taken at any point in time reflect the activity of more than one process. For event response potentials and evoked response fields it ~ common to expect that the measures taken at any point in time represent the activity of more than one putative component. How to d~t~ngu~h these components has been one of the more vexing problems in psychophysiology. Naatanen and Picton (1987), for example, suggested that at least five distinct components coexist in some form with the NIOO in the same segment of the epoch. However, at the same time, their review demonstrates that despite the complexity it is possible, with the appropriate techniques, to study some of these components in isolation. The presence of multiple components complicates the interpretation of results. The challenge presented by this problem has motivated a number of investigators to focus attention on the refinement of both observational and statistical techniques.

OCR for page 43
46 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES Gaps In Neuroscience We lack a fuB understanding of the generating sources of the various components of the ERP. It ~ also the case that we have insufficient understanding of the neurophysiological mechanisms un- derlying the phenomena observed In ERP research. It Is generally agreed that we are observing the summed activity of synchronously activated neurons in ensembles whose fields are "open. But what is the actual role that these ensembles play (Allison, Wood, and McCarthy, 1986~. The most serious difficulty associated with this gap in our knowI- edge is that we are never quite sure how changes In the amplitude of the ERP should be interpreted. Donchin and Coles (in press) comment on this issue, in the context of a discussion of the P300. Their remarks, as quoted below, apply to other components as well: The first assumption is the weakest link in our entire structure. It is that the amplitude of the P300 ~ a measure of the extent to which the processor manifested by the P300 is activated, or ~utilized.n...It must be admitted, however, that we have no direct physiological evidence for this, or any other, interpretation of changes in the amplitude of the P300. It is plausible to assume that the larger the potential the larger the voltages generating the field which is being recorded and that this size is a measure of the activation of the tissue from which we record. This, unfortunately, is not necessarily the case as Allison et al. (1986) point out. ...For the time being, however, we must admit that the assume tion is entertained primarily for its heuristic value and its merit must be evaluated according to its utility as a guide for future research. It should be noted, though, that the assumption does gain credence from the convergence of the very large number of studies conducted under its umbrella. Difficulties With Clinical Studies Malting sense of the psychophysiological literature is a formidable task. The number of papers published annually is very large. Unfor- tunately, the literature contains many poor clinical papers, which are all too often case reports. While they serve an important function in the communication among clinicians, they are a poor source of data for developing theoretical structures. The methodological rigor of these studies is poor and the degree to which they are tied to the

OCR for page 43
CRITICAL PROBLEMS 47 rest of the literature is inadequate. In part, this literature has been generated by the successful development of diagnostic tools based on psychophysiological methods. Instruments designed to serve wed the needs of a clinician proliferate. These machines are often used in research projects even though they are not always designed with the needs of the researcher in mind. Thus, for example, many signal averagers, well designed for the diagnostic clinic, do not provide any means for saving the data associated with each trial in the test. They yield data only In the form of averages obtained over presentation of stimulus blocks. This is entirely adequate, even desirable, when conducting a diagnostic test, yet for many research projects it is critical to sort the data prior to averaging (e.g. to compare the ERP elicited when subjects erred or responded correctly, or to compare ERPs elicited when subjects were fast with those elicited when they were sIow). The Lack of a No~mati~re Data Base As this report shows, there is strong evidence supporting the assertion that psychophysiological measures can be used as tools in the study of cognitive function. However, this assertion is correct largely within the context of the laboratory. That is, the data serve well the needs of the theoretician who is trying to choose among models. It Is the case that the variance among individuals observed in psychophysiological studies is sufficiently small to allow investi- gators to use small groups of subjects. As the research very often uses within-subject designs and is costly, investigators tend to use only 5-15 subjects in any study. One consequence is that there are very few normative data bases that indicate the expected pattern of recording in different nosological, age, and gender groups. This gap in knowledge is a serious impediment to the development of appli- cations based on psychophysiological techniques. There is a serious need to conduct studies that would develop, for a carefully chosen set of experimental designs, a comprehensive normative data base that would allow the determination of the extent of variance in the measures that can be expected from various groups of individuals. The technology of event-related brain potentials is quite similar to that of neuromagnetism. Similarity also exists with regard to fund~nental problems that underly both methods. For example, the inverse problem discussed in the next section is also applicable to ERPs.

OCR for page 43
48 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES NEUROMAGNETISM: TEE MAGNETOENCEPHALOGRAM Magnetic fields accompany all moving electrical charges, includ- ing the ions that flow inside neurons. It is widely accepted that the neuromagnetic field normal to (perpendicular to) the scalp is due to intracellular current flow and ~ not affected significantly by extracel- lular volume currents. Although it is possible to detect the magnetic field associated with action potentials In isolated axons and nerve bundles (Swinney and Wikswo, 1980), the field observed outside the head is due largely to the flow of current in dendrites. Furthermore, the orientation of the path of the flowing current must be tangen- tial to the surface of the skull if an external field radial to the scalp remerging from or reentering the scalp) is to be detected. It is for this reason that the stronger sources of observed fields are believed to lie within sulci of the cortex. Neurons at gyri are predominantly normal in orientation relative to the scalp and would therefore contribute less to the observed external field. The field associated with the flow of current in a single neuron would be too weak to be detected outside the scalp. In fact, it has been estimated that the simultaneous activity of about 10,000 neu- rons may result in a detectable field external to the scalp (Williamson and Kaufman, in press). Some field strengths are consistent with sources composed of 10,000 to 30,000 neurons. This is commensu- rate with the population of a cortical macrocolumn. One motivation for studying neuromagnetic fields is that bone, cerebrospinal fluids, and other media are essentially transparent to low-frequency magnetic fields. These intervening tissues have rad- ically different conductivities and therefore strongly affect the in- tracranial distribution of volume currents that underlie the EEG and the ERP. In analyzing neuromagnetic data it is therefore possible to ignore radial variations in conductivity within the head in computing the location, orientation, and strength of the underlying resources of an observed field. Six problems are discussed in this section. Secondary Sources Despite the claim of an intracellular origin for observed extracra- nial fields, it is theoretically possible for fields to arise from volume currents flowing at boundaries between two regions of different con- ductivity, thus giving rise to Usecondary sources." These are likely to be associated with very weak fields and, if present, would be more

OCR for page 43
CRITICAL PROBLEMS 49 likely to contaminate tangential fields rather than fields that are nor- mal to the scalp. The effects on the radial field would be quite small. Nevertheless, it IS important to precisely determine the magnitudes of contributions of secondary sources since, if they are significant but unaccounted for, they may bias computations of source locations and strengths. Magnetically Silent Sources As mentioned earlier, sources tangential to the surface of the skull are associated with observable fields, while So-called radial sources do not contribute to the external field. Assuming that these radial sources do not contribute significantly to the field external to the head, then many neurons in cortical gyri may be undetectable by magnetic sensors. In fact, Hari et al. (1982) found that the amplitude of the electrical N100 of the auditory evoked potential increases monotonically with interstimulus intervals (ISIs) up to 16 see, while the amplitude of the magnetic NIOO does not increase with IST when it goes beyond ~ sec. This was taken as evidence that radial (magnetically silent) sources contribute to the electrical NtOO and are responsible for the further increase in amplitude with IST beyond 8 see (see Naatanen and Picton, 1987~. However, it should be noted that the magnetic responses were all recorded from over only one hemisphere, while the electrical responses were affected by sources in both hemispheres. Until it is demonstrated experimentally that the responses of the hemispheres are symmetrical at different ISIs, it is not proven that sources that are essentially silent magnetically can be detected electrically and that the behavior of these ~silent" sources is significantly different from that of nonsilent magnetic sources. This is partly an empirical problem, and experiments are needed to cast further light onto it. The problem we have just described may also be addressed in theoretical terms by simulating effects of different head shapes and source configurations on external fields predicted by physical law. Some efforts along these lines suggest that this problem may be of minor significance, but more work is needed. Open Field and Closed Field Nenrone The classical distinction between open field and closed field neu- rons, made orginally by Lorente de No, applies to magnetic fields

OCR for page 43
50 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES just as it does to electrical potentials measured at a distance. Theo- retically, closed field neurons are silent both magnetically and elec- trically. This follows from the observation that their dendritic trees are symmetrical in three dimensions, and because of symmetry there is no net field] to be observed at a distance. Open field neurons have a preferred average orientation, and this makes it possible for them to serve as sources for both magnetic fields and electrical potentials. It is desirable that we obtain some estunate of the proportions of neurons that may be characterized as closed field and the places in the brain where they are most numerous. It should be borne in mind, however, that approx~nate three-~unensional morphological symmetry does not suffice as a criterion for classifying a neuron as closed field. The pattern of current flowing in such neurons is the determining property, and we are not aware of data on how this pattern develops within dendritic trees of neurons of different types. However, it is worth noting that stellate cells predominate In visual cortex and, despite the high degree of three-dimensional symmetry exhibited by their dendritic trees, they probably make a major con- tribution to visual evoked responses. Therefore, the current paths within the dendrites of these cells must have preferred orientations. This observation leaves open the possibility that such cells located within gyri of the cortex may well contribute to external fields. This conjecture requires empirical evaluation. Effects of Anatomical Symmetry While the neurons of the visual cortex have a preferred orienta- tion normal to its surface, stimulation of the entire central fovea will affect neurons on the medial surfaces of both hemispheres and in the roof and floor of the calcarine fissure as well. The roof and floor of the calcarine fissure have opposed orientations, as do the neurons of the two sides of the longitudinal fissure. These will exhibit opposed patterns of current flow, and their electrical and magnetic fields will tend to cancel each other, as do current flows in the dendrites of closed field neurons. Hence, both EP and EF experiments in which visual stimuli are presented to both hemi-retinas produce responses that represent the residual effect of the anatomical asymmetry of the hemispheres. It is possible to avoid this problem in vision by presenting critical stimuli to an octant or quadrant of visual field, but it is not obvious how to accomplish the same end when working in other modalities. Furthermore, this self-canceHation of fields and

OCR for page 43
CRITICAL PROBLEMS 51 potentials due to populations of neurons having opposed orientations would affect both the spontaneous EEG and the MEG and needs to be carefully considered in their interpretation. A Role for Complementary Technologies The foregoing problems of concurrently active and opposed sources of populations of neurons cannot be resolved simply by im- prov~ng the ways in which electrical and magnetic data are analyzed. While we must first discover how pervasive such problems are and where they are most likely to arise, there IS a compelling need for the use of complementary technologies. Since the use of complementary methods is not widespread, the failure to employ them constitutes an important problem for advancing the study of cognitive psychophysi- ology. For example, high resolution and accurately scaled MR! scans together with reviews of histological data on aD areas of the human brain may furnish computational theorists with information useful in estimating the proportion of closed field neurons in the brain and of areas in which there is significant mirror symmetry. furthermore, PET technology applied in experiments in which human subjects are exposed to precisely the same conditions as are subjects used in stud- ies of spontaneous, evoked, and event-related electrical and magnetic activity could provide a great de e] of evidence as to whether signifi- cant areas of the brain are ~invisible" to these methods. By the same token, evidence of electrical or magnetic activity from a region show- ing little metabolic activity would undoubtedly be of considerable interest. The Inverse Problem Many of the foregoing problems are related to the general state- ment that there is no unique solution to what is called the inverse problem. Given a known source or array of sources within a conduct- ing volume whose properties are known, it ~ possible to determine the field or potential that would be detected outside the volume. This forward problem does have a unique solution; however, the con- verse is not true. Given complete knowledge of the external field or the distribution of potentials, it is still not possible to arrive at a unique solution in which a particular source is known to account for the observed phenomena. A large number of sources could produce the same external field. Therefore, there is a need to constrain in- verse solutions by use of knowledge from other domains of science

OCR for page 43
52 B~4IN AND COGNITION: SOME NEW TECHNOLOGIES and, as stated above, also by use of complementary technologies. For example, considerable knowledge already exists of the organization of the primary projection areas in humans, and it Is known that the somatosensory homunculus represents the order of representation of different portions of the body along the posterior bank of the central suIcus. Straightforward measurements of the somatosensory evoked field produce data in excellent agreement with this physiologically established ordering, thus providing face validity for the inverse so- lutions that have been presented. Similarly, MRI scans of human subjects who had served in evoked fields experiments can show that the computed source locations lie at positions in various suIci known to contain neurons that respond to the particular sensory stunuli. Moving a visual st~rnulus into the peripheral retina results In a corresponding migration of the com- puted current dipole source into the depth of the longitudinal suIcus, and the magnitude of this variation with eccentricity is in approx- imate agreement with the known cortical magnification factor. To be sure, ambiguity increases when studying less well-known portions of the brain, and it is difficult to decide among alternative inverse solutions. This problem can also be ameliorated by conjoint use of PET technology. It should be noted that many portions of the brain react in serial rather than paraBe] fashion to a particular sensory input. Thus, for example, N100 and P200 of the auditory evoked fields occur at different times, and their sources have been shown to be separated by a distance of about ~ cm. If both of these portions were to be active at the same time, then it might prove difficult, although not impossible, to resolve them. However, the time differences also serve to ameliorate the problem of source resolution. A Fmal Note on Nenromagnetism The technical problems discussed in this section are critical. Un- ti} further progress is made toward resolving them, it is unlikely that significant progress will be made in applications of the technology to the study of cognitive processes, except perhaps in very fundamen- tal programs as discussed in Chapter 3 in the section on research findings. The use of complementary technologies appears promising: certain problems characteristic of particular technologies can indeed be offset by using several approaches in a particular study or research program. However, it is also necessary to address the technical and

OCR for page 43
CRITICAL PROBLEMS 53 physiological problems in the context of conceptual questions posed by cognitive science. Those questions serve to reinforce the con- nection between cognitive and physiological processes. They should influence decisions about the way in which neuromagnetism and other technologies are developed and used in experimental work. IMAGING T1:CENIQUES I: POSITRON EMISSION TOMOG1lAPlIY Positron emission tomography is a complex technology requiring adequate human and physical resources in the areas of instrumen- tation, radiochemistry, tracer kinetics, computer programming, and neurobiology as well as easy access to patients and normal control subjects. Truly successful programs have managed to gather to- gether into one group individuals with expertise in all of these areas, usually in the setting of a large university teaching hospital. The major physical resources necessary to complement such a team of investigators include a medical cyclotron, a PET scanner, a large minicomputer, and radiochemistry laboratories. At the present time there are approximately 20 PET centers ~ academic medical institu- tions in the United States. Less than half of these centers have yet to assemble the necessary staffing required to do good neurobiological research with PET. The reason for this is twofold: (1) lack of trained personnel and (2) a failure to realize the importance of this need. In addition to the more general needs of PET in terms of staffing, equipment, and patients, there are also several areas of concern tenth regard to the actual performance of PET studies. These include the anatomical localization of regions of interest within a PETscan, the spatial resolution of PET, the temporal resolution of PET, and the statistical analysis of PET data. We deal briefly with each of these issues. Anatomical Localization Determining the relationship between physiology and anatomy is one of the objectives of most functional studies with PET, including all of those designed to understand cognitive processes. Although physiological images of the brain often contain some anatomical in- formation, correspondence between physiology and anatomy cannot be assumed. Despite this fact, some investigators have based their judgments about anatomical localization on the appearance of the

OCR for page 43
54 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES physiological PET unage of blood flow or glucose uptake. This prac- tice is not satisfactory. An anatomical localization procedure for physiological imaging has been developed and validated (Fox, PerEnutter, and Raichle, 1985~. This approach determines the anatomical location of a PET region of interest with the coordinate system of atlases for stereo- taxic neurological surgery. Measurements made from a lateral skull radiograph and from a tomographic transmission scan form the basis of this method. The method is accurate and objective and does not depend on visual inspection of the image. Regions defined by this procedure can be easily compared among subjects In a study and among different subject populations within a laboratory. Compar- isons of data from different laboratories are also possible when this procedure is employed. Acceptance of such a method by people using PET is essential. Spatial Resolution Spatial resolution has always been a concern to investigators con- templating the use of PET for physiological studies. Operationally, the spatial resolution of PET is based on the spatial distribution of measured radioactivity produced by a single point source of ra- dioactivity. This spatial distribution of radioactivity is a blurred representation of the original point source with the highest count- ing rate at its center. The resolution of a PET system is defined as the width of this distribution of radioactivity at one half of the magnum counting rate, the so-called full width at half maximum, usually abbreviated FWHM. One very important consequence of this definition of resolution is that when two point sources of radioactiv- ity occur simultaneously in the field of view of a PET device, they cannot be distinguished as two separate sources if they are closer than a distance of one FWHM. Furthermore, accurate quantifica- tion of radioactivity in a particular region of the brain requires that the region be approximately twice the FWHM in all dimensions. PET devices currently operational have spatial resolutions, defined as the FWHM, In the range of 10~15 mm. The ultunate resolution of PET defined in this way has not been clearly established but will be limited by factors such as the distance traveled in tissue by the positron before annihilation, usually about 2-3 mm; slight deviation of the paths of the two annihilation photons from colinearity; and the statistical quality of the data. It is realistic to anticipate that

OCR for page 43
CRITICAL PROBLEMS 55 reconstruction for some PET images to a resolution of 5-6 rum wiD be possible, although this is still less than optional for some work. Functional studies with PET, including those of importance to cognitive psychophysiology, allow an additional, very important per- spective on the issue of spatial resolution. By functional studies is meant studies in which some type of activation paradigm Is employed to produce a change in local blood flow or metabolism from a resting or control state. Under such circumstances PET data, obtained from the subtractions of a control state image from an image obtained during some type of functional activity, can be used to determine whether areas separated by significantly less than one FWHM are differentially activated by specific alterations in stimulus conditions (Fox et al., 1985~. This expectation is based on signal detection the- ory, which has also been used to explain an equivalent phenomenon in visual processing known as hyperacuity. The occurrence of a source of radioactivity against a background or control state can be local- ized using special computer-based algorithms as a shift in the point of maximum change in radioactivity from the control state (Fox et al., 1985~. Current PET devices permit response localization with an accuracy of 1-2 mm using instruments with an inherent resolution of 18 mm (Fox et al., 1985~. Experimental strategy defines the op- erational resolution of such a study. The ability of PET to spatially discriminate sequential changes in local radioactivity in this manner has important implications for the study of very discrete functional activity in the human brain, provided that experunental protocols are designed to take advantage of it. Anticipated improvements in the inherent spatial resolution of future PET scanners, possibly as good as ~5 mm FWHM, are likely to make future spatial discruninations of this type quite refined (e.g., 30~400 microns). Temporal Resolution The temporal resolution of studies with PET vary greatly de- pending on the choice of tracer strategy. For example, measure- ments of metabolism require 45 minutes, whereas measurements of blood flow can be made in less than ~ minute. Temporal resolu- tion also varies with the choice of radioisotope (i.e., for OFF the hallife is 110 minutes, whereas for }50, the ha~-life is 2 minutes). The measurement of local brain glucose consumption using ~8F- fluorodeoxyglucose (Phelps et al., 1979; Reivich et al., 1979) requires 45 confutes to accomplish once the tracer has been prepared and

OCR for page 43
56 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES administered. The resulting measurement of local glucose utilization is a summary of events during this entire period of time weighted according to the shape of the arterial concentration of the tracer. A feature of any study of functional activation is the need for second measurements on the same individual (e.g., Peterson et al., 19883. Thus if a control state PET unage ~ subtracted from a stimulated state image to localize areas of activation, then the time necessary to make the measurement as well as for radioactivity to clear from the body becomes very important. The time necessary for radioactivity to clear is governed by the physical, as well as the biological, half- life of the tracer. In the case of t8F this time is about nine hours. Times of this magnitude obviously make difficult a second study in the same individual on the same day. As a result, most functional studies of the human brain with PET and fluorodeoxyglucose involve a single measurement in each subject studied with control measure- ments from a separate group of subjects. When a single subject has been studied more than once with fluorodeoxyglucose and PET, the measurements have usually not been made on the same day. Because of the exquisite sensitivity of blood flow to local changes in neuronal activity plus the speed (40 secondsRaichIe et al., 1983) and ease with which it can be measured and repeated in a single subject (up to 10 measurements in a single sitting), it would appear to be the ideal method for functional mapping with PET. St at istica} Analysis The statistical analysis of PET data is an especially challenging problem. The average PET measurement provides 7 or more slices of the human brain. Within such a data set it is possible to identify a very large number of regions of interest. However, in so doing, one runs into the very difficult problem of multiple comparisons leading to the false identification of regions significantly different from the control. In much of the early PET work this problem was often ignored. Simple paired and unpaired l-tests were used to evaluate large data sets. In retrospect, it is very difficult to know the true significance of much of these data, suggesting that caution be taken in making interpretations. Recently, more sophisticated approaches to the statistical analysis of such data have been developed (see Petersen et al., 1988), which now provide an opportunity to use all the available data.

OCR for page 43
CRITICAL PROBLEMS 57 Risks to Participants The risk to participants in PET unaging studies is small but present due to exposure to the effects of ionizing radiation. The U.S. Food and Drug Administration (FDA) has established strict guidelines for the exposure of normal subjects to ionizing radiation in the course of research. These guidelines reHect limits established as safe for workers over age IB exposed to radiation during the course of their employment. Subjects under age IS are allowed to receive only 10 percent of this dose. These limits are strictly enforced by the FDA through local radioactive drug research committees or investigational new drug applications. It is always viewed as a necessary goal of all studies involving PET not only to keep radiation exposure below the FDA limits but also to keep exposure to the lowest possible amount consistent with obtaining adequate data. Gains can be anticipated in reducing radiation by improving the efficiency of PET imaging devices dedicated to functional brain unaging in normal human subjects. This task already is under way: because of the importance of PET to functional unaging of the human brain, there is a great deal of support for research on ways to reduce radiation exposure to normal subjects undergoing PET scanning. This includes instrumentation developments as well as unproved software for image reconstruction. WAGING TECHNIQUES II: MAGNETIC RESONANCE WAGING While in viva MRI provides highly specific information, it sub fers from a lack of sensitivity. The strength of the signal produced by a particular nucleus is the product of its magnetic moment, its abundance in nature, and its concentration In the tissue. McGeer (1983) calculated this product or "imaging index" for a number of nuclei and compared them with hydrogen. The results indicate that the detectable signal from 3iP or 23Na, the next most pro~sing nu- clei in terms of signal strength, is reduced by a factor of more than 10,000, while the potential signal from i3C ~ more than five orders of magnitude less than hydrogen. Reductions in signal intensity of these magnitudes prohibit the reconstruction of images with the spa- tial resolution exhibited for protons. Thus, MR] must be limited to very low resolution studies for nuclei other than protons.

OCR for page 43
58 BRAIN AND COGNITION: SOME NEW TECHNOLOGIES COGNITIVE STUDIES IN BRAIN ALTERATIONS OR DAMAGE Research on both annals and humans with brain lesions has made unport ant contributions to the understanding of the kinds of information processed by large zones of brain tissue. This is particu- larly true when considering sensory systems such as vision, audition, touch, and olfaction. Recent work on primates has also illuminated more complex processes such as how the brain manages attention (Wurtz et al., 1980~. In humans, the brain-damaged patient has also revealed which hemisphere is involved in language and speech and which lobes are involved in memory, thought, sequencing behaviors, emotions, and other mental phenomena (Ness and Gazzaniga, 1987~. The challenge to this approach comes from the greater sophis- tication of current theories about the composition of a particular mental activity. As cognitive theories become more precise, there is a greater interest in identifying more specific brain correlates. A lesion that produces a disorder in language processing must now be categorized in precise cognitive terms. More pointedly, smaller and more defined brain lesions are sought that may disrupt language in a limited and discrete manner in an effort to confirm current theories about the modular organization of language function (Swinney et al., 1989~. It is still the belief of many students of linguistics that discrete brain lesions may disrupt particular grammatical rules or at least disrupt lexical modules versus phrasal modules. Furthermore, seeing how other processes such as attention or spatial mechanisms contribute to language processing is currently of interest. Here, too, the large lesion approach seriously limits progress in this area (Posner et al., 1988~. Another very important problem is how work on anneals and work on humans relate to one another. Does a lesion in the hip- pocampus create the same problems for the rat or the monkey as it does for the human? Some investigators feel there are analogous findings with similar lesions to such structures (Squire, 1987), al- though others disagree (Lynch and Blurry, 1988~. We now know that structures like the superior colliculus contribute different func- tions in the monkey (Wurtz and Albano, 1980) than in humans (see Holtzman, 1984~. The same is true for lesions to the anterior com- missure (Gazzaniga, 1987~. Such differences argue for caution in attempts to extrapolate results obtained from animals to humans. It is even more difficult to develop implications from the animal work for linguistic and other higher-order processes.

OCR for page 43
CRITICAL PROBLEMS 59 Finally, there are serious problems concerning how to interpret the behavioral changes subsequent to a brain lesion with regard to the role of the brain areas affected in cognitive processing modules. At the cellular level, it is now known that altering inputs to a cell can radically change the postsynaptic receptors of a cell and thereby alter its response characteristics. At the level of systems, how can a lesion that produces damage in site A tell you about site A anymore than it telb you about the network that site A ~ a part of? Site A can be connected to dozens of other centers as well ~ receiving inputs from dozens of other sites. Loss of tonic or inhibitory influences in a neural network can have possible profound effects on the kmds of information a system can process. Although this problem ~ generally recognized, it ~ also generally repressed. Yet, as Francis Crick has said, the lesion approach, whether one likes it or not, is one of the few approaches available to the brain scientist.