Neuroimaging techniques have recently been used to study cognitive processes, and since the early 1990s they have been used to study brain processes in learning and memory. This is a field of research on learning and memory that clearly could not have taken place without the discovery of brain-imaging techniques. An early review of this area was that of McCarthy (1995); it included material on 14 brain-imaging studies using working memory tasks and six studies investigating processes involved in long-term memory. Another review focused on hemispheric asymmetry in encoding and retrieval (Nyberg, Cabeza, & Tulving, 1996). It showed that encoding involves especially activation of the left prefrontal cortex, whereas retrieval involves especially activation of the right prefrontal cortex. More reports are continuing to appear, and the next edition of this book may very well include a chapter devoted to such studies.
As McCarthy notes, the majority of neuroimaging studies concerned with cognitive processes have been conducted with positron emission tomography (PET) using radioactive tracers such as H21:,0 to measure blood flow or fluorodeoxyglu-cose to measure glucose metabolism. Images are usually recorded under two or more task conditions presumed to differ in the operation of a single cognitive process. As explained by Posner and Raichle (1994), subtraction of images taken in the two conditions reveals activity associated with the distinctive process.
The appearance of Posner and Raichle's Images of Mind (1994) evoked multiple book reviews with commentaries by 27 authors or pairs of authors (Posner & Raichle, 1995). Most commentaries were positive, but seven stated concerns or problems about use or interpretation of the subtraction procedure with regard to PET images. For example, some raised the problem of choice of an appropriate baseline from which to measure effects. Thus, Halgren (1995, p. 358) states that the subtraction procedure is biased against detection of areas that are activated in virtually all cognitive tasks (including control tasks). Horwitz (1995, p. 360) argues that Posner and Raichle put too much emphasis on localization in discrete neural areas and insufficient emphasis on interaction within networks of widely distributed systems. The poor temporal resolution of PET does not permit visualization of such interactions occurring within a few hundred milliseconds. For this reason, Posner, Raichle, and associates (Snyder, Abdullaev, Posner, & Raichle, 1995) have begun to combine the evoked-potential method, with its excellent temporal resolution, with PET to investigate the succession of processes in generating responses to visually presented nouns.
Dehaene (1996) has extended the subtraction technique to show how the additive-factors method can be employed in an experiment recording event-related potentials. During a task in which subjects responded as rapidly as possible to record their comparisons between numbers presented visually, the potentials indicated six successive processing activations occurring in different brain regions, the whole sequence taking place within 500 msec. I am not aware that the additive-factors method has yet been employed in brain-imaging studies.
In many reports, difference images from individual subjects are normalized to a common stereotaxic coordinate system. Comparison of subtraction images from individual subjects is sometimes used to advantage, however, as in the Logan & Grafton (1995) study of conditioning the eye-blink response in human subjects, mentioned earlier in another context. Logan and Grafton found that, in certain brain regions, the relative metabolic change in PET responses correlated significantly with learning performance.
Investigators have also begun to apply the technique of functional magnetic resonance imaging (FMRI) to studies of cognitive processing. This technique exploits the fact that local changes in blood flow related to neural activity alter the concentration of deoxyhemoglobin and the corresponding magnetic resonance signal. With specialized image acquisition techniques, repeated images of the same brain region can be obtained in less than 1 sec, permitting temporal studies of task-dependent activation. "Because FMRI is fast, noninvasive, and does not use ionizing radiation, subjects can be run repeatedly, allowing for more elaborate experimental designs" (McCarthy, 1995, p. 155).
Black and Greenough in Chapter 2 cite FRMI studies that demonstrate structural changes in the human brain caused by learning, changes that may reflect synaptogenesis. Gershberg and Shimamura in Chapter 9 cite findings from imaging studies in their review of roles of different brain regions in different memory processes.
Was this article helpful?