Short-term and long-term retention of experimentally presented words were compared in a sample of 33 healthy normal volunteers by the [150]H20 method with positron emission tomography (PET). The design included three conditions. For the long-term condition, subjects thoroughly studied 18 words 1 week before the PET study. For the short-term condition, subjects were shown another set of 18 words 60 sec before imaging, with instructions to remember them. For the baseline condition, subtracted from the two memory conditions, subjects read a third set ofwords that they had not previously seen in the experiment. Similar regions were activated in both short-term and long-term conditions: large right frontal areas, biparietal areas, and the left cerebellum. In addition, the short-term condition also activated a relatively large region in the left prefrontal region. These complex distributed circuits appear to represent the neural substrates for aspects of memory such as encoding, retrieval, and storage. They indicate that circuitry involved in episodic memory has much larger cortical and cerebellar components than has been emphasized in earlier lesion studies.Human memory is a complex phenomenon that has been variously conceptualized by cognitive neuroscientists, on the basis of data from lesion studies, animal experiments, and in vivo microelectrode stimulation of the human brain (1-9). Strength and ease of memory encoding and retrieval appear to be linked to a variety of factors, such as novelty of the information, amount of practice, duration and frequency of exposure, depth of processing, or the origin of the information and the relationship of the individual to it (i.e., intrinsic vs. extrinsic, participant vs. nonparticipant) (6,(10)(11)(12). A fundamental question embedded in memory research is whether memory is a unitary construct. Since the original subdivision into primary and secondary by William James, models have been developed to classify it in various ways, invoking constructs such as procedural vs. declarative, implicit vs. explicit, episodic vs. semantic, working vs. reference, or short-term vs. long-term (3, 6, 10-21). Modem cognitive neuroscience seeks to link these models to their neural substrates. In vivo neuroimaging techniques such as positron emission tomography (PET) have provided a new approach to studying these models by permitting measurement of physiological activity while the brain is performing specific tasks (13-27). While lesion studies in animals and humans permit inferences about brain organization by indicating what fails to work when parts of the brain have been damaged, neuroimaging studies permit inferences based on measuring changes in metabolic activity during performance of tasks that can be experimentally controlled.Here we report on a PET experiment designed to examine the neural substrates of recognition of previously learned verbal material. We compared long-term and short-term re-The publication costs of this article were defrayed in part by page charge payment. T...
