Neocortical Memory Circuits

Abstract

P.S. GOLDMAN-RAKIC, S. FUNAHASHI, AND C.J. BRUCE Section of Neuroanatomy, Yale University School of Medicine, New Haven, Connecticut 06510 For many decades, the hippocampus and associated limbic areas of the mammalian central nervous system have been the structures most closely associated with (Scoville and Miner 1957; Mishkin 1982; Amaral et al. 1987). Indeed, the hippocampus is essen- tial for acquiring and consolidating new knowledge about events, people, and places. In the unusual cir- cumstance that both medial temporal lobes are dam- aged, as in the famous case of H.M., is re- duced to that acquired before the injury (Scoville and Milner 1957); after suffering this type of injury, an individual is unable to remember most events for even a few minutes. Other areas of the central nervous system are also important for processes. Among these, the prefrontal cortex has been consistently implicated in mnemonic functions (see, e.g., Warrington and Weis- krantz 1982; Fuster 1985; Kesner 1985; Milner et al. 1985), although the precise nature of these functions has been difficult to define. The question naturally arises as to the character of the prefrontal contribution compared to that of other areas and whether and how it differs from these other areas. In this paper, we review our recent studies on the physiological properties of prefrontal neurons engaged in oculomotor tasks (Funahashi et al. 1989, 1990, 1991). An under- standing of the prefrontal contribution to should allow more precise comparisons with that of the hippocampus and other brain regions. Delayed Response: Basic Probe of Short-term Memory Evidence that the prefrontal cortex is involved in processes comes from the well-established, and often repeated, finding that monkeys with bilateral prefrontal lesions, primarily of the dorsolateral princi- pal sulcus (PS) region of the frontal lobe, are severely impaired on delayed-response tasks (for review, see Goldman-Rakic 1987a; Fuster 1989). As their name implies, these tasks employ a temporal delay to sepa- rate the process of encoding a stimulus from the execu- tion of a response to that stimulus. The simple expedi- ent of introducing a time delay forces the monkey to guide its response on the basis of an internal representa- tion, or memory, of the reward location. In the classic version of the task, after the monkey watches a reward being hidden under the cover of one of two food wells, an opaque screen is lowered to prevent the animal from reaching the hidden reward immediately. Following a 1-5-second delay, the screen is raised and selection of one of the locations is then allowed. Since the place- ment of the concealed item is randomly varied from trial to trial, the animal has to continually update his memory; accordingly, what is relevant on trial n be- comes irrelevant on trial n + 1. Distinct from associa- tive memory, this transient process is analo- gous to the process that cognitive psychologists refer to as memory (Just and Carpenter 1985; Bad- deley 1986). Neuronal Elements of Short-term Memory In a remarkable set of studies, in which monkeys performed the delayed-response task while activity was recorded from individual prefrontal neurons, neuro- physiologists have observed neuronal activity in precise registration with the cue, the delay, or response periods of this task, suggesting that separate and distinct neuro- nal populations encode the input, hold, and output functions required in delayed-response tasks (Kubota and Niki 1971; Fuster 1973; Niki 1974c). These results confirm, at a neuronal level, what could be analytically derived; namely, that performance on a delayed- response task involves several levels of information processing: sensory encoding, temporary storage of in- formation, and response execution. It has been hypoth- esized that the prefrontal cortex contains multiple working each dedicated to a differ- ent information-processing domain, and that within these centers, specific classes of neurons are specialized for carrying out the subfunctions necessary to accom- plish an integrated memory-driven act (Goldman- Rakic 1987a, 1988). Issues in the Study of Short-term Memory Although the basic neuronal elements for construc- tion of a cortical circuit have been identified, some of them are complex, and their role in behavior is not well understood. Fuster (1973) identified at least five types of prefrontal activity in monkeys performing a manual delayed-response task. Subsequently, addi- tional neuronal profiles have been described as inves- tigators have employed a variety of behavioral paradigms and/or variations of delayed-response tasks, including spatial delayed-alternation tasks (Kubota and Niki 1971; Niki 1974a,b), nonspatial delay tasks (Kubota et al. 1980; Rosenkilde et al. 1981; Fuster et al. 1982; Quintana et al. 1988; Yajeya et al. 1988), and