Abstract: A fundamental issue about memory and its different forms is whether learning can occur without the development of conscious knowledge of what is learned. Amnesic patients and control subjects performed a serial reaction time task, exhibiting equivalent learning of an imbedded repeating sequence as measured by gradually improving reaction times. In contrast, four tests of declarative (explicit) knowledge indicated that the amnesic patients were unaware of their knowledge. Moreover, after taking the tests of declarative memory, all subjects continued to demonstrate tacit knowledge of the repeating sequence. This dissociation between declarative and nondeclarative knowledge indicates that the parallel brain systems supporting learning and memory differ in their capacity for affording awareness of what is learned.

Abstract: Some forms of synaptic plasticity depend on the temporal coincidence of presynaptic activity and postsynaptic response. This requirement is consistent with the Hebbian, or correlational, type of learning rule used in many neural network models. Recent evidence suggests that synaptic plasticity may depend in part on the production of a membrane permeant-diffusible signal so that spatial volume may also be involved in correlational learning rules. This latter form of synaptic change has been called volume learning. In both Hebbian and volume learning rules, interaction among synaptic inputs depends on the degree of coincidence of the inputs and is otherwise insensitive to their exact temporal order. Conditioning experiments and psychophysical studies have shown, however, that most animals are highly sensitive to the temporal order of the sensory inputs. Although these experiments assay the behavior of the entire animal or perceptual system, they raise the possibility that nervous systems may be sensitive to temporally ordered events at many spatial and temporal scales. We suggest here the existence of a new class of learning rule, called a predictive Hebbian learning rule, that is sensitive to the temporal ordering of synaptic inputs. We show how this predictive learning rule could act at single synaptic connections and through diffuse neuromodulatory systems.

Abstract: Some forms of synaptic plasticity depend on the temporal coincidence of presynaptic activity and postsynaptic response. This requirement is consistent with the Hebbian, or correlational, type of learning rule used in many neural network models. Recent evidence suggests that synaptic plasticity may depend in part on the production of a membrane permeant-diffusible signal so that spatial volume may also be involved in correlational learning rules. This latter form of synaptic change has been called volume learning. In both Hebbian and volume learning rules, interaction among synaptic inputs depends on the degree of coincidence of the inputs and is otherwise insensitive to their exact temporal order. Conditioning experiments and psychophysical studies have shown, however, that most animals are highly sensitive to the temporal order of the sensory inputs. Although these experiments assay the behavior of the entire animal or perceptual system, they raise the possibility that nervous systems may be sensitive to temporally ordered events at many spatial and temporal scales. We suggest here the existence of a new class of learning rule, called a predictive Hebbian learning rule, that is sensitive to the temporal ordering of synaptic inputs. We show how this predictive learning rule could act at single synaptic connections and through diffuse neuromodulatory systems.

Abstract: I n t r o d u c t i o n The water maze is a widely used spatial learning task (Morris 1981, 1984) in which rodents are trained to escape from water onto a hidden platform located within a large, usually circular, pool. Over a series of trials, rats (or mice) will rapidly learn to take relatively direct paths to the platform, indicating that they may have learned its spatial location. Transfer tests, sometimes called probe tests, in which the platform is removed from the pool, offer further evidence of true spatial learning, as trained animals display a search pattern focusing on the platform location in the training quadrant. Performance in this task has provided a simple and convenient assay of the effects of drugs or brain lesions on cognitive function in rats (for review, see Brandeis et al. 1989; McNamara and Skelton 1993), and its use with mice has revealed covariation between spatial learning and strain-specific differences in neuroanatomical circuitry (Schwegler et al. 1988) and the effects of targeted deletion of genes putatively involved in synaptic plasticity (Grant et al. 1992, Silva et al. 1992). Despite its efficacy, the task as it is usually run has a number of limitations. First, animals may develop strategies for finding the platform, such as circling at roughly the appropriate distance from the sidewall, which do not actually require them to learn where the platform is located. Such strategies can, nevertheless, lead to a steady decline in escape latency that may be falsely interpreted as spatial learning. Second, during a transfer test, animals may initially search the training quadrant but, failing to locate the now-removed platform, move elsewhere in the pool. If this tendency to move elsewhere differs across t reatment groups, an analysis of time (or distance) spent in the training quadrant throughout the entire transfer test will be a confounded index of the extent of spatial learning. Animals that go directly to the platform area and then search elsewhere because they fail to find the now-removed platform may be scored incorrectly as poor learners. Third, because transfer tests, as conventionally run are extinction trials (i.e., no escape is possible), the use of multiple transfer tests may compromise the asymptote reached in learning or even, because learning is so rapid, encourage animals to search outside the trained area when they notice that the platform is absent. Thus, performance with multiple transfer tests, while desirable in pharmacological work, may underest imate the true level of learning. Buresova et al. (1985) designed an ingenious on-demand platform 3Corresponding author, which, they suggested, would distinguish true spatial mapping from other

Abstract: In cerebellar long-term depression (LTD), conjunctive stimulation of parallel fiber (PF) and climbing fiber (CF) inputs to a Purkinje neuron result in a selective depression of PF-Purkinje neuron synaptic strength. This system is attractive for the study of neuronal information storage, both because of its duration and because it demonstrates input specificity. The mechanisms underlying input specificity in this system are not known, but they could involve presynaptic alterations, postsynaptic alterations, or some combination of both. To allow for an unambiguous analysis of postsynaptic processes, an LTD induction protocol has been developed using cultured cerebellar cells in which pulses of quisqualate and direct Purkinje neuron depolarization replace PF and CF stimulation, respectively. Input specificity is retained in this reduced system. When multiple, nonoverlapping quisqualate application sites are used, LTD is confined to those sites that are stimulated during depolarization. This property of LTD induction is also preserved under conditions where both spontaneous and evoked neurotransmitter releases are reduced or eliminated, indicating that postsynaptic alterations are sufficient to confer input specificity. Input-specific LTD may also be induced by local application of a protein kinase C (PKC) activator (1-oleoyl-2-acetylglycerol) together with direct Purkinje neuron depolarization, suggesting that input-specific LTD results from the conjunction of a spatially broad Ca signal mediated by Purkinje neuron depolarization, together with a spatially constrained PKC-activating signal mediated by quisqualate application.

Abstract: We examine the hypothesis that the form representation in the anterior inferotemporal (AIT) cortex is acquired through learning. According to this hypothesis, perceptual aspects of the temporal association area are closely related to its visual representation, in that the response selectivity of AIT neurons can be influenced by visual experience. On the basis of the neurophysiological evidence, we summarize two neuronal mechanisms that subserve the acquisition of form selectivity in AIT neurons. The first mechanism is neuronal tuning to particular stimuli that were learned in a cognitive task. The second mechanism is association, with which relevant information can be retrieved from other stored memories. On the grounds that long-term memory of objects is acquired and organized by at least these two neuronal mechanisms in the temporal association area, we further present a model of the cognitive memory system that unifies perception and imagery.

Abstract: We investigated activation of the two major neuronal protein kinase C (PKC) isoforms in Aplysia, Ca(2+)-activated Apl I and Ca(2+)-independent Apl II, during the induction and maintenance of behavioral sensitization of Aplysia defensive reflexes. Activation of PKC occurred during the training stimulus and persisted for at least 2 hr thereafter but was not maintained for 24 hr. The persistent activation required protein synthesis and was blocked by cyproheptidine, an agent that also blocked the initial activation of PKC. Persistent activation involved both an increase in membrane-associated Apl I and an increase in an autonomous kinase activity that may be related to a post-translational modification of Apl II. These results are consistent with the hypothesis that in addition to its role in producing the presynaptic facilitation of mechanosensory-motor neuron synapses that underlie short-term facilitation, PKC is needed for maintaining synaptic changes in an intermediate period that precedes the modifications accompanying consolidation of memory.

Abstract: A hallmark of many forms of classical conditioning is a precise temporal specificity: Learning is optimal when the conditioned stimulus (CS) slightly precedes the unconditioned stimulus (US), but the learning is degraded at longer or backward intervals, consistent with the notion that conditioning involves learning about predictive relationships in the environment. To further examine the cellular mechanisms contributing to the temporal specificity of classical conditioning of the siphon-withdrawal response in Aplysia, we paired action potential activity in siphon sensory neurons (the neural CS) with tail nerve shock (the US) at three critical time points. We found that CS-US pairings at short (0.5 sec) forward intervals produced greater synaptic facilitation at sensorimotor connections than did either 0.5-sec backward pairings or longer (5 sec) forward pairings, as reflected in a differential increase in both the amplitude and rate of rise of the synaptic potential. In the same preparations, forward pairings also differentially reduced the sensory neuron afterhyperpolarization relative to backward pairings, suggesting that changes in synaptic efficacy were accompanied by temporally specific changes in ionic currents in the sensory neurons. Additional experiments demonstrated that short forward pairings of sensory cell activity and restricted applications of the neuromodulatory transmitter serotonin (normally released by the US) differentially enhanced action potential broadening in siphon sensory neurons, relative to backward pairings. Taken together, these results suggest that temporally specific synaptic enhancement engages both spike-width-dependent and spike-width-independent facilitatory processes and that activity-dependent enhancement of presynaptic facilitation may contribute to both the CS-US sequence and proximity requirements of conditioning.

Abstract: We have previously reported that CCK-A receptor agonists and CCK-B receptor antagonists both enhance memory in an olfactory recognition test. Here, we report that the memory-enhancing effect of the CCK-B receptor antagonist L-365,260 (1 mg/kg i.p.), but not that of the CCK-A receptor agonist caerulein (0.03 mg/kg i.p.), was dramatically decreased following a bilateral transection of the perforant path, a principal source of input to the hippocampal formation. We further confirmed that a significant memory deficit occurred subsequent to this deafferentation of the hippocampus in untreated animals. In contrast, the effect of caerulein, but not that of L-365,260, was abolished following a bilateral subdiaphragmatic vagotomy. These results demonstrate that the hippocampal system plays a role in olfactory recognition and indicate that distinct neuronal pathways underlie the memory-enhancing effects of CCK-A and CCK-B drugs observed in the olfactory recognition test. The former effects (CCK-A) appear to involve a peripheral relay to the brain via the vagus nerve, whereas the latter (CCK-B) are directly central and involve, at least in part, the hippocampal system.

Abstract: Recent studies indicate that stimulation of NMDA receptors in cultured hippocampal cells activates MAP kinase. Although the pathway whereby MAP kinase is activated has been been characterized, little is known about the mechanisms that shut off MAP kinase. In the course of analyzing several immediate-early genes identified previously by differential screen as inducible by seizure activity, we found that one of them, BAD2, encodes dual purpose, threonine/tyrosine phosphates with specific activity directed against MAP kinase (MKP-1). In situ hybridization of BAD2 demonstrates that stimuli that produce seizure, kindling, and long-term potentiation cause a rapid increase in BAD2 mRNA (within 0.5-1 hr after stimulation) that has, in each case, a distinctive pattern of expression in the brain. In these regions, the induction of a MAP kinase-specific phosphatase may provide a negative feedback control associated with long-term synaptic changes.

Abstract: 7-Nitro indazole (7-NI), which is selective for the neuronal isoform of nitric oxide synthase (NOS), was tested in a passive avoidance task in the chick. Injection of 50 mg/kg i.p. pretraining had amnesic effects for the task when tested 30 min, 2 or 24 hr after training. Injections post-training had no effect. Because 7-NI does not inhibit the endothelial isoform of NOS, it does not affect blood vessel relaxation, as nonspecific inhibitors do. This effect on blood vessels could explain the amnestic effects produced by nonspecific NOS inhibitors. The results support the theory that NO is a neuronal transmitter that is important in processes of synaptic plasticity and learning.

Abstract: Long-term desensitization of AMPA receptors (LTDA) is a core mechanism of long-term depression, a model of motor learning in the cerebellum. In this study we investigated the expression of neurotrophic factor genes after induction of LTDA in cultured cerebellar slices. LTDA was induced by application of quisqualate and monitored as a population response with a wedge recording technique. The levels of mRNA were quantified by reverse transcription followed by polymerase chain reaction. Quisqualate, at a dose and duration that reliably induced LTDA, elicited a significant and specific increase in BDNF mRNA with a peak at four hours after the application. By cell fractionation, the major source of BDNF mRNA increase was found to be in granule cells. In addition, a small but significant increase of transcripts with specific exon usage was observed in a Purkinje cell fraction. These results indicate that BDNF may be coinduced with LTDA and suggest that the slow and sustained increase of BDNF mRNA might play a role in later phases of synaptic plasticity in the cerebellum.

Abstract: Programs of gene activation may underlie long-term adaptive cellular responses to extracellular ligands. We have used a differential cDNA cloning strategy to identify genes that are strongly induced by excitatory stimuli in the adult rat hippocampus. Here, we report the rat cDNA sequence of a zinc-finger transcription factor, Egr3/Pilot, and characterize its regulated mRNA expression in brain. Egr3 mRNA is rapidly and transiently induced in neurons of the hippocampus and cortex by electroconvulsive seizure. mRNA levels peak 2 hr after the seizure and remain elevated for as long as 8 hr. Egr3 mRNA is also rapidly induced in granule cells of the dentate gyrus by synaptic NMDA receptor activation elicited by patterned stimulation of the perforant pathway and by drugs that alter dopamine neurotransmission in the striatum. Basal levels of Egr3 mRNA in the cortex appear to be driven by natural synaptic activity because monocular deprivation rapidly decreases Egr3 mRNA in the deafferented visual cortex. Aspects of the protein structure, sequence-specific DNA binding, transcriptional activity, and regulation of Egr3 are highly similar to another zinc-finger transcription factor, Egr1/zif268. Moreover, we demonstrate colocalization of Egr3 and zif268 mRNAs in neurons of normal and stimulated cortex. Our studies suggest that interactions between these coregulated transcription factors may be important in defining long-term, neuroplastic responses.

Abstract: We investigated sensorimotor skill learning, a form of nondeclarative (implicit) memory, in 28 subjects with declarative (explicit) memory defects caused by either mesial temporal (n = 15) or basal forebrain (n = 13) damage and in 66 normal control subjects. All 28 amnesics had normal learning of a rotor pursuit task. We also studied in detail the sensorimotor skill learning of patient Boswell. As a result of bilateral damage to both mesial and lateral aspects of the temporal lobes and to the basal forebrain, Boswell has one of the most severe impairments ever reported for learning of all types of declarative knowledge. Compared to matched controls, Boswell acquired and retained normally the skills associated with performing motor tasks. We conducted a long-term (2-year) followup study of Boswell's retention of the rotor pursuit task, and we found that he retained the skill as well as normal controls. Our study builds on previous work in the following respects: (1) It provides evidence, for the first time, that skill learning is normal in basal forebrain amnesics; (2) it shows that patient Boswell has normal learning and long-term retention of sensorimotor skills, in spite of his extensive damage; and (3) it offers additional evidence that mesial temporal lobe damage spares skill learning. These findings demonstrate unequivocally that sensorimotor skill learning does not require structures in mesial and lateral temporal regions nor in basal forebrain.

Abstract: Certain kinds of learning may be related to potentiation of transmission at specific hippocampal synapses. We investigated whether transmission across the perforant-path/granule-cell synapses of the dentate gyrus is facilitated when rats are learning about novel objects in an open field during exploration. Such studies are complicated by the sensitivity of hippocampal field potentials to brain temperature change. To control for this, we have recorded both brain temperature and field potentials and compared potentials sampled during exploration with potentials taken at corresponding brain temperature in a passive warming situation, with the animals at rest. Relative to these reference potentials, both the f-EPSP slope and the population spike were elevated while the rats explored. The potentiation reached its maximum within < 5 sec after the exploration began. During the first 2 min, the f-EPSP slope was enhanced by 6.5% relative to the control values. The potentiation then decayed, reaching the reference values after 20-30 min of exploration. Significant potentiation required exploration above a certain minimum intensity. Control experiments showed that the changes were neither mimicked by arousal in response to aversive stimuli nor by motor activity. It is suggested that the facilitated transmission across the perforant-path/dentate synapses may be involved in learning during exploration.

Abstract: The effects of hippocampal lesions on the conditioning of fear responses (freezing responses) to contextual stimuli (static, continuously present stimuli) were examined in three conditioning paradigms: forward pairing of a phasic tone conditioned stimulus (CS) with a footshock unconditioned stimulus (US), unpaired presentations of the CS and US, or presentations of the US alone. All three procedures resulted in the acquisition of conditioned freezing to contextual stimuli. Lesions of the dorsal hippocampus prevented the acquisition of contextual conditioning in the Paired procedure, as reported previously, but not in the Unpaired or US Alone procedures. In the Paired procedure, static contextual cues occur in the background, with the phasic tone CS being the primary stimulus that enters into the association with the US. However, in the other two procedures, where there is no phasic CS, the primary associations with the US involve static contextual stimuli, which are therefore in the foreground. We refer to these types of contextual conditioning as background and foreground contextual conditioning, respectively, and argue that the hippocampus is only involved in background contextual conditioning. These results have implications for understanding both fear conditioning and hippocampal function.

Abstract: Amnesic patients and control subjects participated in a study of probabilistic classification learning. In each of three tasks, four different cues were each probabilistically associated with one of two outcomes. On each trial, the cues could appear alone or in combination with other cues and subjects selected the outcome they thought was correct. Feedback was provided after each trial. In each task, the amnesic patients learned gradually to associate the cues with the appropriate outcome at the same rate as control subjects, improving from 50% correct to approximately 65% correct. Presumably because the cue-outcome associations were probabilistic, declarative memory for the outcomes of specific trials was not as useful for performance as the information gradually accrued across trials. Nevertheless, declarative memory does appear to make a contribution to performance when training is extended beyond approximately 50 trials, because with further training control subjects eventually outperformed the amnesic patients. It was also demonstrated that performance on the probabilistic classification task was not the result of holding knowledge of cue-outcome associations in short-term memory, because both control subjects and amnesic patients demonstrated significant savings when testing was interrupted by a 5-min delay (experiment 2). Probabilistic classification learning appears to provide an analog in human subjects for the habit learning tasks that can be acquired normally by animals with hippocampal lesions.

Abstract: Although mild progressive memory impairment is commonly associated with normal human aging, it is unclear whether this phenomenon can be explained by specific structural brain changes. In a research sample of 54 medically healthy and cognitively normal elderly persons (ages 55-87, x = 69.0 +/- 7.9), magnetic resonance imaging (MRI) was used to derive head-size-adjusted measurements of the hippocampal formation (HF) (dentate gyrus, hippocampus proper, alveus, fimbria, subiculum), the superior temporal gyrus (STG), and the subarachnoid cerebrospinal fluid (CSF) (to estimate generalized cerebral atrophy). Subjects were administered tests of primary memory (digit span) and tests of secondary memory with immediate and delayed recall components (paragraph, paired associate, list recall; facial recognition). Separate composite scores for the immediate and delayed components were created by combining, with equal weighting, the subtests of each category. The WAIS vocabulary subtest was used as a control measure for language and intelligence. A highly significant correlation (P < 0.001), independent of age, gender, and generalized cerebral atrophy was found between HF size and delayed memory performance. No significant correlations were found between HF size and primary or immediate memory performance. STG size was not significantly correlated with any of the composite memory variables. These results suggest that HF atrophy may play an important independent role in contributing to the memory loss experienced by many aging adults.

Abstract: In the CA1 region of adult guinea pig hippocampal slices, long trains of theta frequency (5 Hz) stimulation produced a small enhancement of basal synaptic transmission but depressed the strength of synaptic transmission at synapses that had recently undergone long-term potentiation (LTP). Five hertz stimulation delivered immediately prior to high-frequency stimulation also inhibited the subsequent induction of LTP. The depression of potentiated synapses by 5 Hz stimulation (depotentiation) was blocked by 2-amino-5-phosphonovalerate and was observed only during the early phases of LTP. Furthermore, the protein phosphatase inhibitors okadaic acid and calyculin A blocked both depotentiation and the ability of 5 Hz stimulation to inhibit subsequent LTP, suggesting that protein phosphatases are involved in the ability of 5 Hz stimulation to modulate synaptic plasticity in the CA1 region of the hippocampus.