Neuroscience & Biobehavioral Reviews, Vol. 22, No. 2, pp. 181–193, 1998 ᭧ 1998 Elsevier Science Ltd. All rights reserved Pergamon Printed in Great Britain 0149-7634/98 $32.00 + .00 PII: S0149-7634(97)00005-5

Testing the NMDA, Long-term Potentiation, and Cholinergic Hypotheses of Spatial Learning

DONALD PETER CAIN*

Department of Psychology and Graduate Program in Neuroscience, University of Western Ontario, London, Ontario N6A 5C2, Canada

CAIN, D. P. Testing the NMDA, long-term potentiation, and cholinergic hypotheses of spatial learning. NEUROSCI BIOBEHAV REV 22(2), 181–193, 1998.—The problems and issues associated with the use of pharmacological antagonists in studies on learning and memory are considered in a review of the role of N-methyl-d-aspartate (NMDA) receptors, NMDA receptor-mediated long-term potentiation (LTP), and muscarinic receptors in spatial learning in the water maze. The evidence indicates that neither NMDA nor muscarinic receptors, nor NMDA receptor-mediated LTP, are required for spatial learning, although they might normally contribute to it. Detailed behavioral analyses have indicated that the water maze task is more complex than generally has been appreciated, and has a number of dissociable components. Naive rats trained under NMDA or muscarinic antagonism display sensorimotor disturbances that interfere with their ability to acquire the task. Rats made familiar with the general requirements of the task can learn the location of a hidden platform readily under NMDA or muscarinic antagonism. The ability of a rat to acquire the water maze task depends on its ability to apply instinctive behaviors to performance of the task in an adaptive manner. The instinctive behaviors undergo modification as the rat learns the general strategies required in the task. The evidence suggests that at least some of the plastic changes involved in acquiring the task occur in existing neural circuits situated in widespread areas of the brain, including sensory and motor structures in the cortex and elsewhere, and are therefore difficult to distinguish from existing sensorimotor mechanisms. More generally, the findings indicate the difficulty of inferring the occurrence or nonoccurrence of learning from behavior, and the difficulty of causally linking the action of particular receptor populations with the formation of specific memories. ᭧ 1998 Elsevier Science Ltd. All rights reserved.

N-Methyl-d-aspartate NMDA Scopolamine Muscarinic receptor Spatial learning Water maze Sensorimotor disturbances Behavior

CERTAIN PERIODS in the history of brain research have More recently, behavioral neuroscience has experienced been dominated by a search for the mechanisms of learning a strong resurgence of interest in the engram. Two develop- and memory. Karl Lashley, the father of modern behavioral ments during the 1970s were probably important in this neuroscience, spent much of the period from 1920 to 1950 resurgence. First, Eric Kandel and his colleagues set out searching for the engram in a long series of learning experi- to map the nervous system of Aplysia, a simple invertebrate, ments involving rats and monkeys with brain lesions. For and to identify the neurons and events required for simple Lashley the engram represented the sites of lasting change in forms of behavioral learning (14). Second, Bliss and Lomo the nervous system—presumably specific synapses—that (10) discovered long-term potentiation (LTP), a relatively were associated with particular instances of behavioral learn- long-lasting form of neural plasticity that could be artificially ing, and were necessary for it to occur. His approach of beha- induced in the brain of mammals by electrical stimulation. viorally training rats to discriminate visual stimuli or negotiate The first development showed that detailed knowledge of mazes set the tone of such research for years. Lashley’s con- relevant neural circuitry was beneficial and perhaps even tributions are usually, but incompletely, summed up by the necessary, and that this knowledge, together with electro- ‘‘principle of mass action’’, the conclusion that for complex physiological analysis, could identify the nature and location tasks such as learning to traverse a maze, the amount learned of the engram in a simple animal. The second development and remembered was proportional to the amount of remaining provided a laboratory model of the kind of synaptic plasticity neocortex; the location ofthe damage mattered little. Although that could underlie complex forms of learning in mammals the common view that Lashley failed in his attempts to locate (9). LTP appeared to embody the activity of the ‘‘Hebb the engram for some forms of learning is mistaken (104), the synapse’’, a modifiable synapse that Donald Hebb, Lashley’s perceived negativity of his findings, and the accumulating student, proposed in 1949 as a learning mechanism. evidence for the localization of function in the neocortex, led The idea that the hippocampus was uniquely important many researchers to discredit his conclusions. for engram formation stemmed from the discovery in the

* Tel.: +1 519 6792111, ext. 4628; Fax: +1 519 6613961; E-mail: [email protected]

181 182 CAIN

1950s that human patients with hippocampal removals movement by visual and other sensory input, climbing experienced severe disturbances in learning and memory onto and standing on the platform), there must be pre- (87). Later, based on animal research, O’Keefe and Nadel existing neural circuits that the rat brings to the water maze (71) proposed that the hippocampus served as a spatial map. task for its successful acquisition (51,100,102,104,117), in Their hypothesis was supported by the finding that indivi- addition to engrams that might be formed by neuroplastic dual complex spike cells in the hippocampus, called place processes during training. cells, fired when an animal was in a specific location within Successful acquisition of the water maze task requires an an environment (70). This suggested that place cells coded animal to navigate efficiently from any starting point in the spatial locations in the environment, giving anatomical and pool to a specific location containing the hidden platform. functional support to the hypothesis that the hippocampus Whishaw and Mittleman (115) showed that normal rats served as a spatial map that was crucial for spatial learn- made use of a number of different spatial navigation ing—the learning of a location in the environment in strategies in acquiring the water maze task, and that these relation to cues in that environment. The fact that damage strategies were considerably more complex than had been to the hippocampus or related structures severely disrupted thought. They suggested that ‘‘some deficits in spatial acquisition of tasks that required spatial information was navigation that follow brain damage may be as easily consistent with their hypothesis (65,69). attributed to the loss of search strategies as to the loss of a Further impetus for the idea that the hippocampus might hypothetical subsystem of memory’’ (p. 428). This general be important for engram formation came from research approach led to an extensive series of experiments, the showing that LTP occurred most readily in the hippo- conclusion of which was that muscarinic antagonists disrupt campus. Although hippocampal LTP has been theoretically a sensorimotor subsystem important for the selection of linked to various kinds of learning, it has been most closely behavioral strategies required to solve the water maze linked to spatial learning (15,63,64). LTP at most hippo- task, and do not prevent spatial learning per se campal synapses was found to be dependent on activity at N- (109,110,117,118). The question whether experimental methyl-d-aspartate (NMDA) receptors (17,34,64), and treatments affect the selection of behavioral strategies for NMDA receptor antagonists were found to severely disrupt acquiring the water maze task, or even basic sensory acquisition of spatial learning tasks (63,64). This suggested processing or motor control mechanisms, is important for that NMDA receptor-mediated LTP in the hippocampus understanding drug and lesion effects. established the engram for spatial learning. The relevance of this question is that all conclusions This review will evaluate the hypothesis that NMDA about the neural mechanisms of learning, conceived of as receptor-mediated LTP in the hippocampus is necessary changes in conduction through specific neural circuits, are for establishing the engram for spatial learning. The number inferences from observable behavior. For inferences about of laboratory tasks that could potentially involve aspects of spatial learning to be valid, the behavioral changes cannot spatial learning for their acquisition is large, so the discus- have resulted from the disruptive effects of an experimental sion will focus on selected studies involving the water maze. treatment on the mechanisms involved in producing the This task requires a rat or mouse to swim in a circular pool instinctive behaviors required to perform the task. Under- of water 1–2 m in diameter until it finds and climbs onto a standing the effects of treatments on instinctive behaviors, small platform hidden just beneath the surface, which as well as on the engram, is vital for arriving at a useful and provides the only refuge. This task is well suited for the complete understanding of learning and memory (102,104). study of spatial learning because it requires the animal to The difficulty of knowing whether experimental treat- learn to swim to a specific location using visual cues in the ments have affected pre-existing neural circuits that produce room (61,71,96). Normal rats learn to do this quickly, but instinctive behavior, or the engram, or both, is a very diffi- rats given NMDA antagonists or hippocampal lesions do cult problem for even the simplest forms of learning. A poorly on this task (63–65). further complication is the fact that instinctive behaviors In addition to a hypothesized role for NMDA receptor- themselves can be modified by experience. There are mediated hippocampal LTP in spatial learning, there is evi- many examples of this from a large variety of species. For dence that other neurochemical mechanisms are relevant. example, instinctive behavior in insects, birds, and mam- For example, muscarinic receptor antagonists produce mals can depend directly on the animal’s prior experience acquisition deficits that are comparable to those produced with the stimuli that release or evoke it (1,32,86,104). by NMDA antagonists (96). Therefore studies involving Further, the nature and extent of modifications by experi- muscarinic antagonists also will be discussed. ence can depend on the type of instinctive behavior that is measured (80). Some basic human sensorimotor behaviors are well known to be to modifiable (38). All of this suggests LEARNED AND INSTINCTIVE BEHAVIOR IN THE WATER MAZE that both learned and instinctive behaviors involve both pre- Spatial navigation plays a crucial role in many important existing and acquired circuits (104). Said another way, behaviors in the rodent. Foraging for food, reproductive and ‘‘…there is no evidence to justify the a priori assumption parental behavior, escaping predation to a known safe loca- that the control of learned and the control of instinctive tion, and returning to the nest site depend on the ability to behavior are quite separate processes.’’ ((37), p. 169). For navigate accurately in space. The fact that normal rats can the water maze task this means that the instinctive behaviors acquire the water maze task in just a few trials suggests that required in the task might well be modified by experience in their nervous system was shaped during evolution to the pool; indeed, this modification might be a prerequisite produce, among other things, accurate spatial navigation. for successful acquisition of the task. This topic will be dealt As the basic behaviors involved in performing the task are with below in a discussion of the nonspatial pretraining instinctive (movements of the limbs, the guiding of effect. TESTING THE NMDA, LONG-TERM POTENTIATION, 183

WATER MAZE PERFORMANCE AND SENSORIMOTOR instinctive behaviors is required for successful acquisition DISTURBANCES of the water maze task, then this modification might itself involve establishment of an engram related to the instinctive The intertwined nature of learned and instinctive behaviors. This would further complicate the task of separ- behaviors, and of acquired and pre-existing neural circuits, ating the effects of experimental treatments on various com- raises the question of whether it is possible to devise an ponents of behavior that contribute to the overall experimental approach to separate the effects of experimen- performance of the task. However, a beginning in dealing tal treatments on the engram for spatial learning from effects with this question can be made by analyzing a broad battery on instinctive behaviors required to perform the task. If, as of behaviors that includes both sensorimotor function and suggested in the previous paragraph, the modification of instinctive behaviors required in the task, in addition to

NMDA Receptor Muscarinic Receptor 60 Water Maze Trial Blocks 60 Water Maze Trial Blocks

Naive SCO 50 50 Pretrained SCO Naive NPC Pooled Controls 40 Pretrained NPC 40 Pretrained Controls Pooled Controls Pretrained Controls 30 30

20 20 Search Time (sec) Search Time (sec)

10 10

0 0 Hidden Platform Visible Platform Hidden Platform Visible Platform

35 Hidden Platform Probe Trial 35 Hidden Platform Probe Trial Naive NPC Naive SCO 30 30 Pretrained NPC Pretrained SCO Pooled Controls Pooled Controls 25 Pretrained Controls 25


Pretrained Controls Random Platform Random Platform 20 20

15 15

Search Time (sec) 10 Search Time (sec) 10

5
 5 






0 0 Before Training After Training Before Training After Training

100 Percent Direct Swims100 Percent Direct Swims

90  90 

 80 80 70 70 60 60 50 50 Percent Percent 40 40

30
30
20 20 10 10 0 0 Naive Pretr Pooled Rand Naive Pretr Pooled Naive Pretr Pooled Rand Naive Pretr Pooled NPC NPC Contr Platf NPC NPC Contr SCO SCO Contr Platf SCO SCO Contr

Hidden Platform
 Visible Platform
Hidden
Platform Visible Platform FIG. 1. Hidden and visible platform search time (top), hidden platform quadrant dwell time during the probe trial (middle), and percent direct swims to the hidden platform (bottom) for rats given NPC17742, a competitive NMDA receptor antagonist (left) or scopolamine, a muscarinic receptor antagonist (right). Contr, controls; NPC, NPC17742; Platf, platform; Pretr, nonspatially pretrained; Rand, random; SCO, scopolamine. Adapted from Ref. (83). 184 CAIN behavioral measures intended to quantify the particular kind found that atropine sulphate or scopolamine, another of learning that one is interested in. This approach is useful muscarinic antagonist, reliably increased visible platform because it can reveal treatment-induced disturbances in search time ((31,72,83,102); Fig. 1). A possible explanation behavior that have the potential to affect the behavioral for the different results is the nature of the pool and plat- measures of learning, and thus the inferences made from form, which determine the difficulty of the task. The more them. Further approaches to this general problem will be salient the visible platform is against the pool wall or the discussed in the next section. less difficult the task, the less effect muscarinic antagonists Measures for quantifying spatial learning in the water appear to have on the visible platform task (72,117). This maze have included hidden platform search time (elapsed suggests the possible involvement of sensory deficits in the time from release into the water to climbing onto the plat- effect of muscarinic antagonists on the visible platform task, form (60,61)), platform quadrant dwell time (time spent a topic to be discussed below. swimming in the quadrant of the pool that formerly con- Whishaw’s work appears to provide an explanation for tained the hidden platform during training trials (93,96)), some of these findings. He was among the first to emphasize and swim accuracy as indicated by heading error (96) or the the complexity of the water maze task, which requires a rat directness of the swim path to the hidden platform ((118); to learn how to cope with the task by selecting an appro- Fig. 1). Since these measures depend on the instinctive priate strategy before it can learn the spatial location of the behaviors described earlier, they can be affected by treat- hidden platform (109). This approach to understanding how ments that disrupt relevant sensorimotor mechanisms. the rat acquires this task is essentially the same one that A practical way of proceeding is by focusing on sensor- Lashley (51) and Krechevsky (47) developed many years imotor function both in the performance of the task itself, ago (117). The rat must learn a number of things before it and in other sensorimotor tasks conducted outside the spe- can solve the task: to swim; to swim away from the wall; cific learning task. Data from the latter source can provide that the platform provides the only refuge; and to use the independent confirmation of sensorimotor disturbances that platform by climbing onto and remaining on it. Only then might have affected mechanisms for instinctive behaviors can it engage a strategy to learn the spatial location of the required in the task. The more similar the outside sensor- hidden platform in relation to cues in the room. imotor tasks are to behaviors actually required for perfor- However, rats trained under a muscarinic antagonist do mance of the task, the more confident one can be that the not appear to learn these strategies for maze acquisition as information they provide is directly relevant to interpreting readily as normal rats do. Instead, they persist in swimming the effects of an experimental treatment. around the perimeter of the pool, close to the wall The earliest use of a test of sensorimotor function in the (72,109,118). Whishaw and Tomie (118) were the first to water maze was the visible platform task (60,61). Use of a quantify swimming in the periphery of the pool by analyz- platform that protruded above the surface of the water ing the distance swum in the outer 50% of the pool during allowed for the evaluation of basic swimming ability, the posttraining probe trials. This study used a visible platform ability of a rat to guide its swimming using visual cues, and during training, which reduced the rats’ need to use a spatial the ability to climb onto and remain on the platform. An strategy. Rats trained under atropine sulfate swam more impairment in the visible platform task was taken as than 80% of the total distance in the outer portion of the evidence that the experimental treatment affected either pool, a significantly greater distance than controls. The rats sensorimotor abilities other than spatial learning per se, or also displayed fewer search behaviors (orienting move- the motivation to escape the water (61). Unfortunately, not ments, head elevations, tight circles, pauses), and ignored all water maze studies have made use of this important novel cues placed around the pool. These results were control. One of the first to do so reported that hippocampal consistent with the suggestion that atropine impaired the lesions produced a small but reliable increase in visible use of a sensorimotor subsystem for spatial navigation, or platform search time, in addition to a much larger increase impaired vision, or both. Paylor and Rudy (72) obtained in hidden platform search time (65). similar results, and further suggested that the long swim The visible platform task was used in a study with times that resulted from excessive swimming in the periph- MK801, a noncompetitive antagonist of NMDA receptors, ery of the pool would be expected to fatigue the animals, where it was found that MK801 increased visible platform resulting in further increases in platform search time. search time only at a dose (0.08 mg/kg) greater than that In our experiments with NMDA and muscarinic antago- required to increase search time in the hidden platform task nists we measured the time swum in the periphery of the (0.05 mg/kg) (79). The first studies to use the visible plat- pool throughout hidden platform training using a computer- form task with competitive NMDA and non-NMDA excitatory based automatic tracking system. The platform was never in amino acid receptor antagonists reported that the antagonists the periphery, and therefore could not be encountered by a increased visible platform search time at the same doses that rat that spent all of its time swimming there. Rats given an increased hidden platform search time ((11,82,83); Fig. 1), NMDA or muscarinic antagonist spent 70–85% of the time raising the possibility that these treatments caused the swimming in the periphery during training, a significantly hidden platform task deficits by an action on sensorimotor greater amount than controls (11,82,83). They also spent a mechanisms rather than on spatial learning per se. significantly greater proportion of the time swimming in the Use of the visible platform task with muscarinic antago- periphery during the visible platform task. As might be nists has produced mixed results. Whishaw found that high expected, measures of the percent of time swum in the doses of atropine sulfate, a muscarinic antagonist, some- periphery correlated positively with both hidden and visible times increased search time early in testing (118), but that platform search time, and negatively with platform quadrant it generally had little or no effect on performance of the dwell time obtained during the probe trial (11,83). These visible platform task (109,118). However, other researchers results extended the findings of Whishaw and Tomie (118) TESTING THE NMDA, LONG-TERM POTENTIATION, 185 by showing that even when trained in the hidden platform found that most rats quickly jumped off the platform and task, rats given an NMDA or muscarinic antagonist failed to continued swimming (11,83). This behavior seemed to have develop strategies that were appropriate for the task. Some- a strongly ‘‘driven’’ quality; in many cases the rats con- times their failure was especially striking, a point illustrated tinued to make continuous paddling motions with their by rats that persisted in swimming around the perimeter of limbs as they were being picked up and transported to the the pool throughout every training trial, never once encoun- platform and after they were placed onto it. These observa- tering the hidden platform (82). tions seemed consistent with the well-known tendency for The exact cause of this peculiar periphery swimming these treatments to cause behavioral hyperactivity behavior is not known. It is possible that periphery swim- (27,33,107,116,120). When we measured spontaneous ming reflects a specific inability to form a spatial map. activity in rats given NMDA or muscarinic antagonists Alternatively, the fact that NMDA and muscarinic antago- they were markedly hyperactive (11,83). The explanation nists, anxiolytic benzodiazepines (Cain, unpublished obser- seemed to be that the drug-induced hyperactivity caused the vations), hippocampal, thalamic, or neocortical lesions animals to produce continuous swimming movements, (91,114) (Cain and Vanderwolf, unpublished observations), which carried them off the platform and back into the and both hypo- and hyperthermia (75) all increase swim- water. This is an example of how independent measures ming in the periphery could mean that thigmotaxic swim- taken outside the water maze can help clarify observations ming is a general response to the compromised brain made during maze training. function that these treatments cause. A few researchers have noted that rats given an NMDA or If a rat does not encounter the hidden platform on a given muscarinic antagonist displayed unusual and abnormal trial it will not learn anything about the platform’s location behaviors during training in the pool. Some rats had diffi- on that trial, perhaps leading to a biased measure of its culty swimming and interacting adaptively with the hidden spatial learning ability. This is dealt with by placing the platform (111), or fell off the platform after climbing on rat by hand on the platform after any trial in which it failed (63). A striking example was the tendency to swim com- to find the platform during the swim. When we attempted to pletely over the platform and off the other side in one do this with rats given NMDA or muscarinic antagonists we continuous motion as soon as it was encountered (Fig. 2).

FIG. 2. Videotape frames of rats swimming in a pool after being given an NMDA receptor antagonist. The frames proceed from left to right. The top three frames illustrate a deflection; the hidden platform is darkened on the images for clearer visibility. The bottom three frames illustrate a swimover; the hidden platform is indicated by white dots on the images for clearer visibility. 186 CAIN

These ‘‘swimovers’’ had been described but not measured were required for discriminations based on pattern differ- and related to maze acquisition (20,62,63,111). Swimovers ences when luminous flux was held constant. probably resulted from the same hyperactivity that produced These results suggested that NMDA activity was required the jumping-off behavior described above. for optimal performance of the visible platform and pattern We also observed and measured ‘‘deflections’’, which discrimination tasks. The rats’ above-chance visual discri- occurred when rats given an NMDA or muscarinic antago- mination performance under an NMDA antagonist probably nist bumped into the hidden platform, deflected off it, and allowed them to acquire the water maze task effectively continued swimming without attempting to climb on after nonspatial pretraining (see below). The visual discri- (Fig. 2). We found that swimovers and deflections occurred mination deficit might have hampered the maze perfor- far more frequently in the drugged rats than in controls. In mance of naive rats given the same treatment. Similarly, some drugged groups more than half of all contacts with the muscarinic antagonists disrupted visual discrimination hidden platform resulted in a swimover or deflection (7,8,16,25,50,56,77), and in tests conducted in a water (11,83). The incidence of swimovers and deflections corre- maze pool, muscarinic antagonists impaired visual discri- lated negatively with measures of learning in both the mination performance and decreased visual orienting and hidden and visible platform tasks, and positively with search behaviors (31,72,118). Although this review is not other measures of sensorimotor disturbance (11,83). In directly concerned with spatial learning in the radial arm some cases the correlations were very high. For rats given maze, results from visual tests of the kind discussed above an NMDA antagonist the correlation between the percent of would appear to be relevant to this task. The findings contacts with the platform that were deflections or swi- suggest that radial arm maze deficits obtained with movers, and summed platform search time was þ0.99 (83). NMDA or muscarinic antagonists need to be considered in The overall picture that emerged was that rats that had the context of any visual processing deficit the antagonists poor acquisition measures in the maze task exhibited more cause. sensorimotor disturbances, and rats that acquired the task Taken together these studies suggested that the problem well had few if any sensorimotor disturbances. When the of sensorimotor disturbances is an important one in this performance of the same rats was evaluated under the same research. Some researchers have suggested that poor acqui- drug treatment in sensorimotor tasks conducted outside the sition scores in rats given NMDA or muscarinic antagonists pool, rats that exhibited sensorimotor disturbances during could have resulted from the sensorimotor disturbances training in the pool also exhibited sensorimotor disturbances (41,58,72), and have pointed out that animals nevertheless on tasks conducted outside the pool (11,83). This indicated learned a good deal about the task in spite of the drug that the sensorimotor disturbances were not unique to the treatment (41). Certainly the effect of deflections, swimov- training conditions in the pool. ers, and excessive swimming in the periphery would be to The major source of information for the formation of a lengthen search times and reduce the information available spatial map in the water maze task is vision (61,71,96). Both to rats about the location of the hidden platform compared to NMDA and muscarinic antagonists have been reported to rats that swam away from the maze wall and found the disrupt visual discrimination performance (7,8,16,25,50, hidden platform readily, and climbed onto and remained on 56,66,77,98,99), suggesting that these agents might affect the platform whenever it was encountered. Sensorimotor water maze acquisition by producing disturbances in vision. disturbances could be expected to produce poor maze Morris evaluated this possibility for an NMDA antagonist acquisition scores in rats even if the drug treatments had by training rats to discriminate between stable and sinking no direct effect on engram formation. However, this does platforms in the maze pool on the basis of their visual not exclude the possibility that the drug treatments also appearance, but found that an NMDA antagonist had no might have disrupted neural mechanisms of engram forma- effect on the rate of learning this discrimination (63,64). In tion for spatial learning. In other words, the drug treatments an experiment similar to Morris’, NPC17742 did not affect might have had parallel effects on both sensorimotor the rate of learning to discriminate between stable and mechanisms required in the task and on neural mechanisms sinking platforms during a 9-day training phase (13). How- of engram formation. Evidence of sensorimotor distur- ever, this NMDA antagonist reduced the ultimate level of bances due to an experimental treatment can provide choice accuracy in the post-learning phase of testing from important information about treatment effects on neural days 10 through 18, after the learning curves had reached mechanisms for instinctive behavior, but cannot by itself asymptote. The reduction was small, and the animals’ lead to definitive conclusions about drug effects on the choice accuracy remained well above chance ( Ͼ 80% engram. correct), which is consistent with the good performance of How might NMDA or muscarinic antagonists affect rats already familiar with the general strategies required in instinctive behavior? NMDA receptors are widely distribu- the task. In the Morris study rats carried osmotic minipumps ted in the central nervous system, including virtually all of for delivery of the antagonist, which limited testing time to the neocortex and many subcortical sensory and motor areas 10 days. Therefore Morris’ rats had not been tested during including spinal cord (57,59). In animals NMDA antago- the asymptotic phase. nists can cause hyperactivity, stereotypy, ataxia, muscle We also examined the effect of NPC17742 on visual dis- relaxation, catalepsy, and other behavioral abnormalities crimination in a Y-maze and found that the antagonist (11,19,27,33,36,42,83,101,107,111,120). They can also reduced the accuracy of well-established discrimination of disrupt normal behavior-related hippocampal electrical vertical vs. horizontal stripes, but had no effect on a black rhythms in the rat (53), and cause agitation, confusion, vs. white discrimination (13). This suggested that NMDA poor concentration, paranoia, ataxia, delusions and halluci- receptors were not required for simple discriminations nations in humans (30,97). Many sensory and motor based on differences in total luminous flux, but that they mechanisms depend on NMDA receptors for normal TESTING THE NMDA, LONG-TERM POTENTIATION, 187 function (35). NMDA antagonists can markedly alter motor Morris (63) and Whishaw (110) were the first to report system function by altering dopamine release in the basal that prior experience with the general task requirements ganglia and hippocampus (29,119), and can interfere with greatly improved acquisition of the task when rats subse- the processing of visual and somatosensory input (81,88). quently were trained under either NMDA or muscarinic Disruption of these mechanisms by NMDA antagonists antagonism. In the Morris study the prior experience, which could contribute to the sensorimotor disturbances found in was given in the absence of any drug treatment, was termed water maze studies. nonspatial pretraining because black curtains around the Muscarinic receptors also are distributed widely in the pool occluded the room cues, and the hidden platform was central nervous system (26), and muscarinic antagonists moved to a new location after every trial. During the cause a variety of behavioral alterations that can affect a pretraining the rats learned to swim away from the wall rat’s ability to perform the behaviors required in the water and to climb onto and remain on the platform when it was maze task (31,72,102,109,117,120). These findings are encountered (110). consistent with the fact that acetylcholine-sensitive neurons In the Whishaw study (110) rats were first trained in the with muscarinic receptors are frequently found in primary standard hidden platform task in one pool in the absence of sensory areas of neocortex (48,49), and muscarinic antago- drug treatment, then given drug treatment and trained in the nists block a major component of cortical activation (103). hidden platform task in a second pool in another room with a In sum, both NMDA and muscarinic antagonists cause new set of cues. The results were the same in the two similar sensorimotor disturbances in instinctive behaviors studies: rats that were familiar with the general task strate- required for performance of the water maze task. The nature gies had much shorter search latencies when trained under of the disturbances suggests that they might have contribu- NMDA or muscarinic antagonism than naive rats did. ted to the poor acquisition scores reported for animals given Morris (63) measured instances of falling off the platform these agents. However, the existence of sensorimotor dis- and found that falling off the platform was almost elimi- turbances does not exclude possible effects of the antago- nated in the nonspatially pretrained rats, suggesting that the nists on mechanisms of engram formation. pretraining reduced the incidence of sensorimotor distur- bances caused by drug treatment. However, in both studies the experienced rats had slightly but significantly longer MODIFICATION OF INSTINCTIVE BEHAVIOR BY PRETRAINING search times compared to controls. An advantage of the water maze task is its procedural In more detailed dose-response experiments Morris and simplicity and the speed with which normal rats acquire colleagues found that nonspatially pretrained rats could the task. However, as Whishaw and Mittleman (115) have learn the location of the hidden platform when trained emphasized, the water maze task is more complex than it under NMDA antagonism, as indicated by platform quad- appears, and the spatial navigation strategies used by rats are rant dwell data (3,4,20). When we examined the effects of more numerous and complex than had been thought. The rat NMDA or muscarinic antagonists in nonspatially pretrained must learn a number of things before it can deal effectively rats we found that the pretrained rats had few if any with the task, the most important of which are to swim away sensorimotor disturbances, and acquired the task as rapidly from the wall and to climb onto and remain on the platform as controls, as indicated by three different measures of as soon as it is encountered. Only then can the rat effectively acquisition (11–13,82,83). Naive rats given the same drug learn the location of the hidden platform in relation to cues treatments displayed numerous sensorimotor disturbances in the environment. and obtained very poor maze acquisition scores which, as In terms of the earlier discussion, learning these general described above, correlated strongly with the sensorimotor strategies involves changes in instinctive behavior. A naive disturbances. The fact that a small elevation in platform rat normally spends the first few trials swimming close to search time remained in Morris’ nonspatially pretrained rats the pool wall, scratching at it and trying to find a way out. trained under NMDA antagonism might be explained by the This behavior has the obvious goal of getting the rat out of large pool and small hidden platform used in that experi- the water, but it may also reflect the general tendency of rats ment, which could be expected to make the maze task more to locomote thigmotaxically toward the perimeter for pro- difficult than tasks that used a smaller pool and a larger tection whenever they are in a large open space (6). A hidden platform (54). With an especially difficult maze task, requirement for successful acquisition of the water maze is a drug treatment that interfered with sensorimotor mechan- that the rat learns to curtail its instinctive tendency to swim isms could be expected to impair behavioral performance of thigmotaxically, and that it learns to search the open areas of the task to some extent, by increasing thigmotaxic swim- the pool away from the wall. As indicated earlier, rats given ming for example. This could lead to increased search time a variety of drug treatments or brain lesions appear to have without necessarily preventing the rat from learning where great difficulty learning to swim away from the wall, which the platform was located. prevents them from coming into contact with the hidden Supporting this interpretation about task difficulty, work platform. When they do contact the platform, there is a good by Schallert has shown that an easier version of the task that chance they will swim over it or deflect off. prevented the learning of inappropriate strategies was readily Whishaw and Kolb (114) first pretrained decorticate rats acquired by naive rats given either atropine or hippocampal in a visible platform version of the task to reduce their lesions (21,84). Rats were given a treatment and first trained strong tendency to swim thigmotaxically. In subsequent using a very large hidden platform that nearly filled the pool. training in a hidden platform version of the task, the Each day thereafter the hidden platform was made smaller, pretrained rats searched the inner areas of the pool that and it effectively shrunk into one quadrant. Quadrant dwell contained the hidden platform, but could not learn to swim measures showed that the rats had learned the location of the directly to the platform. hidden platform using a spatial learning strategy. 188 CAIN

Taken together, these findings indicate that neither receptor-mediated LTP underlies spatial learning emerged NMDA nor muscarinic activity are required for a rat to be from experiments that showed that the same dose of a able to learn the location of a hidden platform, confirming chronically administered NMDA antagonist that blocked the conclusion initially arrived at by Keith and Rudy (41). LTP in the dentate gyrus also impaired acquisition of the They also suggest that separation of the phase of learning water maze task (63,64). when the rats learn the general strategies required in the task In the earlier experiment (64), which used naive rats, the from the phase when they learn the location of the hidden NMDA antagonist produced a marked elevation of hidden platform is a useful approach to understanding the complex platform search time. In the latter experiment (63), which nature of this apparently simple task. Nonspatial pretraining used nonspatially pretrained rats, the same dose of antago- appears to involve learning that modifies instinctive beha- nist produced a smaller but statistically significant elevation viors in ways that allow rats to perform adaptively in the in hidden platform search time. Subsequent work from the task when challenged with treatments that normally inter- same laboratory indicated that the dose used in these fere with those instinctive behaviors in naive rats. One experiments was higher than required to block LTP, and indication of the learning that rats undergo during nonspatial that if a smaller dose was used LTP continued to be blocked, pretraining is the decrease in platform search time by but nonspatially pretrained rats nevertheless learned the approximately two-thirds from day 1 to day 4 of nonspatial location of the hidden platform (4,20). pretraining (11,20,63,83). We evaluated the role of NMDA receptor-mediated LTP Although the exact neural basis for the striking effects in spatial learning by using two different competitive that result from pretraining is not known, the effects of NMDA antagonists (NPC17742 and CGS19755) in differ- pretraining have been known since at least 1959, when ent groups of rats, each in a dose sufficient to completely Herz showed that prior experience with a pole climbing block LTP in the dentate gyrus. The same rats tested for task eliminated the debilitating effects of muscarinic LTP were then nonspatially pretrained and subsequently antagonists on climbing behavior (39). Similarly, Steinberg trained in the water maze task after being given the same showed that prior experience with the test apparatus elimi- dose of NMDA antagonist that had been found to block nated the effects of combined administration of ampheta- LTP. As expected from earlier work (11,83), the animals mine and barbiturate (90). There have been other similar acquired the maze task as rapidly as controls, as indicated by demonstrations of the protective effects of prior experience three different measures of maze acquisition (11,12,82). on instinctive behaviors that are normally disrupted by This is consistent with the recent finding that fyn gene scopolamine (22) or brain lesions (86). knockout mice, in which hippocampal LTP is blunted or It appears that the pretraining experience must be with the absent, can nevertheless acquire the water maze task (40). specific apparatus that the animals subsequently will be As originally proposed, the NMDA/LTP hypothesis tested with under drug treatment. Nonspatial pretraining in focused on LTP induced in the dentate gyrus of the hippo- the maze pool had no effect on the motor ataxia that was campus. The question arises whether similar findings would caused by NMDA or muscarinic antagonists, as measured be obtained with LTP induced elsewhere in the hippocam- by the rat’s ability to walk on a narrow wooden beam pus, or outside the hippocampus. Other synapses in the hip- (11,83). Similarly, rats that were nonspatially pretrained in pocampus and in the visual neocortex slice preparation also a large circular pool were markedly impaired in a visible support NMDA receptor-mediated LTP (2,34), and it is platform task in a small rectangular glass aquarium when possible that LTP at these sites is important for spatial given scopolamine (Caldji and Vanderwolf, unpublished learning. The concentration of AP5 required to block LTP in observations). The poor performance of nonspatially pre- the CA1 region of the hippocampus (34) and in the visual trained rats under these conditions confirms that the drugs neocortex slice (2) might be less than that required to block were having the expected effects on the nervous system and LTP in the dentate gyrus (64). This suggests that a systemi- on behaviors tested in tasks other than those on which the cally administered NMDA antagonist that blocks LTP in the rats were pretrained. dentate gyrus might also block LTP in CA1 and the visual It is not known which components of the nonspatial pre- neocortex. training experience are important for the pretraining effect Taken together, these results suggest that NMDA recep- in the water maze task. An ongoing experiment designed to tor-mediated LTP in the hippocampus and visual neocortex fractionate the components of the nonspatial pretraining might not be required for spatial learning in the water maze, experience by giving different groups of rats only a part but they do not exclude the possibility that this form of LTP of the total nonspatial pretraining experience (swimming contributes in a non-essential way to spatial learning. If this in the pool with no hidden platform present; climbing is true, there must be an alternative mechanism for estab- from the water onto the platform; being allowed to sit on lishing the engram for the location of the hidden platform. the platform; etc.) suggests that the different parts of the Recently the suggestion was made that an alternative nonspatial pretraining experience are equally beneficial, navigation mechanism based on inertial (dead reckoning) but that no single part of the experience is as beneficial as navigation could help account for spatial learning, and the whole nonspatial pretraining experience (Hoh and Cain, that the recalibration of such a mechanism in the second unpublished observations). room of a two-room water maze task might not require hippocampal LTP (5). The suggestion was based on pre- liminary data indicating that such a mechanism might be NMDA RECEPTOR-MEDIATED LTP AND SPATIAL LEARNING available to rats in the water maze task (108). If true, this A number of the experiments discussed above could explain why NMDA antagonists do not prevent rapid incorporated tests of the NMDA/LTP hypothesis of spatial learning of hidden platform location in the second room (3). learning into the design. The hypothesis that NMDA In a test of this possibility we trained rats in a stable hidden TESTING THE NMDA, LONG-TERM POTENTIATION, 189 platform task, with black curtains around the pool to fact that combined muscarinic and serotonergic antagonism eliminate visual cues. Different groups searched for a plat- often produced a much greater behavioral deficit than form made from acrylic plastic and weighted by a brick, or a antagonism of either system alone suggested that the loss similar platform containing a strong magnet. Both groups of one system was compensated for, at least in part, by the were trained over many days, but neither group gave any other (106). evidence of preferential swimming in the platform quadrant However, to date no water maze study has examined the during probe trials, and platform quadrant dwell time was at effect of combined antagonism of two receptor types in chance (Cain, Beiko and Boon, unpublished observations). nonspatially pretrained rats. In doing these experiments, When the curtains were removed and the rats given further we found that nonspatial pretraining did not protect rats training they all achieved short search times, and signifi- from the effects of combined muscarinic and serotonergic cantly increased their dwell time in the platform quadrant antagonism in the water maze task, in spite of the fact that during the probe trial, indicating that they were capable of nonspatially pretrained rats acquired the task readily if acquiring the task when allowed to use visual cues. This either antagonist was given by itself (Beiko and Cain, suggested that when visual cues were unavailable in the unpublished observations). The same results were obtained water maze task the rats could not make use of inertial with combined antagonism of muscarinic and NMDA navigation or magnetic field information to solve it. receptors (Cain, unpublished observations). In contrast to To date, the focus of research on LTP and its relation to the studies described in the previous paragraph, selective learning has been on the hippocampus. Karl Lashley, speak- damage of both hippocampal cholinergic and hippocampal ing of the role of the primary visual neocortex and learning, serotonergic afferents produced only a mild impairment in concluded that engrams for sensory discrimination tasks the water maze task in naive rats (67). Taken together, the were established in the sensory neocortex, and suggested data suggest that there are multiple systems for production ‘‘that the same cells which bear the memory traces are of the instinctive behaviors required in the water maze task, also excited and play a part in every other visual reaction and that a major part of these systems resides outside the of the animal’’ (52). The possibility that engram formation hippocampus. Further, it appears that hippocampal damage for other forms of learning could occur in the cortex has does not prevent the formation of spatial maps, but instead been widely discussed (89), but to date there have been no interferes with the use of that information to guide locomo- studies of the possible relation between neocortical LTP and tion through the environment (24,84,91,94,112,113). learning. There is considerable evidence that damage to a The question of whether a specific type of synapse is a variety of cortical areas produces acquisition deficits in the site of engram formation for spatial learning is difficult to water maze that are comparable to those produced by answer definitively. In light of the emerging evidence for the hippocampal damage (23,43–46,85,92,93,95,114), and involvement of widespread circuitry involving a number of that some learning deficits formerly attributed exclusively different receptor types in the various components of water to the hippocampus are now known to result from damage to maze acquisition, this becomes an important question. How- cortex (28,55). It seems that it would be fruitful to explore ever, the very requirement that both pre-existing neural cir- the possibility that neocortical LTP might be related to cuits for instinctive behavior, as well as an engram for the spatial learning. specific location of the hidden platform, are required for LTP has been observed in the neocortical slice prepara- acquisition of the water maze makes it impossible to iden- tion (2), but its induction is more difficult than in the tify the specific sites of change that constitute the engram hippocampus, and in an extensive series of experiments in from experiments with neurotransmitter antagonists in awake, behaving rats it was not possible to induce LTP in intact, behaving animals. To illustrate the nature of the pro- the neocortex (74). More recently it has proven possible to blem, the following possibilities are consistent with what is induce LTP in the neocortex, but this required many known about the role of muscarinic and NMDA receptors in stimulations each day for up to 15 days (73). Thus LTP the water maze task. induction in the neocortex of awake, behaving rats requires (1) NMDA and muscarinic receptors function in neural much more stimulation than in the hippocampus. The circuits involved in the processing of visual and other sen- neuropharmacological mechanism of this form of neocor- sory information important for the task. Antagonism of tical LTP is not known. either receptor population alone causes a partial sensory impairment, but enough sensory processing capability remains to allow nonspatially pretrained rats to acquire SEARCHING FOR THE ENGRAM the maze task using a spatial learning strategy (Fig. 1). Study of the involvement of muscarinic and serotonergic When both receptor populations are antagonized, a greater mechanisms in behavior has shown that the effect of com- sensory impairment likely is produced. This leaves insuffi- bined antagonism of both receptor types was generally cient sensory processing capability to allow use of a spatial much greater than the effect of antagonism of either system learning strategy, and the rats fall back on a taxon strategy, alone, even if antagonism of one system produced no deficit such as swimming away from the wall at a distance that (68,76,78,102,106). Combined antagonism of muscarinic would allow interception with the hidden platform. This and serotonergic mechanisms produced a severe deficit that probably requires less sensory processing because it may be analogous to ‘‘global dementia’’ in the rat, which depends on larger, closer cues (e.g., the maze wall). probably resulted from removal of the two major systems (2) NMDA and muscarinic receptors, together with involved in activation of the cerebral cortex (102). These remaining brain mechanisms, are directly involved in learn- results were consistent with the idea that the cerebral cortex ing the general behavioral strategies required in the maze is involved in the organization of the instinctive behaviors task; that is, they are involved in the learning that occurs required in tasks such as the water maze (52,102,105). The during nonspatial pretraining. Both receptor populations are 190 CAIN

associated with other neurons whose synapses are modified (Fig. 3C–F). (3) NMDA or muscarinic receptors are directly involved, together with remaining brain mechanisms, in learning the exact spatial location of the hidden platform. For rats that are already familiar with the general behavioral strategies required in the maze task, only one of these receptor popu- lations is required. NMDA or muscarinic synapses might be modified during this learning process (Fig. 3A,B), or they might be unmodified but associated with other neurons whose synapses are modified (Fig. 3C–F). The available data do not allow a decision about which of these possibilities holds, or which type of neural circuit or synapse undergoes modification during training on the task. There is no reason to believe that these possibilities are exhaustive or mutually exclusive, and it seems likely that all three possibilities might hold. This would not be surpris- ing, given the widespread anatomical distribution of NMDA and muscarinic receptors and the profound disruptions that NMDA and muscarinic antagonists cause in hippocampal and neocortical electrographic activity and instinctive behavior (11,53,83,102,103). Neurotransmitter antagonist or lesion effects by them- selves do not allow conclusions about which specific synapses form the engram for a particular form of learning. Evidence from neuroanatomy or functional electro- physiology, and preferably both, is needed in addition. For example, a series of experiments that obtained behavioral FIG. 3. Hypothetical neural circuits that might be involved in learning the evidence from use of a specific neurotransmitter antagonist general behavioral strategies required in the water maze task or the spatial or brain lesion, together with neuroanatomical evidence location of the hidden platform. Circles represent cell bodies and dendrites; of plastic change in the brain structure of interest, and lines represent axons. Filled terminals and apposed darkened regions on the electrophysiological evidence of a functional change in cell body represent synapses that are modifiable in the course of learning (plastic). Unfilled terminals represent synapses that are effective but unmo- the structure, could be highly informative. The choice of difiable. Ach, acetylcholine-releasing neuron; Glu, glutamate-releasing behavioral task is also important. The specific behavioral neuron; musc, muscarinic receptors; NMDA, N-methyl-d-aspartate recep- strategies required in the task should be well understood, tors; dep, depolarizing (excitatory) synapse; unknown, synapse for which and ideally the task should be amenable to fractionation by the neurotransmitter and receptor type are unknown; T, target neuron. The hypothetical circuits are conceptual representations only, and do not neces- pretraining or some similar approach that would allow a sarily represent synapses proposed to be required for learning. However, clearer understanding of the neural mechanisms of the (B) is similar to the NMDA coincidence detector that is hypothesized to various components of task performance. underlie some forms of learning (9,18). The circuits are intended to provide a simplified representation of which synapses might undergo plastic modification to support learning. Activation of muscarinic and NMDA receptors by acetylcholine or glutamate, respectively, often has excitatory CONCLUSIONS effects (35,48,49). (A) Circuit in which a synapse containing muscarinic receptors undergoes modification during behavioral training. (B) Circuit in The ability of an animal to acquire the water maze task which a synapse containing NMDA receptors undergoes modification depends on its being able to apply instinctive behaviors to during behavioral training. Should this occur when another synapse provides coincident depolarization (‘‘dep’’), this would resemble a coin- performance of the task in an adaptive manner, as well as its cidence detector (18). (C,D) Circuits in which a synapse with unknown ability to learn the specific information required by the neurotransmitter and receptors undergoes modification during behavioral particular task conditions. The instinctive behaviors training. In A–D scopolamine or NPC, as appropriate, will block transmis- undergo modification as the rat learns the general strategies sion, producing behavioral impairments. In (E) and (F) muscarinic or required in the task. Thus, learning the water maze task NMDA receptors do not form modifiable synapses themselves but do play an essential role in a Hebb-synapse learning mechanism (37). depends on activity in pre-existing neural circuits that gen- Simultaneous activation of neurons 1 and 2 leads to increased transmission erate behaviors appropriate for the task, as well as plastic at the modifiable synapse as a result of coincident firing of neuron 1 and the changes in those neural circuits, and the storage of new target neuron. In (E) and (F) scopolamine or NPC, respectively, will information about the spatial location of the hidden plat- interfere with transmission between neuron 2 and the target neuron, thus preventing the neuroplastic modifications that underlie learning. Adapted form. The sites of engram formation for learning the from Fig. 8 in Ref. (102). location of a hidden platform in the water maze are not known. Recent experiments have shown that neither NMDA nor muscarinic receptors are required, although required for rapid learning of the strategies, as antagonism either may normally contribute to this form of learning. of either one alone prevents naive rats from learning the The available evidence suggests that at least some of the strategies within a single session (11,83). NMDA or plastic changes involved in learning the general strategies muscarinic synapses might be modified during this learning required in the task occur in existing neural circuits situated process (Fig. 3A,B), or they might be unmodified but in widespread areas of the brain, perhaps including sensory TESTING THE NMDA, LONG-TERM POTENTIATION, 191 and motor structures in the neocortex and elsewhere, mak- ACKNOWLEDGEMENTS ing them difficult to distinguish from existing sensorimotor Supported by a grant from NSERC. Those familiar with mechanisms. The findings indicate the difficulty of inferring Case Vanderwolf’s career will recognize his many contri- the occurrence or non-occurrence of learning from behavior, butions to the content of this essay. It is a pleasure to and the difficulty of causally linking the action of particular acknowledge these contributions and to have attended the receptor populations with the formation of specific festschrift marking his 60th year. memories.

REFERENCES 1. Alcock, J., Animal Behavior. Sunderland, MA: Sinauer; 1989. place learning in the Morris water task: Spatial mapping deficit or 2. Artola, A. and Singer, W., Long-term potentiation and NMDA inability to inhibit non-place strategies? Behav. Neurosci. in press. receptors in rat visual cortex. Nature, 1987, 330, 649–652. 22. DeVietti, T. L., Pellis, S. M., Pellis, V. C. and Teitelbaum, P., 3. Bannerman, D. M., Good, M. A., Butcher, S. P., Ramsay, M. and Previous experience disrupts atropine-induced stereotyped ‘‘trap- Morris, R. G.M., Distinct components of spatial learning revealed by ping’’ in rats. Behav. Neurosci., 1985, 99, 1128–1141. prior training and NMDA receptor blockade. Nature, 1995, 378, 23. DiMattia, B. D. and Kesner, R. P., Spatial cognitive maps: Differ- 182–186. ential role of parietal cortex and hippocampal formation. Behav. 4. Bannerman, D. M., Good, M. A., Ramsay, M. and Morris, R. G. M., Neurosci., 1988, 102, 471–480. NMDA receptors, long-term potentiation and spatial learning. Brain 24. Eichenbaum, H., Stewart, C. and Morris, R. G. M., Hippocampal Res. Assoc. Abstr., 1995, 12, 107. representation in place learning. J. Neurosci., 1990, 10, 2531–2542. 5. Barnes, C. A., Involvement of LTP in memory: Are we ‘‘searching 25. Evans, H. L., Scopolamine effects on visual discrimination. Mod- under the street light’’?. Neuron, 1995, 15, 751–754. ifications related to stimulus control. J. Pharm. Exp. Therap., 1975, 6. Barnett, S. A., A study in behaviour. London: Methuen 1963. 195, 105–113. 7. Bartus, R. T. and Johnson, H. R., Short-term memory in the rhesus 26. Fibiger, H. C., The organization and some projections of cholinergic monkey: Disruption from the anti-cholinergic scopolamine. Pharm. neurons of the mammalian forebrain. Brain Res. Rev., 1982, 4, 327– Biochem. Behav., 1976, 5, 39–46. 388. 8. Bennett, T. L., Nunn, P. J. and Inman, D. P., Effects of scopolamine 27. Ford, L. M., Norman, A. B. and Sanberg, P. R., The topography of on hippocampal theta and correlated discrimination performance. MK-801-induced locomotor patterns in rats. Physiol. Behav., 1989, Physiol. Behav., 1971, 7, 451–454. 46, 755–758. 9. Bliss, T. V. P. and Collingridge, G., A synaptic model of memory: 28. Gaffan, D., Dissociated effects of perirhinal cortex ablation, fornix Long-term potentiation in the hippocampus. Nature, 1993, 361,31– transection and amydalectomy: Evidence for multiple memory sys- 39. tems in the primate temporal lobe. Exp. Brain Res., 1994, 99, 411– 10. Bliss, T. V. P. and Lomo, T., Long-lasting potentiation of 422. synaptic transmission in the dentate area of the anesthetized rabbit 29. Gauchy, C., Desban, M., Glowinski, J. and Kemel, M. L., NMDA following stimulation of the perforant path. J. Physiol., 1973, 232, regulation of dopamine release from proximal and distal dendrites in 331–356. the cat substantia nigra. Brain Res., 1994, 635, 249–256. 11. Cain, D. P., Saucier, D., Hall, J., Hargreaves, E. L. and Boon, F., 30. Grotta, J., Safety and tolerability of the glutamat CGS1975 antagonist Detailed behavioral analysis of water maze acquisition under APV or 5 in acute stroke patients. Stroke, 1994, 25, 255. CNQX: Contribution of sensorimotor disturbances to drug-induced 31. Hagan, J. J., Tweedie, F. and Morris, R. G. M., Lack of task acquisition deficits. Behav. Neurosci., 1996, 110, 86–102. specificity and absence of posttraining effects of atropine on learning. 12. Cain, D. P.; Saucier, D.; Boon, F., Testing hypotheses of spatial Behav. Neurosci., 1986, 100, 483–493. learning: The role of NMDA receptors and NMDA-mediated long 32. Hailman, J. P., The ontogeny of an instinct. Behav. Suppl., 1967, 15, term potentiation. Behav. Brain Res., 1997, 84, 179–193. 1–159. 13. Cain, D. P.; Saucier, D.; Boon, F., Testing hypotheses of spatial 33. Hargreaves, E. L. and Cain, D. P., Hyperactivity, hyper-reactivity, learning: The role of NMDA receptors and NMDA receptor- and sensorimotor deficits induced by low doses of the N-methyl-D- mediated long term potentiation. ibid. aspartate non competitive channel blocker MK-801. Behav. Brain. 14. Carew, T. J., Pinsker, H. M. and Kandel, E. R., Long term habituation Res., 1992, 47, 23–33. of a defensive reflex in Aplysia. Science, 1972, 175, 451–454. 34. Harris, E. W., Gonang, A. H. and Cotman, C. W., Long-term 15. Castro, C. A., Silbert, L. H., McNaughton, B. L. and Barnes, C. A., potentiation in the hippocampus involves activation of N-methyl- Recovery of spatial learning deficits after decay of electrically D-aspartate receptors. Brain Res., 1984, 323, 132–137. induced synaptic enhancement in the hippocampus. Nature, 1989, 35. Headley, P. M. and Grillner, S., Excitatory amino acids and synaptic 342, 545–548. transmission: The evidence for a physiological function. Trends 16. Clissold, D. B., Ferkany, J. W. and Pontecorvo, M. J., Competitive Pharmacol. Sci., 1990, 11, 205–211. and noncompetitive N-methyl-D-aspartate (NMDA) antagonists, 36. Heale, V. and Harley, C., MK-801 and AP5 impair acquisition, but haloperidol, and scopolamine impair performance in a nonspatial not retention, of the Morris milk maze. Pharmacol. Biochem. Behav., operant discrimination task. Psychobiology, 1991, 19, 332–338. 1990, 36, 145–149. 17. Collingridge, G. L., Kehl, S. J. and McLennan, H., Excitatory amino 37. Hebb, D. O., Organization of Behavior. New York: Wiley 1949. acids in synaptic transmission in the Schaffer-commissural pathway 38. Held, R. and Freedman, S. J., Plasticity in sensorimotor control. of the rat hippocampus. J. Physiol., 1993, 334, 33–46. Science, 1963, 142, 455–462. 18. Collingridge, G. L. and Bliss, T. V. P., NMDA receptors-their role in 39. Herz, A., Uber die Wirkung von scopolamin, benactyzin, und atropin long-term potentiation. Trends Neurosci., 1987, 10, 288–293. auf das verhalten der ratte. Naunyn-Schmiedeberg’s Archiv. Exp. 19. Compton, R. P., Contreras, P. C., O’Donohue, T. L. and Monahan, J. Path. Pharmakol., 1959, 236, 110–111. B., The N-methyl-D-aspartate antagonist, 2-amino-7-phosphonohep- 40. Huerta, P. T.; Scearce, K. A.; Farris, S. M.; Empson, R. M.; Prusky, tanoate, produces phencyclidine-like effects in rats. Eur, J. Pharma- G. T., Preservation of spatial learning in fyn tyrosine kinase knockout col., 1987, 136, 133–134. mice. Neuroreport in press. 20. Davis, S., Butcher, S. P. and Morris, R. G. M., The NMDA receptor 41. Keith, J. and Rudy, J., Why NMDA-receptor-dependent long-term antagonist D-2-amino-5-phosphonopentanoate (D-AP5) impairs potentiation may not be a mechanism of learning and memory: spatial learning and LTP in vivo at intracerebral concentrations Reappraisal of the NMDA-receptor blockade strategy. Psychobiol- comparable to those that blocks LTP in vitro. J. Neurosci., 1992, ogy, 1990, 18, 251–257. 12, 21–34. 42. Koek, W., Woods, J. H. and Winger, G. D., MK-801, a proposed 21. Day, L. B.; Schallert, T., Anticholinergic effects on acquisition of noncompetitive antagonist of excitatory amino acid neurotransmis- 192 CAIN

sion, produces phencyclidine-like behavioral effects in pigeons, rats and mnemonic effects of combined lesions of the medial septal and and rhesus monkeys. J. Pharm. Exp. Therap., 1988, 245, 969–976. serotonin afferents to the hippocampus. Brain Res., 1994, 651,16– 43. Kolb, B., Pittman, K., Sutherland, R. J. and Whishaw, I. Q., 26. Dissociation of the contributions of the prefrontal cortex and dor- 68. Nilsson, O. G., Strecker, R. E., Daszuta, A. and Bjorklund, A., somedial thalamic nucleus to spatially guided behavior in the rat. Combined cholinergic and serotonergic denervation of the forebrain Behav. Brain Res., 1982, 6, 365–378. produces severe deficits in a spatial learning task in the rat. Brain 44. Kolb, B., Sutherland, R. J. and Whishaw, I. Q., A comparison of the Res., 1988, 453, 235–246. contributions of the frontal and parietal association cortex to spatial 69. O’Keefe, J., Nadel, L., Keightley, S. and Kill, D., Fornix lesions localization in rats. Behav. Neurosci., 1983, 97, 13–27. selectively abolich place learning in the rat. Exp. Neurol., 1975, 48, 45. Kolb, B., Buhrmann, K., McDonald, R. and Sutherland, R. J., 152–166. Dissociation of the medial prefrontal, posterior parietal, and posterior 70. O’Keefe, J. and Dostrovsky, J., The hippocampus as a spatial map. temporal cortex for spatial navigation and recognition in the rat. Preliminary evidence from unit activity in the freely moving rat. Cereb. Cortex, 1994, 6, 664–680. Brain Res., 1971, 34, 171–175. 46. Kolb, B. and Whishaw, I. Q., Earlier is not always better: Behavioral 71. O’Keefe, J.; Nadel, L., The hippocampus as a cognitive map. Oxford: dysfunction and abnormal cerebral morphogenesis following neona- Clarendon; 1978. tal cortical lesions in the rat. Behav. Brain Res., 1985, 17, 25–43. 72. Paylor, R. and Rudy, J., Cholinergic receptor blockade can impair the 47. Krechevsky, I., A study of the continuity of the problem solving rat’s performance on both the place learning and cued versions of the process. Psychol. Rev., 1938, 45, 107–133. Morris water task: The role of age and pool wall brightness. Behav. 48. Krnjevic, K. and Phillis, J. W., Acetylcholine-sensitive cells in the Brain Res., 1990, 36, 79–90. cerebral cortex. J. Physiol., 1963, 166, 296–327. 73. Racine, R. J., Chapman, C. A., Trepel, C., Teskey, G. C. and 49. Krnjevic, K. and Phillis, J. W., Pharmacological properties of Milgram, N. W., Post-activation potentiation in the neocortex: IV. acetylcholine-sensitive cells in the cerebral cortex. J. Physiol., Multiple sessions required for induction of long-term potentiation in 1963, 166, 328–350. the chronic preparation. Brain Res., 1995, 702, 87–93. 50. Ksir, C., Scopolamine and amphetamine effects on discrimination: 74. Racine, R. J., Teskey, G. C., Wilson, D., Seidlitz, E. and Milgram, N. Interaction with stimulus control. Psychopharmacology, 1975, 43, W., Post-activation potentiation and depression in the neocortex of 37–41. the rat: II. Chronic preparations. Brain Res., 1994, 637, 83–96. 51. Lashley, K. S., Brain Mechanisms and Intelligence: A Quantitative 75. Rauch, T. M., Welch, D. I. and Gallego, L., Hypothermia impairs Study of Injuries to the Brain. Chicago: U. Chicago Press. 1929. performance in the Morris water maze. Physiol. Behav., 1989, 45, 52. Lashley, K. S., In search of the engram. Symp. Soc. Exp. Biol. no. 4, 315–320. Camb U. Press 1950, pp. 454-482. 76. Richter-Levin, G. and Segal, M., Spatial performance is severely 53. Leung, L. S. and Desborough, K. A., APV, an N-methyl-D-aspartate impaired in rats with combined reduction of serotonergic and receptor antagonist, blocks the hippocampal theta rhythm in behav- cholinergic transmission. Brain Res., 1989, 477, 404–407. ing rats. Brain Res., 1988, 463, 148–152. 77. Rick, J. T., Whittle, K. L. and Cross, S. H., Disruption and facilitation 54. Mactutus, C. F. and Booze, R. M., Accuracy of spatial navigation: of cue discrimination in the rat by cholinergic agents. Neuro- The role of platform and tank size. Soc. Neurosci. Abstr., 1994, 20, pharmacology, 1981, 20, 747–752. 1014. 78. Riekkinen, P. Jr., Sirvio, J. and Riekkinen, P., Interaction between 55. Meunier, M., Bachevalier, J., Mishkin, M. and Murray, E. A., Effects raphe dorsalis and nucleus basalis magnocellularis in spatial learning. on visual recognition of combined and separate ablations of the Brain Res., 1990, 527, 342–345. entorhinal and perirhinal cortex in rhesus monkeys. J. Neurosci., 79. Robinson, G. S. Jr, Crooks, G. B. Jr, Shinkman, P. G. and Gallagher, 1993, 13, 5418–5432. M., Behavioral effects of MK-801 mimic deficits associated with 56. Milar, K. S., Cholinergic drug effects on visual discrimination: A hippocampal damage. Psychobiology, 1989, 17, 156–164. signal detection analysis. Psychopharmacology, 1981, 74, 383–388. 80. Rowley, I. and Chapman, G., Cross-fostering, imprinting, and learn- 57. Monaghan, D. T., Anatomical organization and apparent heteroge- ing in two sympatric species of cockatoos. Behavior, 1986, 96, 1–16. neity of NMDA receptors. In Kozikowski, A. P., ed. Neurobiology of 81. Salt, T. E., Mediation of thalamic sensory input by both NMDA the NMDA Receptor. New York: VCH Publishers, 1991, pp. 149- receptors and non-NMDA receptors. Nature, 1986, 322, 263–265. 171. 82. Saucier, D. and Cain, D. P., Spatial learning without NMDA 58. Mondadori, C., Weiskrantz, L., Buerki, H., Petschke, F. and Fagg, G. receptor-dependent long-term potentiation. Nature, 1995, 378, E., NMDA receptor antagonists can enhance or impair learning 186–189. performance in animals. Exp. Brain Res., 1989, 75, 449–456. 83. Saucier, D., Hargreaves, E. L., Boon, F., Vanderwolf, C. H. and Cain, 59. Monyer, H., Burnashev, N., Laurie, D., Sakmann, B. and Seeberg, P., D. P., Detailed behavioral analysis of water maze acquisition under Developmental and regional expression in the rat brain and functional systemic NMDA or muscarinic antagonism: Nonspatial pretraining properties of four NMDA receptors. Neuron,1994,12, 529–540. eliminates spatial learning deficits. Behav. Neurosci., 1996, 110, 60. Morris, R. G.M., Spatial localization does not require the presence of 103–116. local cues. Learn. Motiv., 1981, 12, 239–260. 84. Schallert, T.; Day, L.; Weisend, M.; Sutherland, R. J., Spatial 61. Morris, R. G. M., Developments of a water-maze procedure for learning by hippocampal rats in the Morris water task. Soc. Neurosci. studying spatial learning in the rat. J. Neurosci. Meth., 1984, 11,47– Abstr. in press. 60. 85. Schenk, F. and Morris, R. G. M., Dissociation between components 62. Morris, R. G. M., Elements of a hypopthesis concerning the partici- of spatial memory in rats after recovery from the effects of retro- pation of hippocampal NMDA receptors in learning. In Lodge, D., hippocampal lesions. Exp. Brain Res., 1985, 58, 11–28. ed. Excitatroy Amino Acids in Health and Disease. New York: Wiley 86. Schulkin, J., The effects of preoperative ingestive events on feeding 1988, pp. 297-320. and drinking behavior following brain damage. Psychobiology, 1988, 63. Morris, R. G. M., Synaptic plasticity and learning: Selective impair- 16, 185–195. ment of learning in rats and blockade of long-term potentiation in 87. Scoville, W. B. and Milner, B., Loss of recent memory after bilateral vivo by the N-methyl-D-aspartate receptor antagonist AP5. J. Neu- hippocampal lesions. J. Neurol. Neurosurg. Psychiat., 1957, 20, 11–21. rosci., 1989, 9, 3040–3057. 88. Sillito, A., Murphy, P., Salt, T. and Moody, C., The dependence of 64. Morris, R. G. M., Anderson, E., Lynch, G. S. and Baudry, M., retino-geniculate transmission in the cat on NMDA receptors. J. Selective impairment of learning and blockade of long-term poten- Neurophysiol., 1990, 63, 347–355. tiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature, 89. Squire, L. R. and Zola-Morgan, S., The medial temporal lobe 1986, 319, 774–776. memory system. Science, 1991, 253, 1380–1386. 65. Morris, R. G. M., Garrud, P., Rawlins, J. and O’Keefe, J., Place 90. Steinberg, H., Rushton, R. and Tinson, C., Modification of the effects navigation impaired in rats with hippocampal lesions. Nature, 1982, of an amphetamine-barbiturate mixture by the past experience of rats. 297, 681–683. Nature, 1961, 192, 533–535. 66. Murray, T. K., Ridley, R. M., Snape, M. F. and Cross, A. J., The 91. Sutherland, R. J., The navigating hippocampus: An individual effect of dizocilpine (MK-801) on spatial and visual discrimination medley of movement, space, and memory. In Buzsaki, G.; Vander- tasks in the rat. Behav. Pharmacol., 1995, 6, 540–549. wolf, C. H., eds. Electrical Activity of the Archicortex. Budapest: 67. Murtha, S. J. E. and Pappas, B. A., Neurochemical, histopathological Akademia Kiado 1985, pp. 255-279. TESTING THE NMDA, LONG-TERM POTENTIATION, 193

92. Sutherland, R. J.; Hoesing, J. M., Posterior cingulate cortex and 107. Venable, N. and Kelly, P. H., Effects of NMDA receptor antagonists spatial memory: A microlimnology analysis. In Vogt, B. A.; Gabriel, on passive avoidance learning and retrieval in rats and mice. M., eds. Neurobiology of Cingulate Cortex and Limbic Thalamus. Psychopharmacology, 1990, 100, 215–221. Boston: Birkhauser 1993, pp. 461-477. 108. Weisend, M. P., Klein, R. L., Hoesing, J. M., Astur, R. S., Koerner, 93. Sutherland, R. J., Kolb, B. and Whishaw, I. Q., Spatial mapping: A., McDonald, R. J., Geving, T., Peinado, J., McWhorter, J., Definitive disruption by hippocampal or medial frontal cortical Weems, M., Schlegelmilch, J., Yeo, R. and Sutherland, R. J., damage in the rat. Neurosci. Lett., 1982, 31, 271–276. Morris water task: Which cues define locations?. Soc. Neurosci., 94. Sutherland, R. J., Martin, G. M., Edwards, C. and Williams, K., Rats 1995, 21, 1939. with hippocampal damage learn spatial relationships which they 109. Whishaw, I. Q., Cholinergic receptor blockade in the rat impairs cannot use to guide navigation. Soc. Neurosci. Abstr., 1995, 20, 1014. locale but not taxon strategies for place navigation in a swimming 95. Sutherland, R. J., Whishaw, I. Q. and Kolb, B., Contributions of pool. Behav. Neurosci., 1985, 99, 979–1005. cingulate cortex to two forms of spatial learning and memory. J. 110. Whishaw, I. Q., Dissociating performance and learning deficits on Neurosci., 1988, 8, 1863–1872. spatial navigation tasks in rats subjected to cholinergic muscarinic 96. Sutherland, R. J., Whishaw, I. Q. and Regehr, J. C., Cholinergic blockade. Brain Res. Bull., 1989, 23, 347–358. receptor blockade impairs spatial localization by use of distal cues in 111. Whishaw, I. Q. and Auer, R. N., Immediate and long-lasting effects the rat. J. Comp. Physiol. Psychol., 1982, 96, 563–573. of MK-801 on motor activity, spatial navigation in a swimming pool 97. Sveinbjornsdottir, S., Sander, J., Upton, D., Thompson, P., Patsalos, and EEG in the rat. Psychopharmacology, 1989, 98, 500–507. P., Hirt, D., Embre, M., Lowe, D. and Duncan, J., The excitatory 112. Whishaw, I. Q., Cassel, J.-C. and Jarrard, L. E., Rats with fimbria- amino acid antagonist D-CPP-ene (SDZ EAA-494) in patients with fornix lesions display a place response in a swimming pool: A epilepsy. Epilepsy Res., 1993, 16, 165–174. dissociation between getting there and knowing where. J. Neurosci., 98. Tan, S., Kirk, R. C., Abraham, W. C. and McNaughton, N., Effects of 1995, 15, 5779–5788. the NMDA antagonists CPP and MK-801 on delayed conditional 113. Whishaw, I. Q., Cassel, J.-C., Majchrzak, M., Cassel, S. and Will, B., discrimination. Psychopharmacology, 1989, 98, 556–560. ‘‘Short-stops’’ in rats with fimbria-fornix lesions: Evidence for 99. Tang, A. H. and Ho, P. M., Both competitive and non-competitive change in the mobility gradient. Hippocampus, 1994, 4, 577–582. antagonists of N-methyl-D-aspartic acid disrupt brightness discrimi- 114. Whishaw, I. Q. and Kolb, B., Decortication abolishes place but not nation in rats. Eur. J. Pharmacol., 1988, 151, 143–146. cue learning in rats. Behav. Brain Res., 1984, 11, 123–134. 100. Tinbergen, N., The Study of Instinct. New York: Oxford U. Press 115. Whishaw, I. Q. and Mittleman, G., Visits to starts, routes and places 1951. by rats (Rattus norvegicus) in swimming pool navigation tasks. J. 101. Trickelbank, M. D., Singh, L., Oles, R. J., Preston, C. and Iversen, S. Comp. Psychol., 1986, 100, 422–431. D., The behavioural effects of MK-801: A comparison with antago- 116. Whishaw, I. Q., O’Connor, W. T. and Dunnett, S. B., Disruption of nists acting non-competitively and competitively at the NMDA central cholinergic systems in the rat by basal forebrain lesions or receptor. Eur. J. Pharmacol., 1989, 167, 127–135. atropine: Effects on feeding, sensorimmotor behaviour, locomotor 102. Vanderwolf, C. H., Near-total loss of ‘‘learning’’ and ‘‘memory’’ as activity and spatial navigation. Behav. Brain Res., 1985, 17, 103– a result of combined cholinergic and serotonergic blockade in the rat. 115. Behav. Brain Res., 1987, 23, 43–57. 117. Whishaw, I. Q. and Petrie, B. F., Cholinergic blockade in the rat 103. Vanderwolf, C. H., Cerebral activity and behavior: Control by central impairs strategy selection but not learning and retention of nonspatial cholinergic and serotonergic systems. Int. Rev. Neurobiol., 1988, 30, visual discrimination problems in a swimming pool. Behav. Neu- 225–340. rosci., 1988, 102, 662–677. 104. Vanderwolf, C. H. and Cain, D. P., The behavioral neurobiology of 118. Whishaw, I. Q. and Tomie, J., Cholinergic receptor blockade pro- learning and memory: A conceptual reorientation. Brain Res. Rev., duces impairments in a sensorimotor subsystem for place navigation 1994, 19, 264–297. in the rat: Evidence from sensory, motor, and acquisition tests in a 105. Vanderwolf, C. H., Kolb, B. and Cooley, R. K., Behavior of the rat swimming pool. Behav. Neurosci., 1987, 101, 603–616. after removal of the neocortex and hippocampal formation. J. Comp. 119. Whitton, P. S., Maione, S., Biggs, C. S. and Fowler, L. J., N-methyl- Physiol. Psychol., 1978, 92, 156–175. D-aspartate receptors modulate extracellular dopamine concentration 106. Vanderwolf, C. H.; Penava, D., Potentiation of the effects of anti- and metabolism in rat hippocampus and striatum in vivo. Brain Res., muscarinic drugs on behavior by serotonin depletion: Specificity and 1994, 635, 312–316. relation to learning and memory. In Levin, E. D.; Decker, M. W.; 120. Wozniak, D. F., Olney, J. W., Kettinger, L., Price, M. and Miller, J. Butcher, L. L., eds. Neurotransmitter Interactions and Cognitive P., Behavioral effects of MK-801 in the rat. Psychopharmacology, Function. Boston: Birkhauser 1992. 1990, 101, 47–56.