This is the unedited penultimate draft of a BBS target article that has been accepted for publication (Copyright 1996: Cambridge University Press) and is being circulated for Open Peer Commentary. This preprint is for inspection only, to help prospective commentators decide whether or not they wish to prepare a formal commentary. Please do not prepare a commentary unless you have received the hard copy, invitation, instructions and deadline information.
For information on becoming a commentator on this or other BBS target articles, write to: bbs@cogsci.soton.ac.uk
For information about subscribing or purchasing offprints of the published version, with commentaries and author's response, write to: journals_subscriptions@cup.org (North America) or journals_marketing@cup.cam.ac.uk (All other countries).


Long-Term Potentiation: What's Learning Got To Do With It?

Tracey J. Shors & Louis D. Matzel
Department of Psychology and Program in Neuroscience,
Princeton University, Princeton,
New Jersey 08544
shors@pucc.princeton.edu
Department of Psychology,
Program in Biopsychology and Behavioral Neuroscience,
Rutgers University, New Brunswick,
New Jersey 08903
matzel@rci.rutgers.edu

KEYWORDS

NMDA, synaptic plasticity, Hebbian synapses, calcium, hippocampus, theta rhythm, spatial learning, classical conditioning, attention, arousal, memory systems

ABSTRACT

Long-term potentiation (LTP) is operationally defined as a long-lasting increase in synaptic efficacy which follows high-frequency stimulation of afferent fibers. Since the first full description of the phenomenon in 1973, exploration of the mechanisms underlying LTP induction has been one of the most active areas of research in neuroscience. Of principal interest to those who study LTP, particularly LTP in the mammalian hippocampus, is its presumed role in the establishment of stable memories, a role consistent with "Hebbian" descriptions of memory formation. Other characteristics of LTP, including its rapid induction, persistence, and correlation with natural brain rhythms, provide circumstantial support for this connection to memory storage. Nonetheless, there is little empirical evidence that directly links LTP to the storage of memories. In this commentary, we review a range of cellular and behavioral characteristics of LTP, and evaluate whether those characteristics are consistent with the purported role of hippocampal LTP in memory formation. We suggest that much of the present focus on LTP reflects a preconception that LTP is a learning mechanism, although the empirical evidence often suggests that LTP is unsuitable for such a role. As an alternative to serving as a memory storage device, we propose that LTP may serve as a neural equivalent to an arousal or attention device in the brain. Accordingly, LTP is suggested to nonspecifically increase the effective salience of discrete external stimuli and thereby is capable of facilitating the induction of memories at distant synapses. In an environment open to critical inquiry, other hypotheses regarding the functional utility of this intensely studied mechanism are conceivable; the intent of this article is not exclusively to promote a single hypothesis, but rather to stimulate discussion about the neural mechanisms that are likely to underlie memory storage, and to appraise whether LTP can reasonably be considered a viable candidate for such a mechanism.


INTRODUCTION

Few topics in neurobiology during the past 20 years have attracted as much attention or commitment of resources as the phenomenon of long-term potentiation (LTP), a putative mechanism for the induction of stable memories in the mammalian brain. Long-term potentiation is typically expressed as an increase in synaptic efficacy lasting from hours to days following brief tetanic (high-frequency) stimulation of an afferent pathway. Thus, following LTP induction, a fixed amount of presynaptic stimulation induces a "potentiated" post-synaptic response, e.g., an increase in excitatory post-synaptic potentials (EPSPs). The phenomenon of LTP was initially observed in 1966 by Terge Lomo, then working in the laboratory of Per Anderson. In 1973, the first full article described LTP in the hippocampus of the rabbit, a collaborative effort between Lomo and Timothy Bliss (see also Bliss & Gardner-Medwin 1973). By 1989, the U.S. National Library of Medicine listed some 312 articles with the term "long-term potentiation" in the title, and in the 1990's alone, over 700 additional articles have appeared. This search vastly underestimates the research effort, because many articles that address LTP do not contain LTP in the title phrase or refer to the same phenomenon with a different name (e.g., "long-term enhancement"; McNaughton et al. 1986).

The concerted attention that LTP attracted over time is perhaps of no surprise to those familiar with the search for the engram (a neural memory store) and the associated mechanism that could account for its formation. Prior to the observation of LTP, the search had produced virtually no viable candidate mechanisms, at least in the vertebrate nervous system (c.f., Kandel & Tauc 1965a, 1965b). In this regard, LTP was and still may be the best candidate. In several recent reviews, different authors have concluded that not only is LTP a viable mechanism for the induction and storage of memories, but is the most promising candidate (e.g., Morris et al. 1991). In one article (Martinez & Derrick, 1996), the authors review recent evidence suggesting that the link between LTP and memory is in some cases tenuous, and in others even contradictory. Nevertheless, they conclude that "most evidence firmly supports a role for LTP in learning and memory" (also see Eichenbaum & Otto 1993). This conclusion is based, in part, on a commonly echoed assertion that although no direct evidence links LTP to memory, no better mechanism has been postulated. This assertion is encompassed by the broader argument that a good theory should not be abandoned until a better one replaces it, an approach with obvious merit. On the other hand, explicit confidence in the validity of a prevailing theory can obfuscate viable alternatives and new approaches to a problem. Einstein once stated that "it is the theory which decides what we can observe" (also see Kuhn 1973). A flawed theory, the explanatory value of which is outweighed by the inconsistencies that it introduces, can only serve as a detriment to empirical progress. To the extent that a theory is maintained by popular consensus, "what we can observe" will necessarily be obscured by the convictions that a theory or its advocates embrace.

Given the vast amount of attention that LTP has generated over the past 20 years, it seems an appropriate time to pause and review the cellular and behavioral characteristics of LTP that led us to consider it a memory device in the first place. We should evaluate whether these properties remain a viable feature of a memory device, and if so, whether LTP remains the most viable mechanism to serve that broader function. Of particular concern here is a distinction that we will draw between LTP and the formation and storage of memories versus a link between LTP and the processes which influence the formation and storage of memories. By influence, we mean that LTP may be neither a necessary nor sufficient condition for the actual storage of memories, but that LTP or an endogenous equivalent could act to facilitate and maintain learning indirectly by altering the organism's responsiveness, or perception of, environmental stimuli. In what follows, we will first review a number of the cellular properties intrinsic to LTP, with a particular emphasis on hippocampal LTP and the characteristics most commonly presented as evidence for its relationship to memory. It is important to stress that even if hippocampal LTP was the "learning mechanism", we would not expect individual synapses to express characteristics of learning and memory processes. Nevertheless, we will discuss them because they are the features commonly cited as evidence for the role of LTP in learning and they will allow us to evaluate the overall consistency of the evidence supporting LTP as a mechanism of memory storage. Second, we will review the behavioral evidence that links LTP in the hippocampal formation to learning and memory in the behaving animal. Finally, we will present an alternative hypothesis in which we propose that LTP is not a memory device per se, but rather, that it can influence the ultimate formation of memories by enhancing attention and the processing of sensory information.

CELLULAR PROPERTIES OF LTP AND THEIR RELATIONSHIP TO MEMORY

Distribution throughout the Nervous System

The contention that LTP might serve as a memory storage device stemmed, at least in part, from its discovery in the hippocampus, a structure that is critical to the formation of certain types of memories. Not only was LTP discovered in the hippocampus, but its distribution, in various forms, is evident at the three major synaptic connections of the structure. It is induced in the dentate gyrus granule cells by stimulation of the perforant path as originally described by Bliss and Lomo (1973), in the CA3 pyramidal cells by stimulation of the mossy fibers (e.g., Alger & Teyler 1976; Yamamoto & Chujo 1978), and in the CA1 pyramidal cells by stimulation of the Schaffer collateral branches of the CA3 neurons (Schwartzkroin & Wester 1975; Anderson et al. 1977). The initial description of LTP in the hippocampus was probably fortuitous for memory research; had it been first witnessed in a brain region with less of a historical link to memory formation (e.g., Olds 1955; Scoville & Milner 1957), LTP may not have received such focused attention. Since 1973 however, LTP has been found to occur in many brain regions, including piriform (Stripling et al. 1988), entorhinal (Wilhite et al. 1986), and prefrontal cortices (Larouche et al. 1990), the septum (Racine, Milgram, & Hafner 1983), the autonomic (Libet et al. 1975) and superior cervical ganglia (Brown & McAfee 1982), as well as in the ventral horn of the spinal cord (Pockett & Figurov 1993). Furthermore, LTP is not limited to the mammalian brain, but has been described in other vertebrates such as the gold fish (Lewis & Teyler 1986; Yang, Korn & Faber 1990), bullfrog (Koyano et al. 1985), bird (Scott & Bennett 1993) and lizard (Larson & Lynch 1985), as well as some invertebrates (Walters & Byrne 1985; Glanzman 1995). Since negative findings are usually not definitive, it cannot be said with certainly that LTP can not be induced in a particular brain region, but it is safe to say that phenomena fitting the general description of LTP occur ubiquitously throughout the nervous system. If LTP is a ubiquitous feature of the nervous system, what might that mean with respect to its potential role in learning and memory? Moreover, if LTP is indeed a learning and memory device, what would such a wide distribution tell us about the neural mechanisms of memory formation?

Most researchers would agree that memory formation requires or at least employs wide and distributed brain regions, and the hippocampus is clearly not the unitary "storage" site for memory; humans and infrahumans do not require a hippocampus to acquire many forms of memory, and even in tasks dependent on the hippocampus for acquisition, the structure is typically not required for later retrieval. If we begin with the premise that many memories are not actually stored in the hippocampus, then what function might LTP play there? Before discussing LTP's role in memory or any behavioral processes, however, we must first form an operational (and functional) definition of LTP.

Multiple Definitions of LTP

A serious impediment to determining or even discussing LTP's putative role in learning is the confusion regarding its definition. As operationally defined by Bliss and Lomo (1973), LTP is a persistent (hours) enhancement of an excitatory postsynaptic potential (EPSP) following brief high-frequency (tetanic) stimulation of afferent pathways, and at least formally, this definition (or a close variant of it) still predominates. For instance, several major text books, in describing LTP, essentially reiterate the earlier definition of Bliss & Lomo (e.g., Kandel, Schwartz, & Jessel 1991; Nicholls, Martin, Wallace, & Kuffler 1992). Similarly, one extensive review states that LTP is "an increase in synaptic efficacy, at monosynaptic junctions, occurring as a result of afferent fiber tetanization" (Teyler & DiScienna 1987).

Although these definitions are generally accepted and often used, they do not capture the range of conditions which are considered to reflect the induction of LTP. For this and other reasons, a number of researchers have either implicitly or explicitly narrowed the definition since its inception. This is, in part, understandable since a number of the properties of LTP were unknown at the time that Bliss and Lomo (1973) first described the phenomenon. For instance, much of the research aimed at elucidating the role of LTP in memory has focused on the hippocampal formation, presumably because LTP was discovered there and for some time was considered unique to that region. In addition, at the time of Bliss and Lomo's original observation, the N-methyl d-aspartate (NMDA) receptor had not yet been identified and thus did not enter into either the conceptualization or operational definition of the phenomenon. Since then (e.g., Collingridge et al. 1983; Harris et al. 1984), it has been determined that LTP at two of the major synaptic regions in the hippocampus (the dentate gyrus and area CA1) is, in part, dependent on calcium influx through the NMDA type of glutamate receptor and channel (details of this mechanism are described below). Thus, some researchers focus on the role of NMDA-dependent forms of LTP in memory (often stating that LTP is a NMDA-dependent phenomenon) despite the numerous instances where long-lasting increases in synaptic efficacy occur in the absence of NMDA receptor activation (Castillo et al. 1994; Jaffe & Johnston 1990; Johnston, Williams, Jaffe, & Gray 1992; Komatsu, Nakajima, & Toyama 1991). To add to the confusion, even in the dentate gyrus and CA1, LTP can be induced in the absence of activation of the NMDA receptor provided there is an alternate means of intracellular calcium accumulation, such as strong depolarization and subsequent influx of calcium through voltage-dependent channels (Malenka, Kauer, Zucker, & Nicoll 1988; Kullmann, Manabe, Perkel, du Lac, & Nicoll 1992; Wierazko & Ball 1993; Malenka 1992) or the release of Ca2+ from intracellular storage pools (Bortollotto et al. 1995). Thus, defining LTP based on its NMDA-dependence seems unnecessarily limiting and may be misleading with regard to a role for LTP in memory. This is not merely a semantic distinction that we are attempting to draw; the significance will become apparent in the discussion (below) of pharmacological manipulations presumed to affect both LTP and memory.

An antithetical, yet potentially more serious impediment to evaluating the link between LTP and memory formation is that the definition of LTP is often expanded to encompass virtually any observation of increased synaptic efficacy. By most accounts, memory storage is likely to involve a strengthening of specific synaptic connections (though these modifications need not be limited to synapses [e.g., Tesauro 1988]). But, in addition to high-frequency stimulation, there are a number of identified mechanisms whereby such synaptic strengthening can occur (see Hawkins, Kandel, & Siegelbaum 1993, for an integrative review). Many of these mechanisms are physiologically relevant and have been linked to memory formation. Thus, the observation of enhanced synaptic efficacy during learning does not necessarily indicate that the enhanced efficacy was induced by a mechanism analogous to that induced by high-frequency stimulation. For example, in one study (Weisz et al. 1984), rabbits were chronically implanted with stimulating electrodes in the perforant path and recording electrodes in the dentate gyrus. The rabbits were subsequently trained to associate a tone (conditioned stimulus, CS) with an aversive air puff to the eye (unconditioned stimulus, US), eventually eliciting a conditioned response (CR) to the tone. The results indicated that neuronal efficacy in the dentate gyrus was enhanced during acquisition of the conditioned response. Although it is tempting to conclude that the potentiation in the dentate gyrus reflected an LTP-like mechanism (c.f. Teyler & DiScienna 1987), there is no evidence that it arose from a stimulation pattern similar to that which would induce LTP. Moreover, given that the increase in learning and neural efficacy was correlational, it cannot be said that the potentiation contributes directly to the expression of the learned response. In fact, extensive experimentation by Thompson and his colleagues (e.g., McCormick et al. 1982; Krupa, Thompson, & Thompson 1993; Lavond, Kim, & Thompson 1993; Swain, Shinkman, Nordholm, & Thompson 1992; Knowlton & Thompson 1992; also see Berthier & Moore 1986; Yeo, Hardiman, & Glickstein 1986) suggests that the necessary and sufficient circuitry for the acquisition of the classically conditioned nictitating membrane response resides in the cerebellum. One mechanism for generating the conditioned response is considered to be a reduction in the activity of Purkinje neurons in response to stimulation of afferent mossy fibres/parallel fibres, the presumed pathway of the conditioned stimulus (CS). This observation is consistent with Ito's (Ito 1984) hypothesis that long-term depression (LTD), rather than long-term potentiation, is the relevant mechanism underlying memory storage in the cerebellum (e.g., Lavond, Kim, & Thompson 1993). If, however, one were to simply record activity induced by the conditioned stimulus in the interpositus or red nucleus (loci in the CR pathway efferent to the Purkinje neurons), an increase in the magnitude of the EPSP magnitude would be observed, due to a release from presynaptic Purkinje cell inhibition. Such an observation could easily lead one to conclude that "LTP" underlies learning in this system, when quite the opposite appears to be true.

It has been suggested that the term LTP actually refers to a presumed endogenous phenomenon and that the laboratory phenomenon is simply a tool to study a more general class of neuronal plasticity. Such an approach is entirely reasonable, but should be made explicit, in order that the operation which produces LTP in the laboratory is not considered a mechanism for storing memories in vivo. It is appreciated that many researchers recognize that the term LTP is generic, but oftentimes, written accounts about the role of LTP in memory storage suggest a more specific function. Statements such as "LTP underlies learning and memory" should perhaps be replaced with "enhanced synaptic efficacy underlies memory storage". Conversely, if the term LTP is simply intended to describe an increase in synaptic efficacy related to learning, then perhaps the "discovery" of LTP should be credited to Kandel and Tauc (1965a, 1965b), who first described heterosynaptic facilitation, an increase in synaptic efficacy which is related to behavioral sensitization. If enhanced synaptic efficacy is, in fact, the mechanism underlying memory formation (a topic that we cannot fully address here), and all forms of enhanced synaptic efficacy are deemed to be LTP, the hypothesis that LTP underlies memory formation cannot be disproved and serves no heuristic value. In the end, the term "LTP" becomes no more than a synonym for memory formation.

The use of the term LTP to describe all forms of enhanced synaptic efficacy might lead a casual observer to conclude that a common mechanism is shared by all. In 1987, Teyler and DiScienna constructed a partial list of 51 compounds (and classes of compounds) or manipulations that induce, prevent or reverse LTP. Since that time, the list has expanded tremendously, with particular emphases on modulators of protein kinases (O'Dell, Kandel, & Grant 1991; O'Dell, Grant, Karl, Soriano, & Kandel 1992; Fukunaga, Stoppini, Miyamoto, & Muller 1993; Kaczmarek 1992a; Malinow, Schulman, & Tsien 1989; Malinow, Madison, & Tsien 1988) and diffusible second messengers, such as arachidonic acid and nitric oxide (Schuman & Madison 1991; Williams, Li, Nayak, Errington, Murphy, & Bliss 1993; Haley, Wilcox, & Chapman 1992; Clements, Bliss, & Lynch 1991; Williams & Bliss 1989; Lynch, Clements, Voss, Bramham, & Bliss 1991; Bohme et al. 1991), as well as platelet activating factors (Goda 1994; Kato, Clark, Bazan, Zorumski 1994). Given the ever-expanding list of agents that are reported to induce an increase in synaptic efficacy referred to as LTP, one might reasonably ask whether all of these agents impinge on a common mechanism. For instance, protein kinase C (PKC) has been reported to play a role in LTP, based on findings that antagonists of the kinase block the induction of LTP (Akers, Lovinger, Colley, Linden, & Routtenberg 1986; Malinow, Madison, & Tsien 1988). Exogenous application of phorbol ester, a synthetic activator of the kinase and potent tumor promoter, can also induce potentiation (Malenka, Madison, & Nicoll 1986; Reymann, Schulzeck, Kase, & Matthies 1988). Likewise, it has been noted that D-alpha Tocopherol (Vitamin E) induces synaptic potentiation, purportedly through its antioxidant, or tumor inhibiting properties (Xie & Sastry 1993). It seems likely that the synaptic potentiation induced by phorbol ester versus that induced by D-alpha-Tocopherol is regulated by different underlying substrates. However, both are referred to as LTP in the titles or abstracts of their respective articles. There are also reports of enhanced LTP through caloric restriction (Hori, Hirotsu, Davis, & Carpenter 1992), and prevention of LTP induction by the sugar substitute saccharin (Morishita, Xie, Chirwa, May, & Sastry 1992), as well as by cocaine (Smith, Browning, & Dunwiddie 1993). Again, the common denominator linking these observations is the term "long-term potentiation" in the title of the reports. The point is that one should not assume that a single mechanism is shared by all; rather, the length and breadth of the list of modulators suggests that they could not impinge on a single mechanism or even a single class of mechanisms.

In summary of this issue, there are at least two approaches to establishing an acceptable definition of LTP. One is to allow the term to encompass all long-lasting forms of potentiation. This approach renders the term almost meaningless and makes the presumed connection between LTP and memory an unfalsifiable construct. The second approach is to partially limit the definition. For the purposes of this review, we have taken the position that all forms of synaptic modifications related to learning and memory are not equivalent. Nevertheless, with regard to experiments that attempt to link LTP to behavior, we review articles that describe manipulations which the authors suggest impinge on "LTP". We recognize that this is not much of a limitation, and on occasion, dispute the authors claims in an attempt to illustrate the necessity for a more precise nomenclature. Moreover, we focus, but do not limit, our discussion to LTP in the hippocampal formation. This is necessary due to both space limitations and also because LTP in the hippocampus has been the most intensely studied with respect to learning and memory.

NMDA Receptors and Postsynaptic Calcium.

One defining feature of LTP is its dependence on high levels of postsynaptic calcium, a common feature of most learning-induced neuronal modifications. In and of itself, a definition which includes "calcium dependence" provides little insight since a wide range of cellular functions require calcium and still more are dependent on elevations of intracellular Ca2+ above basal levels. Although the exact role of calcium in LTP induction is a matter of debate, elevation of postsynaptic calcium is clearly necessary, and may even be sufficient for the induction of hippocampal LTP. Induction of LTP is prevented by a pretetanus injection of calcium chelators into the postsynaptic cell (Lynch et al. 1983; Malenka et al. 1988), and induction occurs when the postsynaptic cell is artificially loaded with the ion (Malenka et al. 1988). A great deal of evidence (e.g., Collingridge et al. 1983; Harris et al. 1984; Jahr & Stevens 1987) indicates that the primary source of calcium influx during the induction of hippocampal LTP occurs through an ion channel that is coupled to the NMDA subtype of glutamate receptor. This receptor is unique in that stimulation of the channel ionophore requires glutamate binding as well as a moderate level of depolarization. At normal resting potentials (~ -70 mV), the channel is blocked by magnesium, and glutamate binding is insufficient to open it. However, at depolarized membrane potentials (> -40 mV), magnesium is expelled from the channel, which can then be opened by glutamate and which displays a high selectivity to calcium ions. Thus, the NMDA receptor complex is said to be dually regulated by two factors: ligand and voltage. These cofactors can be recruited through several means. First, a relatively long, high intensity presynaptic burst of activity (such as a high-frequency train of stimulation) can induce LTP by releasing glutamate onto the postsynaptic receptor, while depolarizing the postsynaptic cell through stimulation of the non-NMDA type of glutamate receptors (AMPA). Second, shorter and more physiologically relevant levels of presynaptic activity can induce hippocampal LTP by stimulating the NMDA receptor with glutamate, while the postsynaptic cell is depolarized via an alternative means such as an input from a second afferent pathway. Other forms of LTP, such as that induced in CA3 pyramidal cells following mossy fiber tetanization, occur independently of the NMDA receptor, and are instead dependent on Ca2+ influx through voltage-gated channels, with some debate as to whether the critical Ca2+ signal occurs pre- (Weisskopf et al. 1994; Castillo et al. 1994) or postsynaptically (Williams & Johnston 1989; Johnston et al. 1992). As alluded to earlier, even in area CA1, LTP can be induced without the participation of NMDA receptors, provided that the tetanus (or postsynaptic depolarization) is of sufficient intensity to activate voltage-dependent calcium channels (Grover and Tyler 1990; Kullman et al. 1992). Thus, activation of the NMDA receptor is critical to many forms of LTP, but it is not necessary for all. In contrast, intracellular calcium appears to be a necessary element for the induction of LTP. A necessary role for calcium in LTP is consistent with LTP's presumed role in learning; calcium plays a critical role in many cellular modifications thought to underlie conditioned behavioral responses (e.g., Abrams & Kandel 1988; Walters & Byrne 1985; Falk-Vairant & Crow 1993; Matzel & Rogers 1993).

Synaptic Efficacy, Specificity and Memory.

The search for the engram has been guided by a number of expectations regarding what features a memory mechanism should possess. Some of these expectations have been strengthened through experimentation, whereas others were ultimately discarded (c.f. Chapouthier 1989; Gaito 1976). One reasonable expectation is that learning is accompanied by an increase in the efficiency of communication between neurons, a concept with long historical antecedents (e.g., Spencer 1870; James 1892; Tanzi 1893). The formalization of this idea is usually attributed to Donald Hebb. In 1949, Hebb wrote in his book The Organization of Behavior, "When an axon of cell A ... excite(s) cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells so that A's efficiency as one of the cells firing B is increased." This particular line from Hebb's treatise, subsequently referred to as "Hebb's Rule", closely resembles the operational definition of LTP and is frequently offered as a theoretical foundation for the presumed role of LTP in learning.

In addition to its expected basis in a modulation of synaptic efficacy, the search for the engram is based on a second expectation, that of synapse specificity. Besides its intuitive appeal, strong empirical support exists for synapse-specific changes that accompany the learning process (Clark & Kandel 1993; for review, see Hawkins, Kandel, & Siegelbaum 1993), and hippocampal LTP itself is considered synapse-specific (Andersen et al. 1977; Dunwiddie & Lynch 1978). But, much like the definition of LTP, the term "synapse specificity" is often used to describe very different phenomena. Under some circumstances, synapse specificity implies that the modifications which underlie LTP are limited to synapses. This type of synapse specificity has intuitive appeal because it provides the necessary structure for a huge memory capacity, well over and beyond that which could be achieved through somatic potentiation. However, the modifications induced by LTP are rarely, if ever, limited to synapses. For instance, in the original description of LTP, Bliss and Lomo (1973) reported a phenomenon since referred to as E/S potentiation. Following LTP induction, an increase in population spike amplitude (S) and a reduced threshold for cell firing can be observed even if the magnitude of the excitatory post-synaptic potential (E) is held constant, indicating that the tetanus-induced modification is not limited to the synapse. Using an extracellular population spike as a dependent measure, an increase in efficacy can reflect changes which occur exclusively in the soma, along the entire membrane, at synaptic terminals, or some combination of the three. Using an extracellular excitatory post-synaptic potential (EPSP) as the dependent measure, on the other hand, does not allow one to measure the changes that may be at other cellular loci. Thus, changes at the soma, such as the nonsynaptic forms of potentiation which typically accompany an increase in synaptic efficacy, are virtually ignored (Anderson et al. 1977; Anderson et al. 1980; Bliss et al. 1987). Given that there is no a priori reason to require that a memory mechanism be limited to modifications of synaptic terminals, these observations do not reflect on the validity of LTP to serve this function. Neither though, should this interpretation of synapse specificity be used as evidence to support LTP's role in learning.

The second and more common use of the term "synapse specificity" is in reference to potentiation that is limited to synapses active during stimulation (as opposed to inactive synapses). Under these conditions, potentiated synapses are proposed to reflect an independent memory store, and those synapses would be preferentially activated during retrieval, a seemingly critical feature of memory. In our opinion and probably many others, this is one of the more compelling aspects of the hypothesis that hippocampal LTP is involved in memory formation. Unlike the prior description of the term "specificity", this usage does not necessarily require that potentiation be restricted to a modification of the synapse. Rather, specificity here suggests that changes will be restricted to those synapses (and possibly other compartments) that are active during the induction of LTP.

At this point, it should be noted that the plastic changes associated with learning often are not limited to the synapses active during the learning event. In fact, many models of complex memory processing explicitly require that memory become distributed among different locations in the nervous system following its initial induction (e.g., Atkinson & Shiffrin 1968), and several lines of empirical evidence support such a view. For instance, in the chick nervous system, the necessary circuitry for the expression of conditioned taste aversion (a form of associative memory) shifts to several anatomically distinct brain areas within days and weeks after initial learning, such that a lesion which disrupts recall at one retention interval may not affect retention at another (Rose 1992; 1995). In the isolated ganglion of the cockroach, an operant leg position response expressed in the prothoracic ganglion of the roach will later be expressed in an untrained, mesothoracic leg, suggesting transfer of information between ganglion (Hoyle 1980; for review, see Eisenstein & Reep, 1985). Evidence for the transfer of "stored" information out of the hippocampus has also been observed. The noted patient H.M. (Scoville & Milner 1957), who underwent bilateral excision of the medial temporal region (including parahippocampal gyrus, amygdala, and anterior portions of the hippocampus), cannot transfer short-term memories into long-term storage, though very short-term memories and many memories established prior to surgery are spared. This, as well as corroborative work with animals (e.g., Kim & Fanselow 1992), indicates that the hippocampus is not in fact a memory "store", but rather a temporary holding site critical to the integration and consolidation of memories which presumably occur in higher cortical areas or elsewhere (e.g., Squire, Cohen & Nadel 1984; Zola-Morgan & Squire 1991). Thus, in general, it does not appear that the brain structure used for acquiring memories is necessarily the site for storage.

Nonetheless, if one accepts the premise that synapse specificity is a defining feature at least of memory induction, then evidence that LTP is not specific to the synapses which were active during stimulation suggests that LTP fails to meet the needs of a memory mechanism. Recently, a number of researchers have reported that the potentiation is not specific to the synapses which were active during afferent stimulation. Boenhoffer, Staiger, & Aertsen (1989) recorded simultaneously from neighboring CA1 pyramidal cells while stimulating the afferent Schaffer collateral fibers at a low frequency, which in itself will not induce LTP. When stimulation was paired with depolarization of one of the pyramidal cells, potentiation was observed not only in that cell, but in a neighboring cell as well. This result was elaborated by Schuman & Madison (1994) who reported that a spread of potentiation could be detected 250 um (but not 500 um) away from the site of induction, and that inhibition of the diffusible gas nitric oxide blocked the spread. These results suggested that the conditions which induce LTP at one synapse could spread over potentially thousands of adjacent synapses, thus abrogating the possibility that tetanized synapses could serve as individual memory storage units. It should be noted, however, that such a dissipating spread of potentiation might well contribute to the generalization gradient which is an ubiquitous feature of most instances of learning (Mackintosh, 1975). [This discussion is subject to two caveats. First, the influence of nitric oxide on the maintenance of LTP maintenance has been studied most intensively in the hippocampus slice preparation where the effects have been equivocal (Cummings et al. 1994). Second, the gas may not play a critical role in memory induction under physiological conditions in vivo. For instance, Bannerman et al. (1994) report that a 90% inhibition of nitric oxide synthesis in the brain had no effect on acquisition in the water maze, although nonspecific behavioral impairments were apparent].

In addition to the spread of overt potentiation to nearby synapses, there are also changes which accompany the induction of LTP, but which are not limited to active synapses or even nearby synapses. For example, unilateral tetanization of the perforant path induces LTP in the dentate gyrus and an increase in messenger RNA (mRNA) for a presynaptic glutamate receptor on the stimulated (ipsilateral) side two hours later (Smirnova et al. 1993). Within five hours of the induction of LTP, however, levels of mRNA are also increased on the contralateral side, in areas that presumably are not exhibiting LTP. Thus, mRNA levels are increased in response to the induction of LTP in regions that do not exhibit enhanced synaptic efficacy and presumably were not active during tetanization. Smirnova et al. concluded that "the induction of LTP at one stage in a neural network may lead to modification in synaptic function at the next stage of the network". In a related example, LTP was again induced in the dentate gyrus following unilateral tetanization of the perforant path. One hour later and in the presence of potentiation in the dentate gyrus, there was a bilateral increase in the binding affinity of the AMPA-type of glutamate receptor (Tocco et al. 1992). This increase was not confined to the dentate gyrus, but occurred in regions throughout both hippocampi. Although not synapse specific, the increase was "specific" in the sense that it was prevented by NMDA antagonists and did not occur in response to stimulation at frequencies too low to elicit LTP. As a third example, unilateral tetanization of the perforant path caused a bilateral increase in mRNA for two neurotrophins; brain-derived neurotrophic factor and nerve growth factor (Castren et al. 1993). These three examples are considered "nonspecific" responses to LTP, and therefore may be of minimal interest to those concerned with understanding the mechanism of LTP induction. Nonetheless, it is clear that a number of changes in neuronal function are occurring in response to the typical induction protocols for LTP, and those changes are not limited to the synapses active during LTP induction.

Although these data indicate that the effects of LTP are not confined to the synapses active during the induction protocol, when viewed from an integrated brain systems approach these transynaptic and "nonspecific" effects may provide some of the most convincing evidence that LTP does have physiological relevance. If one were to assume that information transfer in the nervous system occurs during memory storage, as a number of studies suggest, then the transynaptic modifications may provide clues about the mechanism of that transfer, and thus memory formation itself. For example, the fact that tetanization of the perforant pathway in the hippocampus can increase AMPA binding as remotely as the neocortex (Tocco et al. 1991) suggests that the conditions which induce LTP in the hippocampus also affect structures involved in perception and presumably memory storage. We feel that these nonspecific responses should not be dismissed, but rather, should be appreciated as providing potential clues about the neural mechanisms of information processing.

In summary, the assertion that hippocampal LTP is a memory device because of its limitation to synapses themselves or to synapses active during tetanization reflects a preconception about the nature of memory, rather than an empirically derived observation about memory. For two decades, the presumed "synapse-specific" nature of LTP was cited as support for the argument that hippocampal LTP was a viable substrate of memory. Now that it has been shown that LTP is not necessarily confined to the active synapse, it has been suggested that LTP is a viable substrate of memory because memories are "distributed" (see Barinaga 1994, for commentary). Obviously, these two lines of reasoning are incompatible, and reflect our tendency to validate the theory that LTP is a mechanism of learning and memory with the theory itself.

Long-lasting, but Decremental.

By far, one of the more perplexing issues regarding memory storage in the brain is how a biological representation of a memory can be sustained for such inordinate periods of time. The mechanism suited for such a task must not only be capable of acquiring and encoding the perceived information, but also storing it long after the proteins involved in the initial storage have been degraded and replaced (hours to days). (For a discussion of mechanisms that might underlie long-term modifications of synaptic efficacy, see Lisman, 1994; Miller & Kennedy 1986; Schwartz & Greenberg 1987.) Indeed, relative to most forms of neuronal plasticity, LTP is long-lasting. Other forms of potentiation in the mammalian nervous system persist on the order of seconds, usually not hours and certainly not weeks. In contrast, LTP can persist for weeks in area CA1 in vivo (a median of 10.5 days in CA1; Staubli et al. 1987) and approximately 8 hours, in vitro (Reymann et al. 1985).

There is no doubt that LTP is long-lasting, but is it long enough? Memories can persist intact throughout the life span of the animal (Spear 1978), whereas LTP always decays (c.f. Staubli et al. 1987). To retard the rate of decay, many investigators that have observed the effects of LTP on behavior deliver multiple tetani, sometimes exceeding 100 high-frequency trains over days or weeks (e.g., Castro et al. 1989). But even with this extended "training" regimen, potentiation almost always decays to baseline levels within a week. These results suggest that while LTP is long-lasting, it does not correspond to the time course of a typical long-term memory. It is recognized that many memories do not last a life-time, but taking this point into consideration, we would then have to propose that LTP is only involved in the storage of short-term to intermediate memories. Again, we would be at a loss for a brain mechanism for the storage of a long-term memory.

In order to account for the decremental nature of LTP, some have suggested that a process opposing LTP, such as long-term depression (LTD), can supplant previously potentiated synapses (Pavlides et al. 1988; Sejenowski, 1990). Thus, LTP would decay because of the natural occurrence of LTD in subsets of the potentiated synapses. Similarly, it has been shown that potentiation of one subset of synapses can cause depression of surrounding, non-potentiated synapses ("heterosynaptic depression"; Lynch et al. 1977). While these observations can explain the decremental nature of LTP in vivo, they do not necessarily address the hypothesis that LTP underlies long-term (days-to-years) memory storage; the loss of LTP would degrade the memory regardless of whether it was due to inherent decay or whether it was supplanted by LTD. It is noted that decremental LTP in the hippocampus is only fatal to the hypothesis that LTP is responsible for storage of long-term memories in the hippocampus; it is not necessary for the hypothesis that LTP in the hippocampus serves as a temporary role in the acquisition of sensory information with the memory trace eventually distributed in other brain locations. With respect to LTP in the hippocampus, this last hypothesis is consistent with the empirical evidence suggesting that the hippocampus is preferentially involved in the acquisition of specific types of short-term memory. It must be noted nonetheless that nondecremental LTP has not been observed in any brain structure.

Strengthening through Repetition and Facilitated Reacquisition

Another potential link between LTP and memory is the issue of "strengthening through repetition." Although memory induction can certainly be complete within a single trial (Rock, 1956; Estes, 1970), there are numerous instances in which memory is strengthened by repeated exposure to the learning event. Thus, if LTP was involved in memory formation, it too should be strengthened through repetition. Indeed, synaptic efficacy can be strengthened through repeated exposure to the tetanizing stimulus, provided that the additional tetani are delivered before the potentiation decays back to baseline levels. If the response is allowed to decay back to baseline, however, LTP is neither more easily induced nor more persistent than after the initial induction (de Jonge & Racine , 1985). Therefore, hippocampal LTP does exhibit aspects of "strengthening through repetition", but does not exhibit "facilitated reacquisition", which is a defining feature of most memory processes (e.g., Spear & Riccio, 1993; Miller et al. 1986; Matzel et al. 1992).

Associativity and Cooperativity.

Two additional features of hippocampal LTP, associativity and cooperativity, are often cited as evidence that LTP is involved in the learning process. The terms "associativity" and "cooperativity" derive from the procedures used in Pavlovian conditioning, in which two stimuli or events presented in a temporally contiguous manner tend to become associated with one another. (This is an oversimplification of the characteristics of Pavlovian learning; for further discussion of the subject, see Rescorla 1988). With regard to LTP, cooperativity refers to the observation that an intensity threshold must be met for successful induction (Bliss & Gardener-Medwin, 1973). This threshold can be reached through intense stimulation of a single or a few afferent fibers, or cooperatively through a lower intensity stimulation of many fibers (McNaughton et al. 1978). Similarly, associativity refers to the observation that roughly contiguous, low intensity stimulation of two pathways, or higher intensity stimulation of weak inputs, converging on the same cell are sufficient for the induction of LTP when stimulation of neither pathway alone is sufficient (Barrionuevo & Brown, 1983; Levy & Steward, 1979, 1983). Associativity probably represents the same underlying mechanism as cooperativity, but differs somewhat operationally in that associative interactions can occur across spatially distal regions of a dendrite, and may reflect the contiguous pre- and postsynaptic activity implied by Hebb's rule. Typically, both associativity and cooperativity are explained by the necessity for a sufficient level of postsynaptic activity (McNaughton et al. 1978), and are presumed to reflect the addition of multiple postsynaptic calcium signals. The existence of associativity and cooperativity in LTP are important for several reasons. First, their existence indicates that physiologically relevant levels of stimulation can induce LTP. Second, they suggest that LTP is unlikely to result from normal activity, but rather, might be reserved for detection of spatially and temporally contiguous events. This later point will be recognized as analogous to a defining feature of classical conditioning, and thus has been cited as support for the role of LTP in associative learning (e.g., Brown et al. 1990).

Although the associative and cooperative properties of LTP lend support to its relevance to Pavlovian conditioning, aspects of the two phenomena raise questions about this presumed connection. For example, the stimulation of two converging inputs is most effective in inducing LTP when those two inputs are stimulated in a temporally contiguous or near contiguous (100 ms) manner. In contrast, the optimal interstimulus interval (ISI) between the conditioned (CS) and unconditioned stimuli (US) in classical conditioning varies from several hundred milliseconds in the rabbit eye blink preparation, to several seconds in rabbit conditioned bradycardia, to tens of seconds for many conditioned emotional responses, to hours for conditioned taste aversions (for review, see Mackintosh 1974). Moreover, a constant interstimulus interval may produce inhibitory or excitatory learning depending on the interval between successive trials (Kaplan & Hearst, 1985), and the systematic relationship between the onset of stimuli is entirely absent in the case of context learning. Consequently, the observation that the induction of LTP is most effective at relatively short interstimulus intervals (0 - 200 ms) should not be taken as evidence for its relevance to Pavlovian conditioning, both because there is no universally optimal interstimulus interval, and because the interval used to induce associative LTP is shorter than is optimal for any behavioral conditioning procedure that we are aware. It should be noted though, that near simultaneity of stimuli has been suggested to support most efficient learning in Pavlovian paradigms, while the expression of that learning varies depending on the response system (Matzel et al. 1988; Rescorla 1980, 1988). These issues reflect on the dangers of oversimplifying Pavlovian phenomena in order to make a comparison to a biological system.

A second issue about associativity and cooperativity concerns the neural mechanism which underlies the induction of hippocampal LTP. As described, associativity and cooperativity are thought to arise from a sufficient level of postsynaptic activity and hence, an accumulation of postsynaptic calcium. In essence, the associative feature of LTP is simply the successful expression of what might occur "nonassociatively", that is, with sufficient stimulation of a single afferent fiber. This raises questions regarding its general relevance to Pavlovian conditioning, or even associative learning in general. In summary of this issue, the associative and cooperative features of LTP suggest certain similarities to basic associative learning, but a direct link between the associative features of LTP and associative memory has not been made (or further discussion, see Diamond & Rose, 1994).

LTP AND BEHAVIORAL INDICES OF MEMORY

To this point, we have reviewed a number of the cellular properties of hippocampal LTP that many consider to be indicative, or at least compatible with, its role in learning and memorial processes. Many of the these correlations were based on preconceptions about what the critical features of memory formation should be, and others were indeed consistent with those necessary for memory formation. Others, such as the spread of certain "nonsynaptic" correlates of hippocampal LTP to nonstimulated pathways are inconsistent with certain preconceptions about memory processes, but may provide important clues regarding the role that the induction of LTP in vivo has on behavior. We shall now review a series of experiments that are often cited as evidence in support of a link between hippocampal LTP and memory storage.

Of the 700+ articles published between 1990 and 1996 which refer specifically to LTP in the title, more than 600 either imply or explicitly state in the abstract or introduction that LTP is a memory storage device. The statements range from speculation that LTP "may underlie learning" to definitive statements that it "underlies learning and memory," "is associated with the formation of memory," and "contributes to memory encoding." It was thus surprising to discover that out of 700+ articles, less than 50 contained a behavioral manipulation of memory itself. When the search was extended back to 1974, less than 80 out of 1000+ articles with LTP in the title contained any behavioral manipulation relevant to the assessment of memory. Given these statistics, one might assume that it had been demonstrated that LTP was "the memory mechanism", and that further studies were unnecessary. In fact, many articles provide evidence to the contrary (see Hippocampus, 1993, number 2; Bannerman et al. 1995; Saucier & Cain, 1995).

Pharmacological and Genetic Manipulations of LTP

Three lines of evidence have supported the premise that hippocampal LTP was involved in acquisition and/or storage of memories. One was causal (once removed), and the other two were correlational. The first involved the pharmacological blockade of LTP induction, followed by learning trials and ultimately a test of memory. In response to a competitive NMDA antagonist which prevents the induction of some forms of LTP, rats were impaired in their ability to perform the Morris water maze, a spatial memory task that requires the hippocampus for successful completion (Morris et al. 1986). This experiment and a multitude of similar ones encountered interpretive difficulties, due to the effects of NMDA receptor antagonists on sensory/motor performance; most of these drugs are chemically related to the street drug "angel dust", which can cause profound perceptual distortion, even hallucinations (Julien, 1992). These concerns about performance were expressed in a series of comments and rebuttals published in the journal Psychobiology (Keith & Rudy, 1990, number 3) and we will not thoroughly review them here. However, the debates were based on experiments which remain the most cited evidence for a link between hippocampal LTP and behavioral learning, and thus a brief overview is required.

In the critique by Keith and Rudy (1990), it was noted that in tasks which require the hippocampal formation for acquisition (e.g., the water maze and olfactory discrimination learning), NMDA receptor antagonists in concentrations that do not induce obvious behavioral impairments only mildly disrupt acquisition, and only under a narrow range of conditions. Moreover, drug-treated animals ultimately attain levels of performance equivalent to control animals. Keith and Rudy interpreted these results as evidence that activation of the NMDA receptor (and hence NMDA-dependent LTP) is not necessary and certainly not sufficient for learning these tasks. Further, they suggested that the mild "learning" deficit induced by the NMDA antagonist reflected no more than a subtle sensory or motor impairment, and/or an anxiolytic effect (Bennet & Amrich 1986; Clineschmidt et al. 1982). Staubli (1990) and Lynch and Staubli (1990), interpret these results somewhat differently, suggesting that under normal conditions, NMDA receptor-dependent LTP is the primary mechanism that underlies learning, but that in its absence, a secondary and slower learning mechanism is used. Hence according to Staubli and Lynch, the animals learn, but at a reduced rate. Because biological systems are often redundant, this latter interpretation is certainly plausible, even though it is not an obvious a priori prediction. Moreover, olfactory discrimination learning is possible in neonatal and prenatal rats (Johanson & Hall 1979; Johanson & Hall 1982; Smotherman & Robinson 1991; Smotherman 1982) prior to the expression of NMDA receptors (Harris & Teller, 1984; Baudry et al. 1981; Wilson, 1984; Duffy & Teller, 1978). Although it is not known whether these forms of learning require a hippocampus during early development, the observations suggest that NMDA-dependent forms of plasticity are not the "primary" mechanism of memory during that time.

In response to concerns about performance deficits due to peripheral injection of NMDA antagonists, Morris et al. (1986) injected the antagonist directly into the ventricle surrounding the hippocampus and found that rats were still impaired in their acquisition of the maze. On the first three-trial block (before any substantial learning would normally occur), animals treated with the antagonist exhibited an increased escape latency relative to the untreated group or a group treated with an inactive isomer of the drug. These results suggested that the antagonist did have an effect on processes other than memory formation itself. Indeed, intraventricular administration reduces, but does not necessarily eliminate the possibility that sensory, motor, or motivational processes have been disrupted. It has been reported that the ventricular administration of the NMDA receptor antagonist AP5 evokes subtle anxiolytic and analgesic effects, impairs motor control, and induces muscle flaccidity (Dale 1989; Dale & Roberts 1984; Turski et al. 1985). After administration of the highest dose of antagonist, one study reported that during the first 9 trials, rats "occasionally fell off the escape platform" (Morris et al. 1986). Recent work by Cain et al. (1995) and Caramano and Shapiro (1994) provides more evidence of behavioral abnormalities following intraventricular administration of NMDA receptor antagonists. Using concentrations comparable to those reported by Morris (1991) and Davis et al. (1992), Cain reports that the rats display behavioral hyperactivity and ataxia, a decrease in the rate of swimming, thigmotaxis, and a variety of indirect swim patterns. These behavioral disturbances accounted for over 70% of the variance in acquisition of the water-maze task.

To reduce the influence of motor deficits following the administration of NMDA receptor antagonists, Morris trained his rats to use the escape platform in a water maze prior to actual spatial navigation training in the maze (e.g., Morris, 1991; Davis et al. 1992). With this pretraining, the antagonist impaired spatial learning but did not elicit an obvious motor impairment and did not impair performance on the first trial. In addition, the antagonist did not affect performance on the visual version of task, where a platform is randomly located on each trial and the rat must locate it and escape. Importantly, in companion histological and electrophysiological studies, Morris et al. 1989 demonstrated that the radiolabeled NMDA antagonists did not diffuse out of the hippocampus and only concentrations of the antagonist that impaired LTP impaired spatial learning. This comprehensive series of experiments led Morris to conclude that "these data provide strong support for the now widely accepted view that the neural mechanisms underlying NMDA-dependent hippocampal LTP play a role in spatial and perhaps other kinds of learning" (Morris 1991; see also, Davis et al. 1992).

Several comments should be made regarding this conclusion. First, the antagonist only "slows the rate of learning rather than blocking learning completely", and thus the findings do not support the idea that NMDA receptor-dependent LTP is a singular neural mechanism for the establishment of a neural memory trace. Second, it appears that the procedures employed by Morris (1991) and Davis et al. (1992) may not have been adequate to control for the antagonist's effects on motor performance. Very recently, it was reported by Saucier and Cain (1995) and Morris and colleagues (Bannerman et al. 1995) that prior training with a spatial or nonspatial version of the water maze attenuated deficits in subsequent maze learning conducted under the influence of NMDA receptor antagonists. Two different interpretations of these results were offered, one suggesting that NMDA receptor activation (and by association , LTP) is still involved in learning the spatial maze, but not the learning of spatial location, per se (i.e., nonspatial aspects of the task). The other interpretation, preferred by Saucier and Cain, is that the prior training on a maze precluded the sensory and motor deficits typically encountered during the initial acquisition of spatial learning, and thus NMDA receptor activation (and LTP) is not necessary for hippocampal learning. Thus, in conclusion of this issue, NMDA receptor antagonists can impair performance (and perhaps procedural memory formation) in spatial learning tasks, but it is not clear that the effect is specific to learning or to a disruption of hippocampal LTP.

In contrast to the impaired performance observed in the water maze, others have reported that NMDA antagonists can actually facilitate learning. Mondadori and colleagues (e.g., 1989) have reported an enhancement in passive avoidance learning following peripheral administration of NMDA antagonists that prevent LTP. Consistent with the discussion of the antagonist and spatial learning, it could be argued that the enhancement is due to effects on motor performance. For example, the antagonist could impair performance on spatial navigation task and facilitate performance in the passive avoidance task simply because the first task requires active and coordinated movement while the second task requires passivity. Mondadori et al. (1989) performed a range of control procedures to rule out such an interpretation. For example, rats that received the drug did not exhibit an increase in their latency to initiate normal exploratory behavior. Consequently, we are left to conclude either that the results of these two sets of studies can be explained by motor impairments or that they directly reflect the effect of NMDA antagonists on memory formation. If the later is true, then NMDA antagonists can either retard, facilitate or have no effect on learning.

Adding the complexity of this issue, it is also the case that NMDA antagonists have effects on tasks that have no obvious dependence on the hippocampus. Robinson (1993) reported that the noncompetitive NMDA antagonist MK-801 retards classical conditioning of the eye blink response, a task that is not dependent on an intact hippocampus. Thus, one can obtain decrements in learning with NMDA antagonists even when there is no reason, a priori, to believe that hippocampal LTP is essential to the process. It is interesting to note that MK-801 did not prevent the "LTP-like" increases in synaptic efficacy in the hippocampus that parallel the learning, further supporting our contention that all forms of synaptic potentiation should not be classified as "LTP".

At the behavioral level, attempts to clarify the biological basis of learning through pharmacological manipulation is certain to be plagued by issues of "nonspecific" effects on performance. At a cellular level, these problems are no less enigmatic. Take, for example, studies aimed at establishing a connection between protein kinases in hippocampal LTP and learning (Malinow et al. 1989, Malenka et al. 1989, O'Dell et al. 1991; Zhuo et al. 1994). Two factors (somewhat related) have made it especially difficult to deduce the contribution of these enzymes to LTP: first, kinase inhibitors are relatively nonspecific, due to their effects on enzymes or processes other than those that were intended. Second, the kinases which are targeted are often involved in a multitude of cellular processes which are unrelated to plasticity. The first factor has been, at least in part, addressed with the relatively new technique of gene deletion, where a gene for a specific protein is ablated in the embryonic mouse. This strategy has been used to assess the role of CaM Kinase (Silva et al. 1992a, 1992b) and a subtype of tyrosine kinase (O'Dell et al. 1992) in hippocampal LTP and learning. Both mutations resulted in partial impairments of LTP and corresponding impairments of spatial performance in the Morris water maze. While these techniques may increase specificity for a targeted protein, their interpretation is complicated by the fact that the animal has gone through development without the gene, and many of these genes are critical for normal cellular and behavioral development. For example, deletion of the fyn gene used in one knock-out study (Grant et al. 1992) was later reported to retard the development of myelination in the nervous system (Umemori et al. 1994), and to disrupt normal suckling in neonates (sometimes resulting in death), as well as causing gross abnormalities in the hippocampal formation (Yagi et al. 1993). In addition, the location of the deficit caused by the knockout is not specific to a particular brain region, and the knockouts may cause gross anatomic abnormalities. For example, the fyn gene knockout results in an irregular and undersized olfactory bulb (Yagi et al. 1993). There are also some dissociations between the effects of the gene lesions and learning. For example, deletion of the pcd gene caused degeneration of Purkinje cells in the cerebellum, but has no apparent effect on the gross morphology of the hippocampus. Interestingly, the pcd deletion impaired performance in the Morris water maze (Goodlett et al. 1992).

As suggested, gene mutations can introduce gross behavioral abnormalities that, like NMDA antagonists, can have profound effects on performance irrespective of learning. In one study (Silva et al. 1992), mice lacking the gene for CaM kinase were slow to learn the location of the hidden platform in a Morris water maze, suggesting a role for the kinase in spatial learning. However, the deficient animals were slow to swim to the platform on the first training trial, before any learning could have occurred. The poor performance was attributed to "fatigue" in the mutant mice that was suggested to results from abnormal "jumpiness", a descriptor that has now found its way into several of the reports concerned with the learning capacity of mutant mice (Sakimura et al. 1995). The fyn knock-outs (Grant et al. 1992) displayed a similar performance deficit on the first trial, again prior to the point at which learning could have occurred. Further, the mutants learned at a similar rate as the wild-type controls, and by the sixth trial, were performing at identical levels. In fact, because some mutant mice reached a preimposed 60-sec cutoff on the first trial, the rate of learning for the group may have been underestimated, and may have exceeded that of the wild-type controls. A more detailed account of the interpretative difficulties in these particular studies has been presented by Deutsch (1993). Given the evidence published to date, we are inclined to accept his conclusion that there is "no evidence that the mutant mice in [these] studies suffered from a specific impairment of memory."

Deficient LTP Is Not Necessarily Accompanied by Deficient Memory

Despite the interpretive difficulties in the studies discussed above, one might still consider the overall data set as at least consistent with the supposition that hippocampal LTP is involved in the learning process. However, the convergence of evidence on a single viable hypothesis requires not only that a given data set be consistent with that hypothesis, but also that potentially disconfirming experiments be conducted (studies which could, in principle, either prove or disprove alternative hypotheses). For example, one might ask whether there are chemicals which are known to either block or enhance LTP which do not affect learning. Of the dozens of compounds shown to retard the induction of LTP in the hippocampus, only a few of these directly influence learning, and many others have no effect on learning. For example, saccharin is reported to block the induction of LTP in area CA1 of the hippocampus (Morishita et al. 1992), but chronic and acute administration induces no obvious memory deficits, and in some cases enhances retention (Stefurak & van der Kooy, 1992). Another study which suggested a dissociation between LTP and learning involved the gene deletion technique. Abelovich et al. (1993a) generated mice deficient in the gene for a subtype of the enzyme PKC. At the synaptic level, the mice displayed normal synaptic transmission, but most failed to develop hippocampal LTP. At the behavioral level, despite mild ataxia, the mice exhibited normal learning in the Morris water maze. That is, in the near absence of measurable LTP, learning and memory were not effected. In a subsequent article, some of the same authors (Abelovich et al. 1993b) presented evidence that LTP was not impaired when the synapses were first depressed by a low frequency stimulus train. This set of data did not include any behavior, and thus the evidence that deficient LTP was accompanied by normal learning was not addressed.

As with the studies using NMDA antagonists, we can either assume that perceptual and motor deficits in response to genetic manipulations were not adequately controlled and that they account for the observed effects on learning, or we can be satisfied that these sometimes obvious, and other times subtle deficits were not responsible for the learning deficits and that learning itself was affected. Once again, if we accept the latter position, then it is clear that blocking hippocampal LTP can impair, enhance, or have no effect on learning. Assuming for the time being that LTP is a memory mechanism, then we should consider the possibility that these disparate results reflect the fact that different tasks use different brain structures for memory formation.

Saturation of the Capacity for Plasticity

Another line of evidence linking hippocampal LTP to behavioral learning was correlational and followed logically from the NMDA blockade studies. The rationale for these studies was based on the premise that if synaptic potentiation was necessary for the formation of new memories, then artificially inducing hippocampal LTP at as many synapses as possible should effect subsequent acquisition of new memories. By 1989, this rationale had been systematically applied to two different learning tasks: classical conditioning and spatial maze learning. The effects of the prior induction were initially reported to be bi-directional and thus were reminiscent of the NMDA antagonist studies just discussed. In the first study, Berger (1984) reported that the repeated induction of unilateral LTP in the dentate gyrus over a five day period facilitated acquisition of a classically conditioned nictitating membrane response 24 hours later. Based on these results, Berger concluded that "the cellular mechanisms underlying LTP may be the basis for learning-induced changes in hippocampal unit activity during nictitating membrane conditioning." Since the hippocampus is not necessary for normal acquisition of the nictitating membrane response (McCormick et al. 1982; Berger & Orr, 1983), it is clear that if LTP is affecting learning at all, it is doing so indirectly and is not the sole or even primary mechanism underlying the storage of the learned response. These results are consistent with the finding that NMDA antagonists which block LTP can impair nonhippocampal-dependent learning (Shors and Servatius, 1995). They are also consistent with the idea, developed below (Section IV), that LTP is not a memory storage mechanism, but one which can modify effective acquisition of a learned response.

Instead of a facilitation in learning, however, one might have predicted that the induction of hippocampal LTP would saturate those synapses which are normally recruited during learning, and thus impair acquisition. This prediction was initially tested by McNaughton et al. (1986) and by Castro et al. (1989), using an experimental design conceptually similar to that used by Berger (1984). Animals were chronically implanted bilaterally with electrodes in the perforant path and the dentate gyrus. The perforant path was repeatedly tetanized over a 34 day period, inducing persistent LTP in the dentate gyrus. The rats were then trained on the circular platform task, which requires the animal to use spatial cues to find an escape hole in a circular board (McNaughton et al. 1986). They were also trained in the Morris water maze (Castro et al. 1989), the task used to demonstrate that the pharmacological blockade of hippocampal LTP blocks learning. In both tasks, and in contrast to classical conditioning (Berger 1984), these animals were impaired in their ability to learn the spatial task. Despite the inconsistency, these results were interpreted as indicative of a critical role for hippocampal LTP in memory formation. [It should be noted that LTP induction was more widespread (bilateral as opposed to unilateral) and extensive (34 days as opposed to 5 days) in the study by Castro et al. relative to that of Berger.]

These apparently contradictory results have been interpreted by some to reflect the two task's differential dependence on the hippocampus (e.g., Shors & Dryver 1992). For example, if hippocampal LTP was necessary for learning, as in the spatial task, then inducing it would impair learning; if the hippocampus was not necessary for the task, as in classical conditioning, then learning could be enhanced via some excitatory stimulation of the hippocampus. This idea had some plausibility; memory systems which are dependent on different neuroanatomical substrates can compete for behavioral control (e.g., McDonald & White 1993; 1995), suggesting the possibility that the induction of LTP might impair spatial learning (hippocampal dependent) while facilitating eyeblink conditioning (cerebellar dependent). However, this same logic could be used to suggest that hippocampal lesions (a more invasive analog of LTP-induced saturation) would also facilitate eyeblink conditioning, an effect that has not been observed (e.g., Solomon & Moore 1975). In any case, an explanation based on certain types of learning being dependent on different brain structures has some appeal, even though it is decidedly post hoc, and again raises the question of whether any experimental result could disprove the hypothesis that LTP is a memory mechanism. However, these concerns may have been unwarranted. Since the initial report of Castro et al. (1989), a number of laboratories, including the one in which the observation of Castro et al. originated, reported that tetanization of the perforant path does not impair spatial learning (Robinson 1992; Sutherland et al. 1993; see Bliss & Richter-Levin 1993, for review). In retrospect, it is perhaps not surprising that one cannot easily "saturate" the capacity for potentiation in the hippocampus; if so, one would have to question its adaptability. Taken to the extreme, it would suggest an easily attainable upper limit on memory capacity, something that has not been demonstrated experimentally (see Spear 1978, for discussion). Having acknowledged that saturation may be functionally difficult (Korol et al. 1993), Barnes et al. (1994) used stimulation parameters designed to more completely saturate the capacity for potentiation in the dentate gyrus. Under these conditions, there was no impairment of spatial learning in either the water maze or circular platform task, although there was a "deficit" on the fourth of five trial blocks during reversal training on the circular platform. However, even regarding this relatively minor impairment, there was no correlation between the magnitude of induced LTP and the behavioral deficit on that trial. If the potential for further LTP was indeed occluded, as suggested, these data provide strong evidence that hippocampal LTP is not necessary for spatial learning.

Correlations Between Modulators of LTP and Behavior

Correlational evidence can be powerful when an array of correlations lend support to a given hypothesis, rule out alternative hypotheses, and converge on a single viable conclusion (see Garner, Hake, & Erikson 1956). In the context of hippocampal LTP and memory formation, one could reasonably ask about the correlations between known modulators of LTP and the capacity to store new memories. For example, it is well-established that chronic lead or alcohol consumption is detrimental to memory storage, and they both impair the induction of LTP (e.g., Altman et al 1993; Morrisett & Swartzwelder 1993). To speculate that these compounds retard learning as a result of their effect on hippocampal LTP is misleading, given that there are other established mechanisms by which these substances could affect memory storage. For instance, alcohol, which after consumption has a wide distribution in brain tissue (as well as in the periphery), causes membrane fluidization, depresses both inhibitory and excitatory synaptic activity, and ultimately depresses activity in the cerebral cortex (for review, see Julian 1992). All of these effects disrupt normal CNS function and thus could disrupt the processing of information necessary for memory formation. Specifically, depression of activity in the cerebral cortex retards memory retrieval and storage (e.g., Horel 1993; Martin-Elkins, George, & Horel 1989), and thus a disruption of LTP by alcohol could be superfluous relative to a more gross deficit in cerebral activity. Until it is demonstrated that the alcohol-induced disruption of LTP impedes memory formation independent of its known effects on other processes, there is no reason to conclude that alcohol affects learning through its disruption of LTP. Correlations between manipulations that affect both hippocampal LTP and learning (e.g., anxiolytics; del Cerro et al. 1992; stress; Shors et al. 1990) have begun to pervade the literature and may erroneously reinforce the presumed link between hippocampal LTP and learning.

Natural Stimulation Patterns that Induce LTP

In an effort to more closely approximate endogenous conditions, many researchers study the relationship between hippocampal LTP and learning under more "natural" conditions such as olfactory discrimination. Olfaction is a primary sensory modality in the rat and olfactory information is processed, in part, by the hippocampal formation. Sensory input is relayed from the olfactory bulb to the piriform cortex, and separately to the hippocampus via entorhinal cortex. Lynch and his colleagues have found substantial correlations between hippocampal LTP in this system and olfactory learning. For example, when tetanic stimulation of the lateral olfactory tract was used as a discriminative cue, evoked responses in the piriform cortex were potentiated in the animals which learned the discrimination (Roman et al. 1987). In addition, the unit cellular activity recorded in the piriform cortex during learning was similar in pattern to the tetanic stimulation used for discrimination (McCollum et al. 1991). Animals exposed to the stimulation in a behaviorally irrelevant manner did not learn and did not exhibit LTP. These findings appear to be unique to the olfactory system since such stimulation would typically induce LTP in the hippocampus, regardless of whether the animals learned or not.

Others have established connections between LTP and learning by using tetanic stimulation as a sensory cue. In one study, stimulation of the perforant path (sufficient to induce LTP) was used as a CS to predict the occurrence of a footshock US (LaRoche 1989). Learning was measured as a suppression of lever pressing for food during the fear-evoking CS. Rats that learned the task exhibited a higher level of LTP after training than those who did not learn as well, leading to the interpretation that learning is accompanied by an increase in LTP (or at least that animals which learn most rapidly exhibit an increased capacity for LTP). This facilitation of learning was specific to LTP since it did not occur when 1) LTP was blocked by administration of an NMDA antagonist, 2) when the stimulation was below threshold for LTP induction, or 3) when the induction of LTP was inhibited by concomitant activation of commissural inputs. Using a similar procedure, it was reported that the decay rate of LTP correlated with the amount of forgetting, and that associative training resulted in an increased capacity for LTP 48 hours later (Bergis et al. 1990).

Although these results suggest a link between hippocampal LTP and learning, they are perhaps not surprising. There is a vast literature on brain stimulation and its use as a substitute for sensory stimuli. In particular, it has been repeatedly demonstrated that high-frequency electrical stimulation can be used as a discernible "cue" for the establishment of Pavlovian and other types of conditioning. The stimulation patterns typically used in these studies are very similar to those patterns initially described by Bliss and Lomo to induce LTP: 100 Hz stimulation for one second. The effectiveness of high-frequency stimulation is not limited to its use as a CS. It can also be used as the US; Vandercarr et al. (1970) reported that 100 Hz, 1 second stimulation delivered to the septum and the hypothalamus was an effective unconditioned stimulus (US) in heart rate conditioning and that the conditioned response was indistinguishable from that obtained with a US of peripheral shock (see also Salafia, Chiaia, & Ramirez 1979; Prokasy, Kesner & Calder 1983). It is difficult to evaluate the many studies using brain stimulation and the potential contribution of LTP to any effects on learning. This is, in part, because most studies using brain stimulation do not record electrical activity (much less LTP). However, given the numerous studies, brain sites, and experimental paradigms, it seems likely that there are many instances where brain stimulation mediates behavior (Salafia et al. 1977) without necessarily inducing hippocampal LTP. In conclusion of this point, one explanation for the enhanced learning observed after the induction of LTP is that brain stimulation serves as an effective and salient sensory cue.

One of the more convincing links between learning and hippocampal LTP involves the use of theta-frequency stimulation. As mentioned, when LTP was first described, the inducing stimulus consisted of a 100 Hz train of stimulation for one second (100 pulses total). The relevance of this type of stimulation to learning was questioned because this amount or frequency of activity rarely occurs in the brain. In the 1980's, a connection between a known brain rhythm and LTP was established by Larson, Wong & Lynch (1986) and Larson and Lynch (1988, 1989) and in a related manner by Buzaki, Haas & Anderson (1987), Rose & Dunwiddie (1986) and Greenstein et al. (1988). Patterned after the endogenous "theta" rhythm, one could effectively induce LTP extracellularly with short 100 Hz bursts delivered at 5 to 8 cycles per second (about 50 pulses total). These rhythms are naturally prominent in the hippocampus and fall into two types (Bland et al. 1984; Bland 1986; Kramis et al. 1975; Sainsbury et al. 1987): the first is dependent on motor activity and falls within a range of 8-11.9 Hz. The second type, is not dependent on movement and consists of a slightly lower frequency (Kramis et al. 1975). Further, the second type is dependent on the release of acetylcholine into the hippocampus from the septum (Bland et al. 1984).

Theta rhythms were once considered to be indicative of the learning process, but the consensus view today is that the first type of theta activity occurs in close correlation with concurrent voluntary motor activity, such as head turning, walking, running, forelimb movements, or changes in posture (Vanderwolf 1969; Fox et al. 1986; see Vanderwolf 1988, Vanderwolf & Cain 1994, for review). The second type can be induced without movement during exposure to arousing and stressful stimuli, such as to predators (Sainsbury et al. 1987), tail pinch (Stewart & Vanderwolf, 1987), water deprivation (Berry & Swain 1989; Maren et al. 1994), and tailshock stress (Shors et al. 1996). Neither type appears to be directly involved in memory formation.

With regard to how these rhythms become incorporated into the animal's perception of its environment, rats apparently sniff at a frequency comparable to theta rhythm, and the sniffs are time-locked to hippocampal theta activity (Maorides 1975). Even though theta activity is not observed during sniffing episodes when the animal is motionless (indicating that the rhythm is related to head movement; Vanderwolf 1988), some have suggested that animals processes information at a rhythm very similar to the most effective stimulus for inducing LTP. This correlation provides some behavioral relevance to hippocampal LTP and its induction parameters. Consistent with this view, theta activity in CA1 pyramidal cells occurs during a rat's sampling of a discriminative stimulus in an odor task (Otto et al. 1991). These general connections between theta activity and hippocampal LTP reinforce the presumed relevance of hippocampal LTP to behavior. Nonetheless, since theta occurs in many brain regions and across synapses in those regions, and is similarly associated with a broad range of behaviors, it is unclear how it could contribute to the specificity of synaptic changes presumed to underlie learning.

In addition to the possible association between theta activity and learning, there are also dissociations between theta activity and learning. Black (1975) trained dogs to press a lever in the presence of one stimulus, but to refrain from responding in the presence of a second stimulus, to avoid the onset of a shock. The animals successfully learned both responses, but theta activity was only observed during the active response. Moreover, dogs can be trained to elicit theta activity to terminate one auditory stimulus which signals an impending shock, and to refrain from displaying theta activity in response to a second stimulus which signals shock (Black et al. 1970). These experiments suggest that theta activity is not a prerequisite for the establishment of learning, but may correlate highly with the expression of an appropriate motor response. In summary, the connection between theta activity and LTP is promising in regard to LTP's functional relevance, but does not distinguish between an effect on motor performance versus one on memory storage.

Finally, we return to the observation that potentiation of an evoked response in the dentate gyrus accompanied acquisition of a classically conditioned eyeblink response (Weisz et al. 1984). We suggested that there was no evidence that this type of potentiation was related to the phenomenon of LTP (or at least that induced by high-frequency stimulation of the perforant path). Others apparently share this concern, and refer to these forms of potentiation as "postconditioning potentiation" (Weiss 1988) and "behavioral LTP" (Hargreaves et al. 1990). However, given the proposed link between theta activity and hippocampal LTP induction just discussed, these examples of behavioral LTP deserve further attention. The question is whether the increase in the evoked response following learning is a form of LTP? While no direct evidence is forthcoming, several pieces of evidence suggest that it is not. As mentioned, Robinson (1993) reported that NMDA antagonists which block LTP do not block the increase in the evoked response which accompanied learning. In addition, Krug et al. (1990) compared hippocampal LTP induced by high-frequency stimulation to the potentiation induced by avoidance learning. The potentiation induced by tetanic stimulation caused a decrease in the spike latency, whereas behavioral learning resulted in an increase in spike latency. Analogously, Hargreaves et al. (1990) found no evidence of potentiation in the dentate gyrus despite learning in a radial arm maze or a one-way avoidance task. Moreover, hippocampal LTP could be induced using traditional tetanization techniques following learning in each of these tasks. The results of both Krug et al. and Hargreaves et al. strongly suggest that the learning-induced increase in the evoked response and hippocampal LTP need not share common mechanisms, and that LTP is not the mechanism which underlies learning-induced potentiation.

Summary of Behavioral Evidence Thus Far

Based on the review of the behavioral literature so far, a number of general conclusions can be drawn. In summary, drugs or genetic manipulations that block hippocampal LTP impair performance in some tasks and facilitate performance in other tasks. The interpretation of these effects is confounded by the variability in brain structures necessary for successful completion of the task, the potential effects of the manipulations on sensory and motor performance, as well as the neuroanatomical deformities induced by the genetic manipulation. In addition, while a number of studies have found evidence that "LTP- like" increases in synaptic efficacy occur in the hippocampus during learning of tasks as diverse as spatial learning, associative eyeblink conditioning, conditioned suppression of activity, and olfactory discriminations, other studies find no evidence of potentiation despite robust learning. Still others find evidence of a form of potentiation that was qualitatively different than that which follows the induction of LTP, and artificial induction of LTP (i.e., saturation) has no clear effect on subsequent learning. Finally, while it is intriguing that tetanic stimulation mimicking an endogenous brain rhythm can induce hippocampal LTP, the rhythm is more generally associated with voluntary motor activity and arousal, rather than memory storage, per se, and is clearly not necessary for memory induction. Based on the data reviewed here, it does not appear that the induction of LTP is a necessary or sufficient condition for the storage of new memories.

A NEW AND NONSPECIFIC HYPOTHESIS

In questioning the role of LTP in learning, it is often said that LTP should be considered a memory mechanism until a mechanism is elucidated which is more consistent with our understanding of memory processes. Other candidate forms of plasticity have been described that, while differing in mechanistic detail, maintain a functional similarly, i.e., each would result in an increase in synaptic efficacy (see Hawkins et al. 1993). Whether these mechanisms should be considered unique categories of plasticity or simply subcategories under the larger classification of LTP is debatable (see Section IIB above). As discussed previously, we must be careful about the way in which we apply the term LTP to increases in synaptic efficacy. Most would agree, however, that memory formation involves the modification of synaptic transmission. To the extent that LTP is used only to refer to an enhancement of synaptic efficacy, some form of LTP somewhere in the nervous system is likely to contribute to memory formation. In this regard, we are unable to offer a "better" hypothesis. We have, however, generated several alternative hypotheses for how a mechanism like LTP could be used by the brain to accomplish behavioral objectives other than the storage of memory.

First, we must consider the most extreme alternative; that is, LTP serves no functional role and is an artificially-induced form of synaptic plasticity with no endogenous counterpart in the human brain. Although certainly possible, this alternative seems unlikely due to the fact that LTP-like increases in synaptic efficacy do occur naturally within the brain, as do patterns of activity similar to those used to induce LTP. Instead, we propose here an interpretation of the role of LTP that we feel is consistent with a majority of the data reviewed so far. Specifically, we propose that LTP is the neural equivalent of an arousal or attention device, and that it acts by increasing the gain of neural representations of environmental stimuli. If one were to assume that an environmental cue is represented in the brain as a synaptic response or pattern of responses, then the induction of an LTP-like phenomenon would magnify that response(s), allowing for more rapid detection of stimuli in general. Such an increase in gain (and consequent perceptual awareness) could then modify learning by increasing the likelihood that contingent relationships between stimuli are recognized. Thus, we are proposing that LTP is not a mechanism for memory storage and retrieval, but that it does play an incidental role in memory formation.

Like the discussion of LTP and memory, one must consider the functional relevance of such an arousal mechanism and whether it is consistent with what we know about how animals respond and survive in their natural environment. A time when such a mechanism would be particularly useful is during and after potentially life-threatening events, e.g., encounters with predators. Immediately after an encounter with a predator, the likelihood of another encounter is relatively high (the predator is in the vicinity and knows the prey's location). During this period of time, vigilance (alert watchfulness) should likewise be high in order to process environmental information in a timely and efficient manner. Presumably, a mechanism which is rapidly induced and heightens attentional processes could be critical for survival and would be highly selected for. There are a number of indications that the process that we commonly describe as attention is accompanied by a neural phenomenon which could either induce LTP or enhance the degree of potentiation. For instance, one of the few times that theta activity is observed during immobility is upon presentation of novel sensory stimuli (Kramis et al. 1975), such as a predator (cat or ferret; Sainsbury et al. 1987) or an acute and uncontrollable stressor (Shors et al. 1995). As such, we have postulated that exposure to an aversive and frightening event enhances endogenous theta activity and rapidly induces an LTP-like phenomenon, which then increases attention to environmental stimuli. Once the neural representation of the relevant stimuli are potentiated, responses and/or learning may be facilitated, but again, any facilitation is only incidental to the increase in stimulus processing. Depending on the aversiveness and potential threat of the experience, the potentiation and consequent attention to the environmental cues is maintained from hours to days, and thus represents a typical time course for the decay of LTP. When the chance of another attack or aversive event is once again low, the increase in synaptic efficacy returns to baseline levels, in preparation for subsequent events and to conserve resources.

Before describing our hypothesis in more detail, we should define our use of the terms attention, arousal and vigilance. First, it is recalled that attention is usually treated as a system that is external to and independent of memory, but one that can strongly influence the memory induction process (Norman & Shallice 1986; Posner & Petersen 1990). As a process, attention is divisible into several components (Posner & Pedersen 1990), a primary one being analogous to arousal. Arousal is an overall receptivity to stimuli and is considered the most general and nonspecific form of attention. It prepares the organism to deal with sensory information from multiple modalities and locations. Typically, however, the capacity to process multiple events simultaneously is limited. Thus, we also need a mechanism which allows us to shift attention as a function of what is novel and significant. This process is often referred to as selective attention, which focuses resources on the most critical information for further processing. Such selection can be based on general stimulus properties such as familiarity or particular properties such as its modality, intensity, or spatial location. When selective attention is maintained over time (hours to days), it constitutes a state of alert watchfulness, or vigilance. Here we hypothesize that exposure to novel and/or fearful events induces a form of plasticity similar in mechanism and function to LTP. Such a mechanism could potentially strengthen the neural representation of sensory stimuli, effectively increasing the attention devoted to them (i.e., induction of a state of arousal). Once the relevant stimuli are identified and selected (selective attention), a more sustained state of vigilance may develop, allowing attention to be maintained for an extended period of time.

At least since the empirical demonstrations of Pavlov, it has been known that increasing the intensity of a cue will enhance learning when the cue is relevant. What we are proposing here is that events which arouse an animal's attention do so by inducing a potentiation of neuronal responses which has the functional effect of increasing the intensity of impinging stimuli. Accordingly, manipulations which increase arousal or attention should incidentally facilitate learning when it is dependent on the processing of particular environmental cues. Indeed, Spitzer, Desimone & Moran (1988) reported that increased attention enhances a monkey's neuronal responses to the discriminated stimulus and enhanced the monkey's discriminative ability. Also, in response to an aversive (and presumably arousing) experience of inescapable tailshock, rats become sensitized to discrete sensory cues and independently learn an associative eyeblink response at a facilitated rate (Shors et al. 1992; Servatius & Shors 1994). Like LTP, the effects of the arousing experience on learning can be long-lasting, sometimes persisting for 48 hours after the event has terminated.

As discussed, LTP possesses a number of properties considered desirable in a memory storage device, and we reviewed a number of them. For instance, a memory mechanism should reflect properties central to the formation and storage of memories, i.e., should be rapidly induced, long-lasting, and inducible through natural stimulation patterns. If LTP is not a memory storage mechanism, but rather an attentional device, then its properties should be likewise consistent with those of attention. Therefore, now we review the various characteristics of LTP which are (or are not) consistent with its proposed role as an attention device in the mammalian brain.

Long-Lasting but Decremental and Facilitated Reacquisition

In terms of the temporal characteristics of LTP, they appear to be more consistent with the typical time-course of attention than that of memory storage, per se. In terms of induction, both attention and memory must be rapidly induced. However, in terms of maintenance, LTP decays more rapidly than does memory, a relatively stable process. The time course of LTP does coincide with that of an attentional process which should persist from hours to days, but eventually return to baseline.

It was noted previously that, unlike memories, LTP does not exhibit facilitated reacquisition; following decay, a second induction of LTP is not more easily accomplished than the initial induction. The presence of facilitated acquisition, however, is not a necessary component for the increase in attention or vigilance that occurs in response to an arousing event. To be most effective, it should activate and subsequently inactivate to prepare for exposure to subsequent arousing events. Facilitated reactivation would serve no obvious functional value in an attentional device.

Deficient LTP is Not Necessarily Accompanied by Deficient Memory

In every instance that we are aware, pharmacological, genetic, and neurophysiological manipulations which eliminate the capacity for LTP induction only affect the rate at which some tasks are learned but do not block learning entirely. Such results, though inconsistent with LTP's role as a memory storage device, are consistent with its role in attentional processes associated with effective memory formation. If indeed LTP serves to enhance the effective processing of sensory stimuli, one would expect that the administration of NMDA antagonists which prevent LTP would cause a greater learning deficit when the cues are of low salience. Indeed, Staubli et al. (1989) reported that NMDA antagonists only impaired olfactory learning when the olfactory cues were delivered at low intensities, but did not impair learning when more intense stimuli were used. Also, in the Morris water maze, rats injected with the antagonist are not impaired (or are minimally impaired) in their ability to find the platform when they can either visualize the platform or when they have previously learned the procedures of the task (Saucier & Cain 1995; Bannerman et al. 1995). Thus, it appears that the more subtle aspects of sensory processing are impaired by blocking LTP, not the ability to form new memories. [It is noted that this effect is not universal among tasks: in eyeblink conditioning, NMDA antagonists impaired learning whether the stimuli were delivered at a normal or high intensity (Servatius and Shors, in press)].

Dependence on NMDA Receptor Activation

If one of the factors which can induce LTP is exposure to an aversive and arousing experience, then the persistent behavioral consequences should be prevented by blocking LTP induction during the experience. In a procedural conditioning task, injection of an NMDA antagonist just prior to exposure to an aversive and arousing experience prevented the sensitization to explicit sensory cues and the facilitated learning that normally occurs 24 hours later. However, the antagonist had no effect on normal learning (Shors & Servatius 1994). [As discussed, a similar antagonist with a longer half-life and delivered at a higher dose does impair learning (Servatius & Shors, in press)]. In another task that requires a high degree of vigilance and attention for its completion, differential reinforcement of a low response rate (DRL) performance was severely impaired by NMDA receptor antagonists (Tonkiss et al. 1988). Importantly, the decrease in DRL performance occurred even after the task was well learned, mitigating against the possibility that the NMDA antagonist was interfering with memory storage. These results distinguish between LTP as a mechanism underlying the induction of the neural memory trace and as a modifier of another distinct process that may accompany and influence memory formation. Consequently, we would suggest that the induction of LTP prior to or during early training might be pivotal in determining how fast and the degree to which a particular event is stored in memory.

Correlations between Modulators of Hippocampal LTP and Behavior

If LTP were being used as neural mechanism of arousal or attention rather than memory formation, manipulations intended to increase arousal and vigilance should have effects similar to those of LTP induction. Indeed, exposure to an aversive and stressful event produces biochemical, electrophysiological, and molecular effects that are similar, and in some cases, identical to those following high-frequency stimulation. For example, an increase in binding affinity of the AMPA type of glutamate receptor occurs in the hippocampus in response to the induction of LTP. The increase in binding is virtually indistinguishable from that induced in response to the acute inescapable stressor of restraint and intermittent tailshocks (Tocco et al. 1991, 1992). In addition, a previous induction of LTP and exposure to an aversive and arousing event both impair subsequent LTP induction (Foy et al. 1987; Shors et al. 1989; Shors & Thompson 1992; Shors & Dryver 1994) and enhance the subsequent extracellular response to theta burst stimulation (Shors & Dryver 1994; Shors, Gallegos, & Breindl, submitted). As discussed, exposure to a stressful and arousing event facilitates acquisition of the classically conditioned eyeblink response (Shors et al. 1992; Servatius & Shors 1994) as does the induction of LTP in the dentate gyrus (Berger 1984; see Rioux & Robinson 1995, for exception). Other examples of common targets for LTP induction and the arousing experience include the immediate early genes zif-268 and c-fos in the hippocampus (Cole et al. 1989; Kaczmarek 1992ab; Abraham, Mason, Demmer, Williams, Richardson, Tate, Lawlor, & Dragunow 1993; for exception see Schreiber et al. 1991) and nitric oxide activity (Schuman & Madison 1991; Sadile & Papa, 1993).

When an animal can attain control or perceived control over an aversive experience, the degree of arousal is substantially decreased (Overmier & Seligman 1967; see Seligman & Johnston 1973, for review). If LTP was involved in arousal and attentional processes, then manipulating the degree of arousal should, in turn, alter the amount of potentiation that is induced. Allowing an animal to acquire control over the fearful experience partially ameliorates the impairment of LTP induction in area CA1 of the hippocampus (Shors et al. 1989). Furthermore, a suboptimal induction of LTP can be optimized by exposing the rat to an aversive and uncontrollable stressor such as footshock during exposure to the tetanus (Seidenbecher, Balschun, & Reymann 1995).

Overall, the number of similarities between the effects of an arousing experience and the induction of LTP suggest, at the very least, a convergence on similar neuronal mechanisms, and at the most, that a phenomenon similar in mechanism to LTP is being induced after an animal has experienced a threatening event.

Saturation of the Capacity for Plasticity

In the absence of preconceptions regarding the role of LTP in memory storage, other potential roles for LTP become evident, including the role in attention we propose. In fact, the evidence in support of LTP's role in attention may be as strong as for its role in memory storage. For example, artificial induction of LTP had no effect on subsequent acquisition of water maze learning, but there was a strong positive correlation between the rat's capacity for LTP and speed with which they escaped the aversive environment (Jeffery & Morris 1993; but see Cain et al. 1993). Moreover, rats that were most susceptible to LTP induction spent more time in the quadrant of the maze that previously contained the escape platform when the platform was removed, a common measure of retention. One interpretation of this later effect is that LTP induction alters performance variables such as perseveration, rather than impinging directly on memory formation. The induction of perseverative behavior is common when an animal is in a fearful and attentive state. Following exposure to a fearful event, acquisition of a spatial maze task is impeded by the rat's tendency to perseverate prior to the initiation of a response (Shors & Dryver 1992). At first approximation, these findings may appear inconsistent with our contention that an arousing experience induces LTP and directs attention to sensory stimuli during learning. However, they are not necessarily inconsistent if one recognizes that perseveration (focused attention) might degrade performance (but not necessarily learning) in a task requiring complex, diverse, motor responses in an environment containing both relevant and irrelevant cues (e.g., the radial arm maze), while facilitating acquisition of simple reflex behavior in a task in which irrelevant cues are minimized (e.g., eyeblink conditioning). As discussed, Berger (1984) reported that the previous induction of LTP

in the hippocampus facilitated eyeblink conditioning to a discrete CS. The task was a standard delay task which is not dependent on the hippocampus for acquisition (Weisz et al. 1984), and thus LTP could not be forming the memories themselves. It could, however, enhance the general saliency and perceived intensity of the limited cues presented in the environment, an effect that is known to facilitate acquisition of the conditioned response (Scavio & Gormezano 1974).

Natural Stimulation Patterns that Induce LTP

The possibility that hippocampal LTP is a mechanism for the neural induction of attention or arousal is further supported by the observation that afferent stimulation that mimics the endogenous theta rhythm is an extremely effective stimulation pattern for inducing LTP (Larson et al. 1986; Greenstein et al. 1988). As mentioned, endogenous theta rhythms are associated with voluntary motor activity, but can be induced by exposure to aversive events, even in the absence of an overt motor response. This is not simply an effect of subthreshold (i.e., undetected) muscular exertion, because muscular effort alone is not sufficient to induce theta, e.g., hanging by the forepaws or balancing motionless on the hind limbs does not induce theta rhythm in the rat (Vanderwolf, 1988). Because theta activity correlates highly (and may actually precede) motor behaviors associated with exploration as well as exposure to aversive stimuli (for exception, see Balleine & Curthoys 1991), it seems parsimonious to suggest that it may be an index of arousal. Consistent with such a view, it has been reported that the amount of rhythm can predict the rate of acquisition in a simple associative task (Berry & Thompson 1978). Using a natural inducer of theta activity, water deprivation, Berry and Swain (1989) found that the rate of learning was accelerated. In addition, water deprivation was reported to enhance the induction of LTP, and concurrently fear conditioning to context (Maren et al. 1994).

A prevalent form of theta activity is generated via a cholinergic output from the septum which projects to the hippocampus and elsewhere (Bland et al. 1984; Bland, 1986). The mechanism whereby fearful experiences enhance attention could be initiated by acetylcholine which is released into the hippocampus during the experience (Mark, Rada & Shors, in press). Further, the sensitization to discrete sensory cues produced by an aversive experience is prevented by blocking cholinergic receptors during the aversive experience (Shors et al. 1995). Importantly, if we accept the proposal that theta stimulation induces an LTP-like phenomenon endogenously, then such an induction protocol would most likely have widespread effects. Such widespread effects might be necessary, at least initially, in order to enhance processing of all stimuli. Perhaps these effects reflect the nonspecific responses to LTP induction described earlier, i.e. the increase in mRNA for the glutamate receptor and neurotrophins previously discussed. Some of these effects may likewise coincide with the widespread increase in neuronal activity that occurs in many brain regions during learning.

Synaptic Efficacy and Specificity

The notion that LTP was a substrate of learning and memory arose, in part, from the proposition that enhancing synaptic transmission would be an effective way to "construct" and retrieve specific memories in a neuronal network (e.g., Hebb 1949; James 1892; Spencer 1870; Tanzi 1893). If this idea were valid, one would expect that enhancing synaptic transmission would enhance memory formation. Recently, this hypothesis was tested with a class of drugs known as ampakines. These drugs enhance synaptic transmission by increasing the mean open time of the AMPA type of glutamate receptor. As predicted, the drugs facilitated learning in a number of tasks, from olfactory discrimination to maze learning (Staubli et al. 1994; 1995). One of the questions that arises is whether the drugs facilitate memory formation itself or the processing of sensory information prior to memory storage? We tested one of them in a classical conditioning task and found that rats treated with the drug displayed an enhanced responsiveness to discrete sensory cues. When the sensitization to cues was removed (by lowering the intensity of the CS), facilitated acquisition of conditioned eyeblink response occurred (Shors et al. 1995). It is noted that the drug was injected peripherally and thus, the increase in synaptic transmission occurred throughout the brain (Staubli et al. 1995). Thus, the drug could not by itself form the "memory" because it would not affect specific synapses; it could nonetheless prime the network such that subsequent memories are more easily induced. These results are parsimonious with the idea that enhancing synaptic neurotransmission (inducing and LTP-like phenomenon) enhances the neural representation of cues in the brain (even if below threshold for a sensitized behavioral response) which incidentally enhances learning when the cues are relevant. Consistent with the more general assumption that the efficacy of synaptic transmission should directly influence the rate of learning, Matzel et al. (in press) have reported a strong correlation between the strength of the synaptic integration between two sensory systems and the capacity to form an association between stimuli presented in those sensory modalities. Moreover, Matzel et al. reported that poor learning (which correlated with weak synaptic transmission) could be facilitated simply by increasing the intensity of the sensory stimuli (and thus the amplitude of synaptic potentials).

As previously discusses, LTP is often described as "synapse specific", but in actuality, it is not confined to synapses and spreads to neighboring synapses. The hypothesis proposed here is not dependent on the limitation of LTP to specific synapses. Many responses to fearful stimuli are nonspecific in the sense that they are not confined to one or a set of synapses (Schreiber et. al., 1991; Cullinan et. al., 1995). At a functional level, fear responses are typically induced by a single event (or series) of events, but generalize to subsequent events (Overmier & Seligman 1967; Shors et al. 1992). As one might expect, changes in brain activity associated with attention are initially broad, but narrow considerably as attention becomes focused on the relevant cues. Therefore, as attention is maintained, there is a shift in brain activity from the regions of irrelevance to those necessary for processing of specific stimuli. For example, there is an increase in activity in visual areas when processing a visual cue in a task, and a corresponding decrease in areas such as the auditory cortex that are not involved in the task (Haxby et al. 1994; for review Haxby et al. 1991; Greenwood et al. 1993). One could imagine that a brain-ubiquitous mechanism like LTP could be quite useful as an attention device by initially enhancing the general salience of cues, followed by a more focused attention on relevant stimuli and their associations.

Distribution throughout the Nervous System

A presumed role for LTP in attention might even explain its existence in such a diverse set of brain regions. For example, LTP in the superior colliculus could be involved in directing spatial attention through its involvement with saccadic eye movements (Sheliga et al. 1995). A role for LTP in attention is also consistent with its prominence in the hippocampus, which some consider an on-line processor of incoming sensory information acquired through attentional processes (Mackintosh 1975; Grossberg 1975; Moore & Stickney 1980; Schmajuk 1990; Schmajuk & DiCarlo 1991) and/or exploration (Buzsaki & Czeh, 1992). LTP would be particularly effective in this context when a high state of vigilance has been induced but the relevance of the information has yet to be established. The hypothesis is also consistent with the relatively nonspecific increase in cell excitability observed in the hippocampus during acquisition in a variety of tasks, some of which are clearly not dependent on the hippocampus for memory storage (Berger et al. 1976; McCormick & Thompson 1984). In fact, LTP's wide distribution throughout the nervous system is perhaps one of the most convincing aspects of its proposed role in attention.

The hypothesis presented here is less "specific" than the prevailing alternative. First, it is not constrained by the preconception that LTP serves as a mechanism for synapse-specific modifications underlying memory. The hypothesis does, however, predict that there should be enhanced synaptic efficacy following exposure to aversive and arousing stimuli. In the only direct test of this hypothesis to date, synaptic responses in area CA1 of the hippocampus were not potentiated in response to the aversive event of acute inescapable tailshocks, in apparent disagreement with our predictions (Shors, Gallegos, & Breindl, submitted). However, in another study, synaptic responses elicited by a discrete auditory stimulus were increased following exposure to tetanic stimulation. This synaptic potentiation was observed in a pathway from the thalamus to the amygdala, one that is thought to be involved in fear conditioning (Rogan & LeDoux 1995). These results suggest that LTP-like phenomena can increase the cellular processing of sensory stimuli and perhaps can be induced by previous exposure to fearful experience. As described in the above hypothesis, the increase in synaptic efficacy would initially facilitate sensory processing, and might incidentally impact on future learning. (It is recognized that these data reflect phenomena occurring in brain structures other than the hippocampus, and thus it could be the case that LTP is used for different purposes in different brain structures. This is a difficult position to refute, but there is no a priori reason to accept it without evidence to substantiate LTP's role in memory or any other cognitive process).

In summary, we proposed that when a stressful and arousing stimulus is encountered, a phenomenon mechanistically similar to LTP is induced, functionally increasing the neuronal representation of environmental stimuli. To the extent that those stimuli require storage in memory, that storage may be incidentally facilitated. Such a transition from nonspecificity to specificity is, in our view, a necessary criterion for a system that must be constantly prepared to respond to environmental change without prior knowledge of its significance.

CONCLUDING COMMENTS

In the vast spectrum of experimental results, support for virtually any hypothesis regarding the role of LTP in memory can be found. This may not be surprising, given the prominent role that LTP plays in the hippocampal plasticity, and the debate that has raged for five decades concerning the role of the hippocampus in learning and memory storage (e.g., Cho & Kesner 1995; Bunsey & Eichenbaum 1996; Nadel 1995; McClelland et al. 1995). Here, we have suggested that much of the support for the connection between LTP and memory arose from a preconception that LTP is a learning mechanism. Consequently, the hypothesis that LTP plays a critical role in memory storage has gained a conceptual hold on the field which limits both our capacity to critically evaluate data or to recognize alternative hypotheses as the data suggests them. The task is further complicated by the fact that many forms of plasticity fall under the category of "LTP", yet "LTP" is treated as a unitary phenomenon with respect to its role in behavior.

We recognize that the alternative hypothesis that we have proposed is subject to many of the same criticisms as the prevailing hypothesis that we are critical of, and will ultimately prove as difficult to test. It does, at the very least, pose "how" a mechanism like LTP might be useful to an awake behaving animal under native circumstances. Moreover, it suggests the obvious conclusion that experiments that have purported to establish a link between LTP and memory are subject to multiple, and even equally viable, interpretations.

In the end, we want to again make the distinction between the various roles that LTP might play in the awake behaving animal. The most extreme possibility is that LTP is neither an information-processing device nor a memory mechanism. A less extreme alternative is that LTP plays a critical role in the processing of sensory information necessary for the establishment of stable memories and that it is induced in response to environmental stimuli as well as the organism's response to those stimuli. Information-processing has costs as well as its obvious adaptive benefits, and an efficient brain should employ mechanisms that allocate and reallocate resources depending on current demands, past experiences, and anticipated events. Due to its pervasive and ubiquitous nature, this may be a role particularly suited for LTP. With respect to learning, the lack of conclusive evidence indicating a necessary contribution of LTP to memory induction per se, as well as recent evidence to the contrary, should be sufficient cause for us to stop and reevaluate the conceptual hold that this hypothesis has on current thinking.

ACKNOWLEDGMENTS

This work was supported by the National Science Foundation (IBN-9511027), the Office of Naval Research (N00014-92-J-1897), McDonnel-Pew Program in Cognitive Neuroscience, and the Whitehall Foundation (to TJS) and the US Public Health Service (NIMH Grants MH48387 and MH52314), the Charles and Johanna Busch Memorial Fund, and a Hoechst Celanese Young Investigator Award (to LDM) We thank Gregory Clark, Ernest Greene, Pramod Mathew, Carol Myers, Tim Otto, Kevin Pang, Richard Thompson, and Thomas Walsh for their comments on an earlier version of the manuscript. This article was originally submitted for review in April, 1994. Correspondence may be addressed to either author.

REFERENCES

Abeliovich, A., Chen, C., Goda, Y., Paylor, R., McKee, A., Shuster, L., Wehner, J., Stevens, C.F., & Tonegawa, S. (1993a) PKCg-mutant mice are deficient in hippocampal LTP and display intact spatial learning in the Morris water maze. Society for Neuroscience Abstracts 19: 411.

Abeliovich, A., Chen, C., Goda, Y., Silva, A.J., Stevens, C.F., & Tonegawa, S. (1993b) Modified hippocampal long-term potentiation in PKCg-mutant mice. Cell, 75:1253-1262

Abraham, W.C., Mason, S.E., Demmer, J., Williams, J.M., Richardson, C.L., Tate, W.P., Lawlor, P.A., & Dragunow, M. (1993) Correlations between immediate early gene induction and the persistence of long-term potentiation. Neuroscience 56: 717-727.

Abrams, T.W. (1985) Activity-dependent presynaptic facilitation: An associative mechanism in Aplysia. Cellular and Molecular Neurobiology 5: 123-145.

Abrams, T.W., Kandel, E.R. (1988) Is contiguity detection in classical conditioning a system or cellular property: learning in Aplysia suggests a possible molecular site. Trends in Neuroscience 11: 128-135.

Abrams, T.W., Karl, K.A., Kandel, E.R. (1991) Biochemical studies of stimulus convergence during classical conditioning: amplification of adenylate cyclase by Ca2+/calmodulin and transmitter. Journal of Neuroscience 11: 1655-2665,

Akers, R.F., Lovinger, D.M., Colley, P.A., Linden, D.J., & Routtenberg, A. (1986) Translocation of protein kinase C activity may mediate hippocampal long-term potentiation. Science 231: 587- 589.

Alger, B.E. & Tyler, T.J. (1976) Long-term and short-term plasticity in the CA1, CA3, and dentate regions of the rat hippocampal slice. Brain Research 110: 463-480.

Altmann, L., Weinsberg, F., Sveinsson, K., & Liliental, H. (1993) Impairment of long-term potentiation and learning following chronic lead exposure. Toxicology Letters 66: 105- 112.

Andersen, P., Sundberg, S.H., Sveen, O., & Wigstrom, H (1977) Specific long-lasting potentiation of synaptic transmission in hippocampal slices. Nature 266: 736-737.

Anderson, P., Sundberg, S.H., Sveen, O., Swann, J.W., Wigstrom, H. (1980) Possible mechanisms for long-lasting potentiation of synaptic transmission in hippocampal slices from guinea pigs. Journal of Physiology, 302: 463-482.

Atkinson, & Shiffrin (1968) Human memory: A proposed system and its control processes. The Psychology of Learning and Motivation, Vol. 2. New York: Academic Press.

Balleine, B. W., and Curthoys, I. S. (1991) Differential effects of escapable and inescapable footshock on hippocampal theta activity. Behavioral Neuroscience 105: 202-209.

Bannerman, D.M., Chapman, P.F., Kelly, P.A.T., Butcher, S.P., & Morris, R.G.M. (1994) Inhibition of nitric oxide synthase does not impair spatial learning. Journal of Neuroscience, 14: 7404- 7414.

Bannerman, D.M., Good M.A., Butcher S.P., Ramsay M., and Morris R.G.M. (1995) Distinct components of spatial learning revealed by prior training and NMDA receptor blockade. Nature 378: 182-186

Baudry, M., Arst, O., Oliver, M., Lynch, G. (1981) Development of glutamate binding sites and their regulation by calcium in rat hippocampus. Developmental Brain Research 1: 37-48.

Barnes, C.A., Jung, M.W., McNaughton, B.L., Korol, D., Andreasson, K., Worley, P.F. (1994) LTP saturation and spatial learning disruption: effects of task variables and saturation levels. Journal of Neuroscience 14, 5793-5806.

Baringa, M. (1994) Learning by diffusion: Nitric oxide may spread memories. Science 263: 466.

Barrionuevo, G. & Brown, T.H. (1983) Associative long-term potentiation in hippocampal slices. Proceedings of the National Academy of Sciences (U.S.A.) 80: 7347-7251.

Bennet, & Amrich, (1986) 2-amino-7-phosphonheptanoic acid (AP7) produces discriminative stimuli and anticonflict effects similar to diazepam. Life Sciences 39: 2455-2461.

Berger, T.W., Alger, B.E., & Thompson R.F. (1976) Neuronal substrates of classical conditioning in the hippocampus. Science 192: 483-485.

Berger, T.W., Orr, W.B. (1983) Hippocampectomy selectively disrupts discrimination reversal conditioning of the rabbit nictitating membrane response. Behavioral Brain Research 8: 49-68.

Berger, T.W. (1984) Long-term potentiation of hippocampal synaptic transmission affects rate of behavioral learning. Science 224: 627-630.

Bergis, O.E., Bloch, V., & Laroche, S. (1990) Enhancement of long-term potentiation in the dentate gyrus two days after associative learning in the rat. Neuroscience Research Communications 6: 119-128.

Berry, S.D. & Swain, R.A. (1989) Water deprivation optimizes hippocampal activity and facilitates nictitating membrane conditioning. Behavioral. Neuroscience 103: 71-76.

Berthier, N.E. & Moore, J.W. (1986) Cerebellar Purkinje cell activity related to the classically conditioned nictitating membrane response. Experimental Brain Research 63: 341-350.

Black, A.H. (1975) In The Hippocampus, Vol. 2: Neurophysiology and Behavior (R.L. Isaacson & K.H. Pribram, Eds.). Plenum: New York.

Black, A.H., Young, G.A., & Batenchuck, C. (1970) Avoidance training of hippocampal theta waves in flaxedilized dogs and its relation to skeletal movement. Journal of Comparative and Physiological Psychology 70: 15-24

Bland, B. H., Seto, M. G., Sinclair, B. R., & Fraser, S. M. (1984) The pharmacology of hippocampal theta cells: evidence that the sensory processing correlate is cholinergic. Brain Research 299: 121-131.

Bland, B. H. (1986) The physiology and pharmacology of hippocampal formation theta rhythms. Progress in Neurobiology 26: 1-54.

Bliss, T.V.P., Chavez-Noriega, L.E., Halliwell, J.V., (1987) Long-term potentiation is associated with an increase in excitability of pyramidal cells in area CA1 of the hippocampal slice. Journal of Physiology 390: 260P.

Bliss, T.V.P. & Gardner-Medwin, A.R. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the unanesthetized rabbit following stimulation of the perforant path. Journal of Physiology (London) 232: 357-374.

Bliss, T.V.P. & Lomo, T. (1973) Long-lasting potentiation of synaptic transmission in the dentate area of the anesthetized rabbit following stimulation of the perforant path. Journal of Physiology (London) 232: 331-356.

Bliss, T.V.P. & Richter-Levin, G. (1993) Spatial learning and saturation of long-term potentiation. Hippocampus 3: 123-126.

Boenhoffer, T., Staiger, V., & Aertsen, A. (1989) Synaptic plasticity in rat hippocampal slice cultures: Local "Hebbian" conjunction of pre- and postsynaptic stimulation leads to distributed synaptic enhancement. Proceedings of the National Academy of Sciences (U.S.A.), 86: 8113-8117.

Bohme, G.A., Bon, C., Stutzmann, J.M., Doble, A., Blanchard, J.C. (1991) Possible involvement of nitric oxide in long-term potentiation. European Journal of Pharmacology 199: 379.

Bortolotto, Z.A., Bashir, Z.I., Davies, C.H., Taira,T., Kaila, K., Collingridge, G.L. (1995) Studies on the role of metabotropic glutamate receptors in long-term potentiation: Some methodological considerations. Journal of Neuroscience Methods 59: 19-24.

Brown, T.H. & McAfee, D.A. (1982) Long-term synaptic potentiation in the superior cervical ganglion. Science 215: 1411-1413.

Brown, T.H, Kairiss, E.W., & Keenan, C.L. (1990) Hebbian synapses: Biophysical mechanisms and algorithms. Annual Review of Neuroscience 13: 475-511.

Bunsey, M. & Eichenbaum, H. (1996) Conservation of hippocampal memory function in rats and humans. Nature 379: 255-257.

Buzsaki, G., Haas, H.L., & Anderson, E.G. (1987) Long-term potentiation induced by physiologically relevant stimulus patterns. Brain Research 435: 331-333.

Buzsaki, G. & Czeh, G. (1992) Physiological function of granule cells: A hypothesis. Epilepsy Research (Supplement) 7: 281-290.

Cain, D.P., Hargreaves, F., Boon, F., & Dennison, Z. (1993) An examination of the relations between hippocampal long-term potentiation, kindling, afterdischarge, and place learning in the watermaze. Hippocampus 3: 153-164.

Cain, D.P., Saucier, D., Hall, J., Hargreaves, E.L., & Boon, F. (1996) Detailed behavioral analysis of water maze acquisition under APV or CNQX: Contribution of sensory-motor disturbances to drug-induced acquisition deficits. Behavioral Neuroscience.

Campeau, S., Miserendino, M.J., Davis, M. (1992) Intra-amygdala infusion of the N-methyl-D-aspartate receptor antagonist AP5 blocks acquisition but not expression of fear-potentiated startle to a conditioned stimulus. Behavioral Neuroscience 106: 569-574.

Caramanos, Z., Shapiro, M.L. (1994) Spatial memory and N-methyl-d-aspartate receptor antagonists APV and MK-801: memory impairments depend on familiarity with the environment, drug dose, and training duration. Behavioral Neuroscience 108: 30-43.

Castillo, P.E., Weiskopf, M.G., & Nicoll, R.A. (1994) The role of Ca2+ channels in hippocampal mossy fiber synaptic transmission and long-term potentiation. Neuron 12: 261-269.

Castren, E., Pitkanen, M., Sirvio, J., Parsadanian, A., Lindholm, D., Thoenen, H., Riekkinen, P.J. (1993) The induction of LTP increases BDNF and NGF mRNA but decreases NT-3 mRNA in dentate gyrus. Neuroreport 4: 895-898.

Castro, C.A., Silbert, B.L., McNaughton, E.L., Barnes, C.A. (1989) Recovery of spatial learning deficits after decay of electrically induced synaptic enhancement in the hippocampus. Nature 342: 545- 548.

Chapouthier, G. (1989) The search for a biochemistry of memory. Archives of Gerontology Geriatric Supplement 1: 7-19.

Cho, Y.H., & Kesner, R.P. (1995) Relational object association learning in rats with hippocampal lesions. Behavioral Brain Research 67: 91-98.

Clark, G.A., & Kandel, E.R. (1993) Induction of long-term facilitation in Aplysia sensory neurons by local application of serotonin to remote synapses. Proceedings of the National Academy of Sciences (U.S.A.) 90: 11411-11415.

Clements, M.P., Bliss, T.V., & Lynch, M.A. (1991) Increase in arachidonic acid concentration in a postsynaptic membrane fraction following the induction of long-term potentiation in the dentate gyrus. Neuroscience 45: 379-389.

Clineschmnidt, B.V., Martin, G.E., & Bunting, P.R. (1982) Anticonvulsant activity of (+)-5-methyl- 10,11-dihydro-5H-dibenzo[a,d]cyclo-hepten-5,10-imine maleate (MK-801), a substance with potent anticonvulsant central sympathomimetic, and apparent anxiolytic properties. Drug Development Research 2: 123-124.

Cole, A.J., Saffen, D.W., Baraban, J.M., & Worley, P.F. (1989) Rapid increases of an immediate early gene messenger RNA in hippocampal neurons by synaptic NMDA receptor activation. Nature 340: 474-476.

Collingridge, G.L., Kehl, S.J., McLennan, H. (1983) Excitatory amino acids in synaptic transmission in the Schaeffer-commissural pathway of the rat hippocampus. Journal of Physiology (London) 443: 33-46.

Cummings, J.A., Nicola, S.M., & Malenka, R.C. (1994) Induction in the rat hippocampus of long-term potentiation (LTP) and long-term depression (LTD) in the presence of a nitric oxide synthase inhibitor. Neuroscience Letters 176: 110-114.

Dale, N. (1989) The role of NMDA receptors in synaptic integration and the organization of complex neural patterns. In J.C. Watkins and G.L. Collingridge (Eds.), The NMDA Receptor (pp 93-108) Oxford, U.K.: IRL Press.

Dale, N., Roberts, A. (1985) Dual-component amino-acid-mediated synaptic potentials: excitatory drive for swimming in Xenopus embryos. Journal of Physiology (London) 363: 35-59.

Davis, S., Butcher, S.P., & Morris, R.G.M. (1992) The NMDA receptor antagonist D-2-amino-5-phosphonopentanoate (D-AP5) impairs spatial learning and LTP in vivo at intracerebral concentrations comparable to those that block LTP in vitro. Journal of Neuroscience 12: 21-34.

del Cerro, S., Jung M., and Lynch G. (1992) Benzodiazepines block long-term potentiation in rat hippocampal and piriform cortex slices. Neuroscience 49: 1-6.

de Jong, M.& Racine, R.J. (1985) The effects of repeated induction of long-term potentiation in the dentate gyrus. Brain Research 328: 181-185.

Deutsch, J.A. (1993) Spatial learning in mutant mice. Science 262: 760-761.

Diamond, D.M., & Rose, G.M. (1994) Does associative LTP underlie classical conditioning? Psychobiology, 22: 263-269.

Duffy, C.J. & Teyler, T.J. (1978) Development of potentiation in the dentate gyrus of the rat. Brain Research Bulletin, 3: 425-430.

Dunwiddie, T. & Lynch, G. (1978) Long-term potentiation and depression of synaptic responses in the hippocampus: Localization and frequency dependency. Journal of Physiology, 276: 353-361.

Eichenbaum, H., and Otto, T. (1993) LTP and memory: can we enhance the connection? Trends in Neuroscience 16: 163-164.

Eisenstein, E.M., & Reep, R.L. (1985) Behavioral and cellular studies of learning and memory in insects. In G.A. Kerkut & L.I. Gilbert (Eds.), Comprehensive insect physiology, biochemistry, and pharmacology, vol 9. Oxford: Pergamon Press.

Estes, W.K. (1970) Learning Theory and Mental Development. New York: Academic Press.

Falk-Vairant, J., & Crow, T. (1992) Intracellular injections of BAPTA block induction of enhancement in Hermissenda type B photoreceptors. Brain Research, 147: 45-48.

Fox, S. E., Wolfson S., and Ranck J. B. (1986) Hippocampal theta rhythm and the firing of neurons in walking and urethane anesthetized rats. Experimental Brain Research 62: 495-508.

Foy, M. R., Stanton M. E., Levine S., and Thompson R. F. (1987) Behavioral stress impairs long-term potentiation in rodent hippocampus. Behavioral and Neural Biology 48: 138- 149.

Fukunaga, K., Stoppini, L., Miyamoto, E., & Muller, D. (1993) Long-term potentiation is associated with an increased activity of Ca2++/calmodulin-dependent protein kinase II. Journal of Biological Chemistry 268: 7863-7867.

Gaito, J. (1976) Molecular psychobiology of memory: Its appearance, contributions, and decline. Physiological Psychology, 4: 476-484.

Garner, W.R., Hake, H.W., & Eriksen, C.W. (1956) Operationism and the concept of perception. Psychological Review 63: 149-159.

Glanzman, D.L. (1995) The cellular basis of classical conditioning in Aplysia californica - its less simple than you think. Trends in Neuroscience, 18: 30-36.

Goda, Y. (1994) Long-term potentiation. In pursuit of a retrograde messenger. Current Opinion in Biological Science, 4: 148-150.

Goodlett, C.R., Hamre K.M., West J. (1992) Dissociation of spatial navigation and visual guidance performance in Purkinje cell degeneration (pcd) mutant mice. Behavioral Brain Research, 47: 129-141.

Grant, S.G., O'Dell, T.J., Karl, K.A., Stein, P.L., Soriano, P., & Kandel, E.R. (1992) Impaired long- term potentiation, spatial learning, and hippocampal development in fyn mutant mice. Science, 258: 1903-1910.

Greenstein, Y. J., Pavlides C., and Winson J. (1988) Long-term potentiation in the dentate gyrus is preferentially induced at theta rhythm periodicity. Brain Research 438, 331-334.

Greenwood, P. M., Parasuraman R., and Haxby J.V. (1993) Changes in visuospatial attention over the adult lifespan. Neuropsychologia, 31: 471-485

Grossberg, S. (1975) A neural model attention, reinforcement, and discrimination learning. International Review of Neurobiology 18: 263-327

Grover, L.M., & Tyler, T.J. (1990) Two components of long-term potentiation induced by different patterns of afferent activation. Nature, 347: 477-479.

Haley, J.E., Wilcox, G.L., & Chapman, P.F. (1992) The role of nitric oxide in hippocampal long-term potentiation. Neuron 8: 211-216.

Hargreaves, E.L., Cain, D.P., & Vanderwolf, C.H. (1990) Learning and behavioral-long-term potentiation: importance of controlling for motor activity. Journal of Neuroscience 10: 1472-1478.

Harris, E.W., Ganong, A.H., & Cotman, C.W. (1984) Long-term potentiation in the hippocampus involves activation of N-methyl-D-aspartate receptors. Brain Research 323: 132-137.

Harris, K.M. & Teyler, T.J. (1984) Developmental onset of long-term potentiation in area CA1 of the rat hippocampus. Journal of Physiology (London) 346: 27-48.

Hawkins, R.D., Kandel, E.R., & Siegelbaum, S.A. (1993) Learning to modulate transmitter release: Themes and variations in synaptic plasticity. Annual Review of Neuroscience 16: 625-665.

Haxby, J.V., Horwitz B., Ungerleider L.G., Maison J.M., Pietrini P., Grady C.L. (1994) The functional organization of the human extrastriate cortex: a PET-rCBF study of selective attention to faces and locations. Journal of Neuroscience, 14: 6336-6353.

Haxby, J.V., Grady, C.L., Ungerleider, L.G., & Horwitz, B. (1991) Mapping the functional neuroanatomy of the intact human brain with brain work imaging. Neuropschologia, 29: 539-555

Hebb, D.O. (1949) The Organization of Behavior. John Wiley and Sons, Inc.

Hori, N., Hirotsu, I., Davis, P.J., & Carpenter, D.O. (1992) Long-term potentiation is lost in aged rats but preserved by calorie restriction. Neuroreport, 3: 1085-1088.

Horel, J.A. (1993) Retrieval of a face discrimination during suppression of monkey temporal cortex with cold. Neuropsychologia, 10: 1067-1077

Hoyle, G. (1980) Cellular analysis of operant conditioning of leg position. In C.D. Woody (Ed.), Conditioning: Representation of Involved Neural Functions. New York: Plenum.

Ito, M. (1984) The cerebellum and neural control. New York: Appleton-Century-Crofts.

James, W. (1892) Psychology: Briefer Course. New York: Holt.

Jaffe, D. & Johnston, D. (1990) Induction of long-term potentiation at hippocampal mossy fiber synapses follows a Hebbian rule. Journal of Neurophysiology, 64: 948-960.

Jahr, C.E., & Stevens, C.F. (1987) Glutamate activates multiple single channel conductances in hippocampal neurons. Nature: 325, 522-525.

Jeffery, K.J. & Morris, R.G.M. (1993) Cumulative long-term potentiation in the rat dentate gyrus correlates with, but does not modify, performance in the water maze. Hippocampus, 3: 133-140.

Johanson, I.B. & Hall, W.G. (1979) Appetitive learning in 1-day-old rat pups. Science 205: 419-421. Johanson, I.B. & Hall, W.G. (1982) Appetitive conditioning in neonatal rats: conditioned orientation to a novel odor. Developmental Psychobiology, 15: 379-397.

Johnston, D., Williams, S., Jaffe, D., & Gray, R. (1992) NMDA-receptor-independent long-term potentiation. Annual Review of Physiology, 54: 489-505.

Julien, R.M. (1992) A primer of drug action. New York: W.H. Freeman.

Kandel, E.R., & Tauc, L. (1965a) Heterosynaptic facilitation in neurons of the abdominal ganglion of Aplysia depilans. Journal of Physiology, 181: 1-27.

Kandel, E.R., & Tauc, L. (1965b) Mechanism of heterosynaptic facilitation in the giant cell of the abdominal ganglion of Aplysia depilans. Journal of Physiology, 181: 28-47.

Kato, K., Clark, G.D., Bazan, N.G., & Zorumski, C.F. (1994) Platelet-activating factor as a potential retrograde messenger in CA1 hippocampal long-term potentiation. Nature, 367: 175-179.

Kaczmarek, L. (1992) Expression of c-fos and other genes encoding transcription factors in long- term potentiation. Behavioral and Neural Biology, 57: 263-266.

Kandel E.R., Schwartz, J.H., & Jessel, T.M. (1991) Principles of Neuroscience. New York: Elsevier.

Kaplan & Hearst (1985) Trace conditioning, contiguity, and context. In M.L. Commons, R.J. Hernstein, & J.R. Wagner (Eds.), Quantitative Analysis of Behavior (Vol. 3: Acquisition) Cambridge, MA: Bellinger.

Keith, J.R. & Rudy, J.W. (1990) Why NMDA-receptor-dependent long-term potentiation may not be a mechanism of learning and memory: Reappraisal of the NMDA-receptor blockade strategy. Psychobiology 18: 251-257.

Kim, J.J., & Fanselow, M.S. (1992) Modality-specific retrograde amnesia of fear. Science, 256: 675-677.

Kim J.J., DeCola J.P., Landeira-Fernandez J., & Fanselow M.S. (1991) N-methyl-D-aspartate receptor antagonist APV blocks acquisition but not expression of fear conditioning. Behavioral Neuroscience 105: 126-133.

Knowlton, B.J. & Thompson, R.F. (1992) Conditioning using a cerebral cortical conditioned stimulus is dependent on the cerebellum and brain stem circuitry. Behavioral Neuroscience 106: 509-517.

Komatsu, Y., Nakajima, S., & Toyama, K. (1991) Induction of long-term potentiation without participation of N-methyl-D-aspartate receptors in kitten visual cortex. Journal of Neurophysiology 65: 20-32.

Koyano, K., Kuba, K., & Minota, S. (1985) Long-term potentiation of transmitter releases induced by repetitive presynaptic activities in bullfrog sympathetic ganglion. Journal of Physiology (London), 359: 219- 233.

Kramis, R., Vanderwolf, C.H., & Bland, B.H. (1975) Two types of hippocampal rhythmical slow activity in both the rabbit and the rat: relations to behavior and effects of atropine, diethyl ether, urethane, and pentobarbital. Experimental Neurology, 49: 58-85.

Krug, M., Bergado, J., & Ruethrich, H. (1990) Long-term potentiation and postconditioning potentiation-the same mechanism? Biomed Biochem Acta, 49: 273-279.

Krupa D.J., Thompson, J.K., & Thompson, R.F. (1993) Localization of a memory trace in the mammalian brain. Science 260: 989-991.

Kuhn, T.J. (1973) The Structure of Scientific Revolutions. Chicago: University of Chicago Press.

Kullmann D.M., Perkel, D.J., Manabe, T., & Nicoll, R.A. (1992) Ca2+ entry via postsynaptic voltage-dependent Ca2+ channels can transiently potentiate excitatory synaptic transmission in the hippocampus. Neuron, 9: 1175-1183.

Korol D.L., Abel T.W., Church L.T., Barnes C.A., & McNaughton B.L. (1993) Hippocampal synaptic enhancement and spatial learning in the Morris swim task. Hippocampus 3: 127-132.

Laroche S., Doyere V., & Bloch V. (1989) Linear relation between the magnitude of long-term potentiation in the dentate gyrus and associative learning in the rat. A demonstration using commissural inhibition and local infusion of an N-methyl-D-aspartate receptor antagonist, Neuroscience 28: 375-386.

Larson, J., Wong, D., & Lynch, G. (1986) Patterned stimulation at the theta frequency is optimal for the induction of hippocampal long-term potentiation. Brain Research 368: 347-350.

Larson, J. & Lynch, G. (1989) Theta pattern stimulation and the induction of LTP: the sequence in which synapses are stimulated determines the degree to which they potentiate. Brain Research 489: 49-58.

Larson, J. & Lynch, G. (1988) Role of N-methyl-D-aspartate receptors in the induction of synaptic potentiation by burst stimulation patterned after the hippocampal theta-rhythm. Brain Research 441: 111-118.

Larson, J.R., & Lynch, G. (1985) Long-term potentiation in the lizard cerebral cortex. Society for Neuroscience Abstracts, 11: a225.2.

Lavond, D.G., Kim, J.J., & Thompson, R.F. (1993) Mammalian brain substrates of aversive classical conditioning. Annual Review of Psychology, 44: 317-342.

Levy, W.B. & Steward, O. (1979) Synapses as associative memory elements in the hippocampal formation. Brain Research 175: 233-245.

Levy, W.B. & Steward, O. (1983) Temporal contiguity requirements for long-term associative potentiation/depression in the hippocampus. Neuroscience 8: 791-797.

Lewis, D. & Teyler, T.J. (1986) Long-term potentiation in the goldfish optic tectum. Brain Research 375: 246-250.

Libet, B., Kobayashi, H., & Tanaka, T. (1975) Synaptic coupling into the production and storage of a neuronal memory trace. Nature, 258: 155-157.

Lisman, J. (1994) The CaM kinase II hypothesis for the storage of synaptic memory. Trends in Neuroscience, 17: 406-412.

Lynch, G., Dunwiddie, T., & Gribkoff, V. (1977) Heterosynaptic depression: A postsynaptic correlate of long-term potentiation. Nature, 266: 737-739.

Lynch, G., Larson, J., Kelso, S., Barrionuevo, G., & Schottler, F. (1983) Intracellular injections of EGTA block induction of hippocampal long-term potentiation. Nature, 305: 719-721.

Lynch, M.A., Clements, M.P., Voss, K.L., Bramham, C.R., & Bliss, T.V.P. (1991) Is arachidonic acid a retrograde messenger in long-term potentiation. Biochem Society Trans, 19: 391-396.

Lynch G., Baudry M. (1984) The biochemistry of memory: a new and specific hypothesis. Science 224: 1057-1063.

Lynch G. & Staubli U. (1990) Letter to the editor: Reply to Keith and Rudy. Psychobiology 18: 369.

Mackintosh, N.J. (1974) The Psychology of Animal Learning. London: Academic Press.

Mackintosh, N.J. (1975) A theory of attention: variations in the associability of stimuli with reinforcer. Psychological Review 82: 276-298.

Malenka, R.C. (1992) The role of postsynaptic calcium in the induction of long-term potentiation. Molecular Neurobiology, 5: 289-295.

Malenka, R.C., Kauer, J.A., Perkel, D.J., Mauk, M.D., Kelly, P.T., Nicoll, R.A., & Waxham, M.N. (1989) An essential role for postsynaptic calmodulin and protein kinase activity in long-term potentiation. Nature 340: 554-557.

Malenka, R.C., Kauer, J.A., Zucker, R.S., & Nicoll, R.A. (1988) Postsynaptic calcium is sufficient for potentiation of hippocampal synaptic transmission. Science, 242: 81-84.

Malenka, R.C., Madison, D.V., & Nicoll, R.A. (1986) Potentiation of synaptic transmission in the hippocampus by phorbol esters. Nature 321: 175-177.

Malinow, R., Madison, D.V., & Tsien, R.W. (1988) Persistent protein kinase activity underlying long- term potentiation. Nature 335: 820-824.

Malinow, R., Schulman H., & Tsien, R.W. (1989) Inhibition of postsynaptic PKC or CaMKII blocks induction but not expression of LTP. Science 245: 862-869.

Maorides: F. (1975) Temporal relationships between hippocampal slow waves and exploratory sniffing in hamsters. Behavioral Biology, 14: 295-308.

Maren, S., Tocco G., Standley S., Baudry M., Thompson R.F. (1993) Postsynaptic factors in the expression of long-term potentiation (LTP): increased glutamate receptor binding following LTP induction in vivo. Proceedings of the National Academy of Sciences (U.S.A.) 90 (20): 9654-9658.

Maren, S., DeCola J.P., Swain R.A., Fanselow M.S. and Thompson, R.F. (1994) Parallel augmentation of hippocampal long-term potentiation, theta rhythm, and contextual fear conditioning in water-deprived rats. Behavioral Neuroscience 108: 44-56.

Mark, G.P., Rada P., and Shors T.J. (in press) Stress increases the release of acetylcholine into the hippocampus and prefrontal cortex, but not the amygdala or nucleus accumbens. Neuroscience.

Martin-Elkins, C.L., George, P., & Horel, J.A. (1989) Retention deficits produced in monkeys with reversible cold lesions in the prestriate cortex. Behavioral Brain Research, 32: 219- 230.

Martinez, J.L., & Derrick, B.E. (1996) Long-term potentiation and learning. Annual Review of Psychology, 47, 173-203.

Matzel, L.D., Collin, C., & Alkon, D.L. (1992) Biophysical and behavioral correlates of memory storage, degredation, and reactivation. Behavioral Neuroscience, 106: 954-963.

Matzel, L.D., Held, F.P., & Miller, R.R. (1988) Information and expression of simultaneous and backward associations: Implications for contiguity theory. Learning and Motivation, 19: 317-344.

Matzel, L.D. & Rogers, R.F. (1993) Postsynaptic calcium, but not cumulative depolarization, is necessary for the induction of associative plasticity in Hermissenda. Journal of Neuroscience, 13, 5029-5040

Matzel, L.D., Muzzio, I., & Talk, A.C. (in press) Variations in learning reflect individual differences in sensory processing and synaptic integration. Behavioral Neuroscience.

McClelland, J.L., McNaughton, B.L., O'Reilly, R.C. (1995) Why there are complimentary learning systems in the hippocampus and neocortex: insights from the successes and failures of connectionist models of learning and memory. Psychological Review, 102: 419-457.

McCollum, J., Larson, J., Otto, T., Schottler, F., Granger, R., & Lynch, G. (1991) Short-latency single unit processing in olfactory cortex. Journal of Cognitive Neuroscience, 3: 293-299.

McCormick, D.A., Clark, G.A., Lavond, D.G., & Thompson, R.F. (1982) Initial localization of the memory trace for a basic form of learning. Proceedings of the National Academy of Sciences 79(8): 2731-2742.

McCormick, D.A., Thompson, R.F. (1984) Cerebellum: essential involvement in the classically conditioned eyeblink response. Science 223: 296-299.

McDonald, R.J., & White, N.M. (1993) A triple dissociation of memory systems: Hippocampus, amygdala, and dorsal striatum. Behavioral Neuroscience, 107: 3-22.

McDonald, R.J., & White, N.M. (1995) Information acquired by the hippocampus interferes with acquisition of amygdala-based conditioned cue preference. Hippocampus 5: 189-197.

McNaughton, B.L., Barnes, C.A., Rao, G., Baldwin, J., & Rasmussen, M. (1986) Long-term enhancement of hippocampal synaptic transmission and the acquisition of spatial information. Journal of Neuroscience 6: 563-571.

McNaughton, B.L., Douglas, R.M., & Goddard, G.V. (1978) Synaptic enhancement in fascia dentata: Cooperativity among coactive afferents. Brain Research, 157: 277-293.

Miller, S.G. & Kennedy, M.B. (1986) Regulation of brain type II Ca2+/calmodulin-dependent protein kinase by autophosphorylation: A Ca2+-triggered molecular switch. Cell, 44: 861- 870.

Mondadori, C., Weiskrantz, H., Buerki, H., & Petschke, F. (1989) NMDA receptor antagonists can enhance or impair learning performance in animals. Experimental Brain Research, 75: 449-456.

Moore, J.W., Stickney, K.J. (1980) Formation of attentional-associative networks in real time: role of hippocampus and implications for conditioning. Physiological Psychology 8: 207-217.

Morishita, W., Xie, Z., Chirwa, S.S., May, P.B., & Sastry, B.R. (1992) Blockade of hippocampal long- term potentiation by saccharin. Neuroscience, 47: 21-31.

Morris, R.G., Anderson, E., Lynch, G.S., & Baudry, M. (1986) Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature, 319: 774-776.

Morris, R.G.M., Halliwell R.F. and Bowery N. (1989) Synaptic plasticity and learning: II. Do different kinds of plasticity underlie different kinds of learning. Neuropsychologia 27, 41-59.

Morris R.G.M., Davis, S., & Butcher, S.P. (1991) Hippocampal synaptic plasticity and NMDA receptors: a role in information storage. In J.R. Krebs & G. Horn (Eds.), Behavioral and neural aspects of learning and memory (pp. 89-106) Oxford: Clarendon Press.

Morrisett, R.A., & Swartzwelder, H.S. (1993) Attenuation of hippocampal long-term potentiation by ethanol: A patch-clamp analysis of glutamatergic and GABAergic mechanisms. Journal of Neuroscience, 13: 2264-2272.

Nadel, L. (1994) The role of the hippocampus in declarative memory: A comment on Zola-Morgan, Squire, and Ramus. Hippocampus, 5: 232-239.

Nicholls, J.G., Martin, R.A., Wallace, B.G., & Kuffler, S.W. (1992) From neuron to brain: A cellular and molecular approach. Sunderland: Sinauer.

Norman, D.A., & Shallice, T. (1986) Attention to action: Willed and automatic control of behavior. In R.J. Davidson, G.E. Schwartz, & D. Shapiro (Eds.), Consciousness and self-regulation, Vol. 4 (pp. 1-18) New York: Plenum Press.

O'Dell, T.J., Hawkins, R.D., Kandel, E.R., & Arancio, O. (1991) Tests of the roles of two diffusible substance in long-term potentiation : evidence for nitric oxide as a possible early retrograde messenger. Proceedings of the National Academy of Sciences (USA) 88: 11285.

O'Dell, T.J., Grant, S.G., Karl, K., Soriano, P.M., & Kandel, E.R. (1992) Pharmacological and genetic approaches to the analysis of tyrosine kinase function in long-term potentiation. Cold Spring Harbor Symposium Quantitative Biology, 57: 517-526.

O'Dell, T.J., Kandel, E.R., & Grant, S.G. (1991) Long-term potentiation in the hippocampus is blocked by tyrosine kinase inhibitors. Nature, 353: 558-560.

Olds, J. (1955) Physiological mechanisms of reward. In M.R. Jones (Ed.), Nebraska Symposium on Motivation. Lincoln: University of Nebraska Press.

Otto, T., Eichenbaum, H., Wiener, S.I., & Wible, C.G. (1991) Learning-related patterns of CA1 spike trains parallel stimulation parameters optimal for inducing hippocampal long-term potentiation. Hippocampus, 1: 181-192.

Overmier, J.B. & Seligman, M. (1967) Effects of inescapable shock upon subsequent escape and avoidance learning. Journal of Comparative and Physiological Psychology, 63: 23- 33.

Pavlides, C., Greenstein, Y. J., Grudman, M. & Winson, J. (1988) Long-term potentiation in the dentate gyrus is induced preferentially on the positive phase of theta-rhythm. Brain Research 439: 383- 387.

Posner, M.I., & Petersen, S.E. (1990) The attention system of the human brain. Annual Review of Neuroscience, 13: 25-42.

Prokasy, W.F., Kesner, R.P. & Calder, L.D. (1983) Posttrial electrical stimulation of the dorsal hippocampus facilitates acquisition of the nictitating membrane response. Behavioral Neuroscience, 97: 890-896.

Racine, R. J., Milgram, M. W. & Hafner, S. (1983) Long-term potentiation in the rat limbic forebrain. Brain Research 260: 217-231.

Rescorla, R.A. (1980) Simultaneous and successive associations in sensory preconditioning. Journal of Experimental Psychology: Animal Behavioral Processes, 6: 207-216.

Rescorla, R.A. (1988) Behavioral studies of Pavlovian conditioning. Annual Review Neuroscience 11: 329-352.

Reymann, K.G., Schulzeck, K., Kase, H., & Matthies, H. (1988) Phorbol ester-induced hippocampal long-term potentiation is counteracted by inhibitors of protein kinase C. Experimental Brain Research 71: 227-230.

Reymann, K. G., Malisch, R., Schulzeck, K., Brodemann, R., Ott, T. & Matthies, H. (1985) The duration of long-term potentiation in the CA1 region of the hippocampal slice preparation. Brain Research Bulletin 15: 249-255.

Rioux, G.F. and Robinson, G.B. Hippocampal LTP does not affect either discrimination learning or reversal learning of the rabbit nictitating membrane response. Hippocampus in press.

Robinson, G.B. (1992) Maintained saturation of hippocampal long-term potentiation does not disrupt acquisition of the eight-arm radial maze. Hippocampus 2 (4): 389-396.

Robinson, G.B. (1993) MK801 retards acquisition of a classically conditioned response without affecting conditioning-related alterations in perforant path-granule cell synaptic transmission. Psychobiology 21: 253-264.

Rock, I. (1956) The role of repetition in associative learning. American Journal of Psychology, 70: 186-193.

Roman, F., Staubli, U., & Lynch, G. (1987) Evidence for synaptic potentiation in a cortical network during learning. Brain Research 418: 221-226.

Rose, G. M. & Dunwiddie, T. V. (1986) Induction of hippocampal long-term potentiation using physiologically patterned stimulation. Neuroscience Letters 69: 244-248.

Rose, S.P.R. (1992) On chicks and rosetta stones. In L.R. Squire & N. Butters (Eds.), Neuropsychology of Memory (pp. 547-556) New York: The Guilford Press.

Rose, S.P.R. (1995) Cell-adhesion molecules, glucocorticoids, and long-term memory formation. Trends in Neuroscience 18: 502-506.

Sadile, A.G., & Papa, M. (1993) Arousal and habituation to novelty induce nitric oxide synthase activity in the rat brain. Society for Neuroscience Abstracts, 412.

Sainsbury, R.S., Harris, J.L., and Rowland, G.L. (1987) Sensitization and hippocampal type 2 theta in the rat. Physiology and Behavior 17: 481-483.

Sakimura, K., Kutsuwada, T., Ito, I., Manabe, T., Takayama, C., Kushlya, E., Yagi, T., Alzawa, S., Inoue, Y., Sugiyama, H., Mishina, M. (1995) Reduced hippocampal LTP and spatial learning lacking in mice lacing NMDA receptor 1 subunit. Nature, 373: 151-155.

Salafia, W.R., Romano, A.G., Tynan, T., Host, K.C. (1977) Disruption of rabbit nictitating membrane conditioning by posttrial electrical stimulation of hippocampus. Physiology and Behavior, 18: 207- 212.

Saucier, D& ., Cain, D.P. (1995) Spatial learning without NMDA receptor-dependent long-term potentiation. Nature 378: 186-189.

Scavio, M.J. & Gormezano, I. (1974) CS intensity effects on rabbit nictitating membrane conditioning, extinction and generalization. Pavlovian Journal of Biological Science 9: 25-34.

Schmajuk, N.A., & DiCarlo J.J. (1991) A neural network approach to hippocampal function in classical conditioning. Behavioral Neuroscience 105: 82-110.

Schreiber, S., Tocco, G., Shors, T. J., & Thompson, R. F. (1991) Immediate early gene induction after acute stress. Neuroreport 2: 17-20.

Schreurs, B.G. (1989) Classical conditioning of model systems: A behavioral review. Psychobiology, 17: 145-155.

Schuman, E.M. & Madison, D.V. (1991) A requirement for the intercellular messenger nitric oxide in long-term potentiation. Science 254: 1503-1506.

Schuman, E.M. & Madison, D.V. (1994) Locally distributed synaptic potentiation in the hippocampus. Science, 263: 532-536.

Schwartz, J.H. & Greenberg, S.M. (1987) Molecular mechanisms of memory: Second-messenger induced modifications of protein kinases in nerve cells. Annual Review of Neuroscience, 10: 459- 476.

Schwartzkroin, P. & Wester, K. (1975) Long-lasting facilitation of a synaptic potential following tetanization in the in vitro hippocampal slice. Brain Research, 89: 107-119.

Scott, T.R. & Bennett, M.R. (1993) The effect of ions and second messengers on long-term potentiation of chemical transmission in avian ciliary ganglia. British Journal of Pharmacology 110: 461-469.

Scoville, W.B. & Milner, B. (1957) Loss of recent memory after bilateral hippocampal lesions. Journal of Neurology, Neurosurgery, and Psychiatry 20:11-21.

Seligman, M.E.P., & Johnston, J.C. (1973) A cognitive theory of avoidance learning. In F.J. McGuigan & D.B. Lumsden (Eds.), Contemporary Approaches to Conditioning and Learning. Washington, D.C.: V.H. Winston & Sons.

Seindenbecher, T., Balschun, D., & Reymann, K.G. (1995) A post-tetanic time window for the reinforcement of LTP by appetitive and aversive stimuli. IXth Magdeburg International Neurobiological Symposium on Learning and Memory: Synaptic and Systemic Views.

Sejenowski, T.J. (1990) Homosynaptic long-term depression in hippocampus and neocortex. Seminars in the Neurosciences 2: 355-363.

Servatius, R.J. & Shors, T.J. (1994) Exposure to inescapable stress persistently facilitates nonassociative and associative learning in rats. Behavioral Neuroscience 108: 1101- 1106.

Servatius, R.J. & Shors, T.J. (in press) Early acquisition, but not retention of the classically conditioned eyeblink response is N-methyl-d-aspartate (NMDA) receptor-dependent. Behavioral Neuroscience.

Sheliga, B.M., Riggio, L., Craighero, L., Rizzolatti, G. (1995) Spatial attention-determined modifications in saccade trajectories. Neuroreport, 6: 585-588.

Shors, T. J., Seib, T. B., Levine, S. & Thompson, R. F. (1989) Inescapable versus escapable shock modulates long-term potentiation (LTP) in rat hippocampus. Science 244: 224-226.

Shors, T. J., Foy, M. R., Levine, S. & Thompson, R. F. (1990) Unpredictable and uncontrollable stress impairs neuronal plasticity in the rat hippocampus. Brain Research Bulletin 24: 663-667.

Shors, T. J., Weiss, C. & Thompson, R. F. (1992) Stress-induced facilitation of classical conditioning. Science 257: 537-539.

Shors, T. J. & Thompson, R. F. (1992) Acute stress impairs (or induces) synaptic LTP, but does not affect paired-pulsed facilitation in the stratum radiatum of the rat hippocampus. Synapse 11: 262- 265.

Shors, T. J. & Dryver, E. (1992) Stress impedes exploration and the acquisition of spatial information in the 8-arm radial maze. Psychobiology 20: 247-253.

Shors, T.J. & Dryver, E. (1994) Long-term consequences of stress on subsequent long-term potentiation (LTP) and the theta burst response in the dentate gyrus. Brain Research 666: 232-238.

Shors, T.J., Servatius, R.J., Thompson, R.F., Rogers, G., Lynch, G. (1995) Enhanced glutamatergic neurotransmission facilitates classical conditioning in the freely moving rat. Neuroscience Letters 186: 153-156.

Shors, T.J. and Servatius, R.J. (1995) Stress-induced sensitization and facilitated learning are dependent on NMDA receptor activation. Neuroreport 6: 677-680..

Silva, A.J., Paylor, R., Wehner, J.M., & Tonegawa, S. (1992) Impaired spatial learning in calcium- calmodulin kinase II mutant mice. Science 257: 206-211.

Silva, A.J., Stevens, C.F., Tonegawa, S., & Wang, Y. (1992) Deficient hippocampal long-term potentiation in calcium-calmodulin kinase II mutant mice. Science, 257: 201-205.

Smirnova, T., Laroche, S., Errington, M.L., Hicks, A.A., Bliss, T.V., & Mallet, J. (1993) Transsynaptic expression of a presynaptic glutamate receptor during hippocampal long-term potentiation. Science 262: 433-436.

Smith, D.A., Browning, M., & Dunwiddie, T.V. (1993) Cocaine inhibits hippocampal long-term potentiation. Brain Research, 608: 259-265.

Smotherman, W.P. (1982) Odor aversion learning by the rat fetus. Physiology and Behavior, 29: 769-771.

Smotherman, W.P. & Robinson, S.R. (1991) Conditioned activation of fetal behavior. Physiology and Behavior, 50: 73-77.

Solomon, P.R., VanderSchaaf, E.R., Thompson, R.F., Weisz, D. (1986) Hippocampus and trace conditioning of the rabbit's classically-conditioned nictitating membrane response. Behavioral Neuroscience, 100: 729-744.

Solomon, P.R., & Moore, J.W. (1975) Latent inhibition and stimulus generalization of the classically conditioned nictitating membrane response in rabbits following dorsal hippocampal ablation. Journal of Comparative and Physiological Psychology, 89: 1192-1203.

Spear, N.E. (1978) The processing of memories: Forgetting and retention. Hillsdale: Erlbaum.

Spear, & Riccio (1993) Memory: Phenomena and Principles. Boston: Allyn & Bacon.

Spitzer, H., Desimone, R., & Moran, J. (1988) Increased attention enhances both behavioral and neuronal performance. Science, 240: 338-340.

Squire, L.R., Cohen, N.J., & Nadel, L. (1984) The medial temporal region and memory consolidation: A new hypothesis. In H. Weingartner & E. Parker (Eds.), Memory Consolidation (pp. 185-210) Hillsdale: Erlbaum.

Spencer, H. (1870) The Principles of Psychology. London: Williams and Norgate.

Squire, L.R. & Zola-Morgan, S. (1991) The medial temporal lobe memory system. Science, 253: 1380-1386.

Staubli, U. & Lynch, G. (1987) Stable hippocampal long-term potentiation elicited by "theta" pattern stimulation. Brain Research, 435: 227-234.

Staubli, U., Thibault O., DiLorenzo M., & Lynch G. (1989) Antagonism of NMDA receptors impairs acquisition but not retention of olfactory memory. Behavioral Neuroscience 103: 54-60.

Staubli, U. (1990) Behavioral reflections of the NMDA system. Psychobiology 18: 267-268.

Staubli U., Rogers G., and Lynch G. (1994) Facilitation of glutamate receptors enhances memory. Proceedings of the National Academy of Science (U.S.A.), 91:777-781.

Staubli U. , Perez Y., Xu F., Rogers G., Inguar M., Stone-Elander S., and Lynch G., Centrally active modulators of glutamate (AMPA) receptors facilitate the induction of LTP in vivo. (1995) Proceedings of the National Academy of Science (U.S.A.) in press.

Stefurak, T.L. & van der Kooy, D. (1992) Saccharin's rewarding, conditioned reinforcing, and memory-improving properties: mediation by isomorphic or independent processes? Behavioral Neuroscience 106: 125-139.

Stripling, J.S., Patneau, D.K., & Gramlich, C.A. (1988) Selective long-term potentiation in the pyriform cortex. Brain Research 441: 281-291.

Sutherland, R. J., Dringenberg H. C., and Hoesing J. M. (1993) Induction of long-term potentiation at perforant path dentate synapses does not affect place learning or memory. Hippocampus 3(2): 141-148.

Swain, R.A., Shinkman, P.G., Nordholm, A.F., & Thompson, R.F. (1992) Cerebellar stimulation as an unconditioned stimulus in classical conditioning. Behavioral Neuroscience 106: 739-750.

Tanzi, E. (1893) I fatti e la induzioni nell 'odierna istologia del sistema nervoso. Riv Sper Freniatr Med Leg Alenazioni Ment Soc Ital Psichiatria, 19: 419-472.

Turski, L., Schwartz, M., Turski, W.A., Klockgether, T., Sontag, K.H., & Collins, J.F. (1985) Muscle relaxant action of excitatory amino acid antagonists. Neuroscience Letters, 53: 321-326.

Tesauro, G. (1988) A plausible neural circuit for classical conditioning without synaptic plasticity. Proceedings of the National Academy of Sciences U.S.A. 85: 2830-2833.

Teyler, J.T. & DiScienna, P. (1987) Long-term potentiation. Annual Review of Neuroscience 10: 131- 161.

Thompson, R.F. (1990) Neural mechanisms of classical conditioning in mammals. Philosophical Transactions of the Royal Society of London 29: 161-170.

Tocco, G., Shors, T. J., Baudry, M., and Thompson, R. F. (1991) Selective increase of AMPA binding to the AMPA/quisqualate receptor in the hippocampus in response to acute stress. Brain Research 559: 168-171.

Tocco, G., Maren, S., Shors, T. J., Baudry, M., and Thompson, R. F. (1992) Long-term potentiation is associated with increased [H]AMPA binding in rat hippocampus. Brain Research 573: 228-234.

Tonkiss, J., Morris, R.G.M., & Rawlins, J.N.P. (1988) Intra-ventricular infusion of the NMDA antagonist AP5 impairs performance on a non-spatial operant DRL task in the rat. Experimental Brain Research 73: 181-188.

Umemori, H., Sato, S., Yagi, T., Aizawa, S., & Yamamoto, T. (1994) Initial events of myelination involve Fyn tyrosine kinase signaling. Nature 367: 572-576.

Vandercar, D.H., Elster A.J., Schneiderman N. (1970) Heart-rate conditioning in rabbits in hypothalamic or septal US stimulations. Journal of Comparative and Physiological Psychology, 72: 145-152.

Vanderwolf, C.H. (1969) Hippocampal electrical activity and voluntary movement in the rat. Electroencepholography and Clinical Neurophysiology, 26: 407-418.

Vanderwolf, C.H. (1988) Cerebral activity and behavior: Control by central cholinergic and serotonergic systems. International Review of Neurobiology, 30: 225-340.

Vanderwolf, C.H. & Cain, D.P. (1994) The behavioral neurobiology of learning and memory: a conceptual reorientation. Brain Research Reviews 19: 264-297.

Weisskopf, M.G., Castillo, P.E., Zalutsky, R.A., & Nicoll, R.A. (1994) Mediation of hippocampal mossy fiber long-term potentiation by cyclic AMP. Science, 265: 1878- 1882.

Walters, E.T. & Byrne, J.H. (1985) Long-term enhancement produced by activity-dependent modulation of Aplysia sensory neurons. Journal of Neuroscience 5: 662-672.

Watanbe, Y., Saito, H., & Abe, K. (1993) Tricyclic antidepressants block NMDA receptor-mediated synaptic responses and induction of long-term potentiation in rat hippocampal slices. Neuropharmacology 32: 479-486.

Weiss, D.S. (1988) Membrane potential modulates the activation of GABA-gated channels. Journal of Neurophysiology 59: 514-527.

Weisz, D., Clark, G.A., & Thompson, R.F. (1984) Increased responsivity of dentate granule cells during nictitating membrane response conditioning in the rabbit. Behavioral Brain Research 12: 145-154.

Wierazko, A. & Ball, G.F. (1993) Long-term potentiation in the avian hippocampus does not require activation of the N-methyl-D-aspartate (NMDA) receptor. Synapse, 13: 173-178.

Wilhite, B.L., Teyler, T.J., & Hendricks, C. (1986) Functional relations of the rodent claustral-entorhinal-hippocampal system. Brain Research, 365: 54-60.

Williams, J.H. & Bliss, T.V. (1989) An in vitro study of the effect of lipoxygenase and cyclo-oxygenase inhibitors of arachidonic acid on the induction and maintenance of long-term potentiation in the hippocampus. Neuroscience Letters 107: 301-306.

Williams, J.H., Li, Y.G., Nayak, A., Errington, M.L., Murphy, K.P., & Bliss, T.V. P. (1993) The suppression of long-term potentiation in rat hippocampus by inhibitors of nitric oxide synthase is temperature and age dependent. Neuron 11: 877-884.

Williams, S., & Johnston, D. (1989) Long-term potentiation of hippocampal mossy fiber synapses is blocked by postsynaptic injection of calcium chelators. Neuron, 3: 583-588.

Wilson, D.A. (1984) A comparison of the postnatal development of post-activation potentiation in the neocortex and dentate gyrus of the rat. Developmental Brain Research 16: 61-68.

Xie, Z. & Sastry, B.R. (1993) Induction of hippocampal long-term potentiation by alpha-tocopherol. Brain Research 604: 173-179.

Yamamoto, C. & Chujo, T. (1978) Long-term potentiation in thin hippocampal sections studied by intracellular and extracellular recording. Experimental Neurology, 58: 242-250.

Yeo, C.H., Hardiman, M.J., & Glickstein, M. (1986) Classical conditioning of the nictitating membrane response of the rabbit. IV: Lesions of the inferior olive. Experimental Brain Research 63: 81-92.

Yagi, T., Aizawa, S., Tokunaga, T., Shigetani, Y. (1993) A role for Fyn tyrosine kinase in the suckling behavior of neonatal mice. Nature, 366: 742-745.

Zhou, M., Hu, Y., Schultz, C., Kandel, E.R., & Hawkins, R.D. (1994) Role of guanylyl cyclase and cGMP-dependent protein kinase in long-term potentiation. Nature, 368: 635-639.

Zola-Morgan, S. & Squire, L. (1991) The primate hippocampal formation: Evidence for a time-limited role in memory storage. Science, 250: 288-290.