- Email: [email protected]
Contextual fear, gestalt memories, and the hippocampus
Behavioural Brain Research 110 (2000) 73 – 81 www.elsevier.com/locate/bbr
Contextual fear, gestalt memories, and the hippocampus Michael S. Fanselow * Department of Psychology and Brain Research Institute, Uni6ersity of California, Los Angeles, CA 90095 -1563, USA Accepted 25 November 1999
Abstract This review examines the relationship between exploration and contextual fear conditioning. The fear acquired to places or contexts associated with aversive events is a form of Pavlovian conditioning. However, an initial period of exploration is necessary to allow the animal to form an integrated memory of the features of the context before conditioning can take place. The hippocampal formation plays a critical role in this process. Cells within the dorsal hippocampus are involved in the formation, storage and consolidation of this integrated representation of context. Projections from the subiculum to the nucleus accumbens regulate the exploration necessary for the acquisition of information about the features of the context. This model explains why electrolytic but not excitotoxic lesions of the dorsal hippocampus cause enhanced exploratory activity but both cause deficits in contextual fear. It also explains why retrograde amnesia of contextual fear is greater than anterograde amnesia. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Hippocampus; Fear; Context conditioning; Freezing; Nucleus accumbens; Subiculum; Exploration; Hyperactivity
1. Introduction It is well accepted that the hippocampal formation plays a role in certain forms of learning and memory [41]. Declarative, relational and spatial memory are ‘hippocampal-dependent,’ but there are a large number of other learning/memory tasks (procedural, implicit) that are hippocampally-independent [9,38,42]. In the last few years, contextual fear conditioning has become a popular assay for hippocampal-dependent learning. This is in part because the fear conditioning task has built into it directly comparable forms of hippocampaldependent and hippocampal-independent memory. However, before the initial reports of the role of the hippocampus in context conditioning, there were behavioral data suggesting some unique aspects of context conditioning and these led to the initial investigations of the role of the hippocampus in context conditioning. Therefore, it may be valuable to reconsider some of the basic behavioral features of contextual fear conditioning — especially those relevant to the role of the hippocampal formation in context conditioning. That is * Tel.: +1-310-2063891; fax: + 1-310-2065895. E-mail address: [email protected] (M.S. Fanselow)
where this paper will begin. It will then review some of the findings that relate hippocampal function to context conditioning and contrast three different views of why hippocampal lesions affect behavior in fear conditioning situations. It will conclude with a synthesis of these views that embraces many of the existing empirical complexities. 2. Environmental control of contextual fear In the simplest version of contextual fear conditioning, a rat is placed in a novel environment. After it explores the chamber for a few minutes it receives aversive electric footshock. Typically the shock is brief (e.g. 0.5–2 s), mild (0.3–1.5 mA) and infrequent (one to four shocks). The behavior observed in the situation is striking [18]. The rat shows vigorous locomotor activity during the shock that persists, in a somewhat diminished form, for a period beyond shock termination. However, this activity burst is replaced gradually by a profound immobility called freezing. During freezing all observable movement, including that of the vibrissae ceases, except that of the flanks related to respiration. Freezing is usually, but not necessarily, in a crouching posture against a wall of the enclosure.
0166-4328/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 6 - 4 3 2 8 ( 9 9 ) 0 0 1 8 6 - 2
74
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
The most compelling account of this behavior is in terms of a Pavlovian analysis. The shock is the unconditional stimulus (US) analogous to Pavlov’s meat powder. The unconditional reaction (UR) to the shock is the activity burst, while freezing is the conditional response (CR). In this case the conditional stimulus (CS), which provokes the CR, are the cues that compose the novel chamber (contextual cues). These contextual cues function like the bell in the proverbial Pavlovian conditioning experiment. One striking difference with Pavlov’s experiment is the relationship between the CR and the UR. In Pavlov’s experiment they were both salivation but in fear conditioning the CR (freezing) and UR (activity burst) are strikingly different. When saying that freezing is a Pavlovian response, it means that what controls the response is the relationship between the stimuli in the situation, in this case between context and shock. It also means that freezing
Fig. 1. Rats were tested in either the same or a different context from the one they were shocked in. The data are collapsed over a factor of whether the test occurred immediately or 24 h after shock. Based on [19].
Fig. 2. Rats were tested either immediately or 24 h after shock. The data are from the same test as those of Fig. 1 but are collapsed over the same/different factor. Based on [19].
is not controlled by its effects on the environment, that is, it is not an operant response. With aversively motivated operant behaviors, shock is delivered contingent on a particular behavior. With operant responses such as inhibitory avoidance, rats are given a shock only after they make the response of entering another compartment; shock is under control of the subject. While it is conceivable that rats might freeze in a contextual conditioning experiment because freezing reduced the impact of shock, explicit tests of the influence of operant contingencies on freezing has ruled this out. Even when freezing is explicitly programmed to control shock delivery such contingencies do not control this behavior (see [14]). It should be noted that while other commonly used indices of fear such as potentiated startle are assumed to be Pavlovian, such formal analyses of these responses have never been conducted. But with freezing, we know that we are examining a Pavlovian response. In the above, it was stated that the freezing that follows even a single shock is a CR to the contextual cues present at the time of shock. Another, possibility is that it is an UR but a delayed UR. For example, the UR may have an early and a late component with the activity burst being the early component and freezing the late component [46]. Such an analysis would help reconcile the relationship of the CR and the UR in fear conditioning with that found it other types of Pavlovian conditioning such as salivation and eyeblink. The concept is quite testable. If freezing is a CR it should be CS or context specific; its occurrence must depend on the presence of the cues that were present at the time of shock. If freezing is an UR it should be time locked to delivery of the US; substantial delays between shock and testing should reduce freezing. I undertook an experiment to factorially manipulate these two variables [19]. Rats were placed in a chamber, where they received four mild or moderate intensity shocks; there were also controls that received no shock. The rats were removed from the chamber after the shock. To test for context dependency, the rats were replaced into the same chamber they were shocked in or an alternative (shock and no shock chambers were counterbalanced across animals). To test for time dependency this test occurred either immediately after shock or 24 h later. The data for the 12 groups of this experiment are summarized in Figs. 1 and 2. The first thing that should be noticed is that only the shocked rats froze. This zero baseline for freezing is important. Just as in salivary conditioning where you do not see salivation to the bell before it is paired with food, and eyelid conditioning where there are no eyeblinks to the tone before it is paired with an airpuff, freezing does not occur to contextual cues in the absence of shock. Fig. 1 shows the data for context dependency collapsed over the delay manipulation. While freezing is greater with the
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
Fig. 3. Post-shock freezing and defecation are presented as a function of the time (in seconds) between placement in the chamber and delivery of a single shock. Based on [17].
stronger shock, at both intensities testing rats in a context different from the one they were shocked in greatly reduced freezing. This joins earlier demonstrations [5,8] that clearly show freezing is a CR to the contextual cues present at the time of shock. To reveal time dependency, Fig. 2 shows the same data collapsed over the context manipulation. If freezing was an UR to the shock then more freezing in the immediate test than the 24-h delayed test would be expected and this was clearly not the case. Indeed at the stronger shock intensity there was a significant difference in the opposite direction. Freezing in shock related situations is clearly a Pavlovian CR and not a UR. A compelling convergent line of evidence is provided by manipulation of the time the rat spends in a context prior to a shock [7,16,17,25]. Typically a few minutes are given to explore the chamber prior to shock delivery and this period is essential for context conditioning. If this preshock period is very short there will be no conditioning to the context and savings tests have shown that this is a deficit in acquisition of conditioning [17]. This phenomenon has come to be called the immediate shock deficit, although it is also obtained when a startle stimulus is used as a US to condition fear [24]. Fig. 3 illustrates this effect, which is apparent for both freezing and defecation measures of fear. Note that all animals received the same shock and were observed immediately after the shock. They were undisturbed between shock and testing. If there was a freezing UR the rats that received shock at the short intervals between placement in the chamber and shock should freeze but did not. Again shock does not produce an UR of freezing. Fig. 3 shows that the rat must have a rather substantial period of time in the context prior to shock for robust context conditioning. For Pavlovian conditioning this is somewhat surprising. Usually conditioning is maximal when there is only a very brief delay between presentation of the CS and the US [30]. Even at 27 s
75
context conditioning was weak, less than 1/4 of what it was at 81 s. The picture is very different for fear conditioning when shock is signaled by a tone. For example, there is strong one-trial fear conditioning to a tone whose onset is simultaneous with shock [2]. It seems that context places an information processing demand on the animal that requires an atypically long CS–US interval. I suggested that this demand may be because prior to conditioning the animal must first form an integrated mnemonic representation of the many features of the context and that the rat must have this representation in active memory at the time of shock [16,17]. When a rat receives shock the features it perceives at that moment will differ depending on just where it happens to be in the context. Only if it has formed an integrated representation or ‘Gestalt’ of the context can it associate the general context with shock. But forming an integrated representation alone is not enough. This gestalt memory must be activated at the time of shock. Support for this view was provided by the finding that pre-exposing the context alleviated the deficit at moderate CS–US intervals but had no effect at very short intervals [16]. Pre-exposure allows the rat to form the integrated representation of the context. But the rat must also have adequate time to sample enough features of the context just before shock to activate this representation. Because a simple stimulus like a tone can be quickly and completely sampled prior to shock, it does not require the same long CS–US intervals as a context.
3. Hippocampal dependency of contextual fear While this interpretation of contextual fear conditioning was derived purely on the basis of behavioral data, it does share some commonality with several theoretical views of the function of the hippocampal formation in learning and memory. Sutherland and Rudy [44] suggested that one function of the hippocampus is to form configurations from available stimulus elements and the integrated representation of the context can be thought of as a configuration of its components. Activating the gestalt memory for the entire context after sampling a few of its features can be thought of as a pattern completion problem, a function that has also been ascribed to the hippocampus [39]. Finally, contexts are at least partially defined spatially and processing of spatial context has also been linked to the hippocampus [37,38]. If the hippocampus is essential to the processes required for the formation and activation of the gestalt memory for context, then the most basic prediction is that hippocampal lesions should selectively interfere with freezing to contextual but not tone CSs. Two laboratories in 1992 independently confirmed this prediction [26,27,40]. While the
76
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
two sets of experiments differed in detail, both found that after receiving tone-shock pairings, freezing to the tone was unaffected by electrolytic dorsal hippocampal lesions. However, freezing to the context where shock occurred was dramatically reduced. This happened even though the conditioning occurred at the same instant for both types of stimuli and the same response measure was used to assess fear to both types of stimuli. Thus, these experiments demonstrated hippocampal-dependent and independent memory with the same preparation. The selective effects of hippocampal lesions on context fear can not be attributed to a difference in the level of conditioning to tone and context. Contexts are differentially affected regardless if contextual freezing is the same, greater or less than tone freezing [4]. Fig. 4 shows data, where despite the fact that context fear was significantly greater than tone fear in intact rats, posttraining hippocampal lesions only affected context fear. It seems reasonable that the special features required of context conditioning derived from the behavioral experiments are related to the selective effects of hippocampal lesions on context fear. That is the hippocampus is involved in the formation of the gestalt memory or the memory for the integrated representation of the context. This fits parsimoniously with both the behavioral conditioning data and the data on hippocampal involvement in animal and human memory.
Studies showing that hippocampal lesions produce a temporally graded retrograde amnesia for contextual fear indicate that the hippocampus is necessary for the formation and/or consolidation of this contextual memory, not the use of it. Lesions made shortly after training produce a very pronounced deficit in context fear, while lesions made a month or more after training produce a greatly reduced deficit [4,26,31]. Because the hippocampus is not needed for the retrieval and use of old contextual fear memories, its role seems more related to processes such as the formation, storage and consolidation of contextual fear. Again, this view fits well with the wealth of data derived from studies of humans with damage to their hippocampal memory systems [41]. Studies using contextual pre-exposure manipulations suggest that hippocampus’ role is related to integrating the features of the context and not to forming a context-shock association (see [21] for a review). If a rat is simply pre-exposed to the context, without shock, about a month before a hippocampal lesion it is protected against both anterograde [50] and retrograde amnesia [3] of contextual fear.
4. Activity, exploration and freezing Hippocampally lesioned rats are known to be hyperactive, especially in environments that promote exploration. Rats invariably explore new environments and
Fig. 4. Rats received tone-shock pairings and one day later an electrolytic lesion of the dorsal hippocampus. One week later the scored for freezing to the tone and to the context. The context test was in the original training context without tone or shock. The to the original training tone but in a novel context. Note that under these training parameters the nonlesioned controls (sham) froze to the context than the tone but that the reverse was true in the lesioned animals (i.e. there was a statistically reliable interaction).
animals were tone test was reliably more Based on [4].
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
Fig. 5. Number of line crossings in a dark open field following electrolytic lesions of the dorsal hippocampus. Based on [4]
this exploratory activity decreases over time. Hippocampal hyperactivity is often manifest as a reduction in the rate of decrease in this exploratory activity. Fig. 5 illustrates this point in rats with electrolytic lesions of the dorsal hippocampus. Initially, hippocampal lesioned rats and sham surgery controls show similar high rates of exploration. The exploratory activity of controls decreases over time. Hippocampal lesioned rats are hyperactive, but it is because they maintain a high level of exploration over the 4-min testing period. This point is of considerable interest to the present analysis because there seems to be three ways in which hyperactivity might relate to contextual fear. One is that hyperactivity simply interferes with freezing. According to this view the deficit in contextual freezing has nothing to do with memory, rather the lesioned rats are physically incapable of freezing. The second is that both the hyperactivity and the contextual deficit are a manifestation of the same memory processing deficit. Perhaps the rat keeps exploring because it can not store an integrated representation of the context in memory. The third possibility is that the two deficits are unrelated. I will evaluate each of these three possibilities in turn.
4.1. Interference Blanchard et al. [6] suggested that some aspects of the hippocampal syndrome, especially altered performance in aversive learning situations may stem from the lesioned animal’s inability to ‘hold still.’ This is a version of the view that the hippocampus functions as a response-inhibition system (e.g. [11,12,28]). Note that this view never fit well with studies of humans that indicated that hippocampal damage results principally in a dysfunction of memory [41]. This view fell out of favor with the development of theories and data that could integrate the animal findings with impairment of certain aspects of memory [36,38]. However, recently,
77
this response-inhibition account has been revived as an explanation of the deficit in contextual fear conditioning [22,23,33]. The concept is simple enough; hyperactivity or the rat’s inability to hold still, generates behaviors that interfere with the animal’s ability to freeze. According to this view, the deficit observed in context conditioning studies is not a failure of learning or memory but rather a performance deficit in the index of fear. The evidence for this analysis comes from two sources. One is a brief report that hippocampal lesions did not produce a deficit in context conditioning when potentiated startle was used as an index of fear [33]. McNish et al. [33] speculated that hyperactivity might interfere with freezing more than potentiated startle. The second was the finding that dorsal hippocampal lesions produced anterograde amnesia and hyperactivity but entorhinal cortex lesions [23] produced neither. Thus there appeared to be a correlation between hyperactivity and freezing when the locus of damage to the hippocampal system was varied. This view received systematic evaluation in a series of papers and was found to be completely untenable [3,4,32]. Here I will just briefly underscore some of the evidence that runs counter to this hypothesis. The strongest evidence against the response interference hypothesis is the selectivity of the deficit. Lesions do not affect tone fear and they only affect context fear if the lesion was made shortly after training [26]. Anagnostaras et al. [4] used a within-subjects design were each rat had both a new and old memory for fear of both tones and contexts. We accomplished this by training rats in two different contexts with shocks signaled by two different tones where the two training episodes were 50 days apart. Hippocampal lesions made a day after the second training episode only reduced freezing to the recently trained context. The lesioned rats were clearly capable of freezing at the level of sham operated controls and did so for the old context-fear memory and for both tones. The presence of the deficit was wholly dependent on the nature of the to-be-remembered information. Another line of evidence is that measures of hyperactivity are poor predictors of the level of context conditioning [31,32]. Additionally, pretraining administration of NMDA antagonists produce similar effects to hippocampal lesions even though the rats are tested in the absence of drug [20,50]. It seems clear that performance based accounts can not explain the very selective pattern of results found in conditioned fear. Additionally, a memory based account for contextual fear fits more parsimoniously with the data for hippocampal lesions obtained with other animal and human preparations [41]. The reason that McNish et al. [33] failed to observe a deficit in contextual conditioning is unclear. One possibility that was not assessed by McNish et al. was
78
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
that hippocampal lesions affected baseline startle responding, a result that has been reported previously [10,45]. An enhancement of baseline startle would have compromised the ability to detect a hippocampal lesion-induced deficit with this measure. Furthermore, hippocampal lesions produce deficits in context conditioning with assays such as conditioned taste aversion and fear-induced defecation [1,43]. It is not clear why hyperactivity should influence these measures of context conditioning.
4.2. Common deficit hypothesis Because hippocampal lesions produce deficits in both contextual fear and cause hyperactivity a parsimonious account would be that both stem from the same underlying memory deficit. Hippocampal hyperactivity is not always apparent. It most clearly presents as enhanced exploration in novel environments (e.g. [35]). Indeed, Nadel [35] has suggested that this hyperactivity may be a manifestation of the failure of lesioned animals to form a ‘cognitive map’ of their spatial location [37,38]. From this vantage it is easy to derive the hypothesis that both hyperactivity and the deficit for contextual fear conditioning produced by hippocampal damage derive from a common source. When rats are placed in a novel environment they explore until they have adequate exposure to the context to form an integrated memory of the context. Two things happen when this occurs. One is that exploration decreases and the other is that contextual fear conditioning can proceed if the appropriate conditions are met (e.g. a shock US is delivered). According to this view [15] the lesioned rat is hyperactive because, unable to commit an integrated representation of the context to memory, it continues to explore. This is the same reason context conditioning is impaired. This account anticipates the Good and Honey [23] result because it suggests that structures involved in contextual fear conditioning are those that produce hyperactivity. It also anticipates the exploration data presented in Fig. 5. When placed in a novel open field both lesioned and control rats start out with a high level of exploration because they have not had the opportunity to form a contextual gestalt memory. Exploratory activity diminishes in the controls as that representation is formed. Activity remains high in the lesioned rats because they never form this memory. Another correlation between exploratory activity and contextual fear conditioning adds support to this view. Pretraining treatment with the NMDA antagonist APV prevents contextual fear conditioning and also causes sustained exploration in the open field [15]. Earlier, I suggested that the time course of the immediate shock deficit (Fig. 3) reflects the time it takes for the rat to form the integrated representation of context.
If the formation of this memory is also responsible for the normal decrease in exploratory activity there should be some similarity between the control curve in Fig. 5 and the placement-to-shock interval curve in Fig. 3. Note that the largest change in exploratory activity is between the 1st and 2nd minute (see Fig. 5) and the largest change in the placement-to-shock interval curve is between 20 and 81 s. So there seems to be a least a rough correspondence between the two functions with the bulk of context processing being completed at 1 min. At this point this correspondence is promising but further work is absolutely necessary. First, the two data sets come from different contexts — a large open field and a small conditioning chamber. Furthermore, a more temporally refined index of exploration needs to be obtained simultaneously with generation of the placement-to-shock interval curve.
4.3. Dissociations between exploration and context conditioning The two hypotheses just offered provide causal links between the context conditioning deficit and increased exploratory activity in hippocampally lesioned rats. Therefore, both views predict a high negative correlation between context freezing and exploratory activity. There is reason to question the strength of this relationship. We [32] examined 48 rats with hippocampal lesions for individual differences in exploratory activity and contextual fear in a conditioning chamber. The rats were placed in a novel chamber and crossovers from one side of the chamber to the other were taken for 3 min as a measure of exploration. The rats then received shock and were tested for freezing upon return to the context the next day. As expected the lesioned rats showed increased exploration and reduced freezing relative to an equally large sample of sham controls. However, there was no significant relationship between exploration and freezing (r= − 0.22; PB 0.10). While this is a null result, the nature of the study makes it difficult to dismiss. The sample size was very large. Additionally, the ranges of the exploration and freezing scores were large and not truncated. Hippocampal lesioned rats varied between 100 and 0% freezing and thus used the full range of possible scores. The range of crossover scores of the hippocampal lesioned rats was 157% of the sham controls. Therefore, it seems unlikely that a common mechanism can account for the large individual differences seen in lesion-induced hyperactivity and the deficit in anterograde amnesia for contextual fear. A further dissociation comes from studies that made excitotoxic lesions of the dorsal hippocampus. Such lesions cause deficits in contextual fear memory [31,50] but do not produce the hyperactivity that is manifest
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
with electrolytic lesions. Fig. 6 shows that rats with NMDA-induced lesions of the hippocampus do not differ from shams in the same test of exploration that revealed deficits with electrolytic lesions. Electrolytic lesions may interrupt subiculo-accumbens projections and these may mediate the affects of these lesions on exploratory activity [34,48,49]. These pathways may or may not contribute to contextual fear conditioning. However, cells within the dorsal hippocampus may account for most of the deficit in contextual fear memory that results from either electrolytic or excitotoxic hippocampal lesions.
5. A further complication and potential resolution In the above correlational study, I pointed out that in rats with hippocampal lesions made prior to training were quite variable in the levels of contextual fear. Overall there was a deficit and indeed some rats showed a complete deficit (no freezing). However, there were also many rats that showed substantial levels of conditioning (e.g. up to 100% freezing). This variability seems to be specific to anterograde lesion studies. Lesions made shortly after training produce a consistently large retrograde amnesia for contextual fear conditioning. A clear example of this dissociation of anterograde and retrograde amnesia for contextual fear is strikingly apparent in a study by Maren et al. [31]. We made excitotoxic dorsal hippocampal lesions either 1 week before or one day after training with three tone-shock pairings. While there was a very large retrograde amnesia there was no anterograde amnesia. In some ways this discrepancy between anterograde and retrograde amnesia for contextual fear should not be completely surprising. The fact that rats with hippocampal lesions readily acquire tone fear indicates that rats can acquire fear memories independent of the
79
hippocampal memory system. Under some circumstances the hippocampal-independent system might be able to acquire fear of some aspects of the context. Recently, I proposed a model that suggested that these two forms of memory may compete for acquisition of fear with the hippocampal system normally winning out [13]. According to such a model contextual fear is usually acquired by the hippocampal system and therefore depriving the subject of the hippocampus after training is devastating to retention of contextual fear. If the animal is deprived of the hippocampal system prior to conditioning, the hippocampal-independent system does not suffer from the normal associative competition. This allows the nonhippocampal system to acquire fear, so pretraining lesions have a much smaller impact on conditioning. This model explains many of the complexities that arise from studies examining the effect of lesions on contextual fear [13]. By itself the model does not explain the increased exploratory activity that occurs following electrolytic lesions of the dorsal hippocampus. Above, evidence was reviewed that the increased locomotion following electrolytic lesions of the dorsal hippocampus may result from damage to fibers running from the subiculum to the nucleus accumbens. Indeed, Legault and Wise [29] suggested that glutamatergic projections from the subiculum to the accumbens regulate exploratory behavior and damage to these projections results in enhanced locomotion mediated by increased accumbal dopamine. It seems likely that aberrant exploration would mean that any neural structures processing contextual stimuli would receive degraded information. Consistent with this idea, Westbrook et al. [47] reported that the nucleus accumbens plays a selective role in contextual fear conditioning. If electrolytic lesions of the hippocampus interrupt this pathway it would mean that even the nonhippocampal system would be receiving degraded contextual information. This may explain why electrolytic dorsal hippocampal lesions produce greater anterograde amnesia than excitotoxic lesions of the same structure [31].
5.1. Summary of the model
Fig. 6. Number of line crossings in a dark open field following NMDA-induced lesions of the dorsal hippocampus. Previously unpublished data collected by B. Godsil and B. Wiltgen.
Upon placement in a novel chamber rats explore that chamber and this behavior provides them with information about the myriad features comprising the context. This exploration allows the animal to store an integrated representation of the context and once that integrated representation is formed exploration decreases. This contextual representation is also necessary for contextual fear conditioning, so the animal must be given sufficient time to explore the chamber prior to shock delivery. A subiculo-accumbens pathway regulates this exploratory activity. When that pathway is
80
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
interrupted by fiber destroying electrolytic lesions it results in enhanced exploratory activity. Fiber sparing excitotoxic lesions of the dorsal hippocampus do not damage the pathway and therefore do not cause the increases in exploratory behavior. Normally, the dorsal hippocampus forms and stores the memory of the contextual representation. Therefore, both excitotoxic and electrolytic post-training lesions of the dorsal hippocampus interfere with context conditioning. Over time and with the aid of the hippocampus, this memory becomes permanently stored in other, probably cortical, structures. Thus if sufficient time elapses between training and lesion there is little effect of a hippocampal lesion on contextual conditioning. That is, there is a time dependent retrograde amnesia for contextual fear. The hippocampus rapidly and effectively forms and stores this contextual memory so that when a shock occurs it is associated with the hippocampal representation. In an animal with a hippocampal lesion this representation is unavailable. Under such circumstances, contextual information that is present in other structures can become associated with the shock. Normally, this does not happen because these other sources of contextual information are overshadowed by the hippocampal representation. So pretraining lesions of the hippocampus, be they excitotoxic or electrolytic, serve to free these other sources of contextual information from associative competition. Thus, fiber sparing pretraining lesions of the dorsal hippocampus produce little anterograde amnesia for contextual fear. Because electrolytic lesions disturb contextual exploration even the extrahippocampal system receives poor information about the context. Thus electrolytic lesions produce a greater anterograde deficit than excitotoxic lesions. However, the greatest loss of contextual fear is observed with post-training lesions.
Acknowledgements This research was supported by National Science Foundation Grant IBN-9723295.
References [1] Aguado L, Hall G, Harrington N, Symonds M. Illness-induced context aversion learning in rats with lesions of the dorsal hippocampus. Behav Neurosci 1998;112:1142–51. [2] Albert M, Ayres JJB. One-trial simultaneous and backward excitatory fear conditioning in rats: lick suppression, freezing, and rearing to CS compounds and their elements. Anim Learn Behav 1997;25:210 –20. [3] Anagnostaras SG, Gale GD, Fanselow MS. The hippocampus and contextual fear conditioning: recent controversies and advances, Hippocampus, in press.
[4] Anagnostaras SG, Maren S, Fanselow MS. Temporally graded retrograde amnesia of contextual fear after hippocampal damage in rats: within-subjects examination. J Neurosci 1999;19:1106– 14. [5] Blanchard RJ, Blanchard DC. Crouching as an index of fear. J Comp Physiol Psychol 1969;67:370 – 5. [6] Blanchard DC, Blanchard RJ, Lee EM, Fukunaga KK. Movement arrest and the hippocampus. Physiol Psychol 1977;5:331–5. [7] Blanchard RJ, Fukunaga KK, Blanchard DC. Environmental control of defensive reactions to footshock. Bull Psychonom Soc 1976;8:129 – 30. [8] Bolles RC, Collier AC. The effect of predictive cues on freezing in rats. Anim Learn Behav 1976;4:6 – 8. [9] Cohen NJ, Poldrack RA, Eichenbaum H. Memory for items and memory for relations in the procedural/declarative memory framework. In: Mayes AR, Downes JJ, editors. Theories of Organic Amnesia. East Sussex: Psychology Press, 1997:131–78. [10] Coover GD, Levine S. Auditory startle response of hippocampectomized rats. Physiol Behav 1972;9:75 – 7. [11] Douglas RJ. Pavlovian conditioning and the brain. In: Boakes RA, Halliday MS, editors. Inhibition and Learning. New York: Academic Press, 1972:529 – 53. [12] Douglas RJ. The hippocampus and behavior. Psychol Bull 1967;67:416 – 42. [13] Fanselow MS. Learning theory and neuropsychology: configuring their disparate elements in the hippocampus. J Exp Psychol Anim Behav Process 1999;25:275 – 83. [14] Fanselow MS. Species-specific defense reactions: retrospect and prospect. In: Bouton ME, Fanselow MS, editors. Learning, Motivation, and Cognition: The Functional Behaviorism of Robert C. Bolles. Washington, DC: American Psychological Association, 1997:321 – 41. [15] Fanselow MS. Modality-specific memory of fear: differential involvement of the amygdala and hippocampal formation in Pavlovian fear conditioning. In: Ono T, McNaughton BL, Molotchnikoff, Rolls ET, Nishijo H, editors. Perception, Memory and Emotion: Frontiers in Neuroscience. Oxford: Pergamon, 1996, pp. 499 – 512. [16] Fanselow MS. Factors governing one trial contextual conditioning. Anim Learn Behav 1990;18:264 – 70. [17] Fanselow MS. Associative vs. topographical accounts of the immediate shock freezing deficit in rats: implications for the response selection rules governing species specific defensive reactions. Learn Motiv 1986;17:16 – 39. [18] Fanselow MS. The post-shock activity burst. Anim Learn Behav 1982;10:448 – 54. [19] Fanselow MS. Conditional and unconditional components of post-shock freezing in rats. Pavlov J Biol Sci 1980;15:177–82. [20] Fanselow MS, Kim JJ, Yipp J, De Oca B. Differential effects of an NMDA antagonist, APV, on acquisition of fear of auditory and contextual cues. Behav Neurosci 1994;108:235 – 40. [21] Fanselow MS, Rudy JW. Convergence of experimental and developmental approaches to animal learning and memory processes. In: Carew TJ, Menzel R, Shatz CJ, editors. Mechanistic Relationships Between Development and Learning. New York: Wiley, 1998:15 – 28. [22] Gisquet-Verrier P, Dutrieux G, Richer P, Doyere V. Effects of lesions to the hippocampus on contextual fear: evidence for a disruption of freezing and avoidance behavior but not context conditioning. Behav Neurosci 1999;113:507 – 22. [23] Good M, Honey RC. Dissociable effects of selective lesions to hippocampal subsystems on exploratory behavior, contextual learning, and spatial learning. Behav Neurosci 1997;111:487–93. [24] Kiernan M, Cranney J. Immediate-startle stimulus presentation fails to condition freezing responses to contextual cues. Behav Neurosci 1992;106:121 – 4.
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81 [25] Kiernan MJ, Westbrook RF. Effects of exposure to a to-beshocked environment upon the rat’s freezing response: evidence for facilitation, latent inhibition, and perceptual learning. Q J Exp Psychol B 1993;46:271–88. [26] Kim JJ, Fanselow MS. Modality specific retrograde amnesia of fear following hippocampal lesions. Science 1992;256:675 – 7. [27] Kim JJ, Rison RA, Fanselow MS. Effects of amygdala hippocampus and preiaqueductal gray lesions on short- and long-term contextual fear. Behav Neurosci 1993;107:1093– 8. [28] Kimble DP. Hippocampus and internal inhibition. Psychol Bull 1968;70:285 – 95. [29] Legault M, Wise RA. Injections of N-methyl-D-aspartate into the ventral hippocampus increase extracellular dopamine in the ventral tegmental area and nucleus accumbens. Synapse 1999;31:241 – 9. [30] Mackintosh NJ. A theory of attention: variations in the associability of stimuli with reinforcement. Psychol Rev 1975;82:276 – 98. [31] Maren S, Aharonov G, Fanselow MS. Neurotoxic lesions of the dorsal hippocampus and Pavlovian fear conditioning in rats. Behav Brain Res 1997;88:261–74. [32] Maren S, Anagnostaras SG, Fanselow MS. The startled seahorse: is the hippocampus necessary for contextual fear conditioning? Trends Cogn Sci 1998;2:39–42. [33] McNish KA, Gewirtz JC, Davis M. Evidence of contextual fear after lesions of the hippocampus: a disruption of freezing but not fear-potentiated startle. J Neurosci 1997;17:9353–60. [34] Mogenson GJ. Limbic-motor integration. In: Epstein AN, Morrison AR, editors. Progress in Psychobiology and Physiological Psychology, vol. 12. London: Academic Press, 1987:117– 70. [35] Nadel L. Dorsal and ventral hippocampus lesions and behavior. Physiol Behav 1968;3:891–900. [36] Nadel L, O’Keefe J, Black A. Slam on the brakes: a critique of Altman, Brunner, and Bayer’s response-inhibition model of hippocampal function. Behav Biol 1975;14:151–62. [37] Nadel L, Willner J. Context and conditioning: a place for space. Physiol Psychol 1980;8:218–28. [38] O’Keefe J, Nadel L. The Hippocampus as a Cognitive Map. Oxford: Oxford University Press, 1978, 570 pp.
.
81
[39] O’Reilly RC, McClelland JL. Hippocampal conjunctive encoding, storage, and recall: avoiding a trade-off. Hippocampus 1994;4:661 – 82. [40] Phillips RG, LeDoux JE. Differential contribution of amygdala and hippocampus to cued and contextual fear conditioning. Behav Neurosci 1992;106:274 – 85. [41] Squire LR. Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans. Psychol Rev 1992;99:195 – 231. [42] Squire LR, Knowlton BJ. Memory, hippocampus, and brain systems. In: Gazzaniga MS, editor. The Cognitive Neurosciences. Cambridge, Mass: MIT Press, 1995:825 – 37. [43] Sutherland RJ, McDonald RJ. Hippocampus, amygdala, and memory deficits in rats. Behav Brain Res 1990;37:57 – 79. [44] Sutherland RJ, Rudy RJ. Configural association theory: the role of the hippocampal formation in learning, memory, and amnesia. Psychobiology 1989;17:129 – 44. [45] Tilson HA, Rogers BC, Grimes L, Harry GJ, Peterson NJ, Hong JS, Dyer RS. Time-dependent neurobiological effects of colchicine administered directly into the hippocampus of rats. Brain Res 1987;408:163 – 72. [46] Wagner AR, Brandon SE. Evolution of a structured connectionist model of Pavlovian conditioning (AESOP). In: Klein SB, Mowrer RR, editors. Contemporary Learning Theories: Pavlovian Conditioning and the Status of Learning Theory. Hillsdale, NJ: Erlbaum, 1989:149 – 89. [47] Westbrook RF, Good AJ, Kiernan MJ. Microinjection of morphine into the nucleus accumbens impairs contextual learning in rats. Behav Neurosci 1997;111:996 – 1013. [48] Whishaw IQ, Cassel J-C, Majchrzak M, Cassel S, Will B. ‘Short-stops’ in rats with fimbria-fornix lesions: evidence for change in the mobility gradient. Hippocampus 1994;4:577–82. [49] Whishaw IQ, Jarrard LE. Similarities vs. differences in place learning and circadian activity in rats after fimbria-fornix section or ibotenate removal of hippocampal cells. Hippocampus 1995;5:595 – 604. [50] Young SL, Bohenek DL, Fanselow MS. NMDA processes mediate anterograde amnesia of contextual fear conditioning induced by hippocampal damage: immunization against amnesia by contextual preexposure. Behav Neurosci 1994;108:19 – 29.
Contextual fear, gestalt memories, and the hippocampus Michael S. Fanselow * Department of Psychology and Brain Research Institute, Uni6ersity of California, Los Angeles, CA 90095 -1563, USA Accepted 25 November 1999
Abstract This review examines the relationship between exploration and contextual fear conditioning. The fear acquired to places or contexts associated with aversive events is a form of Pavlovian conditioning. However, an initial period of exploration is necessary to allow the animal to form an integrated memory of the features of the context before conditioning can take place. The hippocampal formation plays a critical role in this process. Cells within the dorsal hippocampus are involved in the formation, storage and consolidation of this integrated representation of context. Projections from the subiculum to the nucleus accumbens regulate the exploration necessary for the acquisition of information about the features of the context. This model explains why electrolytic but not excitotoxic lesions of the dorsal hippocampus cause enhanced exploratory activity but both cause deficits in contextual fear. It also explains why retrograde amnesia of contextual fear is greater than anterograde amnesia. © 2000 Elsevier Science B.V. All rights reserved. Keywords: Hippocampus; Fear; Context conditioning; Freezing; Nucleus accumbens; Subiculum; Exploration; Hyperactivity
1. Introduction It is well accepted that the hippocampal formation plays a role in certain forms of learning and memory [41]. Declarative, relational and spatial memory are ‘hippocampal-dependent,’ but there are a large number of other learning/memory tasks (procedural, implicit) that are hippocampally-independent [9,38,42]. In the last few years, contextual fear conditioning has become a popular assay for hippocampal-dependent learning. This is in part because the fear conditioning task has built into it directly comparable forms of hippocampaldependent and hippocampal-independent memory. However, before the initial reports of the role of the hippocampus in context conditioning, there were behavioral data suggesting some unique aspects of context conditioning and these led to the initial investigations of the role of the hippocampus in context conditioning. Therefore, it may be valuable to reconsider some of the basic behavioral features of contextual fear conditioning — especially those relevant to the role of the hippocampal formation in context conditioning. That is * Tel.: +1-310-2063891; fax: + 1-310-2065895. E-mail address: [email protected] (M.S. Fanselow)
where this paper will begin. It will then review some of the findings that relate hippocampal function to context conditioning and contrast three different views of why hippocampal lesions affect behavior in fear conditioning situations. It will conclude with a synthesis of these views that embraces many of the existing empirical complexities. 2. Environmental control of contextual fear In the simplest version of contextual fear conditioning, a rat is placed in a novel environment. After it explores the chamber for a few minutes it receives aversive electric footshock. Typically the shock is brief (e.g. 0.5–2 s), mild (0.3–1.5 mA) and infrequent (one to four shocks). The behavior observed in the situation is striking [18]. The rat shows vigorous locomotor activity during the shock that persists, in a somewhat diminished form, for a period beyond shock termination. However, this activity burst is replaced gradually by a profound immobility called freezing. During freezing all observable movement, including that of the vibrissae ceases, except that of the flanks related to respiration. Freezing is usually, but not necessarily, in a crouching posture against a wall of the enclosure.
0166-4328/00/$ - see front matter © 2000 Elsevier Science B.V. All rights reserved. PII: S 0 1 6 6 - 4 3 2 8 ( 9 9 ) 0 0 1 8 6 - 2
74
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
The most compelling account of this behavior is in terms of a Pavlovian analysis. The shock is the unconditional stimulus (US) analogous to Pavlov’s meat powder. The unconditional reaction (UR) to the shock is the activity burst, while freezing is the conditional response (CR). In this case the conditional stimulus (CS), which provokes the CR, are the cues that compose the novel chamber (contextual cues). These contextual cues function like the bell in the proverbial Pavlovian conditioning experiment. One striking difference with Pavlov’s experiment is the relationship between the CR and the UR. In Pavlov’s experiment they were both salivation but in fear conditioning the CR (freezing) and UR (activity burst) are strikingly different. When saying that freezing is a Pavlovian response, it means that what controls the response is the relationship between the stimuli in the situation, in this case between context and shock. It also means that freezing
Fig. 1. Rats were tested in either the same or a different context from the one they were shocked in. The data are collapsed over a factor of whether the test occurred immediately or 24 h after shock. Based on [19].
Fig. 2. Rats were tested either immediately or 24 h after shock. The data are from the same test as those of Fig. 1 but are collapsed over the same/different factor. Based on [19].
is not controlled by its effects on the environment, that is, it is not an operant response. With aversively motivated operant behaviors, shock is delivered contingent on a particular behavior. With operant responses such as inhibitory avoidance, rats are given a shock only after they make the response of entering another compartment; shock is under control of the subject. While it is conceivable that rats might freeze in a contextual conditioning experiment because freezing reduced the impact of shock, explicit tests of the influence of operant contingencies on freezing has ruled this out. Even when freezing is explicitly programmed to control shock delivery such contingencies do not control this behavior (see [14]). It should be noted that while other commonly used indices of fear such as potentiated startle are assumed to be Pavlovian, such formal analyses of these responses have never been conducted. But with freezing, we know that we are examining a Pavlovian response. In the above, it was stated that the freezing that follows even a single shock is a CR to the contextual cues present at the time of shock. Another, possibility is that it is an UR but a delayed UR. For example, the UR may have an early and a late component with the activity burst being the early component and freezing the late component [46]. Such an analysis would help reconcile the relationship of the CR and the UR in fear conditioning with that found it other types of Pavlovian conditioning such as salivation and eyeblink. The concept is quite testable. If freezing is a CR it should be CS or context specific; its occurrence must depend on the presence of the cues that were present at the time of shock. If freezing is an UR it should be time locked to delivery of the US; substantial delays between shock and testing should reduce freezing. I undertook an experiment to factorially manipulate these two variables [19]. Rats were placed in a chamber, where they received four mild or moderate intensity shocks; there were also controls that received no shock. The rats were removed from the chamber after the shock. To test for context dependency, the rats were replaced into the same chamber they were shocked in or an alternative (shock and no shock chambers were counterbalanced across animals). To test for time dependency this test occurred either immediately after shock or 24 h later. The data for the 12 groups of this experiment are summarized in Figs. 1 and 2. The first thing that should be noticed is that only the shocked rats froze. This zero baseline for freezing is important. Just as in salivary conditioning where you do not see salivation to the bell before it is paired with food, and eyelid conditioning where there are no eyeblinks to the tone before it is paired with an airpuff, freezing does not occur to contextual cues in the absence of shock. Fig. 1 shows the data for context dependency collapsed over the delay manipulation. While freezing is greater with the
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
Fig. 3. Post-shock freezing and defecation are presented as a function of the time (in seconds) between placement in the chamber and delivery of a single shock. Based on [17].
stronger shock, at both intensities testing rats in a context different from the one they were shocked in greatly reduced freezing. This joins earlier demonstrations [5,8] that clearly show freezing is a CR to the contextual cues present at the time of shock. To reveal time dependency, Fig. 2 shows the same data collapsed over the context manipulation. If freezing was an UR to the shock then more freezing in the immediate test than the 24-h delayed test would be expected and this was clearly not the case. Indeed at the stronger shock intensity there was a significant difference in the opposite direction. Freezing in shock related situations is clearly a Pavlovian CR and not a UR. A compelling convergent line of evidence is provided by manipulation of the time the rat spends in a context prior to a shock [7,16,17,25]. Typically a few minutes are given to explore the chamber prior to shock delivery and this period is essential for context conditioning. If this preshock period is very short there will be no conditioning to the context and savings tests have shown that this is a deficit in acquisition of conditioning [17]. This phenomenon has come to be called the immediate shock deficit, although it is also obtained when a startle stimulus is used as a US to condition fear [24]. Fig. 3 illustrates this effect, which is apparent for both freezing and defecation measures of fear. Note that all animals received the same shock and were observed immediately after the shock. They were undisturbed between shock and testing. If there was a freezing UR the rats that received shock at the short intervals between placement in the chamber and shock should freeze but did not. Again shock does not produce an UR of freezing. Fig. 3 shows that the rat must have a rather substantial period of time in the context prior to shock for robust context conditioning. For Pavlovian conditioning this is somewhat surprising. Usually conditioning is maximal when there is only a very brief delay between presentation of the CS and the US [30]. Even at 27 s
75
context conditioning was weak, less than 1/4 of what it was at 81 s. The picture is very different for fear conditioning when shock is signaled by a tone. For example, there is strong one-trial fear conditioning to a tone whose onset is simultaneous with shock [2]. It seems that context places an information processing demand on the animal that requires an atypically long CS–US interval. I suggested that this demand may be because prior to conditioning the animal must first form an integrated mnemonic representation of the many features of the context and that the rat must have this representation in active memory at the time of shock [16,17]. When a rat receives shock the features it perceives at that moment will differ depending on just where it happens to be in the context. Only if it has formed an integrated representation or ‘Gestalt’ of the context can it associate the general context with shock. But forming an integrated representation alone is not enough. This gestalt memory must be activated at the time of shock. Support for this view was provided by the finding that pre-exposing the context alleviated the deficit at moderate CS–US intervals but had no effect at very short intervals [16]. Pre-exposure allows the rat to form the integrated representation of the context. But the rat must also have adequate time to sample enough features of the context just before shock to activate this representation. Because a simple stimulus like a tone can be quickly and completely sampled prior to shock, it does not require the same long CS–US intervals as a context.
3. Hippocampal dependency of contextual fear While this interpretation of contextual fear conditioning was derived purely on the basis of behavioral data, it does share some commonality with several theoretical views of the function of the hippocampal formation in learning and memory. Sutherland and Rudy [44] suggested that one function of the hippocampus is to form configurations from available stimulus elements and the integrated representation of the context can be thought of as a configuration of its components. Activating the gestalt memory for the entire context after sampling a few of its features can be thought of as a pattern completion problem, a function that has also been ascribed to the hippocampus [39]. Finally, contexts are at least partially defined spatially and processing of spatial context has also been linked to the hippocampus [37,38]. If the hippocampus is essential to the processes required for the formation and activation of the gestalt memory for context, then the most basic prediction is that hippocampal lesions should selectively interfere with freezing to contextual but not tone CSs. Two laboratories in 1992 independently confirmed this prediction [26,27,40]. While the
76
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
two sets of experiments differed in detail, both found that after receiving tone-shock pairings, freezing to the tone was unaffected by electrolytic dorsal hippocampal lesions. However, freezing to the context where shock occurred was dramatically reduced. This happened even though the conditioning occurred at the same instant for both types of stimuli and the same response measure was used to assess fear to both types of stimuli. Thus, these experiments demonstrated hippocampal-dependent and independent memory with the same preparation. The selective effects of hippocampal lesions on context fear can not be attributed to a difference in the level of conditioning to tone and context. Contexts are differentially affected regardless if contextual freezing is the same, greater or less than tone freezing [4]. Fig. 4 shows data, where despite the fact that context fear was significantly greater than tone fear in intact rats, posttraining hippocampal lesions only affected context fear. It seems reasonable that the special features required of context conditioning derived from the behavioral experiments are related to the selective effects of hippocampal lesions on context fear. That is the hippocampus is involved in the formation of the gestalt memory or the memory for the integrated representation of the context. This fits parsimoniously with both the behavioral conditioning data and the data on hippocampal involvement in animal and human memory.
Studies showing that hippocampal lesions produce a temporally graded retrograde amnesia for contextual fear indicate that the hippocampus is necessary for the formation and/or consolidation of this contextual memory, not the use of it. Lesions made shortly after training produce a very pronounced deficit in context fear, while lesions made a month or more after training produce a greatly reduced deficit [4,26,31]. Because the hippocampus is not needed for the retrieval and use of old contextual fear memories, its role seems more related to processes such as the formation, storage and consolidation of contextual fear. Again, this view fits well with the wealth of data derived from studies of humans with damage to their hippocampal memory systems [41]. Studies using contextual pre-exposure manipulations suggest that hippocampus’ role is related to integrating the features of the context and not to forming a context-shock association (see [21] for a review). If a rat is simply pre-exposed to the context, without shock, about a month before a hippocampal lesion it is protected against both anterograde [50] and retrograde amnesia [3] of contextual fear.
4. Activity, exploration and freezing Hippocampally lesioned rats are known to be hyperactive, especially in environments that promote exploration. Rats invariably explore new environments and
Fig. 4. Rats received tone-shock pairings and one day later an electrolytic lesion of the dorsal hippocampus. One week later the scored for freezing to the tone and to the context. The context test was in the original training context without tone or shock. The to the original training tone but in a novel context. Note that under these training parameters the nonlesioned controls (sham) froze to the context than the tone but that the reverse was true in the lesioned animals (i.e. there was a statistically reliable interaction).
animals were tone test was reliably more Based on [4].
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
Fig. 5. Number of line crossings in a dark open field following electrolytic lesions of the dorsal hippocampus. Based on [4]
this exploratory activity decreases over time. Hippocampal hyperactivity is often manifest as a reduction in the rate of decrease in this exploratory activity. Fig. 5 illustrates this point in rats with electrolytic lesions of the dorsal hippocampus. Initially, hippocampal lesioned rats and sham surgery controls show similar high rates of exploration. The exploratory activity of controls decreases over time. Hippocampal lesioned rats are hyperactive, but it is because they maintain a high level of exploration over the 4-min testing period. This point is of considerable interest to the present analysis because there seems to be three ways in which hyperactivity might relate to contextual fear. One is that hyperactivity simply interferes with freezing. According to this view the deficit in contextual freezing has nothing to do with memory, rather the lesioned rats are physically incapable of freezing. The second is that both the hyperactivity and the contextual deficit are a manifestation of the same memory processing deficit. Perhaps the rat keeps exploring because it can not store an integrated representation of the context in memory. The third possibility is that the two deficits are unrelated. I will evaluate each of these three possibilities in turn.
4.1. Interference Blanchard et al. [6] suggested that some aspects of the hippocampal syndrome, especially altered performance in aversive learning situations may stem from the lesioned animal’s inability to ‘hold still.’ This is a version of the view that the hippocampus functions as a response-inhibition system (e.g. [11,12,28]). Note that this view never fit well with studies of humans that indicated that hippocampal damage results principally in a dysfunction of memory [41]. This view fell out of favor with the development of theories and data that could integrate the animal findings with impairment of certain aspects of memory [36,38]. However, recently,
77
this response-inhibition account has been revived as an explanation of the deficit in contextual fear conditioning [22,23,33]. The concept is simple enough; hyperactivity or the rat’s inability to hold still, generates behaviors that interfere with the animal’s ability to freeze. According to this view, the deficit observed in context conditioning studies is not a failure of learning or memory but rather a performance deficit in the index of fear. The evidence for this analysis comes from two sources. One is a brief report that hippocampal lesions did not produce a deficit in context conditioning when potentiated startle was used as an index of fear [33]. McNish et al. [33] speculated that hyperactivity might interfere with freezing more than potentiated startle. The second was the finding that dorsal hippocampal lesions produced anterograde amnesia and hyperactivity but entorhinal cortex lesions [23] produced neither. Thus there appeared to be a correlation between hyperactivity and freezing when the locus of damage to the hippocampal system was varied. This view received systematic evaluation in a series of papers and was found to be completely untenable [3,4,32]. Here I will just briefly underscore some of the evidence that runs counter to this hypothesis. The strongest evidence against the response interference hypothesis is the selectivity of the deficit. Lesions do not affect tone fear and they only affect context fear if the lesion was made shortly after training [26]. Anagnostaras et al. [4] used a within-subjects design were each rat had both a new and old memory for fear of both tones and contexts. We accomplished this by training rats in two different contexts with shocks signaled by two different tones where the two training episodes were 50 days apart. Hippocampal lesions made a day after the second training episode only reduced freezing to the recently trained context. The lesioned rats were clearly capable of freezing at the level of sham operated controls and did so for the old context-fear memory and for both tones. The presence of the deficit was wholly dependent on the nature of the to-be-remembered information. Another line of evidence is that measures of hyperactivity are poor predictors of the level of context conditioning [31,32]. Additionally, pretraining administration of NMDA antagonists produce similar effects to hippocampal lesions even though the rats are tested in the absence of drug [20,50]. It seems clear that performance based accounts can not explain the very selective pattern of results found in conditioned fear. Additionally, a memory based account for contextual fear fits more parsimoniously with the data for hippocampal lesions obtained with other animal and human preparations [41]. The reason that McNish et al. [33] failed to observe a deficit in contextual conditioning is unclear. One possibility that was not assessed by McNish et al. was
78
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
that hippocampal lesions affected baseline startle responding, a result that has been reported previously [10,45]. An enhancement of baseline startle would have compromised the ability to detect a hippocampal lesion-induced deficit with this measure. Furthermore, hippocampal lesions produce deficits in context conditioning with assays such as conditioned taste aversion and fear-induced defecation [1,43]. It is not clear why hyperactivity should influence these measures of context conditioning.
4.2. Common deficit hypothesis Because hippocampal lesions produce deficits in both contextual fear and cause hyperactivity a parsimonious account would be that both stem from the same underlying memory deficit. Hippocampal hyperactivity is not always apparent. It most clearly presents as enhanced exploration in novel environments (e.g. [35]). Indeed, Nadel [35] has suggested that this hyperactivity may be a manifestation of the failure of lesioned animals to form a ‘cognitive map’ of their spatial location [37,38]. From this vantage it is easy to derive the hypothesis that both hyperactivity and the deficit for contextual fear conditioning produced by hippocampal damage derive from a common source. When rats are placed in a novel environment they explore until they have adequate exposure to the context to form an integrated memory of the context. Two things happen when this occurs. One is that exploration decreases and the other is that contextual fear conditioning can proceed if the appropriate conditions are met (e.g. a shock US is delivered). According to this view [15] the lesioned rat is hyperactive because, unable to commit an integrated representation of the context to memory, it continues to explore. This is the same reason context conditioning is impaired. This account anticipates the Good and Honey [23] result because it suggests that structures involved in contextual fear conditioning are those that produce hyperactivity. It also anticipates the exploration data presented in Fig. 5. When placed in a novel open field both lesioned and control rats start out with a high level of exploration because they have not had the opportunity to form a contextual gestalt memory. Exploratory activity diminishes in the controls as that representation is formed. Activity remains high in the lesioned rats because they never form this memory. Another correlation between exploratory activity and contextual fear conditioning adds support to this view. Pretraining treatment with the NMDA antagonist APV prevents contextual fear conditioning and also causes sustained exploration in the open field [15]. Earlier, I suggested that the time course of the immediate shock deficit (Fig. 3) reflects the time it takes for the rat to form the integrated representation of context.
If the formation of this memory is also responsible for the normal decrease in exploratory activity there should be some similarity between the control curve in Fig. 5 and the placement-to-shock interval curve in Fig. 3. Note that the largest change in exploratory activity is between the 1st and 2nd minute (see Fig. 5) and the largest change in the placement-to-shock interval curve is between 20 and 81 s. So there seems to be a least a rough correspondence between the two functions with the bulk of context processing being completed at 1 min. At this point this correspondence is promising but further work is absolutely necessary. First, the two data sets come from different contexts — a large open field and a small conditioning chamber. Furthermore, a more temporally refined index of exploration needs to be obtained simultaneously with generation of the placement-to-shock interval curve.
4.3. Dissociations between exploration and context conditioning The two hypotheses just offered provide causal links between the context conditioning deficit and increased exploratory activity in hippocampally lesioned rats. Therefore, both views predict a high negative correlation between context freezing and exploratory activity. There is reason to question the strength of this relationship. We [32] examined 48 rats with hippocampal lesions for individual differences in exploratory activity and contextual fear in a conditioning chamber. The rats were placed in a novel chamber and crossovers from one side of the chamber to the other were taken for 3 min as a measure of exploration. The rats then received shock and were tested for freezing upon return to the context the next day. As expected the lesioned rats showed increased exploration and reduced freezing relative to an equally large sample of sham controls. However, there was no significant relationship between exploration and freezing (r= − 0.22; PB 0.10). While this is a null result, the nature of the study makes it difficult to dismiss. The sample size was very large. Additionally, the ranges of the exploration and freezing scores were large and not truncated. Hippocampal lesioned rats varied between 100 and 0% freezing and thus used the full range of possible scores. The range of crossover scores of the hippocampal lesioned rats was 157% of the sham controls. Therefore, it seems unlikely that a common mechanism can account for the large individual differences seen in lesion-induced hyperactivity and the deficit in anterograde amnesia for contextual fear. A further dissociation comes from studies that made excitotoxic lesions of the dorsal hippocampus. Such lesions cause deficits in contextual fear memory [31,50] but do not produce the hyperactivity that is manifest
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
with electrolytic lesions. Fig. 6 shows that rats with NMDA-induced lesions of the hippocampus do not differ from shams in the same test of exploration that revealed deficits with electrolytic lesions. Electrolytic lesions may interrupt subiculo-accumbens projections and these may mediate the affects of these lesions on exploratory activity [34,48,49]. These pathways may or may not contribute to contextual fear conditioning. However, cells within the dorsal hippocampus may account for most of the deficit in contextual fear memory that results from either electrolytic or excitotoxic hippocampal lesions.
5. A further complication and potential resolution In the above correlational study, I pointed out that in rats with hippocampal lesions made prior to training were quite variable in the levels of contextual fear. Overall there was a deficit and indeed some rats showed a complete deficit (no freezing). However, there were also many rats that showed substantial levels of conditioning (e.g. up to 100% freezing). This variability seems to be specific to anterograde lesion studies. Lesions made shortly after training produce a consistently large retrograde amnesia for contextual fear conditioning. A clear example of this dissociation of anterograde and retrograde amnesia for contextual fear is strikingly apparent in a study by Maren et al. [31]. We made excitotoxic dorsal hippocampal lesions either 1 week before or one day after training with three tone-shock pairings. While there was a very large retrograde amnesia there was no anterograde amnesia. In some ways this discrepancy between anterograde and retrograde amnesia for contextual fear should not be completely surprising. The fact that rats with hippocampal lesions readily acquire tone fear indicates that rats can acquire fear memories independent of the
79
hippocampal memory system. Under some circumstances the hippocampal-independent system might be able to acquire fear of some aspects of the context. Recently, I proposed a model that suggested that these two forms of memory may compete for acquisition of fear with the hippocampal system normally winning out [13]. According to such a model contextual fear is usually acquired by the hippocampal system and therefore depriving the subject of the hippocampus after training is devastating to retention of contextual fear. If the animal is deprived of the hippocampal system prior to conditioning, the hippocampal-independent system does not suffer from the normal associative competition. This allows the nonhippocampal system to acquire fear, so pretraining lesions have a much smaller impact on conditioning. This model explains many of the complexities that arise from studies examining the effect of lesions on contextual fear [13]. By itself the model does not explain the increased exploratory activity that occurs following electrolytic lesions of the dorsal hippocampus. Above, evidence was reviewed that the increased locomotion following electrolytic lesions of the dorsal hippocampus may result from damage to fibers running from the subiculum to the nucleus accumbens. Indeed, Legault and Wise [29] suggested that glutamatergic projections from the subiculum to the accumbens regulate exploratory behavior and damage to these projections results in enhanced locomotion mediated by increased accumbal dopamine. It seems likely that aberrant exploration would mean that any neural structures processing contextual stimuli would receive degraded information. Consistent with this idea, Westbrook et al. [47] reported that the nucleus accumbens plays a selective role in contextual fear conditioning. If electrolytic lesions of the hippocampus interrupt this pathway it would mean that even the nonhippocampal system would be receiving degraded contextual information. This may explain why electrolytic dorsal hippocampal lesions produce greater anterograde amnesia than excitotoxic lesions of the same structure [31].
5.1. Summary of the model
Fig. 6. Number of line crossings in a dark open field following NMDA-induced lesions of the dorsal hippocampus. Previously unpublished data collected by B. Godsil and B. Wiltgen.
Upon placement in a novel chamber rats explore that chamber and this behavior provides them with information about the myriad features comprising the context. This exploration allows the animal to store an integrated representation of the context and once that integrated representation is formed exploration decreases. This contextual representation is also necessary for contextual fear conditioning, so the animal must be given sufficient time to explore the chamber prior to shock delivery. A subiculo-accumbens pathway regulates this exploratory activity. When that pathway is
80
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81
interrupted by fiber destroying electrolytic lesions it results in enhanced exploratory activity. Fiber sparing excitotoxic lesions of the dorsal hippocampus do not damage the pathway and therefore do not cause the increases in exploratory behavior. Normally, the dorsal hippocampus forms and stores the memory of the contextual representation. Therefore, both excitotoxic and electrolytic post-training lesions of the dorsal hippocampus interfere with context conditioning. Over time and with the aid of the hippocampus, this memory becomes permanently stored in other, probably cortical, structures. Thus if sufficient time elapses between training and lesion there is little effect of a hippocampal lesion on contextual conditioning. That is, there is a time dependent retrograde amnesia for contextual fear. The hippocampus rapidly and effectively forms and stores this contextual memory so that when a shock occurs it is associated with the hippocampal representation. In an animal with a hippocampal lesion this representation is unavailable. Under such circumstances, contextual information that is present in other structures can become associated with the shock. Normally, this does not happen because these other sources of contextual information are overshadowed by the hippocampal representation. So pretraining lesions of the hippocampus, be they excitotoxic or electrolytic, serve to free these other sources of contextual information from associative competition. Thus, fiber sparing pretraining lesions of the dorsal hippocampus produce little anterograde amnesia for contextual fear. Because electrolytic lesions disturb contextual exploration even the extrahippocampal system receives poor information about the context. Thus electrolytic lesions produce a greater anterograde deficit than excitotoxic lesions. However, the greatest loss of contextual fear is observed with post-training lesions.
Acknowledgements This research was supported by National Science Foundation Grant IBN-9723295.
References [1] Aguado L, Hall G, Harrington N, Symonds M. Illness-induced context aversion learning in rats with lesions of the dorsal hippocampus. Behav Neurosci 1998;112:1142–51. [2] Albert M, Ayres JJB. One-trial simultaneous and backward excitatory fear conditioning in rats: lick suppression, freezing, and rearing to CS compounds and their elements. Anim Learn Behav 1997;25:210 –20. [3] Anagnostaras SG, Gale GD, Fanselow MS. The hippocampus and contextual fear conditioning: recent controversies and advances, Hippocampus, in press.
[4] Anagnostaras SG, Maren S, Fanselow MS. Temporally graded retrograde amnesia of contextual fear after hippocampal damage in rats: within-subjects examination. J Neurosci 1999;19:1106– 14. [5] Blanchard RJ, Blanchard DC. Crouching as an index of fear. J Comp Physiol Psychol 1969;67:370 – 5. [6] Blanchard DC, Blanchard RJ, Lee EM, Fukunaga KK. Movement arrest and the hippocampus. Physiol Psychol 1977;5:331–5. [7] Blanchard RJ, Fukunaga KK, Blanchard DC. Environmental control of defensive reactions to footshock. Bull Psychonom Soc 1976;8:129 – 30. [8] Bolles RC, Collier AC. The effect of predictive cues on freezing in rats. Anim Learn Behav 1976;4:6 – 8. [9] Cohen NJ, Poldrack RA, Eichenbaum H. Memory for items and memory for relations in the procedural/declarative memory framework. In: Mayes AR, Downes JJ, editors. Theories of Organic Amnesia. East Sussex: Psychology Press, 1997:131–78. [10] Coover GD, Levine S. Auditory startle response of hippocampectomized rats. Physiol Behav 1972;9:75 – 7. [11] Douglas RJ. Pavlovian conditioning and the brain. In: Boakes RA, Halliday MS, editors. Inhibition and Learning. New York: Academic Press, 1972:529 – 53. [12] Douglas RJ. The hippocampus and behavior. Psychol Bull 1967;67:416 – 42. [13] Fanselow MS. Learning theory and neuropsychology: configuring their disparate elements in the hippocampus. J Exp Psychol Anim Behav Process 1999;25:275 – 83. [14] Fanselow MS. Species-specific defense reactions: retrospect and prospect. In: Bouton ME, Fanselow MS, editors. Learning, Motivation, and Cognition: The Functional Behaviorism of Robert C. Bolles. Washington, DC: American Psychological Association, 1997:321 – 41. [15] Fanselow MS. Modality-specific memory of fear: differential involvement of the amygdala and hippocampal formation in Pavlovian fear conditioning. In: Ono T, McNaughton BL, Molotchnikoff, Rolls ET, Nishijo H, editors. Perception, Memory and Emotion: Frontiers in Neuroscience. Oxford: Pergamon, 1996, pp. 499 – 512. [16] Fanselow MS. Factors governing one trial contextual conditioning. Anim Learn Behav 1990;18:264 – 70. [17] Fanselow MS. Associative vs. topographical accounts of the immediate shock freezing deficit in rats: implications for the response selection rules governing species specific defensive reactions. Learn Motiv 1986;17:16 – 39. [18] Fanselow MS. The post-shock activity burst. Anim Learn Behav 1982;10:448 – 54. [19] Fanselow MS. Conditional and unconditional components of post-shock freezing in rats. Pavlov J Biol Sci 1980;15:177–82. [20] Fanselow MS, Kim JJ, Yipp J, De Oca B. Differential effects of an NMDA antagonist, APV, on acquisition of fear of auditory and contextual cues. Behav Neurosci 1994;108:235 – 40. [21] Fanselow MS, Rudy JW. Convergence of experimental and developmental approaches to animal learning and memory processes. In: Carew TJ, Menzel R, Shatz CJ, editors. Mechanistic Relationships Between Development and Learning. New York: Wiley, 1998:15 – 28. [22] Gisquet-Verrier P, Dutrieux G, Richer P, Doyere V. Effects of lesions to the hippocampus on contextual fear: evidence for a disruption of freezing and avoidance behavior but not context conditioning. Behav Neurosci 1999;113:507 – 22. [23] Good M, Honey RC. Dissociable effects of selective lesions to hippocampal subsystems on exploratory behavior, contextual learning, and spatial learning. Behav Neurosci 1997;111:487–93. [24] Kiernan M, Cranney J. Immediate-startle stimulus presentation fails to condition freezing responses to contextual cues. Behav Neurosci 1992;106:121 – 4.
M.S. Fanselow / Beha6ioural Brain Research 110 (2000) 73–81 [25] Kiernan MJ, Westbrook RF. Effects of exposure to a to-beshocked environment upon the rat’s freezing response: evidence for facilitation, latent inhibition, and perceptual learning. Q J Exp Psychol B 1993;46:271–88. [26] Kim JJ, Fanselow MS. Modality specific retrograde amnesia of fear following hippocampal lesions. Science 1992;256:675 – 7. [27] Kim JJ, Rison RA, Fanselow MS. Effects of amygdala hippocampus and preiaqueductal gray lesions on short- and long-term contextual fear. Behav Neurosci 1993;107:1093– 8. [28] Kimble DP. Hippocampus and internal inhibition. Psychol Bull 1968;70:285 – 95. [29] Legault M, Wise RA. Injections of N-methyl-D-aspartate into the ventral hippocampus increase extracellular dopamine in the ventral tegmental area and nucleus accumbens. Synapse 1999;31:241 – 9. [30] Mackintosh NJ. A theory of attention: variations in the associability of stimuli with reinforcement. Psychol Rev 1975;82:276 – 98. [31] Maren S, Aharonov G, Fanselow MS. Neurotoxic lesions of the dorsal hippocampus and Pavlovian fear conditioning in rats. Behav Brain Res 1997;88:261–74. [32] Maren S, Anagnostaras SG, Fanselow MS. The startled seahorse: is the hippocampus necessary for contextual fear conditioning? Trends Cogn Sci 1998;2:39–42. [33] McNish KA, Gewirtz JC, Davis M. Evidence of contextual fear after lesions of the hippocampus: a disruption of freezing but not fear-potentiated startle. J Neurosci 1997;17:9353–60. [34] Mogenson GJ. Limbic-motor integration. In: Epstein AN, Morrison AR, editors. Progress in Psychobiology and Physiological Psychology, vol. 12. London: Academic Press, 1987:117– 70. [35] Nadel L. Dorsal and ventral hippocampus lesions and behavior. Physiol Behav 1968;3:891–900. [36] Nadel L, O’Keefe J, Black A. Slam on the brakes: a critique of Altman, Brunner, and Bayer’s response-inhibition model of hippocampal function. Behav Biol 1975;14:151–62. [37] Nadel L, Willner J. Context and conditioning: a place for space. Physiol Psychol 1980;8:218–28. [38] O’Keefe J, Nadel L. The Hippocampus as a Cognitive Map. Oxford: Oxford University Press, 1978, 570 pp.
.
81
[39] O’Reilly RC, McClelland JL. Hippocampal conjunctive encoding, storage, and recall: avoiding a trade-off. Hippocampus 1994;4:661 – 82. [40] Phillips RG, LeDoux JE. Differential contribution of amygdala and hippocampus to cued and contextual fear conditioning. Behav Neurosci 1992;106:274 – 85. [41] Squire LR. Memory and the hippocampus: a synthesis from findings with rats, monkeys, and humans. Psychol Rev 1992;99:195 – 231. [42] Squire LR, Knowlton BJ. Memory, hippocampus, and brain systems. In: Gazzaniga MS, editor. The Cognitive Neurosciences. Cambridge, Mass: MIT Press, 1995:825 – 37. [43] Sutherland RJ, McDonald RJ. Hippocampus, amygdala, and memory deficits in rats. Behav Brain Res 1990;37:57 – 79. [44] Sutherland RJ, Rudy RJ. Configural association theory: the role of the hippocampal formation in learning, memory, and amnesia. Psychobiology 1989;17:129 – 44. [45] Tilson HA, Rogers BC, Grimes L, Harry GJ, Peterson NJ, Hong JS, Dyer RS. Time-dependent neurobiological effects of colchicine administered directly into the hippocampus of rats. Brain Res 1987;408:163 – 72. [46] Wagner AR, Brandon SE. Evolution of a structured connectionist model of Pavlovian conditioning (AESOP). In: Klein SB, Mowrer RR, editors. Contemporary Learning Theories: Pavlovian Conditioning and the Status of Learning Theory. Hillsdale, NJ: Erlbaum, 1989:149 – 89. [47] Westbrook RF, Good AJ, Kiernan MJ. Microinjection of morphine into the nucleus accumbens impairs contextual learning in rats. Behav Neurosci 1997;111:996 – 1013. [48] Whishaw IQ, Cassel J-C, Majchrzak M, Cassel S, Will B. ‘Short-stops’ in rats with fimbria-fornix lesions: evidence for change in the mobility gradient. Hippocampus 1994;4:577–82. [49] Whishaw IQ, Jarrard LE. Similarities vs. differences in place learning and circadian activity in rats after fimbria-fornix section or ibotenate removal of hippocampal cells. Hippocampus 1995;5:595 – 604. [50] Young SL, Bohenek DL, Fanselow MS. NMDA processes mediate anterograde amnesia of contextual fear conditioning induced by hippocampal damage: immunization against amnesia by contextual preexposure. Behav Neurosci 1994;108:19 – 29.