1YIPPOCAMPUS, VOL. 1, NO. 3, PAGES 250-252, JULY 1991
Exceptions to the Rule of Space Robert J . Sutherland* and Jerry W. Rudy? *Department of P s y c h o l o g y , The University of Lethbridge, L e t h b r i d g e , A l b e r t a , C a n a d a , T l K 3M4 and ? D e p a r t m e n t of Psychology, Muenzinger Psychology Building, University of Colorado, Boulder, CO 80309 U . S . A .
There are many points in Nadel’s paper with which we heartily agree. At a general level we most strongly agree that a good theory of hippocampal function should be explicit and unambiguous about the kind of information that the hippocampal formation is handling. As Nadel points out. many contemporary positions are limited by their inability to say anything plainly about this issue-they are limited in that they seem to make few clear predictions and seem to lack generality across species and behavioral paradigms. This has not been the case with O’Keefe and Nadel’s position. A key point that should not be lost is their assertion that the hippocampal formation is essential for place learning. that is, for situations in which the topographical relationships i n the environment come to guide an animal’s behavior. This was a very bold claim in the mid-1970s and, to the surprise of most of their contemporaries, the subsequent experimental record is basically unanimous-they were right. Repeatedly, in the hippocampal lesion literature using nonprimates the presence of a specific place learning requirement in a task. as opposed to, for example, working memory, temporary memory buffer. or other nonspatial requirements, has been shown to be critical in detecting a behavioral impairment. For those of us convinced of the basic correctness of this central assertion by O’Keefe and Nadel, more data demonstrating the existence of an intimate link between hippocampal circuitry and place memory do not further sprcjfictil1.v strengthen O’Keefe and Nadel‘s theorv. For example. the demonstrations that certain forms of spatial memory of birds are affected by damage that includes the hippocampal formation will not allow, by themselves, further elucidation of the specific contribution of hippocampal circuitry in solving spatial problems. The same holds for the association between variations in the size of hippocampal components and variations in performance in hippocampal-sensitive tasks. such as two-way active avoidance. These new data do not add a novel kind of support for O’Keefe and Nadel‘s explanation for ~ ’ h the hippocampal formation is important for place learning. The nub of our story is this: O’Keefe and Nadel assert that instances of impaired learning and memory in animals with damage to the hippocampal formation are caused by the elimination of an explicitly topographical, mnemonic representation of the environment, that the kind of information that the hippocampal formation handles is necessarily spatial; we part company with them here. We have chosen a different, although not unrelated, path in accounting for the specific contribution that the hippocampal formation makes to learn-
ing and memory. On reviewing the hippocampal literature, we were impressed with two things: place learning is impaired after HPC damage, and there are several instances of impairments that do not appear to fit well with the idea that the hippocampal formation makes its special contribution to memory by constructing and storing spatial maps of the environment. We have already dealt in detail with these “exceptions” (Sutherland and Rudy, 1989).Nadel acknowledges that the last 15 years of experiments on the hippocampal formation have produced “a small number of exceptions” to his position. It is in this small, unassimilated fringe (which Nadel also acknowledges contains one of the human hippocampi), where we set up camp. We felt motivated t o find some way of characterizing the kind of information unique to the hippocampal formation so that both the place learning impairment and the “exceptions” could be explained in the same manner. It was also important to us that our way of chara c t e r i h g hippocampal function should predict new behavioral situations in which hippocampal circuitry makes an essential contribution. We were also fortunate to have read Hirsh’s (1974) paper, which prompted us to believe that following the path we have chosen was not altogether unreasonable. Our position is that hippocampal circuitry is necessary if a problem cannot be solved on the basis of strengthening o r weakening associations between elementary stimulus events, that is, if an animal must form associations between cue conjunctions and some other event. Topographical relationships are but one example of the many kinds of cue conjunctions or relationships that are possible. Essentially. the hippocampal formation enables an animal to disambiguate the significance of an elemental stimulus o r cue when the meaning of that cue depends upon its relationship to one o r more other cues. Place learning falls into the category of impaircd abilities precisely because if an animal must navigate to a specific location using topographical relationships among cues. then the way that an animal should move when it is facing any particular cue is ambiguous: the cue’s significance for guidance necessarily depends upon its relationship to some other element of the environment. starting location. other perceptible cues, etc. If the animal can solve the spatial problem by using only a single element of the situation t o guide its movements, then by definition this would require only it taxon strategy for solution, or, to use our terminology. a simple association solution. Nadel states quite clearly that he is not averse to the idea that one 01‘ the human hippocampi represents information that is more “abstract” than physical space. We are saying something similar about the hippocampal formation in both hemispheres of all species-primates y to birds to fish. Species differ dramatically in the kinds of perceptual o r motor information represented by activity in ensembles in neocortical zones and other structures that provide input to the hippocampal formation. ’Therefore, the kinds of cues or events that can enter into hippocampal-based configural associations may differ dramatically among species. Our view clearly predicts that it should be possible t o find examples of impaired configural memory after hippocampal damage, in addition to. and quite apart from, those involving spatial mapping-this search is of course hopeless if O’Keefe and Nadel’s theory is more in line with hippocampal function.
250
EXCEPTIONS TO THE RULE OF SPACE / Sutherland and Rudy We will describe three examples of memory impairments that d o not confirm the spatial mapping position, but that are predicted by configural association theory (Sutherland and Rudy, 1989). If our interpretation of these results is correct, it is probably the case that, at least in mammals, the memory deficit after hippocampal damage includes relational information that is both spatial and nonspatial. Our first example is that of negative patterning discrimination. In the version of this discrimination problem that we have studied, animals are rewarded with food for responding in the presence of a light or a tone, but are not rewarded for responding if the light and tone occur together. They readily learn to respond quickly and consistently to either cue alone, and to withhold responding to the compound light + tone. The rats clearly cannot be responding on the basis of the reinforcement history of either of the two cues alone-if that were the case, responding to the compound could never be lower than to the individual cues-they must be using relational information to withhold response to the compound. We predicted that rats with hippocampal damage would be unable to learn or remember this discrimination. We repeatedly have confirmed these predictions of the configural position (Rudy and Sutherland, 1989; Sutherland et al., 1989a; 1989b; Sutherland and Rudy, 1989; Sutherland and McDonald, 1990). Rats with neurotoxin-induced damage to the hippocampal formation respond as readily to the compound as to the individual elements, whether training occurs only after or both before and after surgery. If there is a satisfactory spatial mapping account of these results, we have not heard it. For example, to suggest that in the case of configural discriminations (and not simple discriminations using the same cues) the relevant associations must be “embedded” in an essentially spatial representation and therefore be sensitive to hippocampal damage is the kind of hypothesis drift that Nadel so poignantly abhors. A second example is the transverse patterning discrimination. The problem requires the animal to visually discriminate between pairs of arms in a T-maze set within a swimming pool. On every trial the rat must choose to swim into one of two arms; the arms can be white (W), black (B), or striped (S). Training is divided into three phases. In phase 1 the rats receive a simultaneous choice between white and black arms, with the white arm always containing the goal. In phase 2, the rats continue to receive white vs. black trials but, in addition, they receive black vs. striped trials, with the black arm always containing the goal. In the final phase, rats receive white vs. black trials, black vs. striped trials, and, in addition, striped vs. white trials, with the striped arm always containing the goal (thus, W + B - / B + S - / S + W - ) . It is only by the third phase that the cues are ambiguous for the rat. In order to solve the problem through phase 3 the rat must use the relationship between cues in the arms: prior to that point, the rat can use simple associations involving the cues to discriminate correctly. Damage to the hippocampal formation disrupts acquisition and retention of the solution to the transverse patterning problem, without disrupting the acquisition of the simple associations at early phases of testing (Alvarado and Rudy, 1989). Again, it is difficult to come up with a spatial mapping account of these results, but these
251
experiments were designed to test straightforward predictions of the configural position. A third example comes from Y-maze experiments conducted on dry land. If a rat is required t o choose an arm whose features match the features of the start arm, o r to avoid the arm whose features match the features of the start arm, there appears to be very little, if any, effect of damage to the hippocampal formation (Aggleton et al., 1986; Sutherland et al., 1989b; Sutherland and McDonald, 1990). These experiments have involved simple visual, tactile, and odor cues (obviously, if the relevant feature was spatial, the task would be hippocampal-sensitive). If, on the other hand, the task is modified so as to force the rat to use a nonspatial relationship between features in start and goal arms to make correct choices, then damage to the hippocampal formation disrupts performance. We have modified this basic task to force a configural solution in three different ways in separate experiments: ( I ) Rats had to choose between a black and a white goal arm. If the start box was illuminated at the beginning of a trial, then the white arm contained the goal: if the start box was not illuminated, then the black arm was correct (Sutherland et al.. 1989b). (2) Rats had to choose between arms that were black. white, or striped. If the start box contained odor I , the rat had to avoid black, if odor 2, then avoid white, and if odor 3, then avoid stripes (for more details see Sutherland et al., 1989b; Sutherland and McDonald, 1990). (3) Rats had to learn a configural black vs. white discrimination based upon time of day (Sutherland et al., l989a). In the morning choosing the black and not the white arm was rewarded; in the evening choosing the white and not the black arm was rewarded. In each of these Y-maze experiments the control rats solved the relevant configural discriminations, but in none of them did the rats with hippocampal damage successfully discriminate. These three examples do not provide an exhaustive list of exceptions to the necessity for a spatial mapping requirement in hippocampal-sensitive memory tasks, but they d o illustrate the kind of new data that are problematic for mapping theory. In contrast, these results fit well with a configural position. We wish to address one final point raised by Nadel that may illuminate a difference in a mapping treatment of place learning vs. our configural account. Experiments specifically designed to assess the effects of hippocampal damage on exploration have been few and far between, but we have conducted one that may be worth considering (Sutherland, 1985). Control rats and rats with colchicine-induced damage to the hippocampal formation were allowed to explore on a large, circular table-top on which were placed 10 objects (e.g., cans, bottles, plastic figures, etc.) surrounded by black curtains upon which were hung several large “distal” cues. The rats were always started in the center of the table, after having sat for I minute under an opaque container that was raised by a pulley system from outside the curtains. After the rats had been placed on the table for many days, all of our measures of exploration (rearing, walking, object contacts, etc.) decreased markedly to a low and stable level (habituation). Before one subsequent session, we rotated the table (and all of the objects on it) 180” relative to the cues on the black curtains. According to cognitive mapping theory, the hippocampal damaged rats should not have shown a dishabitua-
252
HIPPOCAMPUS VOI,. 1, NO. 3, JULY 1991
tion of exploration during the session after rotation. In fact, both groups showed dishabituation (1.e.. both groups detected the transformation of object locations). This dishabituation occurred despite the fact that rats with the same hippocampal damage cannot learn to place navigate using the same kind of distal cues. Although we do not hope to compel everyone to the same conclusion as ours with these data alone, they suggest to us that there may be a very important difference. consistent with our account of place learning, between using cue constellations to flexibly navigate and detecting changes in cuc constellations (even if the change is a topographical one). The existence of such a dissociation is explicitly denied by O’Keefe and Nadel’s theory-for them exploration and place navigation are based upon the same mapping system. According to our position (which has clear similarities to the account offered by McNaughton, 1989). the critical feature of place learning situations that makes them sensitive to hippocampal damage is the requirement that the animal conditionally link navigational tmjectories to cue constellations, and not the spatial mapping aspect of the task. At thisjuncture, we are also mindful of a conceptually related dissociation described by McNaughton et al. (1989). In a single unit recording study, they showed that normal-looking place fields with rather good specificity on an eight-arm radial maze were present throughout the hippocampal CA 1 and CA3 subfields after extensive colchicine-induced degeneration of the dentate gyrus granule cells. Surprisingly for a map-based account, these same animals were clearly impaired in three different tasks requiring place novigrlrion. We suggest that the continued study of dissociations of this sort, between map learning and place navigation learning, may be very profitable in distinguishing between the cognitive mapping and configural accounts of hippocampal function. Finally, we note that, at least within the nonprimate literature, at present there may be no compelling reasons to doubt that the configural account may remain relegated to an admittedly still small number of experimental exceptions to the rule of space. It may be that in another IS years of cognitive map history the present troublesome experimental results will come to be understood as not problematic at all for the mapping view. In many ways O’Keefe and Nadel are in the cat-bird seat, and there is no reason why they should not ignore the exceptions. After all, given the present situation,
it is certainly not unreasonable to believe that they may have gotten the basic story right. However, we still hold that the configural account is on the right path and. if it^ are right, the number of acknowledged exceptions should multiply. In that context, it is important that previously articulated positions hold fast and not exhibit the kind of drift that Nadel wishes to avoid; in that way they may become the sort of stationary beacons that mark our progress along the way.
References Aggleton. J . , P. Hunt. and J . N . 1’. Rawlinh (1986) T h e effects of hippocampal lesions upon spatial and non-spatial tests of working memory. Behav. Brain Res. 19:133-146. Alvariido, M . , and J . W. Rudy (19x9) The transverse patterning problem: Configural processing in the hippocampal formation. Soc. Neurosci. Abstr. 15:610. Hirqh. R . (1974) The hippocampus and contextual retrieval of information from memory: A theory. Behav. Biol. 12:421-444. McNaughton. B. ( 19x9) Neuronal mechanism\ for spatial computation and information storage. In Nerrrcil Coiiiicctions arid Mrnrtrl Cornpri/titiolz.s. L. Nadel. L. Cooper. P. Culicover. K. Harnish, eds.. pp. 2x5-350. MIT Press/Bradford Books. Cambridge. M A . McNaughton, B.. J . Meltzer. C . A . Barnes. and R. J . Sutherland ( 19x9) Hippocampal granule cells are necessary for spatial learning but not for spatially-selective pyramidal cell discharge. Exp. Brain Res. 76:485-496. Rudy. J . W.. and K. J . Sutherland (1989) The hippocampal formation is necessary for rats to learn and remember configural discriminations. Behav. Brain Res. 34:97-109. Sutherland. R. J. (19x5) The navigating hippocampus: An individual medley of movement. \pace. and memory. In E/cc.rr.op/i\..ti~i/op?.c!f / l i t Arc.hic,or.I~x, G. Buzsaki and C . H . Vandcrwolf. eds.. pp. 255279. Akademiai Kiadb. Budapest. Sutherland. R . J . . and R . J. Mcllonald (1990) Hippocanipu\. amygdala, and memory deficits in rats. Behav. Brain Kes. 37:.57-79. Sutherland, R. J.. and J . W. Rudy (1989) Configural association theory: The role of the hippocampal formation in learning. memory. and amnesia. Psychobiology 17: 129-144. Sutherland, R. J.. H. J . McDonald. J . M . Hoesing. and J . W. Rudy ( 1989a) Hippocampal formation and configural learning and memory. Soc. Neurosci. Abstr. 15:608. Sutherland, R. J.. J . W . Rudy. R. J . Mcllonald. and C. K . Hill t1989b) Damage to the hippocampal formation in rats selectively impairs the ability to learn cue relationships. Behav. Neural Biol. 5 2 3 3 I 356.
Exceptions to the Rule of Space Robert J . Sutherland* and Jerry W. Rudy? *Department of P s y c h o l o g y , The University of Lethbridge, L e t h b r i d g e , A l b e r t a , C a n a d a , T l K 3M4 and ? D e p a r t m e n t of Psychology, Muenzinger Psychology Building, University of Colorado, Boulder, CO 80309 U . S . A .
There are many points in Nadel’s paper with which we heartily agree. At a general level we most strongly agree that a good theory of hippocampal function should be explicit and unambiguous about the kind of information that the hippocampal formation is handling. As Nadel points out. many contemporary positions are limited by their inability to say anything plainly about this issue-they are limited in that they seem to make few clear predictions and seem to lack generality across species and behavioral paradigms. This has not been the case with O’Keefe and Nadel’s position. A key point that should not be lost is their assertion that the hippocampal formation is essential for place learning. that is, for situations in which the topographical relationships i n the environment come to guide an animal’s behavior. This was a very bold claim in the mid-1970s and, to the surprise of most of their contemporaries, the subsequent experimental record is basically unanimous-they were right. Repeatedly, in the hippocampal lesion literature using nonprimates the presence of a specific place learning requirement in a task. as opposed to, for example, working memory, temporary memory buffer. or other nonspatial requirements, has been shown to be critical in detecting a behavioral impairment. For those of us convinced of the basic correctness of this central assertion by O’Keefe and Nadel, more data demonstrating the existence of an intimate link between hippocampal circuitry and place memory do not further sprcjfictil1.v strengthen O’Keefe and Nadel‘s theorv. For example. the demonstrations that certain forms of spatial memory of birds are affected by damage that includes the hippocampal formation will not allow, by themselves, further elucidation of the specific contribution of hippocampal circuitry in solving spatial problems. The same holds for the association between variations in the size of hippocampal components and variations in performance in hippocampal-sensitive tasks. such as two-way active avoidance. These new data do not add a novel kind of support for O’Keefe and Nadel‘s explanation for ~ ’ h the hippocampal formation is important for place learning. The nub of our story is this: O’Keefe and Nadel assert that instances of impaired learning and memory in animals with damage to the hippocampal formation are caused by the elimination of an explicitly topographical, mnemonic representation of the environment, that the kind of information that the hippocampal formation handles is necessarily spatial; we part company with them here. We have chosen a different, although not unrelated, path in accounting for the specific contribution that the hippocampal formation makes to learn-
ing and memory. On reviewing the hippocampal literature, we were impressed with two things: place learning is impaired after HPC damage, and there are several instances of impairments that do not appear to fit well with the idea that the hippocampal formation makes its special contribution to memory by constructing and storing spatial maps of the environment. We have already dealt in detail with these “exceptions” (Sutherland and Rudy, 1989).Nadel acknowledges that the last 15 years of experiments on the hippocampal formation have produced “a small number of exceptions” to his position. It is in this small, unassimilated fringe (which Nadel also acknowledges contains one of the human hippocampi), where we set up camp. We felt motivated t o find some way of characterizing the kind of information unique to the hippocampal formation so that both the place learning impairment and the “exceptions” could be explained in the same manner. It was also important to us that our way of chara c t e r i h g hippocampal function should predict new behavioral situations in which hippocampal circuitry makes an essential contribution. We were also fortunate to have read Hirsh’s (1974) paper, which prompted us to believe that following the path we have chosen was not altogether unreasonable. Our position is that hippocampal circuitry is necessary if a problem cannot be solved on the basis of strengthening o r weakening associations between elementary stimulus events, that is, if an animal must form associations between cue conjunctions and some other event. Topographical relationships are but one example of the many kinds of cue conjunctions or relationships that are possible. Essentially. the hippocampal formation enables an animal to disambiguate the significance of an elemental stimulus o r cue when the meaning of that cue depends upon its relationship to one o r more other cues. Place learning falls into the category of impaircd abilities precisely because if an animal must navigate to a specific location using topographical relationships among cues. then the way that an animal should move when it is facing any particular cue is ambiguous: the cue’s significance for guidance necessarily depends upon its relationship to some other element of the environment. starting location. other perceptible cues, etc. If the animal can solve the spatial problem by using only a single element of the situation t o guide its movements, then by definition this would require only it taxon strategy for solution, or, to use our terminology. a simple association solution. Nadel states quite clearly that he is not averse to the idea that one 01‘ the human hippocampi represents information that is more “abstract” than physical space. We are saying something similar about the hippocampal formation in both hemispheres of all species-primates y to birds to fish. Species differ dramatically in the kinds of perceptual o r motor information represented by activity in ensembles in neocortical zones and other structures that provide input to the hippocampal formation. ’Therefore, the kinds of cues or events that can enter into hippocampal-based configural associations may differ dramatically among species. Our view clearly predicts that it should be possible t o find examples of impaired configural memory after hippocampal damage, in addition to. and quite apart from, those involving spatial mapping-this search is of course hopeless if O’Keefe and Nadel’s theory is more in line with hippocampal function.
250
EXCEPTIONS TO THE RULE OF SPACE / Sutherland and Rudy We will describe three examples of memory impairments that d o not confirm the spatial mapping position, but that are predicted by configural association theory (Sutherland and Rudy, 1989). If our interpretation of these results is correct, it is probably the case that, at least in mammals, the memory deficit after hippocampal damage includes relational information that is both spatial and nonspatial. Our first example is that of negative patterning discrimination. In the version of this discrimination problem that we have studied, animals are rewarded with food for responding in the presence of a light or a tone, but are not rewarded for responding if the light and tone occur together. They readily learn to respond quickly and consistently to either cue alone, and to withhold responding to the compound light + tone. The rats clearly cannot be responding on the basis of the reinforcement history of either of the two cues alone-if that were the case, responding to the compound could never be lower than to the individual cues-they must be using relational information to withhold response to the compound. We predicted that rats with hippocampal damage would be unable to learn or remember this discrimination. We repeatedly have confirmed these predictions of the configural position (Rudy and Sutherland, 1989; Sutherland et al., 1989a; 1989b; Sutherland and Rudy, 1989; Sutherland and McDonald, 1990). Rats with neurotoxin-induced damage to the hippocampal formation respond as readily to the compound as to the individual elements, whether training occurs only after or both before and after surgery. If there is a satisfactory spatial mapping account of these results, we have not heard it. For example, to suggest that in the case of configural discriminations (and not simple discriminations using the same cues) the relevant associations must be “embedded” in an essentially spatial representation and therefore be sensitive to hippocampal damage is the kind of hypothesis drift that Nadel so poignantly abhors. A second example is the transverse patterning discrimination. The problem requires the animal to visually discriminate between pairs of arms in a T-maze set within a swimming pool. On every trial the rat must choose to swim into one of two arms; the arms can be white (W), black (B), or striped (S). Training is divided into three phases. In phase 1 the rats receive a simultaneous choice between white and black arms, with the white arm always containing the goal. In phase 2, the rats continue to receive white vs. black trials but, in addition, they receive black vs. striped trials, with the black arm always containing the goal. In the final phase, rats receive white vs. black trials, black vs. striped trials, and, in addition, striped vs. white trials, with the striped arm always containing the goal (thus, W + B - / B + S - / S + W - ) . It is only by the third phase that the cues are ambiguous for the rat. In order to solve the problem through phase 3 the rat must use the relationship between cues in the arms: prior to that point, the rat can use simple associations involving the cues to discriminate correctly. Damage to the hippocampal formation disrupts acquisition and retention of the solution to the transverse patterning problem, without disrupting the acquisition of the simple associations at early phases of testing (Alvarado and Rudy, 1989). Again, it is difficult to come up with a spatial mapping account of these results, but these
251
experiments were designed to test straightforward predictions of the configural position. A third example comes from Y-maze experiments conducted on dry land. If a rat is required t o choose an arm whose features match the features of the start arm, o r to avoid the arm whose features match the features of the start arm, there appears to be very little, if any, effect of damage to the hippocampal formation (Aggleton et al., 1986; Sutherland et al., 1989b; Sutherland and McDonald, 1990). These experiments have involved simple visual, tactile, and odor cues (obviously, if the relevant feature was spatial, the task would be hippocampal-sensitive). If, on the other hand, the task is modified so as to force the rat to use a nonspatial relationship between features in start and goal arms to make correct choices, then damage to the hippocampal formation disrupts performance. We have modified this basic task to force a configural solution in three different ways in separate experiments: ( I ) Rats had to choose between a black and a white goal arm. If the start box was illuminated at the beginning of a trial, then the white arm contained the goal: if the start box was not illuminated, then the black arm was correct (Sutherland et al.. 1989b). (2) Rats had to choose between arms that were black. white, or striped. If the start box contained odor I , the rat had to avoid black, if odor 2, then avoid white, and if odor 3, then avoid stripes (for more details see Sutherland et al., 1989b; Sutherland and McDonald, 1990). (3) Rats had to learn a configural black vs. white discrimination based upon time of day (Sutherland et al., l989a). In the morning choosing the black and not the white arm was rewarded; in the evening choosing the white and not the black arm was rewarded. In each of these Y-maze experiments the control rats solved the relevant configural discriminations, but in none of them did the rats with hippocampal damage successfully discriminate. These three examples do not provide an exhaustive list of exceptions to the necessity for a spatial mapping requirement in hippocampal-sensitive memory tasks, but they d o illustrate the kind of new data that are problematic for mapping theory. In contrast, these results fit well with a configural position. We wish to address one final point raised by Nadel that may illuminate a difference in a mapping treatment of place learning vs. our configural account. Experiments specifically designed to assess the effects of hippocampal damage on exploration have been few and far between, but we have conducted one that may be worth considering (Sutherland, 1985). Control rats and rats with colchicine-induced damage to the hippocampal formation were allowed to explore on a large, circular table-top on which were placed 10 objects (e.g., cans, bottles, plastic figures, etc.) surrounded by black curtains upon which were hung several large “distal” cues. The rats were always started in the center of the table, after having sat for I minute under an opaque container that was raised by a pulley system from outside the curtains. After the rats had been placed on the table for many days, all of our measures of exploration (rearing, walking, object contacts, etc.) decreased markedly to a low and stable level (habituation). Before one subsequent session, we rotated the table (and all of the objects on it) 180” relative to the cues on the black curtains. According to cognitive mapping theory, the hippocampal damaged rats should not have shown a dishabitua-
252
HIPPOCAMPUS VOI,. 1, NO. 3, JULY 1991
tion of exploration during the session after rotation. In fact, both groups showed dishabituation (1.e.. both groups detected the transformation of object locations). This dishabituation occurred despite the fact that rats with the same hippocampal damage cannot learn to place navigate using the same kind of distal cues. Although we do not hope to compel everyone to the same conclusion as ours with these data alone, they suggest to us that there may be a very important difference. consistent with our account of place learning, between using cue constellations to flexibly navigate and detecting changes in cuc constellations (even if the change is a topographical one). The existence of such a dissociation is explicitly denied by O’Keefe and Nadel’s theory-for them exploration and place navigation are based upon the same mapping system. According to our position (which has clear similarities to the account offered by McNaughton, 1989). the critical feature of place learning situations that makes them sensitive to hippocampal damage is the requirement that the animal conditionally link navigational tmjectories to cue constellations, and not the spatial mapping aspect of the task. At thisjuncture, we are also mindful of a conceptually related dissociation described by McNaughton et al. (1989). In a single unit recording study, they showed that normal-looking place fields with rather good specificity on an eight-arm radial maze were present throughout the hippocampal CA 1 and CA3 subfields after extensive colchicine-induced degeneration of the dentate gyrus granule cells. Surprisingly for a map-based account, these same animals were clearly impaired in three different tasks requiring place novigrlrion. We suggest that the continued study of dissociations of this sort, between map learning and place navigation learning, may be very profitable in distinguishing between the cognitive mapping and configural accounts of hippocampal function. Finally, we note that, at least within the nonprimate literature, at present there may be no compelling reasons to doubt that the configural account may remain relegated to an admittedly still small number of experimental exceptions to the rule of space. It may be that in another IS years of cognitive map history the present troublesome experimental results will come to be understood as not problematic at all for the mapping view. In many ways O’Keefe and Nadel are in the cat-bird seat, and there is no reason why they should not ignore the exceptions. After all, given the present situation,
it is certainly not unreasonable to believe that they may have gotten the basic story right. However, we still hold that the configural account is on the right path and. if it^ are right, the number of acknowledged exceptions should multiply. In that context, it is important that previously articulated positions hold fast and not exhibit the kind of drift that Nadel wishes to avoid; in that way they may become the sort of stationary beacons that mark our progress along the way.
References Aggleton. J . , P. Hunt. and J . N . 1’. Rawlinh (1986) T h e effects of hippocampal lesions upon spatial and non-spatial tests of working memory. Behav. Brain Res. 19:133-146. Alvariido, M . , and J . W. Rudy (19x9) The transverse patterning problem: Configural processing in the hippocampal formation. Soc. Neurosci. Abstr. 15:610. Hirqh. R . (1974) The hippocampus and contextual retrieval of information from memory: A theory. Behav. Biol. 12:421-444. McNaughton. B. ( 19x9) Neuronal mechanism\ for spatial computation and information storage. In Nerrrcil Coiiiicctions arid Mrnrtrl Cornpri/titiolz.s. L. Nadel. L. Cooper. P. Culicover. K. Harnish, eds.. pp. 2x5-350. MIT Press/Bradford Books. Cambridge. M A . McNaughton, B.. J . Meltzer. C . A . Barnes. and R. J . Sutherland ( 19x9) Hippocampal granule cells are necessary for spatial learning but not for spatially-selective pyramidal cell discharge. Exp. Brain Res. 76:485-496. Rudy. J . W.. and K. J . Sutherland (1989) The hippocampal formation is necessary for rats to learn and remember configural discriminations. Behav. Brain Res. 34:97-109. Sutherland. R. J. (19x5) The navigating hippocampus: An individual medley of movement. \pace. and memory. In E/cc.rr.op/i\..ti~i/op?.c!f / l i t Arc.hic,or.I~x, G. Buzsaki and C . H . Vandcrwolf. eds.. pp. 255279. Akademiai Kiadb. Budapest. Sutherland. R . J . . and R . J. Mcllonald (1990) Hippocanipu\. amygdala, and memory deficits in rats. Behav. Brain Kes. 37:.57-79. Sutherland, R. J.. and J . W. Rudy (1989) Configural association theory: The role of the hippocampal formation in learning. memory. and amnesia. Psychobiology 17: 129-144. Sutherland, R. J.. H. J . McDonald. J . M . Hoesing. and J . W. Rudy ( 1989a) Hippocampal formation and configural learning and memory. Soc. Neurosci. Abstr. 15:608. Sutherland, R. J.. J . W . Rudy. R. J . Mcllonald. and C. K . Hill t1989b) Damage to the hippocampal formation in rats selectively impairs the ability to learn cue relationships. Behav. Neural Biol. 5 2 3 3 I 356.
