Nature © Macmillan Publishers Ltd 19988letters to natureNATURE|VOL 396|5 NOVEMBER 1998|www.nature.com 77were moved, as one, from trial to trial, rather than from session tosession. Moving the platform from trial to trial makes it impossibleto find by reference to the room cues alone, which will have twoconsequences. First, group ‘trial’ cannot show the within-sessionimprovement in performance that would be expected for group‘session’. Second, the tendency to search in the region in which theplatform had been located previously at the start of each sessionshould be less for group ‘trial’ than for group ‘session’. Theimprovement in performance in the first trial of successive sessionsshould thus be more marked for group ‘trial’ than for group‘session’. In other words, if our analysis of the performance of thehippocampally damaged rats is correct, moving the platform fromtrial to trial should have exactly the same effect as a hippocampallesion.In support of the above prediction, the results for group ‘trial’ andgroup ‘session’ (Fig. 5) show a striking similarity to those for,respectively, the hippocampally damaged and control groups(Fig. 3). Group ‘trial’ and the hippocampally damaged rats tookless time to find the platform in the first trial of most sessions thandid their respective control groups. Furthermore, group ‘trial’ andthe hippocampally damaged rats both failed to show the within-session improvement in performance that was characteristic of theother two groups. An ANOVA using individual mean escapelatencies for the 11 sessions combined showed that the latency toescape from the pool was significantly longer in trial 1 than trial 4 forgroup ‘session’ (F ¼ 130:44, d:f: ¼ 1=14), but not group ‘trial’(F ¼ 3:25, d:f: ¼ 1=14). There was also a significant differencebetween the escape latencies for the two groups in trial 1 and trial4(Fvalues . 11:24, d:f: ¼ 1=28).The gradual improvement in performance in the first trial of eachsession for normal rats in experiment 1, and for group ‘session’ inexperiment 2, shows that more reliance was placed on the heading-vector strategy, and less on the cognitive-map strategy, as trainingprogressed. According to theories of learning this change wouldoccur because stimuli, or in this case navigational strategies, mustcompete for the control they acquire over behaviour8. At first,animals may be predisposed to adopt a cognitive-map strategy,based on cues outside the pool. However, because the platform canbe found more reliably by reference to a landmark-based headingvector than by a cognitive map, the theories predict that progres-sively more reliance will be placed on the heading vector. As a resultof this switch in strategy, rats should eventually find the platformswiftly at the start of a session, and when the platform and landmarkoccupy a new position.Both experiments show that normal rats escaped from the pool bymeans of two different navigational strategies when the platformremained in the same place for all the trials of a session. One strategywas based entirely on cues that surrounded the pool, and could haveinvolved information about the geometric relationship between theplatform and a set of these cues. In other words, rats used a cognitivemap to define the position of the platform. The second strategyinvolved the landmark within the pool, and required rats to learnthe direction and distance of the platform from the landmark. Thatis, rats may also have used a heading vector to define the position ofthe platform. The results from experiment 1 show that damage tothe hippocampus disrupts the first of these strategies, but not thesecond.M.........................................................................................................................MethodsThe Morris pool was in a well-lit rectangular room with brightly colouredposters on the wall. The lowest point of the spherical black landmark was 2 cmbelow the surface of the water. For every training trial the midpoint of the 20-cm gap between the sides of the platform and the landmark was located at themidpoint of the radii that pointed towards the eight main points of thecompass (Fig. 1). A different release point on the edge of the pool was randomlyselected for each training trial. Rats remained on the platform for 30 s once theyhad climbed onto it. Further details of the method can be found in ref. 4.Received 29 May; accepted 28 August 1998.1. O’Keefe, J. & Nadel, L. The Hippocampus as a Cognitive Map (Clarendon, Oxford, 1978).2. Collett, T. S., Cartwright, B. A. & Smith, B. A. Landmark learning and visuo-spatial memories ingerbils. J. Comp. Physiol. A 158, 835–851 (1986).3. Morris, R. G. M., Garrud, P., Rawlins, J. N. P. & O’Keefe, J. Place navigation impaired in rats withhippocampal lesions. Nature 297, 681–683 (1982).4. Roberts, A. D. L. & Pearce, J. M. Control of spatial behavior by an unstable landmark. J. Exp. Psychol.Anim. Behav. Proc. 24, 172–184 (1998).5. Jarrard, L. E. On the use of ibotenic acid to lesion selectively differentcomponents of the hippocampalformation. J. Neurosci. Methods 29, 251–259 (1989).6. Morris, R. G. M. Spatial localization does not require the presence of local cue. Learning Motiv. 12,239–260 (1981).7. Galistel, C. R. The Organization of Learning (MIT Press, Cambridge, MA, 1990).8. Rescorla, R. A. & Wagner, A. R. in Classical Conditioning II: Current Research and Theory (eds Black,A. H. & Prokasy, W. F.) 64–99 (Appleton-Century-Crofts, New York, 1972).Acknowledgements. We thank R. Honey, E. S. Redhead and R. G. M. Morris for advice and discussions.This work was supported by a BBSRC studentship, and by a grant from the UK Biotechnology andBiological Sciences Research Council.Correspondence and request for materials should be addressed to J.M.P. (e-mail: [email protected]).Figure 5 Mean latency of escape in the first and fourth trials of each of the 11sessions of training, for a group trained with the platform in the same placethroughout each session (Session-1 and Session-4) and for a group trained withthe platform moved after each trial (Trial-1 and Trial-4).The anti-inflammatory agentsaspirin and salicylate inhibitthe activity of IkB kinase-bMin-Jean Yin, Yumi Yamamoto & Richard B. GaynorDivision of Hematology–Oncology, Department of Medicine, Harold SimmonsCancer Center, University of Texas Southwestern Medical Center, Dallas,Texas 75235-8594, USA. . . . . . .