Brain regions supporting intentional and incidental memory: a PET study

Transcription

1 Learning and Memory NeuroReort 8, (997) REGIONAL brain activity associated with intentional and incidental memory retrieval was studied with PET. Previously studied and new words were resented in either an intentional or an incidental memory task. Tye of task was crossed with an encoding maniulation ( dee vs shallow ) which varied the robability that studied items would be remembered. In both tasks, deely encoded items were associated with greater activation in the left hiocamus than were items that had received shallow encoding, suggesting that the involvement of the hiocamus in memory retrieval is indeendent of whether remembering is intentional or incidental. Right refrontal and bilateral arietal cortex were more activated during the intentional task than during the incidental task, irresective of encoding condition. Thus, these regions lay a more extensive role in memory retrieval when remembering is intentional. Key words: Eisodic memory; Hiocamus; Incidental memory; Prefrontal cortex; Recognition memory; Recollection Brain regions suorting intentional and incidental memory: a PET study M. D. Rugg, CA P. C. Fletcher, C. D. Frith, R. S. J. Frackowiak, and R. J. Dolan Wellcome Brain Research Grou, School of Psychology, University of St Andrews, St Andrews, Fife KY6 9JU; Wellcome Deartment of Cognitive Neurology, Institute of Neurology, Queen Square, London WCN 3BG, UK CA Corresonding Author Introduction The findings from a number of functional neuroimaging studies converge to suggest that exlicit retrieval of secific ast events (eisodic memory) is associated with activation of a distributed network of brain regions, notably the right refrontal cortex, medial arietal cortex (recuneus) and, less consistently, the hiocamal formation and adjacent medial temoral cortex. These findings have invariably been obtained with tasks that require information to be retrieved from memory intentionally. As is clear from everyday exerience, however, memories are often retrieved incidentally, without a rior intention to remember. There is little evidence to indicate which comonents of the network engaged during intentional remembering are also active when retrieval is incidental. The answer to this question is of interest because of the light it would shed both on the neural substrates of incidental memory, and the more general issue of whether neural activity suorting eisodic memory retrieval can be functionally, as well as neuroanatomically, dissociated. The aim of the resent study was to comare the atterns of brain activation associated with incidental and intentional remembering. This was accomlished by measuring regional cerebral blood flow (rcbf) during the erformance of two tasks, both of which emloyed a high roortion of words that had been resented in an immediately receding study hase. One of the tasks forced subjects to adot an intentional retrieval strategy. In the other task memory for the words was indeendent of the attribute that determined the correct resonse, thereby discouraging intentional memory. To maniulate the robability that studied items would be remembered two encoding conditions (dee vs shallow), which varied the degree to which items were encoded semantically, were emloyed. This maniulation exerts a strong effect on the robability of exlicit retrieval on both intentional 2 and incidental 3 memory tests, but has little or no influence on imlicit memory, as measured by indices such as reetition riming. 4 By crossing the factor of memory task (intentional vs incidental) with the factor of encoding condition (dee vs shallow) we were able to study the brain correlates of intentional and incidental remembering under conditions of high and low robability of retrieval, indeendently of any effects due to imlicit memory. Materials and Methods Subjects: Six healthy right-handed male subjects were emloyed, all of whom had given informed consent rior to articiation. The study was aroved by the local hosital ethic committee and the Administration of Radioactive Substances Advisory Committee of the UK. Data acquisition and analysis: Measures of regional rcbf were obtained with 5 O PET, as described reviously. 5 In brief rcbf estimates were obtained with a CTI-953B scanner (Siemens) oerating in 3D Raid Science Publishers Vol 8 No 5 24 March

2 acquisition mode. The resulting images were corrected for movement, transformed into a standard stereotactic sace, and tested for between-condition differences in rcbf with the method of statistical arametric maing (SPM 95, Wellcome Deartment of Cognitive Neurology, UK). A statistical threshold of < 0.00, uncorrected for multile comarisons, was adoted for all contrasts. Stimuli: Twelve airs of study and test lists were created, each with a unique set of words. Each study list consisted of 8 words, of which either 3 (odd numbered lists) or 4 (even numbered lists) were inanimate, and the remainder animate. Test lists contained 24 words, 8 of which were inanimate and six animate. Eighteen of the test words also belonged to the study list and six were new. Of the 8 old words, 3 or 4 were inanimate, and the remainder were animate; either five or four of the new words were also inanimate. Thus, there was no correlation within the lists between study status and the animate/inanimate distinction. Each test list emloyed a different ordering of the four ossible classes of item. Procedure: The exeriment was structured as a series of study test blocks, in which two study tasks (dee and shallow) were crossed with two retrieval tasks (intentional and incidental). In the intentional task an old/new recognition judgement was required on each member of the test list. In the incidental task, an animate/inanimate judgement was required. For each block, the first item of the study list was resented aroximately 5 min before the onset of each scan, and resentation was self-aced thereafter. In the dee encoding condition, subjects were required to incororate each word into a short soken sentence before moving on to the next item. In the shallow condition, they were required to judge whether the first and last letters of each word were in alhabetical order. The words were resented visually on a TV monitor susended from a cradle aroximately 45 cm in front of the subject. The interval between the comletion of the study list and the onset of the test list was filled with counting to revent rehearsal. Shortly before scanning commenced, resentation of the corresonding test list began. Words were resented on the monitor for s, with a stimulus onset asynchrony of 2.5 s. Subjects were required to give a romt verbal resonse to each item. In the intentional memory task, they were required to say yes whenever they saw an item from the study list, and no when they detected a new word. In the incidental task, the requirement was to resond yes to each inanimate word and no to each animate word. Subjects undertook each of the four combinations of task and encoding condition three times (with the excetion of one subject, who, because of technical difficulties, undertook each combination twice). The maing of lists to conditions, and the order in which the different task condition combinations were undertaken, were different for every subject. Results M. D. Rugg et al Performance: Word categorization was near ceiling on both versions of the incidental task. As exected, subjects reorted that they noticed the occurrence of deely studied words more often than they did words subjected to shallow encoding. In the intentional memory task 96% of the deely studied words were correctly recognized, with a false alarm rate of 6%. By contrast, subjects recognized only 65% of the shallowly studied words, and made false alarms to 7% of the new words. As measured by the index Hit False alarm, recognition memory was Table. Z values, x,y,z coordinates, 23 and corresonding brain regions and aroximate Brodmann areas (BAs) of eak activations identified by contrasts testing for the effects of dee vs shallow encoding Co-ordinates Z Cerebral region BA Intentional 6, 34, Left anterior cingulate cortex 32 46,6, Left middle/inferior frontal gyrus 9 52,8,6 4.9 Left inferior frontal gyrus 44 6,42,36 4. Left medial frontal cortex 8 42, 36, Left suerior temoral gyrus 22 32, 38, Left hiocamus/arahiocamal gyrus 0, 2,6 3.2 Left thalamus 50, 2,40 3. Left recentral gyrus 4 22,20, Right caudate/utamen 8, 50, Right arahiocamal/lingual gyrus 29/9 Incidental 44,34,2 3.5 Left inferior frontal gyrus 44 42,4, Left suerior temoral gyrus 22 26, 24, Left hiocamus 20, 82, Right suerior occiital cortex Vol 8 No 5 24 March 997

3 Intentional and incidental memory Table 2. Z values, x,y,z coordinates, 23 and corresonding brain regions and aroximate Brodmann areas (BA) of eak activations identified by contrasts testing for the effects of intentional vs incidental retrieval Co-ordinates Z Cerebral region BA Dee 54,4,6, 4.08 Left inferior frontal gyrus 44 42,28, Left middle frontal gyrus 9 0, 28, Left osteriorcingulate cortex 23/3 8, 68, Left recuneus 7 4, 50, Right retrosenial cortex 30 42,36, Right middle frontal gyrus 46/9 2,2, Right caudate nucleus 6, 72, Right recuneus/cuneus 3/8 4, 74, Right medial cerebellum Shallow 32, 54, Left recuneus/lateral arietal cortex 7 24,50, Right rontoolar cortex 0 40,32, Right middle frontal gyrus 46/9 24, 66, Right recuneus/lateral arietal cortex 3/7 38, 70, Right lateral cerebellum significantly better for deely studied words (t 5 = 4.4, <.0). PET data: Tables and 2 reort the outcomes of tests for the four simle effects that result from the emloyment of a 2 2 factorial design. The outcome of tests for the two main effects (task and encoding condition) are reorted below for those brain regions that were reliably activated in both tests of the aroriate simle effects. The results for the comarisons between the two encoding conditions are given in Table. Three regions of the left hemishere (inferior refrontal cortex, anterior suerior temoral gyrus and hiocamus) were sensitive to the encoding maniulation irresective of task. As illustrated in Fig., all three regions also showed greater activation in the dee than in the shallow encoding condition when the contrast was erformed collased over task (left refrontal: x = 46, y = 34, z = 2; Z = 4.85; suerior temoral gyrus: x = 44, y = 4, z = 2; Z = 4.44, left hiocamus: x= 28, y = 24, z = 8; Z = 4.05). Table 2 shows the outcome of the contrasts between intentional and incidental memory tasks. Irresective of encoding condition, there was more activity during the intentional task in right dorsolateral refrontal cortex and bilateral medial/lateral arietal cortex. Collased over encoding condition, the contrast between the two tasks revealed reliable activation in right dorsolateral refrontal cortex (x = 40, y = 32, z = 24; Z = 5.0), and in the left (x = 8, y = 68, z = 24; Z = 5.40) and right (x = 22, y = 66, z = 24; Z = 5.6) recuneus (see Fig. ). Discussion The observation of greater hiocamal activation during retrieval of deely studied words is consistent with revious reorts that this structure is active during eisodic retrieval. 6,7 It is also consistent with the roosal 7 that hiocamal activation is selectively associated with successful retrieval, rather than with the intention to retrieve. The resent findings go beyond revious results by demonstrating that the FIG.. Statistical arametric mas (threshold < 0.0) suerimosed on to sections of a magnetic resonance brain image which has been transformed into standard stereotactic sace. 23 (A) Sagittal section (x = 40) showing the main effect of encoding condition in left frontal and anterior temoral regions. (B) Coronal section (y = 24) showing main effect of encoding condition in left hiocamus. (C) Transverse section (z = 24) showing main effect of retrieval task on bilateral arietal and right refrontal regions. Vol 8 No 5 24 March

4 retrieval oerations associated with hiocamal activation occur whether or not there is a rior intention to retrieve. This confirms a revious roosal to this effect, 7 and suggests that the everyday exerience of memories seemingly oing to mind may be a reflection of the relative automaticity of hiocamally mediated retrieval. 8 This is only the third PET study (see also Refs 9 and 0) to describe activation of the left hiocamus during recognition memory for words. Notably, studies in which test lists containing high and low roortions of studied words have been contrasted have uniformly failed to find reliable differences in hiocamal activity, 0 3 desite the wealth of neurosychological evidence imlicating this structure in verbal recognition memory. 4 These negative findings may reflect the use of new words as a baseline against which to comare the effects of successful recognition. New test words are robably encoded automatically into memory, and encoding, like retrieval, is suorted by and therefore activates the hiocamus. 5,6 This might exlain why, in a recent study of recognition memory, 0 left hiocamal activation was found when a condition requiring recognition judgements on old words was contrasted with a rest condition, but not when it was contrasted with a condition requiring recognition judgements on new words. The resent findings suggest that a more aroriate baseline is rovided by reviously studied words which fail to be recollected by virtue of their imoverished encoding. As a consequence either of the artial retrieval of eisodic information, or a high level of ercetual fluency, 7 such words may engender sufficient familiarity to forestall further encoding. The contrasts between dee and shallow encoding conditions revealed activations in two extrahiocamal regions left inferior cortex and the left suerior temoral gyrus that were common to both tasks. Activation of these regions has reviously been reorted in a variety of language tasks (e.g. Ref. 8). Given the nature of the dee encoding task (sentence generation), it is ossible that activation of these regions during the retrieval of deely encoded items reflects their role in recaitulating or reinstating the original encoding eisode, erhas in resonse to inut from the hiocamus. 9 The resent findings suggest that these oerations are largely unaffected by whether retrieval is intentional or incidental, although it is noteworthy that the left inferior frontal cortex was more strongly activated in the intentional task (see Table ). In contrast to activity in the hiocamus and left temoral and frontal cortex, activity in right refrontal and medial/lateral arietal regions was unaffected by deth of encoding, but instead differed according to whether retrieval was intentional or incidental. Thus these regions aear to be relatively insensitive to both the robability of successful retrieval (but see Ref. 3), and to the conditions under which retrieved information was encoded. The role of the arietal cortex in memory retrieval is unclear,,20 and other than to indicate an association with intentional memory, the resent findings do little to elucidate this role further. The findings for the right dorsolateral refrontal cortex are consistent with several roosals about the role layed by this region in memory retrieval. The roosed functions include the control of retrieval effort, 7,2 the monitoring and verification of retrieved information, 2 and the use of retrieved information to guide behaviour. 3,22 All of these functions are intrinsic to intentional memory, leading to the rediction that the right refrontal cortex should be engaged to a greater extent during intentional remembering than when memory is incidental to task demands. The resent findings are fully in accord with this rediction. Conclusion Differential activity was observed in several of the brain regions imlicated in eisodic memory retrieval by revious neuroimaging studies. Chief among these regions were the right dorsolateral refrontal cortex, the recuneus and adjacent lateral arietal cortex, and the left hiocamus. By indeendently maniulating the intention to remember, and the robability of successful remembering, we were able to dissociate and hel clarify the roles of these regions in memory retrieval. Whereas the role of the hiocamus aears to be indeendent of whether there is an intention to retrieve, arietal and right refrontal cortex are engaged to a significantly greater extent when retrieval is intentional than when it is incidental. Together with the results of revious studies of the role of the right refrontal cortex in memory, the resent findings suggest that an imortant function of this region is to harness memory retrieval to current behavioural goals. References M. D. Rugg et al. Fletcher PC, Frith CD and Rugg MD. Trends Neurosci (in ress). 2. Craik FIM and Lockhart RS. J Verb Learn Verb Behav, (972). 3. Richardson-Klavehn A, Gardiner JM and Java RI. Memory: task dissociations, rocess dissociations, and dissociations of consciousness. In: Underwood G, ed. Imlicit Cognition. Oxford: Oxford University Press, 996: Roediger HL and McDermott K. Imlicit memory in normal human subjects. In: Boller F, and Grafman J, eds. Handbook of Neurosychology, Vol 8. Amsterdam: Elsevier, 993: Fink GR, Halligan PW, Marshall JC et al. Nature 382, (996). 6. Squire LR, Ojemann JG, Miezin FM et al. Proc Natl Acad Sci USA 89, (992). 7. Schacter DL, Alert NM, Savage CR et al. Proc Natl Acad Sci USA 93, (996). 8. Moscovitch M. J Cogn Neurosci 4, (992). 286 Vol 8 No 5 24 March 997

5 Intentional and incidental memory 9. Nyberg L, McIntosh AR, Houle S et al. Nature (996). 0. Schacter DL, Reiman E and Curran T. Neuron 7, (996).. Kaur S, Craik FIM, Jones C et al. NeuroReort 7, (995). 2. Nyberg L, Tulving E, Habib R et al. NeuroReort 7, (995). 3. Rugg MD, Fletcher PC, Frith CD et al. Brain (in ress). 4. Smith ML. Memory disorders associated with temoral lobe lesions. In: Boller JC and Grafman J, eds. Handbook of Neurohychology, Vol. 3. Amsterdam: Elsevier, 989: Haxby JV, Ungerleider LG, Horwitz B et al. Proc Natl Acad Sci USA 93, (996). 6. Nyberg L, McIntosh AR, Cabeza R et al. Proc Natl Acad Sci USA 93, (996). 7. Jacoby LL and Kelley C. Unconscious influences of memory: dissociations automaticity. In: Milner AD and Rugg MD, eds. The Neurosychology of Consciousness. London: Academic Press, 992: Price CJ, Wise RSJ, Warburton EA et al. Brain 9, (996). 9. McClelland JL, McNaughton BL and O Reilly RC. Psychol Rev 02, (995). 20. Buckner RL, Raichle M, Miezin FM et al. J Neurosci 6, (996). 2. Shallice T. From Neurosychology to Mental Structure. Cambridge: Cambridge University Press, Wilding EL and Rugg MD. Neurosychologia 35, Talairach J and Tournoux P. Co-lanar Stereotaxic Atlas of the Human Brain. Stuttgart: Thieme, 988. ACKNOWLEDGEMENTS: M.D.R. is suorted by a rogramme grant and research leave fellowshi from the Wellcome Trust. P.C.F., C.D.F., R.S.J.F. and R.J.D. are suorted by the Wellcome Trust. Received 2 November 996; acceted 7 January 997 General Summary We comared the regional brain activity associated with intentional and incidental memory. Sequences of reviously studied and new words were resented in two different tasks: one in which the requirement was exlicitly to recognize studied words, and one in which the words study status was irrelevant. Tye of task was combined with an encoding maniulation (dee vs shallow) which varied the robability that studied items would be remembered. In both tasks, deely encoded items were associated with greater activation of the left hiocamus than were items that had received shallow encoding, suggesting that the involvement of the hiocamus in memory retrieval is indeendent of whether remembering is intentional or incidental. Activation of bilateral arietal and right refrontal cortex was greater for the intentional memory task, irresective of how items had been encoded. Thus these cortical regions have more of a role in memory retrieval when remembering is intentional than when it is incidental. Vol 8 No 5 24 March