NEUROREPORT COGNITIVE NEUROSCIENCE AND NEUROPSYCHOLOGY Retention of words in long-term memory: a functional neuroanatomical study with PET Jiro Okuda,1,2,CA Toshikatsu Fujii,1 Atsushi Yamadori,1 Ryuta Kawashima,3,4 Takashi Tsukiura,1 Hiroya Ohtake,1 Reiko Fukatsu,1 Kyoko Suzuki,1 Masatoshi Itoh5 and Hiroshi Fukuda3,4 1 Section of Neuropsychology, Division of Disability Science, Tohoku University Graduate School of Medicine, 2-1, Seiryo-machi, Aoba-ku, Sendai 980-8575; 2 The Japan Society for the Promotion of Science, Chiyoda-ku, Tokyo 102-0083; 3 Department of Nuclear Medicine and Radiology, IDAC, and 5 Cyclotron and Radioisotope Center, Tohoku University, Sendai 980-8575/8; 4 Aoba Brain Imaging Research Center, TAO, Sendai 980-8575, Japan CA Corresponding Author Received 21 October 1999; accepted 25 November 1999 Acknowledgements: We thank Mr Kazunori Sato (IDAC, Tohoku University) for providing a computer program of ANOVA. This study was supported by grants-in-aid to J.O. (1705) and A.Y. (08279103) for scienti®c research from the Ministry of Education, Science and Culture of Japan. This study was also supported by a grant from the Japan Society for the Promotion of Science (JSPS-RFTF97L00202). We used PET to identify brain regions associated with retention of verbal materials in long-term memory. During a PET scan, subjects repeated many sets of words one after another. In a retention condition, they were simultaneously required to retain 10 key words that were irrelevant to the repetition task. Signi®cant increases in regional cerebral blood ¯ow during the retention condition were found in bilateral parahippocampal regions, the left prefrontal and parietal association cortices, the supplementary motor area, the neostriatum and the cerebellum. We clearly demonstrated that retention of verbal materials was accompanied by neural activities in the medial temporal lobes. We also showed that, in the early phase, retention of words in long-term memory recruited left cortical areas surrounding those relevant to verbal short-term memory. NeuroReport 11:323±328 & 2000 Lippincott Williams & Wilkins. Key words: Long-term memory; Neostriatum; Parahippocampal region; Parietal association cortex, Prefrontal cortex; Positron emission tomography; Rehearsal; Retention; Supplementary motor area; Verbal short-term memory INTRODUCTION Many neuroimaging studies on human memory have focused on cerebral systems involved with encoding and retrieval processes (see [1] for a review). However, the retention process, which intervenes between encoding and retrieval periods of memory processing, is also essential in the psychological model of human memory. In the model, the memory system is divided into short-term and longterm, based on the length of the retention period [2]. For short-term memory or working memory, the retention process has been termed short-term maintenance or storage, and its possible psychological structures have been well investigated, especially for verbal materials. Previous studies have proposed two distinct mechanisms involved in the short-term maintenance of verbal materials: a subvocal rehearsal and a phonological store [3,4]. Neural correlates of these mechanisms have been identi®ed around the prefrontal and parietal association cortices by PET [4]. The anatomical basis of the retention process in long-term memory, however, has seldom been investigated with methods of functional neuroimaging. As the abovementioned psychological model implies that short-term 0959-4965 & Lippincott Williams & Wilkins memory is more or less related to long-term memory [2], it is of great interest how the neural substrates of the two are related to each other. The aim of the present study was to identify brain areas relevant to the retention process of verbal materials in long-term memory by using PET. To assess the retention process in long-term memory, we measured regional cerebral blood ¯ow (rCBF) while subjects attempted to retain many verbal items that exceeded the normal span of verbal short-term memory for about 10 min. We minimized the effects of sub-vocal rehearsal by employing an irrelevant word repetition task during the PET scan, which prevented rehearsal of the verbal items to be retained. MATERIALS AND METHODS Subjects: Seven healthy male volunteers (age range 19±24 years, mean 21.0) participated in this study. All subjects were native Japanese undergraduate students of Tohoku University. None had any signs or history of medical or neurological diseases. Each subject's MRI of the brain was normal. They were all right-handed, as assessed by the Edinburgh Handedness Inventory [5]. All subjects were Vol 11 No 2 7 February 2000 323 NEUROREPORT J. OKUDA ET AL. PET measurements: The rCBF was measured by PET (SET2400W Shimadzu, FWHM 4.0 mm) using 15 O-labeled water [7] (35 mCi injected for each PET scan). The transaxial sampling ®eld of view (FOV) was 256 mm, and the axial FOV was 190 mm. The thickness of the slice measured was 3.125 mm. Prior to the PET experiments, each subject had a catheter inserted into the right brachial vein for tracer administration and wore an individual stereotaxic ®xation helmet. PET data acquisition was started simultaneously with bolus injection of H2 15 O. Each acquisition was performed during the 120 s repetition task under each condition. A transmission scan was performed before the experiment and the data were used to obtain corrected emission images. All PET data were reconstructed by using a convolution ®lter (cut-off value 8 mm). informed of the nature of the experiment and they gave written consent in accordance with guidelines approved by Tohoku University and the Declaration of Human Rights, Helsinki 1975. Task procedures: To clarify neural substrates of retention of verbal information when the subjects are prevented from rehearsal, we designed two conditions for a repetition task. One was a retention condition and the other a control condition. Each condition was performed twice in random order. Verbal stimuli were Japanese nouns whose values of psychological property have been fully investigated [6]. The number of syllables in the stimulus words was from 2 to 4 (mean  s.d., 3.47  0.58). Different lists of 50 stimuli consisting of 10 sets of ®ve stimuli were prepared for the repetition task during each PET scan. For the retention condition, ten additional stimuli not used in the repetition task were also prepared as key words. Averaged scores of imageability and ease of learning of the words were controlled so as to be equal in each list. All stimuli were presented to the subjects by a tape recorder through a pair of earphones. The following procedures of the repetition task were common in the two conditions. Within 120 s of one task trial, ten sets of ®ve stimuli were presented to the subjects with inter-set intervals of 7 s. Each stimulus was presented at a rate of 1/s. They were asked to repeat as many of the ®ve stimulus words in each set as possible during the inter-set interval. For the retention condition, the following procedures were added to this repetition task. The subjects were presented with the 10 key words ®ve times in a row at least ®ve minutes and at most eight minutes before the start of the repetition task, and were asked to memorize them. Then they were instructed to try to hold them until the end of the PET scan. They were informed that they would be asked to recall them immediately after the task. Tests of free recall of the key words were performed immediately after the key words had been heard ®ve times and immediately after the repetition task. A schematic representation of the task design is shown in Fig. 1. Data analysis: The rCBF images were transformed into a standard anatomical format using the Human Brain Atlas system [8] and each subject's MRI. They were then smoothed with a three-dimensional Gaussian ®lter (FWHM 10 mm), and normalized for global cerebral blood ¯ow of 50 ml/100 g/min [9,10]. Difference between rCBF patterns during the retention condition and the control one was evaluated by two-way analysis of variance (ANOVA, two tasks and seven subjects as factors) on a voxel by voxel basis. We calculated F-images for task difference and voxels with F values . 11.8 ( p , 0.005, d.f. 1,12) were considered to represent regions of signi®cantly increased rCBF. Each activation focus was superimposed onto an average reformatted MRI of the seven subjects, slice thickness of which was 1 mm. Anatomical localization of areas of activation was performed in relation to this MRI. We also identi®ed the Brodmann area of each focus by referring to the atlas of Talairach and Tournoux [11]. RESULTS Mean ( s.d.) rates of successful repetition of words during the PET scan were not different between the two conditions (0.82  0.08 for the retention condition and 0.84  0.08 for the control condition). In the retention condition, the Task procedures Condition Retention Before the scan (5–8 min before) During PET scan (2 min) Hear and memorize 10 keywords Hear and orally repeat 5 words* (perform 10 times) 1 Hold the keywords in mind Control After the scan (immediately after) Recall the keywords Hear and orally repeat 5 words* (perform 10 times) No tasks No tasks *Time sequence for the repetition task: word1 2 3 4 6 5 repetition of words 1–5 1s 7s Fig. 1. A schematic representation of the experimental design. 324 Vol 11 No 2 7 February 2000 7 8 9 10 repetition NEUROREPORT NEURAL CORRELATES OF MEMORY RETENTION PROCESS subjects could recall 9.4  0.1 words immediately after they had heard the key words ®ve times and 9.5  0.1 words immediately after the repetition task. Regions showing signi®cant increases in rCBF during the retention condition compared with the control are listed in Table 1. The number of activation foci was greater in the left hemisphere than in the right. In the left hemisphere, signi®cant rCBF changes were found in the dorsolateral prefrontal areas including the inferior frontal gyrus (anterior to Broca's area) and the middle frontal gyrus, orbitofrontal areas, the precentral gyrus and the inferior parietal lobule (angular gyrus). In the right hemisphere, activations were observed in the supplementary motor area (SMA) and lateral temporal cortices. Furthermore, the parahippocampal gyri, small regions of the caudate nuclei, inferior occipito-temporal areas and cerebellum were activated bilaterally. It is noteworthy that activation foci of the parahippocampal areas were asymmetrical, i.e. posterior in the left hemisphere and anterior in the right. Representative areas of activation are illustrated in Fig. 2. DISCUSSION In this study, we attempted to localize brain regions related to retention of words in long-term memory. We would like to emphasize that our experiment dealt with long-term memory rather than short-term memory. Reasons for this are as follows; ®rst, the retention interval was around 10 min and therefore exceeded the period during which materials are thought to be stored in the short-term memory. Second, the number of items that the subjects attempted to retain exceeded the normal span of shortterm memory. In addition, this task can be regarded as an episodic memory task, because the subjects had to retain the 10 key words that they had experienced (heard) at a speci®c time before the PET scan. In our design, we asked the subjects to perform the same repetition task in both the retention and the control conditions during PET scans. Effects of word repetition on the rCBFs should be the same between the two conditions, and comparison between the conditions should cancel out these effects. This idea is supported by the behavioral result that the number of correctly repeated words during the scan did not signi®cantly differ between the two conditions. It is important to note that responses on the word repetition were not perfect but rather 85% at the most. This implies that the repetition task in this study was painstaking and that the subjects had to engage in the word repetition attentively during the scan. Thus, in the retention condi- Table 1. Activation foci during the retention condition determined by ANOVA ( p , 0.005). Anatomical structure Left hemisphere Cerebellum Orbital gyrus (11) Rectal gyrus (25) Fusiform gyrus (19) Parahippocampal gyrus (36) Inferior occipital gyrus (18) Inferior frontal gyrus (47) Inferior frontal gyrus (45) Caudate nucleus (tail) Angular gyrus (39) Middle frontal gyrus (9) Inferior parietal lobule (40) Precentral gyrus (4) Right hemisphere Cerebellum Inferior temporal gyrus (20) Parahippocampal gyrus (28) Middle occipital gyrus (19) Middle temporal gyrus (37) Posterior insura Caudate nucleus (tail) Middle temporal gyrus (21) Medial frontal lobe (8) Medial frontal lobe (SMA) (6) Talairach coordinates Peak F-value x y z ÿ8 ÿ16 ÿ36 ÿ5 ÿ9 ÿ40 ÿ26 ÿ31 ÿ30 ÿ44 ÿ26 ÿ40 ÿ40 ÿ49 ÿ34 ÿ44 ÿ36 ÿ26 ÿ32 ÿ66 ÿ52 46 8 ÿ63 ÿ32 ÿ84 22 26 ÿ36 ÿ62 16 10 20 ÿ1 ÿ50 ÿ25 ÿ45 ÿ38 ÿ38 ÿ17 ÿ17 ÿ15 ÿ15 ÿ4 0 3 7 29 30 30 35 36 44 51 17.5 16.4 38.9 26.6 16.1 23.9 16.5 38.3 13.9 13.2 17.9 17.0 22.8 20.9 19.2 22.3 17.4 49.4 28 1 52 18 45 54 31 31 48 11 6 ÿ36 ÿ80 ÿ10 ÿ17 ÿ76 ÿ62 ÿ16 ÿ48 ÿ48 31 ÿ12 ÿ30 ÿ18 ÿ24 ÿ17 ÿ12 0 0 2 7 40 62 14.1 27.7 14.5 16.6 26.3 22.9 19.4 55.9 16.5 19.5 23.1 Stereotaxic coordinates refer to maximal activations indicated by peak F-values in particular cerebral structures. Numbers in parentheses following each anatomical structure refer to the Brodmann areas. Distances refer to the stereotactic space de®ned by Talairach and Tournoux [11]. Coordinates are expressed in mm. Vol 11 No 2 7 February 2000 325 NEUROREPORT J. OKUDA ET AL. Fig. 2. Representative brain activations during the retention condition compared with the control one were superimposed onto the mean reformatted MRI of the seven subjects. The top and second ranks (A and B) show contiguous horizontal sections with a thickness of 1 mm, and the third and bottom ranks (C,D) show contiguous sagittal sections with a thickness of 1 mm. White areas in the far left image of (A) represent locations of the sagittal sections of (C) and (D). In the horizontal section, the right side of the image represents the right side of the brain. Black arrows indicate the activations of (A) the left and right parahippocampal gyri, (B) small regions around the tail of the caudate nuclei, (C) the left inferior frontal, middle frontal and angular gyri and (D) the right SMA. tion, they must have been prevented from articulatory rehearsal of the 10 key words. Nevertheless, the subjects were able to recall most of the key words after the PET scan. This means that the subjects must have retained the key words during the scan period with minimal rehearsal. These behavioral data led us to conclude that comparison of the experimental retention condition with the control condition should reveal the process of retention of words in long-term memory up to 10 min without active rehearsal. As for functional neuroanatomical data, the most intriguing ®nding is the strong bilateral activations of parahippocampal regions in the retention condition, which were not reported in previous studies on the retention process in short-term memory. In previous neuroimaging studies, activation of the medial temporal lobes was observed in relation to encoding and retrieval in episodic memory [1,12]. Our results clearly demonstrate that the medial temporal lobes are involved not only in encoding and retrieval but also in the early retention process in longterm memory. This ®nding is consistent with the idea that the initial stage of the retention process in long-term memory, i.e. the memory consolidation process, requires participation of the medial temporal lobes [13,14], although the duration necessary for memory consolidation is still a controversial issue [15]. 326 Vol 11 No 2 7 February 2000 It is interesting that anterior±posterior dissociation was observed in our activations of the parahippocampal regions, i.e. posterior in the left parahippocampal region and anterior in the right. Recently, a model of distinct anterior± posterior hippocampal activities during episodic encoding and retrieval was proposed considering the ®ndings of many PET activation studies. This so-called HIPER (hippocampal encoding/retrieval) model proposed that encoding in the episodic memory would recruit anterior hippocampal areas and retrieval would recruit posterior areas [16]. If the model is valid, we can hypothesize that the retention process is associated with interaction of the encoding and retrieval processes. Further research is needed on this issue. Neocortical activations in the retention condition were observed in the inferior frontal gyrus, middle frontal gyrus, precentral gyrus, SMA, inferior parietal lobule and inferior occipito-temporal areas. Interestingly, several neocortical regions activated in the retention condition were in the vicinity of cortical areas that have been reported to be involved in maintenance of verbal information in shortterm memory or working memory. For instance, Fiez et al. [17] reported activations in the left frontal lobes (Broca's area, the premotor area, the SMA), the bilateral middle frontal gyri and cerebellum during sub-vocal rehearsal of ®ve words. They concluded that these areas are related to NEUROREPORT NEURAL CORRELATES OF MEMORY RETENTION PROCESS short-term maintenance of verbal materials in an active state. They also argued that Broca's area plays a main role in sub-vocal rehearsal, supported by the SMA and the premotor area. Paulesu et al. [4] compared tasks of auditory and visual discrimination of letters, and suggested that Broca's area had a role in articulatory rehearsal and that the left inferior parietal lobule (the supramarginal gyrus) was related to the phonological store. Their ®ndings were replicated by Salmon et al. [18]. Awh et al. [19] and Smith et al. [20] also con®rmed this idea by comparing a short-term memory task and a rehearsal task with letters. When compared with these neuroimaging results on verbal short-term memory, the left inferior frontal area and the left inferior parietal lobule activated in the present study were similar but not identical. Generally, the activations we observed were located outside of the above-mentioned areas, i.e. area 45/47 of the inferior frontal gyrus and area 39 of the angular gyrus (see Fig. 3). It is plausible that activation in the left inferior frontal gyrus was near but not in Broca's area itself, since our task design prevented the subjects from active rehearsal of the key words. The retention condition in our study concerned with the period beyond short-term memory, but the period was rather short when compared with that in general long-term memory that might be fully consolidated as one's episode. We speculate that the activated neocortical areas in the present study are involved with vary early phase of retention process in long term memory, which may represent a certain transitional process from short-term to longterm memory. When verbal information is being stored y50 z50 into long-term memory from short-term memory, areas near but slightly different from those associated with maintenance in verbal short-term memory may be recruited. The repetition task employed in our study was very demanding even for the young university students. They had to repeat ®ve words each of which had 3.4 syllables on average during 7 s. We assumed that it was too tight to perform any conscious operations in addition to word repetition. Nevertheless, it can not be completely ruled out that the activated inferior frontal areas in the present study are somewhat related to retrieval process of the key words such as attempts for retrieval and the strategic retrieval process. For instance, Petrides et al. [21] reported that the mid-ventrolateral frontal cortex (BA 45) in the left hemisphere was involved in the strategic retrieval of verbal information from long-term memory. The activations in the left inferior frontal gyrus in the present study overlap those observed in the free recall task by Petrides et al. Therefore, it is possible that subtle retrieval of the key words, as well as the transitional process mentioned above, contributes to the activity observed in the left inferior frontal area. The activations in the left middle frontal gyrus might be related to working memory, while activations of bilateral middle frontal gyri were often observed in studies focusing on verbal working memory [22,23]. The retention condition in the present study, in a sense, can be regarded as a working memory task, as an additional demand for retention of the key words was loaded on the repetition task. The present study also showed small but strong activations around the tail of the caudate nuclei. Jueptner et al. [24] reported activations of the caudate nucleus in relation to a motor learning task. The nucleus was strongly activated during new learning of sequences of eight ®nger movements. Although it is dif®cult to discuss the role of the nuclei in verbal memory retention, we suppose that the nuclei have a certain role in the retention process in longterm memory. It has been reported that patients with damage of the bilateral neostriatum showed impaired learning of cognitive skills [25]. The patients' failure in learning might have been due to disturbance of retaining a certain type of memory information for a long time against distraction. CONCLUSIONS Phonological short-term memory [4,18] Item recognition [19, 20] Short-term maintenance [17] Long-term retention (present study) Fig. 3. Relationship between activations in the left frontal and parietal lobes in the present study (closed circles) and in previous reports on verbal short-term memory (open circles, diamonds and squares) is shown on a schematic brain surface. Studies from the same research group or those employing the same task design are represented by the same symbol. A cross represents y- and z-axes in Talairach space. Activations in the present study seem to surround those in the previous reports. Activated areas in the retention process of verbal materials in long-term memory were found in bilateral parahippocampal regions, several neocortical regions (the left inferior frontal gyrus, middle frontal gyrus, precentral gyrus and inferior parietal lobule, the right SMA, bilateral inferior occipito-temporal areas), the tail of the caudate nuclei and the cerebellum. The results provide evidence, for the ®rst time using a neuroimaging technique, that medial temporal lobes are important not only for the encoding and retrieval processes but also for the retention process in human longterm memory. We also found that, in the early phase, the retention process in long-term memory activated the left frontal and parietal regions surrounding those relevant to the retention process in verbal short-term memory. We conclude that, in the early phase, retention of verbal materials in long-term memory requires concerted neural activities in both the medial temporal lobes and the Vol 11 No 2 7 February 2000 327 NEUROREPORT neocortical regions that are located around the regions relevant to verbal short-term memory. REFERENCES 1. Schacter DL and Wagner AD. Hippocampus 9, 7±24 (1999). 2. Baddeley A. Human Memory: Theory and Practice. Massachusetts: Allyn and Bacon, 1990: 39±66. 3. Longoni AM, Richardson JTE and Aiello A. Mem Cogn 21, 11±22 (1993). 4. Paulesu E, Frith CD and Frackowiak RSJ. Nature 362, 342±345 (1993). 5. Old®eld R. Neuropsychologia 9, 97±113 (1971). 6. Ogawa T and Inamura Y. Shinrigaku Kenkyu 44, 317±327 (1974). 7. Fujiwara T, Watanuki S, Yamamoto S et al. Ann Nucl Med 11, 307±313 (1997). 8. Roland P, Graufelds C, Wahlin J et al. Hum Brain Map 1, 173±184 (1994). 9. Raichle ME, Martin WR, Herscovitch P et al. J Nucl Med 24, 790±798 (1983). 10. Herscovitch P, Markham J and Raichle ME. J Nucl Med 24, 782±789 (1983). 328 Vol 11 No 2 7 February 2000 J. OKUDA ET AL. 11. Talairach J and Tournoux P. Co-Planar Stereotactic Atlas of the Human Brain. Stuttgart: Thieme, 1988. 12. Fujii T, Okuda J, Kawashima R et al. Neuroreport 8, 1113±1117 (1997). 13. Scoville WB and Milner B. J Neurol Neurosurg Psychiatry 20, 11±21 (1957). 14. Alvarez P and Squire LR. Proc Natl Acad Sci USA 91, 7041±7045 (1994). 15. Nadel L and Moscovitch M. Curr Opin Neurobiol 7, 217±227 (1997). 16. Lepage M, Habib R and Tulving E. Hippocampus 8, 313±322 (1998). 17. Fiez JA, Raife EA, Balota DA et al. J Neurosci 16, 808±822 (1996). 18. Salmon E, Van der Linden M, Collette F et al. Brain 119, 1617±1625 (1996). 19. Awh E, Jonides J, Smith EE et al. Psychol Sci 7, 25±31 (1996). 20. Smith EE, Jonides J and Koeppe RA. Cerebral Cortex 6, 11±20 (1996). 21. Petrides M, Alivisatos B and Evans AC. Proc Natl Acad Sci USA 92, 5803±5807 (1995). 22. Petrides M, Alivisatos B, Meyer E et al. Proc Natl Acad Sci USA 90, 878±882 (1993). 23. D'Esposito M, Detre JA, Alsop DC et al. Nature 378, 279±281 (1995). 24. Jueptner M, Frith CD, Brooks DJ et al. J Neurophysiol 77, 1325±1337 (1997). 25. Saint-Cyr JA, Taylor AE and Lang AE. Brain 111, 941±959 (1988).