- Email: [email protected]
The birth of a memory
Research Update
viable offspring following the mutation or deletion of a gene. Finally, it should allow more rapid investigations of the interactions between neuronal signalling pathways, by testing for synergistic effects between knockouts and drugs. References 1 Hebb, D.O. (1949) The Organization of Behavior, John Wiley & Sons 2 Thompson, R.F. (1986) The neurobiology of learning and memory. Science 233, 941–947 3 Collingridge, G.L. et al. (1983) The antagonism of amino acid-induced excitations of rat hippocampal CA1 neurones in vitro. J. Physiol. 334, 19–31 4 Morris, R.G.M. et al. (1986) Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature 319, 774–776
TRENDS in Neurosciences Vol.25 No.6 June 2002
5 Silva, A.J. et al. (1992) Deficient hippocampal long-term potentiation in α-calcium–calmodulin kinase II mutant mice. Science 257, 201–206 6 Silva, A.J. et al. (1992) Impaired spatial learning in α-calcium–calmodulin kinase II mutant mice. Science 257, 206–211 7 Grant, S.G.N. et al. (1992) Impaired long-term potentiation, spatial learning, and hippocampal development in fyn mutant mice. Science 258, 1903–1910 8 Mayford, M. and Kandel, E.R. (1999) Genetic approaches to memory storage. Trends Genet. 15, 463–470 9 Tsien, J.Z. et al. (1996) Subregion and cell type restricted gene knockout in mouse brain. Cell 87, 1317–1326 10 Garrick, D. et al. (1998) Repeat-induced gene silencing in mammals. Nat. Genet. 18, 56–59 11 Ohno, M. et al. (2001) Inducible, pharmacogenetic approaches to the study of learning and memory. Nat. Neurosci. 4, 1238–1243
279
12 Picciotto, M.R. (1999) Knock-out mouse models used to study neurobiological systems. Crit. Rev. Neurobiol. 13, 103–149 13 McKernan, R.M. et al. (2000) Sedative but not anxiolytic properties of benzodiazepines are mediated by the GABA(A) receptor α1 subtype. Nat. Neurosci. 3, 587–592 14 Orban, P.C. et al. (1999) Is RAS-dependent signalling necessary for long-term plasticity? Trends Neurosci. 22, 38–44 15 Sweatt, J.D. (2001) The neuronal MAP kinase cascade: a biochemical signal integration system subserving synaptic plasticity and memory. J. Neurochem. 76, 1–10
Paul F. Chapman Cardiff School of Biosciences, Cardiff University, PO Box 911, Cardiff, UK CF10 3US. e-mail: [email protected]
The birth of a memory Leun J. Otten and Michael D. Rugg Laying down new memories has long been thought to involve interactions between the hippocampus and multiple regions of the neocortex. Functional neuroimaging studies performed over the past four years provide evidence for this proposal. A recent electrophysiological study offers a possible mechanism by which interactions between brain regions take place during memory formation.
Since the description of the famous patient H.M. in the late 1950s [1], it has been recognized that the medial temporal lobe (MTL) plays a crucial role in human memory. Following bilateral resection of the MTL to alleviate persistent epilepsy, H.M. became unable to remember new facts and events for more than a few minutes. Neuropsychological findings such as these indicate that an intact MTL is necessary for normal long-term memory functioning, although it is difficult to discern from such data the extent to which the MTL contributes to memory formation, as opposed to memory storage and retrieval. Studying the neural correlates of memory formation
The question of which neural structures are associated with the formation of lasting human memories can be addressed by recording neural activity while items are initially encoded into memory. A particularly powerful method makes use of the ‘subsequent memory procedure’. http://tins.trends.com
In this approach, neural activity is recorded while volunteers study a sequence of items, after which memory for the items is tested. The activity elicited by items at the time of study is then sorted according to whether the items are remembered or forgotten in the subsequent memory test. Differences between these two types of activity (‘subsequent memory effects’) are taken as candidate neural correlates of encoding. Thus, the procedure provides information about neural structures in which activity is correlated with successful encoding. However, it is important to note that it does not identify which structures are necessary for encoding to occur. The subsequent memory approach has been used with scalp-recorded electrophysiological activity (event-related potentials, or ERPs) for >25 years [2,3]. These studies have provided valuable
information about temporal aspects of the formation of memory traces, but they have been less helpful in identifying the associated brain regions. This is because of both the difficulties of localizing sources of scalp electrical activity and the relative insensitivity of scalp recordings to neural activity in deep brain structures such as the MTL. Non-invasive detection of neural activity with high spatial resolution became possible with the introduction of functional neuroimaging techniques. Memory encoding has been studied since the earliest days of the application of these techniques to cognition [4], but it was with the advent of ‘event-related’functional MRI (fMRI) [5] that their full potential was realized. A key development came in 1998, when two papers using the subsequent memory approach with event-related fMRI were published back-to-back in Science [6,7]. These papers described studies of the encoding of visually presented words and pictures, respectively. They showed that relatively greater fMRI signals were elicited by subsequently remembered items in the MTL, among other regions, highlighting a role for the MTL in successful memory encoding. Evidence has now accumulated to suggest that the MTL (both cortex and hippocampus proper) plays a role in encoding across a wide range of task conditions and stimulus materials [3,8–10]. Notwithstanding the puzzling failure to identify encodingrelated MTL activity in some studies
0166-2236/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0166-2236(02)02176-8
280
Research Update
(e.g. see Ref. [11]), these findings provide support for the idea that the MTL plays a key role in memory formation. Recent findings regarding the role of the medial temporal lobe
The precise role played by the MTL in memory formation has, however, remained elusive. A recent study, which combined the temporal resolution of electrophysiological recordings with the spatial resolution usually available only with functional neuroimaging, could provide important insights. Fell et al. [12] asked unilateral temporal-lobe epilepsy patients to study short word lists, and to attempt to recall the words after a short distraction task. As the words were being studied, electrical activity was recorded from bilateral, multicontact depth electrodes that had been inserted into the MTL for use in clinical investigations. The authors analysed the oscillatory activity at around 40 Hz (socalled ‘gamma’ activity) that was induced in the unaffected hemisphere by the presentation of the words. The degree to which these oscillations were synchronized between the anterior MTL cortex (perirhinal and entorhinal regions) and the hippocampus was found to depend on whether the words were subsequently remembered or forgotten. For remembered words, synchronization was relatively higher during the first few hundred milliseconds after word onset and relatively lower around one second after onset. In both structures, the initial increase in synchronization for remembered words was accompanied by a reduction in the amount of gamma activity. The paper by Fell et al. follows an earlier report on a related data set. Fernández et al. [13] focused on temporal aspects of the interaction between the anterior MTL cortex and the hippocampus. Whereas the ERPs elicited by subsequently remembered and forgotten words differed in both regions, the onset of this difference began some 200 ms earlier (at around 300 ms) in cortex than in the hippocampus. Regardless of their other implications, the findings from both studies point to involvement of the anterior MTL cortex as well as the hippocampus in the initial stages of memory formation. Anterior MTL activations have not generally been reported in event-related fMRI studies of encoding. However, using the same experimental procedure as Fernández http://tins.trends.com
TRENDS in Neurosciences Vol.25 No.6 June 2002
et al., and an fMRI method optimized for the detection of activity in the anterior MTL, Strange et al. [14] recently described subsequent memory effects in both hippocampus and perirhinal cortex. Hippocampal involvement in memory encoding has also been observed at the single neuron level in epilepsy patients implanted with depth electrodes. Cameron et al. [15] found that when patients were asked to remember word pairs, the firing rates of hippocampal neurons predicted whether the pairs would subsequently be remembered. Interestingly, the majority of the neurons showed relatively lower firing rates in response to subsequently remembered items – a finding opposite to what most researchers would have predicted on the basis of fMRI findings. Implications
Together, the studies discussed here contribute to our understanding of the neural bases of memory formation in at least two ways. First, the data provide strong evidence that the MTL is directly involved in human memory encoding, and, thus, that its role in memory is not confined to storage or retrieval functions. Second, the findings suggest that different MTL regions act in concert to support memory encoding, and that the nature of their interaction within the first second or so after a stimulus event influences its subsequent memorability. Future directions
Of course, many questions remain regarding the role of the MTL during memory encoding. One can speculate about the functional significance of phase locking of gamma activity [16], but its functional significance, origin and relationship to subsequent memory effects seen in other measures are uncertain. It is also unclear whether the phase locking observed between MTL structures is restricted to the 40 Hz frequency band or, indeed, whether it is restricted to structures within the MTL. In this respect, it is noteworthy that the most consistent subsequent memory effects identified in fMRI studies have been in the inferior prefrontal cortex. It has been suggested that the prefrontal cortex supports working memory processes engaged by the demands of the encoding task, and that the products of these processes are relayed to the MTL (e.g. see Ref. [17]). In light of these suggestions, it would be of
considerable interest to determine whether synchrony between frontal and MTL regions in the gamma band, or in other frequency ranges, varies according to subsequent memory performance. What type of memory is reflected by subsequent memory effects in the MTL? Most studies using the subsequent memory approach have tested memory using either recall or recognition tasks. It remains to be established whether MTL subsequent memory effects are found when memory is tested in other ways – most importantly, when it is tested with ‘indirect’ tests that tap ‘implicit’, rather than ‘explicit’, memory for the study items. According to prevailing views of hippocampal function [18], this structure is unnecessary for implicit memory, and thus should not exhibit subsequent memory effects for indirect tests. Studies along these lines would therefore be of considerable theoretical interest. Acknowledgements
Our research is supported by the Wellcome Trust and a cooperative award from the UK Medical Research Council. References 1 Scoville, W.B. and Milner, B. (1957) Loss of recent memory after bilateral hippocampal lesions. J. Neurol. Neurosurg. Psychiatry 20, 11–21 2 Friedman, D. and Johnson, R., Jr (2000) Eventrelated potential (ERP) studies of memory encoding and retrieval: a selective review. Microsc. Res. Techn. 51, 6–28 3 Paller, K.A. and Wagner, A.D. (2002) Observing the transformation of experience into memory. Trends Cognit. Sci. 6, 93–102 4 Fletcher, P.C. et al. (1997) The functional neuroanatomy of episodic memory. Trends Neurosci. 20, 213–218 5 Rosen, B.R. et al. (1998) Event-related functional MRI: past, present, and future. Proc. Natl. Acad. Sci. U. S. A. 95, 773–780 6 Wagner, A.D. et al. (1998) Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. Science 281, 1188–1191 7 Brewer, J.B. et al. (1998) Making memories: brain activity that predicts how well visual experience will be remembered. Science 281, 1185–1187 8 Fernández, G. and Tendolkar, I. (2001) Integrated brain activity in medial temporal and prefrontal areas predicts subsequent memory performance: Human declarative memory formation at the system level. Brain Res. Bull. 55, 1–9 9 Rugg, M.D. Functional neuroimaging of memory. In Handbook of Memory Disorders (2nd edn) (Baddeley, B. et al., eds), John Wiley & Sons (in press) 10 Brewer, J.B. and Moghekar, A. Imaging the medial temporal lobe: exploring new dimensions. Trends Cognit. Sci. (in press) 11 Otten, L.J. and Rugg, M.D. (2001) Taskdependency of the neural correlates of episodic
Research Update
encoding as measured by fMRI. Cereb. Cortex 11, 1150–1160 12 Fell, J. et al. (2001) Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nat. Neurosci. 4, 1259–1264 13 Fernández, G. et al. (1999) Real-time tracking of memory formation in the human rhinal cortex and hippocampus. Science 285, 1582–1585 14 Strange, B.A. et al. (2002) Dissociable human perirhinal, hippocampal, and parahippocampal
TRENDS in Neurosciences Vol.25 No.6 June 2002
roles during verbal encoding. J. Neurosci. 22, 523–528 15 Cameron, K.A. et al. (2001) Human hippocampal neurons predict how well word pairs will be remembered. Neuron 30, 289–298 16 Tallon-Baudry, C. and Bertrand, O. (1999) Oscillatory gamma activity in humans and its role in object representation. Trends Cognit. Sci. 3, 151–161 17 Buckner, R.L. et al. (2000) Cognitive neuroscience of episodic memory encoding. Acta Psychol. 105, 127–139
281
18 Squire, L.R. and Zola, S.M. (1997) Amnesia, memory and brain systems. Philos. Trans. R. Soc. London Ser. B 352, 1663–1673
Leun J. Otten* Michael D. Rugg Institute of Cognitive Neuroscience and Dept of Psychology, University College London, London, UK WC1N 3AR. *e-mail: [email protected]
Response: The birth of a memory Guillén Fernández, Jürgen Fell and Pascal Fries Otten and Rugg have elegantly integrated the results of the study by Fell and colleagues [1] into the current knowledge about how experiences are transformed into memories. One of their central conclusions is that the findings ‘…point to the involvement of anterior MTL [medial temporal lobe] cortex as well as the hippocampus in the initial stages of memory formation.’ Despite these important aspects of localization and of the temporal sequence of MTL substructure involvement [2], we would like to highlight another central aspect of that study: namely, that a particular parameter of neuronal activity, rhinal–hippocampal phasesynchronization in the gamma-frequency band (~40 Hz), correlated with subsequent recall. This is conceptually important because memories might reside in neuronal assemblies rather than in individual neurons [3], and synchronization is an ideal mechanism to bind neurons into assemblies [4]. In early visual processing, neurons that encode features of a complex visual percept are associated in functional assemblies through gamma-frequency synchronization [5]. Furthermore, when sensory stimuli are perceptually or attentionally selected and the respective neurons are bound together to raise their saliency, gamma-frequency synchronization among these neurons is also enhanced [5]. Gamma-mediated coupling, and its modulation by attention, is not limited to the visual modality: it is also found in the auditory [6] and somatosensory domains [7]. Moreover, gamma oscillations allow visuo–motor binding between posterior and central brain regions [8] and are involved in higher order cognitive operations, such as http://tins.trends.com
working memory [9] or learning of new associations in a conditioning task [10]. In addition to being a means for dynamically binding neurons into assemblies, gamma-frequency synchronization appears to be the prime candidate mechanism for stabilizing cortical connections among members of a neural assembly over time. On the one hand, neurons increase or decrease the strength of their synaptic connections depending on the precise coincidence of their activation [11] and gammafrequency synchronization provides exactly the required temporal precision. On the other hand, strengthened connections among neurons in a ‘memory assembly’ might facilitate its later recall. In general, EEG signals reflect postsynaptic potentials, which are mainly determined by the average activity of local neuronal populations [12]. In other words, EEG oscillations of ~40 Hz are based on clusters of discharges occurring about every 25 ms. Although the exact mechanisms underlying the generation of gammafrequency synchronization are as yet unclear, several studies have begun to shed light on this issue. In the hippocampus, gamma-frequency synchronization is driven by interneuron network oscillations and intrinsic membrane resonance, as revealed in slice preparations [13]. Bragin and colleagues [14] have identified gammaactivity in the hippocampus of behaving rats and have shown that gamma-activity is most prominent in the dentate gyrus, the main hippocampal recipient of input from the neocortex via the entorhinal cortex and perforant path [15]. With regard to ‘phase locking of gammaactivity’, Otten and Rugg state that its ‘functional significance, origin and relationship to subsequent memory effects
seen in other measures are uncertain.’ However, taking the previously mentioned findings into account, we arrive at a different conclusion. Several lines of evidence suggest that gamma-frequency synchronization plays a general role in binding neurons into assemblies over short and long distances. In memory formation, gamma activity has the optimal frequency to support the transformation of a temporary representation into a durable memory trace by strengthening synaptic connectivity. Therefore, we hypothesize that the rhinal–hippocampal coupling observed by Fell and colleagues [1] enables information transfer to the hippocampus and initiates the mnemonic operation of memory encoding, which leads to synaptic plasticity and occurs within the first second of an event that can be remembered later on. References 1 Fell, J. et al. (2001) Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nat. Neurosci. 4, 1259–1264 2 Fernández, G. et al. (1999) Real-time tracking of memory formation in the human rhinal cortex and hippocampus. Science 285, 1582–1585 3 Rumelhart, D.E. and McClelland, J.L. (1986) Parallel Distributed Processing: Explorations in the Microstructure of Cognition, MIT Press 4 Singer, W. (1999) Neuronal synchrony: a versatile code for the definition of relations? Neuron 24, 49–65 5 Engel, A.K. et al. (2001) Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716 6 Tiitinen, H. et al. (1993) Selective attention enhances the auditory 40-Hz transient response in humans. Nature 364, 59–60 7 Desmedt, J.E. and Tomberg, C. (1994) Transient phase-locking of 40 Hz electrical oscillations in prefrontal and parietal human cortex reflects the process of conscious somatic perception. Neurosci. Lett. 168, 126–129 8 Rodriguez, E. et al. (1999) Perception’s shadow: long-distance synchronization of human brain activity. Nature 397, 430–433 9 Tallon-Baudry, C. et al. (1998) Induced gammaband activity during the delay of a visual
0166-2236/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0166-2236(02)02177-X
viable offspring following the mutation or deletion of a gene. Finally, it should allow more rapid investigations of the interactions between neuronal signalling pathways, by testing for synergistic effects between knockouts and drugs. References 1 Hebb, D.O. (1949) The Organization of Behavior, John Wiley & Sons 2 Thompson, R.F. (1986) The neurobiology of learning and memory. Science 233, 941–947 3 Collingridge, G.L. et al. (1983) The antagonism of amino acid-induced excitations of rat hippocampal CA1 neurones in vitro. J. Physiol. 334, 19–31 4 Morris, R.G.M. et al. (1986) Selective impairment of learning and blockade of long-term potentiation by an N-methyl-D-aspartate receptor antagonist, AP5. Nature 319, 774–776
TRENDS in Neurosciences Vol.25 No.6 June 2002
5 Silva, A.J. et al. (1992) Deficient hippocampal long-term potentiation in α-calcium–calmodulin kinase II mutant mice. Science 257, 201–206 6 Silva, A.J. et al. (1992) Impaired spatial learning in α-calcium–calmodulin kinase II mutant mice. Science 257, 206–211 7 Grant, S.G.N. et al. (1992) Impaired long-term potentiation, spatial learning, and hippocampal development in fyn mutant mice. Science 258, 1903–1910 8 Mayford, M. and Kandel, E.R. (1999) Genetic approaches to memory storage. Trends Genet. 15, 463–470 9 Tsien, J.Z. et al. (1996) Subregion and cell type restricted gene knockout in mouse brain. Cell 87, 1317–1326 10 Garrick, D. et al. (1998) Repeat-induced gene silencing in mammals. Nat. Genet. 18, 56–59 11 Ohno, M. et al. (2001) Inducible, pharmacogenetic approaches to the study of learning and memory. Nat. Neurosci. 4, 1238–1243
279
12 Picciotto, M.R. (1999) Knock-out mouse models used to study neurobiological systems. Crit. Rev. Neurobiol. 13, 103–149 13 McKernan, R.M. et al. (2000) Sedative but not anxiolytic properties of benzodiazepines are mediated by the GABA(A) receptor α1 subtype. Nat. Neurosci. 3, 587–592 14 Orban, P.C. et al. (1999) Is RAS-dependent signalling necessary for long-term plasticity? Trends Neurosci. 22, 38–44 15 Sweatt, J.D. (2001) The neuronal MAP kinase cascade: a biochemical signal integration system subserving synaptic plasticity and memory. J. Neurochem. 76, 1–10
Paul F. Chapman Cardiff School of Biosciences, Cardiff University, PO Box 911, Cardiff, UK CF10 3US. e-mail: [email protected]
The birth of a memory Leun J. Otten and Michael D. Rugg Laying down new memories has long been thought to involve interactions between the hippocampus and multiple regions of the neocortex. Functional neuroimaging studies performed over the past four years provide evidence for this proposal. A recent electrophysiological study offers a possible mechanism by which interactions between brain regions take place during memory formation.
Since the description of the famous patient H.M. in the late 1950s [1], it has been recognized that the medial temporal lobe (MTL) plays a crucial role in human memory. Following bilateral resection of the MTL to alleviate persistent epilepsy, H.M. became unable to remember new facts and events for more than a few minutes. Neuropsychological findings such as these indicate that an intact MTL is necessary for normal long-term memory functioning, although it is difficult to discern from such data the extent to which the MTL contributes to memory formation, as opposed to memory storage and retrieval. Studying the neural correlates of memory formation
The question of which neural structures are associated with the formation of lasting human memories can be addressed by recording neural activity while items are initially encoded into memory. A particularly powerful method makes use of the ‘subsequent memory procedure’. http://tins.trends.com
In this approach, neural activity is recorded while volunteers study a sequence of items, after which memory for the items is tested. The activity elicited by items at the time of study is then sorted according to whether the items are remembered or forgotten in the subsequent memory test. Differences between these two types of activity (‘subsequent memory effects’) are taken as candidate neural correlates of encoding. Thus, the procedure provides information about neural structures in which activity is correlated with successful encoding. However, it is important to note that it does not identify which structures are necessary for encoding to occur. The subsequent memory approach has been used with scalp-recorded electrophysiological activity (event-related potentials, or ERPs) for >25 years [2,3]. These studies have provided valuable
information about temporal aspects of the formation of memory traces, but they have been less helpful in identifying the associated brain regions. This is because of both the difficulties of localizing sources of scalp electrical activity and the relative insensitivity of scalp recordings to neural activity in deep brain structures such as the MTL. Non-invasive detection of neural activity with high spatial resolution became possible with the introduction of functional neuroimaging techniques. Memory encoding has been studied since the earliest days of the application of these techniques to cognition [4], but it was with the advent of ‘event-related’functional MRI (fMRI) [5] that their full potential was realized. A key development came in 1998, when two papers using the subsequent memory approach with event-related fMRI were published back-to-back in Science [6,7]. These papers described studies of the encoding of visually presented words and pictures, respectively. They showed that relatively greater fMRI signals were elicited by subsequently remembered items in the MTL, among other regions, highlighting a role for the MTL in successful memory encoding. Evidence has now accumulated to suggest that the MTL (both cortex and hippocampus proper) plays a role in encoding across a wide range of task conditions and stimulus materials [3,8–10]. Notwithstanding the puzzling failure to identify encodingrelated MTL activity in some studies
0166-2236/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0166-2236(02)02176-8
280
Research Update
(e.g. see Ref. [11]), these findings provide support for the idea that the MTL plays a key role in memory formation. Recent findings regarding the role of the medial temporal lobe
The precise role played by the MTL in memory formation has, however, remained elusive. A recent study, which combined the temporal resolution of electrophysiological recordings with the spatial resolution usually available only with functional neuroimaging, could provide important insights. Fell et al. [12] asked unilateral temporal-lobe epilepsy patients to study short word lists, and to attempt to recall the words after a short distraction task. As the words were being studied, electrical activity was recorded from bilateral, multicontact depth electrodes that had been inserted into the MTL for use in clinical investigations. The authors analysed the oscillatory activity at around 40 Hz (socalled ‘gamma’ activity) that was induced in the unaffected hemisphere by the presentation of the words. The degree to which these oscillations were synchronized between the anterior MTL cortex (perirhinal and entorhinal regions) and the hippocampus was found to depend on whether the words were subsequently remembered or forgotten. For remembered words, synchronization was relatively higher during the first few hundred milliseconds after word onset and relatively lower around one second after onset. In both structures, the initial increase in synchronization for remembered words was accompanied by a reduction in the amount of gamma activity. The paper by Fell et al. follows an earlier report on a related data set. Fernández et al. [13] focused on temporal aspects of the interaction between the anterior MTL cortex and the hippocampus. Whereas the ERPs elicited by subsequently remembered and forgotten words differed in both regions, the onset of this difference began some 200 ms earlier (at around 300 ms) in cortex than in the hippocampus. Regardless of their other implications, the findings from both studies point to involvement of the anterior MTL cortex as well as the hippocampus in the initial stages of memory formation. Anterior MTL activations have not generally been reported in event-related fMRI studies of encoding. However, using the same experimental procedure as Fernández http://tins.trends.com
TRENDS in Neurosciences Vol.25 No.6 June 2002
et al., and an fMRI method optimized for the detection of activity in the anterior MTL, Strange et al. [14] recently described subsequent memory effects in both hippocampus and perirhinal cortex. Hippocampal involvement in memory encoding has also been observed at the single neuron level in epilepsy patients implanted with depth electrodes. Cameron et al. [15] found that when patients were asked to remember word pairs, the firing rates of hippocampal neurons predicted whether the pairs would subsequently be remembered. Interestingly, the majority of the neurons showed relatively lower firing rates in response to subsequently remembered items – a finding opposite to what most researchers would have predicted on the basis of fMRI findings. Implications
Together, the studies discussed here contribute to our understanding of the neural bases of memory formation in at least two ways. First, the data provide strong evidence that the MTL is directly involved in human memory encoding, and, thus, that its role in memory is not confined to storage or retrieval functions. Second, the findings suggest that different MTL regions act in concert to support memory encoding, and that the nature of their interaction within the first second or so after a stimulus event influences its subsequent memorability. Future directions
Of course, many questions remain regarding the role of the MTL during memory encoding. One can speculate about the functional significance of phase locking of gamma activity [16], but its functional significance, origin and relationship to subsequent memory effects seen in other measures are uncertain. It is also unclear whether the phase locking observed between MTL structures is restricted to the 40 Hz frequency band or, indeed, whether it is restricted to structures within the MTL. In this respect, it is noteworthy that the most consistent subsequent memory effects identified in fMRI studies have been in the inferior prefrontal cortex. It has been suggested that the prefrontal cortex supports working memory processes engaged by the demands of the encoding task, and that the products of these processes are relayed to the MTL (e.g. see Ref. [17]). In light of these suggestions, it would be of
considerable interest to determine whether synchrony between frontal and MTL regions in the gamma band, or in other frequency ranges, varies according to subsequent memory performance. What type of memory is reflected by subsequent memory effects in the MTL? Most studies using the subsequent memory approach have tested memory using either recall or recognition tasks. It remains to be established whether MTL subsequent memory effects are found when memory is tested in other ways – most importantly, when it is tested with ‘indirect’ tests that tap ‘implicit’, rather than ‘explicit’, memory for the study items. According to prevailing views of hippocampal function [18], this structure is unnecessary for implicit memory, and thus should not exhibit subsequent memory effects for indirect tests. Studies along these lines would therefore be of considerable theoretical interest. Acknowledgements
Our research is supported by the Wellcome Trust and a cooperative award from the UK Medical Research Council. References 1 Scoville, W.B. and Milner, B. (1957) Loss of recent memory after bilateral hippocampal lesions. J. Neurol. Neurosurg. Psychiatry 20, 11–21 2 Friedman, D. and Johnson, R., Jr (2000) Eventrelated potential (ERP) studies of memory encoding and retrieval: a selective review. Microsc. Res. Techn. 51, 6–28 3 Paller, K.A. and Wagner, A.D. (2002) Observing the transformation of experience into memory. Trends Cognit. Sci. 6, 93–102 4 Fletcher, P.C. et al. (1997) The functional neuroanatomy of episodic memory. Trends Neurosci. 20, 213–218 5 Rosen, B.R. et al. (1998) Event-related functional MRI: past, present, and future. Proc. Natl. Acad. Sci. U. S. A. 95, 773–780 6 Wagner, A.D. et al. (1998) Building memories: remembering and forgetting of verbal experiences as predicted by brain activity. Science 281, 1188–1191 7 Brewer, J.B. et al. (1998) Making memories: brain activity that predicts how well visual experience will be remembered. Science 281, 1185–1187 8 Fernández, G. and Tendolkar, I. (2001) Integrated brain activity in medial temporal and prefrontal areas predicts subsequent memory performance: Human declarative memory formation at the system level. Brain Res. Bull. 55, 1–9 9 Rugg, M.D. Functional neuroimaging of memory. In Handbook of Memory Disorders (2nd edn) (Baddeley, B. et al., eds), John Wiley & Sons (in press) 10 Brewer, J.B. and Moghekar, A. Imaging the medial temporal lobe: exploring new dimensions. Trends Cognit. Sci. (in press) 11 Otten, L.J. and Rugg, M.D. (2001) Taskdependency of the neural correlates of episodic
Research Update
encoding as measured by fMRI. Cereb. Cortex 11, 1150–1160 12 Fell, J. et al. (2001) Human memory formation is accompanied by rhinal-hippocampal coupling and decoupling. Nat. Neurosci. 4, 1259–1264 13 Fernández, G. et al. (1999) Real-time tracking of memory formation in the human rhinal cortex and hippocampus. Science 285, 1582–1585 14 Strange, B.A. et al. (2002) Dissociable human perirhinal, hippocampal, and parahippocampal
TRENDS in Neurosciences Vol.25 No.6 June 2002
roles during verbal encoding. J. Neurosci. 22, 523–528 15 Cameron, K.A. et al. (2001) Human hippocampal neurons predict how well word pairs will be remembered. Neuron 30, 289–298 16 Tallon-Baudry, C. and Bertrand, O. (1999) Oscillatory gamma activity in humans and its role in object representation. Trends Cognit. Sci. 3, 151–161 17 Buckner, R.L. et al. (2000) Cognitive neuroscience of episodic memory encoding. Acta Psychol. 105, 127–139
281
18 Squire, L.R. and Zola, S.M. (1997) Amnesia, memory and brain systems. Philos. Trans. R. Soc. London Ser. B 352, 1663–1673
Leun J. Otten* Michael D. Rugg Institute of Cognitive Neuroscience and Dept of Psychology, University College London, London, UK WC1N 3AR. *e-mail: [email protected]
Response: The birth of a memory Guillén Fernández, Jürgen Fell and Pascal Fries Otten and Rugg have elegantly integrated the results of the study by Fell and colleagues [1] into the current knowledge about how experiences are transformed into memories. One of their central conclusions is that the findings ‘…point to the involvement of anterior MTL [medial temporal lobe] cortex as well as the hippocampus in the initial stages of memory formation.’ Despite these important aspects of localization and of the temporal sequence of MTL substructure involvement [2], we would like to highlight another central aspect of that study: namely, that a particular parameter of neuronal activity, rhinal–hippocampal phasesynchronization in the gamma-frequency band (~40 Hz), correlated with subsequent recall. This is conceptually important because memories might reside in neuronal assemblies rather than in individual neurons [3], and synchronization is an ideal mechanism to bind neurons into assemblies [4]. In early visual processing, neurons that encode features of a complex visual percept are associated in functional assemblies through gamma-frequency synchronization [5]. Furthermore, when sensory stimuli are perceptually or attentionally selected and the respective neurons are bound together to raise their saliency, gamma-frequency synchronization among these neurons is also enhanced [5]. Gamma-mediated coupling, and its modulation by attention, is not limited to the visual modality: it is also found in the auditory [6] and somatosensory domains [7]. Moreover, gamma oscillations allow visuo–motor binding between posterior and central brain regions [8] and are involved in higher order cognitive operations, such as http://tins.trends.com
working memory [9] or learning of new associations in a conditioning task [10]. In addition to being a means for dynamically binding neurons into assemblies, gamma-frequency synchronization appears to be the prime candidate mechanism for stabilizing cortical connections among members of a neural assembly over time. On the one hand, neurons increase or decrease the strength of their synaptic connections depending on the precise coincidence of their activation [11] and gammafrequency synchronization provides exactly the required temporal precision. On the other hand, strengthened connections among neurons in a ‘memory assembly’ might facilitate its later recall. In general, EEG signals reflect postsynaptic potentials, which are mainly determined by the average activity of local neuronal populations [12]. In other words, EEG oscillations of ~40 Hz are based on clusters of discharges occurring about every 25 ms. Although the exact mechanisms underlying the generation of gammafrequency synchronization are as yet unclear, several studies have begun to shed light on this issue. In the hippocampus, gamma-frequency synchronization is driven by interneuron network oscillations and intrinsic membrane resonance, as revealed in slice preparations [13]. Bragin and colleagues [14] have identified gammaactivity in the hippocampus of behaving rats and have shown that gamma-activity is most prominent in the dentate gyrus, the main hippocampal recipient of input from the neocortex via the entorhinal cortex and perforant path [15]. With regard to ‘phase locking of gammaactivity’, Otten and Rugg state that its ‘functional significance, origin and relationship to subsequent memory effects
seen in other measures are uncertain.’ However, taking the previously mentioned findings into account, we arrive at a different conclusion. Several lines of evidence suggest that gamma-frequency synchronization plays a general role in binding neurons into assemblies over short and long distances. In memory formation, gamma activity has the optimal frequency to support the transformation of a temporary representation into a durable memory trace by strengthening synaptic connectivity. Therefore, we hypothesize that the rhinal–hippocampal coupling observed by Fell and colleagues [1] enables information transfer to the hippocampus and initiates the mnemonic operation of memory encoding, which leads to synaptic plasticity and occurs within the first second of an event that can be remembered later on. References 1 Fell, J. et al. (2001) Human memory formation is accompanied by rhinal–hippocampal coupling and decoupling. Nat. Neurosci. 4, 1259–1264 2 Fernández, G. et al. (1999) Real-time tracking of memory formation in the human rhinal cortex and hippocampus. Science 285, 1582–1585 3 Rumelhart, D.E. and McClelland, J.L. (1986) Parallel Distributed Processing: Explorations in the Microstructure of Cognition, MIT Press 4 Singer, W. (1999) Neuronal synchrony: a versatile code for the definition of relations? Neuron 24, 49–65 5 Engel, A.K. et al. (2001) Dynamic predictions: oscillations and synchrony in top-down processing. Nat. Rev. Neurosci. 2, 704–716 6 Tiitinen, H. et al. (1993) Selective attention enhances the auditory 40-Hz transient response in humans. Nature 364, 59–60 7 Desmedt, J.E. and Tomberg, C. (1994) Transient phase-locking of 40 Hz electrical oscillations in prefrontal and parietal human cortex reflects the process of conscious somatic perception. Neurosci. Lett. 168, 126–129 8 Rodriguez, E. et al. (1999) Perception’s shadow: long-distance synchronization of human brain activity. Nature 397, 430–433 9 Tallon-Baudry, C. et al. (1998) Induced gammaband activity during the delay of a visual
0166-2236/02/$ – see front matter © 2002 Elsevier Science Ltd. All rights reserved. PII: S0166-2236(02)02177-X