Brain(2001).124,849-881 INVITED REVIEW Frontal lobes and human memory Insights from functional neuroimaging P.C.Fletcher!and R.N.A.Henson2 itive Neurolo and Instinte of Cognitive Neuroscience.Universiry College London,UK E-mail:pcf22@cam.ac.uk The new functional neuroimaging techniques,PET and lateral.dorsolateral and anterior -that are consistently to the updatingl e sel s They have had a y We also understanding of the role of the frontal cortex in memory acknowledge a number of empirical inconsistencies asso- ciated with this synthe sis,and suggest possible reasons a attem耳 di synt or the fir the More generally e pred at the r subdivisions of the frontal cortex will ultimately depend on hese different aspects of memory have usually beer a fuller cognitive psychological fractionation of memory isol n,w suggest that there is suf proc an en erpr e that wi通 e guided an uch h rthwhile.We conc e in s will pr ovide nt rt of this on three regions of the lateral frontal corte entro Keywords:frontal;memory:functional MRI;PET Abbreviations:AFC anterior frontal cortex;DLFC dorsolateral frontal cortex;ERP =event-related potential;FC functional MRI;HERA =hemispheric encoding-retrieval asymmetry:LTM long-term memory: ed c tions vary widely,ho their pre tions within FC.In this artic vever,th idity of this level takes heed of the broad anatomical variation of activations within FC thesis is that functional specialization maging offers a umber of advantages over to unc s deal at least three anatomically distinct frontal regions.This -anatomical specialzation ha s prove across different patients PET and in particular RI offer deficits Oxford University Press 2001
Brain (2001), 124, 849–881 INVITED REVIEW Frontal lobes and human memory Insights from functional neuroimaging P. C. Fletcher1 and R. N. A. Henson2 1Research Department of Psychiatry, Cambridge University, Correspondence to: P. C. Fletcher, Box 189, Research Addenbrooke’s Hospital, Cambridge and 2Wellcome Department of Psychiatry, Cambridge University, Department of Cognitive Neurology and Institute of Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK Cognitive Neuroscience, University College London, UK E-mail: pcf22@cam.ac.uk Summary The new functional neuroimaging techniques, PET and lateral, dorsolateral and anterior—that are consistently functional MRI (fMRI), offer sufficient experimental activated in these studies, and attribute these activations flexibility and spatial resolution to explore the functional to the updating/maintenance of information, the selection/ neuroanatomical bases of different memory stages and manipulation/monitoring of that information, and the processes. They have had a particular impact on our selection of processes/subgoals, respectively. We also understanding of the role of the frontal cortex in memory acknowledge a number of empirical inconsistencies assoprocessing. We review the insights that have been gained, ciated with this synthesis, and suggest possible reasons and attempt a synthesis of the findings from functional for these. More generally, we predict that the resolution imaging studies of working memory, encoding in episodic of questions concerning the functional neuroanatomical memory and retrieval from episodic memory. Though subdivisions of the frontal cortex will ultimately depend on these different aspects of memory have usually been a fuller cognitive psychological fractionation of memory studied in isolation, we suggest that there is sufficient control processes, an enterprise that will be guided and convergence with respect to frontal activations to make tested by experimentation. We expect that the neurosuch a synthesis worthwhile. We concentrate in particular imaging techniques will provide an important part of this on three regions of the lateral frontal cortex—ventro- enterprise. Keywords: frontal; memory; functional MRI; PET Abbreviations: AFC anterior frontal cortex; DLFC dorsolateral frontal cortex; ERP event-related potential; FC frontal cortex; fMRI functional MRI; HERA hemispheric encoding–retrieval asymmetry; LTM long-term memory; VLFC ventrolateral frontal cortex; WM working memory Introduction PET and functional MRI (fMRI) have demonstrated consistent specialization might not exist within FC. We believe this activations of the frontal cortex (FC) in a number of memory approach will prove more fruitful than attempting to define tasks. Interpretations of these activations vary widely, how- a general and abstract function for FC as a whole. Ultimately ever, as do their precise locations within FC. In this article, however, the validity of this level of functional specialization we review these findings and offer a new interpretation that is best judged by its success in explaining extant neuroimaging takes heed of the broad anatomical variation of activations and neuropsychological data. within FC. Neuroimaging offers a number of advantages over Our main hypothesis is that functional specialization, neuropsychology with regard to understanding the functional within the context of memory-related processes, exists across parcellation of FC. First, neuropsychological studies deal at least three anatomically distinct frontal regions. This with lesions that often differ markedly in size and location principle of functional–anatomical specialization has proved across different patients. PET and in particular fMRI offer a remarkably successful in, for example, the study of the visual more precise spatial characterization of functional cortex, and we see no a priori reason why analogous differentiation across FC. Secondly, the memory deficits © Oxford University Press 2001
850 P.C.Fletcher and R.N.A.Henson produced by frontal lesions tend to be subtle,and it is likely only meaningful to the extent that the psychological theory that the sorts of memory processes subserved by FC are of task performance is accurate.A specific example of this upstre of observe haviours (Burg isthe assumption that a task manipulation changes behavioural performance with varving degrees of frontal mediation and compensatory strategies.Functional neuro 1996:Donders,1969)is particularly relevant to simple ctive methods of analysing imaging data,in whicl of rol)is subtr memory process.For example,they can examine separately another task that is assumed to differ only in the single psychological process of interest.The difference between the on mad wo ta sks may in to function inde endently of other brain systems with which measures aone)This is why the activations cannot be evaluated without but no m may b h。 er a regi .This pro cquisition of whole-hrain im lated ved by FC tio of spatially distributed functional networks of activity Moreove analytical techniques have been developed tha allow the cha on connectivity others const but do so witho cha ha ions rather than the stimuli). (e.g the 1997) It is important to raise this problem -that neuroimaging I acti vation obs are only context of 0 subjects show onta I activation in ass tion with recogn re discussed in the Conclusions section).This is b au ing et al. Rugg et c e recent neurom ng nndi that such tasks may be performed relatively ontext of specific theories In the final section.howeve normally even in the face of widespread frontal damage.On possibility is that such activations are theoretical note also thet the pproa ontain important additional information about the wa 000).When the Talairach of activation healthy subjects perform the task.If so. the failure of naxima these authors found a ural m res region that asks.but failed to ind evidence for discrepancies between functional imaging and neuropsycho of the maxima within these regions.Our approach begins logical data may point to flaws in our cognitive models o with prior,anatomically defined regions and. while accepting asks are pert mance is me the functional imaging techniques rather than,as has beer unt emeroes from diff ntial activations of these re suggested,a weaknes We propose to distinguish be een activatio use imaging to addre nctiona regi y state of e nfined to the lateral understanding of the types of processes subserved by FC.In they are the regions most commonly activated in memory- most functional neuroimaging experiments,changes in the elated tasks.DLFC con of pon tal gyrus E the to a specific psychological process sedly isolated hy the fron olar area lying anterior to the anteriormost extent the task manipulation.The pattern of brain activity is therefore of the inferior frontal gyrus (Fig.1).We make these
850 P. C. Fletcher and R. N. A. Henson produced by frontal lesions tend to be subtle, and it is likely only meaningful to the extent that the psychological theory that the sorts of memory processes subserved by FC are of task performance is accurate. A specific example of this some distance ‘upstream’ of observed behaviours (Burgess, problem is the assumption that a task manipulation changes 1997). Patients may, for example, achieve comparable only a single cognitive process, leaving other processes behavioural performance with varying degrees of frontal unaffected. This assumption of ‘pure insertion’ (Friston et al., mediation and compensatory strategies. Functional neuro- 1996; Donders, 1969) is particularly relevant to simple imaging offers the possibility of detecting differences in the subtractive methods of analysing imaging data, in which strategies that subjects or patients employ. Thirdly, functional mean brain activity during the performance of one task (the neuroimaging techniques can elucidate different stages of a control) is subtracted from that during the performance of memory process. For example, they can examine separately another task that is assumed to differ only in the single the encoding and retrieval of memories, a dissociation that psychological process of interest. The difference between the cannot be made with confidence from anterograde memory two tasks may in fact be accompanied by numerous cognitive deficits following frontal lobe lesions. Finally, FC is unlikely changes (which may not be evident from behavioural to function independently of other brain systems with which measures alone). This is why the ‘activations’ reported it interacts (Fuster, 1997). Neuropsychological study can by neuroimaging experiments cannot be evaluated without show whether a region is necessary for a given task, but not reference to the control task. This problem may be particularly usually the broader system of which that region forms a part. relevant to the relatively high-level (non-automatic) and interAcquisition of whole-brain images enables the characteriza- related processes generally believed to be subserved by FC. tion of spatially distributed functional networks of activity. Isolating such processes requires experimental manipulations Moreover, analytical techniques have been developed that that not only engage each of them to different degrees while allow the characterization of the effective connectivity holding the others constant, but do so without changing between different brain regions during task performance lower-level (e.g. perceptual) processes (e.g. changing the (McIntosh and Gonzales-Lima, 1994; Bu¨chel and Friston, instructions rather than the stimuli). 1997). It is important to raise this problem—that neuroimaging It has been suggested that a regional activation observed ‘activations’ are only interpretable in the context of a in functional imaging tells us little about the necessity of particular theory of task performance and often with respect that region for task performance (Price and Friston, 1999; to a specific control—at the outset of this review (other Fletcher, 2000). For example, a number of studies of healthy problems associated with current neuroimaging experiments subjects show frontal activation in association with recogni- are discussed in the Conclusions section). This is because we tion memory (e.g. Tulving et al., 1994b; Rugg et al., 1996) describe and organize recent neuroimaging findings initially in while neuropsychological studies (e.g. Stuss et al., 1994) terms of one or more conventional labels and within the have indicated that such tasks may be performed relatively context of specific theories. In the final section, however, we normally even in the face of widespread frontal damage. One offer a re-evaluation of the prominent findings within a possibility is that such activations are epiphenomenal, in the modified theoretical framework. We note also that our sense that they are not directly task-related. A more interesting approach differs from formal meta-analyses, such as that possibility, however, is that the functional imaging data recently performed by Duncan and Owen (Duncan and Owen, contain important additional information about the way 2000). When plotting the Talairach coordinates of activation healthy subjects perform the task. If so, the failure of maxima from a number of studies, these authors found a behavioural measures to distinguish between the performance subset of lateral and dorsomedial FC regions that were of a task in patients and in controls may indicate a limitation commonly activated across a range of different cognitive or insensitivity in the behavioural measures. That is, tasks, but failed to find evidence for functional segregation discrepancies between functional imaging and neuropsycho- of the maxima within these regions. Our approach begins logical data may point to flaws in our cognitive models of with prior, anatomically defined regions and, while accepting how tasks are performed and how performance is measured. some errors in the attribution of functional activations to In this sense, such discrepancies may represent a strength of these regions, examines whether a consistent theoretical the functional imaging techniques rather than, as has been account emerges from differential activations of these regions. suggested, a weakness. We propose to distinguish between activations occurring The use of functional imaging to address functional in the following FC regions: ventrolateral FC (VLFC), specialization within FC is, however, problematic. The most dorsolateral FC (DLFC) and anterior FC (AFC). We chose fundamental problem lies in the rudimentary state of current these regions, confined to the lateral aspect of FC, because understanding of the types of processes subserved by FC. In they are the regions most commonly activated in memorymost functional neuroimaging experiments, changes in the related tasks. DLFC consists of the area lying superior to the haemodynamic response of a region are correlated with a inferior frontal gyrus and VLFC to the area below it, i.e. the manipulation of the subject’s task. This change is attributed inferior frontal gyrus. AFC is defined more arbitrarily as to a specific psychological process supposedly isolated by the frontopolar area lying anterior to the anteriormost extent the task manipulation. The pattern of brain activity is therefore of the inferior frontal gyrus (Fig. 1). We make these
Frontal lobes and human memory 851 DLFC AFC Fig.1 Le ving our working subdivi marked by a ine drawn vertically at the anterior edge of the inferior frontal gyrus. distinctions (in addition to the left-right lateralization of the we cannot be certain of the precise relationship between al landma am subiects and of the likelihood that ultimately,these broad areas will themselves be shown to be The nature of the contribution of the frontal lobe to ally subdivided heaona behind thi ory is cloude d by the divis of the exp inf to ack ntain info nds,and long-term memory (LTM),the ability to retain separate regional 'frontal information for much longer periods.While there are good Our d these two types of men ovide some clues to imaging studies of encoding and retrieval in LTM.for to Brodmann areas 44. 5 and 47.DLFC to areas 9 and 4 and 10.It is our inter likelyto entail maintaining and manipulating ever, on in WN ded into LTM. pic sulcal/gyral features (ont which th that similar FC dissociat ons of function have been PET and fMRI activations are mapped)and the bounda both LTM and WM imaging studies, yet these (Rolan et eptions (Wagner 1999) diffe in the size of Broca's are oduce the evidence is to consider each field sep 10 individuals,the microscopic boundaries bearing litt before subsequently discussing how they may converge stent relationship to mac scopic landmarks (Amunts We therefore address the patterns of memory-related F et al,1999 exerc n re Pirst,we Brodmann's definitions.The chosen subdivisions are also LTM retrieval).Secondly.in the likely to reflect differences in patterns of connectivity oncluding section.we attempt a more general interpretation (Passingham.1993:Fuster. 1997).0nce more,however. that extends to FC activations across the different domains
Frontal lobes and human memory 851 Fig. 1 Left-sided view of human brain showing our working subdivisions of lateral FC. The border between VLFC and DLFC is marked by the inferior frontal sulcus. The posterior border of AFC is marked by a line drawn vertically at the anterior edge of the inferior frontal gyrus. distinctions (in addition to the left–right lateralization of the we cannot be certain of the precise relationship between regions) with due consideration of the imperfect spatial connectivity and macroanatomical landmarks, and we refrain resolution of the techniques, of the enormous anatomical from further speculation. Finally, we confine our review to variability among subjects, and of the likelihood that, studies of groups of young, healthy individuals. ultimately, these broad areas will themselves be shown to be The nature of the contribution of the frontal lobe to functionally subdivided. The rationale behind this division memory is clouded by the division of the experimental is, on the one hand, an attempt to acknowledge the limited literature into two broad fields: working memory (WM), the spatial information provided by group studies (particularly ability to maintain information temporarily over periods of with PET) and, on the other hand, to avoid treating clearly seconds, and long-term memory (LTM), the ability to retain separate regional responses as undifferentiated ‘frontal’ information for much longer periods. While there are good activations. Our particular subdivisions are based on existing reasons for distinguishing between these two types of memfunctional imaging data rather than microstructural findings, ory, it is also likely that considerable overlap exists between although they may be considered to provide some clues to the frontally mediated processes involved in each. Many the underlying anatomy. Thus, VLFC corresponds loosely imaging studies of encoding and retrieval in LTM, for to Brodmann areas 44, 45 and 47, DLFC to areas 9 and 46 example, are likely to entail maintaining and manipulating and AFC to areas 8 and 10. It is our intention, however, to information in WM. Conversely, information maintained in avoid relying upon the uncertain and inconsistent relationship WM may be encoded into LTM. It is interesting, therefore, between macroscopic sulcal/gyral features (onto which the that similar FC dissociations of function have been proposed PET and fMRI activations are mapped) and the boundaries in both LTM and WM imaging studies, and yet these of the Brodmann areas (Roland et al., 1997; Zilles et al., findings, with certain exceptions (Wagner, 1999), are not 1997). Amunts and colleagues, for example, noted a 10-fold often considered together. Nonetheless, a convenient way to difference in the size of Broca’s area across a group of introduce the evidence is to consider each field separately, 10 individuals, the microscopic boundaries bearing little before subsequently discussing how they may converge. consistent relationship to macroscopic landmarks (Amunts We therefore address the patterns of memory-related FC et al., 1999). Caution must therefore be exercised in relating activation in two stages. First, we consider interpretations of macro- to microanatomy, and we will avoid the use of FC activations offered by researchers within each domain Brodmann’s definitions. The chosen subdivisions are also (WM, LTM encoding, LTM retrieval). Secondly, in the likely to reflect differences in patterns of connectivity concluding section, we attempt a more general interpretation (Passingham, 1993; Fuster, 1997). Once more, however, that extends to FC activations across the different domains
P.C.Fletcher and R.N.A.Henson Frontal function in working memory tasks The term 'working memory'is ger rally used to refer to the currently maintain ability to maintain information on-line,often in the service as he onitoring and h her-level planning of a particular be illustrated or goa ever,the term has diffe om often used to describe the ability of an animal to remembe simply require that the subject decides whether a prob a stimulus for a short period after it is removed (in orde matches one of a set to perform e.g.delayed matching-to-sample tasks).In th cognitive stimulus at a time fror ature.on th the set presented previously.such that,over trials,every sources of information are order to perform timulus has been selected once (without repetition). Thi requires that the wo perspectives etore consi revious responses.According to ol stigation eas a sel tent with thi defici tasks but not typi ally on delayed-matchin Perspectives from animal studies:domain in patients which typical lly include (Petrides nal specializ 98 19 0).Fur theories. These theories co entrate in r particular on dis with DLFC lesions w re imnaired on simple verhal o sociations between ventral and dorsal regions of lateral FC span tasks that require only maintenance of a stimulus on-line ry. hou ymanipula on)(D'Esposito oppo Moreover,the fo mainte prec n the is 0 largely on electrophysio tical recordings and is 3 imnle spatial delaved r sponse tasks or only in more comple extension of the object-spatial (what'versus'where)visua situations,such as self-ordering tasks.Nonetheless,we will n posterior regions(Mishkin ompare the e two genera theories for their ability to accoun ore spec n and the imaging dat The data are i from the human psychological literature. on WM e dorsal to the principal sulcus code for r spatial information al. 1993 gene tial VLEC-DLEC distinction refects all com Perspectives from human cognitive psychology: WM:the'attentional,memorial and response control mechan multiple-component models isms'(Goldman-Rakic 1998).That is,there is no suggestio hich th 194 nerate develoned to account for a range of different wM functions The altemative,process-specific theory proposes that the from temporary maintenance of a single stimulus to the e the type of nipulation of multiple types of inform ratine on that material (Pe Shiffrin.1968).which acted simply as ag derives mainly from animal lesion data (Petrides,1994)and perception and LTM,to a multicomponent system in which has been extende to human lesion data (Petride a number of subsidiary 'slave'systems are coordinated by a wen et e ding to this th nmo centra ex The ave systen tion in WM.This infommation may have been perccived limited-capacity.material-specific stors.concemed with the recently or retrieved from LTM.DLFC.however.supports maintenance of verbal and visuospatial material respectively
852 P. C. Fletcher and R. N. A. Henson Frontal function in working memory tasks more complex processes operating on information that is The term ‘working memory’ is generally used to refer to the currently maintained in WM. These include processes such ability to maintain information on-line, often in the service as monitoring and higher-level planning. of a particular task or goal. However, the term has different The process-specific distinction can be illustrated by connotations in different fields. In the animal literature, it is comparing two types of WM task. ‘Delayed matching tasks’ often used to describe the ability of an animal to remember simply require that the subject decides whether a probe a stimulus for a short period after it is removed (in order stimulus matches one of a set of stimuli held in WM. This task requires maintenance only. In ‘self-ordered tasks’, to perform e.g. delayed matching-to-sample tasks). In the cognitive psychological literature, on the other hand, WM however, the subject must select one stimulus at a time from frequently refers to a mental workspace in which multiple the set presented previously, such that, over trials, every stimulus has been selected once (without repetition). This sources of information are manipulated in order to perform complex problem-solving tasks. We begin by introducing the requires that the subject not only selects stimuli from a set background to these two perspectives, before considering maintained in WM but also updates and monitors the set of recent imaging studies that have attempted to synthesize ideas previous responses. According to Petrides and colleagues, a delayed matching task would engage VLFC, whereas a self- from these traditionally quite distinct fields of investigation. ordering task would engage DLFC. Consistent with this view, DLFC lesions in primates produce deficits on self-ordering tasks but not typically on delayed-matching tasks (Petrides, Perspectives from animal studies: domain- 1995). Self-ordering deficits are also seen following frontal versus process-specific theories lesions in patients, which typically include DLFC (Petrides Two competing ideas concerning functional specialization and Milner, 1982; Owen et al., 1990). Furthermore, a review of FC in WM are ‘domain-specific’ and ‘process-specific’ by D’Esposito and Postle found no evidence that patients theories. These theories concentrate in particular on dis- with DLFC lesions were impaired on simple verbal or spatial sociations between ventral and dorsal regions of lateral FC. span tasks that require only maintenance of a stimulus on-line According to the domain-specific theory, FC is the primary (without any manipulation) (D’Esposito and Postle, 1999). site of WM processes and different regions within FC process Though often placed in opposition, the domain-specific different types of information (Goldman-Rakic, 1987, 1998). and process-specific theories are not necessarily incompatible. Specifically, VLFC is believed to be responsible for the FC may be functionally dissociable according to both the maintenance of stimulus form (object information), whereas type of material and the type of process. Moreover, the DLFC is believed to be responsible for the maintenance of precise site of lesions in the primate DLFC (e.g. Brodmann stimulus location (spatial information). This theory is based areas 9 or 46) can affect whether impairments are seen in largely on electrophysiological recordings and is an simple spatial delayed response tasks or only in more complex extension of the object–spatial (‘what’ versus ‘where’) visual situations, such as self-ordering tasks. Nonetheless, we will processing streams found in posterior regions (Mishkin et al., compare these two general theories for their ability to account 1983). More specifically, Wilson and colleagues found that for the human imaging data. The data are introduced later, FC cells ventral to the principal sulcus code for object after considering an alternative perspective on WM deriving information during a delay, whereas frontal cells within and from the human psychological literature. dorsal to the principal sulcus code for spatial information during a delay (Wilson et al., 1993). More generally, Goldman-Rakic and colleagues suggested that the object– Perspectives from human cognitive psychology: spatial VLFC–DLFC distinction reflects all components of WM: the ‘attentional, memorial and response control mechan- multiple-component models isms’ (Goldman-Rakic, 1998). That is, there is no suggestion Baddeley and Hitch’s theoretical model of WM function of specialization for different WM processes across FC, only (Baddeley and Hitch, 1974) has been highly influential in specialization for the domains over which these processes framing functional neuroimaging studies. This model was operate. developed to account for a range of different WM functions, The alternative, process-specific theory proposes that the from temporary maintenance of a single stimulus to the difference between VLFC and DLFC lies not in the type of manipulation of multiple types of information. It evolved from material being maintained but in the type of processes earlier conceptions of a single short-term buffer (Atkinson and operating on that material (Petrides, 1994, 1995). This theory Shiffrin, 1968), which acted simply as a gateway between derives mainly from animal lesion data (Petrides, 1994) and perception and LTM, to a multicomponent system in which has been extended to human lesion data (Petrides and Milner, a number of subsidiary ‘slave’ systems are coordinated by a 1982; Owen et al., 1990). According to this theory, VLFC common ‘central executive’. The slave systems, the supports processes that transfer, maintain and match informa- ‘phonological loop’ and ‘visuospatial scratch-pad’, are tion in WM. This information may have been perceived limited-capacity, material-specific stores, concerned with the recently or retrieved from LTM. DLFC, however, supports maintenance of verbal and visuospatial material respectively
Frontal lobes and human memory 853 age an n et c nents of distinguished a phond logical store' verbal WM.Paulesu and colleagues (Paulesu) control process'(Baddeley,1986).Verbal material is assu compared a verbal Sternberg task with a control task in enter the p store wher ymed with a target by (subvocal)rehearsal via the articulatory control process omparison revealed left inferior parietal activation but no The proposal of a 'visual cache'and an 'inner scribe (L0g1 995)repre ents an istinctior FCcivatioRAwhandcoleaguescompared2backn age-rehearsa (Fig.2D)with a continuous subvocal repetition taskwithno bdsof WM that cmploy storage requirement(Awh er al..1996).Again,activation of pariet ved. bu nd in the left inferior p of distraction).This would correspond to use of the slave store,which was engaged in the memory tasks,and the left systems of the WM model.Manipulation refers to the VLFC in subvocal articulatory rehearsal,which was assumed of th is bein the WM model.We ng mainte Early imaging studies of such tasks have tended to support the neuropsychological evidence Maintenance of spatial and object ial Both have hee ma of s to the left for verbal material and to the right for spa and object information.In a study using a spatial Ste hav task (Fig.2B).activations were seen in everal een the ject of more recent in to th 10 03 thus relevant to the domain versus process-specific parietal cortex,right dorsal debate outlined above. cortex and right VLFC.(The FC activations resulting from ompa the h its Maintenance of verbal infor A common test of maintenance in WM is the Sternberg task and (Fig.2).Subjects are presented with a'memory set' spatial material are made.)Smith and colleagues re typically three which a n ren ved dings in V15 -and right-late lized the probe stimulus was one of the stimuli in the memory set. To isolate brain areas involved in maintenance from tho involved in perceptual or m ta the 20 pes (o contrasted against those obtained in a control task in which the memory set and probe item are presented simultaneously lateralized pari etal cortex, any mem 24S por ated with two abstract sha and hemisphere regions,including parietal,dorsal premotor and after a 3-s delay,a single probe shape prompting a yes-no 1996).Similar regions test of the task was to decide were y when the the p one of th e In the te non-verbalizable symbols (Fig.2A and C)(Paulest WM.the task was to decide whether the probe matched one et al 1993).This left hemisphere network of the VLFC f the memory set in its location (regardless of its shape) parietal and motor areas (plus right cerebellum)is a consistent The regions mon active in the object task than in the spatial
Frontal lobes and human memory 853 An important distinction within the slave systems of the finding in studies of maintenance in verbal WM (Smith and WM model is between passive storage and active rehearsal. Jonides, 1997; Henson et al., 2000b). In the case of the phonological loop, for example, Baddeley To distinguish the storage and rehearsal components of distinguished a ‘phonological store’ from an ‘articulatory verbal WM, Paulesu and colleagues (Paulesu et al., 1993) control process’ (Baddeley, 1986). Verbal material is assumed compared a verbal Sternberg task with a control task in to enter the phonological store, where it is vulnerable to which subjects judged whether letters rhymed with a target interference and/or rapid decay over time. The rapid decay letter [a task that is believed to require the same articulatory of material in the phonological store can be offset, however, processes as those used in rehearsal (Besner, 1987)]. This by (subvocal) rehearsal via the articulatory control process. comparison revealed left inferior parietal activation but no The proposal of a ‘visual cache’ and an ‘inner scribe’ (Logie, FC activation. Awh and colleagues compared a 2-back task 1995) represents an analogous storage–rehearsal distinction (in which a positive response is required whenever the current within the visuospatial scratchpad. stimulus matches the stimulus presented two trials previously) For the purpose of this review, we make a coarse distinction (Fig. 2D) with a continuous subvocal repetition task with no between imaging studies of WM that employ maintenance storage requirement (Awh et al., 1996). Again, activation of tasks and those that employ manipulation tasks. Maintenance the inferior parietal cortex was observed, but no difference refers to the process of keeping information in mind in the in FC activation was seen. Both studies therefore implicate absence of an external stimulus (and perhaps in the presence the left inferior parietal cortex as the locus of a phonological of distraction). This would correspond to use of the slave store, which was engaged in the memory tasks, and the left systems of the WM model. Manipulation refers to the VLFC in subvocal articulatory rehearsal, which was assumed reorganization of the information that is being maintained, to be engaged in both memory and control tasks and therefore and would correspond to the use of the central executive in not observed in the subtractions. the WM model. We begin by considering maintenance tasks. Early imaging studies of such tasks have tended to support the neuropsychological evidence for a role of posterior Maintenance of spatial and object information regions in the passive storage of material and of posterior Imaging studies of WM using non-verbal material have FC in the rehearsal of material. Both have been lateralized focused on differences between the maintenance of spatial to the left for verbal material and to the right for spatial and object information. In a study using a spatial Sternberg material. We then consider manipulation tasks. These have task (Fig. 2B), activations were seen in several right been the subject of more recent imaging studies which have hemisphere regions, broadly homologous to those seen in focused on dissociations between VLFC and DLFC and are verbal maintenance tasks (Jonides et al., 1993). These thus relevant to the domain-specific versus process-specific included the right parietal cortex, right dorsal premotor debate outlined above. cortex and right VLFC. (The FC activations resulting from comparison of the Sternberg task with its control are sometimes bilateral for both verbal and spatial material. Maintenance of verbal information However, the left–right verbal–spatial lateralization is A common test of maintenance in WM is the Sternberg task normally clearer when direct comparisons of verbal and (Fig. 2). Subjects are presented with a ‘memory set’ of spatial material are made.) Smith and colleagues reported typically three to nine stimuli, which are then removed for similar findings in a direct comparison of visuospatial and several seconds before the appearance of a single probe verbal Sternberg tasks, the networks of parietal, dorsal stimulus. The goal of the subject is to decide whether or not premotor and VLFC regions being left- and right-lateralized the probe stimulus was one of the stimuli in the memory set. for verbal and spatial tasks, respectively (Smith et al., 1996). To isolate brain areas involved in maintenance from those Smith and Jonides used an object version of the Sternberg involved in perceptual or motor components of the task, task that tested memory for abstract shapes (for which spatial functional images obtained during the Sternberg task can be location was irrelevant) (Fig. 2C) (Smith and Jonides, 1995a). contrasted against those obtained in a control task in which This task produced activations that were predominantly leftthe memory set and probe item are presented simultaneously, lateralized, including the inferior parietal cortex, inferior alleviating any memory requirement. Using a verbal Sternberg temporal cortex and left VLFC. In a direct comparison of task in which the stimuli were letters (Fig. 2A), Awh and object and spatial maintenance (Smith and Jonides, 1995b), colleagues reported significant activations in several left participants were presented with two abstract shapes and, hemisphere regions, including parietal, dorsal premotor and after a 3-s delay, a single probe shape prompting a yes–no ventral premotor/VLFC (Awh et al., 1996). Similar regions response. In the test of object WM, the task was to decide were implicated by Paulesu and colleagues when they whether the probe matched one of the memory set in shape compared two Sternberg tasks, one using letters and one (regardless of its location on the screen). In the test of spatial using non-verbalizable symbols (Fig. 2A and C) (Paulesu WM, the task was to decide whether the probe matched one et al., 1993). This left hemisphere network of the VLFC, of the memory set in its location (regardless of its shape). parietal and motor areas (plus right cerebellum) is a consistent The regions more active in the object task than in the spatial
P.C.Fletcher and R.N.A.Hensor (A)Typical Verbal Maintenance Task Maintenance BHR☐b 阿□丽 (B)Typical Spatial Maintenance Task Maintenance 口可□□0 .□可9□ C)Typical Object Mainte Task Maintenance 回□￥ ✉回☐啊 D)n-back Task 2-hack Task 0-back Task(target=X) BHCRC... BHC RX Alphabetize BHC2H Fig.2 Schematic representation of working memory tasks. vell as the ventral FC in the m WM.Howe ver the differ (Smith and Jonides,1995a.b).The areas more active in the object and spatial information appears more likely to reflect spatial task k were the right VLFC and the right posterior a left-right lateralization than a ventral-dorsal one.spatial tasks (Belger).the spatial task activated the righ arise when the objects are faces.for which obiect tasks tend to produce VLFC activation and spatial tasks DLFCactivation ry si (e.g.Courtney er )One possit that face the partici nsiomainainmoreha1gdhni2o ical studies suggest that face-selective sha pes,which is beyond the normal visuospatial memor neurones are restricted to ventral FC regions (O'Scalaidhe span)(MeCa rthy et al.,1996).Finally,in a study compar etal,19971. esp 【a objec has proved difficul to isolate storage activation was observed in the former and left DLFC in the latter.These studies suggest a role for the dorsal as movements has little support.because activations of frontal
854 P. C. Fletcher and R. N. A. Henson Fig. 2 Schematic representation of working memory tasks. task were the left posterior parietal cortex and left inferior well as the ventral FC in the maintenance of information in temporal cortex, a subset of the areas implicated in the study WM. However, the difference between the maintenance of (Smith and Jonides, 1995a, b). The areas more active in the object and spatial information appears more likely to reflect spatial task were the right VLFC and the right posterior a left–right lateralization than a ventral–dorsal one, spatial parietal, right anterior occipital and right premotor cortices. tasks activating the right FC and object tasks activating the In another study comparing spatial and object Sternberg left or bilateral FC. One exception to this pattern appears to tasks (Belger et al., 1998), the spatial task activated the right arise when the objects are faces, for which object tasks tend DLFC, whereas the object task activated bilateral DLFC and to produce VLFC activation and spatial tasks DLFC activation left VLFC. A very similar pattern was reported by McCarthy (e.g. Courtney et al., 1996). One possibility is that faces and colleagues (though in this case the memory task required constitute a special class of visual objects [e.g. the participants to maintain more than 18 different locations/ electrophysiological studies suggest that face-selective FC shapes, which is beyond the normal visuospatial memory neurones are restricted to ventral FC regions (O’Scalaidhe span) (McCarthy et al., 1996). Finally, in a study comparing et al., 1997)]. a spatial delayed response task with an object delayed It has proved difficult to isolate storage from rehearsal matching task (Baker et al., 1996a), greater right DLFC processes in spatial and object maintenance tasks. The hypoactivation was observed in the former and greater left DLFC thesis that visuospatial rehearsal corresponds to planned eye in the latter. These studies suggest a role for the dorsal as movements has little support, because activations of frontal
Frontal lobes and human memory 855 mply to heneve T a prespe of information viewed as propo rtional to theworking memory load'the istent with neuroimaging studies of spatial attention total demand placed on the maintenance and/or manipulation of the pro nd colle Coull and 1998)A tentativ hypothesis is tha visuospatial information is stored as abstract or object visua representations in the occipital cortex and inferor tempora linearly increasing functi aps com well as nted by as between these areas and the righ rea On the ad parietal cortex. associations that may be refreshed by a process above.the VLFC.posterior parietal and motor activations sequenti network of areas involved in th arsal of the mo In summary,imaging studies have produced good evidence findings implicate the additional bilateral activation of DLFC for material-specific in manipulation (e.g.updating of the particular letters being udy Smit y to sus studies ver ther ted bilat DLFC/AFC activations in both a verbal is little imaging evide e for ventral-dorsal object -spatia no task in human vation i obiect inf left-lat tudy owen Sn6 nd c and that for the maintenance of spatial information.Th back tasks (Owen).Although differences betv een gion mos consistently the spatial and object memory-relate we as th p infor for the object task the coordinates of the DLFC is also sometime peaks of the bilateral DLFC/AFC activations for the two et al.,1996a:Belger et al..1998) SKS within a pr ght al int Manipulation in working memory ses in vlEC but that maninulation p he Manipulation of the contents of WM involves n array of nmon to visual- tial and visual-obiect WM.These two processe that may b unde the headin que the M sed and a huse nge of diffe onmateral-specine en tasks have been examined.Without attempting a precis FC lateralization of different executive processes. we Cohen and colleagues attempted to dissociate maintenanc on broad have be 'dual'and planning tasks.We emphasize that these term (Cohen et al 1997)Brain regions involved in transien are descriptive of the type of task employed and are not such ceiving stimu meant to imply different sets of executive processe isual and m N-back tasks such as maintenance,were predicted to show an uask that combine and manipulation isthe These regions inclu ask (Fig tas requires the right DLF Region whe ever the current stimulus matches the stimulusp mantain.were predicted to show an inter on between load tions back in the quence.For n 0.this task requires t an I time (i.e.greater transient effects at higher loads).The maintenance of the last n stimuli (in order)and updating of only lateral prefront region to shov this pattem was
Frontal lobes and human memory 855 eye fields, the pulvinar nucleus or superior colliculus are not these stimuli each time a new stimulus occurs (for n 0 the typically observed in neuroimaging studies of visuospatial task is simply to respond whenever a prespecified target WM. Another possibility, that rehearsal of visuospatial occurs, thus no updating is required). The value of n is often information involves an internal attentional mechanism, is viewed as proportional to the ‘working memory load’—the consistent with neuroimaging studies of spatial attention, total demand placed on the maintenance and/or manipulation which activates similar areas of the right superior parietal processes. cortex independently of eye movement (Corbetta et al., 1993; Braver and colleagues varied the verbal WM load by Coull and Nobre, 1998). A tentative hypothesis is that increasing n from 0 to 3 in a letter version of the n-back visuospatial information is stored as abstract or object visual task (Braver et al., 1997). Areas in which activity was a representations in the occipital cortex and inferior temporal linearly increasing function of load included DLFC, VLFC cortex, respectively (perhaps corresponding to visual caches). and the parietal cortex, bilaterally in each case, as well as a The (egocentric) spatial organization of the stimuli may be number of left motor, premotor and supplementary motor represented by associations between these areas and the right areas. On the basis of the maintenance studies reviewed parietal cortex, associations that may be refreshed by a process above, the VLFC, posterior parietal and motor activations of sequential, selective attention (perhaps corresponding to are likely to reflect the network of areas involved in the an inner scribe) that engages the right superior parietal cortex, maintenance of verbal information (e.g. the storage and right premotor cortex and right FC. rehearsal of the most recent n letters). If this is so, then these In summary, imaging studies have produced good evidence findings implicate the additional bilateral activation of DLFC for material-specific stores in posterior brain regions and in manipulation (e.g. updating of the particular letters being some evidence for a left–right lateralization of FC regions for maintained). the rehearsal of verbal and spatial information, respectively. In another study, Smith and colleagues (Smith et al., 1996) Contrary to suggestions from primate studies, however, there reported bilateral DLFC/AFC activations in both a verbal is little imaging evidence for ventral–dorsal object–spatial and spatial 3-back task, though there is a tendency for greater distinction in non-verbal maintenance tasks in humans. left DLFC activation in the former and greater right DLFC Rather, FC activation associated with the maintenance of activation in the latter (Smith and Jonides, 1997). In a similar object information appears to be more left-lateralized relative study, Owen and colleagues compared spatial and object 2- to that for the maintenance of spatial information. The back tasks (Owen et al., 1998). Although differences between FC region most consistently associated with the simple the spatial and object memory-related activations were maintenance of verbal material is the left VLFC. The VLFC observed in posterior regions, such as the posterior parietal is often associated with the maintenance of spatial and object cortex for the spatial task and the middle and anterior information (on the right for spatial information), though temporal cortex for the object task, the coordinates of the DLFC is also sometimes activated in these cases (e.g. Baker peaks of the bilateral DLFC/AFC activations for the two et al., 1996a; Belger et al., 1998). tasks were within 2 mm of each other. These data suggest that manipulation processes in DLFC are left–right lateralized for verbal versus spatial information, as for maintenance Manipulation in working memory processes in VLFC, but that manipulation processes may be Manipulation of the contents of WM involves an array of common to visual–spatial and visual–object WM. These two processes that may be loosely grouped under the heading studies again question the specific dorsal–ventral spatial– of executive processes. Many different types of executive object FC dissociation suggested by Goldman-Rakic, processes have been proposed and a huge range of different though support a material-specific left–right verbal–spatial tasks have been examined. Without attempting a precise FC lateralization. definition of different executive processes, we concentrate Cohen and colleagues attempted to dissociate maintenance below on broad categories of manipulation task that have been and manipulation in an n-back task by using event-related used in neuroimaging: ‘n-back’, ‘reordering’, ‘generation’, fMRI to measure activity at four intervals after each trial ‘dual’ and ‘planning’ tasks. We emphasize that these terms (Cohen et al., 1997). Brain regions involved in transient are descriptive of the type of task employed and are not processes, such as perceiving stimuli and producing meant to imply different sets of executive processes. responses, were predicted to show an effect of time but no effect of load (n). As expected, these regions included the visual and motor cortices. Regions involved in sustained N-back tasks processes, such as maintenance, were predicted to show an A task that combines maintenance and manipulation is the effect of load but not time. These regions included bilateral N-back task (Fig. 2D). This task requires the monitoring of VLFC and right DLFC. Regions associated with transient a continuous sequence of stimuli; a positive reponse occurs manipulation processes, such as updating the n items to whenever the current stimulus matches the stimulus n posi- maintain, were predicted to show an interaction between load tions back in the sequence. For n 0, this task requires both and time (i.e. greater transient effects at higher loads). The maintenance of the last n stimuli (in order) and updating of only lateral prefrontal region to show this pattern was left
856 P.C.Fletcher and R.N.A.Hensor VLFC.Though this was not the FC region that might have to detect which digit was omitted,the same bilateral DLFC DLCthisd on the bas of the above ee th activation was observed (Petrides er al..1993b) re (Baddeley as well as condition. forming to any rule or pattern.Tasks like these involve no only internal monitoring of previous responses (as in self but also The use of eventa clated fMRI to distinguish transient and random ke sustained effects in wM tasks is clearly an important method pressing was compared with reactive.stimulus-driven key ological advance.and one that is likely to prove valuable in pressing(Frith et al. 1991).Jahanshahi and colleagues apart pen on anc ma on when ra technique to isolate brain regionsre nsive during the s at higher of WM trials (see also (ahanshahi e) (D'Eposi whe for ex with a sea of fiv indices of random s or the ation rate orting the letters.followed by either a 'forward or an'alphabetize instruction.After a d ented tha are required by,but not related to,random generation and h as ih gen ney.a com osition denoted by the ce of fiv stimuli from much lar sets.The ve task ters were maintained in the (original)forward order (in requires generation without repetition of.for example.as Oin th the s).or five lette ge alpha of five as the latter additional manipulation GpTcntoncWtaceiestoaidgeaernlion(eg ordering).Both VLFC resp sive durin en a subject is require to gene rate as many animals a del D sho may by the consistent with the ecific FC modelo left di fC activation when letter fluency was compared with Petride these word repetition () are invo CO rable only DL and the evi Ge Dual asks task m tasks simulta rated withoutr tition one at a tim y 1986)mo e This task has been explored in neuropsychological (Petride between information appropriate for one or other task.patient and Milner.1982 and neur maging studies Petrides with frontal lesions may be disproportiona es er a 199 ersus singl Dowell er a durins contro task in which participants res nded WM.D'Esposito and collea exte nally produced stim h participants performed two brain activit tasks concurrently with ab was ting the atial totatio task and a semantic iude nt task roduced lateralization of manipulation ses in visuos patial wm ignificant activation of DLFC when performed alone:only s),DLFC activation as bi they were combir activa of did ed was significant bilate not dep activation w when an extemally ordered condition was tested in which omhined hecause a second exneriment in which perform participants listened to a random sequence of digits in orde ance of the rotation task was impaired by decreasing the
856 P. C. Fletcher and R. N. A. Henson VLFC. Though this was not the FC region that might have to detect which digit was omitted, the same bilateral DLFC been expected on the basis of the above studies (i.e. the activation was observed (Petrides et al., 1993b). DLFC), this experiment illustrates the opportunity afforded A related task is random number generation (Baddeley, by event-related studies to dissociate FC processes by time 1966), in which numbers must be generated without conas well as condition. forming to any rule or pattern. Tasks like these involve not only internal monitoring of previous responses (as in selfordering tasks), but also inhibition of prepotent responses Reordering tasks and well-learned routines. Frith and colleagues reported The use of event-related fMRI to distinguish transient and bilateral DLFC activations when generative, random key sustained effects in WM tasks is clearly an important method- pressing was compared with reactive, stimulus-driven key ological advance, and one that is likely to prove valuable in pressing (Frith et al., 1991). Jahanshahi and colleagues teasing apart perception and maintenance, and maintenance observed left DLFC activation when random number generaand manipulation. D’Esposito and colleagues have used this tion was compared with counting, and this activity was technique to isolate brain regions responsive during the negatively related to indices of randomness at higher generapresentation, delay and probe phases of WM trials (see also tion rates (Jahanshahi et al., 2000). Interestingly, VLFC Courtney et al., 1997). D’Esposito and colleagues (D’Esposito activation was also seen when random number generation et al., 1999) and Postle and colleagues (Postle et al., 1999), was compared with counting, but did not correlate with for example, presented subjects with a sequence of five indices of randomness or the generation rate, supporting the letters, followed by either a ‘forward’ or an ‘alphabetize’ proposal that this region is involved in maintenance processes instruction. After a delay of 8 s, a probe was presented that that are required by, but not related to, random generation. consisted of a letter and a digit (Fig. 2E). The subject’s task Other generation tasks, such as verbal fluency, a common was to indicate whether the probe letter would appear in the clinical test of frontal lobe damage, involve the selection of position denoted by the probe digit if the sequence of five stimuli from much larger sets. The verbal fluency task letters were maintained in the (original) forward order (in requires generation without repetition of, for example, as the ‘forward’ trials), or if the five letters were rearranged many animal names (category fluency) or words beginning into alphabetical order (in the ‘alphabetize’ trials). The former with a specified letter (letter fluency) as possible in a short trials require only the maintenance of five letters in order, period of time. This task involves not only monitoring but whereas the latter trials require additional manipulation (i.e. also the development of new strategies to aid generation (e.g. reordering). Both VLFC and DLFC were responsive during when a subject is required to generate as many animals as the delay period, but DLFC showed a greater response they can, they may begin by thinking of pets, then safari during the alphabetize trials (bilaterally in all cases). Though animals, etc.). The PET study of Frith and colleagues found broadly consistent with the process-specific FC model of left DLFC activation when letter fluency was compared with Petrides and colleagues, these studies suggest a nested word repetition (Frith et al., 1991). organization in which both VLFC and DLFC are involved Considerable evidence thus exists for a role of DLFC, on in maintenance, but only DLFC is additionally involved in the left for verbal and the right for visuospatial information, manipulation. in the manipulation processes necessary for generation tasks. Generation tasks Dual tasks In the self-ordering task mentioned earlier, stimuli must be Performing two tasks simultaneously makes demands on generated without repetition, one at a time, from a finite set. WM (Baddeley, 1986), most probably reflecting the switching This task has been explored in neuropsychological (Petrides between information appropriate for one or other task. Patients and Milner, 1982) and neuroimaging studies. Petrides and with frontal lesions may be disproportionately impaired in colleagues (Petrides et al., 1993a, b) compared brain activity dual-task versus single-task performance (McDowell et al., during the performance of a self-ordering task with activity 1997), again suggesting a frontal role in these aspects of during a control task in which participants responded to WM. D’Esposito and colleagues compared brain activity externally produced stimuli, without the requirement to order when participants performed two tasks concurrently with their own responses. When abstract figures were used, the the brain activity when each task was performed alone self-ordering task produced greater activation in right DLFC, (D’Esposito et al., 1995). Neither of the two tasks, a as predicted (Petrides et al., 1993a), supporting the right spatial rotation task and a semantic judgement task, produced lateralization of manipulation processes in visuospatial WM. significant activation of DLFC when performed alone; only With verbal stimuli (digits), DLFC activation was bilateral when they were combined was significant bilateral activation (Petrides et al., 1993b). This FC activation did not depend of this area observed. This activation was unlikely to be due solely on the self-generated nature of the ordering task: simply to the impaired performance of both tasks when when an externally ordered condition was tested in which combined, because a second experiment in which performparticipants listened to a random sequence of digits in order ance of the rotation task was impaired by decreasing the
Frontal lobes and human memory 857 bilateral id while when a stimulus was of lower luminance or pitch than the achieving subgoals.Imp y the previous stimulus.Klingberg and colleagues found no cortical AFC activation was selective to the branching condition,and rea that was a al in the dua-task in compa ble control ng. Card-Sorting task was actually diminished when combined with an auditory verbal shadowing task (Goldberg and et al..1998).participants watched a sequence of words and and Plc che kept a rnning count of the number of w that were name was diminished when the task was combined with a visuo condition.activations were seen in both right DLFC and motor secondary task (Fletcher).One possible right AFC.Like the branching task.this task might also be explan tha or even when evaluation)is pe activating DLFC.This might leave less scope for additional In a recent PET study,Burgess and colleagues observed sa set of different prospective wh the performance of b k (Burges 【asks aga1 whil differ ent task Thus AFC activation may reflect a third of dual-tasking.see Adcock et al..2000:Bunge et al.2000) level of executive control,beyond the manipulation in DLFC and m intenance Though this level Planning tasks ing.it is likely to nt of yday life Shallice introduced the Tower of London task in order to (Burgess er al.2000b).such as when we are inter test planning deficits in patients with frontal lesions(Shallice with a question while performing a complex task like reading. 98 must ge a set constraints on legal movements of the balls.this task re Other working memory tasks er te Yet more complex problem-solving tasks have been g02 Sta aedwith this gu AFC a simple visual-motor control,as well as several regions in DLFC (for a review,see Christoff and Gabrieli.2000).The the right premotor and parietal cortices that may be a compon ent processes of such complex tasks remain even less nainte ance (Owen er al. understood,however,and we do not discuss them requirins no movement (particinants were shown an initial state and a goal state and simply indicated the minimum number of moves from the initial to the goal state)(Bake Summary 199 s)The that subtrac FC suserve This evidence is summarized right AFC.These s are at least suggestive of a (perhap n Table 1.VLFC,for example,is more often activated FCinmepulhtion.cvenifmaipulai durin吗as srequiring maintenance anc DLF The study by Baker and collea es (1996h)is also or the few WM studies we have considered thus far,apart from that of Goldman-Rakic (Goldman-Rakic.1987).Nonetheless. 1996:0we here also appears to be a lateralization of FC processes hough the of London task.which includes up and maintainin ons of verbal and spatial tasks suggest that left multiple subgoals at the same time as making (or imagining movements between states.A more recent study showed an
Frontal lobes and human memory 857 interval between stimuli did not reveal any significant increase association between bilateral AFC activation and a in DLFC activity. However, in another dual-task study, using ‘branching’ task (Koechlin et al., 1999). This task also a visual and an auditory task in which participants indicated required the participant to maintain an overall goal while when a stimulus was of lower luminance or pitch than the concurrently setting and achieving subgoals. Importantly, the previous stimulus, Klingberg and colleagues found no cortical AFC activation was selective to the branching condition, and area that was activated specifically in the dual-task condition was not seen in comparable control conditions that required (Klingberg et al., 1998). Moreover, Goldberg and Berman either switching attention between goals (dual-tasking, which found that the DLFC activation associated with the Wisconsin activated right DLFC instead) or simply maintaining a single Card-Sorting task was actually diminished when combined goal. In another WM study that activated AFC (MacLeod with an auditory verbal shadowing task (Goldberg and et al., 1998), participants watched a sequence of words and Berman, 1998), and Fletcher and colleagues found that the kept a running count of the number of words that were names DLFC activation associated with elaborative verbal encoding of dangerous animals. Relative to a passive word-viewing was diminished when the task was combined with a visuo- condition, activations were seen in both right DLFC and motor secondary task (Fletcher et al., 1998b). One possible right AFC. Like the branching task, this task might also be explanation for these results is that one or both tasks, viewed as entailing the maintenance and periodic updating unlike the tasks used by D’Esposito and colleagues, included of one type of goal information while a demanding task manipulation requirements even when performed alone, (semantic evaluation) is performed concurrently. activating DLFC. This might leave less scope for additional In a recent PET study, Burgess and colleagues observed DLFC activation when the tasks are combined, or even a bilateral AFC activation across a set of different prospective decrease in DLFC activation when the performance of both memory tasks (Burgess et al., 2000a). These tasks again tasks suffers under dual-task conditions (for arguments against required delayed realization of an intention while performing the association of specific regions with the executive demands a different task. Thus AFC activation may reflect a third of dual-tasking, see Adcock et al., 2000; Bunge et al., 2000). level of executive control, beyond the manipulation in DLFC and maintenance in VLFC. Though this level of executive control is difficult to isolate and control in the laboratory Planning tasks setting, it is likely to be a vital component of everyday life Shallice introduced the Tower of London task in order to (Burgess et al., 2000b), such as when we are interrupted test planning deficits in patients with frontal lesions (Shallice, with a question while performing a complex task like reading. 1982). Participants in this task must rearrange a set of balls in order to match a specified goal state. Because of the constraints on legal movements of the balls, this task requires Other working memory tasks advance planning of a number of separate moves in order to Yet more complex problem-solving tasks have been attain the goal state, often via various subgoals, in the investigated with functional imaging, such as Wisconsin minimum number of moves. Owen and colleagues found Card-Sorting, Raven’s matrices, and inductive reasoning. activation of left DLFC associated with this task relative to These tasks have also tended to activate AFC as well as a simple visual–motor control, as well as several regions in DLFC (for a review, see Christoff and Gabrieli, 2000). The the right premotor and parietal cortices that may be associated component processes of such complex tasks remain even less with visuospatial maintenance (Owen et al., 1996). Baker well understood, however, and we do not discuss them and colleagues used a version of the Tower of London task further here. requiring no movement (participants were shown an initial state and a goal state and simply indicated the minimum number of moves from the initial to the goal state) (Baker Summary et al., 1996b). They found that subtraction of easy (two or Functional imaging of human WM has provided considerable three moves) from difficult (solutions involving four or five evidence that broad anatomical divisions within the lateral moves) conditions revealed activation in bilateral DLFC and FC subserve different processes. This evidence is summarized right AFC. These studies are at least suggestive of a (perhaps in Table 1. VLFC, for example, is more often activated bilateral) role of DLFC in manipulation, even if manipulation during tasks requiring maintenance and DLFC is more often was not completely dissociated from maintenance in this task. activated during tasks requiring manipulation. This is more The study by Baker and colleagues (1996b) is also one of consistent with the view of Petrides (Petrides, 1994) than with the few WM studies we have considered thus far, apart from that of Goldman-Rakic (Goldman-Rakic, 1987). Nonetheless, some n-back tasks with large n (Smith et al., 1996; Owen there also appears to be a lateralization of FC processes et al., 1998), in which AFC was activated. This activation is according to the type of material. Though the FC activations perhaps related to the complex planning required in the Tower are often bilateral (relative to baseline tasks), direct of London task, which includes setting up and maintaining comparisons of verbal and spatial tasks suggest that left multiple subgoals at the same time as making (or imagining) VLFC is primarily concerned with the maintenance of verbal movements between states. A more recent study showed an information and right VLFC with the maintenance of spatial
P.C.Fletcher and R.N.A.Henson Table 1 Working memory studies VLFC DLFC AFC Left Right Left Right Left Right Sn th et al. verba-p ds(199 h er af.(19 Be er et al (1998 ct-spatial Sternberg ne spa nngs Baker et al (996a Spatialob Br vera-back Spatial 3-back ,1997 ng m al. erbal n-back Frith er al.(1991) a2000 Random ke -tasking n er at.(19 et al.(19 96a Ko chlin etal.(1999 B ss et al (2000a) Semantic monitoring son is descriptive ony
858 P. C. Fletcher and R. N. A. Henson Table 1 Working memory studies VLFC DLFC AFC Left Right Left Right Left Right Awh et al. (1996) Verbal Sternberg Paulesu et al. (1993) Verbal Sternberg Jonides et al. (1993) Spatial Sternberg Smith et al. (1996) Spatial–verbal Sternberg Verbal–spatial Sternberg Smith and Jonides (1994) Object Sternberg Smith et al. (1995b) Spatial–object Sternberg Belger et al. (1998) Spatial–object Sternberg Object–spatial Sternberg McCarthy et al. (1996) Spatial–object running span Object–spatial running span Baker et al. (1996a) Spatial–object delayed Object–spatial delayed Braver et al. (1997) Increasing n in verbal n-back Smith et al. (1996) Verbal 3-back Spatial 3-back Owen et al. (1998) Spatial 2-back Object 2-back Cohen et al. (1997) Increasing n in verbal n-back D’Esposito et al. (1999) Alphabetization Petrides et al. (1993a) Spatial reordering Petrides et al. (1993b) Verbal reordering Frith et al. (1991) Random key-pressing Letter fluency Jahanshahi et al. (2000) Random key-pressing D’Esposito et al. (1995) Dual- versus single-tasking Owen et al. (1996) Planning versus difficult control Baker et al. (1996a) Planning versus control Koechlin et al. (1999) Branching versus dual-tasking Burgess et al. (2000a) Prospective memory MacLeod et al. (1998) Semantic monitoring Significant activation detected; no significant activation detected. Note that the label for each comparison is descriptive only; for more details see text. Studies are ordered according to their order of appearance in the text.