Frontal lobes and human memory 859 and dienc ove thus atter supporte ventral-dorsal obje tasks.FC lesions do not cause the same global amnesia that for thi temporal/die apping in the two species,contributed to by the effects of tasks such as r anguage evolution i humans.Indeed one problem with )and tasks with high levels of interference (Incisa experime for peop Della Rocchetta and Milner. 1993). FC activations during T therefore,likely to controlp to imagine for very abstract visual stimuli).One solutio than more automatic storage processes may b comb spatial tasks Most neuroimaging experiments on LTM consist of two of the stim the bilit is that hoth es:a study pha which multiple stimuli are pr pecific and omain-spe ecific specializations exist thin the human DLF but the ent resolution of imaging or recognized from among other stimuli.The majority of inguish them dies have used familiar words as stimuli o typ hetween storage and rehearsal of verbal and spatial d by cognitive psych logical models (e.g and ass 1974 memory that word was presented in a barticular context in whe cinital lobe whe en the regio that include VLF As me bov the cor and CAT) is likely to storag left- e retrieved from long-term emanti me nd this with diff ombined with contextua tion (consis rent slave syste studies have aso highlighted the shortcomings of this mode a ne epis c me executive.[This i retrieved tral executive has he from episodic memory.Given the long history of labora tudy heterogeneous set of executive ng an emory and th sts that DLFC regions ar Manipu focus here on neuroimaging studies of verhal emory.engage DLFC ne eas more complex pro methodo on entail maintaining the goals and products pe ing anoth ociating the and retr val stages of e given that it i difficult to attribute a patient's evidence for left dominance for verbal mate ial and righ nterograde memory del specifcally to either an encoding e for spa erial,th ough the evid ng ar s they may share a number of subp For ex ostle and D'Es sito 2000)1 Nonethele while we oth of semant the imaging date re likely to involve searche memory,firs evidence fo ome func me encoding episod pre of the commonalities and differences between the comp apt to retrieve a wordfrom result in a train of associative thought that will become the ubstrate of a further encoding episod Thus the encod by the Frontal function in long-term memory tasks esses involved. Non one goal of Neuropsychological studies of patients with focal brain functional naging researchers over the last few years has lesions have highlighted the importance of medial temporal been to isolate more specifically the cognitive processes that
Frontal lobes and human memory 859 information. The maintenance of object information is and diencephalic structures in human long-term declarative sometimes left-lateralized, though the overall pattern is less memory (Squire and Cohen, 1984). Functional neuroimaging clear. There is thus an apparent discrepancy between the studies of healthy subjects, however, have emphasized the findings from human and primate studies, because some of engagement of FC structures during the performance of LTM the latter have supported a ventral–dorsal object–spatial tasks. FC lesions do not cause the same global amnesia that distinction. There may be several reasons for this. One can result from medial temporal/diencephalic lesions, but possibility is a difference in the functional–anatomical they are associated with impairments in more complex mapping in the two species, contributed to by the effects of memory tasks, such as memory for temporal order (Janowsky language evolution in humans. Indeed, one problem with et al., 1989) and tasks with high levels of interference (Incisa human experiments is the potential for people to recode Della Rocchetta and Milner, 1993). FC activations during visuospatial stimuli verbally, effectively converting an object LTM tasks are, therefore, likely to reflect control processes task, for example, into a verbal one (though this is difficult that aid and optimize memory encoding and retrieval, rather to imagine for very abstract visual stimuli). One solution than more automatic storage processes. may be to combine imaging of visuospatial tasks with Most neuroimaging experiments on LTM consist of two concurrent articulatory suppression, to prevent verbal phases: a study phase, in which multiple stimuli are presented recoding of the stimuli. Another possibility is that both (with or without explicit instruction to remember the stimuli), process-specific and domain-specific specializations exist and a test phase, during which these stimuli must be recalled, within the human DLFC, but the current resolution of imaging or recognized from among other stimuli. The majority of techniques is unable to distinguish them. these studies have used familiar words as stimuli. These Imaging studies have also supported the dissociations studies allow a clear distinction between two types of LTM between storage and rehearsal of verbal and spatial (Tulving, 1983): semantic memory, the knowledge of the information proposed by cognitive psychological models (e.g. words’ meanings and associations, and episodic memory, the Baddeley and Hitch, 1974). The demands of storage engage memory that a word was presented in a particular context in posterior brain regions, including the parietal, temporal and the past (i.e. the study phase). Thus, when a word such as occipital lobes, whereas rehearsal engages a network of DOG is presented during the study phase, information about regions that include VLFC. As mentioned above, the cortical its meaning and close associates (e.g. CAT) is likely to loci of both storage and rehearsal are left-lateralized for be retrieved from long-term semantic memory, and this verbal information and right-lateralized for spatial information may be combined with contextual information information (consistent with different slave systems). Imaging in the encoding of a new episodic memory. If the word is studies have also highlighted the shortcomings of this model, presented again during a recognition test phase, information particularly regarding the central executive. [This is a problem about its prior occurrence in the study phase may be retrieved acknowledged by Baddeley (1996), who admitted that the from episodic memory. Given the long history of laboratory central executive has been a rag-bag containing a potentially study of verbal learning and memory and the high level of heterogeneous set of executive functions.] The evidence experimental control afforded by such material (e.g. physical summarized above suggests that DLFC and AFC regions are form, frequency, imageability and semantic associations), we associated with executive control of WM. Manipulation focus here on neuroimaging studies of verbal episodic processes, operating on information already maintained in memory. memory, engage DLFC, whereas more complex processes A clear methodological advantage of functional that entail maintaining the goals and products of one task neuroimaging over neuropsychology is the possibility of while performing another, appear to engage AFC. These dissociating the encoding and retrieval stages of episodic higher-level processes may also be lateralized; there is some memory, given that it is difficult to attribute a patient’s evidence for left dominance for verbal material and right anterograde memory deficit specifically to either an encoding dominance for spatial material, though the evidence is less or a retrieval problem. Attempts to dissociate encoding and clear in this respect than for maintenance processes [perhaps retrieval by neuroimaging are rarely straightforward, however, because there have been fewer direct comparisons (see also as they may share a number of subprocesses. For example, Postle and D’Esposito, 2000)]. Nonetheless, while we regard both are likely to involve searches of semantic memory, first the imaging data as good evidence for some functional to produce a rich memory trace of the encoding episode, and specialization within FC, more precise definition of these later to generate cues that aid access to that trace. Furthermore, functions remains constrained by our limited understanding an attempt to retrieve a word from episodic memory may of the commonalities and differences between the component result in a train of associative thought that will become the cognitive processes involved in the range of different WM substrate of a further encoding episode. Thus the encoding– tasks that have been studied. retrieval distinction is driven more by the format of the typical episodic memory task than by consideration of the Frontal function in long-term memory tasks executive processes involved. Nonetheless, one goal of Neuropsychological studies of patients with focal brain functional imaging researchers over the last few years has lesions have highlighted the importance of medial temporal been to isolate more specifically the cognitive processes that
860 B.C.Fletcher and R.N.A.Hensor differentiate encoding from retrieval.The encoding-retrieval analyses afforded by the faster acquisition times of fMRI,it distinction,therefore,provides a useful means of organizing is possible to scan people during the study phase of a memor our review of previous neuroimaging research k,9 scanning. hoc,into those that were e remembered and those that Frontal function in long-term memory or an ding 'unsucce Man only sterior VLFC was higher du ng the r words that were subsequently remembered confidently than review,we define encoding as the process(es)that allo ntation of those that were forgotten (Wag quent exp of m 1998 nat th studyas the engagement of left FC.This functional necifically to successful encoding in a similar ever lated formed part of the influential hemispheric encoding-retrieval tudy using visuospatial rather than verbal material,Brew (HE A)generalizat howed hat right上 right than left Fc activatio with isodic (Tulving er al 1994a).Furthermore,the left lateralization uring encoding tasks is specific to the use of verbal mate We return to this que tion later evi sug h goal is to understand the es that c The evidence for FC activation in incidental encoding to this success.More generally.we know from behavioural hat manipulate the degre studies that ding de nd er ship the 'depth of pr sing'effect (Craik and Lockhart.1972) which visual imager sed.Indeed.the depth of reby tasks produ Pcesinganddege Ican be shown to are ger emph ve ser 199 an also nd on the nature of the elated e deserves close task ().Kapur eral,for example,compare original et o of len C g or non- 199 1994)They found the Thon son-Schill et al,1997).In brief,the main po ons of VLE C to be a ssociated with dee gard to the of left FC encoding mer as:(i) n a study of intentional din the maintenance (in semantic WM')of these attribute colleagues used a paired associate task in which subjects nd associates (Gabrieli et al.1998):(ii)the selection of were instructed to learn the pairings (e.g task-appropriat or associat from among thos er al.. activation was seen in an anterior gion of left VLFC.Thi asis of those attributes (Fletcher et al.1998a).We shall efer to these ositions ,“maintenanc h contribution of the iated with left VLEC aluating fou e that ortant for sful encodine the picture is complicated by the fact that they seem to form arly PET studie d a link bet hierarchy semantic information cannot be maintair on eft FC(predo VLFC)an gene th correlated brain activity during a study task with subsequer ffective organization of multinle items is unlikely to proceed retrieval performance.Using the event-related or trial-specific unless appropriate attributes have been selected.Therefore
860 P. C. Fletcher and R. N. A. Henson differentiate encoding from retrieval. The encoding–retrieval analyses afforded by the faster acquisition times of fMRI, it distinction, therefore, provides a useful means of organizing is possible to scan people during the study phase of a memory our review of previous neuroimaging research. task, measure their retrieval performance after scanning, and then use their memory performance to sort stimuli, post hoc, into those that were remembered and those that were forgotten. That is, each stimulus can be associated with either Frontal function in long-term memory a ‘successful’ or an ‘unsuccessful’ encoding event. Using encoding this approach, Wagner and colleagues showed that activity Many moment-to-moment experiences are forgotten; only in left posterior VLFC was higher during the presentation of some can later be remembered. For the purposes of this words that were subsequently remembered confidently than review, we define encoding as the process(es) that allow during presentation of those that were forgotten (Wagner subsequent explicit (conscious) retrieval of memories. Early et al., 1998c). Given that the study task remained constant, studies examining memory encoding showed evidence for this is more direct evidence that the left FC region is related the engagement of left FC. This functional lateralization specifically to successful encoding. In a similar event-related formed part of the influential hemispheric encoding–retrieval study using visuospatial rather than verbal material, Brewer asymmetry (HERA) generalization, which associates greater and colleagues showed that right FC activity is associated left than right FC activation with episodic encoding, and with encoding success (Brewer et al., 1998). This raises the greater right than left FC activation with episodic retrieval question of whether the left lateralization of FC activations (Tulving et al., 1994a). Furthermore, the left lateralization during encoding tasks is specific to the use of verbal material. during encoding is found whether or not subjects are aware We return to this question later. that their recall will be tested later. That is, left FC activation Assuming, as the evidence strongly suggests, that is found in both incidental (Kapur et al., 1994) and intentional successful encoding of verbal material engages left FC, the (Shallice et al., 1994) memory encoding. goal is to understand the component processes that contribute The evidence for FC activation in incidental encoding to this success. More generally, we know from behavioural comes from studies that manipulate the degree of semantic studies that successful encoding depends not just on semantic processing of verbal material. The relationship between processing, but also on, for example, the level of attention, semantic processing and memory encoding is referred to as the degree to which material is organized, and the extent to the ‘depth of processing’ effect (Craik and Lockhart, 1972), which visual imagery is used. Indeed, the depth of semantic whereby tasks producing better recall (‘deep encoding tasks’) processing and degree of divided attention can be shown to are generally those that emphasize the meaning of items have separable effects (Craik et al., 1996). The relationship rather than their surface features (although recall performance between left FC activations and these different encodingcan also depend on the nature of the subsequent retrieval related processes therefore deserves closer consideration. The task (Morris et al., 1977). Kapur et al., for example, compared original suggestion of Tulving and colleagues that left FC a deep encoding task (judging whether words referred to activation is associated with the retrieval of semantic living or non-living entities) with a shallow encoding task attributes of studied material (Tulving et al., 1994a) has (judging whether words contained the letter ‘a’) (Fig. 3A) been both expanded (Gabrieli et al., 1998) and re-evaluated (Kapur et al., 1994). They found the anterior and posterior (Thompson-Schill et al., 1997). In brief, the main positions regions of VLFC to be associated with deep encoding, despite with regard to the role of left FC at encoding can be the fact that subjects were unaware that their memory would summarized as: (i) the generation/retrieval of semantic be tested subsequently. attributes and associates of a word (Tulving et al., 1994a); In a study of intentional memory encoding, Shallice and (ii) the maintenance (in ‘semantic WM’) of these attributes colleagues used a paired associate task in which subjects and associates (Gabrieli et al., 1998); (iii) the selection of were instructed to learn the pairings of a category (e.g. task-appropriate attributes or associates from among those ‘poet’) with a specific exemplar (‘Owen’) (Fig. 3B) (Shallice associated with the word (Thompson-Schill et al., 1997); and et al., 1994). When contrasted with a passive listening task, (iv) the organization of multiple words or associates on the activation was seen in an anterior region of left VLFC. This basis of those attributes (Fletcher et al., 1998a). We shall activation was reduced when learning occurred in the presence refer to these positions as the ‘generation’, ‘maintenance’, of a distracting motor task. Moreover, this distraction ‘selection’ and ‘organization’ views of the contribution of produced an impairment in subsequent cued recall, suggesting the left FC to encoding. that the processes associated with left VLFC activation are Before evaluating these four processes, we observe that important for successful encoding. the picture is complicated by the fact that they seem to form Early PET studies, therefore, established a link between a hierarchy: semantic information cannot be maintained onleft FC (predominantly VLFC) and successful word encoding. line until it is first generated, and it cannot provide the Converging evidence comes from studies that have explicitly basis for selection without on-line maintenance. Furthermore, correlated brain activity during a study task with subsequent effective organization of multiple items is unlikely to proceed retrieval performance. Using the event-related or trial-specific unless appropriate attributes have been selected. Therefore