Until fairly recently, retrieval was an uncharted terrain in the neurobiology of memory. This was particularly striking when compared with the rich contribution of experimental psychology and modelling to the phenomenology and theory of retrieval (Semon 1904; Shiffrin and Artkinson 1969;Anderson 1983;Tulving 1983). This situation was also in sharp contrast to the signal role of retrieval in behavioural "assays of memory, for, at the end of the test, even if the intention is to study "acquisition, "consolidation, or retention of memory, it is retrieval that is tested. Hence many brain scientists study retrieval without even being aware of it. In a way, they are in the same position as Mr Jourdain, Moliere's bourgeois gentleman (Molière 1670), who suddenly realized ("insight) that he was speaking prose for 40 years without ever knowing it.
Definition 1 is molar, and pertains to both the phenomenon and the process of retrieval. Definitions 2 and 3 refer to the subprocesses and the brain state, respectively, that underlie 'retrieval' as defined in 1. Definition 3 includes also those situations in which the attempt to retrieve a particular item fails, or is only partially successful (e.g. Hart 1965; Brown and McNeill 1966).
The following points highlight selected attributes of retrieval:
1. At the behavioural "level, a memory unretrieved is undetected.1 It is actually possible to go a step further and claim that there is no such thing as a behaviourally-meaningful "engram in a nonre-trieved state (Tulving 1991; "persistence).2
2. Similarly to other memory "phases, retrieval refers to a heterogeneous group of processes that share a function. The computational theories, "algorithms and circuits that subserve retrieval of simple "reflexive memories are clearly different from those that subserve the actualization of abstract thoughts. Some operational principles and elementary building blocks of the molecular and cellular hardware may, however, be universal.
3. Success in retrieval depends on the availability of appropriate "cues. These cues are either external or self-generated. Cues are instrumental in retrieval even if unidentified by the "subject or the experimenter. Cues that are part of the response (e.g. a digit in a multidigit number) prompt gradual reconstruction of the memory; the product is termed 'redintegrative memory (from 'reintegra-tive'; Horowitz and Prysulak 1969).
4. Retrieval is more effective when attempted in the presence of cues that were present in acquisition ("context, "state-dependent memory). This idea is reflected in the 'encoding specificity principle', which refers to the overall relations between acquisition and retrieval: retrieval occurs if and only if properties of the trace are sufficiently similar to the properties of the information available in retrieval (Tulving 1983). In other words, for the "system to retrieve, it must resonate with the input. A related idea is that retrieval is more effective when the subject processes the information in retrieval in the same way that it was processed in acquisition; this is termed 'transfer appropriate processing' (Morris et al. 1977). For example, in verbal tests, semantic processing of verbal material in learning favours success in retrieval using semantic but not phonological processing, and vice versa. Whereas 'encoding specificity' emphasizes the information encoded, 'transfer appropriate processing' emphasizes the processing of that information by the subject. 5. Retrieval is not merely a passive readout of information, it is also an experience; therefore, once retrieved, the engram is unlikely to remain exactly the same. This is evident from studies at the behavioural, brain system, cellular, and molecular levels (Bartlett 1932; Schacter et al. 1998; Sara 2000; "false memory; see also below).
How does retrieval work in the nervous system? A simple case is illustrated by "classical conditioning of the withdrawal reflex in *Aplysia. A simplified "model of learning in this system involves use-dependent facilitation of sensory-to-motor "synapses. This is expressed as the enhanced release of "neurotransmitter in response to the sensory stimulus. The facilitation is embodied in chemical changes in the presynaptic terminal. Retrieval is the readout of the new state of the synapse by the action potentials that encode the conditioned stimulus (CS) in the sensory neurons. The cue for retrieval is the CS. The neurons that retrieve are those that learn and retain at least part of the trace.3 The conditions at acquisition influence retrieval, because whether the animal was already "habituated or naive determines the nature and extent of the subsequent cellular modifications in the sensory synapse, hence the synaptic state encountered by the action potentials that encode the CS in retrieval (Byrne and Kan del 1996). And, finally, retrieval, being an experience in the same neuron that retains the memory, might disrupt "homeostasis and induce new "plastic changes.
Most of what we now know about neuronal mechanisms of retrieval, however, owes to systems that are far more complex than Aplysia. "Functional neuroimag-ing, combined with smart design of behavioural tests, has made it possible to identify neuronal players in the retrieval of explicit ("declarative) memory in humans. Before the introduction of functional neuro-imaging, data on the involvement of specific brain structures in retrieval in humans were based predominantly on clinical cases (Shallice 1988; Schacter et al. 1996b). As brain lesions, once formed, are permanent, and could affect the formation or 'storage' of the memory trace,whereas retrieval is restricted to a brief time window of memory expression, conclusions about retrieval that are based solely on the effect of static pathology are problematic to start with. In contrast, functional neuroimaging could dissociate the neuronal events of retrieval from those of earlier memory phases (e.g. Buckner et al. 1995; Nyberg et al. 1996; Buckner and Koustaal 1998; Fletcher et al. 1998; Wagner et al. 1998a,b; Schacter and Wagner 1999; Lepage et al. 2000; Rugg and Wilding 2000). This is currently one of the most dynamic fields in memory research, and the picture keeps changing. A tentative model of explicit retrieval in the mammalian brain can nevertheless be portrayed.
It is methodologically convenient, and probably correct, to describe the brain system that subserves retrieval as composed of two main types of components: item-specific and item-invariant. The item-specific component subserves the actual recovery of the particular memory item. This process is referred to as 'ecphor/ (Greek for 'to be made known'). The term was coined by Semon (1904), the same person who brought us 'engram', and later retrieved by Tulving (1983). Ecphory involves circuits that 'store' ("metaphor) memories (Markowitsch 1995, Schacter et al. 1996a; Schacter and Wagner 1999; Eldridge et al. 2000; Nyberg et al. 2000; Wheeler et al. 2000). In the case of declarative memory, these circuits are distributed over areas in the "cerebral neocortex, and connect with paleocortical and subcortical structures ("amnesia, "hippocampus, "limbic system).4 In addition, there is an item-invariant system, which searches for items in memory, allocates the mental resources, controls the process, and verifies the outcome. This system is said to put the brain into a retrieval 'state', 'mode', or 'set'.5
Where is this retrieval-mode system located? It has been demonstrated by a number of groups that prefrontal cortex is differentially activated in retrieval (Buckner et al. 1995; Schacter et al, 1996a; Rugg et al. 1996; Buckner and Koustaal 1998; Fletcher et al. 1998; Wagner et al. 1998; Lepage et al. 2000). The prefrontal cortex subserves multiple executive functions in the brain ("planning, "working memory), and is therefore a likely candidate for setting the stage for ecphory. There is hemispheric asymmetry in the retrieval-related activity of prefrontal cortex. This finding has led to a model, the hemispheric encoding/retrieval asymmetry (HERA), which proposes that the left prefrontal cortex is differentially involved in retrieval of semantic information and in encoding novel aspects of the retrieved information into "episodic memory, whereas the right pre-frontal cortex is involved in retrieval of episodic memory (Nyberg et al. 1996). The possibility was further raised that the hemispheric asymmetry observed in retrieval in the prefrontal cortex is contributed by the memory retrieval mode, but not by the ecphory (Lepage et al. 2000). Prefrontal areas probably contribute differentially to retrieval mode functions, as well as to the "metamemory processes that monitor and verify retrieval; but again, the picture is incomplete (Schacter and Wagner 1999; Rugg and Wilding 2000).
The main picture of the layout and function of retrieval system(s) in the human brain, including the executive role played by the prefrontal cortex, are supported by lesion studies and cellular physiology in the "monkey (Hasegawa et al. 1998; Tomita et al. 1999). The investigation of the mechanisms of retrieval at more reduced levels of analysis depends much on the use of additional species. These include rodents (Moser and Moser 1998a; Riedel et al. 1999; Maren and Holt 2000; Nader et al. 2000; Sara 2000; Berman and Dudai 2001), "Drosophila (Dubnau et al. 2001), and Aplysia (see above).
An intriguing proposal, hinted already above, is that retrieval, being an experience, is followed by a new "phase of "consolidation (Spear and Mueller 1984; Sara 2000). Even more provocative is the suggestion that in some systems and conditions the original trace, or at least a part of it, could become markedly labile for a while, after its retrieval (Nader et al. 2000). The 'reconsolidation' hypothesis, if validated, could guide the development of new drugs and behavioural methods to alter specific items in memory. This might be used to erase unwanted memories (Dudai 2000), or enhance desired ones. Until such treatments are identified, down-to-earth tricks could be tried to improve retrieval. Well, on the one hand, retrieval appears to be enhanced by glucose (Manning et al. 1998; "nutrients); on the other, stress and corticosteroids impair retrieval (De Quervain et al. 1998; "lotus). A good dessert in relaxed company might hence provide a reasonable alternative to medical intervention.
Selected associations: Performance, Persistence, Phase, Recall, Recognition
1Lewis (1979) proposed to call unretrieved memories 'inactive', and retrieved as well as short-term memories 'active' ('taxonomy). This terminology, however, did not catch on.
2Retrieval is not obligatory evidence for memory at 'reduced levels of analysis. Hence one could identify 'persistent memory-related 'plastic changes at the molecular, cellular, or circuit level in the absence of actualization of the internal representation. Still, for the engram to acquire its representational meaning, it must be retrieved.
3It is possible, though, that other parts of the nervous system of Aplysia, which are not involved in acquisition or consolidation of the information, evoke or control retrieval.
4The role of the hippocampus in retrieval of declarative memories is limited in time (Teng and Squire 1999; Haist et al. 2001; 'consolidation).
5For more on the general notion of 'set', see 'learning set.
Was this article helpful?