You have to begin to lose your memory, if only in bits and pieces, to realize that memory is what makes our lives. Life without memory is no life at all, just as an intelligence without the possibility of expression is not really an intelligence. Our memory is our coherence, our reason, our feeling, even our action. Without it, we are nothing. - —Luis Buñuel (Spanish filmmaker)
No doubt, the question of whether memory is important to human function is indisputable: it is essential to the human experience, allowing us to integrate our past into who we are today and to imagine what we might be in the future.1 How this happens, though, raises many questions, one of which is “What is memory?” In answering this specific question, however, historically there has been considerable debate. Much of this debate centers on the issue of whether memory is best understood as a single, unitary system that engages different processes (ie, modes of operation) as a function of the type of task (a processing approach) or whether it is made up of many systems (ie, different neural networks or neurological substrates) that can be distinguished based on specific criteria (a systems approach). While proponents exist on both sides of the debate, it has become evident that memory is composed of both multiple systems and multiple processes.2 Unfortunately, there is no clear resolution as to the exact number of systems/processes that make up memory,3 with a variety of contrasting models existing.4–7
In answering the question “What is memory?,” we have chosen to rely on an approach advocated by Schacter and Tulving,5 primarily because it is the predominant approach within cognitive psychology addressing the memory systems issue. To preview, we anchor our discussion on distinctions made between kinds of memory that involve consciousness (ie, declarative memory) versus the kinds of memory that express previous experiences through changes in task performance, without reference to or conscious recollection of the previous experience (ie, nondeclarative memory; Fig. 1). There is a vast literature (especially research conducted with amnesics) supporting dissociations between these two broad types of memory. What follows is an overview of these systems.
Declarative memory refers to a collection of memory systems that together are responsible for the acquisition, retention, and retrieval of information that can be consciously and intentionally recollected. This definition is what most people think of when the term memory is used. Traditionally,6 two major distinctions have been made within declarative memory: memory for events (episodic memory) and memory for facts (semantic memory).
Episodic Memory: General Characteristics
Episodic memory refers to memory for personally experienced events, 8 including information about the particular time or place in which these events occurred (eg, remembering your high school graduation or what you had for dinner last night; Table 1). Consequently, when these events are recollected, individuals often experience a mental reliving of the event (a form of conscious awareness termed autonoetic, or self-knowing9), including the emotions that were experienced during the original event. This type of recollective experience sets episodic memory apart from other memory systems and provides important information about the nature of episodic memory traces: the representations within this system are rich with details about experienced events, including the specific spatial/temporal contexts associated with them, as well information about the internal states of the person (eg, emotional, cognitive) when the event was encoded. This is an important characteristic of episodic memory as it affects the success with which information stored in this system can be retrieved. Generally, the extent to which information present during retrieval matches information incorporated into an episodic trace, then retrieval will be successful, a principle known as encoding specificity.10 Similarly, successful retrieval is also dependent upon the degree of overlap between encoding and retrieval processes, a concept known as transfer-appropriate processing.11 Both these principles emphasize how important context is to episodic memory: representations are more readily retrieved when the retrieval context matches what is present in the encoded representation.
Effects of Brain Injury and Aging on Episodic Memory Function
A tremendous amount of behavioral evidence has been accumulated to demonstrate the existence of episodic memory. Indeed, Tulving12 notes that there are “literally hundreds of studies that are concerned with ‘what,' ‘where,' and ‘when,'” so only a few selected case studies that examine various forms of episodic memory loss are presented here to illustrate episodic memory function. One particularly interesting set of cases is discussed by Vargha-Khadem et al,13 who describe anterograde amnesia (an inability to acquire new event information) in three individuals who suffered bilateral hippocampal damage as children. As a consequence of this damage, all three individuals have a severe loss of episodic memory for recently experienced events, but otherwise intact cognitive function. Specifically, all three patients (young adults at the time studied) report difficulty recollecting the location of objects that they have placed down, require frequent reminders of scheduled appointments and events, and report difficulty providing reliable accounts of their daily activities, phone or other personal conversations, television programs, stories, etc. These impairments are reported to be so severe that all of these individuals remain dependent on their parents and other caretakers. Clearly, these individuals demonstrate a loss of the “where” and “when” information that is characteristic of episodic memory. Despite this impairment, however, all three individuals are able to acquire factual knowledge (eg, “What is the capital of Italy?”) and are competent readers and writers, demonstrating spared knowledge of vocabulary/language, all aspects of semantic memory (discussed in the next section). Similar types of dissociations have been noted by Zec et al14 in a group of patients with closed head injuries (CHIs), with the locus of damage being limited to hippocampal and related medial temporal lobe structures. When compared with normal controls and a group of spinal cord–injured (SCI) patients, the CHI group displayed significantly poorer performance on measures of episodic memory. Specifically, the CHI group demonstrated severe impairment in learning and retention of episodic information over many years, deficits that correlated with the hippocampal and medial temporal lobe damage they had experienced as a result of their injuries.
While the studies just described illustrate the importance of the hippocampal and medial temporal lobe structures in episodic memory function, other brain regions that are critical to retrieval processes employed by episodic memory have been implicated as well. For example, Levine et al15 present findings from a patient with isolated retrograde amnesia following severe traumatic brain injury (TBI). Imaging of the patient revealed damage to the right frontal cortex and underlying white matter, including the uncinate fasciculus, a frontotemporal band of fibers hypothesized to mediate episodic events from one's personal past. While this individual was densely amnesic for events that occurred prior to the injury, he showed normal performance on recall and recognition measures of information acquired/experienced since the injury. Upon further examination, however, it was noted that the patient did not process postinjury events in the same way as normal controls. Specifically, he appeared to be relying on different retrieval strategies than normal controls to distinguish between events he had experienced and those he had not. He appeared to rely on familiarity processing rather than recollection of event details, highlighting the fact that episodic memory relies on at least two types of retrieval processes. Various theories exist to characterize the difference between these two processes (ie, familiarity versus recollection), but in general they are consistent in their description of familiarity processing as being relatively fast and automatic, but often lacking in perceptual/contextual detail, while recollection is described as an effortful, slow, controlled process in which perceptual/contextual details of stored memories are retrieved (see Yonelinas16 for a review of these processes). The patient described by Levine et al15 appears to have greater difficulty relying on this latter process in his retrieval of episodic information, suggesting that this episodic memory process may be supported by areas of the frontal lobe rather than by medial temporal lobe regions.
The above discussion illustrates how deleterious are the effects of select types of brain damage to episodic memory function, with damage to multiple brain regions (both medial temporal lobe as well as frontal lobe regions) affecting one's ability to accurately encode and retrieve episodic information. Episodic memory function is also impaired in normal aging, but in contrast to the severe deficits that are characteristic of the types of brain damage just described, the normal aging process produces less severe, but nonetheless notable, deficits. Indeed, it is this aspect of memory function that is most affected by aging. A large number of empirical investigations confirm what older adults often report anecdotally; ie, the ability to consciously recollect past events declines with advancing age (see Backman et al17 for a review). Many theoretical arguments have been proposed to address the issue of why this aspect of memory function appears to be so vulnerable to the effects of aging (see Luszcz and Bryan18 for an overview), with one in particular highlighting age-related breakdowns in the ability to integrate episodic context as underlying the observed age differences in episodic memory.
The underlying assumption of this model19 is that event representations are composed of multiple kinds of information that become integrated during encoding: the event's meaning, the time it occurred, where it took place, who was present, etc. To distinguish this model from other similar models, it is argued that the linking together of these events with their contexts requires associative memory processing (a process in which individual information units are bound together into a larger representation) to produce the integrated representation. The ability to remember an episode is dependent on retrieval of both these individual components, as well as how they are related to each other. Older adults are hypothesized to have an associative memory deficit, in which the ability to create the associations between event information units no longer operates efficiently, limiting their ability to adequately encode and later retrieve episodes. Support for this argument comes from data19,20 demonstrating that older adults are particularly impaired at remembering relationships among attributes of studied information (eg, the relationship between stimulus words and the type of font in which they were presented or face-name associations where specific names have to be tied to specific faces), even though they can remember specific units of information (eg, a name or a face by itself is recognized) as well as young adults. These observations illustrate how episodic memory binds together general event information with its specific context to create rich episodic traces. A failure in the ability to bind these two types of information will lead to impairments in the ability to recollect episodic memories.
Episodic Memory: Summary
The impairments and deficits in episodic memory function just described illustrate the important role that the episodic memory system has in cognition. It is a very personal memory system, tying together the various spatial, temporal, and other contextual features of our experiences to allow us to relive these events when they are recollected. As will be seen from the remaining discussion, this function of episodic memory makes it distinct from other memory systems, all of which have important roles in cognitive function, but do not maintain the direct connections to personal experience. Damage to the underlying brain structures that support this memory system, whether through traumatic brain injury, or to a lesser extent normal aging, disrupts one's ability to accurately recollect the events of one's past and encode new episodes.
Semantic Memory: General Characteristics
Semantic memory refers to a person's memory for facts, ie, their general world knowledge,8 including conceptual knowledge (eg, knowing that tigers, but not leopards, have stripes) and language (eg, vocabulary knowledge). This knowledge is argued to be abstract and lacks associations to a learning context (ie, it is “timeless”21). As such, there is no mental reliving of an event during semantic memory retrieval as there is in episodic memory retrieval. Rather, retrieval of information from semantic memory allows for efficient language production, comprehension, reasoning, and other higher order cognitive processes. This system is, then, distinct from episodic memory, with which it is often compared, as both comprise what is typically thought of as long-term memory (Table 1).
Effects of Brain Injury and Aging on Semantic Memory Function
Evidence supporting the dissociation between episodic and semantic memory is rich and compelling, coming from both experimental laboratory studies of healthy individuals across the life span,22 as well as work done with amnesics. In this latter category, a few case studies23,24 have demonstrated that significant impairments in semantic memory can exist, while leaving episodic memory (specifically, autobiographical memory) relatively spared. For example, Yasuda et al24 describe patient M.N., who following surgery to remove a brain tumor, demonstrated significant impairments in the ability to recollect well-known historical events, historical figures, popular proverbs, and other types of information typically ascribed to semantic memory. She was, however, able to freely recall events from her own past (eg, former boyfriends, school field trips). Similar semantic memory impairments have been noted in four patients with nonprogressive brain injury, two with severe head injury, and two with herpes simplex virus encephalitis.25 What is most important about these cases is that despite these different etiologies, the patients all exhibited similar damage to temporal neocortical areas, although the damage was much more diffuse in the encephalitis patients, affecting hippocampal structures as well. Notably, all four patients demonstrated poor performance on a semantic memory battery, but impaired episodic memory function was only present in the encephalitis patients, whose damage spread beyond temporal neocortical areas into frontal lobe regions. These findings not only illustrate dissociations between episodic and semantic memory, but also point to the role of the temporal neocortex specifically in supporting semantic memory function.
While the studies just described illustrate how specific types of localized damage can affect semantic memory function, diffuse brain injury appears to impair semantic memory processing efficiency, specifically affecting the ability to access related concepts. Such damage, often observed in CHI patients, is argued to impair the ability to intentionally access semantic knowledge, making it appear as if this group is experiencing semantic memory loss, when, in fact, their semantic memory system is intact.26 Support for this latter point comes from studies examining semantic priming, in which normal individuals produce faster or more accurate decisions about a target item (eg, Is nurse a word?) when a related word (eg, doctor) precedes it, as a result of spreading activation across an associative network. If there is a true loss of semantic memory, such priming effects are not likely to be observed. Haut et al27 report data from CHI patients who, while they showed slower responding to target items relative to normal controls, had a magnitude of semantic priming effects that was indistinguishable from that of normal controls. These findings suggest that semantic memory organization was intact in their patients, despite impairments in the ability to deliberately access it to make category judgments. Importantly, these data demonstrate that semantic memory processing relies on both automatic processes that allow access to semantic knowledge, as well as effortful, controlled processes to intentionally use this knowledge. This latter process seems to be particularly susceptible to head injury, while the underlying structure of semantic memory is more resistant to damage.
The examples just described from the brain injury literature highlight the dissociation between episodic and semantic memory function, as well as underscore differences in the types of processes that make utilization of semantic knowledge possible. Similar dissociations have been reported in older adults, with episodic memory consistently showing significant age-related declines (as described earlier), but semantic memory showing little appreciable change across the life span.28 It has been consistently noted across the literature that the organization and structure of semantic memory remain stable, even well into old age.29 Indeed, some aspects of semantic memory (eg, vocabulary knowledge) may even increase with advancing age.30 One argument as to why semantic memory declines are generally not present in old age, while episodic memory declines are so noticeable (despite the fact that both systems are part of a larger declarative memory system) suggests that the deficits observed in episodic memory are linked to deficits in processing context-dependent memories (as was described earlier). Because semantic memory is not a context-dependent system, those processing deficits are likely to have little impact on semantic system function, making semantic memory a relatively stable and well-functioning aspect of memory in old age.
Semantic Memory: Summary
While semantic and episodic memory are both considered long-term memory systems in which information can be consciously accessed, semantic memory serves a very different function than does episodic memory. It is not an experientially based memory system like episodic memory, but rather is a conceptually based system that allows us to comprehend what goes on around us, without reference to the context in which this information was first encountered. Unlike episodic memory in which information is organized temporally, the organization of semantic memory is based on meaningful associations among concepts, such that related information is connected in associative networks to allow facilitated access of concepts that share meaning. When the information from the episodic and semantic memory systems is activated and brought into awareness, a different memory system is responsible for manipulating this information in the performance of cognitive tasks. This system is described next.
Working Memory: General Characteristics
In contrast to the episodic and semantic memory systems just described, which refer to long-term memory systems, working memory refers to an integrated memory system responsible for short-term storage of information used for cognitive processing activities. While this system is not typically listed as part of declarative memory, Eichenbaum and Cohen4 have argued that is should be characterized as part of this system because of its role in holding stimuli in consciousness. While this system can be compared to earlier notions of short-term memory (STM), it is a much more complex system that works with both episodic and semantic memory in its role as facilitator of higher order cognitive activities (eg, reasoning). This system can be considered, then, as the workhorse of cognition. While multiple models of working memory exist (see Miyake and Shah,31 for an overview), all have in common the notion that a system exists both to support the maintenance (storage) of task-relevant information and to engage the appropriate processes that operate on this information during cognitive task performance. For example, the ability to perform complex multiplication problems (eg, 173 × 42) mentally requires that the entire problem, as well as the solutions occurring at each step, be held in memory while the mathematic operations are being carried out. The ability to hold the information in memory while also carrying out processes on those items is the function of working memory.
One well-known conceptualization of working memory comes from Baddeley,7 who describes this system has having multiple components, some responsible for the temporary storage and rehearsal of verbal information (through a component known as the phonological loop) or visuospatial information (through the visuospatial sketchpad; collectively, he terms these systems slave systems), with the entire system controlled by a limited-capacity attentional system known as the central executive. The central executive's role is to coordinate the operation of the other components of working memory by allocating resources to various tasks (both storage and processing) and setting and adjusting priorities to reduce potential conflict between the storage and processing functions of the entire system. Recently, this model has expanded to include a fourth component, the episodic buffer, which is assumed to be a limited-capacity temporary storage system that is capable of integrating multimodal information32 (Fig. 2). The episodic buffer plays an important role in the transfer of information from episodic memory to the central executive component of working memory, with the other slave systems assisting in the transfer of information to semantic long term memory.
A common approach for investigating both the structure and function of working memory has been the use of dual-task paradigms, in which individuals are asked to perform two tasks that either engage different components of working memory (eg, the phonological loop for the primary task and the visuospatial sketchpad for the secondary task) or similar components (eg, the phonological loop is required for both tasks). To the extent that interference effects are observed (ie, poorer performance on the primary task is noted when it is performed with the secondary task relative to the efficiency with which it is performed by itself), claims about the involvement of each of these systems in a task can be made. For example, Logie et al33 demonstrated that the efficiency with which participants could solve mental arithmetic tasks varied as a function of both the difficulty of the secondary task, and the specific slave systems involved in each of the primary and secondary tasks. They noted that when participants performed a difficult secondary task that relied on the same slave system as did the primary task (eg, the phonological loop for both tasks), significantly more errors were made as compared to conditions where the secondary task involved a different slave system (eg, the visuospatial sketchpad). These findings demonstrate the important role of the central executive in coordinating processing activities across multiple short-term storage systems. When tasks are dependent on the same slave system, performance declines relative to those conditions where coordination is required across different slave systems. Working memory is, then, much more than a simple short-term storage system. Although part of its role in memory is to act as a short-term storage buffer, its role also includes the manipulation of information necessary for reasoning, comprehension, problem solving, and other complex cognitive activities that require coordinating information flow across other memory systems.
Effects of Brain Injury and Aging on Working Memory Function
Within the context of brain injury, working memory has demonstrated specific susceptibility and impairment. In particular, the TBI literature has extensively examined how various working memory systems play a role in cognitive dysfunction,34 with the central executive component being highly implicated in contributing to the working memory deficits seen in brain injury. Several studies have found that individuals with TBI perform poorly in dual-task paradigms relative to controls.35 The general argument is that poor performance in these situations reflects impairment in the central executive system's ability to adequately regulate and coordinate the slave systems. This argument is supported by studies in which TBI patients demonstrated poor performance on tasks that specifically tapped the central executive. For example, Haut et al36 used a Sternberg task in which participants must remember a string of digits of varying lengths to make a decision about whether a probe item was part of the string to assess the speed and accuracy of STM scanning in brain-injured individuals. Overall response times to make these decisions were longer for TBI patients than they were for controls, but more importantly, these response times increased disproportionately with increasing memory load. These results suggest that when TBI patients must coordinate decision making and storage, the ability to do this efficiently becomes more problematic as storage demands increase; were the coordination function operating efficiently, then response times to make probe decisions would have increased proportionately. The general conclusion to draw from this and similar findings (see Vakil37 for an overview) is that the central executive component of working memory is more sensitive to the effects of brain injury than the storage components alone. Further support for this comes from neuroimaging studies38 in which differences between TBI patients and controls in patterns of brain activation (particularly in dorsolateral prefrontal cortex) are noted on tasks that tap the associative or strategic processing elements of working memory, but not on tasks that tap the active maintenance of stimulus representations (Table 2 for an overview of brain areas supporting the various working memory components).
The findings from the brain injury literature illustrate the important role played by the central executive in working memory and how impairments in its ability to function can disrupt coordinated task performance. There is a vast cognitive aging literature that reveals similar points. This literature documents two main areas of working memory deficit in older adults. First, research examining memory span indicates that there are age-related declines in both the short-term storage and processing components of working memory. Span tasks, depending on the specific nature of the task, can measure either STM storage or both the storage and processing functions characteristic of working memory. A typical STM span task requires a participant to repeat a series of stimuli (letters, words, or digits, presented at the rate of one per second) in the order they were presented. The initial series length is usually two items, increasing by one after every correct set of two trials, continuing until the participant can no longer correctly repeat two series of the same length. This type of span task, which appears to tap the temporary storage component of memory, can be distinguished from a working memory span task, in which the participant must not only store a series of items in memory, but also must manipulate or transform them in some way when responding. For example, in a reading span task,39 participants read sets of sentences while remembering the last word of each sentence, which they must recall in order after a series of sentences is read (series typically range from two to six sentences, with three trials for each series). This type of span task requires both storage (remember the last word from each sentence) as well as online processing (eg, sentence comprehension) and therefore is argued to measure working memory function. A recent meta-analysis of the memory span literature40 notes that while there are age differences in performance favoring young adults for both short-term and working memory span tasks, these differences are more pronounced for working memory span measures.
The additional age-related costs in performance on working memory span tasks relative to simple short-term storage tasks appear to reflect a greater age-related decline in working memory processing functions than storage functions, impairments similar to those that occur with TBI. Efforts to understand the exact nature of these processes has produced considerable investigation. Salthouse41 argues that age-related declines in processing are due to age-related reductions in the speed with which these processes can be executed. These speed reductions affect both the number of items that can be processed and the number of processes that can be executed in a given amount of time. Individuals with slower processing speed (eg, older adults) consequently would not be able to perform as well on tasks that required rapid synchronization and coordination of task activities that are typically seen in working memory span measures. Others argue that these processing declines are the result of an age-related decrease in inhibitory control in which older adults become less able to inhibit the processing of irrelevant information that enters working memory, leaving them less able to rid working memory of “mental clutter.”42 In essence, this reduces the functional capacity of working memory and leaves older adults more subject to interference effects from irrelevant information during processing.
It has also been proposed that age differences in working memory function are linked to age-related declines in the ability to coordinate tasks and processing streams.43 Indeed, the second major area of focus regarding working memory function in older adults has been an examination of this coordinating aspect of executive control. A popular paradigm for this investigation is the dual-task paradigm, where the simultaneous processing of two (or more) streams of information must be coordinated (as was described earlier). Verhaeghen et al44 conducted a meta-analysis of this literature and concluded that this control process is particularly sensitive to the effects of aging, with older adults generally performing more slowly (even taking into account general cognitive slowing) and less accurately than young adults in a variety of dual-task paradigms. Further, they argue that the process underlying this ability to coordinate multiple streams of information is a focus-switching process,45 a process responsible for moving information in and out of focused attention.
Recently, Oberauer46 has argued that the focus of attention within working memory serves to select information for further cognitive action. This focus has an extremely limited capacity (perhaps as small as one item), and the ability to switch focus from one item to another is time-consuming.47 While an item is in this focus of attention, it can be actively processed. Other task-relevant information is maintained outside the focus of attention and can be brought into focus when the focus-switch process is executed in working memory. It is this specific control process (whose roles include the scheduling and regulation of focused attention) that appears to be particularly sensitive to the effects of aging, making it more difficult for older adults to perform concurrent tasks.
Working Memory: Summary
The above discussion points out a number of specific changes in working memory function that can result from brain injury or aging. As a group, though, the primary point to glean from this discussion is that while working memory plays a role in memory storage, it is really a processing system that serves complex cognition. As a “memory” system then, it is unique relative to the other memory systems, maintaining a close relationship between its storage and processing functions, the latter of which can be argued to be its central role. While other memory systems have both storage and processing functions, the relationship between these two is not always apparent when long-term memory systems (whose primary function can be argued to be storage) are discussed. The working memory system is critical to cognition, relied on for a variety of everyday cognitive tasks (eg, reading the newspaper, determining the better bargain for products on sale at the supermarket) with multiple steps and intermediate results that must be maintained and monitored moment to moment so that the task can be successfully completed. When this system cannot operate quickly enough or cannot efficiently switch focus across these steps, important information relevant to the task can be lost, and the result is that these everyday tasks cannot be completed.
In contrast to our discussion so far, in which memory has been described as a set of dissociable, consciously accessible systems, nondeclarative memory refers to a set of systems that cannot be accessed through consciousness, but rather express their contents through task performance.48 Historically, this system has been most closely aligned with the concept of procedural memory,49 a system designed to support skill learning. More recently, however, nondeclarative memory has been argued to comprise a number of systems that support learning in a variety of ways, including classic conditioning and perceptual processing of word and object forms. The discussion focuses on two of these systems: procedural memory and the perceptual representation system (PRS) (Table 3).
Procedural Memory: General Characteristics
As stated previously, procedural memory is the memory system that supports cognitive, motor, and perceptual skill learning. This system is characterized by gradual, incremental learning that produces representations lacking information about experiences. Rather, its representations comprise the algorithms, rules, and actions that allow performance of various skills (eg, riding a bike, playing a musical instrument). As such, its output, unlike the output from the declarative memory systems, is noncognitive, ie, cannot be provided verbally. This content makes its existence known through task performance, ie, faster or more accurate performance of various skilled tasks across trials. With repetition, the performance of skills supported by this system can become automatic.
Effects of Brain Injury and Aging on Procedural Memory Function
It is well documented that procedural memory for tasks learned prior to injury (or well-practiced skills) are minimally affected by varying types of TBI.50,51 In fact, some rehabilitation protocols use a patient's spared procedural memory to aid in memory remediation.52 There are, however, some reports in the literature that question whether all aspects of skill learning are preserved in TBI. Tasks that appear to be mediated by frontal lobe regions are more likely to show poor performance than are perceptual tasks. These types of tasks tend to be those that require planning and coordination (eg, the Tower of Hanoi task) as opposed to requiring perceptual processing (eg, search-detection tasks) and therefore may be relying on working memory function. As was described earlier, this aspect of memory (and more specifically the central executive) is very sensitive to brain injury and so may have a consequent impact on a TBI patient's skill-learning ability.
Similarly, procedural memory shows little decline across the life span relative to declarative memory systems. While older adults are slower and less accurate relative to young adults when performing perceptual-motor tasks,53 and older adults often do not reach the same levels of skilled performance as young adults,54 they do appear to be able to acquire and retain new skills over an extended period of time. For example, Smith et al55 assessed long-term memory for a newly learned motor skill in a group of subjects aged 18 to 95 years. They noted that while performance times and learning of the task were slower with increasing age, all age groups showed significantly faster performance times on the task after a 2-year delay between initial learning and retesting, suggesting that the procedural memory traces acquired during training were well preserved and intact despite the age-associated differences in the amount of time needed to acquire and perform the task.
Overall, the observations of relatively spared procedural memory in TBI and aging indicate that this system is controlled by different brain regions (primarily the neostriatum and cerebellum56) than those supporting declarative memory functions (primarily the medial temporal lobe and prefrontal cortical regions) and relies on different processes than does declarative memory. This system is a “know how” as opposed to a “know that” or “know what” system that is critical to the acquisition and execution of various skills. As such, it is considered a primitive form of memory, developing earlier than the other systems57 to support everyday human learning.58
PRS: General Characteristics
While the procedural memory system supports skill learning, the PRS supports identification of perceptual features, allowing facilitated identification of previously encountered objects via a phenomenon known as repetition priming. Priming is demonstrated when there is an increased probability of identifying, or a reduced latency in identifying, a previously encountered stimulus relative to a new stimulus. In a typical priming experiment, participants encounter a set of stimuli (eg, words, line drawings of objects) in a study phase, and later in a test phase, they are asked to perform a task in which these stimuli reappear, along with new stimuli. The decreased speed and/or increased accuracy for performance on old stimuli relative to new stimuli reflects the magnitude of priming. This type of memory has been termed implicit memory, although this term is really only a descriptive label that refers to the manner in which past experience is expressed (ie, implicitly or unintentionally rather than explicitly or deliberately). Reliable priming effects have been observed in both amnesic patients59 and healthy older adults,60 despite deficits (of varying degrees) in episodic memory performance (as was described earlier). It has been argued that the primary function of this system is to support episodic memory function by allowing for perceptual processing of stimuli prior to formation of episodic memory representations.61
The PRS, like working memory, is a multicomponent system, with separate domain-specific subsystems designed for processing of visual word forms, structural description of objects, and auditory word forms. All three of these subsystems have been argued to operate presemantically,61 ie, with no access to word or object meaning and therefore support perceptual repetition priming phenomena. Importantly, however, while many of the tasks employed in research to examine repetition priming effects are perceptually based, some tasks (eg, category exemplar generation tasks in which individuals list members of categories such as KIWI for the category FRUIT) tap into semantic processing and are considered to be conceptually based.62 The observation of both perceptual and conceptual repetition priming indicates that this phenomenon is not exclusively a perceptual process, but that the system and mechanisms that support repetition priming may also access semantic memory. Regardless of these interactions with other memory systems, however, these type of priming effects are expressions of a memory system that is functionally different than those systems that support recollection. This point becomes clearer when the brain injury (amnesic) and aging literature are examined.
Effects of Brain Injury and Aging on the PRS
As was mentioned earlier in the general description of the PRS, priming has been shown to be generally intact in patients with brain injury. To illustrate this point, consider the case of L.H.,63 who has bilateral occipital lobe lesions. On a variety of perceptual priming tasks, L.H. demonstrates impaired performance relative to H.M.,64 who has bilateral medial temporal lobe lesions, but no impairments in perceptual priming. Unlike H.M., however, who has significant impairments in the ability to recollect recently experienced events and recently encountered stimuli, L.H.'s recollective abilities are relatively normal, as is also his conceptual priming. These findings suggest that perceptual priming, as an aspect of the PRS, is mediated by different neural substrates than is episodic memory, as well as illustrate how the PRS is a memory system with different functions than the other memory systems described earlier.
As further evidence for this argument, it is important to look at the effects of aging on the operation of the PRS. Meta-analyses of the repetition priming literature60,65 indicate that while small reliable age differences exist in some types of repetition priming, these age differences are reliably smaller than those observed for recall and recognition tests. These latter findings suggest that there may be some overlap in the mechanisms that support repetition priming and episodic memory performance, with arguments being made66 that this mechanism is a fluency mechanism that produces facilitated reprocessing of the perceptual features of previously experienced stimuli, thus supporting repetition priming. This perceptual fluency is also believed to support familiarity processing within episodic memory, leading to increased stimulus recognition. As a caveat, however, Wagner and Gabrieli67 have obtained evidence to refute this argument, so it is still unclear as to why older adults' priming performance exhibits small decrements relative to young adults. Regardless of these findings, however, the generally consistent findings from the amnesic and aging literature suggest that priming represents a form of memory that is maintained despite severe brain injury or normative aging processes.
This discussion and description of various memory systems points to the complexity of human memory function. To support the hypothesis that memory comprises multiple systems rather than being a single system, we presented evidence from the brain injury and aging literature that reveal important performance dissociations across memory measures that assess different types of recollections (eg, recollection of past events versus recollection of facts). Whether or not this evidence is conclusive, it nonetheless points to the fact that human memory function is more complicated than the early philosophical notion that memory serves to record all that we experience, and when damaged, all is lost. Clearly, some aspects of memory do function this way, but the multiple systems coexist and work together to prevent an entire loss of function. Together, they help us acquire a sense of self by integrating our experiences into meaningful recollections and allow us to comprehend the world around us by tapping into knowledge stores, functions that go beyond simple, passive storage. In answering the question, then, “What is human memory?,” we can tentatively answer it by describing the various functions of each of its systems. Memory is the storage and retrieval of our experiences (episodic memory); it is the encoding and application of concepts, knowledge, and language (semantic memory); it is the acquisition and execution of rules, algorithms, and procedures that allow us to perform cognitive, motor, and other skills (procedural memory); it is our conscious problem solver and mediator of other cognitive activities (working memory); finally, it is the processor of perceptual experiences (the PRS).
1. Klein SB, Loftus J, Kihlstrom JF. Memory and temporal experience: the effects of episodic memory loss on an amnesic patient's ability to remember the past and imagine the future. Soc Cogn
2. Tulving E. Study of memory: processes and systems. In: Foster JK, Jelicic M, eds. Memory: Systems, Processes, or Function
. New York: Oxford University Press, 1999:11–30.
3. Foster JK, Jelicic M, eds. Memory: Systems, Processes, or Function
. New York: Oxford University Press, 1999.
4. Eichenbaum H, Cohen NJ. From Conditioning to Conscious Recollection: Memory Systems of the Brain
. New York: Oxford University Press, 2001.
5. Schacter DL, Tulving E. What are the memory systems of 1994? In: Schacter DL, Tulving E, eds. Memory Systems 1994
. Cambridge, MA: MIT Press, 1994:1–38.
6. Squire LR. Declarative and nondeclarative memory: multiple brain systems supporting learning and memory. In: Schacter DL, Tulving E, eds. Memory Systems 1994
. Cambridge, MA: MIT Press, 1994:203–232.
7. Baddeley A. Working Memory
. Oxford: Clarendon Press, 1986.
8. Tulving E. Episodic and semantic memory. In: Tulving E, Donaldson W, eds. Organization of Memory
. New York: Academic Press, 1972:381–403.
9. Tulving E. Memory and consciousness. Can Psychol
10. Tulving E, Thomson DM. Encoding specificity and retrieval processes in episodic memory. Psychol Rev
11. Morris CD, Bransford JD, Franks JJ. Levels of processing versus transfer appropriate processing. J Verb Learn Verb Behav
12. Tulving E. Episodic memory: from mind to brain. Am Psychol
13. Vargha-Khadem F, Salmond CH, Watkins KE, et al. Developmental amnesia: effects of age at injury. Proc Natl Acad Sci U S A
14. Zec R, Zellers D, Belman J, et al. Long-term consequences of severe closed head injury on episodic memory. J Clin Exp Neuropsychol
15. Levine B, Black S, Cabeza R, et al. Episodic memory and the self in a case of isolated retrograde amnesia. Brain
16. Yonelinas AP. The nature of recollection and familiarity: a review of 30 years of research. J Mem Lang
17. Backman L, Small BJ, Wahlin A. Aging and memory: cognitive and biological perspectives. In: Birren JE, Schaie KW, eds. Handbook of the Psychology of Aging
. San Diego, CA: Academic Press; 2001:349–377.
18. Luszcz MA, Bryan J. Toward understanding age-related memory loss in late adulthood. Gerontology
19. Naveh-Benjamin M. Adult age differences in memory performance: tests of an associative deficit hypothesis. J Exp Psychol Learn
20. Naveh-Benjamin M, Guez J, Kilb A, et al. Associative memory deficit of older adults: further support using face-name associations. Psychol Aging
21. Tulving E. What kind of a hypothesis is the distinction between episodic and semantic memory? J Exp Psychol Learn
22. Mitchell DB. How many memory systems? Evidence from aging. J Exp Psychol Learn
23. De Renzi E, Liootti M, Nichelli P. Semantic amnesia with preservation of autobiographic memory. A case report. Cortex
24. Yasuda K, Waanabe O, Ono Y. Dissociation between semantic and autobiographical memory: a case report. Cortex
25. Wilson BA. Semantic memory impairments following non-progressive brain injury: a study of four cases. Brain Inj
26. Levin HS, Goldstein FC. Organization of verbal memory after severe closed-head injury. J Clin Exp Neuropsychol
27. Haut MW, Petros TV, Frank RG. Speed of processing within semantic memory following severe closed head injury. Brain Cogn
28. Lindenberger U, Baltes PB. Intellectual functioning in old and very old age: cross-sectional results from the Berlin Aging Study. Psychol Aging
29. Laverm GD, Burke DM. Why do semantic priming effects increase in old age? A meta-analysis. Psychol Aging
30. Verhaeghen P. Aging and vocabulary scores: a meta-analysis. Psychol Aging
31. Miyakem A, Shah P. Models of Working Memory: Mechanisms of Active Maintenance and Executive Control
. New York: Cambridge University Press, 1999.
32. Baddeley A. The episodic buffer: a new component of working memory? Trends Cogn Sci
33. Logie RH, Gilhooly KJ, Wynn V. Counting on working memory in arithmetic problem solving. Mem Cogn
34. Leclercq M, Couillet J, Azouvi P, et al. Dual task performance after severe diffuse traumatic brain injury or vascular prefrontal damage. J Clin Exp Neuropsychol
35. McDowell S, Whyte J, D'Esposito M. Working memory impairments in traumatic brain injury: evidence from a dual-task paradigm. Neuropsychologia
36. Haut M, Petros T, Frank R, et al. Short-term memory processes following closed head injury. Arch Clin Neuropsychol
37. Vakil E. The effect of moderate to severe traumatic brain injury (TBI) on different aspects of memory: a selective review. J Clin Exp Neuropsychol
38. Christodoulou C, DeLuca J, Ricker J, et al. Functional magnetic resonance imaging of working memory impairment after traumatic brain injury. J Neurol Neurosurg Psychiatry
39. Daneman M, Carpenter PA. Individual differences in working memory and reading. J Verb Learn Verb Behav
40. Bopp KL, Verhaeghen P. Aging and verbal memory span: a meta-analysis. J Gerontol B Psychol Sci Soc Sci
41. Salthouse TA. The processing-speed theory of adult age differences in cognition. Psychol Rev
42. Hasher L, Zacks RT. Working memory, comprehension, and aging: a review and a new view. In: Bower GH, ed. The Psychology of Learning and Motivation. Volume 22
. San Diego, CA: Academic Press, 1988:193–225.
43. Mayr W, Kliegl R. Sequential and coordinative complexity: age-based processing limitations in figural transformations. J Exp Psychol Learn
44. Verhaeghen P, Steitz DW, Sliwinski MJ, et al. Aging and dual-task performance: a meta-analysis. Psychol Aging
45. Verhaeghen P, Basak C. Ageing and switching of the focus of attention in working memory: results from a modified N
-back task. Q J Exp Psychol A
46. Oberauer K. Access to information in working memory: exploring the focus of attention. J Exp Psychol Learn
47. Garavan H. Serial attention within working memory. Mem Cogn
48. Squire LR. Memory systems of the brain: a brief history and current perspective. Neurobiol Learn Mem
49. Cohen NJ, Squire LR. Preserved learning and retention of pattern-analyzing skill in amnesia: dissociation of knowing how and knowing that. Science
50. Schmitter-Edgecombe M. Effects of divided attention on implicit and explicit memory performance following severe closed head injury. Neuropsychology
51. Schmitter-Edgecombe M, Nissley HM. Effects of divided attention on automatic and controlled components of memory after severe closed-head injury. Neuropsychology
52. Donaghy S, Williams W. A new protocol for training severely impaired patients in the usage of memory journals. Brain Inj
53. Harrington DL, Haaland KY. Skill learning in the elderly: diminished implicit and explicit memory for a motor sequence. Psychol Aging
54. Touron DR, Hoyer WJ, Cerela J. Cognitive skill acquisition and transfer in younger and older adults. Psychol Aging
55. Smith CD, Walton A, Loveland AD, et al. Memories that last in old age: motor skill learning and memory preservation. Neurobiol Aging
56. Thac WT, Goodkin HP, Keating JG. The cerebellum and the adaptive coordination of movement. Annu Rev Neurosci
57. Bachevalier J. Ontogenetic development of habit and memory formation in primates. Ann N Y Acad Sci
58. Gupta P, Cohen NJ. Theoretical and computational analysis of skill learning, repetition priming, and procedural memory. Psychol Rev
59. Cave CB, Squire LR. Intact and long-lasting repetition priming in amnesia. J Exp Psychol Learn
60. Light LL, Prull M, LaVoie DJ, et al. Dual process theories of memory in aging. In: Perfect T, Maylor E, eds. Theoretical Debates in Cognitive Aging
. New York: Oxford University Press, 2000:238–300.
61. Tulving E, Schacter DL. Priming and human memory systems. Science
62. Roediger HL, McDermott KB. Implicit memory in normal human subjects. In: Spinnler H, Boller F, eds. Handbook of Neuropsychology. Volume 22
. Amsterdam: Elsevier, 1993.
63. Keane MM, Gabrieli JDE, Mapstone HC, et al. Double dissociation of memory capacities after bilateral occipital-lobe or medial temporal-lobe lesions. Brain
64. Corkin S. What's new with the amnesic patient H.M.? Nat Rev Neurosci
65. LaVoie DJ, Light LL. Adult age differences in repetition priming: a meta-analysis. Psychol Aging
66. Jacoby LL, Dallas M. On the relationship between autobiographical memory and perceptual learning. J Exp Psychol Gen
67. Wagner AD, Gabrieli JDE. On the relationship between recognition familiarity and perceptual fluency: evidence for distinct mnemonic process. Acta Psychol
68. Baddeley A. Working memory: looking back and looking forward. Nat Rev Neurosci
69. Smith EE, Jonides J. Working memory: a view from neuroimaging. Cogn Psychol
70. Poldrack RA, Gabrieli JDE. Functional anatomy of long-term memory. J Clin Neurophysiol