Age-Related Changes in Memory: Experimental Approaches

Susan R Old & Moshe Naveh-Benjamin. Handbook of Cognitive Aging: Interdisciplinary Perspectives. Editor: Scott M Hofer & Duane F Alwin. Sage Publications, 2008.

It is commonly thought that aging leads to an overall decline in memory performance. It is interesting, however, that the effects of age are highly dependent upon the type of memory task being administered, with some tasks showing large impairment yet others showing no decline or even improvement into old age. An important goal of research on memory and aging is to explain this variability. In the first section of this chapter, we review empirical findings regarding age-related differences in several domains of memory performance, distinguishing between short- and long-term memory, explicit and implicit memory, and semantic and episodic memory. We also examine age-related differences in specific processes involved in episodic memory—encoding, retention, and retrieval—as well as various other manifestations of age-related episodic memory differences. Next, we discuss several general theoretical approaches that have been offered to explain age-related cognitive changes, examining how these approaches can explain age-related differences in memory performance. Finally, we note ways in which future researchers can use empirical findings to generate memory-preserving interventions for older adults, and we address the need for the integration of various lines of research.

Empirical Evidence

Numerous studies have been carried out in the different domains of memory and aging. Because we cannot review all of those studies, we have chosen to first present relevant meta-analytical findings, when such findings are available. Meta-analyses provide quantitative summaries of a large number of studies on a given topic, thus allowing researchers to reach generalizable conclusions. After citing such meta-analytical studies in each domain, we provide examples of representative empirical works, in order to describe typical methods and results in that area of study. Most of these examples employ cross-sectional designs, in which different age groups of adults (usually, young ones in their 20s, and older ones, in good health, between 65 and 85 years of age) are tested at a given point in time. The cross-sectional approach allows researchers to determine overall age differences on a variety of tasks. It should be noted, however, that it is also important to investigate intraindividual variability in the effects of age on memory (see, e.g., Lindenberger & von Oertzen, 2006; see also chaps. 27 and 28, this volume, for a discussion of longitudinal designs in aging research, which can determine changes in single individuals across time). Furthermore, it is important to address individual differences in change across middle adulthood (e.g., Martin & Zimprich, 2005); it has been suggested, for example, that there are subtypes of “stables,” “decliners,” and “gainers” in terms of memory performance in middle age (see Willis & Schaie, 2005). Despite the limitations of the cross-sectional approach, it does provide advantages, such as controlled conditions and a relatively brief time commitment. Readers interested in individual differences in cognitive aging are referred to chapter 2 in this volume.

Short-Term Memory and Working Memory

Short-term memory (STM) involves the ability to retain a recently experienced event for a brief period of time, such as repeating a phone number until it can be dialed. The original conceptualization of STM was based on a mechanism that allows temporary storage of information (Atkinson & Shiffrin, 1968). The later concept of working memory (WM; Baddeley & Hitch, 1974) involves the simultaneous maintenance and active manipulation of information. For example, when multiplying two 2-digit numbers, you must maintain some of the digits and products while multiplying other digits. Thus, whereas STM requires only maintenance, WM requires both maintenance and processing.

Aging seems to affect WM more negatively than STM. For example, a meta-analysis by Bopp and Verhaeghen (2005) examined age differences in several verbal tasks and indicated relatively small age differences in tasks requiring simple temporary maintenance of materials. Forward digit span, for instance, showed modest age-related differences (d = -0.53, where d is the mean effect size in terms of unit standard deviations over all the studies included). As a processing component was added to the task, however, and as processing became more dominant relative to the maintenance component, age differences became much larger (ds = -1.01, -1.27, and -1.54 for sentence span, listening span, and computation span tasks, respectively).

One relevant empirical work was conducted by Park et al. (2002), who asked 345 people, ranging in age from 20 to 92 years, to complete a series of tasks involving visuospatial and verbal measures of both STM and WM. STM tasks included Corsi blocks tasks, in which participants tried to replicate patterns tapped out on raised blocks by the experimenter, as well as a digit span task, in which strings of presented digits were to be recalled in order. WM tasks included reading span and computation span tasks, in which participants answered series of questions or math problems, respectively. Following an entire series, they were asked to recall the last word from each question or the last number from each math problem. Results showed that, whereas all measures of WM and STM declined with age, this decline was larger in the WM than in the STM tasks. It seems, then, that older adults are more impaired on tasks that require both processing and storage than on tasks requiring only storage and that age-related differences increase as the processing required becomes more demanding.

Implicit/Indirect Memory Versus Explicit/Direct Memory

Another notable distinction made by memory researchers is that between explicit (or declarative/direct) and implicit (or nondeclarative/indirect) memory systems (e.g., Squire, 1986). The former is involved when conscious intentional retrieval is used (e.g., trying to remember an address that was given to you yesterday), whereas the latter involves memories that can be inferred by subsequent behavior without any intention of memory retrieval (e.g., responding quickly to a familiar face). Research conducted during the last 20 years shows that aging affects these two forms of memory differently. A meta-analysis carried out by Light and La Voie (1993), for example, concluded that age-related declines on implicit memory measures (d = -0.18) are much smaller than those on explicit measures, such as recognition or recall, which, in other meta-analytical studies, reviewed below, show d values ranging from -0.5 to -1.5.

Such a pattern was obtained by D. B. Mitchell and Bruss (2003). In their study, young, middle-aged, and older adults named a combination of presented words and pictures, unaware of the implicit and explicit memory tests they would be given at a later point. In one of the implicit tests, participants were given a category name and were asked to name six exemplars from that category. Words and pictures from the previous list fit into some of the categories, but participants were not aware of this and therefore did not intentionally try to report previously named items (targets). Implicit memory could therefore be measured by comparing participants’ reporting of targets to a baseline measure produced by people who had not been exposed to the initial list of words and pictures. Participants also took part in an explicit memory task, in which they were asked to recall the two words and two pictures from the original list that belonged to a given category. The results indicated no significant age differences in the implicit memory tasks—in fact, age was slightly related to an improvement in this measure—whereas the explicit test showed significant age-related declines (see also Light & Albertson, 1989; Light & Singh, 1987).

The major conclusion based upon such studies is that older adults are impaired more on explicit than on implicit memory measures. Hence, in order to assess and predict potential age-related changes, it is important to take into account the degree to which a given task involves intentional and nonintentional memory retrieval.

Long-Term Memory: Episodic Versus Semantic Memory

Within the declarative memory system, a distinction has been made (Tulving, 1972) between episodic memory, which involves one’s personal memories tied to a particular time and place (e.g., remembering who came to your recent birthday party), and semantic memory, which involves knowledge of general facts not related to a specific time and place (e.g., naming the capital of Germany or defining a given word). Like the explicit/implicit distinction discussed above, episodic and semantic memory seem to be affected differently by aging.

In one study, Rönnlund, Nyberg, Bäckman, and Nilsson (2005) analyzed data collected from 1,000 participants between 35 and 80 years of age, on two occasions separated by 5 years. Episodic memory measures included recall of self-performed and other-performed actions, recall of nouns, and recall of statements. Semantic measures included tests of general knowledge, vocabulary, and word fluency. After the researchers adjusted the results for practice effects and education levels, the longitudinal and cross-sectional data led to converging conclusions. Episodic memory stayed fairly stable until around the age of 55-60, at which point there was a large decline in performance. Semantic memory performance, on the other hand, increased between the ages of 35 and 55, leveled off, then declined beginning around the age of 65; this decline, however, was less substantial than that for episodic memory (see also Park et al., 2002).

Using other episodic and semantic memory tasks in a cross-sectional design, Spaniol, Madden, and Voss (2006) asked older and younger adults to judge the pleasantness of a series of words, half of which described living things. After a 1-minute retention interval, each participant took part in either an episodic test, in which they indicated whether a given word was from the study list, or a semantic memory test, in which they decided if a given word described a living or nonliving thing. In both tests, they were asked to respond as quickly as possible. Results indicated that older adults performed less accurately than younger adults on the episodic task, but not on the semantic task. Furthermore, response time data showed that, although both age groups were slower to respond to the episodic than to the semantic task, this difference was larger in the older than the younger adults.

This finding of differential effects of age on episodic and semantic memory seems to be a robust finding. For example, it applies not only to explicit memory, as described above, but also to implicit memory, with older adults showing semantic priming effects comparable to—and often larger than—those of younger adults (e.g., Laver & Burke, 1993; Madden, 1988). Overall, then, it seems that semantic memory is relatively spared of age-related decline. Episodic memory, however, can be quite impaired with advancing age; we will discuss this impairment in more detail in the following section.

Episodic Memory

In this section, we examine research on several factors that affect age-related changes in episodic memory. We have classified many of those factors as occurring at encoding (i.e., the changing of outside information into an internal representation), at retention/maintenance, or at retrieval (i.e., accessing from memory), of the information. We also highlight various other manifestations of age-related differences in episodic memory.

Encoding Processes and Age-Related Memory Changes

One distinction frequently used to address both theoretical and applied questions is that between incidental and intentional learning. In the context of aging, the question is whether the magnitude of age-related differences is similar when information is learned without expectation of later memory tests (e.g., when witnessing a crime) as when learning occurs with the knowledge that this information will be required at a later time (e.g., when trying to commit to memory a phone number for future use). A meta-analysis by Spencer and Raz (1995) addressed this issue and found larger age effects in studies involving memory for intentionally learned (d = -0.62) than incidentally learned (d = -0.41) materials. R. E. Johnson (2003) reached a similar conclusion in a meta-analysis of memory for text (d = -0.85 with advance knowledge of an upcoming test, and d = -0.55 when this information was withheld).

One relevant study was conducted by Troyer, Hfliger, Cadieux, and Craik (2006), who compared age-related differences in incidental and intentional learning of names. Older and younger adults viewed a series of surnames and were told that they would need to remember only specified names. Before each presented name, participants were given encoding instructions that involved physical processing (stating the first letter of the surname), phonemic processing (stating a word that rhymed with the name), semantic processing (defining or creating an association with the name), or intentional learning (simply trying to remember the name for a later test). Whether testing involved recall or recognition, younger adults outperformed older adults in memory for intentionally learned names, but there were no significant age differences in any of the other encoding conditions for either test form (i.e., for incidentally learned names).

Such results indicate that older adults can encode information incidentally quite well but that they are highly impaired when intentional learning is required. One possible reason for this finding, supported by Dunlosky and Hertzog (2001), is that older adults do not spontaneously use strategies that are as effective as those employed by younger adults at encoding and at retrieval.

Retention Processes and Age-Related Memory Changes

Whereas the differential effects of encoding instructions on age-related memory differences are largely consistent, the picture regarding forgetting rates is much less clear. A meta-analysis by Verhaeghen, Marcoen, and Goossens (1993) found similar age differences on immediate as on delayed memory measures; this applies to tests of list recall (ds = -1.01 and -0.83 for immediate and delayed recall, respectively), paired-associate recall (ds = -0.84 and -0.97), and prose recall (ds = -0.66 and -0.85). R. E. Johnson (2003), however, conducted a meta-analysis of studies involving memory for text and found that age differences were significantly larger when testing occurred between 1 and 10 minutes after study (d = -0.97, with one mean effect size per study), compared to both immediate testing (d = -0.78) and testing after more than 10 minutes (d = 0.70). Thus, the breakdown of retention intervals seems to be an important factor in these investigations.

This lack of converging evidence is shown by specific empirical studies as well. Park, Royal, Dudley, and Morrell (1988) conducted a study to determine if there are age differences in rates of forgetting picture information over an extended period of time. Older and younger adults studied a series of line drawings. Each participant took recognition tests over the drawings 3 minutes, 48 hours, 1 week, 2 weeks, and 4 weeks after study, with no stimulus appearing in more than one test. Results supported previous work (e.g., Rybarczyk, Hart, & Harkins, 1987) in that there was no age-related difference in forgetting rates when testing occurred within 48 hours of study. However, older adults exhibited larger forgetting rates than younger adults beyond this 48-hour interval. These findings point to the importance of the retention intervals studied and offer one explanation for why only some studies find differential effects of age on forgetting.

Another suggested reason for discrepant findings was put forth by Wheeler (2000), who tested word recall of older and younger adults either 3 minutes or 1 hour after incidental learning took place. Whether testing occurred between or within subjects, results showed larger rates of forgetting in the older than in the younger participants. However, although age differences were consistent in these experiments, effect sizes were quite modest; thus, Wheeler offered this as an explanation for why many previous studies have found similar rates of forgetting in younger and older adults. It seems likely that age differences do increase along with retention intervals but that this effect is small and may differ according to the precise retention intervals being tested.

Retrieval from Memory

Test Type. For explicit memory tasks, processes involved in accessing information when it is needed—that is, retrieving it—are as important as those involved in the encoding of information. One factor that affects age-related memory differences is the degree to which a test employs retrieval cues. For example, free-recall tasks provide no cues to participants, whereas cued-recall tasks do provide cuing. In recognition tests, participants receive copies of original stimuli as cues; thus recognition involves a greater degree of cuing than either free recall or cued recall. By comparing age-related differences on each type of test, it is possible to determine the degree to which older and younger adults utilize these cues.

Spencer and Raz (1995), in their meta-analysis mentioned above, found larger age differences in tests of recall (d = -1.01, which includes both free and cued recall) than in recognition tests (d = -0.57). Similarly, in a meta-analysis involving memory for text, R. E. Johnson (2003) showed that age effects were smaller in recognition tests (d = -0.67) than in free recall (d = -0.82) or cued recall (d = -0.88) tests. Such an effect is exemplified by the aforementioned study by Troyer et al. (2006; Experiment 1). At study, older and younger participants were presented with surnames. In a recall test, participants were simply asked to write down all of the names they could remember. In a recognition test, they responded as to whether a presented name had appeared in the study list. Results showed an age-related impairment in the ability to recall names, but there was no significant age difference in terms of recognition test performance.

Cued recall also seems to be more impaired with age than is recognition. For example, Craik and McDowd (1987) reported a large age-related deficit in the ability to use a cue phrase (e.g., “a body of water) to recall a target word (e.g., pond) following study of the phrase-target word pairs. However, there were no significant age differences in the ability to make yes-no responses to target and distractor words.

One proposed explanation for this differential effect of test type on age-related memory differences involves Craik’s (1986) notion of self-initiated operations. According to this idea, older adults have difficulty carrying out tasks that require a high degree of self-initiated processing—that is, tasks not accompanied by sufficient environmental support. Because recognition tasks include more environmental support than recall tasks, and cued-recall tasks include more environmental support than free recall, Craik’s idea is supported by empirical evidence such as that mentioned above. A related explanation for such findings is the idea that aging has a more negative impact on tasks requiring controlled, effortful processes than on tasks mediated by more automatic processes (Hasher & Zacks, 1979).

Recollection and Familiarity. There has been a recent distinction made between two types of retrieval processes (e.g., Yonelinas, 2002). The first, recollection, requires retrieval of contextual details of an episode, whereas the second, familiarity, is based on a feeling of having previously experienced an event without remembering any specific contextual details. Three major paradigms used to measure recollection and familiarity are the process dissociation procedure (PDP), the Remember/Know (R/K) method, and the assessment of receiver operating characteristics (ROC) curves.

Studies using the PDP (Jacoby, 1991) show an age-related decline in retrievals based on recollection but not on familiarity (e.g., Hay & Jacoby, 1999). One experiment using this method was conducted by Jennings and Jacoby (1993, Experiment 2). In the first phase of the study, older and younger adults read a list of words, unaware that memory for the words would later be tested. In the next phase, participants were told to remember auditorily presented words for a later test. Finally, in each of two tests, participants were presented with a word from one of the previous phases and a new word. In the exclusion test, participants were falsely told that all test pairs included one auditorily presented word and that they should choose that word; because no word had been presented in both study lists, recollecting that one of the test words had appeared in the visual phase would lead participants to choose the other word in the test pair. Thus, incorrectly choosing a visually presented word would indicate familiarity in the absence of recollection. In the inclusion test, participants were correctly informed that each word pair included one old and one new word and were asked to choose the old word. With these instructions, choosing words from the visual phase could indicate either recollection or familiarity without recollection. The researchers used performance on the tests to calculate estimates of recollection and familiarity. Younger adults’ recollection estimates were higher than those for older adults, whereas there were no significant age differences in familiarity estimates. The authors thus concluded that aging impairs recollection but leaves familiarity intact. Variations of the PDP have been developed in more recent studies (e.g., Toth & Parks, 2006).

The R/K method makes use of participants’ reports of whether they “remember” a particular stimulus (indicating recollection) or simply feel or “know” that the stimulus was previously presented (representing familiarity). The ROC curves method involves plotting hit rates against false alarm rates at various levels of confidence. Like the PDP method, studies using these two methods show a definite age-related decline in recollection. The picture regarding familiarity is less clear, however. Prull, Dawes, Martin, Rosenberg, and Light (2006) used all three procedures discussed here; the PDP method found no age-related deficit in familiarity, but the R/K and ROC methods did show such a deficit. This finding is consistent with other studies that have used R/K (e.g., Light, Prull, La Voie, & Healy, 2000) and ROC (e.g., Healy, Light, & Chung, 2005) methods.

Overall, then, although age-related patterns in familiarity seem to depend on the measurement method used, there is a great deal of evidence that recollection declines with age. It has been claimed (e.g., Jacoby, Jennings, & Hay, 1996) that this recollection deficit—an inability during retrieval to access the details of an episode—is largely responsible for older adults’ memory deficit. This notion can explain the relatively adequate performance of older adults in tasks involving implicit or semantic memory, because such tasks do not require detailed conscious recollection of an original event.

Other Manifestations of Age-Related Episodic Memory Change

Memory for Content Versus Context-Source. A defining characteristic of episodic memory is that it is tied to a time and place. If you are asked whom you met last Thanksgiving, you can use various contextual aspects of the episode—for example, perhaps the time (Thanksgiving dinner) and the place (your brother’s house) where you had the festive dinner—in order to retrieve the relevant information about the people you have met there. Studies show that older adults demonstrate poor memory for such contextual elements, including the external source of presented information (e.g., the voice in which information was presented) and whether it was presented or imagined (e.g., M. K. Johnson, Hashtroudi, & Lindsay, 1993), relative to their memory for the content of the event (e.g., what was said or presented). Spencer and Raz (1995) reviewed evidence from 46 studies involving both young and old participants and found larger age-related differences in memory for context (d = – 0.90) than for content (d = -0.72).

A representative study was conducted by Simons, Dodson, Bell, and Schacter (2004). Older and younger adults heard a series of sentences read by one of four speakers (two male and two female) while viewing the written sentence and a photograph of the speaker. A surprise test indicated that older adults were just as able as young adults to distinguish between old and new sentences. However, they were less able, compared to younger participants, to identify the speaker of earlier presented sentences. The authors concluded that although older adults can remember information presented in sentences (content), they are impaired in the ability to remember the source (voice and face) associated with that sentence. Similar results have been reported in memory for words and their spatial locations (Puglisi, Park, Smith, & Hill, 1985) and for words and the case (upper-or lowercase) in which they appeared at study (Kausler & Puckett, 1980).

This age-related contextual-source memory deficit may explain many aspects of older adults’ memory decline (e.g., Kausler & Puckett, 1980; Naveh-Benjamin & Craik, 1995; Spencer & Raz, 1995). Contextual details, if remembered, can serve as retrieval cues when relevant information must be recalled. Without such cues, memory suffers. Furthermore, the contextual-source deficit suggestion is in line with findings discussed earlier in this chapter. For example, episodic memory and tests of recall depend heavily on contextual information and are more impacted by age than are semantic memory and recognition memory, which rely less on context.

Memory for Items Versus Associations. Another distinction related to episodic memory involves the difference between item and associative memory (e.g., Naveh-Benjamin, 2000; Naveh-Benjamin, Guez, Kilb, & Reedy, 2004; Naveh-Benjamin, Hussain, Guez, & Bar-On, 2003). Item information may involve a single contextual feature, such as remembering that a certain voice has been heard before, or basic content information, such as recalling a word. Associative information, on the other hand, involves the binding together of two or more items (e.g., a name and a face) or contexts, or an item and its context, such as remembering that a specific word was heard in a particular voice.

It has been shown that age-related memory deficits are much larger when associative information is tested than when item information is tested. For example, Naveh-Benjamin (2000; Experiment 2) presented younger and older adults with a series of unrelated pairs of words, with instructions to remember either the words or their pairings. All participants then took both item and associative recognition tests. In the item test, they indicated whether they had seen a given word during study. In the associative test, they were presented with word pairs as they had appeared at study and with pairs consisting of two words taken from different studied pairs, and indicated which pairs were the same as at study. Results revealed that, in each of the study instruction conditions, age differences in performance were larger on the associative than on the item test.

Older adults’ particular difficulty in binding components of an episode into a cohesive unit has led to the suggestion that this age-related associative deficit can partially explain overall age differences in memory (Chalfonte & Johnson, 1996; K. J. Mitchell, Johnson, Raye, Mather, & D’Esposito, 2000; Naveh-Benjamin, 2000). Naveh-Benjamin termed this idea the associative-deficit hypothesis. Such an age-related impairment in the ability to associate or bind together units of information can accommodate several of the results reviewed earlier. For example, semantic memory does not require the specific binding of information to a place and time and should therefore be less affected by aging than is episodic memory; this is in alignment with results discussed earlier. It is interesting to note that the associative-deficit hypothesis may partially reflect similar age-related deficits as the recollection- and context-deficit hypotheses discussed previously in this chapter, in that all three ideas attribute older adults’ episodic memory impairment to the inefficient episodic encoding and retrieval of detailed bound pieces of information.

Autobiographical Memory. Episodic memory can also be assessed by asking people of different ages to remember events in their personal pasts. In contrast to events presented for study in the laboratory, researchers cannot verify the accuracy of these memories. Nevertheless, the study of autobiographical memory can indicate a great deal about the aging process. Research on this topic indicates that the age at which memories are established is quite important. Not surprisingly, people tend to remember recent events. More interesting is that there seems to be an increase in memory for events that happened between the ages of 10 and 30—an effect known as the reminiscence bump. This bump is quite robust and has been found using various stimuli as cues. For example, in a study by Willander and Larsson (2006), older adults between the ages of 65 and 80 years were presented with retrieval cues that could take the form of odors, words, or pictures, and were asked to retrieve an autobiographical memory evoked by each cue. Results showed a reminiscence bump in the second decade of life in response to picture cues and word cues. There was also an odor-evoked reminiscence bump, but this occurred during the first decade of life.

One explanation for the reminiscence bump, proposed by Holmes and Conway (1999) is that privileged encoding occurs for events that are especially important within one’s state of development. For example, identification with society occurs during childhood, whereas a focus on personal relationships occurs during adolescence and early adulthood. In support of this idea, the researchers found that the reminiscence bump in reported public events occurred during the first decade of life, whereas that in reported personal events occurred during the second decade.

False Memory. Another line of research focuses on the memory errors that people commit. A review article by Jacoby and Rhodes (2006) concludes that older adults are more susceptible than younger adults to misinformation— that is, to incorporating new materials into their memory for an original event—although, it is interesting to note, they are more confident than younger adults in the accuracy of these false memories. Karpel, Hoyer, and Toglia (2001), for example, showed that older adults were more likely than younger adults to incorporate misinformation, presented in a questionnaire, into their memories for a crime viewed in slides, yet were highly confident in these false memories.

False memories are not always externally generated but may be internally generated as well. Watson, Balota, and Sergent-Marshall (2001), for example, asked older and younger adults to learn lists of words presented on a screen. Each list consisted of words that were associated— either semantically or phonetically—with an unpresented critical lure. After the presentation of a list, participants were asked to freely recall all the words they could remember from that list. Results showed that older adults falsely recalled more critical lures than the younger group in each relatedness condition. This study aligns with and extends findings based on the Deese-Roediger-McDermott paradigm (Deese, 1959; Roediger & McDermott, 1995), which elicits false memories of critical lures that are semantically related to a given list of words.

One suggested reason for the increase in false memories by older adults (e.g., Smith, Lozito, & Bayen, 2005) involves the inability to link content information to its context. This may lead older adults to experience difficulty in remembering whether a familiar item was externally presented or internally generated (Hashtroudi, Johnson, Vnek, & Ferguson, 1994), or in remembering the sources of conflicting pieces of information. Such a notion is in alignment with the recollection deficit, contextual-deficit, and associative-deficit hypotheses discussed earlier in this chapter.

Emotional Materials. Some recent studies indicate that older adults tend to remember positively valenced information quite well; this is true for both working memory (e.g., Mikels, Larkin, Reuter-Lorenz, & Carstensen, 2005) and long-term episodic memory (see Carstensen, Mikels, & Mather, 2006, for a review). For example, Charles, Mather, and Carstensen (2003) asked young, middle-aged, and older adults to view a series of 32 images— some neutral in valence and others positive or negative. When later asked to recall the images, middle-aged and older adults remembered more positive than negative images, whereas the young adults reported equal numbers of images from those two valences. On a recognition test, younger adults performed better on the negative stimuli than on the positive or neutral stimuli, whereas the middle-aged and older adults performed similarly across all three valence levels. Thus, age-related memory deficits seem to be reduced when positive stimuli are involved.

One suggestion raised to explain this effect (e.g., Carstensen, Isaacowitz, & Charles, 1999) is that, as people approach the end of life, emotional goals—as opposed to knowledge-based goals—become increasingly salient. This, in turn, leads to successful encoding and retrieval of emotional—especially positive— information in older adults.

Explaining Age-Related Memory Changes Using General Theoretical Approaches

Several theoretical frameworks and hypotheses have been suggested to explain empirical findings regarding cognitive aging in general. In this section, we discuss briefly how these ideas may apply to age-related changes in memory.

Speed of Processing

Several researchers have proposed that the execution of mental processes slows down in old age (e.g., Birren, 1965). Salthouse (1996), for example, holds that older adults’ slowing of processing speed is a major factor involved in general cognitive decline. According to his processing-speed theory, there are two separate mechanisms contributing to the speed-cognition relationship. The limited time mechanism involves slowing at an early stage of basic information processing, which leaves less available time for later processing. The simultaneity mechanism involves the notion that products formed early in processing are lost before later processing is finished. When multiplying two 2-digit numbers together, for example, slow performance of late operations may lead to the forgetting of products found earlier.

Salthouse (1996) analyzed the outcomes of several studies in order to show the relationship between speed and various measures of memory performance, including free recall, associative learning, and working memory. After controlling for reaction-time speed, age-related variance on the memory tasks was reduced by over 75%. Furthermore, Verhaeghen and Salthouse (1997) conducted a meta-analysis of cross-sectional studies and found that speed can account for over 70% of age-related variance in episodic memory measures. Thus, processing speed seems to be a key factor involved in age-related memory impairment.

Working Memory Capacity

Several researchers (e.g., Hasher & Zacks, 1988; Welford, 1980) claim that reduced WM capacity is a major factor in the age-related declines in many cognitive tasks, including memory tasks. To effectively encode information, such as spoken sentences, one must use WM to store previously learned information while simultaneously processing and integrating new information into a cohesive meaningful event (Baddeley, 1986). According to one suggestion, older adults do not possess the control processes necessary to switch between processing and storage tasks (e.g., Light & Albertson, 1988). For example, Hogan, Kelly, and Craik (2006) presented older and younger adults with a set of words, with instructions to report either the color of each word or whether the word described a living or nonliving thing. After six to nine words had appeared with the same instructions, the instructions changed to the other task. Results showed that older adults were more impaired than younger adults by this task switch, both in terms of accuracy and response time to the encoding task.

Other executive processes involved in WM may also help to account for episodic memory decline in older adults. Hasher and Zacks (1988) and Hasher, Zacks, and May (1999) have claimed that older adults have trouble using inhibitory processes to block irrelevant information from entering WM. For example, Hartman and Hasher (1991) found that older adults tend to hold disconfirmed information in WM to a greater extent than do young adults. This notion extends to long-term episodic memory as well; Salthouse, Siedlecki, and Krueger (2006) found that older adults were quite impaired in recall of items they had been instructed to remember but were just as able as younger participants to recall items they had been told to forget.

Attentional Resources

There have been some suggestions that a decline in attentional resources—the limited pool of resources available for processing—is reduced in old age and that this results in age-related changes in a variety of cognitive functions (Craik, 1986; Craik & Byrd, 1982). According to this notion, older adults lack the attentional resources required to effectively perform certain tasks—especially difficult tasks, which are highly demanding of resources. Evidence is consistent with this idea (e.g., Craik, 1986; Craik & Byrd, 1982; Craik & McDowd, 1987).

One relevant line of research involves the effects of divided attention (DA) on memory. There have been several studies using memory tasks that indicate a larger reduction in overall performance of older than younger adults when under DA conditions (e.g., Anderson, Craik, & Naveh-Benjamin, 1998). Furthermore, studies have shown that young people whose attentional resources are reduced through use of DA conditions exhibit a pattern of memory performance decline similar to that of older adults (e.g., Craik & Byrd, 1982; Jennings & Jacoby, 1993), suggesting that older adults’ resources are in fact depleted even without a secondary task.

This suggestion of a possible mediating role of reduced attentional resources is in line with the age-related patterns of memory performance described earlier. For example, explicit memory requires more attentional resources than does implicit memory, and the age-related impairment is larger on explicit tasks. Likewise, the larger effect of age on recall than on recognition measures could be due to the substantial amount of resources needed to search for a target stimulus. Thus, although there have been criticisms of the reduced attentional resources notion on the grounds of vagueness (e.g., Salthouse, 1988), it provides a heuristic functional explanation of age-related differences in memory performance.

Future Directions

The empirical evidence reviewed in this chapter points to interesting differential effects of aging on memory. In the following section, we discuss some issues reflecting ways in which these findings can be applied. For example, researchers should actively search for factors that positively and negatively affect older adults’ memory performance in order to develop practical measures to enhance memory functioning in old age. In this search, various lines of research must be unified in an effort to determine the causes of age-related memory loss and ways to avoid such loss.

Maximizing Older Adults’ Episodic Memory Strengths

One issue that should be pursued in future research deals with the factors that reduce older adults’ declines in memory performance. One area of study showing some positive effects of age involves prospective memory (PM; i.e., the ability to remember to perform a future action; see chap. 10 for an in-depth discussion of PM and aging). Although older adults are impaired in laboratory-based PM tasks, they seem to outperform younger adults on naturalistic PM tasks, such as remembering to make phone calls or to take medications, even when participants are asked not to use external aids (e.g., Henry, MacLeod, Phillips, & Crawford, 2004). Perhaps future researchers will be able to determine the precise mechanisms behind this effect and to help older adults apply those mechanisms to retrospective memory tasks. Seniors might be trained, for example, to make new information highly personally relevant, as with naturalistic PM tasks, to see if this aids retrospective memory performance.

Another area of memory in which older adults are relatively unimpaired is semantic memory, as discussed earlier (e.g., Park et al., 2002). An important question is whether they can use this semantic knowledge to improve episodic memory performance. There is some evidence that this is the case. For example, a meta-analysis by Verhaeghen et al. (1993) provides an indication that increasing categorizability (the ability to group information into previously learned semantic categories) of episodically to-be-remembered information leads to a decrease in age differences. Furthermore, Naveh-Benjamin (2000) found larger age differences in memory for unrelated than for related word pairs, suggesting that older adults can use previously learned semantic information to support new episodic knowledge, thus reducing age-related deficits in episodic memory (see also Naveh-Benjamin, Craik, Guez, & Kreuger, 2005). Thus, various factors rather than purely mechanistic ones can affect older adults’ memory performance. By focusing not just on memory impairment but on factors positively related to memory and aging, future research may discover ways in which the findings can be practically applied.

Minimizing Negative Influences on Older Adults’ Episodic Memory Performance

One factor associated with older adults’ memory impairment involves stereotype threat; that is, older adults may not perform at their best because they are aware of (and threatened by) the common notion that aging impairs memory abilities. Hess, Auman, Colcombe, and Rahhal (2003), for example, examined free-recall performance of older and younger adults who had just read articles indicating either positive or negative effects of aging on memory. Whereas the younger adults were unaffected by the stereotype threat manipulation, the older adults were quite impaired in the negative, relative to the positive stereotype condition. Manipulations presumably eliciting stereotype threat, or the related feelings of anxiety, may also be surprisingly subtle. For example, Rahhal, Hasher, and Colcombe (2001) found that age differences in memory were eliminated when the terms memory and testing were completely omitted from task instructions.

Although without further assumptions, anxiety and stereotype threat may not be able to explain interaction effects (i.e., cases in which age deficits are larger on one episodic task than on another), these factors can increase the overall differences observed between young and old. Thus, one practical way to reduce age-related episodic memory impairment in the laboratory is to take measures to reduce older adults’ anxiety and to alleviate stereotype threat. Such measures should be extended to naturalistic settings, to aid older adults’ memory performance in their everyday lives.

The above sections provide a sample from the wealth of research investigating some of the factors that are positively and negatively associated with older adults’ memory functioning, but more work is needed to shape the results into a more applicable form. Future researchers should integrate the results of this basic research into effective interventions for older adults.

Effects of Training on Older Adults’ Episodic Memory Performance

One especially applicable area of research involves the effects of special training on older adults’ memory performance. Results of past studies seem to be quite promising. Verhaeghen, Marcoen, and Goossens (1992), for example, conducted a meta-analysis of studies comparing pre-and postmnemonic training performance of older adults. Groups receiving this training improved significantly more than those not receiving training (d = 0.73 for trained groups, compared to ds = 0.38 and 0.37 for control and placebo groups, respectively); all types of mnemonic training (method of loci, name-face, and organization) were equally effective. The authors concluded that “even in old age memory remains plastic” (Verhaeghen et al., 1992, p. 248).

An example of an empirical study involving memory training was conducted by Cavallini, Pagnin, and Vecchi (2003), who compared pre-and posttraining performance of young (20-35 years), younger elderly (60-70 years), and older elderly (70-80 years) participants. First, participants completed a test battery, which included various ecological tasks, such as recalling a short story after reading it, studying and then recalling items from a shopping list, and remembering names paired with faces. Next, each participant underwent one of two types of memory training, which were completed in five separate sessions. Some participants received training in the loci mnemonic, associating images along a familiar route with to-be-remembered items. The remaining participants underwent “strategic training,” in which they learned a variety of memory techniques and were taught to choose the method most appropriate for a given task. After finishing all training sessions, the participants completed a test battery similar to the one from the pretraining session. Results showed that the types of training had similar effects on memory performance. Importantly, all age groups improved to a similar extent. On one task—recall of stories— trained older adults were actually able to improve their performance to the levels of untrained young adults.

There is also evidence that older adults maintain a benefit of memory training over long periods. For example, Derwinger, Neely, and Bckman (2005) found that, 8 months after being trained to generate memory strategies, older adults maintained a boost in performance. There are some indications, however, that although older adults benefit greatly from intense mnemonic training, their “developmental reserve capacity,” found through testing the upper limits of training benefits, is relatively small (see, e.g., Baltes & Kliegl, 1992; Kliegl, Smith, & Baltes, 1989). Thus, memory training may be one highly useful tool through which researchers and practitioners can work to improve older adults’ memories, but training is unlikely to completely remove age differences. Training that is used should teach older adults how to create and evaluate their own memory strategies, because this creates lasting benefits (Derwinger et al., 2005)

Integration of Research: Combining Perspectives

To create useful interventions as mentioned above, various lines of research must be brought together. In a narrow sense, this involves the integration of findings regarding aging’s effects on various specific aspects of memory. This has already been done to some extent; for example, researchers have noted that aging negatively impacts episodic memory but does not greatly affect semantic memory (e.g., Park et al., 2002). Such comparisons can help to determine the precise mechanisms behind age-related memory impairment.

In more broad terms, researchers should also strive to establish causal links between various factors, such as health issues, and memory performance. Work on this task, too, has already begun. Factors such as physical exercise, diet, and stress levels have been suggested to influence the degree to which memory declines with age (e.g., Small, 2002). For example, data from the Nurses’ Health Study showed that self-reported physical activity was positively associated with performance on immediate and delayed recall of 10-word lists and East Boston Memory Tests (Weuve et al., 2004) in women between 70 and 81 years of age. It is also important for researchers to consider disease as a potential contributor to memory decline. Diabetes, for example, appears to impair performance on a variety of memory tasks (Bent, Rabbitt, & Metcalfe, 2000). By considering physiologically related changes and their relationship to memory performance, it may be possible to determine new ways to reduce age-related memory deficits through simple lifestyle changes (see chap. 16 for more on the relationship between health and cognition). Although this is a monumental task, future researchers should seek to integrate information provided from diverse areas of study.

The Relationship Between Age-Related Memory Changes and Cognitive Changes in General

The preceding few paragraphs provide a brief review of approaches needed to demonstrate the effects of various factors on memory and aging. We would now like to examine in greater detail the relationship between age-related changes in memory and in general cognition. The question is whether there is uniqueness in changes that occur in memory as people age or whether these changes are just a manifestation of more general changes in cognition.

Certain statistical approaches are quite valuable in the assessment of this question, especially because relevant studies tend to involve correlation rather than manipulation. Salthouse, Berish, and Miles (2002) assessed performance of different age groups on several variables, some related to memory and others to other cognitive tasks. After statistically controlling the variance in some non-memory-related tasks, the proportion of age-related variance in free-recall memory performance was greatly reduced, implying that age-related effects on memory measures and on other cognitive variables are not independent of each other. Similar results have been obtained for source memory and other episodic memory measures (e.g., Salthouse et al., 2006; Siedlecki, Salthouse, & Berish, 2005).

The above results raise interesting questions regarding different research approaches in the study of age-related changes in memory. The main approach discussed in this chapter involves the effects of various experimental manipulations on memory performance and reveals a variety of differential effects of those manipulations on the memory performance of younger and older adults. The other approach discussed here looks simultaneously at relationships between age and several memory and cognitive indexes, and often shows, as in the examples above, that the effects of age on memory are not unique but are shared with other cognitive factors. If this is the case, then it is crucial for future researchers to integrate findings from these experimental and psychometric approaches. This would provide a better understanding of the absolute age-related changes in different tasks as well as the degree to which these changes are independent from those in other areas.

The issues raised in this section regarding the applicability of memory research for older adults, as well as the integration of various approaches and statistical methods used in this research, should be further investigated. Consideration of these issues and approaches will help researchers to better understand the phenomena involved in the effects of aging on memory processes. Finally, although the empirical work described throughout this chapter has proven valuable in the search for mechanisms behind age-related memory change, there is clearly a need for greater theoretical development (see, e.g., chap. 2, this volume). Such advancement may be generated, in part, through the integration of various lines of research.


In this chapter, we first reviewed empirical evidence for differential patterns of age-related declines in memory. Certain types of tasks, especially those involving the encoding, retention, and explicit retrieval of detailed, bound episodic information, seem to be quite impaired in old age. In contrast, implicit and semantic memory processes remain relatively intact. We also described several theoretical approaches to the study of general cognitive change and showed how each of these perspectives can be applied more specifically to age-related changes in memory performance. Finally, we discussed directions for future research on memory and aging, including the practical application of findings involving factors that positively and negatively affect older adults’ episodic memory performance, as well as the integration of diverse lines of research. Such directions should allow researchers to progress in the understanding of developmental changes that occur in memory.