Memory

Another view of memory that has both behavioral and neurological support considers memory not as a unitary construct but as a collection of component systems. The general view is that memory consists of sensory memory, short-term memory, and long-term memory. Short-term memory is further divided into primary memory and working memory, and long-term memory is divided into episodic memory, semantic memory, and procedural memory. These memory systems differ in the nature of how memories are represented and in how these representations are maintained and retrieved. They also differ considerably in how they are affected by adult aging.

Sensory memory. After an event is experienced, it is first represented very briefly in the sensory system. Here the information is represented as it is processed and analyzed by the attentional and perceptual systems. There has been very little work on this type of memory and aging, but some research does suggest that older adults are less efficient in this early type of processing, especially in the visual system. It should be pointed out, however, that differences in sensory-perceptual processing would be an unlikely explanatory construct for memory differences found in later memory systems because later memory differences vary according to what system is examined (Craik). Adequate perceptual processing of the to-be-remembered stimulus would seem to be a requirement for all types of long-term memory, and the fact that some long-term memory systems are minimally affected by aging while others show large effects would not support an explanation that relied solely on faulty processing at the very early stage of processing.

Primary memory. Primary memory refers to the number of items that can be represented in the mind at one time. Primary memory is typically measured by digit- or word-span tasks. For example, when strings of numbers are presented one at a time at one-second intervals, primary memory would be the number of items that could be repeated back without errors. The digit-span test found on most intelligence tests is a measure of primary memory. When primary memory is tested by digit-span tests with individuals of different ages, no reliable age differences are found (Craik). Another measure of primary memory is recall of the last few items presented on a word list. Again, research finds no age differences in the ability to recall the most recently presented items in a word list.

Working memory. Age differences are found, however, on a measure of short-term memory called working memory, which takes into account both the ability to keep things in mind and, at the same time, the ability to process information. Working memory, unlike primary memory, requires individuals to keep information in mind while engaging in another processing task. One commonly used working memory task is reading span. Individuals read a series of short sentences (e.g., "The girl hit the ball.") and answer questions about the sentence (e.g., "Who hit?"). At the same time, individuals have to remember the last word in each sentence (e.g., "ball") and report the words when told to do so. The number of sentences presented varies, and reading span is the greatest number of words that can be remembered without making an error. Because working memory tasks require simultaneous storage (remembering the words) and processing (reading the sentences and answering the questions), it is a better simulation of everyday information processing.

Working memory is used to understand a conversation or to write an e-mail message on the computer. Other working memory tasks have been developed that involve different kinds of information processing, such as arithmetic calculations (computational span) and spatial manipulations (spatial working memory). Regardless of the type of information involved, however, large, reliable age differences are found on working memory tasks (Zacks et al.). Therefore, while the more passive primary memory tasks, such as simple digit span, do not show age differences, differences are found with working memory measures. As will be discussed later, working memory capacity is considered by many researchers to be a fundamental mechanism for more complex memory processing.

Long-term memories are not kept in conscious awareness, as short-term memories are, but instead have to be retrieved into consciousness when they are needed.

Episodic memory. Episodic memories are recollections that are actively retrieved as previous personal experiences. For this reason, episodic memory is sometimes called autobiographical memory. The memory is a reconstruction of an earlier experience very much like looking something up in an internal cognitive diary. For this reason, contextual information about when and how an event was originally experienced is often used to guide retrieval. "What did I have for dinner last night?" "Where did I park my car?" "Did I take my medicine this morning?" "Did you see Joan at the party last night? To answer each question, one attempts to reconstruct the original event.

Older adults do worse on most episodic memory tasks than do younger adults (see Figure 1). In the laboratory, typically a list of words or some other to-be-remembered information is presented to individuals, and later, after a retention interval, memory is tested. Because the items presented are commonly used words, they are already known to the individuals. The memory task, therefore, is to remember the words in the particular context of the original list. Episodic memory requires one to put what one is trying to remember in a specific context. Even though the magnitude of age differences varies considerably among different memory tasks using different materials and types of tests, older adults tend to have greater problems with episodic remembering than do younger adults.

As will be discussed later, the variable that seems important in determining the magnitude of age differences is the degree to which the memory task involves deliberate processing by the individual. At both encoding and retrieval, the more intentional processing required to perform the task, the larger the age differences that will be found with that task. For example, free recall of a word list requires more deliberate processing than recognition tasks, and age differences are larger on free recall (Craik). Instead of trying to actively recall the items, individuals in a recognition task only have to pick out the words they saw earlier on a longer list of words.

Because episodic memory is so dependent on contextual information, one of the reasons for older adults' poorer performance on these tasks is probably their inability to encode and remember contextual information easily. Older adults, for example, do not do as well as younger adults in identifying the way in which information was presented to them (i.e., source memory). They do worse when asked to remember whether a word was presented in uppercase or lowercase, spoken by a male or a female voice, in one color versus another, or in the upper part of a computer screen or the lower part (Zacks et al.). These tasks require individuals to remember contextual detail. Because older adults encode less contextual detail, they do not do as well on tasks in which contextual detail provides the cues for retrieval. In fact, because older adults encode less context, they have problems distinguishing events they actually experience from those they have only thought about, a phenomenon called "reality monitoring." In a reality-monitoring task, individuals either read words at encoding or generate words in response to some cue. Older individuals have greater problems in determining whether remembered events were the ones read or the ones imagined in response to the cue (Norman and Schacter).

In addition to the problems associated with remembering context itself, older adults have problems with binding the context with the to-be-remembered information. In one study, for example, younger and older adults were presented with pairs of words and asked to generate a sentence that included both words. There were no age differences in the nature of sentences generated, but older adults did have more problems generating sentences, especially for unrelated word pairs that required them to generate the binding sentence through deliberate processing (Smith et al.). Older adults were better able to generate sentences when the two words were related to one other. They were also better able to recall one word from the word pair when given the other word as a cue. By having related word pairs, there was less need to bind the two words together because their relationship provided an existing bond. Again, because less deliberate processing was required in both the encoding and the retrieval conditions when related pairs were used, age differences were smaller. In fact, research has shown that age differences in the ability to recall a target picture when another picture is given as a cue depends on the relationship of the cue to the target. Age differences are large when the two pictures are unrelated, but smaller when the two pictures are either semantically related or presented as perceptually interacting (Park, Smith, et al. 1990).

One interesting type of episodic memory, "prospective memory," does not involve remembering something from the past, but instead involves intending to do something in the future. "Stop by the store when you leave work, and bring home some milk." "When you see Wanda, tell her to look at my new Web page." "Take two of these pills every other day after lunch." These are examples of prospective memory tasks. In the laboratory, prospective memory tasks simulate these real-world examples (e.g., "Press the key when you see a word with an 'R' in it" or "Press the key every ten minutes"). The prospective task is combined with some other cognitive task, such as trying to study a word list for a later memory test. With simple laboratory tasks, however, such as pressing a key when a certain letter is found in a word, age differences in prospective memory are often not found. Again, the determinant of whether age differences are found seems to be the degree of deliberate recollection required to perform the task, either for the prospective task or for the background task. As the difficulty of either task is increased, requiring more deliberate processing, age differences are increased.

Age differences are often larger in time-based tasks (e.g., individuals are asked to press a computer key every ten minutes while performing another computer cognitive task) than in event-based tasks (e.g., individuals are asked to press the key when a certain cue word appears). Because event-based prospective remembering involves less deliberate processing, given the external cue, age differences sometimes were not found. It is also clear that prospective memory errors in older adults increase when the background task they have to perform becomes more demanding in terms of processing resources (Einstein et al.).

Semantic memory. Not all remembering requires one to reconstruct the experience of encoding. There are many examples of remembering without reference to how or when what one is trying to remember was originally learned. There is access to a great deal of knowledge that has lost all connection to the context of its original episodic learning. "What is the capital of North Dakota?" "What bug has eight legs and weaves webs?" "Where were you born?" Answering these questions requires semantic memory. Semantic memories are retrieved conceptually rather than contextually, and represent accumulated knowledge. Instead of using a cognitive diary, semantic memory is like looking something up in an internal cognitive encyclopedia. Of course, the information is not alphabetically organized, but instead organized conceptually or semantically.

As mentioned earlier, tests of semantic memory typically show either no age differences or improvement over the adult life span (see Figure 1). Vocabulary tests and tests of general knowledge (such as found in Trivial Pursuit games) show either no age differences or increases for older adults up until very late in life (eighties or nineties). In Figure 1, vocabulary knowledge increased steadily through the seventies and showed only a slight decline in the eighty-year-old group.

There have also been attempts to examine the architecture of semantic memory (i.e., the way semantic organization is conceptually associated). Free association tests are one way to look at semantic organization. Individuals are given a word and asked to generate another word, the first word that comes to mind when thinking about the word given. A category name is given in another type of test (e.g., vehicle), and individuals are asked to generate the first five vehicles they can think of. If semantic information in memory is organized in different ways by different age groups, then there should be qualitative differences in the nature of the responses given on free association tests or category generation tests. If older adults' semantic memories are organized differently, then the organization should produce differences in the strength of associations between different concepts. Norms of free associations, however, as well as norms of generating instances in categories, show no differences between age groups (Smith and Earles).

One aspect of semantic memory that does seem to decline with aging is the ability to find a word, given its definition. This phenomenon extends to finding proper names and to the "tip-of-the-tongue" phenomenon (Craik). A tip-of-the-tongue state is created when a person knows that he or she knows something but cannot think of it at the moment. Often one can generate information about the answer that is correct but cannot think of the answer itself. Some of this effect (word finding, name finding, tip-of-the-tongue), however, is associated with older adults simply being slower to respond. Some research suggests equivalent word finding in different age groups with difficult words (Craik). Other research suggests that older adults eventually can resolve tip-of-the-tongue states when given enough time (MacKay and Abrams).

Procedural memory. Several times in this entry it has been stated that age differences in memory seem to increase when the degree of deliberate processing required to remember increases. Procedural memories are assumed not to require any deliberate, intentional processing at all. They instead involve only automatic processing and, in fact, do not even require conscious awareness of the effects.

Procedural memory tasks, sometimes called implicit memory tasks, often use repetition priming as a measure. For example, individuals first examine a list of words (stand, chair, . . . radio), not aware that a later memory test is involved. Rather, they are told to make some judgment about the words, such as to rate the pleasantness of each word on a five-point scale. Then several other word tasks are performed. Finally, a series of word stems is presented (fe___, ra___, . . . bl___) and the individuals are asked to complete the stem with the first word that comes to mind. Some of the word stems could be completed with a word presented earlier on the list. Even though they are not aware that the word stem list contains stems for words seen earlier, individuals will use the presented words to complete the stems at a level greater than chance. This increase in "remembering" previously presented items represents implicit or procedural memory.

As might be expected, given the lack of deliberate recollection, age differences are typically not seen on implicit memory tasks (Zacks et al.). If they are found, they are very small, especially when compared with the large effects seen in explicit episodic recall.

In summary, the memory systems approach has been very useful in describing when significant age differences in memory are found. Whether or not memory systems eventually are supported by neuroscientific data, they have provided a useful conceptual framework for organizing memory phenomena and for showing dissociations with age.