Ageing memory: ageing memories
DOI:10.1093/acprof:oso/9780199228768.003.0007
Abstract and Keywords
This chapter examimes developments in the study of ageing memory during the past fifty years. It suggests that the field of cognitive ageing was significantly influence by Donald Broadbent's Perception and Communication and Alan Welford's Ageing and Human Skill. These books caused a shift from stimulus-response behaviourism and this paradigm shift had a strong influence on studies of age-related changes in human performance.
Keywords: ageing memory, Donald Broadbent, Perception and Communication, Alan Welford, Ageing and Human Skill, stimulus-response behaviourism, cognitive ageing
In 2008, fifty years will have passed since the publication of two books that had a major impact on the field of cognitive ageing: Broadbent's Perception and Communication (1958) and Welford's Ageing and Human Skill (1958). Broadbent's classic book was a main source of the shift from S-R behaviourism to information-processing approaches. This paradigm shift had a strong influence on studies of age-related changes in human performance, as the early work in this area was carried out by mainstream experimental psychologists who simply added individual differences in ageing to their studies. Welford's book is a classic in the narrower field of ageing research, reporting a systematic programme of laboratory and field studies carried out in Cambridge between 1946 and 1956. The many findings are grouped under such headings as skill, speed and accuracy, translation processes, perception, memory, and problem-solving; the findings are also linked by reference to underlying theoretical concepts and how these concepts change in the course of ageing. Welford's book is quite theoretical in fact, and in that sense stands in contrast to the mostly descriptive studies of age-related differences in memory and cognition of the time. I will survey that empirical scene after saying some words about my own entry to the field.
A personal perspective
To my mild embarrassment, the year 2008 also marks the 50-year anniversary of my own involvement in experimental psychology. I completed my undergraduate degree at the University of Edinburgh in 1960, having taken courses in memory from Ian Hunter and in perception from James Drever. My first experiment (looked back on as fondly as a first love!) was on time perception. I was very impressed by George Miller's ‘Magical number 7’ paper (Miller 1956), and so measured information-processing rates in bits per second while subjects made absolute perceptual judgements involving various degrees of (p.74) complexity for about 12 minutes, and were then unexpectedly asked to estimate how long the task had taken. I obtained a nice function relating rate of processing to subjective time estimates, although I never got around to publishing the data. But I did keep remembering the result, to the point that Janine Hay and I conducted and published a more sophisticated version of it with younger and older adults many years later (Craik and Hay 1999).
After Edinburgh I joined the Medical Research Council (MRC) Unit on Occupational Aspects of Ageing, attached to the Psychology Department at the University of Liverpool, established to carry on Welford's useful work on ageing, but in a more applied industrial setting. The Director, Alastair Heron, asked me to explore the idea that older adults' performance may be limited by lack of confidence and feelings of inadequacy. It's an interesting but complex problem, undoubtedly involving personality, developmental, social, and contextual factors. The last aspect was made obvious to me while testing a very anxious and timid older woman who, when tea was served during a break, immediately took command of the situation and ordered everyone else about, only to revert to her previous ‘personality’ when testing resumed! After inconclusive experiments using Rotter's ‘level of aspiration’ framework, I discovered signal detection theory (Swets et al. 1961) and investigated age-related differences in caution-riskiness by examining criterion placement (β) in a tone-detection paradigm. The results suggested that older adults require stronger evidence of a signal before committing themselves to a response, a result that was in line with an influential study by Botwinick and colleagues (1958). These authors had subjects judge which of two simultaneously exposed lines was the shorter; as the difference in length between the lines was reduced, decision latencies increased—substantially more so for older than for younger adults. This result was obtained when the lines were exposed for 2 seconds, but when exposure time was shortened to 0.15 seconds the age difference in latencies greatly diminished. It thus seemed that older adults preferred to take more time before committing themselves to a decision, but that this extra time was not necessary.
Being a member of an MRC unit enabled me to make contact with the MRC Applied Psychology Unit (APU) in Cambridge, directed by Donald Broadbent. I visited the Cambridge unit on several occasions, making the acquaintance of such youthful researchers as Alan Baddeley, John Morton, and Patrick Rabbitt, as well as the demi-gods Broadbent and Conrad! In their dichotic listening studies, Broadbent and Gregory (1963) had shown that diversion of attention away from a stimulus reflects a decrease in d′—a reduction in effective signal strength. Inglis and Caird (1963) had carried out dichotic listening experiments on younger and older adults, with the main finding that older adults (p.75) showed a substantial age-related loss in recall of items from the second ear reported, but with no loss in first ear recall, concluding that ageing increases memory loss from simultaneous interference, in this case recall of the first half-set. This was in line with Welford's (1958) earlier suggestion that an age-related decrement in short-term memory, attributable to simultaneous interference, is a major cause of age-related declines in learning, memory, and problem-solving. Broadbent and Gregory's result made me wonder whether the reduction in d′ associated with divided attention might not be greater in older adults; in other words, whether the Inglis and Caird result might reflect an attentional as opposed to a memory problem.
This thought led me to design a dual-task study in which one task was auditory signal detection and the other was short-term recognition memory. Both tasks were presented auditorily, but from different loudspeakers. The memory task consisted of seven successive digit pairs followed immediately by a probe pair; the subject had to decide whether the probe was one of the preceding pairs. The signal detection task consisted of 10 seconds of white noise overlapping the digit-pair presentation; the signal was a 1000-Hz tone lasting for 1 second and occurring at an unpredictable time during the noise. The signal strength was adjusted for each subject to yield 75% detectability, and the signal was present on 50% of the trials, independently of the presence or absence of the memory probe. Thus, on each trial the subject decided yes or no with regard to both tasks, yielding a d′ value for each task. The tasks were performed both separately (full attention) and together (divided attention), and we compared groups of young, middle-aged, and older subjects.
One reason for running the study was to see whether divided attention (DA) was particularly detrimental to performance in older adults, but there was a further interest in examining the hypothesis with signal detection measures. Taylor and colleagues (1967) suggested that d′2 may be considered a measure of processing capacity, and that when two or more discrimination tasks are performed simultaneously the sum of individual values of d′2 should add to a constant total. More accurately, for tasks T1 and T2, performed either alone or together, Taylor et al. suggested that
The results of the signal detection/memory study are shown in Table 6.1A. Neither the memory task nor the detection task showed age differences when carried out alone, but the Σd′2 values showed a systematic drop with age. The fact that Σd′2 values are greater than 1.0 suggested that these particular tasks did not require extensive cognitive analysis—this would be particularly true of the tone detection task which (following Treisman 1964, 1969) could presumably be carried out by relatively peripheral analysing mechanisms. Accordingly, I carried out a more ‘cognitive’ version of the experiment in which one task was
Table 6.1 Values of d′ and 2d′2 for young, middle-aged, and older adults in two dual-task experiments (Craik 1973)
Experiment 1
|
Detection
|
Memory
|
Σd′2
|
|---|---|---|---|
Young
| |||
Together
|
1.35
|
1.08
|
1.58
|
Alone
|
1.39
|
1.35
| |
Middle-aged
| |||
Together
|
1.02
|
1.28
|
1.39
|
Alone
|
1.82
|
1.23
| |
Older
| |||
Together
|
1.22
|
0.94
|
1.05
|
Alone
|
1.59
|
1.38
| |
Experiment 2
|
Letters
|
Digits
|
Σd′2
|
Young
| |||
Together
|
0.62
|
1.14
|
0.42
|
Alone
|
2.38
|
1.93
| |
Middle-aged
| |||
Together
|
0.76
|
1.00
|
0.31
|
Alone
|
2.21
|
2.29
| |
Old
| |||
Together
|
0.28
|
0.56
|
0.11
|
Alone
|
1.70
|
1.91
|
Detection = detection of tone in noise; memory = short-term recognition memory for digit pairs; letters = short-term recognition memory for visual letters; digits = detection of three successive odd digits, presented auditorily
I have described this study at some length as it gives a sense of the British interest in attention and short-term memory in the 1960s and early 1970s. I presented the two experiments shown in Table 6.1 at an American Psychological Association meeting in 1973, in a session devoted to the application of signal detection theory to ageing. The person doing the commentary was a psychophysical fundamentalist unfortunately, and he heavily criticized other speakers for using stimuli that were not pure sinusoids. Turning with some distaste to my paper, he commented, ‘And then there was the person who used verbal stimuli…’—further words failed him! I found this such a traumatic experience that I never followed up on the ideas, although Martin Taylor later comforted me by saying that d’ is just a statistical measure, like t or F.
My own interests in cognitive ageing shifted gradually from studies of attention and dichotic listening to experiments on age differences in primary and secondary memory. The general finding (e.g. Craik 1968) was that the amount recalled from primary memory is not sensitive to ageing, and indeed the same insensitivity is found in amnesia (Baddeley and Warrington 1970). These results make sense from the perspective that primary memory largely reflects recall of items that have been held in mind by means of subvocal rehearsal; they are held in the articulatory loop in terms of the Baddeley and Hitch (1974) working memory model. Recall from secondary memory (or ‘long-term memory’) is sensitive to ageing, however, presumably owing to age-related inefficiencies in both encoding and retrieval. One early observation, (p.78) following from Murdock's (1960) classic work on factors influencing free recall, was that recall (R) is a linear function of list length (L); that is,
I left the Liverpool Unit in 1965 to take a job as lecturer at Birkbeck College, London, where I remained until 1971 when I moved to Toronto. During the late 1960s and early 1970s I focused largely on mainstream studies of memory and attention, as I was no longer required to study ageing. However, in 1973 I was invited by Jim Birren and Warner Schaie to contribute a chapter on age-related differences in memory to the first edition of their Handbook of the Psychology of Aging (1977), and this rekindled my interest in the area. I received some helpful comments on the first draft from the English-Canadian psychologist David Schonfield who, among other points, noted the difficulty that older adults have in ‘remembering to remember’—a notion that grew into the current research topic of prospective memory. It is interesting to reflect that Schonfield was thinking about such issues in the 1970s. I had also asked Alan Welford whether he had some new memory data to include in the review. He replied that he did not, but that he was surprised to see that my address was now Toronto and regretted that I was now apparently ‘lost to Britain’. The guilt that this mild admonishment might have induced was mitigated, however, by the fact that Welford's letter came from Adelaide, South Australia!
Attention, memory, and ageing, 1960–1980
Studies of ageing memory in the 1960s tended to describe some difference in performance between younger and older adults without much consideration of mechanism. Many experiments were carried out in the verbal learning tradition, although mercifully few involved nonsense syllables—perhaps older learners had too much robust good sense to tolerate such pointless exercises! This early work thus dealt more with learning than with memory, in keeping with the dominant paradigm of the times. Many results were interesting and informative. Wimer and Wigdor (1958) found that age differences occurred mainly at acquisition, and that if initial learning was equated retention and rates of forgetting did not differ. There was much interest in the (p.79) conditions of learning. Rate of presentation in the recall phase of paired-associate learning is age sensitive, for example, with older learners especially penalized by fast testing rates (Taub 1967). Hulicka and Grossman (1967) discovered that, whereas older learners could profit from the use of imagery, they did not generate image mediators spontaneously. This echoes similar findings in young children, studied under the label of production deficiency. Experiments from this era are well described in Donald Kausler's comprehensive Experimental Psychology and Human Aging (1982).
The new information-processing approach radically changed topics and paradigms, with a new focus on age differences in the ability to process and retain information. Experiments showed that older adults processed information less rapidly (Welford 1958), were less able to ignore irrelevant information (Rabbitt 1965), and were more likely to concentrate on one channel rather than share processing resources between two channels in a dual-task situation (Broadbent and Heron 1962). Memory comparisons now focused on possible differences in the capacity, coding characteristics, and forgetting functions associated with various memory stores—sensory, short-term and long-term. The answers in brief, and with the benefit of hindsight, are that age differences are slight in both sensory and short-term memory—with the exception of an age-related decrement in dichotic listening, as mentioned previously. The idea of limited capacity did not seem relevant to long-term memory, so researchers shifted focus gradually to the processes of encoding and retrieval. Organizational processes were in vogue at the time, following the pioneering work of George Mandler, Endel Tulving, and others; in this connection Hultsch (1974) showed that older adults were less effective organizers, and that this processing inefficiency was one factor contributing to their poorer long-term memory performance. Other evidence pointed to age-related retrieval deficits, however. Laurence (1967) found that older adults showed a disproportionate benefit from receiving cues at retrieval, and Schonfield and Robertson (1966) reported that, whereas recall of verbal material fell off progressively with increasing age, recognition memory did not—implying that the information is still available in the older memory system, but is not accessible, except with the help of powerful cues. A strong case for an age-related retrieval deficit being the root cause of memory loss in older adulthood was made by Burke and Light (1981).
The 1960s and 1970s were thus characterized by a move away from ‘verbal learning’ paradigms and from its roots in learning theories. Compared with America, British psychology was much less under the thrall of learning theory, so information theory approaches made an earlier start there. Once cognitive theories became established in the 1960s there was a great liberalization of (p.80) ideas and methods, and these gradually trickled down to studies of cognitive ageing. A particularly fresh and insightful programme of studies was carried out by Harry Bahrick and his collaborators (e.g. Bahrick 1979; Bahrick et al. 1975). He assessed very long-term memory for real-life information, such as names and faces of high-school colleagues (recognition holds up well for at least 25 years but recall falls off earlier—c.f. Schonfield and Robertson, 1966), and for knowledge of places and languages learned many years before. There was also a move towards models of memory (e.g. Atkinson and Shiffrin 1968; Baddeley and Hitch 1974; Craik and Lockhart 1972; Crowder and Morton 1969; Waugh and Norman 1965), with the models acting, not so much as representations of ‘the truth’ (although perhaps they were considered to be so at the time!), but more as points of view that suggested particular lines of experimentation. Two other shifts in the general area were first an attempt to specify mechanisms of memory, and second a gradual awakening of interest in the brain and a realization that our behavioural and cognitive notions must ultimately be compatible with brain processes and mechanisms. Two examples of this latter trend are, first, Marcel Kinsbourne's (1980) speculation that interference effects increase with age owing to a reduction in functional distance between cortical representations because of age-related cellular and synaptic loss; competing representations are less easily held apart in the ageing brain. Second, Albert and Kaplan (1980) drew attention to the fact that frontal lobe efficiency declined with age and that this in turn was likely to reduce the efficiency of cognitive control—a very current idea at the present time!
Slowing, inhibition, and resources: cognitive ageing in the 1980s and 1990s
The area of cognitive ageing really took off in the 1980s. New theoretical ideas were proposed and a multitude of experiments conducted to test them. Age-related declines were variously attributed to slowing of cognitive operations (Salthouse 1982, 1996), to a decline in inhibitory processes (Hasher and Zacks 1979, 1988), and to a reduction in available processing resources (Craik and Byrd 1982). Although proponents of these different perspectives argued in favour of their point of view and against alternative positions, the strong likelihood is that all of these theoretical ideas have some merit. We are still awaiting an adequate ‘unified theory’ of cognitive ageing in 2008, however.
My lab in the 1980s was located at Erindale College in the University of Toronto; we worked both on general problems of memory, stemming from the levels of processing ideas suggested by Craik and Lockhart (1972), and on (p.81) problems of memory and ageing, exploring the hypothesis that ageing is associated with a reduction in available processing resources. The concept of ‘processing resources’ was left a little vague (to say the least!), but has a lot in common with both general arousal and with the information-processing notions of selective attention and channel capacity. One line we pursued is that, if reduced attentional resources is a major problem of ageing, the effects of ageing might be mimicked by reducing resources in young adults—by placing them in a dual-task situation, for example (Craik 1982). After a number of successful studies on this theme Moshe Naveh-Benjamin has persuaded me that older adults also have an additional memory problem over and above resource attenuation, namely a deficit in establishing and utilizing associative information (Naveh-Benjamin 2000, 2002). Further, the notion that resources must be allocated efficiently between tasks is obviously similar to the concept of cognitive control, and my friend Larry Jacoby has patiently told me over the years that ‘control’ is really the relevant variable. Perhaps, but back in the 1980s we believed firmly in resources, suggested how the idea applied to ageing and memory (e.g. Rabinowitz et al. 1982) and to older adults working in dual-task situations (e.g. McDowd and Craik 1988).
If one consequence of ageing is reduced ability to process information rapidly and powerfully, how might that affect memory? It seems plausible that difficult, complex operations requiring a great deal of supervision and control will suffer badly while easy automated operations will be relatively unaffected. This was the message of an important paper by Hasher and Zacks (1979), who presented persuasive evidence that ageing spares automatic mental processes but severely impairs controlled processes. Procedural learning and priming are two types of retention that can apparently be executed in the relative absence of cognitive control, and it was shown by Light and Singh (1987) that priming declined little with age. Two other notions that fit this general analysis are, first, the proposal by Jacoby and colleagues (e.g. Dywan and Jacoby 1990; Jacoby 1991; Jennings and Jacoby 1993) that memory processes can be dissociated into the components of recollection and familiarity. Recollection requires conscious control whereas familiarity does not, and Jacoby has demonstrated that recollection is age sensitive whereas familiarity is not (Jennings and Jacoby 1993). Second, I have suggested that, when remembering is not well supported by either the external or internal environments, a person can nevertheless run off appropriate and effective retrieval operations by mustering ‘self-initiated’ mental activities (Craik 1983, 1986). Such self-initiated activities are effortful, however; they probably involve frontal lobe processes and apparently decline in the course of ageing (Craik 2002). I am therefore suggesting that ageing is associated with a reduction in either (or both) processing (p.82) resources or cognitive control, that this simple notion fits well with observed differential age-related losses, and also fits the basic ideas of a number of theorists (Craik 1983, 1986; Hasher and Zacks 1979, 1988; Jacoby 1991; Jennings and Jacoby 1993). There may be more latent agreement in the area than we have all acknowledged!
The current scene
I will end this nostalgic exercise in retrospective memory by pointing to a few topics of current interest in the area of age-related changes in memory, again focusing on topics that have been of greatest interest to me and my collaborators. Undoubtedly the main shift in research activity from previous decades is the current interest in brain mechanisms and their relations to behavioural and experiential variables. One aspect of this approach is neuropsychology—studying the effects of neural pathology on cognitive measures—and one focus here is to find parallels with the normal ageing process. Theories and findings in the area are well described in review chapters by Moscovitch and Winocur (1992) and Prull et al. (2000). Current conjectures include the notion that many age-related problems in working memory and cognitive control can be related to inefficiencies of frontal lobe functioning (West 1996), although this remains controversial (Phillips and Henry 2005). Rubin (1999) has suggested that Parkinson's disease and subcortical damage might provide a better model for cognitive ageing; this suggestion may be quite compatible with the frontal hypothesis, however, given the strong links provided by frontal–striatal circuits.
The area of cognitive neuroscience has been revolutionized by the development of various neuroimaging technologies, and this type of evidence is now an obligatory component of any current description of memory functioning. It still makes perfect sense to develop models based entirely on behavioural data, but it will be progressively more necessary for such models to be compatible with the rapidly accruing evidence from functional magnetic resonance imaging (fMRI) and other brain-scanning techniques. Evidence and theoretical notions relating to the cognitive neuroscience of memory and ageing are still in the early stages of development, although interesting work is being reported by Cheryl Grady, Roberto Cabeza, Denise Park, and other colleagues. Useful reviews of the current scene are provided by Grady and Craik (2000) and in a book edited by Cabeza, Nyberg, and Park (2005).
But many interesting things are happening in purely behavioural studies. One topic that has generated lively debate is ‘prospective memory’—remembering to carry out some planned action at a future time—‘remembering to remember’ (p.83) in Schonfield's (1982) phrase. I had suggested in 1983 that prospective memory necessarily involved self-initiated activities and should therefore be one type of memory that falls off steeply with age (actually, John Harris made this point to me after hearing the talk on which the paper was based). However, Einstein and McDaniel (1990) reported a study finding no age-related decrement in prospective memory, which clearly went against Schonfield's and Craik's (and Harris's!) intuitions. Later work by Einstein and McDaniel's group did find an age difference in time-based tasks (remembering to carry out an action at some nominated time), which the investigators characterized as being high in self-initiated retrieval, but no age differences in event-based tasks (remembering to carry out a planned action when a nominated cue is encountered later) (Einstein et al. 1995). Studies from my own lab have found age-related differences in event-based prospective memory, however (Rendell and Craik 2000), so the issue is still unresolved.
Context is an important concept in memory studies, and my own work has related it to ageing in two rather different ways. First, encoding context may be regarded as the source of some piece of acquired information—you know the facts, but were they learned from a friend, read in a newspaper, or seen on TV? Following work by Dan Schacter, a study by McIntyre and Craik (1987) showed that older adults are less able to recollect where and when they learned some information; that is, older adults show a mild form of ‘source amnesia’. Further, the degree of source amnesia is related to their performance on tests of frontal lobe functioning (Craik et al.1990). This latter study thus ties in difficulties of recollection to frontal lobe efficiency, and possibly again to age-related problems with self-initiated activities. Given this evidence of age-related problems in recollecting context, it is apparently paradoxical to claim that older adults' memory is more dependent on reinstatement of the encoding context than is the case for younger adults. Yet the evidence also shows that older people benefit more than do their younger counterparts from ‘environmental support’ at the time of retrieval (Craik 1983, 1986; Luo et al. 2007). This is not a paradox in my opinion, however; both cases reflect an age-related difficulty in initiating appropriate encoding and retrieval operations in the absence of external cues and guidance.
Age-related inefficiencies of cognitive control are still regarded as the root cause of cognitive decline at the present time, even though the precipitating factors remain a matter of fierce debate. Candidates include general slowing (Salthouse, 1996), less effective inhibitory or control powers (Hasher et al. 1999; Hay and Jacoby 1999), and a decline in working memory operations (Park et al. 2002). Even the vaguely defined notion of attentional resources is beginning to receive support from some neuroscientists (e.g. Sarter et al. 2006), (p.84)so this too remains a viable candidate. The next phase in theoretical development will be to show how these various constructs map on to one another; my own bet is that many will be shown to be aspects of the same basic neurobiological mechanism. One last puzzle is the role of consolidation in age-related memory problems. I have taken the view that, if processing is sufficiently rich and meaningful during encoding, good memory follows automatically, but this position was not supported in some studies of divided attention (Craik and Kester 2000). So it seems possible that, as with amnesic patients, older adults may process information deeply and elaborately, yet still show later memory decrements. Alan Baddeley has suggested this to me on several occasions.
Figure 6.1 Levels then and now. Part (A) shows recall and recognition scores for younger and older adults after encoding in terms of case, rhyme, and semantic incidental orienting tasks or intentional learning (Craik 1977). Part (B) shows recall and recognition performance after generation of the first letter, a rhyming word or a semantic association to presented names, or intentional learning of such names (Troyer et al. 2006).
Conclusion
So what has changed in 50 years? My retrospective scan through the area suggests that terminology and methods have changed a great deal, but that fundamental ideas have changed much less. Welford's (1958) point that many problems of ageing are attributable to an increased vulnerability to interference in short-term memory can be readily translated into age-related problems with working memory and executive functions. In my 1977 review paper, I reported a levels-of-processing experiment with younger and older adults. The main result of interest was that older adults performed as well as younger adults when a semantic orienting task at encoding was combined with a well supported (recognition) task at retrieval (Figure 6.1A). In recent work with Angela Troyer and others, we again used levels of processing in an attempt to bolster name learning in older adults, and again contrasted later recall with later recognition. The similarities with the earlier study are striking (Figure 6.1B), and I leave it to the reader to decide whether this represents brilliant foresight in 1977 or total stagnation in 2007!
No comments:
Post a Comment