50 Years of Cognitive Aging Theory

Abstract

Objectives:

The objectives of this Introduction to the Journal of Gerontology: Psychological Sciences special issue on “50 Years of Cognitive Aging Theory” are to provide a brief overview of cognitive aging research prior to 1965 and to highlight significant developments in cognitive aging theory over the last 50 years.

Method:

Historical and recent theories of cognitive aging were reviewed, with a particular focus on those not directly covered by the articles included in this special issue.

Results:

Prior to 1965, cognitive aging research was predominantly descriptive, identifying what aspects of intellectual functioning are affected in older compared with younger adults. Since the mid-1960s, there has been an increasing interest in how and why specific components of cognitive domains are differentially affected in aging and a growing focus on cognitive aging neuroscience.

Discussion:

Significant advances have taken place in our theoretical understanding of how and why certain components of cognitive functioning are or are not affected by aging. We also know much more now than we did 50 years ago about the underlying neural mechanisms of these changes. The next 50 years undoubtedly will bring new theories, as well as new tools (e.g., neuroimaging advances, neuromodulation, and technology), that will further our understanding of cognitive aging.

This special edition commemorates the roughly 50 years of theoretical work on cognitive aging. As an Introduction to this issue, we first briefly review cognitive aging efforts prior to the mid-1960s, then describe the mid-1960s as a pivotal point for cognitive aging theory, and finally comment on how the field has evolved since then.

Cognitive Aging Research Prior to the Mid-1960s

Observations of cognitive aging date back millennia. Take, for instance, the sixth century BC poet Solon who, in writing about life’s hebdomads, or periods of 7 years, “… in the ninth, though he’s still capable, his tongue and expertise have lost some of their force” (Baars, 2012). Empirical proof of cognitive aging appeared in the 1930s, when Miles evaluated the perceptual, motor, and cognitive abilities of 1600 people aged 6 to 95 years old, and reported declines after age 30 in these skills, including learning ability (e.g., Miles, 1933). Age-related slowing emerged as a salient feature in the 1930s; the fact that most intelligence tests were speeded led Lorge (1940) to correct intelligence test data for speed of processing, finding that intelligence did not, in fact, decline with age. Lorge’s discovery of the relationship between speed and intellectual capacity reminds us of Salthouse’s later and more encompassing theory that age-related changes in speed underlie changes in a number of cognitive domains (e.g., Salthouse, 1996).

This early literature was predominantly descriptive, comparing task performance across age groups in cross-sectional designs, and typically used the same intelligence tests that were used in the study of child development. In 1955, Wechsler identified the maintenance (“hold”) and decline (“no hold”) of what was later termed crystallized and fluid intelligence, respectively (Cattell, 1963). Uncovered too was a decline in what we now call working memory, thanks to the inclusion of backward digit span in intelligence testing (Bromley, 1958).

Emergence of Cognitive Aging Theory: The Mid-1960s

1965 marked the beginning of a sea change in cognitive aging theory. This was the year that Welford and Birren (1965) published Behavior aging and the nervous system, a collection of chapters on slowing, attention, memory and their relationship with biological and health factors that together provide a strong foundation for cognitive aging theory. Patrick Rabbitt (1965) reported a study in which younger and older adults sort cards as quickly as possible into piles. Sorting was done once into two piles and once into eight piles, defined by the number of relevant (target) stimuli in that condition. Moreover, each card contained 0, 1, 4, or 8 irrelevant stimuli. Older adults’ sorting times were more vulnerable to the increasing amount of irrelevant information, particularly when fewer stimuli were relevant. Rabbitt concluded that older adults have greater difficulty ignoring irrelevant information compared with their younger counterparts and suggested based on some of his previous work that this may be due to a reduced efficiency of perceptual grouping. Although the way in which we carry out and describe research has certainly changed since then (e.g., in his paper, stimuli were stenciled onto cards that participants manually sorted, participants were referred to as Ss in the manuscript), the ability of older adults to ignore irrelevant information and both the hindrances, as well as the surprising advantages of them doing so is still a topic of much interest today (e.g., Biss, Ngo, Hasher, Campbell, & Rowe, 2013; Weeks & Hasher, 2014).

1965 also brought us Schonfield’s letter to the journal Nature. Until that time, cognitive aging researchers had focused on the acquisition and short-term storage of new information and reported equivocal findings. Schonfield showed for the first time age-related deficits in retrieval from long-term memory—free recall performance decreased across age decades (20s to 60s+), but recognition did not. These early ideas were later fleshed out in Craik’s notions of the depletion of attentional resources with age and the effects of environmental support, in which a task environment supporting elaborative encoding or guiding retrieval can compensate for older adults’ reduced attentional resources that render encoding and retrieval less effective (Craik, 1986).

In 1966, Bromley made the very prescient observation: “Intellectual processes appear to become more differentiated as we grow up. There is some evidence – too little to support firm conclusions – that during maturity and old age the process of intellectual differentiation reverses…” (p. 205). One early finding perhaps contributing to this observation was Balinsky’s (1941) report of a U-shaped function of correlations between performance on various tests from the Wechsler–Bellevue Scale, where inter-test correlations were high in 9- and 12-year-old children, but then decreased across age groups until the mid-40s, after which correlations increased again in the 50- to 59-year-old age group (the oldest age group included). Bromley’s reflection presages the notion of dedifferentiation—the heightened correlation among cognitive, perceptual, and neural processes—that was revived in cognitive aging research in the 1980s (e.g., Baltes, Cornelius, Spiro, Nesselroade, & Willis, 1980) and still remains a topic of interest today, particularly in relation to the dynamic activity of the brain (e.g., Ferreira et al., 2015).

Cognitive Aging Theory: Evolution Since the Mid-1960s

Two notable trends in cognitive aging research over the past 50 years reflect changes in the parent discipline of cognitive psychology; the first is an increasing differentiation among components of cognitive performance (accompanied by a growing realization that life-span changes in these components peak at different ages) and the second is a shift in emphasis to the cognitive neuroscience of age-related changes—attributing cognitive changes to age-related changes in brain structure and function (e.g., Raz, 2000). With regard to changes with age in general intelligence, some influential researchers argued that there is little or no decline (Baltes & Schaie, 1976) and that popular conceptions of age-related losses are a “myth.” This optimistic position was hotly contested, however, withone line of argument being that fluid intelligence declines markedly with age after peaking in the 20s or 30s, whereas crystallized intelligence is maintained until the 60s or later or even improves across the life span (Cattell, 1971; Horn, 1970; Horn & Donaldson, 1976). The analysis in terms of component abilities was taken further by Salthouse (1982) who showed that measures of vocabulary and general information hold up well into the 70s at least, whereas speeded measures such as digit-symbol substitution decline precipitously from the 20s or early 30s. This account in terms of differential age-related decline, presumably linked to differences in vulnerability to aging of the relevant brain regions, is still the current view. For example, Park and colleagues (2002) presented compelling data showing that whereas working memory and speed of processing decline steadily for the 20s to the 80s, forward digit span shows only a moderate decline, and verbal knowledge actually increases from the 20s to the 70s. Later work has shown this useful categorization of crystallized and fluid abilities to be too broad, however. A recent online study of almost 50,000 participants showed that the age of peak performance on a large sample of intelligence tests varied from the late teens to the 50s (Hartshorne & Germine, 2015). In line with previous formulations, the tests that peaked late (e.g., vocabulary and comprehension) tended to reflect crystallized intelligence, whereas those that peaked early (e.g., digit-symbol coding, letter-number sequencing) reflect aspects of fluid intelligence, but there is no clear categorical shift from one type to another type of test.

Paul Baltes (who sadly died in 2006) was a major contributor to the field of cognitive aging. He took a very positive, optimistic view of aging, acknowledging that cognitive abilities do decline, but that older adults possess some degree of “latent reserve” that can be activated with the investment of additional time and energy. Together with Margret Baltes he argued that older individuals cope with a decrease in environmental adaptability by strategically choosing classes of behavior that are most adequate for their personal lifestyle. They termed this the process of “selective optimization” (Baltes & Baltes, 1990). Paul Baltes also directed a large-scale study of memory training in Berlin, demonstrating that by implementing strategies older adults could improve their memory dramatically for specific materials (Kliegl, Smith, & Baltes, 1989). In line with other results, however, such gains tended to be restricted to the practiced situation and not show “far transfer” to other situations (Baltes & Willis, 1982). One other interesting result from the Baltes group was the discovery of a strong link between sensory functioning (visual and auditory acuity) and various measures of intellectual ability (Baltes & Lindenberger, 1997). Further, these relations were much stronger in an older group (70–103 years) than in a group of younger adults (25–69 years). The authors considered various possible accounts of these findings, somewhat favoring a “common cause” hypothesis—the notion that age-related decrements in brain structure and function are associated with correlated losses in a variety of sensory and cognitive functions. It seems possible that the increased linkage with age between sensory and cognitive functions could be another example of the general age-related loss of differentiation among abilities noted earlier.

Age-related changes in attention and memory have been major topics in the field of cognitive aging during the last 50 years. Continuing on the trend mentioned earlier in this Introduction, a further sign of increasing differentiation of cognitive abilities is the distinction among various types of memory (e.g., episodic, semantic, implicit, explicit, and prospective) and aspects of attention (e.g., selective attention, sustained attention, inhibition, and set switching). Recent evidence shows that these varied aspects of attention and memory also appear to age differentially. Research on memory is well covered by Park and Festini (2016), so we will describe some highlights of the “attention and aging” literature in the present introductory article. Early theoretical statements concerning the nature of human attention were put forward by Broadbent (1958), Treisman (1964), and Kahneman (1973), with the first two focusing on selection and the last on arousal and effort. As described earlier, Rabbitt (1965) demonstrated that older adults were less able to ignore irrelevant information when selecting targets, thereby implicating both selection and inhibition as aspects of attention and choice. Other aspects include sustained attention, divided attention, set switching, executive control, and (probably) working memory. One clear trend over the last half century has been the study of age-related changes in these types or components of attention, rather than simply “attention” as such. This trend, in both the aging literature and the literature based on young adults, has continued despite clear evidence that these various types of attention are often highly interrelated. Because much of the work on attention involves frontally based cognitive control mechanisms, it might be expected that all aspects of attention should decline with age. This follows from work showing that the frontal lobes are areas of the brain known to be particularly vulnerable to the effects of aging (Raz, 2000; West, 1996). Teasing out age-related effects has been surprisingly complex, however, as shown by the following brief accounts.

Following the work of Rabbitt (1965), further research has generally confirmed an age-related decrement in selective attention (e.g., Plude & Hoyer, 1985) with later work by Plude and colleagues suggesting that the problem lies specifically in the feature-integration phase of perception (Plude & Doussard-Roosevelt, 1989). Such difficulties in selection are compounded by age-related inefficiencies in inhibition (Hasher & Zacks, 1988; although see also Kramer, Humphrey, Larish, Logan, & Strayer, 1994). In an influential article, Hasher and Zacks (1988) proposed that older adults are less able to inhibit unwanted or irrelevant material, which then “occupies space” in working memory, thereby reducing temporary storage and further processing abilities. Further work has suggested that inhibition is not a unitary construct, however, and that this approach may help to understand when age-related decrements are and are not found (Kramer & Madden, 2008).

The ability to manipulate top-down control of selection, for example, selecting sometimes based on color and sometimes on shape, is referred to as task switching or set switching, and this literature shows some agreement with respect to age differences. Paradigms typically contain some blocks of trials in which the selection criterion is stable (e.g., select by color only) and some blocks in which participants must switch unpredictably from color to shape from trial to trial. The longer latencies associated with mixed blocks versus pure blocks are referred to as general task-switching costs; these costs increase with age. The longer latencies associated with switch versus nonswitch trials within a block are called specific task-switching costs, and there is good agreement that general costs increase throughout adulthood whereas specific costs remain relatively stable (Kray & Lindenberger, 2000; Reimers & Maylor, 2005; Verhaeghen & Cerella, 2002). The Reimers and Maylor study was conducted online on more than 5,000 participants aged between 10 and 66 years; they found that general costs were lowest for participants in their teens and increased steadily from the age of 18 to 66 years.

Sustained attention over several minutes is another aspect of attention that remains stable with age—at least until the mid-40s. Berardi, Parasuraman, and Haxby (2001) and also Carriere, Cheyne, Solman, and Smilek (2010) reported minimal adult age differences in aspects of sustained attention. However, in another large online study, Fortenbaugh and colleagues (2015) reported results from more than 10,000 participants and found that performance peaked in the early 40s and then declined. As always, of course, the precise pattern may depend on the exact task chosen; Fortenbaugh and colleagues used a city versus mountain scene discrimination task, whereas Berardi and colleagues (2001) used a digit discrimination task and Carriere and colleagues (2010) used the sustained attention to response task.

With regard to divided attention, Craik (1977) made the bold assertion that “One of the clearest results in the experimental psychology of aging is the finding that older subjects are more penalized when they must divide their attention, either between two input sources, input and holding, or holding and responding.” This claim was challenged, however, in a study by Salthouse and Somberg (1982) who controlled for age differences in single-task performance and found no significant age difference in divided attention ability. A later study from Salthouse’s lab (Salthouse, Rogan & Prill, 1984) did find an age-related decrement in divided attention, however; the authors suggest that the discrepancy between the results may reflect the fact that task complexity was greater in the second study. Verhaeghen and Cerella (2002) reviewed the results of a series of meta-analyses and concluded that age deficits in dual-task performance are in fact generally found.

Finally, the topic of age-related differences in “executive processes” or “cognitive control” has received progressively more attention in the course of the last 50 years. An important paper by Hasher and Zacks (1979) drew the distinction between automatic and effortful cognitive operations, illustrating their claim that effortful processing requires considerable amounts of processing capacity and that age decrements are typically found on such tasks, in contrast to the absence of such decrements on automatic tasks. The notion that conscious, effortful processing requires cognitive control which in turn becomes less efficient with age is now generally accepted. Two brief examples are first the work from Larry Jacoby’s lab (e.g., Jennings & Jacoby, 1993); in this study, the authors found that older adults were impaired on consciously controlled memory processing. The second example is the suggestion from Todd Braver and colleagues that one major aspect of cognitive control consists of maintaining a task-related context in working memory and that older adults are less adept at this form of processing (Braver et al., 2001). Braver’s approach emphasizes the point that various aspects of attention—in this case working memory and executive control—are closely interwoven. More generally, when attempting to understand age-related problems of attention, it is increasingly clear that while attention is not a unitary construct, its components (inhibition, executive control, working memory etc.) are also not unitary constructs—possibly right down to the level of specific tasks. This understanding calls for a new framework for the interpretation of “attention” and its changes with age.

The Future

We can forecast a few trends in the future of cognitive aging research. We anticipate a greater focus on cognitive aging throughout the adult life span, not from cross-sectional studies but rather from the many large longitudinal studies that are currently ongoing. Some of this work might be interventional—our understanding of midlife predictors of later life cognitive ability is growing, but what if we intervene at midlife in some of those harbingers of cognitive decline? More sophisticated neuroimaging techniques with better spatial-temporal resolution, and perhaps ways to track neurotransmitter activity or gene expression in real time, may be developed that will afford a better understanding of the neural contributors to cognitive aging and provide another avenue to intervene. Technological advances will not only support cognitive functioning but could also help us move our labs into the real world and track cognitive functioning as it is influenced by people’s environments. Whatever the future may be, it will stand on the solid foundational theoretical understanding of cognitive aging that has developed over the last 50 years.

The Special Issue

Several influential cognitive aging theorists have contributed to this special edition. Salthouse highlights how an individual difference approach has advanced our understanding of cognitive aging. Three empirical papers applied individual difference approaches. Thorvaldsson, Karlsson, Skoog, Skoog, and Johansson demonstrate marked cohort effects in both baseline cognitive functioning and cognitive decline. Gow, Pattie, and Deary identify associations of midlife leisure and late life physical activity with cognitive ability level and cognitive decline, respectively. A paper from the late Humphreys and colleagues demonstrates the utility of technology to facilitate cognitive assessment in low income, low literacy settings, where the effects of age and education on various cognitive domains are remarkably similar to what the earlier studies demonstrated. Harkening back to the dawn of theoretical cognitive aging, Rabbitt reflects on the evolution of research on age-related changes in speed of visual recognition. Kensinger and Gutchess place cognitive aging in a social context with their description of the progression of research on social and affective cognitive aging, whereas Fischer, O’Rourke, and Loken Thornton report age-invariant and age-varying influences of various cognitive domains, sex, and pulse pressure on cognitive and affective theory of mind. Park and Festini guide us through the evolution of theories of memory and aging. Two empirical papers demonstrate the influence of memory in other cognitive domains. Wallin, Gajewski, Teplitz, Mihelic Jaidzeka, and Philbeck demonstrate that memory for a preview of a scene facilitates judgments of distance to a spatial target after a later brief glimpse, more so for older than younger adults. Finally, Shafto, James, Abrams, Tyler, and Cam-CAN provide convincing evidence that, contrary to previous assertions, greater acquisition of lifelong knowledge (crystallized intelligence) does not lead to more frequent tip-of-the-tongue experiences. This collection of review and empirical papers provides a broad perspective on both historical and contemporary cognitive aging theory.

Funding

This work was supported by a grant awarded to NDA by the Canadian Institutes of Health Research (MOP 123484) and by a grant awarded to FIMC by the Natural Sciences and Engineering Research Council of Canada (A8261).