Humans have recognized the difference between having knowledge and being able to use it for at least 2,000 years. In 1963, Raymond B. Cattell was the first to propose a psychometric model for these distinct abilities (1963). Horn and Cattell (1966) further developed his theory and Horn (Horn 1968, 1982; Horn and Noll 1997) applied it to changes in cognitive abilities in old age. Horn showed that our abilities to solve problems rapidly and accurately (and so achieve high scores on intelligence tests), to respond fast to simple signals, and to quickly learn unfamiliar material such as lists of random words peak in our early twenties but decline as we grow older. These changes are slight from our 30s through our 50s but accelerate throughout our 60s, 70s, and 80s. Horn and Cattell termed these “fluid abilities” or “fluid intelligence” (gf) because they are not specific to particular problems but support performance in all mental tasks. In contrast to waning fluid abilities, Horn showed information that we have learned throughout our lives is relatively age robust. Following Cattell, Horn called such bodies of acquired information “crystallized intelligence” (gc). Their choice of the word “intelligence” emphasizes that Cattell and Horn did not regard gc only as a mental archive of semantic information, such as vocabulary, or collections of names of birds or trees, or athletic records but also as tool kits that we have invented or learned to carry out “procedures.” Some examples of these procedures might be constructing grammatical sentences, doing algebra, working out lines of play in chess or go, or managing a bank, a business, or a kitchen. This distinction is clear in Horn’s discussions but while his experiments do show that vocabulary and other kinds of semantic knowledge can survive well into old age with little loss, he did not systematically explore how far this is also true of complex procedural skills that we have learned over our lifetimes. Horn’s early experiments and discussions also do not make the important point that, in contrast to “fluid intelligence” or “fluid abilities,” the “crystallized” skills are intensely “domain specific.” That is, mastery of a particular skill may be of little help in learning or using others, even if these may seem quite similar. An analogy from information technology is that computer programs written to efficiently carry out particular tasks are usually useless for any others.

Horn and Cattell’s distinction also implies a contrast between the difficulty of inventing, learning, and using a difficult procedure for the first time and its easy and automatic deployment once it has become familiar. This applies to cultural as well as individual accomplishments. Clay tablets, papyrus, libraries, and the World Wide Web can all be envisaged as means of “crystallizing” and indefinitely preserving semantic and procedural information that most individual humans could not discover or invent on their own: For example, it took Leibniz and Newton years of hard thought to invent the Calculus that schoolboys can now easily learn from textbooks in a few weeks.

The Cattell/Horn distinction raises interesting practical questions as to how we should view our likely trajectories of competence in our everyday lives throughout our lifespans. Though our fluid intelligence declines as we get older, can we still continue to practice demanding professions supported by information and skills that we have learned throughout our lives and still retain in our old age? An associated question is whether some kinds of crystallized abilities may be more age robust than others so that professional competence based on these durable abilities can be maintained longer than on more age-fragile skills?

A key issue is that retaining effective procedures is one thing, but carrying them out is quite another. Another helpful metaphor from information technology is the difference between developing and storing a program that is ideal to perform a particular task and having a system that is adequate to run it. As the “benchmark” bandwidth and memory capacity of an information processing system reduce, so will the maximum complexity of the programs that it can run. For humans “working memory” (seealso entry Crystallized Intelligence in this volume) is a convenient term coined by Alan Baddeley and Graham Hitch (1974) as a blanket label for our abilities to rapidly shift attention from less to more important information, to process new information and to relate it to other information that has been recently registered or held in memory for many years, and to reorder all of this old and new information so as to decide and implement what we should do next. All of the abilities implicit in the general concept of “working memory” are, in Cattell and Horn’s terms, “fluid” and age fragile. However, without a well-functioning “fluid” working memory system, we cannot manage to do complicated things for which we, long ago, learned reliable procedures: e.g., to produce a long, grammatical sentence, to understand and solve a complex business problem, to cook a complicated dish, or to plan and carry out apt sequences of moves in chess. So, as the efficiency of working memory sharply declines with age (Salthouse et al. 1989), we may still be able to perfectly describe effective procedures for completing complex tasks but become unable to meet the demands these make on our diminishing working memory capacity when we attempt to put them into practice.

A neat illustration of this is Susan Kemper’s (1990) analyses of diaries written by citizens of Kansas during the late nineteenth and early twentieth centuries. These often covered 40 or 50 years of their authors’ lives. As diarists aged the ranges of words that they used only slightly reduced (as Horn and Cattell’s empirical results predicted). Nevertheless, although their youthful diary entries were often long sentences and complex grammatical constructions, as they grew old, their sentences became shorter and their grammatical constructions increasingly simple. Retaining large numbers of words and retaining the ability to assemble them into complex sentences are different things.

This raises the interesting practical question whether some “crystallized abilities,” that is, kinds of learned knowledge and skills, can survive later in life than others. In 1935 Lehman (1935) pioneered studies of the ages at which distinguished mathematicians, scientists, poets, novelists, musicians, and artists had made their most remarkable contributions (Lehman 1942; Lehman and Ingerham 1939). Ages of greatest productivity and achievements were the early 20s and 30s for mathematicians, physicists, and chemists; the 40s and 50s for historians, philosophers, and novelists but might be as late as the 60s, 70s, or even 80s for some visual artists and musicians. Does this mean that some skills are more age resistant than others?

Lehman pointed out that his data were not ideal to address these questions. Many of the careers documented took place in the eighteenth and nineteenth centuries when life expectancy was much shorter and career trajectories were very different. Recent studies confirm early age-related losses in scientific productivity but suggest that these now happen much later than in the historical periods for which Lehrman collated data. Studies of British psychologists in the 1970s and 1980s (Over 1982); of large groups of less eminent physicists, geologists, physiologists, and biochemists in 1989 (Levin and Stephan 1989); and of the careers of economists and other scientists (Cohen 1991; Bayer and Dutton 1977), all found that, as they grow older, all academics publish less and in less prestigious journals. Recent studies of average or slightly above-average scientists find that their plateaux of greatest productivity last more than a decade longer than Lehman’s analyses suggested. Studies of artistic productivity also revise his conclusions. A 1999 analysis of the number of paintings produced by 739 graphic artists, works by 719 musicians, and books by 229 authors found that, like most contemporary scientists, their periods of maximum output were in their 30s and 40s. Unfortunately literary skills are not immune to changes that come with old age and with approaching death. Suedfield and Piedrahita (1984) analyzed the late work of distinguished novelists and found that the quality of writing in their correspondence declined during the 10 years before they died.

Other recent studies find that while the learned skills of bankers and business executives allow them to competently do their jobs in late middle age and even give them some advantage over younger colleagues with less experience, their ability to correctly analyze and cope with novel problems tends to have decline by their late 40s and 50s (Colonia-Willner 1998, 1999).

The current consensus is that while competence, even at learned and highly practiced skills, does decline with age, these changes are much smaller and slower than the earliest surveys suggested. The contrast between “early flowering and early decline” in the hard sciences and “late flowering and late decline” in the humanities and visual arts now seems less clear-cut. One problem is finding comparable standards across different disciplines. Assessments of quality in the arts are much more contentious and differ sharply between various kinds of achievement. Standards of comparison are more elusive than the earliest studies assumed. For example, a tally of the year 2000 market value of paintings by 51 modern US artists found that for painters born before 1920, the average peak age for the valuation of their paintings was 50.6, but for those born after 1920, it was only 28.8. Changes from a cautious to a speculative market account for similar discrepancies (Colonia-Wilner 1999). From current sale prices, we might conclude that artists who are now elderly are painting much better (or at least much more profitably!) than their young contemporaries or, indeed, than themselves when young.

Ideal data to examine differences between “fluid” and “crystallized” abilities would be the achievements of large numbers of extremely gifted people on the same, difficult, mental skill on which they can be compared against each other in terms of a common objective standard. The careers of chess masters are as close to this as we get. Chess requires high levels of both fluid abilities, such as working memory and intelligence. It also requires crystallized knowledge because it has been so exhaustively researched and documented that, even for young prodigies, success at the highest levels needs long and intensive study. Chess play requires an ability to simultaneously hold many variables in mind and to recognize, as rapidly as possible, how patterns of relationships between these variables will alter if particular moves are made. Clearly there is a “natural talent” for chess because some prodigies play at a remarkably high standard at ages as early as 6–12 years. However, a study by Charness and colleagues found that even maintaining success at much lower levels than “grandmaster chess” requires 5,000 or more hours of deliberate practice over 10 years (Colonia-Wilner 1999). In terms of John Horn’s dichotomy, chess mastery requires both considerable fluid intelligence and a formidable body of learned crystallized knowledge of tactics and strategy.

Statisticians such as Arpad Elo developed very sensitive and reliable systems for rating the relative strengths of different chess players. Because these have been used and validated for at least 50 years, even small changes in the playing strengths of individual grand masters can now be tracked from their 20s through to their 70s. Elo’s initial studies (Charness et al. 2005) found that nearly all the careers he compared showed improvements until 30s or 40s, a plateau of the best achievement until the late 50s or, in some few cases, early 60s but then a significant decline. During the historical spans of this and later analyses, chess at the highest level evolved so rapidly that players could not keep or improve their ranks unless they continually revised a vast body of theory on openings and end games. It is cheering to find that gifted individuals can, though perhaps with gradually increasing effort, remain at the very peak of an extraordinarily demanding profession until their seventh decades. Evidently, acquired knowledge and endless practice can support even a skill that is, essentially, computational and demanding of fluid intelligence until late in life. We must also remember that long after they had retired from competitive chess, these remarkable people could still play at a level that most humans cannot hope to reach at any age. We should also remember the wise comment of Gary Kasparov, perhaps the greatest player yet: “Excelling at chess has long been considered a symbol of more general intelligence. That is an incorrect assumption in my view, as pleasant as it might be.” Crystallized intelligence is intensely task specific.