Abstract
Over 100 years have passed since Binet and Simon proposed scales for assessment of intelligence of children to predict academic success and failure. The extension of these assessments to adults largely resulted from efforts of psychologists to provide insights for military selection in World War I. At the time, relatively little thought was given to how adult intelligence might differ from child and adolescent intelligence. Traditional approaches for assessing adult intelligence have largely survived. However, there is little reference to adult intellectual functioning outside of laboratory-based tasks and clinical assessments of pathology. The result is that there are insufficient criterion measures for adult intelligence. Moreover, researchers have shifted from treating intelligence tests as predictors to treating them as criterion measures. The result is a disconnection between basic research on one hand and understanding adult intelligence on the other hand. This lack of connection is a serious impediment for predicting individual differences in performance on tasks which adults perform in their day-to-day work and nonwork lives. This article explores how the field has come to the current situation, and what remedies might be explored. Ultimately, a fundamental reexamination of how adult intelligence is studied and applied is suggested.
Adult intelligence as a construct, its assessment, and associated applications have represented vexing problems for psychologists for nearly 100 years. Early efforts toward assessing adult intelligence were largely an almost accidental offshoot of assessing child intelligence. In 1905, Binet and Simon introduced the first “modern” intelligence test for children. The Binet-Simon scales (Binet & Simon, 1905), and the various translations and extensions (most notably perhaps the Stanford-Binet; Terman, 1916), were highly successful in predicting the academic success of children and young adolescents. Two innovative aspects of the Binet-Simon method represented a critical contribution to the assessment of child and adolescent intelligence but provided serious obstacles to assessment of adult intelligence, namely age differentiation and the mental age construct. To determine which items to include in the scales, Binet and Simon first assumed that older children, on average, were more intelligent than younger children. Thus, items were specifically designed or selected such that a higher percentage of older children were likely to answer the items correctly than younger children. If the items did not evidence age differentiation, they were discarded. In a related fashion, the mental age construct has as its core the proposition that absolute intellectual ability levels are a monotonically increasing function of age.
The Measurement of Adult Intelligence
The first efforts toward developing tests of intellectual ability that could accommodate older adolescents and adults (e.g., Yerkes, Bridges, & Hardwick, 1915) dispensed with both the age differentiation and mental age constructs in favor of an absolute or “point scale.” Nonetheless, Yerkes et al. argued, on the basis of a modest-sized sample of adults, that for all intents and interests, maximal average intelligence was probably reached at age 16. Terman (1916), while generally keeping with both age differentiation criteria and mental age constructs, essentially fudged a solution for adults. The original Stanford-Binet provides age-graded tests from age 3 to 14, and then two levels of adult intelligence—an “average adult set” of scales/norms, and a “superior adult.” The scales to assess the superior adult in the Stanford-Binet included (a) Vocabulary, (b) Binet’s Paper-Cutting Test, (c) Repeats 8 Digits, (d) Repeats Thought of Passage Heard, (e) Repeats 7 Digits Backward, and (f) Ingenuity Test (p. 61). The Ingenuity Test is essentially a water pitcher math problem—figuring out how to obtain seven pints of water using only a three-pint and a five-pint vessel.
Thus, the Stanford-Binet approach to adult intelligence was to extend the types of tests from child and adolescent intelligence assessments, but to “fix” mental age to a maximum roughly at age 15 and to ignore mental age as a reference beyond that point. Moreover, the adult items were not selected on the basis of age differentiation, but rather appear to have been selected with reference to increasing test difficulty (from the adolescent-level tests) and also with reference to normative samples of individuals who were either assumed to be “normal” or “superior” from other sources (e.g., education, occupational level).
Through the efforts of Yerkes and his colleagues, roughly 1,700,000 Army conscripts were administered intelligence assessments during World War I. The shortcomings of the mental age construct when extended to adults were well known in the 1920s when some researchers attempted to determine the average “mental age” of the U.S. Army conscripts by comparing their scores to mental age scores on the Stanford-Binet, a project that yielded the questionable conclusion that the mean mental age of these adults was 13.14 (e.g., Brigham, 1923). Although the arguments were quite lively at the time (e.g., Lippmann, 1922; Terman, 1922), the existing data from these early efforts toward assessment of adult intelligence did indeed point to a peak level of average absolute scores on existing intelligence tests somewhere between about age 18 and age 21. 1
One important finding from explicit studies of age and adult intelligence in the 1920s was that, at the level of individual intelligence test scales, there was an uneven pattern of growth and decline across the scales. In particular, tests of vocabulary and general knowledge were more likely to show small or negligible differences across much of the adult age range. In contrast, tests of short-term memory, mathematics, reasoning, and spatial ability showed much greater differences across age groups, with older adults performing, on average, much worse than adolescents and young adults (e.g., Conrad, 1930; Hsiao, 1927). Although one could not rule out cohort differences as a major source of these results, supporting evidence for the preservation of verbal knowledge and abilities across a wide range of adult ages was provided in a classic longitudinal study by Owens (1953). Owens found that in a sample of 127 men who had originally been tested on the Army Alpha test in 1919 (around age 18), and then retested in 1950 (around age 49), there were significant gains in the verbal/information subtests and mostly stable scores on the other subtests.
Wechsler’s scales
The first major omnibus individual test of intelligence specifically developed for assessment of adults was created by Wechsler (1939). The test was first named the Bellevue Intelligence Scales and then later revised to become the Wechsler Adult Intelligence Scales (WAIS). In his discussion of the need for an intelligence test for adults, Wechsler (1939) noted three major justifications: (a) The Binet inspired tests contained material that was unsuitable for adults, (b) the Binet tests had too great an emphasis on speed of responding to questions, rather than accuracy, and (c) there was a problem with using mental age as a reference for intelligence quotient (IQ) classifications when there was clearly an absence of monotonically increasing scores with increasing age once late adolescence was reached.
It is important to note that most of the underlying intellectual demands (e.g., short-term memory, general information, comprehension) from the scales that make up the WAIS can be traced to Binet-Simon and Terman’s intelligence tests. That is, although Wechsler’s tests are not identical in content to the Binet-inspired tests, they have a great deal of overlap, such that rank ordering of individuals on both the Wechsler and Binet-type tests are typically quite similar in nonclinical samples (Anastasi & Urbina, 1997). The important point, for the current discussion, is that Wechsler’s tests—though they included material that was more suitable and relevant for adults—did not fundamentally change the assessment of intelligence (from a content perspective) from that of Binet and Terman, except perhaps for the separate reporting of verbal and performance composites of intelligence. The principal advantage of the WAIS has been its historical use as part of clinical diagnoses and neurological assessments, especially prior to the widespread availability of magnetic resonance imaging (MRI) assessments, which provide more accurate diagnoses of neurological impairments in adults.
Adult intelligence and aging
In the 1950s, Schaie started a long-running lagged cross-sectional study of cohort differences and adult aging-related intelligence changes. Schaie (1958) rejected the use of the Wechsler-Bellevue scales, because although they were deemed useful for clinical applications, they did not provide the level of precision in assessment of lower order mental abilities that constitute overall intelligence levels. Thus, in his study, Schaie focused on assessing the abilities defined by Thurstone as “primary” and used primary mental ability tests that were originally designed for adolescents, namely (a) Verbal-Meaning, (b) Space, (c) Reasoning, (d) Number, and (e) Word Fluency (Schaie, 1958). In subsequent decades of study, Schaie established that there were different patterns of growth and decline for these abilities and also documented cohort differences that had been long suspected to confound aging interpretations from cross-sectional studies of adult intelligence (e.g., Schaie, 1996, 2005).
Demming and Pressey (1957) noted that the existing literature of the time indicated well-documented lower average IQ scores of middle-aged and older adults when compared to adolescents and young adults. However, they conjectured that standard intelligence assessments included a substantial amount of content that was not clearly relevant to the day-to-day intellectual activities that were “indigenous” to adults. These authors created a set of test items that would be both interesting and relevant to the kinds of intellectual tasks that adults were more likely to confront, such as functional knowledge of legal terms, occupations, and where to find information in a telephone directory (i.e., an archaic periodical that contained both residential/business phone numbers by names and business advertisements in a local city or town). The results of their study were that even as older age groups performed less well compared to young adults on standard intelligence tests, the older adults performed better on average than the younger adults on the tests that Demming and Pressey designed to be “indigenous” to adults.
More recent investigations have taken myriad ap-proaches to expanding the construct of intelligence beyond the traditional Binet-Simon approach, such as research programs that have focused on constructs such as of wisdom, everyday cognitive tasks, practical intelligence, and emotional intelligence. Each of these programs has provided important perspectives on adult cognition, yet they appear to have limited traction in accounting for the bulk of adult intellectual activities, at least until later life (i.e., after retirement). Explorations of wisdom (Baltes & Staudinger, 2000; for a review, see Staudinger & Glück, 2011) identify the construct as a complex of subjective insights and objective capabilities to solve problems that require balancing “good and bad, positivity and negativity, dependency and independence, certainty and doubt,” and so on (Staudinger & Glück, 2011, p. 217). Assessments of wisdom from this framework focus on problems like advice giving on existential issues (e.g., suicide). Investigators have concluded that general knowledge and experience, which are integral to crystallized intelligence, are associated with performance on such measures (Jordan, 2005). But, it is not clear that these kinds of problems represent common intellectual activities for most adults. Indeed, if they were common problems, then it seems that the judgment and insight demands of these problems would be supplanted by reference to external sources. In contrast, there have been examinations of the relations between intelligence and performance on everyday tasks (e.g., food preparation, medication intake, telephone use), mostly for elderly populations (e.g., Diehl, Willis, & Schaie, 1995; for a review, see Thornton & Dumke, 2005). Together, these investigations may provide the basis for considering some nonwork components of intelligence, especially for older adults.
Other abilities
Other researchers have attempted to broaden the construct of intelligence in a manner that reconceptualizes the construct for both children and adults, such as Gardner’s multiple intelligences framework (Gardner, 1983) and Sternberg’s (1985) triarchic theory. Sternberg’s theory in particular, includes a significant component of practical intelligence, which includes concepts of tacit knowledge (see, e.g., Polanyi, 1966/1983), or what Broudy (1977) referred to as “knowing with”—which essentially is the individual’s repertoire of thinking, perception, and judgment skills used to solve problems, even though the individual might not be able to remember or articulate the specific strategies or skills. Some efforts have been made to develop assessments of tacit knowledge for adults (see, e.g., Sternberg et al., 2000), though there have been relatively few developments in this area in the past decade. In addition, though there are some findings that tacit knowledge/practical intelligence increases with tenure in a job (Sternberg & Hedlund, 2002), these particular approaches do not directly address adult intellectual development issues.
There has also been quite a bit of discussion in the literature of the past two decades about the construct of “emotional intelligence.” Although there appears to be overlap between some aspects of emotional intelligence and traditional intellectual abilities (R. D. Roberts, Zeidner, & Matthews, 2001), extant measures of emotional intelligence are limited in convergence with traditional measures, and they have yet to reach a useful level of criterion-related validity for criteria of academic and occupational performance measures (see, e.g., Murphy, 2006; though see Mayer, Salovey, Caruso, & Sitarenios, 2001, for a different perspective). Because of these concerns, discussion of emotional intelligence and related variables reflects constructs beyond the scope of the current discussion.
Domain knowledge
Some time ago, Dixon and Baltes (1986) called for research on professional knowledge and specialization as an important means toward understanding adult intellectual development. Along these lines, but with the additional inclusion of nonoccupational domain knowledge, Ackerman’s (1996) framework for adult intellectual development, called PPIK (for intelligence as process, personality, and interests, leading to intelligence as knowledge) was an attempt to reframe the discussion of adult intelligence away from the traditional Binet-inspired construct, toward a construct representation for adult intelligence that was based on individual differences in the depth and breadth of knowledge and skills. Most important to this conceptualization is that it focused on a broader representation of knowledge than is assessed on IQ tests. That is, IQ tests are designed to find the lowest common denominator for a population’s intellectual repertoire, mainly including information/knowledge that is common to a wider culture. Ackerman (1996) proposed that individual differences in adult intelligence must represent what an adult can do, and such a conceptualization included domains that were not common to the culture, but instead represented domain-specific knowledge that might be common only to those with have similar educational backgrounds, occupational experiences, or avocational interests (e.g., hobbies). This framework for adult intelligence recognized that musicians, carpenters, physicists, and sales clerks might have both overlapping and nonoverlapping domain knowledge, but to assess their relative intelligence levels, one must sample knowledge very broadly (consistent with Cattell’s, 1971/1987, notion of “current” crystallized intelligence). Ultimately, the individual’s repertoire of knowledge and skills is likely to be strongly related to the intellectually demanding tasks that the individual can successfully perform in the world outside of the laboratory.
When adult intelligence is conceptualized as largely based on an individual’s repertoire of knowledge and skills, cross-sectional studies of age and intellect as knowledge show a strikingly different pattern from cross-sectional studies of adult intelligence measured with traditional IQ tests, such as the WAIS (e.g., Ackerman, 2000; Ackerman & Rolfhus, 1999). Instead of the average peak level of performance at late adolescence or early adulthood (see, e.g., Wechsler, 1939), assessments of domain knowledge in many areas showed increasing or level performance well into middle age (see, e.g., Ackerman, 2000).
However, characterizing adult intelligence as solely “knowledge and skills” is an oversimplification. To maintain professional expertise, for example, a doctor, lawyer, professor, or electrician must acquire new knowledge in accordance with developments in the particular occupational field. Even in the area of avocational activities, there are differences between individuals who merely attempt to retain earlier knowledge and those who seek out and acquire new insights or broaden/deepen their knowledge in the particular domain. Thus, as suggested by Hebb (1942), some “process” aspects of intelligence may be critically important to many adult intellectual endeavors (e.g., those that require new learning or short-term memory).
Rationality
An alternate viewpoint for adult cognitive functioning has recently been offered by Stanovich and his colleagues (for a review, see Stanovich, West, & Toplak, 2016). These investigators explicitly avoid the terminology of intelligence. Instead, they place the challenges of cognitive functioning in adults as determined by a set of rational reasoning abilities and skills (including statistical reasoning, scientific reasoning, reflection versus intuition, avoiding reasoning biases, financial literacy, and so on). As noted by Stanovich et al., these constructs are mostly “domain specific” in the areas of “probabilistic reasoning, causal reasoning, and scientific reasoning” (p. 35)—areas that are not covered by traditional measures of either fluid or crystallized intellectual abilities, even though the authors report a substantial correlation between their rationality quotient and IQ (.69). Expanding conceptualizations of adult intellectual functioning in this manner represents a promising direction, but this approach does not currently address issues of growth and decline in the context of adult development.
The Current State of Adult Intelligence Research
The literature on adult intelligence up to the present day has come a long way since it was “established” after World War I that the mean mental age of adult men was 13.14. The key findings are as follows:
The mental age construct is largely meaningless after adolescence.
There are (or at least have occurred) large birth cohort differences in intellectual abilities, as measured with traditional Binet-inspired intelligence tests.
Different lower order intellectual abilities show different patterns of growth and decline throughout adulthood. In particular, verbal/information/crystallized intellectual abilities are the most robust in the face of adult aging, while highly speeded tests and tests of short-term/working memory and abstract reasoning abilities show the greatest declines in average performance as adult age increases.
In many domain knowledge assessments, middle-aged and older adults perform, on average, better than adolescents and young adults.
Intelligence scores as criteria
Somewhere along the line, intelligence tests scores were transformed from being considered only as a predictor variable (e.g., for predicting academic success or job performance) to themselves becoming criterion variables. Some of the earliest uses of IQ scores as potential criterion variables are found outside of psychological research, such as in the eugenics-inspired efforts to limit immigration into the United States (see, e.g., discussion in Fancher, 1985). As noted earlier, the data from the Army Alpha experience resulted in many controversies. The discussion occasionally touched on IQ test scores as providing revealing information about various national origin or ethnic groups, but there was no consequential empirical treatment of academic achievement or work performance issues—the IQ became the criterion instead of the predictor.
One of the most striking early research programs that focused on changing IQ from a predictor variable to a criterion variable was the series of studies by Wellman and her colleagues in the 1930s (see, e.g., Wellman, 1940). In these studies, the authors purported to demonstrate that the IQ of children, especially those with initially low IQ scores, could be raised as a consequence of attending a university-run nursery school. The essence of the studies (and the controversies associated with documented methodological and statistical issues; see Goodenough & Maurer, 1940; McNemar, 1940) focused exclusively on changes of IQ scores from pretest to posttest as the criteria for raising intellectual abilities of the young children in the study. That is, nowhere in these extensive discussions was the presence or absence of any concomitant effects of the nursery school program on later academic performance. Thus, the IQ scores again had become the criterion, in and of themselves, rather than serving as predictors for academic success or failure. In hindsight, the important research question was the one that was not asked: Did the nursery school have an effect on later academic success rates? Whether the IQ increased, on average, is an imperfect, indirect measure of gains in intellectual ability. One could very reasonably assert (as McNemar and others did) that IQ changes could have partly been a result of practice effects on the tests. If all that changed as a function of nursery school attendance was IQ test performance, there would be little to report, except that the tests were perhaps no longer useful as predictor measures, given repeated exposure to testing on the same items.
As the Wellman et al. research indicates, though, making IQ scores the criterion of interest removes any consideration of what the IQ test scores actually mean. The implicit belief is that, if there is a relationship between IQ scores and academic success, then raising IQ should lead to higher levels of academic achievement (an argument explicitly discussed by Jensen, 1969, in his otherwise controversial article, “How Much Can We Boost IQ and Scholastic Achievement?”) But the correlations between IQ scores and academic achievement measures, while substantial (see, e.g., Anastasi & Urbina, 1997), are certainly not even close to unity. Moreover, in the aggregate there is little evidence to suggest that such a causal relationship exists between IQ test score change and academic achievement change. Or if such an association were to be found, it is unknown whether it might instead be the result of academic achievement increases causing changes in IQ scores, or some third variable intervention having an effect on both IQ scores and academic achievement. 2 The extant research suggests that not attending school has a negative effect on both academic achievement indicators and IQ test scores (see, e.g., Ceci, 1991).
The effect of changing intelligence test scores from predictor variables to criterion variables is that researchers have lost sight of what intelligence is. As Wechsler (1975) noted, intelligence tests are “only a means to an end.” The critical issue is to determine the individual’s capabilities to “understand the world about him [/her] and his [/her] resourcefulness to cope with its challenges” (p. 139). When IQ (or other intelligence test) scores become the criteria, one is implicitly assuming that such scores are the essence of intelligence, but as Wechsler argued, how one performs on these various puzzles and tasks is not the same thing as acting intelligently at work, at home, or in the society at large. Also, given that IQ-type tests have much smaller correlations with indicators of these behaviors for adults (Schmidt & Hunter, 1998), when compared to predicting academic success of children and adolescents, equating such scores to be representative of adult intelligence is a somewhat tenuous proposition.
Ultimately, the problem is that several programs of research have essentially adopted intelligence tests themselves as criterion indicators of adult intelligence, without any reference to the behaviors of the individuals outside of the laboratory. Even more pernicious, perhaps, is the selection of a very limited set of tests (e.g., Raven’s Progressive Matrices) as the sine qua non tool to assess adult intelligence, even though there is extensive research that supports the notion that the Raven is a relatively poor sampling of the content of a test such as the WAIS (Burke, 1958), and arguably even less representative of the intellectual challenges adults actually encounter in their day-to-day work and home life. One implicit justification for the use of nonverbal reasoning tests such as the Raven is that it is characteristic of the kinds of tasks that experimental psychologists use in the laboratory ever since Ebbinghaus’s (1885/1913) use of nonsense syllables as learning stimuli. That is, by making the stimuli either extremely familiar or extremely novel, experimental psychologists attempt to minimize the contribution of individual differences in knowledge and skills that might affect task performance. Such an approach is in keeping with a reductionistic perspective: one that has as an ultimate aim of boiling down the construct of intelligence into a single construct or entity. The history of intelligence testing provides an interesting contrast between reductionists (e.g., Spearman, 1904) and antireductionists (e.g., Guilford, 1967). The consensus opinion among theorists is that traditional measures of intellectual abilities are hierarchically organized (as in the Cattell-Horn-Carroll model—see Schneider & McGrew, 2012), but the general factor itself accounts for only 20% to 40% of individual differences in intellectual abilities (Vernon, 1950), and narrow tests (such as Raven’s Progressive Matrices; see Burke, 1958, 1985) or simple information processing tasks (Deary, 2000) are, at best, markedly inferior estimators of general intelligence, compared to omnibus tests based on the Binet-Simon approach.
Ultimately, reductionistic strategies that focus on narrow aspects of intelligence may be reasonably justified for some learning and information processing studies, but for a construct such as intelligence, which in the real world is almost always contextually embedded, such an approach is more than likely to miss important sources of variance that lead to better or poorer task performance outcomes.
One recent example is especially illustrative. In a meta-analysis and review of “brain training programs,” Simons et al. (2016) reported that the vast majority of the studies had little or no evaluation of effectiveness of such programs beyond practice on the tasks themselves or near-transfer to similar tasks. A small number of the studies examined effects on tests of intellectual ability (but none examined intelligence at the level of thoroughness of the Stanford-Binet or Wechsler scales), and only a single large-scale study of elderly participants involved examination of “everyday task performance” and “everyday problem solving.” The results of that particular study were assessed to be inconclusive in terms of objective performance changes in real-world task performance. The authors’ conclusions from the review is that brain-training programs have little transfer effects on the limited samples of intelligence-type tests and no demonstrable effects on real-world behaviors that involve intellectual challenges. Even if these research programs were to focus on transfer to omnibus intelligence tests, such as Wechsler’s scales, it is, as argued above, beside the point. If the training effects were to result in IQ increments, but no noticeable change in success at daily work and nonwork activities (see, e.g., Tucker-Drob & Salthouse, 2011), then why would one recommend engaging in brain training to begin with? The existing corpus of psychological research on learning and transfer over the past hundred years suggests instead that if one wants to improve performance at a particular task, then the best strategy is to practice and engage with that task. Engaging in so-called “brain training” in the hope that one could be a better foreign-language speaker is highly unlikely to be as effective as actually practicing the foreign language itself—a point entirely consistent with Thorndike and Woodworth’s (1901) classic evaluation of learning and transfer effects.
The Criterion Problem
The elephant in the room for adult intelligence assessments has been the question of what criterion or criteria should be used for validating the assessments? The Binet-inspired tests give us little useful guidance, given that they were created for and have been used mainly for prediction of academic performance of children and adolescents. College and graduate entrance exams provide a good example of the problem in scaling-up to adult intellect, in that the dominant theme of such tests is to sample mainly knowledge/skills/abilities that were acquired by the individuals in high school, when there is some common core curriculum (i.e., this is what Cattell, 1971/1987, termed “historical crystallized intelligence”). Thus, the math content of the SAT, ACT, and GRE tests is limited to algebra and geometry—when most examinees are many years removed from such topics, while other examinees have gone far beyond these topics in the domain of mathematics knowledge and skills. In that sense, it should come as no surprise that such tests provide useful predictive criterion validity but leave substantial individual-differences variance in academic criterion performance unaccounted for.
Early adult intelligence tests (most notably the Army Alpha) were designed with the hope that they would predict individual differences in occupational performance, first in the military, and later to businesses (and for a period of time, for college/university academic selection—see Whipple, 1922). Later adult intelligence tests, such as the Wechsler scales, found important uses in the diagnosis of neurological incidents or illnesses—a theme that carries into consideration of older-aged adults in the Mini-Mental State assessments, which are clearly related to performance of older adults on basic tasks of everyday functioning (see, e.g., Tucker-Drob, 2011). Other measures of intelligence, such as multibattery tests like the General Aptitude Test Battery, Armed Services Vocational Aptitude Battery, and so on, do have respectable psychometric properties, but their actual validities for predicting occupational performance are relatively limited, typically accounting for only about one-quarter of the variance in occupational performance measures (see, e.g., Schmidt & Hunter, 1998). Criteria beyond occupational performance include myriad variables ranging from physical health and mortality (e.g., Batty, Gale, Tynelius, Deary, & Rasmussen, 2009) to mental health (e.g., Gale, Batty, Tynelius, Deary, & Rasmussen, 2010). In many cases, IQ and related intelligence composites show significant positive associations with indicators of physical and mental health (consistent with assertions made by E. L. Thorndike, 1940). Although these findings support the idea that traditional measures of intelligence are related to important outcomes—it has not been established that efforts to raise intelligence levels in adults translate to positive effects on health and related variables. Additional criteria that have been investigated include things like socioeconomic status and annual income—interesting on one level, but clearly inadequate for any serious consideration as the major criteria for adult intelligence, because there are examples why many or even most adults might not see income level as the ultimate source of personal intellectual achievement.
Domain Knowledge and Expertise
Finally, Ackerman (1996) has argued that occupational success and avocational competencies are largely related to individual differences in domain knowledge, which in many cases may have been acquired long before the assessments of adult intellect are made (e.g., the domain knowledge and expertise of doctors or professors may have been mostly acquired decades before the adult is assessed on current intellectual functioning). Expertise—defined as an integrated set of knowledge and skills is the foundation on which many adults perform their daily work activities and may also be instrumental in avocational activities ranging from completing crossword puzzles to discussing current events. Expertise can include declarative (factual) knowledge, procedural knowledge (how to do a task), or tacit knowledge, and typically includes some combination of all three types of knowledge. There are four aspects of expertise that should be noted with respect to adult intellectual development and expression. First, acquisition of expertise inevitably requires investment of intellect over an extended period of time, which is arguably why it is much easier to acquire expertise during adolescence and young adulthood, compared to middle age and beyond. Second, expertise is generally domain-limited. That is, expertise in one domain (e.g., carpentry) does not readily transfer to other domains (e.g., neurosurgery). Third, once acquired, expertise tends to be resistant to degradation, except from lack of use over a long period of time. Finally, once a foundation of knowledge and skills has been developed, an individual can be expected to acquire new knowledge in the same domain with much lower intellectual investment, compared to a novice learner. Individual differences in domain knowledge and expertise are likely to provide a partial, but not complete view of the current intellectual functioning of adults, except when advanced ages are reached, and the primary issues relate to skill maintenance or forgetting phenomena (for an extended discussion of these issues, see Ackerman, 2008).
The Challenge of Defining and Assessing Adult Intelligence in the 21st Century
There appear to be two related considerations missing from a more complete assessment of adult intelligence: (a) a recognition that in the 21st century, what constitutes intellect across adult lives may be different from what was deemed to constitute intelligence across most of the 20th century; and (b) a better appreciation/assessment of the intellectually demanding tasks that adults are likely to encounter in their day-to-day work and nonwork lives. Each of these is discussed in turn.
Before proceeding with considering the construct of adult intelligence in the 21st century, it is worth highlighting that intelligence assessments for children have not changed markedly over the last century in both content and criterion-related validity, partly because the common elementary and early secondary school curricula have not fundamentally changed from that of the early 20th century. The “three Rs” (reading, writing, and arithmetic) are still core competencies for early school instruction. The content of physical and social sciences curricula has changed sufficiently, such that an IQ test from the 1910s (e.g., the Army Alpha) would be somewhat incomprehensible, and an assessment of academic knowledge from the 1930s (see, e.g., Learned & Wood, 1938) would yield abysmally low scores among today’s adolescents. However, the core competencies that make up a modern Binet-inspired IQ test are largely the same as those that were highly predictive of child and early adolescent academic success in the first part of the 20th century.
Technology and intelligence
Consideration of arithmetic/math abilities provides a glimpse of the issues involved in defining intelligence for adults. One hundred years ago, academic math tests required students to compute answers to test items either mentally or with the aid of paper and pencil. Today, especially when considering tests of algebra, geometry, trigonometry and more advanced mathematics, computers and electronic calculators are used in the classroom and calculators are allowed on many tests of math ability, including college entrance examinations. A high school senior from 1930 confronted with a logarithm problem would perhaps be almost as befuddled by a computer or an electronic calculator as a high school senior from 2016 would be with a printed table of logarithms and mantissas.
In a prescient article, published just two years after the invention of the World Wide Web, yet before most of the public had much access to, or use for, the Internet, Salomon, Perkins, and Globerson (1991) outlined the potential for technological advances to either enhance or degrade (deskill) human intellect. The authors’ hopes were that advanced technology, especially in the classroom, could yield “cognitive residue” that represented an enhancement of the learner’s capabilities. Yet, they also acknowledged that as future technology developed, there was a substantial risk of degrading abilities, once the technology replaced the need for human intellectual operations to yield effective results.
In an earlier period (roughly from 1972), when the first inexpensive handheld calculators became available for the public, there was extensive discussion in the educational community about whether to allow calculator use in the mathematics classroom. Ultimately, the ubiquity of calculators, and later computers, in the classroom has made the controversy a wistful footnote. However, reviews and meta-analyses conducted on the question of whether calculator use had an effect on student achievement yielded results that suggested positive effects—but only when students could use calculators during tests (Ellington, 2003; D. M. Roberts, 1980). In the terminology of Salomon et al., the “cognitive residue” beyond what could be accomplished with the technology was quite small or nonexistent.
Defining Adult Intelligence
To attempt a current definition of adult intelligence, perhaps it is worthwhile to consider Thorndike’s (1921) proposition that “realizing that definitions and distinctions are pragmatic, we may then define intellect in general as the power of good responses from the point of view of truth or fact” (p. 124). From this perspective, it seems somewhat quaint to think that many of the items on existing intelligence tests are actually representative of intelligence to the degree that they demand processes that resemble how adults actually reach “good responses” in the current environment. Questions of general knowledge (common cultural information) can be readily answered by most adults in the space of a few seconds to search the Internet (e.g., “Who was Neil Armstrong?” “Who wrote the play Waiting for Godot?”). Similarly, although adults of a “certain age” can easily add, subtract, multiply, or divide a string of numbers in their heads or with paper and pencil, less error-prone (and typically faster) answers can be derived with a calculator, computer, or the ubiquitous smartphone application. So what does it mean to be “intelligent” in this context? Do we give credit only to those who can retrieve correct information from long-term memory, or is the person who has achieved the requisite skills at rapid information search and retrieval from the Internet or other sources equal in ability? How would different adults perform on an “open-book” test of intelligence, where they could attempt to answer the test questions from their own memory and mental skills, but they also had access to the kinds of tools they typically use when confronted with similar tasks on a day-to-day basis?
Some items on current intelligence tests will be resistant to online solutions. Items of abstract reasoning, spatial ability, and comprehension are not readily answered with these kinds of external aids, but the nature of some of these items (e.g., abstract reasoning) essentially raises the question of their relevance to the expression of adult intelligence. This leads to the main issue: What activities do adults typically confront that require intelligence? Table 1 provides a few examples along these lines, illustrating the content of real-world intellectually demanding tasks and changes that have taken place over the past 50 years. For an applied geometry example, the task might involve determining the number of ceramic tiles necessary to cover a shower stall, or estimating how much paint will be needed to cover a room with two coats. Faced with such problems, many adults would refer to Internet sources, rather than attempting to work out the answers in their heads or use an electronic calculator. Or, consider the installation of a graphics board in a computer or needing to fix a lawn mower. Adults unfamiliar with either of these tasks are more likely to search YouTube for an audio and video demonstration of how to accomplish the task, rather than to attempt to solve the problem by first reading through a manual. Under such circumstances, the intellectual task the adult faces is to take note of the critical steps in the demonstration video and work that into an action plan for accomplishing the goal. One might expect that higher intelligence adults will require fewer viewings or review occasions with the YouTube video, compared to lower intelligence adults, but this is an empirical question for which there are not adequate data yet to draw a conclusion.
Example Work and Nonwork Intellectually Demanding Tasks and Changes Over the Past Half Century
In addition, transfer of training, as noted by Ferguson (1956), is another critical component of modern adult intellect. When confronted with a new self-service checkout system at a grocery store, higher intelligence adults might be faster and more effective in transferring their prior knowledge and skills to the new system. For children, higher levels of intelligence are associated with more effective “far transfer” (when the to-be-learned task is only remotely related to the individual’s existing knowledge), though not with “near transfer” (when the to-be-learned task is similar to the individual’s existing knowledge—see Sullivan & Skanes, 1971). As noted earlier, the brain-training literature does not support far transfer effects in adults, so there remain critical questions: What are the boundary conditions for real-world transfer effects in adults? And what role is played by individual differences in traditional or other measures of intellect in these aspects of transfer of training and knowledge?
Intelligence applied
One theme that arises incidentally across many domains (e.g., wisdom, rationality) and explicitly in others (e.g., Sternberg’s conceptualization of “foolishness”; see Sternberg, 2005) is the observation of a mismatch between an individual’s level of intelligence and incidents of unintelligent actions performed by the same individual. That is, the question is often asked, “Why do smart people do stupid things?” Leaving aside issues of motivated ignorance, 3 it seems that two of the most important issues have to do with a lack of insight into transfer of training or transfer of knowledge, and/or a lack of interest in expending intellectual resources for problem solving. When considering adult intelligence in action in the real world, in contrast to performance in the laboratory or in high-stakes testing situations, the field would be well served by attending to an individual’s desire to engage problems with investment of his or her intellect. Whether this is an aspect of adult intelligence or more of personality (e.g., the Goff & Ackerman, 1992, conceptualization of “typical intellectual engagement”) remains an open question.
Improving Assessment of Adult Intellectual Abilities
To build a more comprehensive assessment of adult intellectual abilities, one might start with the basic proposition of Demming and Pressey (1957)—that is, select/create intellectual tasks that are “indigenous” to adults. Such an approach, though, requires that we delineate three different types of knowledge/skills/abilities: (a) content/processes that adults “need” to know—that is, to have either the information at their fingertips or be able to solve a problem quickly using only their mental processes (e.g., perhaps knowledge of current events, ability to memorize names and personal details of newly met others, reading comprehension); (b) important or useful content/processes that adults “should” know/be able to do—that is, things for which possessing the knowledge/skills/abilities for problem solution will either result in solutions that are faster, more accurate, or both—when compared to individuals who do not have these capabilities/skills (e.g., planning a budget with an Excel spreadsheet, performing cost-benefit analyses for purchases online vs. in a brick-and-mortar store; planning a meal for a dinner party with an odd number of attendees; evaluating and weighing online reviews of multiple product choices), along with elements of rationality abilities and skills used in critical thinking; and finally (c) knowledge/skills/abilities that adults either never/rarely (at least in the past 100 years) needed to solve the kinds of problems they are likely to confront (abstract reasoning, some working memory tasks), or knowledge/skills/abilities that were once useful/required to solve intellectually demanding problems in the past century, but are like vestigial organs today (perhaps, e.g., tasks like finding square roots of large numbers by hand, reading a road map, or mechanical drawing).
In the final analysis, the overarching message of this article is that, after 100 years of modern intelligence assessment, many researchers who study adult intelligence have adopted a subset of traditional assessments, originally designed for children and adolescents, and validated as predictors for academic achievement—turning them from imperfect predictor measures to idealized criterion measures of intelligence. The fundamental problem with this approach is an incomplete understanding of what constitutes adult intellect and an inability to generalize results to behaviors that are directly relevant to intellectual challenges in the day-to-day activities of adults. What the field needs is a more comprehensive understanding of the intellectual tasks adults are likely to encounter, assessments that reflect these tasks, and validation of findings from various sources (e.g., fMRI, working memory investigations, brain-training effects) against the criteria of day-to-day intellectually demanding activities. Predicting individual differences in these behaviors can be expected to yield better understanding of the construct of adult intelligence as fundamentally and qualitatively different from child intelligence, more comprehensive understanding of the effects of aging on adult intellectual development, improved prediction of individual differences in work and nonwork behaviors, 4 and satisfactory criteria for evaluating the effectiveness of interventions designed to improve adult intelligence. Consistent with Wechsler (1975), the main point here is that it is critical to keep in mind that what we are interested in as psychologists is not the test scores themselves, but the adults’ capabilities to act intelligently in the world outside of the laboratory or examination room.
Footnotes
Declaration of Conflicting Interests
The author declared no conflicts of interest with respect to the authorship or the publication of this article.