This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation

Domain-Specific Knowledge as the "Dark Matter" of Adult Intelligence

Gf/Gc, Personality and Interest Correlates

^a School of Psychology, Georgia Institute of Technology, Atlanta

Phillip L. Ackerman, School of Psychology, Georgia Institute of Technology, Psychology Building, 274 5th Street, Atlanta, GA 30332-0170 E-mail: phillip.ackerman{at}psych.gatech.edu.

Decision Editor: Toni C. Antonucci, PhD

	Abstract

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
An enduring controversy in intelligence theory and assessment, the argument that middle-aged adults are, on average, less intelligent than young adults, is addressed in this study. A sample of 228 educated adults between ages 21 and 62 years was given an array of tests that focused on a broad assessment of intelligence-as-knowledge, traditional estimates of fluid intelligence (Gf) and crystallized intelligence (Gc), personality, and interests. The results indicate that middle-aged adults are more knowledgeable in many domains, compared with younger adults. A coherent pattern of ability, personality, and interest relations is found. The results are consistent with a developmental perspective of intelligence that includes both traditional ability and non-ability determinants of intelligence during adulthood. A reassessment of the nature of intelligence in adulthood is provided, in the context of a lifelong learning and investment model, called PPIK, for intelligence-as-Process, Personality, Interests, and intelligence-as-Knowledge (Ackerman 1996).

WHILE a few researchers have been willing to address the issue directly (see e.g., Baltes and Schaie 1976; Horn and Donaldson 1976, Horn and Donaldson 1977; Labouvie-Vief 1977) across some 80 years of research on adult intelligence, one important implication of the findings in the literature is that middle-aged adults are, on average, less intelligent than young adults. This controversial implication of adult intelligence assessment was first raised in the 1920s by a nonpsychologist in the popular press (Lippmann 1922) and has stirred substantial disagreement ever since. A possible conclusion from the corpus of data is that middle-aged adults are, on average, less intelligent than the average 21-year-old, and older adults are, on average, less intelligent than younger adults (Wechsler 1958). One way of addressing the controversy has been to appeal to cohort differences as a possible explanation of lower scores among middle-aged samples (Schaie 1970, Schaie 1996; Schaie and Strother 1968; Flynn 1984). For example, results of longitudinal studies have tended to indicate that intelligence is much better preserved into middle age than the cross-sectional studies have indicated (Bayley and Oden 1955; see also a review by Bloom 1964). Other ways of addressing the controversy have included investigations of testing conditions or other artifacts (e.g., speededness of testing, speed vs accuracy of responses; see for example, the work by Jones and Conrad 1933, Miles 1934, and more recently, Schwartzman, Gold, Andres, Arbuckle, and Chaikelson 1987).

One way out of this apparent conundrum is to assert that middle-aged adults are equally intelligent as younger adults (or perhaps more intelligent than 21-year-olds), but that there are missing sources of individual differences in intelligence variance—domains that are not assessed by traditional measures. There have been several attempts to find this "dark matter" (see A), which at least in theory would shift the balance of the advantage of intelligence away from the 18–25-year-old group to an older group of adults. Theories of practical intelligence (Demming and Pressey 1957; Sternberg and Wagner 1989; Wagner and Sternberg 1985) and wisdom (Baltes and Staudinger 1993; Baltes, Smith, and Staudinger 1992) have been offered, but provide little demonstrable evidence for advantages to middle-aged adults. Similar approaches have taken a broader perspective of the intellectual content of adult life situations, especially in the context of adults aged 70 years and older (see Schaie 1977–78; Schaie 1978; Scheidt and Schaie 1978). Other approaches suggest that older adult intelligence is qualitatively different from the intelligence of younger adults. For example, some studies suggest the notion that adult intelligence is more heavily dominated by postformal reasoning—an ability that is more highly predicated on interpersonal skills and nonlinear problem solving (see Sinnott 1998 for an extensive discussion).

The major premise of the current perspective (Ackerman 1996) is that the locus of the "dark matter" of adult intelligence is in domain-specific knowledge—especially knowledge that is not common to a particular dominant culture. Such domains include: knowledge associated with academic study (e.g., science, music, art); knowledge associated with active engagement in society (e.g., knowledge about the operation of the government); knowledge about the world around us (technology, law); knowledge associated with occupations; and knowledge associated with avocational hobbies. These domains are differentiated from the kinds of general cultural knowledge assessed in traditional one-on-one IQ tests (such as the Information test of the Wechsler Adult Intelligence Scale). It has been proposed (Cattell 1957) that knowledge structures (crystallized intelligence, in Cattell's terms) may be as different between individuals as carpentry knowledge differs from chemistry, and as computer programming knowledge differs from child care knowledge. The proposition is that if adults are given credit for what they know—it will be possible to better assess individual differences in intelligence from a real-world perspective (e.g., improving prediction of academic and occupational success), better take account of the nature of lifespan changes in intellect, and finally, provide a direct assessment of whether middle-aged and older adults are indeed less intelligent, on average, compared with young adults (even in the face of cohort differences). That is, when the respective groups are given credit for those intellectual tasks they can perform—including, for example, abstract reasoning (the traditional fluid intelligence measures) and neurosurgery skills (something that extant intelligence tests cannot assess), a fair comparison between age groups may be possible.

In a more colloquial real-world sense, the current approach provides an explanation for the success of many middle-aged and older adults in a variety of occupations, such as medical surgery, aircraft piloting, carpentry, and so on, where it appears to be quite sensible to conjecture that there are many situations where it is far more efficient and more effective for the performer to "recall" the correct answer to a pending real-time problem, than it is to "generate" an optimal solution, based on deductive reasons. For example, when Michael DeBakey assisted in Boris Yeltsin's heart bypass surgery in 1996, the fact that Dr. DeBakey was 86 years old at the time perhaps had far less influence on his performance than the fact that he had, at the time, performed some 60,000 cardiovascular procedures (DeBakey 1999).

The proposed location for the "dark matter" of adult intelligence is simultaneously more revolutionary and more conservative than these other approaches. The revolutionary aspect of an approach to adult intelligence that gives broader credit for knowledge, is that mainstream adult intelligence research has come to focus on a factor of general intelligence (g), or fluid intelligence (Gf), or abstract reasoning, originally referred to by Spearman as eduction of relations and correlates (see for example, extensive theoretical and empirical discussions by Salthouse 1996; Jensen 1998). That is, extant theorizing about intelligence has come to discount domain-specific knowledge in favor of process aspects of intelligence, a perspective that exacerbates the belief that middle-aged adults are less intelligent than young adults. On the conservative side, placing knowledge into a broader consideration of adult intelligence is historically grounded (e.g., Cattell 1943; Hebb 1942); consistent with the broad specification of crystallized intelligence proposed by Cattell 1957; and it is concordant with Demming and Pressey 1957 research on the intelligence of adults. Moreover, it is concordant with numerous developments in cognitive psychology that have demonstrated the importance of knowledge structures in determining success or failure on both laboratory and real-world tasks (Chi and Ceci 1987; Chi, Glaser and Rees 1982; see Voss, Wiley, and Carretero 1995 for a review). This approach is also concordant with the artificial intelligence literature, which has nearly universally determined that domain-specific knowledge is a major determinant of the success of so-called expert systems (Crevier 1994). An approach to adult intelligence that focuses on knowledge is even concordant with lay understanding of the term "intelligence" (see Goodnow 1980; Sternberg, Conway, Ketron, Bernstein 1981). My view (see Ackerman 1996) emphasizes the definition of intelligence offered by Henmon 1921, that is: "Intelligence, then, involves two factors—the capacity for knowledge and knowledge possessed" (p. 195). From this perspective, efforts toward focusing on Gf to the exclusion of other factors, especially knowledge, may overlook significant and substantial aspects of adult intelligence. These justifications for assigning an important role to knowledge in the application of intelligence in both adult educational and occupational settings are extensive (Ackerman 1996, Ackerman 1998). A brief description of the theory, called PPIK (for intelligence-as-Process, Personality, Interests, and intelligence-as-Knowledge) is provided below.

	PPIK Theory

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
The PPIK theory (Ackerman 1996) is derived from the work of Hebb 1942 and Cattell 1943, who divided intelligence into a physiological-based set of abilities (Intelligence A and fluid intelligence [Gf], respectively) and a set of educational/experiential abilities (Intelligence B and crystallized intelligence [Gc], respectively). According to these views, Gf shows greater declines in performance across the adult age span, whereas Gc is much better preserved, with adults showing gains on Gc well into middle age. However, it is important to note that the greater losses in Gf during middle age do not fully offset gains in traditional Gc measures. Thus, a composite with equal weighting of Gf and Gc measures will typically show lower scores among 40-year-olds than 20-year-olds (e.g., see Wechsler 1958, Figure 5). Although Hebb did not provide much follow-up to his ideas of intelligence, Cattell 1957 made a critical distinction between historical crystallized and present crystallized intelligence Gc_Historical and Gc_Present: a distinction between what knowledge persons acquire in high school (and retain, to whatever degree, in adulthood) and the range and depth of knowledge that persons have when actually tested. The traditional approach to assessing Gc has been to measure mainly individual differences in Gc_Historical (the kinds of knowledge and skills tapped in the SAT and GRE general tests). Only when knowledge assessments, such as domain-specific achievement tests (e.g., the GRE subject tests), are administered to adults, does the investigator have a direct measurement of Gc_Present. Moreover, because adults differ substantially in academic, occupational, and avocational knowledge, it is critical to conduct a wide assessment of knowledge, rather than administer a single in-depth domain-specific test.

Cattell's investment hypothesis (Cattell 1971/1987) provided a rationale for considering knowledge in a larger context of non-ability traits, such as personality and interests. In the current context, investment of cognitive (ability), affective (personality), and conative (motivational) resources is what drives the acquisition and maintenance of domain-specific knowledge over the lifespan. These influences can be seen in an illustration of the PPIK theory, shown in Fig. 1. Consistent with Cattell's theory, intelligence starts off as a relatively amorphous Gf (or in terms of the PPIK theory, intelligence-as-Process [g_p]). Personality traits (such as Typical Intellectual Engagement and Openness to Experience) and interests (especially Realistic, Investigative and Artistic) together determine the intensity and direction of intellectual investment that is available to the individual. The traits of interests are believed to be both influenced by and in turn influence respective ability levels (see Holland 1959). The momentary intensity of intellectual effort is arguably determined by intelligence-as-process; but, the cumulative investment of intellectual effort is proposed to substantially influence the development of intelligence-as-knowledge, as a result of a series of individual choices. The result, for adults through middle age, is a differentiation of knowledge according to the direction and intensity of effort.

View larger version (25K): [in this window] [in a new window]   Figure 1. Illustration of Ackerman's PPIK theory, outlining the influences of intelligence-as-Process, Personality, Interests, and intelligence-as-Knowledge during adult development, covering academic and occupational knowledge. Arrows represent correlational influence. The representation indicates that measured fluid intelligence (Gf) develops out of intelligence-as-process (gp), and that measured crystallized intelligence (Gc) develops out of (or is a consequence of) intelligence-as-knowledge (gk). Interests (Realistic, Investigative, and Artistic) and personality traits (Openness and TIE) are influenced by intelligence to some degree, and in turn, influence knowledge. Adapted and expanded from Ackerman 1996. (Adapted from Ackerman, 1997, Figure 5, Copyright Duke University Press).

View larger version (26K): [in this window] [in a new window]   Figure 2. Trait complexes, including abilities, interests, and personality traits showing positive commonalities. Shown are: (1) Social, (2) Clerical/Conventional, (3) Science/Math, and (4) Intellectual/Cultural trait complexes. (Figure 7 on Page 239 from Ackerman and Heggestad 1997, "Intelligence, personality, and interests: Evidence for overlapping traits." Psychological Bulletin, 121, 219–245). Copyright American Psychological Association. Reprinted by permission.

1. Are middle-aged adults more knowledgeable than younger adults (or at least equally knowledgeable)? Or more generally, what is the relationship between age and individual differences in knowledge?
   
2. Are individual differences in knowledge well accounted for by traditional measures of Gf, Gc or both?
   
3. What are the non-ability correlates of individual differences in knowledge?
   

To address these questions we developed a set of scales (with the assistance of the College Board) that assessed knowledge across several domains: arts and humanities, physical and social science, civics, and technology (Rolfhus 1998; Rolfhus and Ackerman 1999). In addition, we administered these scales to normative samples of college students and to a group of 135 middle-aged adults (Ackerman and Rolfhus 1999). In general, the middle-aged adults did not perform as well as the young adults on standardized tests of math and spatial abilities and performed better than the young adults on the verbal ability tests. The middle-aged adults performed, on average, better than the younger adults on most of the knowledge tests. The knowledge scales that held the highest advantage for the middle-aged adults were in the domains of arts and humanities, civics, and technology. The knowledge scales that showed the least relative advantage to the middle-aged adults were in the physical sciences (such as chemistry, physics, and biology). Within the group of middle-aged participants (aged 30 to 59 years), there were no significant negative correlations between knowledge scale test performance and age. Moreover, although significant correlations were found between a higher-order general ability (g) factor and performance on many knowledge tests, verbal ability (independent of g) was also significantly and substantially correlated with many knowledge tests. This indicates that knowledge is more than gsomething.

Nonetheless, there were limitations of the Ackerman and Rolfhus 1999 study in evaluating the age and ability correlates of individual differences in knowledge. First, although a higher-order g was determined and verbal ability measures were used to identify traditional assessment of Gc, measures were not selected to specifically assess Gf versus Gc. Second, the middle-aged adult sample was highly heterogeneous in educational background; older and younger sample comparisons may have yielded artifactual results because of the large educational differences between the two groups.

	The Current Investigation

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
The current study is a refinement and extension of the previous norming study (Rolfhus and Ackerman 1999) and the middle-aged comparison sample (Ackerman and Rolfhus 1999). First, in the current study, a battery of ability tests was administered that were explicitly chosen to assess Gf and Gc factors. Second, I decided to focus on a population that was more homogeneous in educational background than the participants in the Ackerman and Rolfhus 1999 study, by requiring that all participants had obtained an education at the baccalaureate level.

Third, relevant personality traits were assessed at a narrower level of specificity than the previously used Five-Factor Model measures. A previous study (Ackerman and Rolfhus 1999) had indicated that both Extroversion and Conscientiousness as measured by the Costa and McCrae 1992, NEO-Five Factor Inventory (NEO-FFI) showed negative correlations with several knowledge scales. In contrast, positive correlations were found between Openness and knowledge in arts and humanities, and several positive correlations were found between the Goff and Ackerman 1992 TIE scale and knowledge. The negative results for Extroversion were unanticipated (based on the Ackerman and Heggestad 1997, meta-analysis), and the Conscientiousness correlations, though consistent with the meta-analysis, were somewhat counter to the relations expected on the basis of the industrial/organizational psychology literature (Barrick and Mount 1991; Tett, Jackson, and Rothstein 1991). One possibility to be considered is that the NEO-FFI instantiation of both factors is somewhat convoluted. In the case of extroversion, it seemed possible that social closeness and social potency (or dominance) are confounded. In the case of conscientiousness, it seemed possible that plodding aspects of the trait may have dominated the measure, rather than other aspects. To shed more light on these issues, I selected a different set of personality scales that explicitly separated these facets of the broader trait configurations.

Based on the previous investigations and the PPIK theory, the following predictions were made:

1. First, adult age would be negatively and substantially correlated with scores on Gf tests, and positively (but less substantially) correlated with Gc tests. Note that this is a relatively straightforward prediction in accord with the extant literature and the fact that this is a cross-sectional study subject to both aging and cohort effects (e.g., Birren and Morrison 1961).
   
2. Overall, knowledge scale performance would be more highly correlated with Gc than with Gf. Higher positive correlations would be found between Gc and knowledge in the arts and humanities, civics, and technology. The highest relative correlations for Gf would be found in the physical sciences.
   
3. Age would be positively correlated with performance on the knowledge scales, with the largest correlations in the domains that are most highly related to Gc (e.g., humanities). In contrast, age correlates of knowledge would be lowest for knowledge scales most highly correlated with Gf.
   
4. One additional construct is found at the interface between Gc and personality, called Typical Intellectual Engagement (TIE) (Ackerman, 1994; Goff and Ackerman 1992). TIE refers to the degree to which individuals prefer to engage in intellectually demanding leisure tasks, such as reading, and attending cultural activities such as concerts, lectures, and the theater. In concert with the PPIK theory and the trait complexes shown in Fig. 2, Typical Intellectual Engagement was hypothesized to be positively and significantly related to knowledge scales in the Humanities, with smaller positive correlations for Civics, and negligible correlations with physical sciences knowledge. Furthermore, correlations between social closeness and traditionalism were expected to be negatively and substantially correlated with knowledge, whereas absorption was predicted to have positive and significant correlations with humanities knowledge (especially music and art). Social potency was predicted to have negligible correlations with knowledge (as a contrast to social closeness). See Ackerman and Rolfhus 1999 and Rolfhus and Ackerman 1996 for additional empirical evidence specific personality trait-knowledge associations.
   
5. A concordant pattern of correlations would be found between interest traits, personality traits, and both knowledge and abilities, consistent with the PPIK theory (Ackerman 1996) and the recent meta-analysis by Ackerman and Heggestad 1997, as follows:
   

a. Artistic interests were predicted to have the highest correlations with knowledge in arts and humanities.

1. Investigative interests would have the highest correlations with knowledge in the physical and social sciences.
   
2. Realistic interests would have significant positive correlations with knowledge scales that included technology items (e.g., Electronics, Astronomy).
   

Thus, the major aims of current study are to evaluate age differences in intellect-as-knowledge across several domains, to evaluate the relations between Gf/Gc and individual differences in knowledge, and to assess the role of non-ability traits (namely selected personality and interest traits) in explaining individual differences in knowledge.

	Method

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
Participants
Participants were recruited through flyers and newspaper advertisements at the University of Minnesota. Two hundred thirty individuals participated in the study; the data from two participants were excluded because of a failure to follow instructions. Of the 228 participants in the final sample, 78 were men and 150 were women. The sample had a mean age of 34.2 years, SD age of 10.6 years, and a range from 21 to 62 years of age. Participants were required to have attained at least a B.A./B.S. level of education, have normal or correct-to-normal hearing, vision, and motor coordination. Thus, 100% of the sample had attained education at least at the baccalaureate level. The distribution of educational attainment reported by the participants was: B.A./B.S. only (75%), M.A./M.S. level (19%), J.D. (2%), M.D. or D.V.M. (1%), Ph.D. (4%).

Apparatus
Paper and pencil tests were administered at tables, in a laboratory room with up to 16 participants at a time. Instructions and start/stop timings were administered with prerecorded minidiscs over a public address system. Knowledge tests were administered using IBM-compatible Pentium computers with standard keyboards and monitors, running under Windows 3.1 or Windows NT. Up to 18 participants were tested at a time in individual carrels. The participants were instructed to complete the take-home questionnaire at home in a quiet, undisturbed environment.

Measures
Ability battery.
The ability battery contained 14 tests, selected to provide assessment of Gf and Gc. The tests are listed and described in Table 1 . To assess Gf, seven tests were administered. First included was the traditionally embraced Raven Advanced Progressive Matrices (although there are some reservations about the assertions that this test univocally assesses g—see Burke 1958). Other reasoning tests were included: Spatial Analogy, Number Series, and Diagramming Relations; also administered were an auditory Number Span test, and two math problem solving tests, called Problem Solving and Necessary Facts.

Personality measures.
Selected personality traits were assessed with scales selected from the Tellegen 1982 Multidimensional Personality Questionnaire, including measures of Social Closeness, Social Potency, Absorption, and Traditionalism. In addition, the Goff and Ackerman 1992 TIE questionnaire was administered. These measures are described in Table 1 .

Interest measures.
Interests were assessed with the 90-item Unisex Edition of the American College Testing Interest Inventory (UNIACT—Lamb and Prediger 1981). The inventory consists of Likert-type items to assess the six interest themes identified by Holland 1959, Holland 1973. Participants were asked to rate specific job tasks on a scale ranging from strongly like to strongly dislike. The six themes include: Realistic, Investigative, Artistic, Social, Enterprising, and Conventional interests. These measures are described in Table 1 .

As part of another study, additional non-ability measures were administered, but are not reported here.

Knowledge Scales
Development of the knowledge test battery is described in Rolfhus 1998 and Rolfhus and Ackerman 1999. The battery is based on a series of tests provided by the College Board (Advanced Placement [AP] and College Level Experience Program [CLEP] tests). Extensive item development, testing and validation was undertaken with over 700 individuals. The result is a series of objective knowledge tests, administered in a power format. Items covered the entire range of difficulty. The easiest items were designed to be easily completed by all participants; the most difficult items were intended to be beyond reach for all participants (except perhaps by career experts in that domain). Tests began with the easiest item in a domain. Items were presented in order of difficulty. When an individual responded incorrectly to three consecutive items, he or she was moved to the next domain. An individual's score was the number of items completed correctly.

Item presentation was accomplished with a program written specifically for the purpose. Items were displayed as bit-mapped files. Audio files were also played over headphones for the music test. Different response types (e.g., True/False, fill-in-the-blank, numerical response) were collected, depending on the requirements of particular items. Most important, the program scored responses on-line. It recorded verbatim input, response time, and whether the response was correct or incorrect. It also tracked the number of consecutive incorrect answers. When this count equaled three, the program terminated questions in that knowledge test and moved on to the next test.

Eighteen domains were selected from a set of 20 used in Ackerman and Rolfhus 1999 and Rolfhus and Ackerman 1999. The tests represent domains of physical and social sciences, arts and humanities, Western civilization, technology, business and law. They are described in Table 2 .

	Results

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
The analyses are presented in six sections, as follows: (1) An examination of the classification of ability tests into Gf and Gc markers; (2) Relations between knowledge scale scores and abilities; (3) Factors and composites of the Knowledge tests; (4) Relations between personality traits and knowledge tests; (5) Relations between interests and knowledge tests; and (6) A multivariate view toward evaluating the relative influence of abilities, personality, interests and demographic variables in predicting individual differences in knowledge.

Ability factors.
Initial analysis of the ability tests is necessary to examine whether an appropriate representation of Gf and Gc ability factors has been obtained. Means, standard deviations, and cross-correlations of the ability tests are shown in Table 3 . To assess the factors underlying the ability tests, the correlation matrix was first subjected to a principal axis factor analysis, with squared multiple correlations as initial communality estimates. Parallel analysis (Humphreys and Montanelli 1975; Montanelli and Humphreys 1976) was used to evaluate the number of latent factors. The parallel analysis clearly indicted that two factors should be extracted from the matrix, concordant with the initial selection of tests as markers for two primary ability factors. The factors were rotated to an oblique simple-structure criterion, using the Tucker and Finkbeiner 1981 Direct Artificial Personal Probability Factor Rotation (DAPPFR) procedure. The resulting factor pattern matrix is also shown in Table 3 . Examination of the factor patterns clearly allows the identification of a fluid intelligence factor (Gf), dominated by the Raven Progressive Matrices test and the Spatial Analogy test, as well as the numerical reasoning tests. The second factor is also easily identified as crystallized intelligence (Gc), dominated by Vocabulary, the Wechsler Information test, and the verbal tests. The factors are substantially correlated , a finding that is concordant with most extant analyses of abilities in adults of varying ages (e.g., see reviews by Carroll 1993; Horn 1989).

A confirmatory factor analysis with LISREL was also conducted on the ability test data, using a minimal number of specifications. That is, with the exception of the Word Problem Solving Test (which had both verbal and numerical content) and the Nelson Denny Comprehension measure (which taps both verbal abilities and working memory), all tests were specified to have zero loadings on one factor (e.g., Raven Progressive Matrices was specified to have a zero loading on gc). The results of that analysis are also shown in Table 3 . The factor model yielded and Non-Normed Fit Index and a Root Mean Squared Residual (RMR) of .078. (See Bentler and Bonett 1980, for discussion of these fit statistics and their interpretation.) The confirmatory analysis yielded essentially an equivalent solution as the exploratory analysis.

There are two major methods of comparing Gf and Gc ability estimates with the other measures of interest. The first would be to compute Gf and Gc scores on the basis of the derived factor loadings of the individual tests—either by computing factor scores or by extension analysis (e.g., see Ackerman and Rolfhus 1999). The second method, and the one adopted here, is simpler, and demonstrably less affected by the influence of communality inflation that occurs when similar tests are included in a factor solution (e.g., the Raven and the Spatial Analogy tests had a correlation of .775, yielding possible overestimates of their respective loadings on the underlying Gf factor). That is, the method used here was to compute summed z-score composites of the various tests that were a priori chosen to reflect Gf and Gc factors (see Thorndike 1986, for a discussion of this issue). Computation of these composites avoids factor score problems (e.g., see Tucker 1971), and loses little, if any, information. The composites showed a generally similar correlation to that from the derived factor solution . Correlations with the Gf and Gc abilities discussed below are based on these composites.

Abilities, knowledge, age and education.
Correlations between Gf and Gc composites and the 18 Knowledge Scale scores are shown in Table 4 , along with correlations between Knowledge Scales and participant age and participant education (where education was coded into three levels, . In addition, t tests (for dependent correlations) for the difference between respective Gf and Gc correlations were computed, and are shown in the last column of Table 4 . The Knowledge Scales are presented in order of the largest positive Gf-Gc differences to the largest negative Gf-Gc differences. Correlations between Gf, Gc, Age, and Education are shown at the bottom of the table.

Hypothesis #2 indicated that higher correlations would be found between Gc and Knowledge Scales than between Gf and Knowledge Scales, with the greatest differences in arts and humanities, civics, and technology. From the last column of Table 4 , it appears that this hypothesis was mostly, but not completely confirmed. Of the 18 knowledge scales, 10 showed significantly greater Gc correlations than Gf correlations, and 5 showed higher Gc correlations, but were not statistically significant. Two scales showed a nonsignificant advantage to Gf, and one test showed significantly higher Gf correlation than the Gc correlation. The pattern of results, though, was consistent with the expectations. That is, the physical science domains (Chemistry and Physics) showed the highest Gf correlations in comparison to Gc; social sciences showed nearly equal correlations with Gf and Gc; and Arts and Humanities and Civics showed the greatest Gc correlations, relative to Gf. The largest overall differences, though, clearly favored Gc.

Hypothesis #3 concerned the relations between age and knowledge. Age was predicted to be most highly positively correlated in the domains that also showed high loadings on Gc. In general, this hypothesis was supported. Although significant negative correlations were found between age and knowledge in the physical sciences (which also had the relatively lowest correlations with Gc), age had positive and significant correlations with knowledge in the social sciences, humanities, and civics.

Although no specific hypotheses were made for education (because of the expectation of a restricted range of education in this sample), the results for educational level appeared to mirror the results for age. That is, generally higher positive correlations were found between educational level and knowledge in the social sciences and civics (though not in the humanities), than in the physical sciences domains.

Factors and composites of knowledge tests.
Several facets of the knowledge test data are shown in Table 5 , including means (number of items correct), standard deviations, intercorrelations, and the results of an orthogonal factor analysis of the scales. Just as with the ability tests, the ubiquitous positive correlations among the knowledge tests illustrate positive manifold and indicate that one could derive a general knowledge factor (e.g., see Rolfhus and Ackerman 1999). Thus, it is important to keep in mind that there is common variance among the knowledge scales, just as there is among Gf and Gc measures. However, the purpose in factoring the 18 knowledge scales was to examine whether subsets of knowledge scales could be derived to simplify categorization and interpretation, instead of attempting to resolve underlying hypothetical constructs. The analysis was performed with an oblique DAPPFR rotation of the factors, which showed that it was possible to roughly identify four broad sets of knowledge scales (Science, Civics, Humanities, and Business/Law). It is important to note that only two factors correlated significantly—Humanities and Civics, . In order to minimize the number of comparisons, correlations are reported for composite scores based on the domain grouping of the knowledge scales. This is accomplished by forming unit-weighted z-score composites of the scales that constitute the identified knowledge factors.

Neither Social Potency nor Absorption revealed any salient correlations with knowledge domains (though in an analysis of individual knowledge scales, small, but significant positive correlations were found between Absorption and Art and American Literature ). Consistent with earlier studies (Ackerman and Rolfhus 1999; Rolfhus and Ackerman 1996, Rolfhus and Ackerman 1999), salient positive correlations were found between Humanities knowledge and TIE . Smaller, though significant, correlations were found for some of the other knowledge areas. No significant negative correlations were found between TIE and knowledge. TIE was also positively correlated with Gc , again consistent with previous findings.

The personality-knowledge relations found in this data set were in broad agreement with Hypothesis #4, though there were a few notable exceptions. In accordance with the predictions, Social Closeness and Traditionalism were broadly negatively associated with knowledge, and TIE was broadly positively associated with knowledge. Also, Social Potency, in contrast to Social Closeness, had minimally negative correlations with knowledge, supporting the importance of separating these two components of introversion-extroversion. In contrast to the hypotheses, Absorption failed to show the correlations that exceeded the salience criterion.

Interests—knowledge relations.
Correlations between the six interest traits and knowledge, ability, age and education are shown in Table 7 . Concordant with previous literature (Ackerman and Heggestad 1997), positive associations were found between knowledge and ability for Realistic, Investigative, and Artistic interests, whereas negligible or negative associations were found for Social, Enterprising, and Conventional interests. Most notable among these correlations were the large positive correlations between Investigative interests and knowledge in the Science category , and Artistic interests and knowledge in the Humanities category . In addition, a positive association was found between Gc and Investigative interests . Significant positive, but smaller correlations were found between Realistic interests and knowledge. At the individual knowledge scale level, only Electronics scale exceeded the salience criterion . The results were thus generally supportive of Hypothesis #5, indicating convergent validity among various indicated knowledge scales and Investigative, Realistic, and Artistic interests.

Multiple regression prediction of knowledge.
Although the individual correlations shown in Table 4 , Table 6 , and Table 7 provide important trait-knowledge comparisons, bivariate correlations do not allow for an assessment of the presence of overlapping variance (communality) among the various trait measures in predicting individual differences in knowledge. Multivariate procedures, such as multiple correlation and regression, provide one method for evaluating the relative influence specific predictor variables have in the context of other predictor variables. In order to assess the influence of various trait measures to arrive at a prediction of individual differences in knowledge, a series of multiple correlations were computed. Three facets of such analyses are of interest to understanding the nature of individual differences in knowledge: (1) Whether traditional assessments of Gc provide any incremental validity in predicting individual differences in knowledge beyond the influence of Gf; (2) Whether personality, interests and demographic variables of educational attainment and age add significant incremental predictive validity after traditional Gf and Gc measures are entered into the prediction equation; and (3) How much variance in the separate knowledge domains can be explained by abilities, non-ability traits, and demographic variables.

To address these issues the multiple correlations were computed in a multiple-step analysis plan. For the present purposes, it was decided to first enter (Step 1) Gf (which many intelligence theorists believe will account for all of the meaningful variance in adult intellect). In Step 2, Gc was added to the equation, to examine whether any significant incremental prediction of individual differences in knowledge can be accounted for by a traditional assessment of Gc. After the two ability factors, (Step 3) the three Personality trait measures that had been hypothesized (and demonstrated) as saliently correlated with knowledge scores (Social Closeness, Traditionalism, and TIE) were added, followed by the three Interest traits also hypothesized (and generally shown) to be related to individual differences in knowledge (Step 4). Then, the estimate of Educational attainment was added (Step 5). Finally, Age (Step 6) was added after all of the remaining predictor measures were entered in the prediction equation as a means of assessing whether there is a trait strongly associated with chronological age, that is both unmeasured and related to knowledge. The results of these analyses are shown in Table 8 . The table first indicates the amount of variance in knowledge composite scores that is accounted for by Gf (Step 1). Second, the table shows both the amount of incremental variance that is accounted for (noted as "R² to add") by each additional variable (or set of variables) at Steps 2–6, and the cumulative total amount of variance accounted for at each step of the analysis (shown as "Total R²").

The prediction of Humanities knowledge follows a similar pattern to that of Civics, though the role of Gf in predicting knowledge is diminished (7.5%) and the influence of Gc is larger (42.8%). Personality (4.2%) and Interests (8%) also provide incremental predictive validity, though Education does not. Age does provide a final incremental prediction, accounting for 3.1% of the variance in Humanities knowledge. For prediction of Business/Law knowledge, Gf (10.9%) and Gc (12.6%) are both significant predictors, though not large in comparison to the other knowledge areas. Personality and Educational attainment do not provide significant incremental prediction, but Interests (6.6%) and Age (2.4%) do.

Overall, the ability and non-ability traits provide both separate and aggregate predictions of knowledge in all of the domains that were assessed. Knowledge in Humanities was best predicted from these measures (62.6% of the knowledge variance accounted for), followed by Civics and Science knowledge (54.1% and 54.6% respectively). Educational attainment provided additional predictive power only in the area of Civics, and then only a small amount of variance (1.4%). Age accounted for significant increment in predictive validity for all of the areas except for Science, but in no case did age account for more than 5% of the remaining variance.

	Discussion

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
This paper was started with three broad questions, which are now addressed.

1. Are middle-aged adults more knowledgeable (or at least equally knowledgeable) than younger adults?
   

The results of this study appear to provide supporting evidence to a coherent view of adult intelligence-as-knowledge that is quite different from the extant data and discussion of adult intelligence as abstract reasoning or g. First, there were positive correlations between age and knowledge scores in 10 of the 18 domains investigated, across the sample of adults aged 21 to 62 years. Five of the remaining correlations between age and knowledge showed no significant relationship with age, and only the remaining three knowledge scales showed significantly negative correlations with age. All three were science tests (Chemistry, Physics, and Biology) that were also the most highly correlated with Gf (in contrast to Gc). Overall, a single composite score computed across all of the knowledge scales (Gk) yielded a correlation of , p < .01 between age and knowledge, indicating that (at least across the domains and participants that were sampled) older adults, were on average more knowledgeable than younger adults. For comparison purposes: Gf yields a correlation of p < .01 with age; Gc yields a correlation of +.14, p < .05, a general intelligence score (Gf + Gc) yields a correlation with age of , p < .05, and a composite of Gf + Gc + Gk yields a correlation of , nonsignificant association between a comprehensive intelligence composite and age. In those terms, older adults may not be more intelligent than younger adults, but neither are they on average, less intelligent.

1. Are individual differences in knowledge well-accounted-for by traditional measures of Gf and/or Gc?
   

The results of the analyses (to determine the respective contributions of Gf and Gc to predicting individual differences in knowledge) were differentiated by knowledge domain. Gf had a quite considerable explanatory power in predicting knowledge in the science domain, especially for Chemistry, Physics and Biology, accounting for 38.5% of the variance in the Science composite scores. Gf had a diminished role in accounting for individual differences in any of the other areas tested (less than 15% of the variance). In contrast, Gc accounted for an additional (i.e., beyond the variance accounted for by Gf) 34% of the variance in Civics knowledge and 42.8% of the variance in the Humanities, with a lesser role in Science and in Business/Law. Even in the aggregate, Gf and Gc together accounted for only 50% of the variance in any broad knowledge domain (except for Business/Law, where they only accounted for 23.5% of the variance). A reasonable conclusion from these results is that Gf is mostly related to science knowledge, Gc is mostly related to Civics and Humanities knowledge; however, there is much variance in knowledge that is unaccounted for by these traditional intelligence assessments.

1. What are the non-ability correlates of individual differences in knowledge?
   

In addition to the influence of Gf and Gc, but at a diminished overall magnitude, selected personality traits of Social Closeness, Traditionalism, and TIE accounted for significant variance in knowledge in every domain except for Business/Law. Individual differences in Realistic, Investigative, and Artistic interests accounted for significant amounts of variance in knowledge for all of the broad domains assessed. After trait measures were considered, individual differences in educational attainment and age provided relatively little additional explanatory power to predicting knowledge scale performance. This suggests that age may only be a useful predictor of knowledge in the absence of measures of relevant traits. As such, the influence of chronological age on individual differences in knowledge may be substantially overemphasized.

Conclusions
The PPIK theory is essentially an "investment" theory of adult intellectual development (in the general spirit of Cattell 1971[/1987] investment hypothesis). Numerous studies of Gf across the age span have given substantial support to the assertion that there is substantial decline in adult intellect once the individual reaches early adulthood. According to these views there is a stable period of performance followed be an inevitable decline in intellect as the individual reaches middle age and beyond (e.g., Horn 1982; Horn and Donaldson 1976; Salthouse 1996). There is little doubt as to the outcome of a comparison between 20-year-olds and 40-year-olds on a test of abstract reasoning (such as the Raven Progressive Matrices). I assert that such a demonstration misses the "dark matter" of adult intellect. That is, because knowledge is an important determinant of success in occupational pursuits (e.g., Hunter 1983), and middle-aged adults perform at least equally well on knowledge tests compared to young adults, I predict that middle-aged adults may perform as well or better than young adults on intellectual tasks that are fundamentally situated in the real-world. It is important to note, however, that this is not directly a "compensation" approach, where knowledge compensates for fluid intelligence. Rather, the assertion is more fundamental: many intellectually demanding tasks in the real world cannot be accomplished without a vast repertoire of declarative knowledge and procedural skills. The brightest (in terms of IQ) novice would not be expected to fare well when performing cardiovascular surgery in comparison to the middle-aged expert, just as the best entering college student cannot be expected to deliver a flawless doctoral thesis defense, in comparison to the same student after several years of academic study and empirical research experience. In this view, knowledge does not compensate for a declining adult intelligence; it is intelligence!

Moreover, the importance of personality and interests as determinants of the direction and amount of effort expended in the acquisition and maintenance of intelligence-as-knowledge should not be underestimated. Small correlations at the micro-level, when aggregated as influence over time (Abelson 1985; Rushton, Brainerd, and Pressley 1983), may help us predict and understand why some adults continue to acquire knowledge in particular areas and others do not. Longitudinal study is required; however, the salient correlations found among these various traits and individual differences in knowledge show potential for future study. Focusing on diverse domains of knowledge, and ultimately validating such measures for the prediction of academic and vocational success, may be more difficult than administering tests of abstract reasoning and predicting scores on decontextualized laboratory tasks of learning and memory. However, it seems time to recall why modern intelligence tests were created to begin with: to predict individual differences in real-world learning and performance. The program of research on adult intellect described in this paper is just one small step in this direction.

	Acknowledgments

 
This research was supported by a grant from the National Institute on Aging (1R01AG16648-01) and additional support from the Air Force Office of Scientific Research (F49620-98), Phillip L. Ackerman, principal investigator.

I thank the College Board for making available CLEP and AP tests for use in development of the knowledge scales. I also gratefully acknowledge the assistance of the Kanfer-Ackerman Laboratory staff for assisting in item creation, experiment running and data coding, especially the efforts of Eric Rolfhus in coordinating the data collection and preprocessing; and Chris Hertzog for useful discussions regarding data analysis strategies.

Received for publication November 25, 1998. Accepted for publication October 4, 1999.

	Appendix

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 
In cosmology, substantial controversy exists about the nature of the universe that traditionally hinges on mass estimates of unobserved "dark matter"—that if measured, would make various other observations confirm or refute extant theories.

	References

TOP Abstract PPIK Theory The Current Investigation Method Results Discussion Appendix References

 

• Abelson R. P., 1985. A variance explanation paradox: When a little is a lot. Psychological Bulletin 97: (1) 129-133.
• Ackerman P. L., 1996. A theory of adult intellectual development: Process, personality, interests and knowledge. Intelligence 22:227-257.
• Ackerman P. L., 1998. Adult intelligence: Sketch of a theory and applications to learning and education. Smith M. C., Pourchot T., , ed.Adult learning and development: Perspectives from educational psychology 145-158. Lawrence Erlbaum Associates, Mahwah, NJ.
• Ackerman P. L., Heggestad E. D., 1997. Intelligence, personality, and interests: Evidence for overlapping traits. Psychological Bulletin 121:219-245. [Medline]
• Ackerman P. L., Rolfhus E. L., 1999. The locus of adult intelligence: Knowledge, abilities, and non-ability traits. Psychology and Aging 14:314-330. [Medline]
• Ackerman P. L., Kanfer R., 1993. Integrating laboratory and field study for improving selection: Development of a battery for predicting air traffic controller success. Journal of Applied Psychology 78:413-432.
• Baltes P. B., Schaie K. W., 1976. On the plasticity of intelligence in adulthood and old age: Where Horn and Donaldson fail. American Psychologist 31:720-725. [Medline]
• Baltes, P. B., Smith, J., & Staudinger, U. M. (1992). Wisdom and successful aging. Nebraska Symposium on Motivation, 39, 123–167.
• Baltes P. B., Staudinger U. M., 1993. The search for a psychology of wisdom. Current Directions in Psychological Science 2: (3) 75-80.
• Barrick M. R., Mount M. K., 1991. The Big Five personality dimensions and job performance: A meta-analysis. Personnel Psychology 44:1-26.
• Bayley N., Oden M. H., 1955. The maintenance of intellectual ability in gifted adults. Journal of Gerontology 10:91-107. [Free Full Text]
• Bentler P. M., Bonett D. G., 1980. Significance tests and goodness of fit in the analysis of covariance structures. Psychological Bulletin 88:588-606.
• Birren J. E., Morrison D. F., 1961. Analysis of WAIS subtests in relation to age and education. Journal of Gerontology 16:363-368. [Free Full Text]
• Bloom B. S., 1964. Stability and change in human characteristics Wiley, New York.
• Brown J. I., Fishco V. V., Hanna G. H., 1993. Nelson-Denny Reading Test Riverside Publishing Company, Chicago, IL.
• Burke H. R., 1958. Raven's progressive matrices: A review and critical evaluation. The Journal of Genetic Psychology 93:199-228. [Medline]
• Carroll J. B., 1993. Human cognitive abilities: A survey of factor-analytic studies Cambridge University Press, New York.
• Cattell R. B., 1943. The measurement of adult intelligence. Psychological Bulletin 40:153-193.
• Cattell R. B., 1957. Personality and motivation structure and measurement World Book Company, Yonkers-on-Hudson, NY.
• Cattell, R. B. (1971/1987). Intelligence: Its structure, growth, and action. (Revised and reprinted in Abilities: Their structure, growth, and action, Boston: Houghton-Mifflin). Amsterdam: North Holland.
• Chi M. T. H., Ceci S. J., 1987. Content knowledge: Its role, representation, and restructuring in memory development. Advances in Child Development and Behavior 20:91-142. [Medline]
• Chi M. T. H., Glaser R., Rees E., 1982. Expertise in problem solving. Sternberg R. J., , ed.Advances in the psychology of human intelligence (Vol. 1) 7-76. Erlbaum, Hillsdale, NJ.
• Costa P. T., Jr. McCrae R. R., 1992. Revised NEO Personality Inventory and Five-Factor Inventory professional manual Psychological Assessment Resources, Odessa, FL.
• Crevier D., 1994. AI: The tumultuous history of the search for artificial intelligence NY: Basic Books, New York.
• DeBakey M. E., 1999. Timeline and accomplishments Author, Houston, Texas.
• Demming J. A., Pressey S. L., 1957. Tests "indigenous" to the adult and older years. Journal of Counseling Psychology 4:144-148.
• Ekstrom R. B., French J. W., Harman H. H., Derman D., 1976. Kit of factor-referenced cognitive tests Educational Testing Service, Princeton, NJ.
• Flynn J. R., 1984. The mean IQ of Americans: Massive gains from 1932 to 1978. Psychological Bulletin 95:29-51.
• Goff M., Ackerman P. L., 1992. Personality-intelligence relations: Assessment of typical intellectual engagement. Journal of Educational Psychology 84:537-552.
• Goodnow J. J., 1980. Everyday concepts of intelligence and its development. Warren N., , ed.Studies in cross-cultural psychology (Vol. 2) 191-219. Pergamon Press, Oxford, England.
• Hebb D. O., 1942. The effect of early and late brain injury upon test scores, and the nature of normal adult intelligence. Proceedings of the American Philosophical Society 85:275-292.
• Henmon V. A. C., 1921. Intelligence and its measurement: A symposium. Journal of Educational Psychology 12:195-198.
• Holland J. L., 1959. A theory of vocational choice. Journal of Counseling Psychology 6: (1) 35-45.
• Holland J. L., 1973. Making vocational choices: a theory of careers Prentice Hall, Englewood Cliffs, NJ.
• Horn J. L., 1982. The aging of human abilities. Wolman B. B., , ed.Handbook of developmental psychology 847-870. Prentice-Hall, Englewood Cliffs, NJ.
• Horn J. L., 1989. Cognitive diversity: A framework of learning. Ackerman P. L., Sternberg R. J., Glaser R., , ed.Learning and individual differences. Advances in theory and research 61-116. W. H. Freeman, New York.
• Horn J. L., Donaldson G., 1976. On the myth of intellectual decline in adulthood. American Psychologist 31:701-719. [Medline]
• Horn J. L., Donaldson G., 1977. Faith is not enough: A response to the Baltes-Schaie claim that intelligence does not wane. American Psychologist 32:369-373.
• Humphreys L. G., Montanelli R. G., Jr. 1975. An investigation of the parallel analysis criterion for determining the number of common factors. Multivariate Behavioral Research 10:193-206.
• Hunter J. E., 1983. A causal analysis of cognitive ability, job knowledge, job performance, and supervisor ratings. Landy F., Zedeck S., Cleveland J., , ed.Performance measurement and theory. Hillsdale, NJ: Erlbaum
• Jackson D. N., 1985. Multidimensional Aptitude Battery Sigma Assessment Systems, London, Ontario, Canada.
• Jensen A. R., 1998. The g factor: The science of mental ability Praeger, Westport, CT.
• Jones H. E., Conrad H. S., 1933. The growth and decline of intelligence. Genetic Psychology Monographs 13:223-298.
• Labouvie-Vief G., 1977. Adult cognitive development: In search of alternative explanations. Merrill-Palmer Quarterly 23:227-263.
• Lamb R. R., Prediger D. J., 1981. The Unisex Edition of the ACT Interest Inventory American College Testing, Iowa City, IA.
• Lippmann W., 1922. The mental age of Americans. New Republic 32:213-215.
• Miles C. C., 1934. Influence of speed and age on intelligence scores of adults. Journal of General Psychology 10:208-210.
• Montanelli R. G., Jr. Humphreys L. G., 1976. Latent roots of random data correlation matrices with squared multiple correlations on the diagonal: A Monte Carlo study. Psychometrika 41:341-348.
• Raven J. C., Court J. H., Raven J., 1977. Raven's Progressive Matrices and Vocabulary Scales Psychological Corporation, New York.
• Rolfhus, E. L. (1998). Assessing individual differences in knowledge: Knowledge structures and traits. Unpublished doctoral dissertation, University of Minnesota, Minneapolis.
• Rolfhus E. L., Ackerman P. L., 1996. Self-report knowledge: At the crossroads of ability, interest, and personality. Journal of Educational Psychology 88:174-188.
• Rolfhus E. L., Ackerman P. L., 1999. Assessing Individual Differences in Knowledge: Knowledge, Intelligence and Related Traits. Journal of Educational Psychology 91:511-526.
• Rushton J. P., Brainerd C. J., Pressley M., 1983. Behavioral development and construct validity: The principle of aggregation. Psycho-logical Bulletin 94:18-38.
• Salthouse T. A., 1996. The processing-speed theory of adult age differences in cognition. Psychological Review 103:403-428. [Medline]
• Schaie K. W., 1970. A reinterpretation of age related changes in cognitive structure and functioning. Goulet L. R., Baltes P. B., , ed.Life-span developmental psychology 423-466. Academic Press, New York.
• Schaie K. W., 1977–1978. Toward a stage theory of adult cognitive development. International Journal of Aging and Human Development 8:129-138. [Medline]
• Schaie K. W., 1978. External validity in the assessment of intellectual development in adulthood. Journal of Gerontology 33:695-701. [Abstract/Free Full Text]
• Schaie K. W., 1996. Intellectual development in adulthood: The Seattle longitudinal study Cambridge University Press, New York.
• Schaie K. W., Strother C. R., 1968. A cross-sequential study of age changes in cognitive behavior. Psychological Bulletin 70:671-680. [Medline]
• Scheidt R. J., Schaie K. W., 1978. A taxonomy of situations for an elderly population: Generating situational criteria. Journal of Gerontology 33:848-857. [Abstract]
• Schwartzman A. E., Gold D., Andres D., Arbuckle T. Y., Chaikelson J., 1987. Stability of intelligence: A 40-year follow-up. Canadian Journal of Psychology 41:244-256. [Medline]
• Sinnott J. D., 1998. The development of logic in adulthood: Postformal thought and its applications Plenum, New York.
• Sternberg R. J., Conway B. E., Ketron J. L., Bernstein M., 1981. People's conceptions of intelligence. Journal of Personality and Social Psychology 41:37-55.
• Sternberg R. J., Wagner R. K., 1989. Individual differences in practical knowledge and its application. Ackerman P. L., Sternberg R. J., Glaser R., , ed.Learning and individual differences. Advances in theory and research 255-278. W. H. Freeman, New York.
• Tellegen, A. (1982). Brief manual for the Multidimensional Personality Questionnaire (MPQ). Author.
• Tett R. P., Jackson D. N., Rothstein M., 1991. Personality measures as predictors of job performance: A meta-analytic review. Personnel Psychology 44:703-742.
• Thorndike R. L., 1986. The role of general ability in prediction. Journal of Vocational Behavior 29:332-339.
• Thurstone T. G., 1962[Primary Mental Abilities]. PMA Science Research Associates, Chicago.
• Tucker L. R., 1971. Relations of factor score estimates to their use. Psychometrika 36:427-436.
• Tucker L. R., Finkbeiner C. T., 1981. Transformation of factors by artificial personal probability distributions. (Report No. RR-81-58) Educational Testing Service, Princeton, NJ.
• Voss J. F., Wiley J., Carretero M., 1995. Acquiring intellectual skills. Annual Review of Psychology 46:155-181. [Medline]
• Wagner R. K., Sternberg R. J., 1985. Practical intelligence in real-world pursuits: The role of tacit knowledge. Journal of Personality and Social Psychology 49: (2) 436-458.
• Wechsler D., 1958. The measurement and appraisal of adult intelligence Williams and Wilkins, Baltimore, MD.
• Wechsler D., 1981. Manual for the Wechsler Adult Intelligence Scale-Revised Psychological Corporation, New York.

This article has been cited by other articles:

				P. L. Ackerman Cognitive Ability and Non-Ability Trait Determinants of Expertise Educational Researcher, November 1, 2003; 32(8): 15 - 20. [Abstract] [PDF]

This Article Abstract Full Text (PDF) Services Download to citation manager Citing Articles Citing Articles via HighWire PubMed PubMed Citation