This chapter begins with a description of the pilot studies that led to the selection and validation of the measures used in our research program. I then present the 1956 baseline study and compare its findings with the six cross-sectional replications. For purposes of an orderly presentation, I start with the analyses of the basic cognitive battery that is common to all study cycles. I then present data for the fifth, sixth, and seventh cycles for the extended cognitive battery and the practical intelligence measures. Finally, I consider the cross-sectional findings for the measures of cognitive style (Test of Behavioral Rigidity, TBR).
Our inquiry began by questioning whether factorially defined measures of intellectual abilities would show differential age patterns. Before this question could be examined parametrically, it was necessary to examine the applicability of the Primary Mental Abilities (PMA) test to an older population, with respect to both its level of difficulty and whether the low correlations among abilities observed in childhood would continue to prevail for adults. Two pilot studies concerned with these questions are described in this section. In addition, our interest in cognitive style as a concomitant of intellectual aging required the development of a set of psychometrically sound measurement instruments for the multiple dimensions of rigidity–flexibility. A third pilot study was concerned with demonstrating the construct validity of the resultant measure (the TBR).
Sixty–one study participants, gathered from the geriatric practice of a family physician and from the membership of the small first cohort of the San Francisco Senior Citizen Center, were given the PMA tests under standard conditions. For purposes of analysis, participants were arbitrarily divided into four approximately equal groups: ages 53 to 58, 59 to 64, 65 to 70, and 71 to 78. In the absence of adult norms and to permit comparison across the five abilities measured by the test, raw scores were converted into percentiles employing norms for 17-year-old adolescents (L. L.Thurstone & T. G. Thurstone, 1949). Thus, for example, if our participants, on average, were at the 50th percentile, this would imply that their level of functioning would be similar to that of 17-year-olds.
The results of this study are shown in figure 4.1. For the group in their 50s, stability is suggested for Verbal Meaning and Number (performance is slightly above the adolescents’ 50th percentile), but performance appears to be substantially lower for the other three tests. Indeed, it was lowest for Spatial Orientation and Inductive Reasoning, measures of the kind of ability later to be termed fluid by Cattell (1963). The differential pattern was observed for all age groups, with some further lowering of scores into the 60s and apparent maintenance of the lower level for the group in their 70s (Schaie, Rosenthal, & Perlman, 1953).
On the off chance that the differential pattern might be caused by unequal effects of the slightly speeded instructions for the performance of older individuals, four of the tests were administered to 31 participants without a time limit. Results shown in figure 4.2 indicate that, if anything, differential performance levels were greater and in the same order as under the standard conditions of instruction.
The first pilot study also investigated the construct validity of the PMA 11–17 when used with older individuals. Intercorrelations between the tests were computed and were quite low, ranging from .07 for the correlation between Spatial Orientation and Number to .31 for the correlation between Spatial Orientation and Inductive Reasoning. These correlations did not differ significantly from those obtained for an adolescent comparison group (Schaie, 1958d). Split-half reliabilities computed under the power test condition were also quite satisfactory, with all above .92 after Spearman–Brown correction.
FIGURE 4.1. Age differences in primary mental abilities for the adult pilot sample.
FIGURE 4.2. Age differences in primary mental abilities under unspeeded conditions of test administration.
A second pilot study was conducted in 1954 as part of an investigation of community-dwelling older persons (more completely described in Schaie & Strother, 1968a, 1968d). A campus and community appeal resulted in the selection of 25 men and 25 women, all college graduates with professional careers, ranging in age from 70 to 88 years (mean age 76.5 years). These study participants were all in fair-to-superior health and free of diagnosable psychiatric symptoms. The differential ability pattern shown in the first pilot study was replicated, with performance on Number, Word Fluency, and Verbal Meaning substantially above that observed for Spatial Orientation and Inductive Reasoning. Also noteworthy was the finding that some of the octogenarians in this study still equaled or exceeded the adolescent mean on some of the verbal tests, even though it was most likely that this performance represented a decrement from previously higher levels, suggested by the unusually advantaged demographic characteristics of this sample. In this study, findings also indicated sex differences in favor of the males for the Spatial Orientation and Number abilities and in favor of the females for Verbal Meaning, Inductive Reasoning, and Word Fluency (Strother, Schaie, & Horst, 1957).
The third pilot study was concerned with demonstrating the construct validity of a set of measures defining the multiple dimensions of rigidity-flexibility. This study began as my master’s thesis (Schaie, 1953), in which I factor-analyzed a number of tests representing the functional, structural, and attitudinal approaches to the study of rigidity-flexibility (see Chown, 1959). Although use of an unorthodox method of rotation (Horst, 1956b) made the result of the initial analysis somewhat tenuous, it provided the basis for selecting the variables to be included in the more definitive construct validation study (Schaie, 1955). Because I intended to use the final battery for studies of aging, only those tests were included that could be adapted for use with both adult and elderly populations. For practical reasons, only tests suitable for group administration were retained, and those tests were selected for the social status and education that were minimally important in influencing results in the initial study.
The measures included in the construct validity study, for what eventually became the TBR (Schaie & Parham, 1975), are described in the section that discusses measures of rigidity-flexibility in chapter 3. Not described there are the Jar test (Luchins, 1942) and the Alphabet test (Bernstein, 1924), which were subsequently dropped from the final battery. The Jar test involves participants correctly identifying the use of jars of different sizes in measuring a given quantity of water, with different methods of solution appropriate at various times. The so-called Einstellung effect is tested by first conditioning the participant to employ a complex method of solution. Critical problems are then presented for which a more direct solution is available. The rigidity measure then becomes the number of problems unnecessarily solved in the more complex manner. The Alphabet test involves the letter combinations abcde and lmnopq first written forward and then backward.
The validation sample consisted of 216 participants who were drawn from day and evening classes at the University of Washington (Seattle, WA), from a social club for older people, from the membership of a Unitarian church, from a Rotary club, and from a group of YMCA members. The sample ranged in age from 17 to 79 years, with a mean age of 38 years. Educational level ranged from 4 to 20 years, with a mean of 14.2 years. The participants’ occupational levels averaged 6.3 on a 10-point scale from unskilled to professional. For purposes of cross-validation, a second sample of 200 participants was drawn from a restricted population of college students between the ages of 19 and 26 years, with a mean age of 21.4 years.
The correlation matrix for the first sample was factored using a simplification of Thurstone’s multiple group method (Horst, 1956a; L. L. Thurstone, 1947, pp. 170 ff.). Note that this approach is an early forerunner of modern confirmatory factor analysis. The first hypothesis specified the existence of a single rigidity factor and a motor speed factor. This hypothesis had to be rejected, and instead a three-factor combination emerged that, on appropriate oblique rotation (Horst & Schaie, 1956), yielded an acceptable simple structure solution (see table 4.1).
TABLE 4.1. Factor Loadings and Factor Intercorrelations After Oblique Rotation
Factor I was originally named Motor-Cognitive Speed (the current term is Psychomotor Speed [PS]), Factor II was identified as Personality-Perceptual Rigidity (now called Attitudinal Flexibility [AF]), and Factor III was thought to be a representation of Motor-Cognitive Rigidity (now called Motor-Cognitive Flexibility [MCF]). These factors were next cross-validated by factoring the correlation matrix for the second sample and rotating it to the same factor pattern. As indicated by table 4.1, the resulting pattern replicates that obtained for the original sample. The major difference between results from the two samples can be seen in the factor intercorrelations. All three factors are moderately correlated in the heterogeneous sample, whereas the factor correlations are quite small in the homogeneous sample of college students.
As a result of this study, we decided to retain the four subtests (Capitals, Opposites, R scale, P scale) that provided the best factor definition for inclusion in the final version of the TBR used throughout the SLS.
In our effort to determine the pattern of age differences, we sampled 500 individuals from our health maintenance organization population frame distributed evenly by gender and 5-year age interval from 21 to 70. For ease of comparison, we standardized all variables across the entire sample to T-score format (M = 50, SD = 10). We have followed this procedure throughout, always restandardizing on the basis of the largest total sample of individuals’ scores at first test (entry into the study). Age difference findings from the baseline study are shown in figure 4.3.
What is most noteworthy about the baseline study is that, although negative age differences are found on all five abilities, peak ability ages occur generally later than had been observed in the previous literature (see Jones & Conrad, 1933; Wechsler, 1939), and that the differential ability patterns noted in our first pilot study could be confirmed in this reasonably representative and age-continuous investigation.
More specifically, we noted that peak levels of the abilities were reached for Reasoning by the 21- to 25-year-old group, for Space by the 26- to 30-year-olds, and for Verbal Meaning and Word Fluency by the 31- to 35-year-olds, but for Number only at ages 46–50. Similarly, there are differences among abilities for the first occurrence of a significant age difference from performance at the peak age. Such significantly lower average performance levels were observed for Verbal Meaning, Space, Reasoning, and Word Fluency with respect to the 36- to 40-year-olds, but for Number only for the 56- to 60-year-olds. Differences were found also in the absolute magnitude of the age difference between the group at peak age and the oldest observed group (ages 66–70). This difference amounted to 1.5 SD for Verbal Meaning, Space, and Reasoning, but slightly less than 1 SD for Number and Word Fluency. The absolute differences between the youngest and oldest age groups was greatest for Reasoning (1.5 SD), next largest for Space and Verbal Meaning (1.2 and 1 SD, respectively), and smallest for Word Fluency and Number (0.6 and 0.2 SD, respectively). These differences could be interpreted as of substantial magnitude for Reasoning, Space, and Verbal Meaning; moderate for Word Fluency; and near trivial for Number.
FIGURE 4.3. Baseline age differences on the primary mental abilities.
This section begins with the basic cognitive data collected throughout the study. I then turn to the expanded battery and consider the age difference patterns within ability domains as well as the age difference patterns in the factor scores for the latent ability constructs. Finally, cross-sectional results are provided for the measure of practical intelligence and the measures of cognitive style.
Given the 7-year interval between our data cycles, all data were reorganized into 7-year age intervals. Thus, the baseline study for purposes of comparison with the subsequent replications contains seven age groups, with mean ages from 25 to 67 years. In the second cross-sectional sample (1963, N = 998), we included eight age groups, with mean ages from 25 to 74 years. The third sample (1970, N = 705) and the remaining three cross-sectional samples (1977, N = 609; 1984, N = 629; 1991, N = 691; and 1998, N = 719) include nine age groups, with mean ages from 25 to 81 years.
Again, for ease of comparison, all raw scores were converted to T-scores with means of 50 and standard deviations of 10 based on the entire set of 4,851 observations at first test.
Mean scores by age and gender for the five PMA subtests and the two composite measures of Intellectual Ability and Educational Aptitude are presented in table 4.2. For a more dramatic presentation of differences across abilities in age difference pattern, we graphed mean values by gender for the first (1956) and last (1998) cross-sectional comparisons (see figure 4.4).
Age difference patterns for males have remained relatively constant over the course of the study, with the exception that peak ages for Verbal Meaning and Number shift to age 60 and Word Fluency shifts to the late 30s. For the female study participants, much greater shifts can be observed. In young adulthood, females now show higher performance on Spatial Orientation than was the case some 42 years ago. For this ability, the peak age for women is now in midlife. For the women, age differences in Number ability have virtually disappeared, and the peak age for Verbal Meaning has moved to the early 50s.
TABLE 4.2. Means and Standard Deviations for the Basic Ability Battery by Sample and Gender
FIGURE 4.4. Age difference patterns of the five primary mental abilities by gender for the first (1956) and last (1998) cross-sectional comparisons.
Absolute age differences across the adult life span, observed at any given point, have been reduced over time but remain substantial for most abilities. However, there has been a sharp reduction of age differences in performance observed until the late 60s are reached. For example, for Verbal Meaning the absolute difference between ages 25 and 81 currently amounts to 0.8 SD for both men and women. But the absolute difference between ages 25 and 67 has been reduced from 0.8 to 0.2 SD for men and from 1.2 to 0.1 SD (or virtually no difference) for women.
Age difference data are not directly relevant to testing propositions about ontogenetic change. Such data, however, when examined in the context of cross-sectional sequences, are quite appropriate for testing the proposition of invariance in age difference patterns over time. Given certain assumptions, they are also the data of choice to evaluate the magnitude of cohort differences and time-of-measurement (period) effects. The average age difference patterns across all measurement occasions from 1956 to 1998 are graphed in figure 4.5.
Throughout our study, we have questioned whether age difference patterns remain invariant over time and have concluded that statistically significant shifts in such patterns can be observed. This conclusion is based in part on the findings of Age × Time interactions in time-sequential analyses and of Cohort × Time interactions in cross-sequential analyses (Schaie & Hertzog, 1983; Schaie, Labouvie, & Buech, 1973; Schaie & Strother, 1968c).
To examine shifts in age profiles as well as the peak ages across the seven cross-sectional studies, mean values for each study across age for the five PMAs and for the composite indexes are graphed in figure 4.6. What seems most apparent is that means observed at the same ages tend to fall at progressively higher levels for successive cohorts attaining a given age. This is clearly the case for Verbal Meaning, Spatial Orientation, and Inductive Reasoning as well as the composite indexes. Successively lower levels are attained for successive cohorts on Number; the cohort pattern for Word Fluency is less clearly defined.
FIGURE 4.5. Age difference patterns of the five primary mental abilities summed across all measurement occasions.
For Verbal Meaning there has been a general increase in performance level at all ages. Most noteworthy, however, are performance increases at the older ages. For ages 60 to 74 years, these increases amount to a full standard deviation over a 42-year period. Equally dramatic, at 81 years of age, performance has increased by approximately 1 SD over the 28 years monitored for this age group.
Somewhat smaller increases across time, averaging approximately 0.3 SD, were observed for Spatial Orientation until 53 years of age. At ages 60 and 67 years, there was a gain of 0.5 SD, with less improvement (0.2 SD) for the two oldest age groups. Inductive Reasoning also showed large rises in performance over time at comparable ages (averaging about 2/3 SD), whereas Number and Word Fluency showed complex changes involving curvilinear patterns of age differences.
FIGURE 4.6. Age difference patterns of the expanded battery by test occasion. ETS = Educational Testing Service.
The summary indexes of Intellectual Ability and Educational Aptitude also showed significant reductions in age differences at all comparable ages. Changes in performance level are discussed in greater detail in chapter 6, with explicit discussions of cohort and period effect.
Table 4.2 also contains means and standard deviations by age and gender summed across all observations. These values represent the averages for all seven replications over a 42-year period. Correlations among the five basic abilities by age group and for the total sample are reported in appendix A-4.1.
The expanded cognitive battery described in chapter 3 was administered in Cycles 5, 6, and 7, except for the Perceptual Speed measures of Finding A’s and Identical Pictures, which were first introduced during the fourth cycle. Cross-sectional analyses for these cycles were consequently done both at the level of individual measures and at the latent construct level to determine within-ability and across-ability age difference patterns (see Schaie & Willis, 1993).
Most of the major longitudinal studies of adult development in the past collected only very limited data that speak to the issues of generalizability of findings within and across domains in the area of intellectual functioning (e.g., Busse, 1993; Costa & McCrae, 1993; Eichorn, Clausen, Haan, Honzik, & Mussen, 1981; Rott, 1990; Schaie, 1983b; Schmitz-Scherzer & Thomae, 1983; Shock et al., 1984; Siegler, 1983). Cross-sectional data may actually be quite instructive with respect to this issue because such data allow us to draw concurrent comparisons of age difference patterns within and across domains without requiring attention to the thorny methodological issues associated with comparisons across time (see chapter 2).
There are likely to be substantial life stage differences in adulthood in the degree to which levels of performance for different ability markers are equivalent both within and across ability domains. First I examine the extent to which patterns of age differences are congruent within a particular ability domain by describing age difference patterns for six psychometric abilities (see Schaie & Willis, 1993). These abilities broadly sample higher-order constructs, such as those espoused by Horn (1982). Thus, fluid intelligence is represented by the ability of Inductive Reasoning. Verbal Ability and Numeric Ability stand as representatives of crystallized intelligence, and mental rotation is represented by our Spatial Ability construct. Verbal Memory and Perceptual Speed are examined as ability samplers for the memory and speed domains, respectively. Next, consideration is therefore given to the age difference patterns across the various ability domains.
What is at issue is the question of whether patterns of age differences in ability remain invariant for multiple markers of the primary mental abilities. This issue is addressed again in chapter 8, in which several studies are reported that examine invariance across age and time by means of structural analyses (see Schaie, Maitland, Willis, & Intrieri, 1998; Schaie, Willis, Jay, & Chipuer, 1989), and it is shown that configural invariance (i.e., number of factors and factor pattern) is preserved across widely differing age groups. Here, we examine shifts of different marker variables at different life stages, an effect observed a long time ago in cross-sectional studies of the Wechsler Adult Intelligence Scale (Cohen, 1957). The reasons for such shifts may be sought in such contextual variables as shifts in educational exposure to the skills embodied in a particular marker variable or latent construct and the impact of slowing in perceptual and/or motor speed that may differentially affect various markers. For example, we know that conditions of instructions and speededness imposed by time limits differentially affect performance on the PMA and Educational Testing Service vocabulary tests (Hertzog, 1989). Likewise, it is known that there have been generational shifts in instruction in quantitative skills that should affect numerical performance for different cohorts. For other abilities, congruence would be expected across the entire life span until the 80s are reached. At that late stage, the differential memorization demands as well as the relative motor complexity of answer sheet as compared with disposable booklet formats might result in differential efficiency of a given marker.
Age difference patterns appear generally invariant across sex within domains (albeit there is strong evidence for overall gender differences in level of performance), but it is not a foregone conclusion that such invariance holds for all abilities or for all markers of a given ability. Gender differences will therefore be examined, and results are reported separately by gender when warranted.
The issue of the generalizability of markers within domains is particularly important for age comparative studies. For valid cross-sectional comparisons, it must be shown that an observed variable provides a reasonable representation of the developmental trajectory of the latent construct to be marked. If this is the case, then a single marker may suffice. But if there is wide discrepancy in developmental trajectories for multiple markers, we would then be forced to multiply mark the construct, providing differential weights for the markers at different life stages. The data presented here provide some guidance on these matters.
All scores on the observed variables were rescaled into T-score form (M = 50; SD = 10) using parameters for the total sample at first test (N = 2476). Factor scores for the six intellectual abilities were computed using factor regression weights based on a previously determined best-fitting factor model (Schaie, Dutta, & Willis, 1991).
The age difference patterns for the observed markers of the six primary ability domains (in standardized form) are reported in table 4.3. Means and standard deviations are given separately by gender and for the total age/cohort groupings. This table also reports averages over the three data points. Correlations among the variables in the extended battery by age group and for the total sample can be found in appendix A-4.2.
The age difference gradients are graphed in figure 4.7, comparing the markers for each latent construct representing the averages over the three cohorts assessed at the same age. Results with respect to congruence among the patterns for the different operations measuring the ability constructs are described in the following paragraphs.
The new markers of the Inductive Reasoning factor have very similar age profiles, with an overall age difference of about 2 SD from the youngest to the oldest age cohort. Significant age differences from the youngest (peak) age appear by age 46 for all three tests. Gender differences were not significant for Letter Series but favored women for Word Series and men for Number Series. A significant overall increase in performance level across the three assessment times was found only for Letter Series.
TABLE 4.3. Means and Standard Deviations for the Expanded Ability Battery by Sample and Gender
FIGURE 4.7. Age difference patterns of the latent ability constructs.
The ability profiles for the new markers of spatial orientation also showed a difference of just less than 2 SD from the youngest to the oldest group. As for the original marker, there were significant gender differences favoring males for all three tests. Significant age differences from the youngest group were observed by age 53 for Alphanumeric Rotation, but already by age 46 for Object Rotation and Cube Comparison. On the last measure, there was also a significant gain across measurement occasions as well as a significant Sex × Age interaction that reflected the absence of sex differences in midlife (ages 39, 46, and 53).
The expanded battery contains three markers for the Perceptual Speed factor that were not measured earlier. Age differences from the youngest to the oldest group range from 1 SD for the Finding A’s test to almost 2.5 SD for the Identical Pictures test. Interestingly, Identical Pictures appears to be the easiest measure for the young groups, whereas Finding A’s is easier for the older groups. Gender differences favor women for all three tests. Significant age differences from the youngest group were found by age 46 for Identical Pictures and by age 53 for Number Comparison and Finding A’s. Significantly higher performance levels across the three assessment points were found for Identical Pictures and for Number Comparison.
The new markers of Numeric Ability attain a peak at age 46 and then show a negative age difference of approximately 2/3 SD to the oldest age group. No significant gender differences were observed for either test, but there was a significant increase in performance level in 1991 for the Addition test.
The new measures of Verbal Ability have rather different profiles from the original PMA marker because they are virtually unspeeded. The profile for the 1984 testing is slightly concave for both measures, with virtually no difference in level between the youngest and oldest sample and a peak plateau from 39 to 67 years of age. In contrast, the 1991 profile shows positive age differences to age 53, followed by a virtual plateau to the oldest group for both measures,
In 1998, peaks were reached at age 53 for the easier and at age 60 for the more difficult test, followed by virtual plateaus to the oldest ages. There were no significant gender differences on either test, but significantly higher overall performance levels were observed for the Advanced Vocabulary test in 1991.
Also, newly included were two measures of the Verbal Memory factor. They have quite similar profiles, with an age difference of just under 2 SD from the youngest to the oldest age/cohort. Significant gender differences on these measures favor women (Immediate Recall). Significant age differences from the level of the youngest group appeared by age 39 for Delayed Recall and by age 46 for Immediate Recall. Significantly higher levels of performance on the second measurement occasion were also observed.
Both the original markers and the new tests were combined as described above, and factor scores were computed for the resulting six latent constructs. Means and standard deviations separately by gender and for the total in each age group are given in table 4.4 and are graphed across gender in figure 4.8. Table 4.4 also has means and standard deviations for the factor scores averaged over the three test occasions. Intercorrelations among factor scores by age group and for the total sample are reported in appendix A-4.3.
For four of the six factors, there are consistent negative age differences. They are statistically significant for Inductive Reasoning, Spatial Orientation, and Perceptual Speed at age 46 and for Verbal Memory at age 39. The magnitude of age difference from the youngest to the oldest group amounts to approximately 2 SD on average. The remaining two factors, Numeric Facility and Verbal Comprehension, have a very different profile. They both show positive age differences until midlife, with less than 0.5 SD negative differences thereafter, such that persons in advanced old age, on average, are at a higher level than the youngest age group. Gender differences on the latent construct measures were observed in favor of men on Spatial Orientation and in favor of women for Perceptual Speed and Verbal Memory. A higher overall performance level in 1991 was shown for Inductive Reasoning and Verbal Memory.
The Basic Skills–Reading test was developed by the Educational Testing Service (1977) to examine whether graduating high school students had attained knowledge that was relevant to real-life tasks. The test therefore simulates textual materials representing everyday problem-solving activities. This test was given in both the fifth and sixth cycles. An overall analysis of variance of participants at first test did not detect any significant gender differences but did result in a significant age/cohort main effect (F[df = 8, 1953] = 142.62, p < .001). As shown in table 4.5 and figure 4.9, the age difference profile is virtually flat until age 60; the first significant age difference occurs between 60 and 67 years of age. There was no significant Age × Gender interaction. Overall, there was a significantly higher level of performance in 1991 than in 1984 (F[df = 1, 1935] = 34.47, p < .001) as well as a significant Age × Time interaction (F[df = 1, 1935] = 4.42, p < .001). However, for a specific age group, it was only the 81-year-old level that showed a statistically significant gain across cohorts in 1991.
TABLE 4.4. Means and Standard Deviations for the Latent Construct Scores by Sample and Gender
FIGURE 4.8. Age difference patterns of the measure of practical intelligence by test occasion. EPT, Everyday Problems Test; ETS, Educational Testing Service.
Because of significant ceiling effects in the Basic Skills test, we decided to substitute the Everyday Problems Test (EPT; Marsiske & Willis, 1995; Willis, 1992, 1996) in the 1998 cycle. Cross-sectional data by age and gender are reported in table 4.6, and the age difference patterns by individual IADLs are graphed in figure 4.9. Intercorrelations among the seven subscales are reported by age group and for the total sample in appendix A-4.4.
As on the cognitive ability measures, there are age differences of approximately 2.5 SD from young adulthood to advanced old age. Overall age differences become significant by age 60. Women generally do better than men except at the oldest age.
TABLE 4.5. Means and Standard Deviations for the Tests of Practical Intelligence by Sample and Gender
Data on cognitive style, as measured by the TBR, were collected from the beginning of the SLS. A summary of cross-sectional data through the fourth cycle can be found in Schaie (1983b). Detailed data on the TBR subscores were provided by Schaie and Willis (1991). In this section, I provide a summary of the cross-sectional data for the latent dimensions of MCF, AF, and PS. The structural relationship between the primary mental abilities and the cognitive style constructs (see Dutta, 1992; Schaie, 2005a; Schaie et al., 1991) are examined in chapter 9.
Cross-sectional data on the cognitive style variables were obtained for mean ages 25 to 67 years in Cycle 1, for mean ages 25 to 74 years in Cycle 2, and for mean ages 25 to 81 years in the remaining cycles. To provide appropriate comparisons with the ability data, the TBR factor scores were restandardized across all seven cycles (N = 4837). Table 4.7 presents means and standard deviations for the six test cycles separately by gender and for the total sample for each age level. Table 4.7 also provides data by gender and age summed across all seven measurement occasions. Intercorrelations of TBR factor scores by age group and for the total sample are reported in appendix A-4.5. Figure 4.10 provides a graphic representation of shifts in age differences over time combined across genders, and the average age difference patterns across all occasions are graphed in figure 4.11.
TABLE 4.6. Means and Standard Deviations in T-Scores for the Everyday Problems Test by Sample and Gender
FIGURE 4.9. Age difference patterns for the separate instrumental activities of daily living for the Everyday Problems Test.
The cross-sectional data on cognitive style suggest that there is a decrease in MCF and PS for successive age/cohort groups. What is most noteworthy, however, is the fact that, until the 80s are reached, there has been an increase in flexibility and speed for successive cohorts at the same ages. This trend has led to successively later ages at which significant declines are observed. For the age groups represented on each test occasion (ages 25 to 67), there is a statistically significant Time × Age effect for both MCF and PS. MCF seems to peak in young adulthood, with lower levels prevailing as early as age 39 in our 1956 sample, but beginning only with the 60s in the most recent samples. A similar pattern is shown for PS, which in the earliest sample peaked in young adulthood but now peaks in the early 50s. AF seems to be fairly level across age groups until the mid-40s, with decline below young adult levels observable beginning with the 60s. Again, recent cohorts show average performance that is above earlier cohorts at the same ages.
Significant overall gender differences are found for all three cognitive style factors. Women exceed men on PS, and men exceed women on MCF and AF. These gender differences generalize across age and measurement occasions.
I begin this chapter by describing three pilot studies: The first two were designed to demonstrate the applicability of the PMA test to an older population, with respect to both its level of difficulty and whether the low correlations among abilities observed in childhood would continue to prevail for adults. A third pilot study was concerned with the development of a set of psychometrically sound measurement instruments for the multiple dimensions of rigidity–flexibility that we wished to relate to cognitive abilities.
TABLE 4.7. Means and Standard Deviations for the Rigidity-Flexibility Factor Scores by Sample and Gender
FIGURE 4.10. Age difference patterns of the cognitive style variables for the total sample by test occasion.
FIGURE 4.11. Age difference patterns of the cognitive style variables for the total sample across all occasions of measurement.
I next report findings from the 1956 baseline study, which found negative age differences on all five primary mental abilities but showed that peak ability ages occur later than observed in the previous literature and that the differential ability patterns noted in our first pilot study could be confirmed in this representative and age-continuous investigation. Cross-sectional findings are then reported for the six measurement occasions from 1956 to 1998, involving a total of 4,852 participants. For these data sets, I describe the differential ability patterns as they have changed in magnitude and pattern across time. Whereas there are increased performance levels at most ages in successive data sets, this shift is particularly noteworthy for women. Young adult females now show markedly higher performance on Spatial Orientation than was the case 42 years earlier, age differences in Numeric Ability have virtually disappeared, and Verbal Meaning now peaks in the early 50s. The magnitude of age differences, at least until the mid-60s and early 70s, has markedly decreased. The last finding, of course, has provided a strong rationale for the abandonment of mandatory retirement ages in all occupations.
Similar data are next provided for the expanded test battery given in 1984, 1991, and 1998, including the cross-sectional analysis of age differences on the latent constructs of Inductive Reasoning, Spatial Orientation, Verbal Ability, Numeric Ability, Perceptual Speed, and Verbal Memory. For four of these more broadly sampled constructs, the earliest evidence of reliable (although modest) negative age differences was found at somewhat earlier ages: Inductive Reasoning, Spatial Orientation, and Perceptual Speed at age 53 and Verbal Memory at age 46. The remaining two factors, Numeric and Verbal Ability, showed positive age differences until midlife, and even in advanced old age were at a higher level than for the youngest age group.
Cross-sectional data are also reported for the Educational Testing Service Basic Skills test, our measure of practical intelligence, in 1984 and 1991. No significant age differences were found until age 60, but increasingly severe age differences occurred from then on. Similar findings occurred for our current measure of practical intelligence, the Everyday Problems Test.
Finally, cross-sectional data are presented for our measures of cognitive style for the latent constructs of motor-cognitive flexibility (MCF), attitudinal flexibility (AF), and psychomotor speed (PS). Negative age differences in MCF and AF are currently first observed at age 53, whereas such differences appear already at 46 years of age for PS. Recent cohorts show average performance that is above earlier cohorts in the flexible direction at the same ages. At all ages, women exceed men on PS, whereas men exceed women on MCF and AF.
