This chapter reviews the central core of our study, the results from the longitudinal inquiries. As indicated in the description of the database, the longitudinal studies consist of seven 7-year follow-ups, six 14-year follow-ups, five 21-year follow-ups, four 28-year follow-ups, three 35-year follow-ups, two 42-year follow-ups, and one 49-year follow-up. The presentation of each individual segment of the longitudinal studies or the data and conclusions presented elsewhere are not repeated here (Hertzog & Schaie, 1986, 1988; Schaie, 1979, 1980a, 1980b, 1980c, 1983a, 1989a, 1996b; Schaie & Hertzog, 1983, 1986; Schaie & Labouvie-Vief, 1974; Schaie & Parham, 1977; Schaie & Strother, 1968b; Schaie &Willis, 2010). Instead, similar to the approach I adopted previously (Schaie, 1996b, 2005a), I again attempt to integrate the entire longitudinal database to provide estimates of age changes based on the largest available number of study participants for each age interval.
One of the major objectives of attempting to forecast ontogenetic change in individuals and of generating normative data on age changes in behavior is to be able to determine whether a particular individual change remains within the average range of interindividual differences in such change or whether the observed change is excessive and thus may provide a possible indicator of behavioral pathology or neuropathology. To obtain the requisite longitudinal estimates, it seems best to average over as many cohorts and times of measurement as possible to yield data whose stability is maximized by basing it on the largest possible number of observations. In the following sections, consequently, data are aggregated, whenever possible, across two or more samples observed at the same age (Nunnally, 1982).
To permit comparison with the cross-sectional findings (chapter 4), the base of our mean-level estimates was set to the observed average values across all cohorts for participants tested at age 60 (the median age of our total sample). The average intraindividual age changes aggregated across all cohorts for whom each age interval is available were then cumulated and added to or subtracted from these base values. In the following sections, these predicted values are provided for the total sample as well as separately by gender.
The five primary mental abilities and their two composite indices for which data are available over the entire study are considered first.
Intraindividual change estimates were computed by aggregating over all participants with data for 7-year intervals from mean ages 25 to 95 years based on 6,143 test records that were available for two points 7 years apart. Table 5.1 provides the resulting average within-participant age changes in T-score units, with positive values indicating gain from the age listed in the row to that listed in the column and negative values indicating age-associated decrement. The values in the diagonals of this table represent the observed within-group age changes. The off-diagonals are the cumulated changes obtained by summing the appropriate successive within-group values. These estimates are required to determine the ages at which decrement from some previously observed base age reaches statistical significance.
One can observe immediately that statistically significant cumulative age decrements from any previous age do not occur for any variable except Number prior to age 60. Several variables were found to have modest increments in young adulthood and middle age. The increment above the performance level observed at age 25 remained significant for Verbal Meaning until age 67, for Spatial Orientation until age 60, for Inductive Reasoning until age 53, and for Word Fluency until age 46. It was also found that cumulative age decrement, when taken from age 25, attained statistically significant magnitudes only at age 67 for Number and Word Fluency, at age 74 for Inductive Reasoning, and at age 81 for Verbal Meaning and Spatial Orientation.
The composite Index of Intellectual Aptitude showed statistically significant gain until age 60 and decline by age 67. The Index of Educational Aptitude showed gain until age 67 and significant decline by age 81. However, from the peak performance at age 46, decline was observed by age 67.
Significant gender differences were found for all five abilities (p < .01), with women excelling in Verbal Meaning, Inductive Reasoning, and Word Fluency, and men doing better than women in Spatial Orientation and Number. There are no statistically significant gender differences in the shapes of the age gradients. However, because of the level differences, cumulative decline over the entire adult age range is somewhat greater for women than for men on Verbal Meaning and Inductive Reasoning.
A visual representation of the resultant longitudinal age gradients from age 25 to age 95 is provided in figure 5.1 for the entire sample and separately for men and women. The longitudinal gradients are centered on the aggregated mean value at age 60 (the median age of our sample).
TABLE 5.1. Cumulative Age Changes for 7-Year Longitudinal Data in T-Score Points
FIGURE 5.1. Estimated age changes from 7-year data for the primary mental abilities for (a) the total sample and (b, c) separately by gender.
Longitudinal change in individuals can also be estimated using 14-year estimates. After aggregating age changes across the equivalent age ranges from the four 14-year data sets, estimates are derived similar to those given for the 7-year data. The major difference here is that all data come from 3,375 individual records that extend over 14 years; hence, these estimates are somewhat less sensitive to possible changes in rates of aging across successive cohorts but have the disadvantage of being based on smaller samples. The resultant estimates of age changes are given in table 5.2.
When age changes are examined over 14-year segments, such change becomes statistically significant for Number as early as age 53, for Word Fluency at age 60, and for the remaining three abilities at age 67. The 14-year changes were found to be significant for the Index of Intellectual Ability and the Index of Educational Aptitude by age 60.
The longitudinal age gradients resulting from these estimates are provided in figure 5.2. Note that, in contrast to figure 5.1a, the 7-year segments represent a rolling average obtained from the within-participant 14-year age changes. As a consequence, the resultant age gradients show a somewhat later attainment of peak levels of performance (in late middle age), and except for Number, decline does not become steep until the mid-70s are reached. Interestingly, in advanced old age, decline is now steepest for Verbal Meaning and Number, the two crystallized abilities.
Next, longitudinal changes over a 21-year period were estimated. Data used in these estimates were limited to the test records of individuals extending over 21 years (n = 1266). The individual age change estimates were based on even smaller samples. The resultant estimates of age changes are given in table 5.3.
TABLE 5.2. Average 14-Year Longitudinal Age Changes in T-Score Points
FIGURE 5.2. Estimated age changes from 14-year data for the primary mental abilities.
Although we recognize that those participants remaining in the study for as long as 21 years may be an increasingly select sample, it is still interesting to point out that the rate of average decline was somewhat less for these persons. For the 21-year segments, modest but significant decrements are noted for Number and the Index of Intellectual Ability by age 60 and for the remaining variables by age 67. Cumulative decrements estimated from the three samples that cover the entire age range from 25 to 88 years amount to 0.5 SD for Verbal Meaning, 0.7 SD for Spatial Orientation and Inductive Reasoning, 1.0 SD for Number, and 0.9 SD for Word Fluency.
The longitudinal age gradients resulting from these estimates, averaging across 7-year segments, are provided in figure 5.3. Because of the longer within-cohort age ranges covered by the same participants, these gradients are even smoother than for the 14-year data; the major difference is a somewhat less-steep decrement for Verbal Meaning.
TABLE 5.3. Average 21-Year Longitudinal Age Changes in T-Score Points
FIGURE 5.3. Estimated age changes from 21-year data for the primary mental abilities.
Three data sets are available for participants who were followed for 28 years (n = 578). Again, data were aggregated for the comparable age ranges, and average longitudinal changes across the available 28-year ranges are given in table 5.4. In this even more select group, significant decrements over the 28-year segments are first observed for Number and the Index of Intellectual Ability by age 60; for Inductive Reasoning, Word Fluency, and the Index of Educational Aptitude by age 67; and for Verbal Meaning and Spatial Orientation by age 74. The estimated longitudinal gradients are shown in figure 5.4. For these participants, there was only modest average decline by age 74, with steep decline first observed by age 81.
TABLE 5.4. Average 28-Year Longitudinal Age Changes in T-Score points
FIGURE 5.4. Estimated age changes from 28-year data for the primary mental abilities.
Three data sets contain individuals who have been followed over 35 years (n = 377). Data aggregated for comparable age ranges are provided in table 5.5 and are charted in figure 5.5. In these data, significant decline is observed for Number, Word Fluency, and the Index of Intellectual Ability only by age 67 and for the remaining variables by age 74.
Two data sets contain individuals followed over 42 years; data are provided for the small residual sample that has participated in our entire study (n = 149; see table 5.6). Findings are quite similar to those for individuals followed for 35 years. In these data, significant decline is observed by age 67 for Verbal Meaning and Inductive Reasoning, and by age 88 for Spatial Orientation, while no significant decline is noted for Number, Word Fluency, and the indices of Intellectual Ability and Educational Aptitude even by age 88. This group of study participants undoubtedly represented an unusual group of the successfully aging.
TABLE 5.5. Average 35-Year Longitudinal Age Changes in T-Score Points
FIGURE 5.5. Estimated age changes from 35-year data for the primary mental abilities.
Finally, 49-year data are provided for the small residual sample that has participated in our entire study (n = 26; see Table 5.7). Findings are similar for those in the 42-year data, and because of the small number, statistically significant onset of decline is difficult to interpret.
Longitudinal data over 7 and 14 years are now available on the additional variables that entered the expanded test battery in 1984. In addition, this section contains the results for the longitudinal estimates of the latent ability constructs.
Intraindividual change estimates were computed for all variables added in the 1984 cycle. It should be noted that the longitudinal estimates provided in table 5.8 and the resultant longitudinal gradients over the age range from 25 to 95 years shown in figure 5.6 were limited to four data points (1984, 1991, 1998, and 2005); that is, each age segment was based on three samples followed over 7 years. However, there are now a total of 3,374 observations over 7 years.
TABLE 5.6. Average 42-Year Longitudinal Age Changes in T-Score Points
TABLE 5.7. Average 49-Year Longitudinal Age Changes in T-Score Points
All three additional markers of Inductive Reasoning ability show significant 7-year decline only by age 67. For Spatial Orientation ability, the parallel forms for PMA Space and Object Rotation, as well as the three-dimensional rotation test, Cube Comparison, show decline by age 67. However, the parallel form of Alphanumeric rotation declines significantly only by age 74. Of the new Perceptual Speed measures, Identical Pictures declines as early as age 46, Number Comparison declines by age 67, but Finding A’s declines significantly only by age 88. Both new markers of Numeric Facility, Addition, and Subtraction and Multiplication, show a significant decline by age 53. The new unspeeded tests of Verbal Comprehension only show significant decline by age 81. Finally, the new Verbal Memory markers show decline by age 67.
Data are also provided on the new marker variables for the 1,464 study participants for whom 14-year data were available (see table 5.9). Conclusions as to age of onset of cognitive decline from these smoother data are as follows:
Inductive Reasoning and Spatial Orientation. All new markers of these factors show initial statistically significant decline by age 67.
Perceptual Speed. Initial significant decline is observed as early as age 53 for the Identical Pictures test, at age 67 for Number Comparisons, but as late as age 81 for the Finding A’s test.
Numeric Facility. Both new markers of this ability show earliest statistically significant decline by age 53.
Verbal Memory. Initial significant decrement is observed for Immediate Recall by age 67, and for Delayed Recall by age 74.
Verbal Comprehension. Both new measures show gain in level of performance through most of adulthood with first significant decline beginning only at age 81.
TABLE 5.8. Intraindividual Age Changes Over 7 Years for the Extended Cognitive Test Battery
FIGURE 5.6. Estimated age changes from 7-year data for the expanded test battery. ETS = Educational Testing Service.
TABLE 5.9. Intraindividual Age Changes Over 14 Years for the Extended Cognitive Test Battery
Given the availability of multiple markers, 7-year longitudinal estimates of change within participants were computed for the latent ability constructs. The resulting estimates are provided in table 5.10, and longitudinal gradients for the total sample and separately by gender can be found in figure 5.7. These gradients are centered on the actually observed mean for the median age group in our sample (age 60). With respect to these latent construct factor score estimates, earliest reliably observed decline over 7 years occurs for Numeric Ability by age 53; for Perceptual Speed by age 60; for Inductive Reasoning, Spatial Orientation, and Verbal Memory by age 67; and for Verbal Ability by age 81.
When age changes for the latent ability constructs are examined for the smaller numbers of study participants who were available over 14-year segments (N = 1464), such change becomes statistically significant for Number Facility as early as age 53; for Perceptual Speed at age 60; and for Inductive Reasoning, Spatial Orientation, and Verbal Memory at age 67. However, statistically significant decline is observed for Verbal Comprehension only as late as age 81 (see table 5.11 and figure 5.8).
The magnitude of decline for the longitudinal data is substantially lower for several latent abilities than would be suggested by cross-sectional data. Thus, there is only a modest decline from young adulthood to advanced old age for Verbal Ability (0.4 SD). For Inductive Reasoning and Spatial Orientation, longitudinal change from age 25 to age 88 amounts to 0.8 SD. However, longitudinal estimates of change are close to cross-sectional estimates for Perceptual Speed (1.2 SD) and Verbal Memory (1.1 SD) and exceed the cross-sectional estimates for Numeric Ability (1.5 SD).
This section reports intraindividual change estimates for our measures of practical intelligence. These estimates were based on only two data points because the Basic Skills test was replaced by the Everyday Problems Test (EPT) in 1998. The longitudinal estimates for the practical intelligence measures are reported in table 5.12 and graphed in figure 5.9. What is most noteworthy about these data is that peak performance for these measures is reached only by age 60 and that steep decline is noted only by age 81. Thereafter, decline in average performance is quite dramatic, amounting to approximately 2 SD from the 60s to the 80s.
Longitudinal data are also presented for the Everyday Problems Test (Willis, 1992b), based on two-point data from 1998 to 2005. Figure 5.10a presents the estimated means for the total EPT scores, while Figure 5.10b gives the longitudinal means for the seven individual activity dimensions. Note that total EPT scores rise slightly into early midlife and remain stable until age 74. Thereafter there is rapidly accelerating decline as advanced ages are reached. Similar patterns are shown for the separate activities of daily living. Note, however, that the asymptotic level of meal preparation occurs in the mid-20s and consumer activities in the late 30s. In very advanced old age, financial activities and health promotion activities decline the most, while meal preparation is most preserved.
TABLE 5.10. Intraindividual Age Changes Over 7 Years for the Latent Construct (Factor) Scores
FIGURE 5.7. Estimated age changes from 7-year data for the latent ability constructs.
TABLE 5.11. Intraindividual Age Changes Over 14 Years for the Latent Construct (Factor) Scores
FIGURE 5.8. Estimated age changes from 14-year data for the latent ability constructs.
TABLE 5.12. Cumulative Age Changes for the Measures of Practical Intelligence in T-Score Points
FIGURE 5.9. Estimated age changes from 7-year data for the ETS measure of practical intelligence.
The longitudinal estimates for the cognitive style data again involve changes averaged over six time periods and are based on all participants for whom 7-year data were available. Table 5.13 provides the longitudinal estimates, and figure 5.11 graphs the longitudinal gradients for the total sample and separately by gender. First, examining the age gradients for the total sample, we note a small longitudinal increment (cumulatively about 0.2 SD) for both Motor-Cognitive Flexibility and Attitudinal Flexibility to age 60 and a moderate cumulative decline of 2/3 SD thereafter. Psychomotor Speed increases by approximately 1/2 SD from age 25 to a peak at age 60. In contrast to the flexibility factor, this is followed by a decline of about 1 SD by age 88.
FIGURE 5.10. Estimated age changes from 7-year data for the Everyday Problems Test (EPT).
Examining the longitudinal findings for the cognitive style data by gender, we note continuing gain by males on Motor-Cognitive Flexibility until about age 60, with moderate decline thereafter. Women, on the other hand, show a more modest gain until age 60, with greater decline than men thereafter. As for Attitudinal Flexibility, men show early virtual stability until about 60, with moderate decline noticeable by age 67. For women, there is a small increase in Attitudinal Flexibility from 25 to 60, with modest decline (about 2/3 SD) thereafter. Both men and women peak on Psychomotor Speed at age 60, with decline thereafter. Over the entire age range, men decline somewhat more than women (2/3 SD for men, 1/3 SD for women).
TABLE 5.13. Cumulative Age Changes for 7-Year Cognitive Style Data in T-Score Points
FIGURE 5.11. Estimated age changes from 7-year data for the cognitive style variables for the total sample and separately by gender.
The presentation of the longitudinal findings begins with an examination of the within-participant estimates obtained from aggregating across participants whose data are available over 7, 14, 21, 28, 35, 42, and 49 years. Next, 7-year longitudinal data for the additional marker variables added beginning with the 1984 data collection are considered. Longitudinal estimates are provided for the latent ability construct factor scores, and for the measures of practical intelligence they are also provided based on 7-year data. Similar longitudinal data are then presented for the measures of cognitive style.
Longitudinal age changes are generally less pronounced than the cross-sectional data for most variables, with modest decline beginning in the early 60s and marked decline not occurring until the 80s are reached. The major exceptions to these findings occur for the Number ability, which begins to decline in the 50s. Cumulative decline is somewhat larger for men than for women on Verbal Meaning and Inductive Reasoning.
The 7-year data represent the most conservative estimates of within-participant change because they are based on large samples. Data for the same individuals collected over a longer period of time (up to 49 years), who may be the select survivors of our study, show average maintenance of many abilities into the mid-70s.
For the more broadly marked ability constructs, there is an even more dramatic difference between the cross-sectional and longitudinal findings. In the longitudinal data, there is only modest decline from young adulthood to advanced old age for Verbal Ability; somewhat greater declines are seen for Inductive Reasoning, Verbal Memory, and Spatial Orientation until the 80s are reached. However, longitudinal estimates of change equal the cross-sectional ones for Perceptual Speed and exceed cross-sectional estimates for Numeric Ability. The profound age decline on Numeric Ability in particular seems hidden by negative cohort trends. A peak is reached for our measure of practical intelligence by age 60, but steep decline on that measure is not observed prior to the 80s.
Finally, on the measures of cognitive style, increment is noted for Motor-Cognitive and Attitudinal Flexibility to age 60, with modest decline thereafter. Psychomotor Speed also shows small increases until age 60 and decline thereafter. However, when examined by gender, women have greater decline on Motor-Cognitive and Attitudinal Flexibility, but less decline than men on Psychomotor Speed.
