Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
00001.tif
(USC Thesis Other)
00001.tif
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
INFORMATION TO USERS
This manuscript has been reproduced from the microfilm master. UMI
films the text directly from the original or copy submitted. Thus, some
thesis and dissertation copies are in typewriter face, while others may be
from any type of computer printer.
The quality of this reproduction is dependent upon the quality of the
copy submitted. Broken or indistinct print, colored or poor quality
illustrations and photographs, print bleedthrough, substandard margins,
and improper alignment can adversely affect reproduction.
In the unlikely event that the author did not send UMI a complete
manuscript and there are missing pages, these will be noted. Also, if
unauthorized copyright material had to be removed, a note will indicate
the deletion.
Oversize materials (e.g., maps, drawings, charts) are reproduced by
sectioning the original, beginning at the upper left-hand comer and
continuing from left to right in equal sections with small overlaps. Each
original is also photographed in one exposure and is included in reduced
form at the back of the book.
Photographs included in the original manuscript have been reproduced
xerographically in this copy. Higher quality 6” x 9” black and white
photographic prints are available for any photographs or illustrations
appearing in this copy for an additional charge. Contact UMI directly to
order.
UMI
A Bell & Howell Information Company
300 North Zeeb Road, Ann Arbor MI 48106-1346 USA
313/761-4700 800/521-0600
COGNITIVE PROCESSES AND LIFESTYLE INDICATORS OF FLUID
REASONING DECLINE AND CRYSTALLIZED KNOWLEDGE
ENHANCEMENT IN ADULTHOOD
by
Jennie Gale Noll
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSITY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(Psychology)
August 1995
Copyright 1995 Jennie Gale Noll
UMI Number: 9621623
UMI Microform 9621623
Copyright 1996, by UMI Company. All rights reserved.
This microform edition is protected against unauthorized
copying under Title 17, United States Code.
UMI
300 North Zeeb Road
Ann Arbor, MI 48103
UNIVERSITY OF SOUTHERN CALIFORNIA
THE GRADUATE SCHOOL
UNIVERSITY PARK
LOS ANGELES. CALIFORNIA 90007
This dissertation, written by
under the direction of h£.C Dissertation
Committee, and approved by all its members,
has been presented to and accepted by The
Graduate School, in partial fulfillment of re
quirements for the degree of
DOCTOR OF PHILOSOPHY
CP—
Dean of Graduate Studies
Date
S u^t/C M S '
DISSERTATION COMMITTE [ON COMMITTEE
Chairperson
ii
For Kek
Acknowledgments
1 thank the American tax payer for providing the research funds allocated to
Forming Science Careers in Developmental Neurocognition, NIA T32 AGOO156-05.
and Adulthood Age Differences in Cognitive Abilities, NIAR01 AG09936-05.
making both this project and my graduate education possible
Special thanks go to Jeff, Ruby, Jack, Mary, Mark, Bruce, Sam, Monette,
Allan, Mitch, Penny, Emily, Mom, and Pop for helping to smooth out some of my
rough edges and for providing much needed shelter from the storm
Thanks to the several artists whose work provides unending inspiration— Amy
Taylor, Charlie Papazian, and Glen Gould.
I am grateful to Professor John Horn who has the heart of a scientist and the
soul of a poet His scientific and psychometric genius pales only in comparison to his
integrity, generosity, persistence, and loyalty.
To Dr Lief Noll I extend gratitude beyond words Thank you for bringing
humor to chaos, hope to despair, confidence to doubt, and love to life I cannot
imagine a better soul-mate. As we reflect on the dreams of our first ten years, we
begin our lives together again.
iv
Table of Contents
Dedication .................................................................................................................. ii
Acknowledgments ..................................................................................................... iii
List of Tables ............................................................................................................ vii
List of Figures ........................................................................................................... ix
Abstract ..................................................................................................................... x
STUDY 1: THE ROLE OF COGNITIVE PROCESSES IN THE DECLINE OF FLUID
REASONING AND THE ENHANCEMENT OF CRYSTALLIZED KNOWLEDGE
Introduction .................................................................................................. 1
Historical Overview ......................................................................... 1
Evidence From Studies of the WAIS ............................................. 4
The When and Why of Vulnerable Ability Decline ........................ 5
Short-term Apprehension Retrieval .................................... 9
Cognitive Speediness .......................................................... 10
Concentration/Attention ..................................................... 11
Processes that Aid Performance ...................................................... 12
Summary ........................................................................................... 14
Method .......................................................................................................... 16
Subjects ............................................................................................. 16
Ability Measures ............................................................................... 19
Fluid Reasoning (Gf) ............................................................ 19
Crystallized Knowledge (Gc) ............................................... 21
Speed of Processing or Cognitive Speed (Gs) ................... 21
Concentration (CON) ......................................................... 22
Short-term Apprehension Retrieval (SAR) ........................ 23
Procedures ........................................................................................ 25
Analyses ........................................................................................... 25
Results .......................................................................................................... 29
Psychometric Considerations .......................................................... 32
Partitioning the Age Related Decline of Fluid Ability ................... 34
Power Letter Series as an Indicator o f Fluid Ability ........ 34
Common Word Analogies as an Indicator of Fluid Ability 41
The Age Related Enhancement of Crystallized Knowledge ........ 46
Vocabulary 2 as a Crystallized Ability Indicator .............. 47
Vocabulary 1 as a Crystallized Ability Indicator .............. 52
Esoteric Analogies as a Crystallized Ability Indicator ..... 56
Summary .......................................................................................... 59
Discussion ..................................................................................................... 62
V
STUDY 2 LIFESTYLE INDICATORS OF COGNITIVE ABILITIES
Introduction .................................................................................................... 67
The Operational Definition of Lifestyle ............................................ 67
Physical Health ....................................................................... 72
Activity Level ......................................................................... 76
Psychological Health ............................................................. 79
Personal Control .................................................................... 84
The Unidimensional Theory ...................................... 87
The Multidemensional Theory ................................. 87
The Development of Perceived Control .................. 91
Personal Control Related to Psychopathology ....... 94
Evidence for an Eight Factor Theory ...................... 96
Perceived Control and Cognitive Functioning ........ 97
Expectations (Hypotheses) ............................................................... 100
Methods ......................................................................................................... 105
Subjects ............................................................................................... 105
Measures ............................................................................................. 106
Physical Health Measures ......................................................107
Activity Measures ................................................................. 109
Psychological Health Measures ............................................110
Personal Control Measures ..................................................111
Analyses ............................................................................................. 112
Missing Data Imputation ..................................................... 112
Measurement Invariance ..................................................... 114
Part and Partial Correlational Procedures ........................... 115
Understanding the Portion of Ability Variance that is not
Associated with Aging ..................................................... 117
Results ........................................................................................................... 118
Factor Analyses and Measurement Invariance ............................... 118
Relationships Among Lifestyle and Cognitive Variables ............... 127
Interpretation of the Part Correlational Results ............................. 134
Short-term Apprehension Retrieval ..................................... 135
Concentration ...................................................................... 141
Cognitive Speediness .......................................................... 141
Residual Fluid Reasoning .................................................... 141
Residual Crystallized Knowledge ....................................... 145
Non-Overlap Multiple Variable Relations for Ability Components 145
Discussion ..................................................................................................... 151
Lifestyle Indicators of the Development of Cognitive Abilities ... 152
Lifestyle Variables that Explain Age-Independent Cognitive
Abilities .............................................................................................. 161
Possible Age Cohort Explanations .................................................. 163
“Semi-Real” Ability Differences .......................................... 165
Summary and Conclusions ............................................................. 166
References ................................................................................................................ 170
Appendix A: Summary Statistics for all Cognitive Variables and Lifestyle Indicators
used in Analyses .......................................................................................... 189
Appendix B: Lifestyle Factors Used in Analyses .................................................... 192
Appendix C: Procrustes Reference Vector Structure for 145 Variables and 16
Factors ......................................................................................................... 200
Appendix D: Causal Inferences ............................................................................... 206
Vll
List o f Tables
Table 11. Means, Standard deviations and Ranges fo r Cognitive Variables and
Age. Breakdown fo r Gender also shown (N=577) .................................... 24
Table 12. Zero-order Correlations Between Age, Gender, and Cognitive Ability
Variables (N 577) ....................................................................................... 31
Table 1.3. Effects o f Parting Cognitive Processes on the Correlations Between Age
and Indicators o f G f Based on Power Letter Series (PLS) ........................ 36
Table 1.4. Effects of Parting Cognitive Processes on the Correlations Between Age
and Indicators o f G f Based on Common Word Analogies (CWA) ............ 42
T able 15 Effects of Parting Cognitive Processes on the Correlations Between Age
Indicators o f Gc Based on the Second Vocabulary (V0C2) Test ............. 48
Table 1.6. Effects o f Parting Cognitive Processes on the Correlations Between Age
Indicators ofGc Based on the First Vocabulary (VOCJ) Test ................. 54
Table 1.7. The Effect o f Parting Cognitive Sub-processes on the Correlations
Between Age and Various Esoteric Analogies (ESA) Criterion Variables 57
Table 2.1. Intercorrelations Among Lifestyle Factors (N=577) .......................... 121
Table 2.2 Measurement Invariance Tests fo r Lifestyle Factors .......................... 124
Table 2.3. Zero-order Correlations o f Lifestyle Variables with Age, Memory for
Paired Associates (MPA), Concentration (CON), Speeded Letter Comparison
(SLC), residual Fluid Reasoning (rGf), and residua! Crystallized Knowledge
(rGc) .............................................................................................................. 128
Table 2.4. Lifestyle Variables that Account fo r the Age Related Decline o f Memory for
Paired Associates (MPA) .............................................................................. 136
Table 2 5 Lifestyle Variables that Account fo r the Age Related Decline o f
Concentration (CON) ................................................................................... 140
Table 2.6 Lifestyle Variables that Account fo r the Age Related Decline o f Speeded
Letter Comparison (SLC) ............................................................................ 142
VM
Table 2.7 Lifestyle Variables that Account fo r the Age Related Decline o f Residualized
Fluid Ability (rGf)— Fluid Ability with Short-term Apprehension Retrieval,
Concentration, Clerical Speed, and Crystallized Ability Variance
Removed .......................................................................................................... 143
Table 2.8. Lifestyle Variables that Account fo r the Age Related Decline o f Residualized
Crystallized Ability (rGc)— Crystallized Ability with Short-term Apprehension
Retrieval, and Fluid Ability Variance Removed ........................................ 144
Table 2 9. Correlations o f Lifestyle Variables with Age, Memory for Paired
Associates (MPA), Concentration (CON), and Speeded Letter Comparison
(SLC)— Part Correlations where Age has been Removed from Lifestyle
I'ariables Also Shown ................................................................................... 146
Table 2.10. Correlations o f Lifestyle Variables with Age, Residualized Fluid Ability
(rGf), and Residualized Crystallized Ability (rGc)— Part Correlations where
Age has been Removedfrom Lifestyle Variables Also Shown .................. 147
Table 2.11 Regression Models that Account fo r the Most, Non-Redundant Variance
o f Memory fo r Paired Associates (MPA), Concentration (CON), and Speeded
Letter Comparison (SLC), Residual Fluid Ability (rGf), and Residual
Crystallized Knowledge (rGc) Above and Beyond Age ............................. 149
Table 2.12. Summary o f Findings ........................................................................... 153
T able A-1 Means and Standard Deviations fo r Standardized (mean 0, sld I)
Cognitive Variables Across Several Different Age Groups ...................... 190
Table A-2 Means and Standard Deviations o f Variables Used in A nalyses for Three
Age Groupings—Variables standardized (mean=0, std 1) over entire sample
(N 577) ...... 191
Table C-1 . Procrustes Reference Vector Structure for 145 Variables and 16
Factors .......................................................................................................... 200
Table D-2 Cross-lag Causa! Parameters (fi j) when Autoregressive Effects ( 0 ,3,
02,d are Fixed to a Range o f Plausible Test-retest Correlations— f}t 2
Parameters for 5 Values o f 0 i3 and 2 Values o f 02,4 are Shown ...............213
IX
List of Figures
Figure 1.1 Controlling Various Cognitive Processes to Account for (Decrease)
Fluid Ability (Gf) and Speed (Gs) Decline— After Horn, Donaldson, Engstrom
(1981) ............................................................................................................ 8
Figure 1.2. Controlling fo r Persistence (PER), Carefulness (CAR) and Crystallized
Knowledge (Gc) Aging to Increase the Decline o f Fluid (Gf) Ability— After
Horn, Donaldson, Engstrom (1981) ..............................................................13
Figure 1.3a The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Decline o f Fluid Ability as Measured by Power Letter Series
(PLS) ............................................................................................................. 37
Figure 1 3b. The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Decline o f Power Letter Series with Vocabulary Test 2
Controlled (PLS. VOC2) .............................................................................. 40
Figure 1 4a. The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Decline o f Common Word Analogies (CWA) as an Indicator
o f Fluid Ability ............................................................................................. 44
Figure 1 4b. The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Decline o f Common Word Analogies with Vocabulary I and
2 Controlled (CWA :VOC2 VOC1) ............................................................ 45
Figure 1.5a. The Effect o f Removing the Variance o f Memory for Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Enhancement o f Vocabulary Test 2 (VOC2) as a Crystallized
Ability Indicator .......................................................................................... 49
Figure 1.5b. The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Enhancement o f Vocabulary Test 2 when Power Letter Series
is Controlled (VOC2 :PLS) .............................................................................51
Figure 1.6 The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Enhancement o f Vocabulary Test 1 (VOC1) alone and with
Power Letter Series Controlled (VOCJ. PLS) ............................................. 55
Figure 1.7 The Effect o f Removing the Variance o f Memory fo r Paired Associates
(MPA), Concentration (CON), and Speeded Letter Comparison (SLC) from
the Age Related Enhancement o f Esoteric Analogies (ESA) as an Indicator of
Fluid Ability .................................................................................................. 58
Figure D 1 Latent Longitudinal Model (Golloh & Reichardt, 1987) 208
Abstract
The principal aim of this research was to advance the understanding o f the
development of Fluid Reasoning (Gf) and Crystallized Knowledge (Gc) abilities.
Previous research suggested that there are cognitive processes involved in the aging
decline of Gf and the aging enhancement of Gc. Replication of these findings
provided a foundation for the examination of how lifestyle indicators influence the
development of abilities
In Study 1, the extent to which cognitive abilities such as Short-term
Apprehension Retrieval (SAR), Concentration (CON), and Cognitive Speed (Gs) can
be thought of as processes involved Gf and Gc functioning was examined. Results
indicated that SAR, CON, and Gs are processes that, when controlled, help explain
the aging decline of Gf ability in a cross-sectional study of 577 adults ranging in age
from 22 to 92. Significant Gf decline remained even after the variance of these
processes was removed. SAR was shown to be a process that, when controlled,
affects the aging enhancement of Gc ability. Results also suggested that G f and Gc
are not independent from one another.
In Study 2, the relationships between indicators of lifestyle and cognitive
abilities were examined. Income, Information Seeking, Attention to Healthy Diet,
Secure Attachment, Confiding in Others, a lack External Belief Systems, and a large
Number of Children were shown to be lifestyle indicators that account for the aging
decline of the vulnerable cognitive processes that are involved in Gf functioning.
Decreased residual Gf functioning was shown to be a behavioral consequence
of anoxia resulting from alcohol abuse suggesting that the part of Gf that cannot be
accounted for by SAR, CON, Gs, or Gc is closely tied to biological functioning.
The age-related enhancement of residual Gc was shown to be positively
related to Information Seeking, Optimism, Tolerance for Disorder, Education, and
negatively related to External Belief Systems. The part of Gc that cannot be
explained by Gf or SAR is chiefly related to cultural immersion, goal directedness,
and engagement.
STUDY 1: THE ROLE OF COGNITIVE PROCESSES IN THE DECLINE OF
FLUID REASONING AND THE ENHANCEMENT OF CRYSTALLIZED
KNOWLEDGE
Introduction
A major purpose of this research is to describe age differences in cognitive
abilities. To do this it is necessary to describe the procedures employed by
researchers who study such differences. Conclusions about intellectual development
will depend on which abilities are studied, under what conditions these abilities are
assessed, and the theory under examination. Theory dealing with general intelligence
"g" must be distinguished form theory of multiple cognitive capabilities. Similarly,
measurement under timed (speeded) conditions must be distinguished from
measurement under untimed (power) conditions. The distinction between cross-
sectional and longitudinal data collection is particularly important for understanding
inferences about development. Analysis of how these matters of theory, design,
perspective, and measurement pertain to understanding this research will be
considered in the next section prior to discussion of the extant evidence.
Historical Overview
Inspired by Gabon's Inquiries into the Human Faculty (1883), Spearman
[1863-1945] explored the notion that abilities commonly taken to be "intellectual" are
intercorrelated with each other In a seminal paper Spearman (1904) demonstrated
how such correlations might be explained in terms of what he referred to as a “g”
factor (usually interpreted as indicating general intelligence), and a number specific
factors. Spearman observed that the correlations among various ability tests were
positive, a condition he labeled positive manifold. He demonstrated conditions under
which this positive manifold could be accounted for, to some quantifiable extent, by a
single factor common to all the variables. These conditions specify a rigorous set of
hypotheses— a model— that can be tested with the intercorrelations obtained from a set
of ability measures. Spearman devoted a major part of his professional life to looking
for evidence in support of this model and attempting to explicate the psychological
nature of cognitive capacities.
Many researchers and the results of many studies questioned the findings and
theory of Spearman (e.g., Anderson, 1939; Burt, 1909; El Koussey, 1935; Kelly,
1928; Rimoldy, 1948; Thomson, 1916; andTryon, 1933). The most concerted
questioning, leading to a quite different formulation, was conducted by Thurstone
(1938, 1947). The structural system of Primary Mental Abilities (PMA) put forth by
Thurstone specified that the organization of mental abilities is not unifactorial; it is
multiple factored. Thurstone found support for a model in which no fewer than nine
common factors were required to describe the reliable individual differences variance
obtained with tests designed to measure salient features of intelligence. With the
advent of the theory of primary mental abilities, more and more evidence accumulated
to indicate the multiple ability nature of human intelligence. Systems of no fewer than
28 factors emerged. It became clear that several factors at the same general level as
Thurstone’s were needed to explain the relationships among ability tests.
But it was not possible to study such large systems with the resources
available for developmental research. A simplified system that retained the essential
information of multiple factor findings was needed. Gf-Gc theory (Cattell, 1957,
1971; Horn, 1965, 1968, 1988; Horn & Cattell, 1966) was put forth partly in
response to this need. This is a theory that organizes the many PMA first-order
factors into a system of second-order factors. It deals with a relatively small number
of basic cognitive processes that can be studied with reasonable resources.
3
Gf-Gc structural theory has been replicated in many studies (Carroll, 1989;
Hastain & Cattell, 1978; Horn & Bramble, 1967; Horn & Stankov, 1982), under
many different conditions of small and large samples— up to 6,400 (Woodcock, 1990),
different cultures (Gustafsson, 1984; Undheim, 1987), and in different times (e.g.,
Horn & Cattell, 1967: Carroll, 1989). Results indicate that the PMA system can be
organized into nine dimensions that are almost as broad as the sets of abilities people
refer to when they use terms such as intelligence or IQ
Briefly, the nine broad abilities of Gf-Gc theory are. Fluid Reasoning (Gf),
Acculturation Knowledge (Gc), Quantitative Knowledge (Gq), Short-term
Apprehension-retention (Gsm), Fluency of Retrieval from Long-term Storage (Glr),
Visual Processing (Gv), Auditory Processing (Ga), Processing Speed (Gs), and
Correct Decision Speed (CDS).
These different abilities have different developmental patterns. Basically,
there are two broad classes: the class of vulnerable abilities (Gf, Gs, Gv, and Gsm)
and maintained abilities (Gc and Glr; Horn, 1985; Horn, Donaldson, & Engstrom,
1981; Horn & McArdle, 1980; Schaie, 1989, 1990). The vulnerable abilities decline
(on average) in adulthood (Gf, Gs and Gsm starting in the early 20’s, Gv and Ga
starting in the 40’s). At the same time in the same subjects, averages for the
maintained abilities either increase or do not decline until late in adulthood, and then
only moderately relative to the decline of vulnerable abilities (Schaie & Baltes, 1977).
If vulnerable and maintained abilities are combined in a mixture (which might
be called ”g” or IQ), and the distinction between abilities that decline with age and
abilities that improve with age is lost. Depending on the amounts of each ability in
the mixture, it may appear that such a conglomerate ability improves with age,
declines with age, or does not change with age.
4
Evidence From Studies o f the WAIS
The confounding of abilities having different developmental courses occurs in
the IQ measure of the Wechsler Adult Intelligence Scales (WAIS). The Verbal (V)
and the Performance (P) subscales rather imperfectly measure maintained and
vulnerable abilities respectively. Older subjects do worse (on average) than do
younger subjects on the P subscale Over much of adulthood there is no aging decline
in the V scales (up through about age 60-70 years). These different aging trends have
been described as the "classic aging pattern" present for both men and women, for
white and black subjects, for different SES levels, and for persons classified as
abnormal (Botwinik, 1977; Eisdorfer, Busse, & Cohen, 1959). When the V and P
measures are combined, resulting in the full scale IQ, the averages for older subjects
usually are found to be below the averages for younger samples (Doppelt & Wallace,
1955).
Recent analyses (McArdle, 1984) of both longitudinal and cross-sectional
adult differences in the V and P subscales of the WAIS indicate that the slope for the
curve of the averages for P is consistently and rather strongly negative, but the slope
for the curve of the averages for V is near zero or slightly positive in most of
adulthood, and weakly negative over the most advanced ages. The variances are
similar at each age for P, but they increase substantially from age 40 to 75 for V.
V and P are not isomorphic with Gc and Gf respectively. Although the V
subscale is often interpreted as indicating Gc and the P subscale is interpreted as
indicating Gf, the aging curves for V do not show as prominent an aging increase as
Gc, and the age decline for P is not as steep as for Gf. O f the six tests that define the
V subscale, only four are indicative of Gc: (Information, Similarities, Vocabulary,
and Comprehension). Digit Span, although part of V, is indicative of the vulnerable
5
ability Gsm (short-term memory). Arithmetic, also a part of V, is indicative of Gsm
and Gs, both vulnerable abilities, as well as Gq (quantitative ability). Combining
vulnerable abilities with maintained abilities masks the age increases that have been
shown to be present for more nearly pure factor measurements of Gc.
Four of the five tests that make up the P subscale (Picture Completion, Picture
Arrangement, Block Design, and Object Assembly) may be more indicative of broad
visualization (Gv) than they are of Gf The remaining P test (Digit Symbol) is most
indicative of speed (Gs). In general, the age curves for P reflect those that are
commonly seen for pure factors of vulnerable abilities.
Factor analytic studies of the WAIS-R further support the need for careful
interpretation of V and P age curves A one factor model indicative of the full scale
IQ score (McArdle & Horn, 1983) does not fit the data (x2=2982, df=l 52) A two
factor model (Kaufman, 1979) also shows a lack of fit (x2=1034, df=142). A model
that does seem to fit the data reasonably well (Woodcock, 1990, McArdle & Horn,
1983) is a four factor solution which indicates Gc, Gv or Gf, Gq, and Gs (the latter
two are severely underdetermined). Combining the ability tests of the WAIS-R to
measure V or P subscales thus can result in inadequate interpretations of age curves.
The When and Why of Vulnerable Ability Decline
Are age differences in abilities in adulthood indicative of age changes? When
in the lifespan does decline of vulnerable abilities begin to become prominent7
Results from studies of Horn and his colleagues, whose results are mainly derived
from cross-sectional studies, indicate that Gf, Gs, and Gsm begin to show decline in
early adulthood (mid to late 20's) Schaie (1979, 1983) found such results for Gs, but
for Induction, a marker for Gf, the longitudinal evidence suggested decline occurring
much later in adulthood.
6
The differences in these findings needs to be interpreted cautiously. Test-
retest effects produce an improvement in scores on the second (or later) testing in
longitudinal data. Usually these effects have not been partialed out in longitudinal
results (Horn & Donaldson, 1976, 1980, 1992). Also, if time between testings is
short, no age differences can be detected unless sample sizes are very large. When
care is taken to ensure that times between retesting are large enough for change to
occur, and samples are large enough to provide adequate statistical power,
longitudinal results are basically in agreement with the cross-sectional results.
Why do abilities change with age? The maintained abilities reflect
accumulated knowledge, experience, and exposure to culture. Such knowledge and
experience continues to accumulate over adulthood. Older adults benefit from this
accumulation and thus score higher on tests of Crystallized Knowledge (Gc) than do
younger adults. Since it is required that one posses information before it can be
recalled, and one must be able to recall information in order for knowledge to be
demonstrated, Retrieval from Long-term Storage (Glr), is closely tied to Gc
performance. Glr also reflects the continuing restructuring of knowledge (Broadbent,
1966) that makes it increasingly more accessible. Older adults, when given enough
time, do better, on average, on Glr tests than do younger subjects.
The story is not as clear cut for the vulnerable abilities. It is important to
distinguish between speed and power tests in the study of cognitive capabilities. A
speeded test is one that emphasizes the quickness of the respondent-the faster one
can correctly respond to the items, the higher the score. A power test is given under
untimed conditions that allow all subjects to have an equal chance at addressing all
levels of difficulty. Such power measures of Gf ensure that individual differences are
not due to speediness, but instead indicate the true level of functioning.
7
In order to do well on a power measure of Fluid Reasoning (GO one must be
able to hold information in immediate awareness, manipulate that information
accurately, integrate that manipulation into the novel sequence quickly before it leaves
immediate awareness, and persist in this action until the problem is solved. Short
term apprehension retrieval, encoding organization, working memory, cognitive
speed, concentration and attention abilities have all been shown to account for the
aging decline of Gf (Horn, Donaldson, Engstrom, 1981). These results are
summarized in Figure 1.1, which depicts the relationships between age, Gf, and
cognitive speed when other cognitive process variables are statistically controlled
The figure illustrates the how the procedure of parting out (or removing the variance
of) process variables accounts for the aging decline of Gf and Gs.
The figure summarizes results showing that the aging decline of G f is partially
accounted for when Memory Span Backward (MSB) is controlled. When the
variance of MSB is removed from Gf, the age related decline associated with the
residual variable (Gf:MSB control) is significantly less than the age related decline of
Gf alone. This result suggests that part of the reason older adults do worse on power
measures of Gf is due to the fact that they (on average) have lower working memory
capacities than do younger adults. This kind of result is seen also when cognitive
speed (Gs), dividing attention (ATD), and encoding organization (EOG) are
controlled. The age related decline of Gf thus can be partially accounted for by
cognitive process variables indicating short-term memory, cognitive speed, and
concentration/attention. To study the age related decline of Gf is to study the
cognitive processes involved in such decline. Evidence for age related decline of
these processes will be discusses next in some detail.
Ability
Level
in
IQ
Units
Figure 1.1
Controlling Various Cognitive Processes to Account for (Decrease) Fluid Ability
(Gf) and Speed ((is) Decline. After Horn, Donaldson, Iingstiom (1981).
Gs: ATD Out
Gs: CON Out
Ke\
Gf=Fluid Ability
Gs=Speed
ATD=Attention
CON=Concentration
HOG=Hncoding Organization
MSB=Memory Span Backward
MSF=Memor>' Span Forward
SAR=Short-tenn Apprehension Retrieval
f: MSF Out
f alone
s alone
'Gf: Gs, ATD or CON Out
Gf: SAR or LOG Out
Gf: MSB Out
Short-term Apprehension Retrieval (SAR). There is considerable evidence
that links short-term memory ability to both concrete and formal reasoning skills in
childhood and adulthood (Liben, 1977; Trabasso, 1977; Case& Globerson, 1974,
Neimark, 1982; Swinton, Sipple, Hooper, & Hougum, 1972). There may be virtually
no aging decline in apprehension,/w .vt?~i e very short-term memory of a few
milliseconds (Botwinick & Storant, 1974; Craik, 1977), but when care is taken to
ensure reliable measurement and sample sizes are large, there is evidence for the
decline of memory over periods of a minute or two (Craik & Lockhart, 1972; Horn,
Donaldson, & Engstrom, 1981; Horn, 1978; Labouvie-Vief & Shell, 1982; Reece,
1977) With measures of forward memory span (MSF), or primary memory, there is
a significant decline with age (see Figure 1.1), but this decline does little to explain
the decline of Gf with age. On the other hand, backward memory span (MSB), which
represents working memory (holding information in memory while mentally
manipulating it), does account for a significant part of the aging decline of Gf. Thus,
G f decline seems to be associated with a loss of working memory capacity
Figure 1.1 illustrates how encoding organization (EOG) and short-term
apprehension retrieval (SAR) each produce about the same effect on Gf decline.
EOG is a measure of grouping. Subjects view objects that can be organized into sets.
In addition to its encoding component, this task is a good measure of how many
things are remembered. EOG accounts for almost all o f the aging decline of Gf that is
also accounted for by SAR. This suggests that most o f the decline of Gf that is
associated with short-term memory is well represented by losses in abilities to
organize information for encoding.
Speediness (Gs). Jensen (1980) has suggested that general intelligence (g) is
primarily indicated by processing speed, as measured in reaction time experiments
He concludes that speed is the essential element of g and a battery to assess g should
be made up of reaction time tests. The assumption is also that g, to a large extent, is
biologically based and inherited Eysenck (1982) has put forth a similar theory
The notion that processing speed is the central, biologically based mechanism
o f intelligence needs to be critically examined. Some measures of processing speed
slow with aging and are related to some aspects of intellect (Salthouse, 1985), but this
does not indicate decreased quality of all cognitive capabilities (Carroll, 1987,
Herzog, 1994; Horn, 1985) Jensen’s results are not clear in indicating this. He
correlates reaction time with mixture measures of g that include both vulnerable and
maintained abilities. This makes it impossible to sort out what exactly accounts for
the correlation. Mixtures that are heavily weighted toward maintained abilities show
little or slightly negative correlations with reaction time. Mixtures that contain many
vulnerable measures show positive correlations
Schaie (1979, 1983) has been criticized for emphasizing speed as much as
power on all cognitive tests The observed declines in these studies indicate, in part,
declines of speed as well as declines in level of functioning (Carroll, 1993). When
care is taken to partial out speed (Gs) in Schaie's data (Schaie, 1989), the aging
curves are found to be similar to those reported by other researchers (Horn,
Donaldson, & Engstrom, 1981, Horn, 1985, Carroll, 1993) who rely strictly on power
measures.
The results of the Horn research group indicate that speed is involved even in
power measures of Gf, but that Gf, and the aging decline of Gf, involves substantially
more than merely cognitive speed (as in the Jenson-Eysenck theory) or cognitive
speed and short-term memory (as in the Salthouse theory). When the variance due to
processing speed (Gs) is partialed out of power measures of Gf, the age effect on
decline is reduced significantly (see Figure 1.1). This suggests that some aspect of
cognitive speed is an important component in determining the level of Gf functioning
even when tests are untimed. The ability to quickly integrate the information of
working memory before that information leaves immediate awareness is essential to
performing well on a Gf tasks But Gs control does not account for the total Gf
decline
Concentration Attention (CON, ATD). Figure 11 shows that Gs speediness,
and the aging decline of such speediness, are linked to abilities in maintaining close
concentration (CON) and attention (ATD). Decline in capacities for ATD and CON
are mainly responsible for both the aging decline of Gs and, therefore, the aging
decline of Gf. The ability to sustain attention while inhibiting potentially distracting
stimuli has been shown to decline with age (McDowd & Fillion, 1992) and account
for some of the aging decline of other vulnerable abilities (Stankov, 1988, Stankov,
Roberts, & Spilsbury, 1994). The ability to hold information in immediate awareness
or short-term apprehension is influenced by attentiveness, and attentiveness is also
important in the speediness of doing simple clerical-perceptual tasks. Concentration
ability also requires attentiveness. Thus attentiveness seems to be the common
process in these otherwise diverse measures of cognitive processes that are related to
G f decline.
Processes that Aid Performance
Controlling for cognitive processes such as short-term memory, attention,
speed, and concentration result in the decrease of the age related Gf decline. This is
only part of the story. There are different processes that, when controlled, result in an
increase of the aging decline of Gf (Horn, Donaldson, Engstrom, 1981; Stankov,
1988) This represents an important feature of performance on ability tasks. Older
adults give fewer incorrect answers than do younger adults when there is opportunity
to abandon a problem if one is not sure about an answer— that is, older adults are
more careful (CAR) to give the correct answer, or indicate that they cannot find an
answer, rather than risk choosing a wrong answer through guessing. Older adults
also work longer before abandoning difficult problems and thus are more persistent
(PER), than younger adults. If assessment does not reward guessing and speediness,
the older person— or any person high on CAR and PER— is not penalized and has an
advantage. If this advantage is taken away, this person will score lower.
115
Figure 1.2
Controlling for Persistence (PER), Carefulness (CAR) and Crystallized Knowledge
(Gc) Aging to Increase the Decline of Fluid (GO Ability. After Horn, Donaldson.
Engstrom (1981).
106 - -
Ability
Level
in 88
Alone
Units
Key:
Gf=Fluid Abilitty
Gc=Crystallized Ability
CAR=Carefulness
PER=Persistence
PER Out
CAR Out
CAR+PER Out
Gc Out
It can be seen in the Figure 1.2 that when CAR and PER are controlled, not
only is there no decrease in the aging decline of Gf, there is actually an increase. Such
results indicate that carefulness and persistence are advantages that accrue to older
adults, on the average, more than younger adults. These advantages enable older
adults to do relatively better at Gf problems than they would if the advantages were
removed. Carefulness and persistence thus can be viewed as compensatory skills
adopted by older adults that enable them to overcome some of the deficiencies
encountered when attempting inductive reasoning tasks. When these tools are not
allowed to operate, older adults do worse at Gf tasks than they would otherwise.
Tests which allow a "no answer" option and do not have time limits are tests that
allow the tools of CAR and PER to be utilized Carefulness and Persistence, then, are
processes that work against, or forestall the aging decline of fluid ability.
This kind of process is also represented by controlling for crystallized
knowledge (Gc). As seen in Figure 1.2, controlling for Gc results in an increase of
the age related decline in Gf which suggests that the older person's Gc "advantage" is
a compensatory tool (like CAR and PER) working against greater decline.
Summary
Thus there are two kinds of process variables that aid in our understanding of
fluid ability decline; those that are part of the ability and those that aid performance.
Older people (on average as compared to younger people) lack important processes
of Gf that, when controlled, remove some of the decline. Older people (on average as
compared to younger people) possess important advantages in performance on some
kinds of Gf ability measures that, when taken away through control, result in
increased decline.
Speed of processing (Gs), concentration (COS) and attention (ATD),
encoding organization (EOG), and short-term apprehension retrieval (SAR) are
processes that, when removed, significantly decrease the aging decline of G f and thus
appear to be part of the fluid reasoning process. Crystallized knowledge (Gc),
carefulness (CAR) and persistence (PER), on the other hand, appear to be
compensatory processes that aid older persons. When these processes are removed,
there is a significant increase in the aging decline of Gf
The main purpose for this study is to provide a foundation for further study of
cognitive development. It is an attempt to replicate the results of previous research,
primarily the findings of Horn, Donaldson, and Engstrom (1981), showing that
various cognitive processes account for age related decline of fluid ability. The extent
to which the maintenance or enhancement of crystallized knowledge can be explained
in terms of cognitive processes will also be explored. This study will provide a
theoretical framework for subsequent studies in which lifestyle variables will be
examined as potential contributors to individual differences in the age related fluid
ability decline and crystallized ability enhancement above and beyond the development
of other cognitive processes Once the relationships between cognitive variables are
understood, the extent to which cognitive variables are related to lifestyle and
personality variables can be explicated with greater clarity.
Method
Subjects
There were two phases of sampling
Phase One Sampling. Subjects were volunteers recruited from the
membership of the Fraternal Order of Elks, residents of a retirement complex , an
organization of volunteers of the Andrus School of Gerontology, participants of a job
training center , and members of a church. Subjects were paid $30.00 for three hours
of taking cognitive tests and filling out questionnaires. In most cases the money was
donated as a charitable contribution of the organization from which the volunteers
were solicited
In all, N=379 persons were administered the battery o f tests and
questionnaires in the first phase of data gathering. Persons in the church group
(N= 13) and the job training program (N=19) served as pilot subjects Instructions
and tests were tried out and in some cases modified in accordance with information
obtained from these subjects. These subjects were not used in further analyses. Also,
it turned out that seven subjects either did not read English or reported having a
stroke or brain damage and thus were dropped in further analyses.
The first phase of sampling thus yielded 340 subjects for analysis O f these,
48.9% (N=166) were male. The age range was 26 to 92 years, with mean =67.9,
17
SD=12 .6 ; and median 70 (i.e., 50% of this sample was over 70 ). The sample was
85% white, 9.9% black, 3 7% Hispanic, and 0.5% Asian. The interquartile range for
income was between $20,000 and $40,000 per year. Eighty percent had graduated
from college.
Phase Two Sampling. Most of the subjects obtained in the first phase of
testing were in a range of between 50 to 90 years of age. The second phase of
sampling was directed at filling out the sample in the range of between 20 and 50
years of age. Subjects with demographic features similar to those sampled in the first
phase were recruited (i.e. mostly white, college educated and in the $20,000-$40,000
per year income bracket) These subjects, also, were recruited through organizations—
the Elks, residents of a retirement complex , staff and members of a YMCA , staff of
city offices, staff and parents of a chapter of the Girl Scouts of America, parents of an
aquatics center swim team, members of a philanthropic organization, and staff and
members of a parks and recreation club. In each case the amounts paid for a subject's
participation ( $ 2 0 0 0 ) was donated to a charity sponsored by the organization
through which the subjects were recruited.
In the second phase of sampling 267 subjects were gathered. Again, there
were subjects for whom English was not a language in which they were proficient,
subjects on heavy medication, and subjects who failed to complete most of the tests
and questionnaires. In all, N=30 subjects were lost for these reasons. The second
phase of sampling thus netted N=237 subjects for further analyses.
Final Sample. The samples from the two phases of data gathering were
combined to provide a total sample of N=577 subjects. The age range in this sample
is 22 to 92 years (mean=58.42, SD=16.30). Females constituted 55.5% of the
sample, 88.2% were White, 5.1% were Hispanic , 3.4% were Black, 2.1.% were
Asian, and 1.2% were of other ethnicity The mean for family income before taxes
was a bracket (supplied in the questionnaire) of $40,000-$49,000. Similarly, the
mean education was for a bracket described as representing "some college "
Approximately 64 .5% of the sample was married, 13 .8 % single, 12 .1% divorced,
2.1% separated, and 13.7% widowed. Forty point two percent of the sample was
retired, 25.7% homemakers, and 2.1% unemployed. All subjects were recruited from
the greater Los Angeles area— N= 170 (29.5% ) recruited through a = fraternal
organization, N=138 (23 9%) were residents of a retirement complex, N 58 (10.2%;)
were staff and members of a YMCA, N=47 (8 .1%) were community volunteers of
local cities, N=44 (7 .6 %;), were staff and parents of a chapter of Girl Scouts of
America, N=32 (5.5%;) were from the Andrus School of Gerontology Volunteers,
N=31(5.3%;) were parents of an aquatics center swim team N=17 (2.9%) were
members of philanthropic organization, N=16 (2 8 %) were staff and members of
several local parks and recreation groups, N=16 (2 .8 %;) were members of the League
o f Women Voters, and N=8(1.4%;) were firefighters and interns.
19
Ability Measures
To reduce subject fatigue due to the length of the testing session, lull common
factors measuring each cognitive ability were not obtained, but particular tests known
to be good, salient markers of abilities were assessed.
Fluid Reasoning (Gf). The Power Letter Series (PLS) task (Horn, 1967,
1985) was administered as the chief marker for Gf. The 38-item test requires subjects
to detect the implicit pattern in a string of letters and choose a letter that completes
(or continues) the pattern. A set of three items at a low level of difficulty is presented
first to provide a warm-up. These items are followed by seven sets each containing
five items from five levels of difficulty arranged in order from “most simple” to “most
difficult.” Instructions are designed to teach subjects that a problem may have no
good solution, in which case they should select a NA (no answer) option and move on
to the next item The NA possibility thus enables subjects to give up on a difficult
problem even when the problem has an answer. This, plus the fact that the subjects
must produce an answer, not select an answer from among several choices, decreases
the likelihood that subjects will guess when they have doubts. Also, learning that the
problems may have no acceptable answer, plus the NA option, discourages
perseveration on particular items at the expense of attempting other items. Subjects
were given twenty minutes to complete this task.
Although the PLS test is timed, the measure obtained with it is not: the
measure is the level of difficulty successfully resolved, not the number of problems
2 0
solved within the time limit. The score is the sum of the difficulty weights for items
correct divided by the number of items attempted within each level of difficulty.
Formally, the Li score within any of the seven (i=l . .7) sets of five items is
'wc
AveL, = 2
j=i
N.
where W is a weight for difficulty level (Wi=l, W2=2 , Wj=3, W4=4, W 5=5),
C=correct answer (l=correct, 0=incorrect), N=number of items attempted at
difficulty level j. The L, scores o f each set are summed and divided by the number of
sets in which at least one item was attempted. Scores can range form zero (incorrect
answers at all levels of difficulty) to 15 (correct answers for items at all levels;
1+2+3+4+5).
Common Word Analogies (CWA; Horn, 1985) is a verbal test marker for Gf
It requires subjects to figure out the relationship between two common words and
choose a word from a list of common words that represents an analogous relationship
for a third word. For example:
SOON is to NEVER as NEAR is to: (choose one word from the following
list)
NOWHERE FAR AWAY DISTANT SOMEWHERE
Given that the meaning of the words in each item is known to the respondent,
the task is one of reasoning to determine the relationship among the meanings. That
is, the principal difficulty of each item was determining relationships among concepts.
Subjects were given five minutes to complete the 15-item test. Time was not cut off
until all subjects had attempted all items. Subjects were encouraged to choose “no
answer” if there appeared to be no analogous relationship. Possible scores range
from zero (none correct) to 15 (all correct).
Crystallized Knowledge (Gc). Two 15-item Vocabulary tests (VOC1 and
VOC2) were given to indicate Gc Subjects were given time to attempt all items.
Typically this was accomplished within five minutes. Possible scores range from zero
to 15.
The Esoteric Analogies (ESA) test (Horn, 1985) was also used to indicate
Gc This test is the same as common word analogies, but the words are esoteric.
Understanding the relationship between the words requires the rather advanced
knowledge that comes with acculturation. For example:
GUSTATORY is to TASTE as OLFACTORY is to:
SMELL TOUCH FEEL HEAR BALANCE
requires that one has acquired meaning for the words gustatory and olfactory. In
each Gc task, subjects were encouraged to utilize the “no answer” option rather than
skipping items or guessing
Speed of Processing or Cognitive Speed (Gs). In the Speeded Cross-out
(SCO) task the subject was required to quickly cross out all the numbers in a row that
22
are identical to the first number in that row. The following is an example of a row of
the test:
6 1 1 9 6 9 0 4 4 6 2 6 4 5 7 4 7 7 7 4 5 192433729
Scores indicate the number of correct cross-outs in a 45-second trial. Possible scores
range from zero (no correct cross-outs) to 73.
The Speeded Letters Comparison (SLC) task requires subjects indicate
whether sets of letters are the same or different. The following are examples of such
sets:
Same Different
almpq almpp _____ _____
tfstd tfstd _____ _____
Such comparison sets were arranged in columns. The subjects moved down the
columns making same/different judgments as quickly as possible. Scores indicate the
number of correct comparisons in a 30-second trial. Possible scores range from zero
(no correct comparisons) to 25 (number of comparisons on the page).
Concentration (CON). The Slow Tracing task (Horn, 1985) was utilized to
assess concentration (CON). Here, the subject is asked to trace an irregular line as
slowly as possible. The administrator very deliberately watches a stopwatch tick
down long after all subjects have completed tracing the line. With emphasis, the
subjects are again asked to trace a line (new item) as slowly as possible, keeping their
pencil moving at all times. Again, the administrator does not call time until a minute
23
after all subjects have completed tracing the (rather short) line and again emphasized
that the task is to trace as slowly as possible. On subsequent trials subjects try vary
hard to trace very slowly. Two further trials were given to obtain the measure.
Scores indicate the shortness of the line traced. That is, the traced line was measures
in centimeters and subtracted from a constant Thus, the smaller the number of
centimeters, the larger the score. The higher the score the more the ability to
concentrate A separate score was given for the quality of the tracing.
Short-term Apprehension Retrieval (SAR). Memory fo r Paired Associates
(MPA) (Horn 1985) was used to indicate SAR. Here, subjects were given three
minutes to study the 21 pictures paired with numbers. They were then given a page
containing only the pictures and were given two minutes to recall the numbers that
were paired with the pictures One point was given for each correct pairing. Possible
scores range from zero (no correct pairing) to 2 1 (all possible correct pairing)
Table 1 1 provides a summary of psychometric information on the ability
measures The reliability for CON was obtained as the internal consistency (alpha
reliability) for trials 3 and 4 of the slow tracing task. The or the reliability estimates
were obtained as the multiple correlation of a measure with all other ability measures.
This, an estimate of the communality of the variable, is a lower bound estimate of the
reliability (Guttman, 1957; Kaiser & Caffrey, 1965; Tyron, 1957).
T able 1.1. Means, Standard deviations and Ranges fo r Cognitive Variables and Age. Breakdown for Gender also shown
(N=577)
Ability Variable Name Min Max Mean SD
Males
Mean(SD)
N=257
Females
Mean(SD)
N=320
SMR Reliability
Estimate
Age in years 22.08 91.75 58.42 16.30 57.63(15.95) 59.06(16.56)
— —
G f Power Letter Series 0 15 4.87 2.42 485(2.43) 4.88(243) 51 .71
Common Analogies 0 14 8.50 2 61 8 53(2.66) 8.48(2 57) 58 76
Gc Esoteric Analogies 1 15 961 2 93 9.43(2.99) 9.76(2.88) 59 .77
Vocabulary 1 2 15 11.62 2 . 1 1 11.45(2.02) 11.77(2.16) 59 .77
Vocabulary 2 0 15 8 58 2.97 8.39(291) 8.75(3.00) .59 .77
SAR Memory for Paired
Associates
0 2 1 9 1 1 5.07 9.24(5 32) 9.03(4.90) .59 .77
CON Slow Tracing (cm) .75 36.75 15 39 11 92 13 .98(12 24) 16 35(11 63) 46 96
Gs: Speeded Cross-out 3 47 2 2 . 6 6 6.16 22.56(5.72) 13.37(4.35) .49 .70
Speeded Letter
Comparison
1 25 13.54 4.31 13.56(4.29) 22.74(6.45) .51 .71
Gf=Fluid Reasoning
Gc=Crystallized Knowledge
SAR=Short-term Apprehension Retrieval
CON=Concentration
Gs=General Speediness
25
Procedures
Subjects were tested in groups ranging in size from four to 32 They were
first made aware that their responses to the tests would remain anonymous. They
were told to not put their name anywhere on the testing materials. Instead, they were
asked to make up a six-digit number and record this number as their “name” on the
testing materials and on the envelope into which they placed their completed tests.
These procedures were followed to enable subjects to freely express themselves
without feeling that their responses would be known to the investigators and
associated with them personally. The ability tests were administered in the first 90-
minutes of testing. First, subjects filled out a demographic information form and
completed several of the group-administered ability tests. A ten-minute break was
given after 50 minutes of testing. Another period of 30 minutes of ability testing
followed. Subjects then completed questionnaires for another 20 minutes before
being given another break with refreshments. In a final session of 50 minutes,
subjects completed questionnaires Subjects were encouraged to stay after the testing
session for discussion and debriefing with the investigators
Analyses
The procedures for analyses were described in Horn, Donaldson, and
Engstrom (1981). These consist primarily of part correlational (or semi-partial)
analyses. The extent to which a process variable can predict (linearly) individual
differences in an ability say, Gf is removed from that ability. That residual is then
26
correlated with age. The difference between this correlation for the residual and the
unresidualized ability correlation with age is an indication of the extent to which the
process variable is involved in (accounts for) the ability correlation with age. For
example, consider the hypothesis that short-term apprehension memory (SAR)
accounts for some of the age related decline of Gf. Let Age=A, Gf=G, and SAR=S.
The extent to which S estimates G (linearly) is obtained as:
G = rc s *S [1.1]
where rG s is the correlation between G and S. The residual component is calculated:
D = G - G [1.2]
This residual’s correlation with age, t a d , represents the extent to which Gf is related
to age after what Gf shares with SAR has been parted out (or removed) from Gf
The resulting relationship is the correlation of age with the part of Gf that is not
explained by SAR. The difference between rG S and rD s indicates how much of the rG s
relationship is accounted for by S. If the T a d relationship is not zero, then it can be
concluded that there is a part of Gf (beyond the part shared with SAR) that declines
with age and remains to be explained.
Part correlation is used in these analyses in preference to partial correlation.
In partial analysis the component S is removed from A ( E = A - A, where
A = rA S * S ) as well as from G and the correlation, rE D , is between the two residuals
( A - A ) and ( G - G ) But residualization in this study of development simulates
control. It is logically sensible to control for one behavioral variable in another
27
behavioral variable— e.g., SAR in Gf— but it is dubious to control for a behavioral
variable in age, which represents the conditions through which the behavioral variable
has developed. Age indicates conditions o f cause, not behavior as such. If partial
correlation were utilized to simulate control, the behavioral process variable SAR
would be removed from both the behavioral variable, Gf, and the developmental
variable, age, thus eradicating developmental inferences.
The part correlation can be tested for significance using the programs of part
correlation in which age is the dependent variable. This can be seen algebraically.
First, estimate A of the above example as a linear combination of G and S.
Assuming that all variables are in standard score form, the b-weights in this multiple
regression are
where in general rj< k .p ) represents the part correlation between variable j and the
residual of k when p has been removed, and Vkp represents the variance of this
residual. The test of a b-weight for significance is the test of the part correlation for
significance. For example, the test of bi for significance is the test that r^c.s) is
A = b,G + b; S [1.3]
[1.4]
and
_ r AS f At; r o s
1 — r'c.s
[15]
28
significant. But the part correlation between age and the residual G - G when
G = rG S * S , as in equation 1 . 2 and 1 . 2 respectively, is:
_ *AG ~ f A S f C.S _ _ r i /r-l
A (G S ) — I ;--- A (G S ) L, 0J
V I - r ' G S
This is precisely the correlation that is tested for significance when bi in the multiple
correlation of G and S with A is tested for significance. Thus, multiple regression
analyses can be used to accomplish part correlation controls To provide an
indication that the control variable has a significant (different from zero) effect, the
significance of the difference between the multiple correlation before control and after
control can be calculated The R-square change formula (Cohen & Cohen, 1983)
assesses the extent to which the addition of each subsequent component adds
significantly to the multiple correlation
For purposes of communicating results, correlations and part correlations
were converted to regression coefficients expressed in IQ units in which the mean was
set to 100 and the standard deviation was set to 15 In this sample, the zero-order
correlation of age with Power Letter Series is -.469 and the mean and standard
deviation for age are 58.42 and 16.30 respectively (Table 11). Thus, again using the
abbreviations A=Age, G=Gf, the regression of Gf (estimated by Power Letter Series)
and age in IQ units can be expressed as:
The slope, -.431, indicates the amount of change of G for a unit of (year of) age. Ten
years of age would thus yield 4 .31 IQ units of decline in G. To put it more simply,
the zero-order correlation between age and G f (-.469) is multiplied by the IQ unit
standard deviation of Letter Series (- 469* 15=-7.05). This is the Gf decline, in IQ
units, for one standard unit of age (16 30 years) or roughly 4.31 IQ units Gf decline
per decade. IQ units are familiar to most psychologists and conveniently illustrate
how change (in IQ units) is affected by the control of process variables. In previous
studies this decline has been between 3.30 and 4.80 IQ units.
Results
Means, ranges, and standard deviations for the total sample and separately for
males and females are given for age and all cognitive variables in Table 1.1 The
gender differences are not significant (at the .05 level) for on any of the cognitive
measures. As indicated earlier, Table 1.1 also includes SMR’s of each ability with all
the others and estimates of the reliabilities. In Appendix A means and standard
deviations are provided for all cognitive measures across several age groups. The
averages for measures of vulnerable abilities (Power Letter Series, Common Word
Analogies, Memory for Paired Associates, Slow Tracing, Speeded Cross-out, and
Speeded Letter Comparison) decrease with age. There is, in general, an increasing
monotonic trend for the maintained abilities (Vocabulary trials 1 and 2, and Esoteric
Analogies). That is, the means drop slightly for the 56-65 age group, but increase
from age 6 6 on The trend is thus generally monotonic. A linear function fits the data
for different Gc indicators: quadratic and cubic functions do not describe the curve.
Table 1.2 contains zero-order Pearson product-moment correlation coefficients for
the cognitive variables, gender, and age
Table 12. Zero-order Correlations Between Age, Gender, and Cognitive Ability Variables (N 577)
1 Age 2. Gender
l=m, 2 =f
SEX
3. Power
Letter
Series
PLS
4. Common
Analogies
CWA
5 Esoteric
Analogies
ESA
6 .Vocab
1
VOC1
7.Vocab
2
VOC2
8 . Memory
Paired
Associates
MPA
9. Slow
Tracing
CON
1 AGE 1.000
2 SEX .043 1.000
3 PLS -469** 0 0 1 1.000
4.CWA -.231** . 0 0 2 601** 1.000
5.ESA .071 .026 .341** .511** 1 0 0 0
6.VOC1 .251** 072 .2 2 1 ** .410** .641** 1.000
7.VOC2. .156** 061 .288** .430** 661** .720** 1 0 0 0
8 .MPA. -.469** .0 0 1
4 4 9 **
.331** .230** .072 141** 1.000
9.CON -.497** 082 .405** .262** .091 -.071 .030 .379** 1.000
IOSCO -.531** . 0 1 0 291** .181** -.081 -.1 2 2 * .060 .271** 310**
11 SLC -.629** . 0 2 0 .468** .300** .130* -.061 .068 .406** .360**
10 Speeded 11.Speeded
Cross-out Letter
SCO Comparison
________________SLC
10 SCO 1.000
11.SLC .561 1 . 0 0 0
* = Significant at p<05
** = Significant at p< 0 1
32
Psychometric Considerations
Full common factors were not assessed for each cognitive ability. Instead,
reliable markers were assessed for different manifestations of abilities Two measures
of fluid ability were included in the battery; Power Letter Series (PLS) and Common
Word Analogies (CWA). The CWA test indicates Gf, to be sure, but it also contains
a substantial crystallized knowledge (Gc) component (Horn, 1985; Horn & Cattell,
1967). Indeed, in the present sample the correlation between common word
analogies and Gc markers is .41 and .45 for VOC1 and VOC2 and 51 with esoteric
analogies. Common factor estimates of Gf and Gc typically correlate about .30
(Horn & Noll, 1994). A linear composite comprised of CWA and PLS would thus
represent an estimate of Gf that was tipped toward indicating Gc as well The two
measures were, therefore, not combined into one common factor, but instead were
studied as two separate indicators of fluid reasoning When the CWA test is used in
subsequent analyses, partialling is used to help purify it by removing variance
associated with Gc, thus making the measure more nearly an indicator o f fluid ability.
Esoteric Analogies (ESA) was designed to measure Gc, but it involves an
element of Gf. Though acculturational knowledge of vocabulary and knowledge of
the association between concepts is called for in solving ESA problems, one weak in
these abilities but strong in reasoning with only limited information— i.e., Gf--can also
discern the right answer The zero-order correlation between ESA and Power Letter
series is .34 (see Table 1.2). To combine ESA with other measures of Gc (e.g.,
33
vocabulary tests) would be to “contaminate” the composite with an element of Gf. It
is therefore desirable to analyze this and the other indicators of Gc separately and to
partial out variance associated with Gf indicators in estimating effects related to Gc.
Of the two Vocabulary trials (VOC1 and VOC2 ), the first was much simpler
than the second. The variance for VOCl is truncated. There is a ceiling effect
reflecting the fact that many subjects give correct responses to all items. The measure
is thus skewed The distribution for V0C2, on the other hand, is nearly symmetrical,
is not skewed, and has substantial variance VOC2 is probably the better indicator of
Gc, but analyses with VOC 1 rnay provide indications of how ceiling effects and
skewness affect results.
The scores on the two Slow Tracing trials were combined into a composite
measure of concentration (CON) for further analyses. This composite has an alpha
internal consistency reliability o f 96
Speeded Letter Comparison (SLC) and Speeded Cross-out (SCO) provide
somewhat different estimates o f cognitive speed. SLC requires attentiveness, very
simple reasoning, and quick decision action. The SCO task involves simple search
and motor reaction time While cognitive speed (as measured by SLC) relates to the
aging decline of Gf, such a relationship for motor reaction time (as measured by SCO)
appears to be very weak (Horn et al, 1981). SCO thus can be expected to be more
lowly related to Gf decline than SLC.
34
Partitioning the Age Related Decline o f Fluid Ability
Analyses were directed at describing the extent to which cognitive process
variables that are indicative of Gf-short-term apprehension retrieval, concentration,
and cognitive speed— are related to aging decline of fluid ability. To further describe
the age related decline of Gf the effects of parting indicators o f crystallized
knowledge from fluid ability indicators was examined.
Power Letter Series (PLS) as an Indicator o f Fluid Ability. Summarized in
Table 1.3 are correlations of age with fluid ability as estimated by PLS under
conditions of control for processes hypothesized to affect Gf. In each case, the
variance of the process variable is removed from the fluid ability indicator and the
resulting residual is then correlated with age. Let us focus, first, on the first column
of numbers in the Table. In this column are summarized the age correlations with
PLS alone and with various residualized PLS scores. Here it can be seen that the
correlation of age with PLS alone is -.469, but when MPA is parted form PLS, the
correlation with age and the residual PLS is -.299. This reduction is significant at the
05 level This result suggests that the aging decline of PLS can be partly accounted
for, or explained, by the process assessed in MPA. Similarly, the Table shows the
effects on the PLS age correlation when CON, SLC, and various combinations of
these three processes indicators are removed. The correlation between age and PLS
when CON alone is removed is -.291. It is -.210 when SLC is removed. The results
35
in column 1 of Table 1.3 thus suggest that MPA, CON, and SLC, each considered
separately, represent a process of fluid reasoning that declines with age in adulthood
When combinations of these three process variables are controlled, the aging
effect is further reduced. When MPA+CON is removed, the resulting age correlation
is - 201 which is significantly smaller than the .299 or .291 resulting when MPA or
CON alone is removed. Thus, MPA and CON represent somewhat separate
processes of Gf that decline with age When MPA+SCL is controlled the age
correlation goes to -. 142. CON+SLC control results in a - 131 age correlation
When all three process variables are controlled (MPA+CON+SLC) the age
correlation is reduced to -.090. Each of these correlational shifts is significant relative
to the previous model in which the shift is nested That is, the R2’s for the models in
which a process variable is added (e.g., CON added to MPA) are significantly larger
than R2’s for simpler models (MPA or CON alone). Though the variance of these
processes overlap somewhat each contributes significantly, alone and in company with
the other two, to the aging decline of fluid ability.
Figure 1.3a depicts the results of Table 1.3 in graphic form using IQ unit
transformations of the age slopes. A slope o f- .469 is equivalent to a 4 .31 IQ unit per
decade decline in PLS alone When MPA is controlled, the per decade decline drops
to 2 .75 IQ units. Thus, a 1.56 IQ unit decline (4.31-2.75=1.56) per decade in Gf (as
estimated by PLS) is associated with a process measured in MPA. Similarly, when
SLC alone or CON alone is parted out the decline of Gf is significantly reduced.
36
Table 1.3. Effects o f Parting Cognitive Processes on the Correlations Between Age
and Indicators o f G f Based on Power Letter Series (PLS)
PLS PLS: PLS: PLS: PLS: PLS
Alone VOC2 ESA VOC2+ VOC2+ VOC2+
Out Out VOC1 ESA VOC1 +
Out Out ESA
Out
1 2 3 4 5 6
Age Correlation
BASE MODELS: -.469 -.531* -.524* -.540* -.539 -.538
REDUCED
MODELS:
MPA Out -.299* -.365’ -.372* -379’ -.388’ -.377’
CON Out -.291* -.358* -.356* -.337’ -.371’ -.361*
SLC Out - 2 1 0 * -.268* -.274* -.280* -.289* -.281’
MPA+ CON Out -.2 0 1 * -.273* -.282* -.286* -.279* -.287*
MPA+ SLC Out - 142* -.206’ -.2 2 0 * -.2 2 1 * -.235* -.227*
CON+ SLC Out -.131* -.190’ - 199* -.2 0 2 * -213* -.206’
MPA+ CON+SLC -090* -.155* -.171* -.169* -.185’ - 178*
Out
* = Significant R2 change at the .05 level from the R2 of the just previous model in
which the effect is nested. Example: The R2 for PLS:VOC2+VOCl is significantly
greater than the R2 for the model PLS:VOC2 at p<05.
* = Significant R2 change from the just previous model in which the effect is nested at
p<05. Example: the R2 for the model in which that includes MPA is parted is
significantly greater than the R2 for the models in which MPA+CON,
MPA+SLC, MPA+ CON+SLC are parted
CON=Concentration; CWA=Common Word Analogies; ESA=Esoteric Word
Analogies; MPA=Memory for Paired Associates; PLS=Power Letter Series;
SLC=Speeded Letter Comparison, V0C1= Vocabulary Test 1; VOC2=Vocabulary
Test 2
Figure 1.3a
The L fleet of Removing the Variance of Memory for Paired Associates (Ml’A), Concentration
(CON), and Speeded Letter Comparison (SLC) from the Age Related Decline of Fluid Ability
as Measured by Power Letter Series (PLS).
116
IQ nt/decade decline
110
0.83
'LS.(MPA+CON+SLC Out)
104
>LS:(SLC Out) 1.93
Ability
Level
in 98
>LS:(CON Out)
>LS:(MPA Out)
Units 2.68
2.75
Legend:
PLS=Po\ver Letter Series
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
’ LS alone
92 32 72 82 22 42
38
Thus, one of the reasons why fluid ability declines with age is because MPA
(or CON or SLC) is also declining. But MPA (or CON or SLC) alone does not
explain all this G f decline. When CON and SLC, as well as MPA, are parted out the
decline of Gf is further reduced to .83 units per decade. Nearly three and a half IQ
points per decade of fluid ability decline is due to the fact that short-term
apprehension retrieval, concentration, and cognitive speed are also declining (4.31-
0.83=3.48). When these three processes of fluid ability are controlled, older subjects
look like younger subjects Less than one IQ point per decade fluid ability loss would
be observed if aging did not also involve the decline of these cognitive processes of
fluid ability.
Horn et al. (1981) report that controlling for a maintained ability (Gc)
produced an increase in the aging decline of fluid ability. This effect is demonstrated
in columns 2 through 6 in Table 1.3. The second column of numbers shows that
when the maintained ability VOC2, an indicator of Gc, is controlled in PLS, the
absolute valued o f the age correlation for the residual PLS increases to -.531. The
aging decline for this residual variable PLS VOC2 is significantly larger than the aging
decline of PLS alone at the .05 level When the Gc indicator is removed from Gf, the
resulting measure is more nearly “true” fluid ability. The aging decline for this
estimate of Gf is larger than for the less pure measure.
This effect is depicted in Figure 1 3b. The figure indicates that the aging
decline for “purified” fluid ability is 4.88 IQ units per decade (as compared to 4.31
39
units if VOC2 is not controlled). Removing the processes associated with MPA,
CON and SLC in this purified measure also reduces the aging decline, as in the case
for the non-purified measure. In each case the nested reductions are statistically
significant, as for the non-purified measure. A notable differences is that after full
control of all three processes of Gf (MPA, CON, and SLC), there is still notable aging
decline for the purified measure-1 .43 IQ units per decade (corresponding to an age
correlation of -. 155 in Table 1 3). Controlling for all three process variables accounts
for roughly the same magnitude of Gf decline for the purified measure (4 8 8 -
1.43=3.45) as for the non-purified measure (4.31-0.83=3.48).
The third column of Table 1 3 indicates that when ESA is used as an indicator
of Gc to purify PLS, the purification o f PLS via ESA control is not as effective as
when VOC2 is used to purify PLS— age correlates -.531 with PLS:VOC2 and - 524
with PLS:ESA. As can be seen in Table 1.2, this is because ESA correlates higher
(.341) with PLS than does VOC2 (.288). These correlations indicates that ESA
measures Gc only slightly better than it measures Gf and to control for ESA in PLS
does not effect the purest of Gf measurement
The fourth column of Table 1.3 summarizes results for aging decline when
VOC1 as well as VOC2 are removed from PLS. These results indicate that the
addition of VOC1 as a second indicator o f crystallized knowledge affects the aging
decline of fluid ability. VOC1 accounts for significant, non-redundant fluid ability
decline. This suggests that there is something about the measurement achieved with
I
116
110
104
Ability
Level
in 98
IQ
Units
92
86
80
Figure 1.3b
The Iiffect of Removing the Variance of Memory' for Faired Associates (.MPA), Concentration
(CON), and Speeded Letter Comparison (SLC) from the Age Related Decline of Power Letter
Series with Vocabulary Test 2 Controlled (PLS:VOC2)
Legend:
PLS:VOC2=Power Letter Series with
Vocabulary 2 Controlled
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
IQ nt/decade decline
’LS.VOC2 (MPA+CON+SLC Out) 1.43
LS:VOC2 (SLC Out)
’ LS:VOC2 (CON Out)
’ LS:VOC2 (MPA Out)
2.46
3.29
3.36
’ LS:VOC2 alone
4.88
22 32 42
52 AGP, 62 ■ fe.
o
this simple, skewed estimator of Gc that overlaps with the measurement of PLS over
and above the measurement overlap of VOC2 and PLS. When these two Gc
indicators are controlled in the PLS indicator of Gf, the age correlation for the doubly
purified measure is -.540, which corresponds to 4.97 IQ units of decline per decade
Controlling for the three process variables (MPA, CON, and SLC) in the
doubly purified measure, the age correlation is 169 corresponding to 1.55 IQ units
of decline per decade The magnitude of fluid ability decline that can be accounted
for when VOC2 and VOCl have been removed (4.97-1.55=3 42) is roughly
equivalent to the magnitudes when PLS is analyzed alone, and when only VOC2 is
removed from PLS.
The sixth column in Table 1.3 shows that the additional removal of ESA from
PLS (after VOC2 and VOC1 are removed) does little to change the age correlations
over and above what VOC2 and VOCl can do. This result thus again indicates that
ESA correlates near zero with age and measures Gf as well as Gc.
Common Word Analogies as an Indicator o f Fluid Ability. In Table 1.4 the
age correlations for the Common Word Analogies (CWA) indicator of Gf are
summarized. As for the PLS indicator, the first column contains age correlations for
CWA alone and columns 2 through 6 summarize results for analyses in which the Gf
estimate is “purified” by controlling for indicators of Gc.
42
Table 1.4. Effects o f Parting Cognitive Processes on the Correlations Between Age
and Indicators o f G f Based on Common Word Analogies (CWA)
CWA CWA: CWA: CWA: CWA: CWA
Alone VOC2 ESA VOC2+ VOC2+ VOC2 +
Out Out VOCl ESA VOC1+
Out Out ESA
Out
1 2 3 4 5 6
Age Correlation
BASE MODELS: -.231 -.333* -.314* -.371* -.340 -.353
REDUCED
MODELS:
MPA Out - 080* -.2 0 1 ’ - 199* -.235* -.224* -.235’
CON Out - 155* -.214’ -.206* -.245* -.229* -.239*
SLC Out
l
o
u >
« ♦
-.136* - 135* -.164* -.157* -.166*
MPA+CON Out -.033* - 145* - 149* -.175* -.171* - 180*
MPA+SLC Out .025* -.091* -098* -.116* -.1 2 0 ’ -.128’
CON+SLC Out 013* -.089* -.094* - 113* -.114* -.1 2 2 *
MPA+CON+SLC .047’ -064* -074’ -.087* -095* -.1 0 0 *
Out
* = Significant R2 change at the 05 level from the R2 of the just previous model in
which the effect is nested Example: The R2 for PLSVOC2+VOC1 is significantly
greater than the R2 for the model PLS:VOC2 at p<05.
* = Significant R2 change from the just previous model in which the effect is nested at
p<05 Example: the R2 for the model in which that includes MPA is parted is
significantly greater than the R2 for the models in which MPA+CON,
MPA+SLC, MPA+ CON+SLC are parted.
CON=Concentration; CWA=Common Word Analogies; ESA=Esoteric Word
Analogies; MPA=Memory for Paired Associates, PLS=Power Letter Series,
SLC=Speeded Letter Comparison, VOCl=Vocabulary Test 1; VOC2=Vocabulary
Test 2
The correlation o f age with CWA alone is -.231, corresponding to 2.12 IQ
units o f decline per decade (see Figure 1,4a). When process variables MPA, CON,
and SLC are removed from CWA alone, the aging decline practically disappears
Indeed, the relationship is slightly positive ( 047) when all three processes are
removed, suggesting that all that is left in CWA is a small crystallized knowledge
component after memory, concentration, and cognitive speed have been removed.
Purifying the CWA estimate of Gf by controlling for VOC2 increases (absolute value)
the decline correlation to -.333, controlling for both VOC2 and VOCl increases the
correlation to -.371 When the best indicator of crystallized knowledge, VOC2, is
removed from CWA, and when both VOC2 and VOCl are removed, the resulting age
correlations are highly similar to those when Power Letter Series alone is the fluid
ability indicator
Figure 1,4b illustrates that when VOC2 and VOCl are removed from CWA
the decline is 3 .41 IQ units per decade (see Figure 1 4b). The effects of controlling
for the MPA, CON, and SLC processes of Gf in this purified CWA estimate of Gf are
smaller in magnitude but the same in principle as the effects for such controls for the
PLS estimates of Gf. This is seen most clearly for the twice purified measure
obtained when both VOC2 and VOC1 are partialled The decline o f-.371 is reduced
to between - .164 and - 245 by removal of one of the processes (SLC, CON, or
MPA). Removal of one additional process further reduces the age and residual Gf
correlations to between -.113 and -. 175; removal of all three process variables
Figure 1.4a
The Effect of Removing the Variance of Memory for Paired Associates (MPA), Concentration
(CON), and Speeded Tetter Comparison (SLC) from the Age Related Decline of Common
Word Analogies (CWA) as an Indicator of Fluid Ability.
120
IQ nt/decade change
JWA. (MPA+CON+SI.C Out) 0 4y
115
CW A: (SIC Out)
CW A: (CON Out)
-0.29
-0.74
110
.’WA: (MPA Out)
- 2.12 JWA alone
Units
Legend:
CWA=Common Word Analogies
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
92 72 82
52 AGE 62
22 32 42
Figure 1 4b
The KiTcct of Removing the Variance of Memory for Paired Associates (MPA), Concentration
(CON), and Speeded Letter Comparison (SLC) from the Age Related Decline of Common
Word Analogies with Vocabulary 1 and 2 Controlled (CWA: VOC2+VOCI).
116
IQ nt/decade decline
CWA: VOC2+VOCI
(MPA+CON+SLC Out)
0.79
110
104
1.51 ,'WA:VOC2+VOCT (SLC Out)
:WA:VOC2 +VOC1 (MPA Out)
2WA:VOC2+VOCl (CON Out)
2.16
Ability
Level
in 98
Units
Legend:
CWA:VOC2+VOC1 =Common Word
Analogies with Vocabulary' 1 and 2
Controlled
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
5.41
:WA:VOC2+VOCl alone
92 22 52 42 72 82
46
reduces the correlation to -.087 The control of process variables in this purified
CWA measure of Gf results in a 2.62 IQ unit per decade reduction (3.41-.79=2.62).
This is slightly less (absolute value) than was seen when the purified PLS measures
were studied.
The removal of Esoteric Analogies (ESA) as a third Gc indicator does little to
account for additional Gc variance in CWA above and beyond that which VOCl and
VOC2 can produce
Overall, the PLS measure is a purer indicator of Gf than is CWA: it contains
less measurement overlap with indicators of crystallized knowledge. When Gc is
controlled in CWA the declines with age are smaller than those for PLS. This may
reflect lower reliability for the Gf component of CWA than for the comparable
component of PLS. It may as well reflect differences in the validity of measures the
two Gf estimates. CWA is a measure of Gc as well as Gf and it shows less decline
begin with Regardless of which indicator is used, and which indicators of Gc are
parted out, some decline remains even after the processes involved in fluid ability
have been are taken into account. A portion of fluid ability decline with age remains
to be accounted for Perhaps control by other variables can help us to understand this
remaining portion of fluid reasoning decline
The Age Related Enhancement o f Crystallized Knowledge
The different estimates Gc (Esoteric Analogies— ESA, and two Vocabulary
trials— VOC1 and VOC2) have different measurement properties and probably tap
47
different aspects o f Gc. These measures, too, are not pure indicators of the Gc
construct. They may involve different amounts of Gf and other contaminants. To
study such effects different Gf indicators (Power Letter Series— PLS and Common
Word Analogies— CWA) were removed from the different indicators of Gc to
establish more nearly “pure” measures of crystallized ability.
Vocabulary 2 as a Crystallized Ability Indicator. Table 1.5 summarizes
correlations of age with VOC2 alone and with VOC2 when various different
indicators of Gf contamination are removed to purify this Gc estimate. For VOC2
alone, the age correlation is 156. This corresponds to a 1.43 IQ unit per decade
increase in Gc over the age range from 22 to 92 years (see Figure 1 5a).
The Gf contaminants of Gc can be seen to be of two types, each represented
by parting out results in Table 1.5. The first kind of contaminant is a Gf process, such
as MPA, CON, or SLC. The effects of removing these contaminants are summarized
in the rows of Table 1.5. The second kind of contaminant is that of measures of Gf,
per se, as with PLS or CWA. The effects of controlling for these contaminants are
summarized in columns 2 and 3 of Table 1.5
Considering, first, the G f process contaminants, it can be seen in examining
the rows of Table 15 that of the three Gf processes, MPA affects the age-Gf
correlation most. Control of MPA alone on VOC2 alone increases the age
correlation to 2.03—1.87 IQ units enhancement per decade (see Figure 1.5a).
Removing both MPA and SLC increases the correlation with age to .220 with age
48
Table 1.5. Effects o f Parting Cognitive Processes on the Correlations Between Age
Indicators o f Gc Based on the Second Vocabulary (V0C2) Test
VOC2 VOC2: VOC2:
Alone PLS PLS+CWA
Out Out
1 2 3
Age Correlations
BASE MODELS: .156 .281* .291
REDUCED
MODELS:
MPA Out .203* .302* .296*
CON Out .171* .251° .256°
SLC Out 198* .255° .245°
MPA+CON Out .2 0 1 .269° .261°
MPA+SLC Out .2 2 0 * .268° .250°
CON+SLC Out . 2 0 0 .263 .223
MPA+CON+SLC .213
O
0 0
< N
.230°
Out
* = Significant R2 change at the .05 level from the R2 of the just previous model in
which the effect is nested. Example: The R2 for VOC2:PLS is significantly greater
than the R2 for the model VOC2 alone at p<05.
' = Significant correlation increase from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which MPA is parted is
significantly greater than the R2 for the model where VOC2 is alone.
° = Significant correlation decrease from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which CON is parted is
significantly greater than the R2 for the model where VOC2 is alone.
CON=Concentration; CWA=Common Word Analogies; ESA=Esoteric Word
Analogies, MPA=Memory for Paired Associates, PLS=Power Letter Series;
SLC=Speeded Letter Comparison; V0C1= Vocabulary Test 1; VOC2=Vocabulary
Test 2
I
I
Figure 1.5a
The effect of Removing the Variance of Memory for Paired Associates (Ml’A), Concentration
(CON), and Speeded Fetter Comparison (SFC) from the Age Related enhancement of
Vocabulary Test 2 (VOC2) as a Crystallized Ability Indicator.
IQ pt/decade enhancement
115 - i -
,VOC2. (MPA+SLC Out)
■VOC2: (MPA Out)
-VOC2: (SLC Out)
11 0 - 2.02
1.87
1.82
.VOC2: (CON Out) 1.57
105
'OC2 alone
1.43
Ability-
Level
in 100
Units
Legend:
VOC2=Vocabulary Test 2
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
92 82 42 72 22
50
(2.02 IQ units per decade). CON control does little to affect aging enhancement
above and beyond the effect produced by MPA and SLC. These results indicate that
if the Gf processes represented by MPA and SLC do not operate in the measure of
Gc, older subjects who already score higher than younger subjects on the VOC2
measures of crystallized knowledge, will score even higher. In part, the Gc advantage
that older people have over younger people is “masked” by cognitive speediness and
short-term memory that operate in the measurement of Gc with the fallible VOC2
test.
When the contamination associated with a direct measure of Gf (PLS) is
removed from VOC2, the resulting residual correlates .281 with age. This
corresponds to a 2.58 IQ unit increase per decade (see Figure 1 5b). This is an
addition of 1.15 IQ units over the enhancement seen with VOC2 alone (2.58-
1.43=1.15) Looking down column 2 in Table 1.5 it can be seen that the 281 age-Gc
correlation for VOC2 purified of PLS increases to .302 when short-term memory
(MPA) is removed. This correlation does not further increase, however, with
removal of concentration or cognitive speed or both. First, these results suggests that
there is a considerable portion of Gf operating in the VOC2 measure of Gc.
Secondly, these results show that the concentration and cognitive speed portions are
represented in the direct measure of Gf with PLS, but the short-term memory process
(MPA) that is involved in Gc functioning is not fully represented in PLS. This
suggests that there is a part of short-term apprehension memory involved in Gc that is
Figure 1,5b
The Effect of Removing the Variance of Memory for Paired Associates (MPA), Concentration
(CON), and Speeded Letter Comparison (SLC) from the Age Related Enhancement of
] ] 5 Vocabulary Test 2 when Power Letter Series is Controlled (VOC2:PLS).
✓VOC2:PLS (MPA Out)
/OC2 PLS alone
,VOC2:PLS (MPA+CON or SLC Out)
_-VOC2:PLS (CON or SLC Out)
/OC2:PLS (MPA+CON+SLC Out)
OC2 alone
in 00
IQ nt/decade change
2.72
2.58
2.47
2.33
2.28
.43
90
Legend:
VOC2=Vocabulary Test 2
VOC2:PLS=Vocabulary 2 with Power Letter
Series Controlled
MPA=Memory for Paired Associates
CON=Concentration
SLC=Speeded Letter Comparison
85
22 32 42 52
AGE
62 72 82 92
52
unrelated to the short-term memory functioning involved in the measurement of PLS.
When this short-term memory is controlled, the aging enhancement of Gc in increases
over the enhancement indicated by control for PLS alone. As seen in the lower rows
of columns 2 and 3 of Table 1.5 and as depicted in Figure 1.5b, controlling for SLC
and CON after control of short-term memory, results in a decrease of the aging
enhancement of VOC2 with PLS controlled. These decreases are statistically
significant. These results suggest that a pure measure of crystallized intelligence
(purified by removal of Gf and short-term memory) involves capacity for
concentration and cognitive speed That is, these processes enhance pure Gc. When
they are removed this enhancement is removed
The addition of CWA to PLS to further purify Gc and does not materially
change the results seen when only PLS is removed to purify the measure of Gc. The
results are the same with or without this further purification
Vocabulary 1 as a Crystallized Ability Indicator. Table 1.6 indicates the age
correlations when the relatively easy, skewed VOCl measure is used to indicate Gc.
This estimate of Gc actually correlates higher with age than does the VOC2 estimate
(see Figure 1 6 ). The enhancement per decade is 2.30 IQ units, compared with 1.43
for the VOC2 estimate This indicates that it is in respect to basic vocabulary, rather
than advanced vocabulary, that older persons are most knowledgeable relative to
younger persons. As is depicted in Figure 1.6 , when MPA is removed from the
VOCl estimate for Gc, the aging enhancement increases to .291, corresponding to
53
2 . 6 8 units of enhancement per decade. This is a significant increase from the base
model. Although to a lesser extent, this increase is maintained when either CON or
SLC is controlled in addition to MPA. When short-term memory is controlled, older
people, relative to younger people, score even higher on this estimate of Gc than
when MPA is controlled. However, as with the VOC2 estimate, control by CON
alone or SLC alone significantly decreases the enhancement for the VOC1 estimate
alone and for the VOCl estimate purified by removal of PLS and/or MPA. Thus,
again there is evidence that CON and SLC measure aspects of concentration and
cognitive speed that are parts of the Gc that improves with age in adulthood.
When PLS is removed from VOCl, the aging enhancement of VOCl
increases from 2.30 to 3.41 units per decade (see Figure 1.6 ). This increase is
approximately equal to the increase when PLS is removed from VOC2 (see Figure
1 5b). Thus, PLS has the same effect on the two estimates of Gc even though one is
highly skewed. Unlike the results for control of MPA in the VOC2 estimate of Gc
purged of Gf, the removal of MPA in the VOCl estimate of Gc results in decreases in
the age enhancement of Gc. This decrease is furthered by removal of CON and SLC
Thus, the evidence suggests that Gc improvement in adulthood involves maintenance
of short-term memory as well as a capacity for concentration and cognitive speed
However, the removal of any of the process variables MPA, CON, or SLC alone or in
combination results in decreased enhancement. If processes are involved in the
enhancement of VOCl, this involvement can be accounted for by removing the
54
Table 1.6. Effects o f Parting Cognitive Processes on the Correlations Between Age
Indicators o f Gc Based on the First Vocabulary (lr OCI) Test
VOCl VOCl VOCl
Alone PLS PLS+CWA
Age Correlation
BASE MODELS: .251 .371* .375
REDUCED
MODELS:
MPA Out .291* .357° .346
CON Out .234°
O
O
.304
SLC Out .242 .292° .278
MPA+CON Out .263’ .301 .300
MPA+SLC Out .263’ .297 .277
CON+SLC Out .229 .261° .247
MPA+CON+SLC .245° .270° .251
Out____________________________________________
* = Significant R2 change at the .05 level from the R2 of the just previous model in
which the effect is nested. Example: The R2 for VOCl :PLS is significantly greater
than the R2 for the model VOC 1 alone at p<05.
f = Significant correlation increase from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which MPA is parted is
significantly greater than the R2 for the model where VOCl is alone.
° = Significant correlation decrease from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which CON is parted is
significantly greater than the R2 for the model where VOCl is alone.
CON=Concentration; CWA=Common Word Analogies; ESA=Esoteric Word
Analogies; MPA=Memory for Paired Associates; PLS=Power Letter Series;
SLC=Speeded Letter Comparison; VOCl=Vocabulary Test 1; VOC2=Vocabulary
Test 2
Figure 1.6
The Kflect of Removing the Variance of Memory for Paired Associates (MPA). Concentration
(CON), and Speeded Letter Comparison (SI.C) from the Age Related enhancement of
Vocabulary Test I (VOC I) alone and with Power Letter Scries Controlled (VOC I PLS).
115
110
105
Ability
Level
in 100
IQ
Units
95
OC1 PLS alone
IQ ot/decade change
3.41
-VOCl: (MPA Out) 2.68
-VOC 1 : (MPA * CON or SI.C Out) 2.42
OC1 alone 2.30
-VOC 1 : (MPA - CON - SLC Out) 2 .25
O C l: (SLC Out) 2.23
% -V-
Legcnd:
VOC 1 ’Vocabulary Test 1
VOCl :PI-S’ Vocabulary 1 with Power Letter
Series Controlled
MPA Memory for Paired Associates
CON Concentration
SLC’ Speeded Letter Comparison
85
56
measurement overlap associated with fluid ability. The simpler measure of vocabulary
does not seem to require the same memory function than does the more difficult
measure.
Removal of the CWA estimate of Gf does nothing to change the results
indicated by removal of the PLS estimate of Gf CWA control does not further
“purify” Gc measurement.
Esoteric Analogies as a Crystallized A bility Indicator. Table 1.7 summarizes
age correlations for the Esoteric Analogies (ESA) estimate of Gc. Since ESA
correlates nearly zero with age (.071), Gc enhancement occurs only when the part of
it that involves Gf or a Gf process is removed (see Figure 1 .7). The results under
conditions of control are very similar in form and only slightly smaller in magnitude as
the results for other estimates of Gc. The removal of MPA increases the correlation
with age for ESA, as such. Removal of the PLS estimate of fluid ability also increases
the Gc correlation with age enhancement from .071 to .248. Further, the control of
this purged measure of Gf with MPA significantly further increases the aging
enhancement of the ESA estimate This suggests that short-term memory is involved
in the Gf portion of ESA performance and when this involvement is damped, Gc is
improved. This is the result seen for the VOC2 estimate of Gc. As with both the
VOC2 and VOC 1 estimates of Gc, when CON and SLC are controlled in the Gc
measure already purified by removal of Gf and MPA, the correlation with age actually
decreases Also as with the VOC2 and VOCl estimates, additional purification with
57
Table 1.7 The Effect o f Parting Cognitive Sub-processes on the Correlations
Between Age and Various Esoteric Analogies (ESA) Criterion Variables
ESA ESA: ESA
Alone PLS PLS+CWA
Age Correlations
BASE MODELS: .071 .248* .245
ALTERNATIVE
MODELS:
MPA Out .177’ .284* 266’
CON Out 113f .2 2 1 ° .215°
SLC Out .159* .234 .214°
MPA+CON Out .180* .253° .237°
MPA+SLC Out .206’ .256° .231°
CON+SLC Out .172 .219 . 2 0 0
MPA+CON+SLC .204 .238° .215
Out
* = Significant R2 change at the 05 level from the R2 of the just previous model in
which the effect is nested. Example: The R2 for ESA:PLS is significantly greater
than the R2 for the model ESA alone at p<05.
* = Significant correlation increase from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which MPA is parted is
significantly greater than the R2 for the model where ESA is alone.
° = Significant correlation decrease from the just previous model in which the effect is
nested at p<05. Example: the R2 for the model in which CON is parted is
significantly greater than the R2 for the model where ESA is alone
CON=Concentration; CWA=Common Word Analogies; ESA=Esoteric Word
Analogies, MPA=Memory for Paired Associates; PLS=Power Letter Series;
SLC=Speeded Letter Comparison; VOCl =Vocabulary Test 1; VOC2=Vocabulary
Test 2
I
I
Figure 1.7
The Effect of Removing the Variance of Memory’ for Paired Associates (MPA). Concentration
(CON), and Speeded Letter Comparison (SLC) from the Age Related Enhancement of Esoteric
Analogies (ESA) as an Indicator of Fluid Ability.
IQ pt/decade enhancement
115
ESA: (MPA+SLC Out)
ESA: (MPA Out)
ESA: (SLC Out)
ESA: (CON Out)
1.89
1.63
1.46
110
1.04
105
(SA alone
0.65
Ability
Level
in 100
Units
Legend:
ESA=Esoteric Analogies
MPA=Memory for Paired Associates
CONConcentration
SLC Speeded Letter Comparison
92 22 32 72 42 82
59
CWA control does nothing to change the correlations with age above and beyond
changes seen when PLS is controlled.
Summary
These analyses indicate, first, that Memory for Paired Associates (MPA),
Concentration (CON), and Speeded Letter Comparison (SLC) can be considered
cognitive processes of Fluid Reasoning (Gf). When these process are controlled, the
aging decline of Gf is significantly decreased This indicates that the differences
between younger and older persons in measures o f fluid reasoning, which differences
suggest aging decline of Gf, are partly related to differences in short-term memory,
cognitive speed, and capacity for maintaining concentration This finding is indicated
not only for the Power Letter Series (PLS) estimate of Gf, but also for a Common
Word Analogies (CWA) indicator. The PLS indicator contains less measurement
overlap with indicators of Crystallized Knowledge (Gc) and requires less Gc ability
than does the CWA indicator, but the results for the two indicators are in major
respects the same. Controlling Vocabulary estimates of Gc in G f indicators
demonstrates that age differences in Gf may be larger for the pure construct of Gf
than is indicated by the fallible measures obtained with a lone test such as Power
Letter Series. Though controlling for the processes of short-term memory, cognitive
speed, and concentration measures with fallible measures in this study significantly
reduces the aging decline of Gf, unexplained decline in Gf remained after these
6 0
controls. A significant part of Gf and aging decline o f Gf was not accounted for by
MPA, CON, and SLC.
Results for analyses of measures of crystallized knowledge indicate that for
each o f three estimates o f Gc there are differences between younger and older persons
suggesting that with increase in age in adulthood there is increase in knowledge This
is true both for a basic vocabulary indicator of Gc fo r which the distribution was
skewed to the ceiling— many people identified the meanings of most o f the words— as
well as for a vocabulary test for which the distribution was more nearly symmetrical
The age increase in G c was larger for the skewed measure than for the more
symmetrically distributed measure An Esoteric Analogies (ESA) measure of Gc w as
found to contain a substantial portion o f Gf, but when this was controlled, it indicated
essentially the same increase with age as was indicated for vocabulary estimates of
Gc
The results from these analyses suggest that short-term apprehension retrieval,
cognitive speed, and capacity for focused concentration are processes of Gf which,
when controlled in Gc, result in an increase of the ag inS enhancement of Gc
indicators, just as when Gf indicators, as such, are controlled in Gc That is, when the
fallible measures of G c are not contaminated with G f processes, then older subjects
show an even greater Gc advantage over their younger counterparts than when fallible
Gc measures are analyzed.
61
The involvement of short-term memory in the measurements of Gc was found
to persist even when direct indicators of fluid ability (PLS and CWA) were removed.
That is, the correlation of Gc with age increases with control of MPA even after
control of Gf indicators had been effected. This result suggests that an aspect of
short-term memory is involved in Gc enhancement (just as some aspect of MPA is
involved in the aging decline of Gf). This result thus suggests that there is a part of
short-term apprehension retrieval that operates as a process in Gc
Control of cognitive speed and capacity for concentration in Gc purged of Gf
and MPA reduces the correlation between the purged measure and age. This
suggests that the Gc construct, as such (not involving Gf or short-term apprehension
retrieval), involves capacity for concentration and cognitive speed. That is, the Gc
capacity that improves with age in adulthood may involve capacity for concentration
and cognitive speed that cannot be accounted for by measures of Gf. If true— i.e., if
this finding is established in replication— this questions current theory which stipulates
that cognitive speed and attentiveness, per se, are features of intellect that decline
with age in adulthood It must be recognized, though, that the current finding
pertains to a construct— Gc purged of Gf and short-term memory— that is more refined
than constructs represented by the simple test scores on which present theory is
based
62
Discussion
On average, the results of this research indicate an age related decline of fluid
reasoning ability of approximately 3.06 to 4.97 IQ units per decade. This finding is
well within the bounds o f the declines found in other studies. Horn, et al. (1981) and
Horn & Donaldson (1980) reported between 3 and 7 IQ units per decade decline;
Hofer (1994) found between 2.5-5.0 units. Similarly, the range of 1.56 to 3.41 IQ
unit per decade age related improvement of Gc found in this study is comparable to
the ranges (.05 to 2.10) reported in the Horn et al. (1981), Horn & Donaldson (1980)
and Hofer (1994) studies. Thus, the findings of previous research are replicated here
The results show that there are aspects of cognition that develop with fluid
ability (Gf) along a the same trajectory. These aspects of cognition-short-term
apprehension retrieval, concentration, and cognitive speed— appear to be processes of
Gf. They are distinctly different and function independently, but all contribute to
individual differences generally and to age related differences in Gf. In particular,
they indicate that fluid reasoning requires that one hold bits of information in
immediate awareness, be able to quickly manipulate information before it leaves
immediate awareness, and persist at attending to these processes in order to
comprehend relationships and arrive at correct conclusions. When there is damage to
any one process, the entire system is affected. The evidence of this study adds to the
evidence o f other research to suggest that these three processes decline with age.
63
Part of the observed age related decline of Gf is accounted for by aging decline of
these processes.
The results here are consistent with findings from previous research on these
same processes in finding that age related decline remains even after the variance of
process variables is removed. Depending on which indicator of Gf is analyzed there is
approximately 0.59 to 1 43 IQ units decline not explained with cognitive process
variables. It is logical to assume that this residual decline is associated with a part of
Gf that is uniquely inductive reasoning. It is logical, also, to suppose that causal
factors operating through development produce the observed decline in this residual
as well as the observed decline in the processes. But what are these factors? What
aspects of human functioning outside the realm of cognition can aid in understanding
the processes and the residual construct. What might influence the development of
these factors? These are questions to be studied in further analyses of this research
The present results indicate that control for a Gc component in Gf increased
age-related decline by between 1.11 and 1.63 IQ units per decade. This result comes
about because Gf and Gc are positively correlated, but the age correlation is positive
with Gc and negative with Gf.
The positive correlation between Gf and Gc may come about for either of two
basic reasons. Gf and Gc may stem from the same influences (e.g., a common set of
genetic determiners or environmental conditions), or, there may be some overlap in
the operations of measurement of the two— both depend to some extent on language
64
proficiency. If Gf and Gc stem from the same influences, which is assumed under the
theory of general intelligence, then even under the most ideal circumstances of
measurement and subject sampling, the two would be positively correlated. On the
other hand, if Gf and Gc are separate and independent, then their positive correlation
may result only because tests used to measure Gf involve some of the same
requirements as tests used to measure Gc
In any study concepts are imperfectly represented by measurement operations.
The partialling procedures used in this research are ways of studying phenomena
under an assumption that operational measurement definitions are confounded.
However, the control achieved with these procedures remove any common-influence
effects that produce a true correlation Since we don’t know the extent to which Gf
and Gc are produced by the same influences, we can’t accurately assess the extent to
which control for confounding of measurement operation is over-control that
suppresses true correlation.
Thus, control of Gc in Gf decline estimates decline for a psychometrically
purified construct but may remove common-influence variance In sum, to control Gc
in assessing the aging decline of Gf (1) removes aspects of Gf measurement that
involve common (with Gc) psychometric factors— verbal protocol, test-taking
sophistication, etc., (2) removes Gc factors of acculturational knowledge and
education that confound measurement of Gf, and (3) removes factors that are
intrinsically common to the development and manifestation of Gf and Gc.
65
As concerns the manifestation of Gc itself, removing the influence associated
with short-term memory in Gc results in an increase in the age correlation even when
Gc has been purged of Gf variance. Thus, just as Gf can be regarded as a
contaminant that confounds measurement of pure Gc, so short-term apprehension
retrieval (SAR) is a contaminant in the measurement of Gc. When this part of SAR is
removed, the aging enhancement of Gc increases substantially. The difference
between old and young Gc functioning is even greater than the differences observed
when SAR is operating in the fallible measurement of Gc
There is a connection between SAR, which declines with age in adulthood and
Long-term Storage and Retrieval (Glr), which increases over much o f adulthood
(Horn, 1985; Horn et al, 1981). Glr is to some extent determined by SAR
(Broadbent, 1966). In order for information to be consolidated in long-term storage
and later retrieved, short-term apprehension must operate. Glr, in turn, is vital to Gc
functioning— one cannot demonstrate the accumulated knowledge of Gc if it is not
consolidated and then retrieved from long-term storage. Because SAR helps produce
Glr, and Glr is functionally related to Gc, it is reasonable that individual and age
differences in SAR are related to individual and age differences in Gc. It is reasonable
that Gf and Gc are linked through this kind of interdependent set of relationships
What the present evidence indicates is that this kind of linkage is separate for SAR
and Gf
There is still much about the decline of vulnerable abilities that we do not
understand and there is little known about the portion of Gf that cannot be explained
by cognitive processes. The aim of the research to be described in subsequent
sections o f this report is to determine if non-cognitive variables of lifestyle can help
account for individual differences in age related differences in Gf and Gc.
67
STUDY 2 LIFESTYLE INDICATORS OF COGNITIVE ABILITIES
Introduction
The Operational Definition o f Lifestyle
Alfred Adler [1870-1937] was one of the first social scientists to develop a
theory in which lifestyle was the central feature of personality. He proposed a
concept of Lebensstil (lifestyle) to replace concepts he had described earlier such as
“guiding line” and “lifeplan.” He aimed to refine the idea that an individual is a
purposive actor in life Style of life in his writings refers to the “ unity, goal-
directedness, and uniqueness of a person's actions, which are for the most part
subjectively determined” (Adler, 1968)
Adler split from Freud because he became increasingly reluctant to think of
human behavior in terms of unconscious and unrationalized drives or motives. He
was developing a more self-determined, self-rationalized conception of human
behavior, one that characterized the human as a holistic organism that behaves as a
totality of mind and body. “A human being is not a machinelike structure first and a
moving system second, as when pushed by libido [or other psychosexual motives].
People appear to have set structures only because their behavioral movements take on
a consistent style or form” (Adler, 1964)
Adler reasoned that it was not correct to characterize a person by some
limited feature of behavior-such as in Freud’s concept of anal personality. While
anality does convey a quick impression of the person, and may characterize a limited
68
portion of a person’s behavior, it does not provide a complete picture of the
personality or describe fully the person’s behavior. In order to get a comprehensive
understanding of personality, we have to consider the person’s early life setting, the
“prototype” (be it conscious or unconscious) he or she puts down as the essential
goal for life movement, the resultant law of movement, and the success or failure of
this strategy in pursuit of the projected goal (Adler, 1968). The important factor here
that what separates Adler’s theory from that of Freud is the emphasis on goals for life
and the consciousness of successes and failures in pursuit of those goals According
to Adlerian theory a person behaves the way he or she does for reasons he or she
thinks are rational in moving toward life goals.
When we collate (in some way) the rationalized pursuits of a human being,
and refer to them collectively, the result is the person’s lifestyle. The essential aspects
of personality manifest themselves in the behaviors generated by lifestyle. To attempt
to understand a person is to assess that person’s lifestyle.
Adler's idea of lifestyle is holistic. It encompasses a person's ideas, behaviors,
thoughts and actions. The theory implies that there are individual differences in
lifestyle, but Adler was never clear in specifying how such differences can be, or
should be, assessed.
In theories of those who followed Adler, there are efforts to describe lifestyle
as a feature of individual differences in personality, in some cases as the chief
determinant of personality. Coleman (1964) defined lifestyle as “the general pattern
69
of assumptions, motives, cognitive styles, and coping techniques that characterize the
behavior of a given individual and give it consistency.” This definition is provocative,
but, like Adler, Coleman doesn't indicate how individual differences in lifestyle might
be measured.
In the last twenty years most studies concerned with operational definitions of
lifestyle have been referenced in the socio-medical literature These operational
definitions have most often pertained to “health behavior” such as that relating to diet,
smoking, and exercise. Variables used to define lifestyle have been specified in terms
of “sedentary versus active” (Green, 1984) or “health promoting verses health
threatening” (Pope, 1982). Usually no more than one or two such types have been
specified These concepts of lifestyle have not had the mult-faceted quality suggested
by the theories of Adler or Coleman. Particularly in cognitive psychology, lifestyle
has been narrowly defined, theory has not been well developed, and measurement has
been sparse.
Gribbin, Schaie, and Parham (1980) constructed an interview form [the Life
Complexity Inventory (LCI)] to obtain data on the “lifestyle and activities” of adults.
With factoring, they defined eight clusters in data gathered with the LCI on 140
subjects The clusters included: a) Homemaker Activities; b) Level of Social Status;
c) Subjective Dissatisfaction with Life Status; d) Disengagement; e) Semi
engagement, f) Noisy Environment; g) Degree of Family Solidarity; andh)
Maintenance of Acculturation. Correlations of scores derived from the clusters with
70
scores on several cognitive tests were presented Verbal ability, a marker of Gc, was
found to be positively correlated with Social Status (r=. 52) and Acculturation (r=. 19)
and negatively correlated with Disengagement (r=-.48). Inductive reasoning ability, a
marker of Gf, was found to correlate positively with Social Status (r=.55) and
Acculturation (r=. 16), and negatively with Disengagement (r=-.46), and Family
Dissolution (r=-.21). Word Fluency correlated positively with Social Status (r=.34),
Acculturation (r=.25), Noisy Environment (r=.24) and negatively with Disengagement
(r=-.22) The items that make up each cluster in the LCI are diverse and ambiguous
The concepts represented by the clusters are not clear. The cluster labels are
somewhat misleading. The theory on which these measures are based has not been
clearly stated
Arbuckle, Gold, Chaikelson, Schwartzman, and Andres (1992) examined the
relationship between maintenance of intellectual abilities and what they called
“lifestyle” measures. Lifestyle in this case included such general personality variables
as Introversion vs. Extroversion, negative affectivity (illness, depression, anxiety, and
stress), and engagement (education, socioeconomic status, social activities and
reading activities). Rationale and theory in this study, also, was lacking. Results
indicated that engagement was associated with an increase vocabulary scores across
age Negative affectivity was associated with decline in arithmetic scores.
Both Gribbin, et al. (1980) and Arbuckle, et al. (1992), neglected to report
reliabilities of the measures and factor structures. Such incomplete reporting provides
little basis for using these studies to build theory.
The present study is part of a larger, ongoing study in which it is assumed that
lifestyle is made up o f four broad domains of life goals. These are discussed under the
short-hand headings of Physical Health, Psychological Health, Activity Level, and
Personal Control. The domains are intended to encompass Adler’s notion of rational
striving and thus lend a holistic quality to the concept of lifestyle, even though merely
specifying the domains does not constitute a theory. The domains provide a rationale
for specification of variables, however, and thus provide a basis for construction o f
operations of measurement of variables. Within each domain several constructs are
specified in terms of measures that indicate the extent to which aspects of lifestyle are
promoted.
Unlike the larger study, the present study does not aim to develop a theory of
lifestyle. Rather the aim is to describe how variables of lifestyle relate to cognitive
abilities and age differences in cognition across the adulthood period of the lifespan.
The present study draws from the broad theory of lifestyle proposed by the larger
study and samples a manageable subset of constructs from the Lifestyle Assessment
Battery (LAB) of the larger study. Constructs for which there is evidence of
relationship to vulnerable or maintained abilities, or for which there is plausible
hypothesis of such relationship, were sampled from each of the four domains o f
72
lifestyle The extant evidence and the plausible hypotheses specifying how variables
o f lifestyle relate to cognition and cognitive development will be outlined in the
following sections.
Physical Health. Under this heading are considered thoughts, strivings,
behaviors and martialling of motives aimed at attaining and maintaining physical
health Whether a person fully realizes it or not, maintenance of physical health is a
necessary prerequisite for life itself and for the energy that can be devoted to other
activities If individual differences in dealing with this reality exist and can be
measured, it is reasonable to suppose that they relate to human abilities— as causes and
as consequences. A basic hypothesis of cognitive development stipulates that the loss
in cognitive functioning is due to a loss of neurons in the brain. Many things can
produce such loss-high fever, anoxia, carbon monoxide, drugs, head trauma, etc.
Particular lifestyles may predispose one to factors that produce neural loss
Neuronal overdetermination may prevent neural losses from registering as
cognitive losses. If 100 neurons can produce a cognitive capability, this skill is said to
be overdetermined relative to a skill in which only 10 neurons interact to produce the
behavior (Hebb, 1949) When there is substantial overdetermination, then notable
amounts of neural tissue can be lost without loss of the skill. The maintained abilities
(Gc and Glr) can be understood as neurologically overdetermined abilities. Factors of
lifestyle may predispose one to develop such overdetermined abilities.
73
If the capacities called upon in a task are not overdetermined, a small amount
of tissue damage o r loss can result in a tangible loss of functioning. The processes of
intellect that depend on this kind o f neural support (the vulnerable abilities, Gf, Gv,
Gs, Gsm) are notably weakened with the loss o f brain tissue
The loss o f neurons and subsequent loss of ability may not be evident when
only short periods o f adulthood development are considered because the amount of
neural loss is small relative to the billions of neurons that support human intellectual
functioning (Thompson, 1993) Large losses, however, as indicated by brain damage
or over long periods of aging, can be accompanied by notable behavior and
intellectual changes (Horn, 1968; 1979).
Loss o f health generally is expected to be both a direct and indirect indicator
o f neural tissue loss. It is a direct indicator when loss of neural function produces
declines in health, as when, at an extreme, the neural losses o f Alzheimer’s Disease
produce declines in respiratory efficiency Similarly, uncontrolled hypertension
produces brain damage which produces loss o f cognitive capacity.
Loss of health may be an indirect indicator of loss o f neural tissues and loss of
cognitive capability if, for example, it results in lack of motivation to learn and think
and “exercise” intellect, which, in turn, results in decreases in neural processor
functioning and dendritization. this decrease many, in turn, cascade into further
decrease in motivation and “exercise” of intellect.
74
Health can be assessed with several kinds of measures— those of direct reports
to an investigator or physician, those of diagnostic measures, such as blood pressure
recordings, and those of health practices such as diet and exercise habits. Results
from several studies indicate relationships between intellectual performance and both
transitory and chronic health problems across the life span (Biren & Schaie, 1977;
Elias, Elias, & Elias, 1990; Harris & Frankel, 1977; Hebb, 1942; Horn, 1979; Horn &
Donaldson, 1980; Kahn& Antonucci, 1980; Perlmutter& Nyquist, 1990; Seigler,
1989).
Although health problems can negatively affect intellectual functioning at any
age, health problems are more prevalent at older ages than at younger ages.
Individuals who show significant cognitive decline report significantly greater
numbers of illness diagnoses, medical risks factors, and more frequent doctor visits
(Palmore, 1985; Schaie, 1989). Hertzog, Schaie, & Gribbin (1978) found that over a
14-year period individuals at a high risk of cardiovascular disease, on average,
declined earlier in primary factor indicators of perceptual speed, inductive reasoning,
and visualization than did those who showed low risk. Poor nutrition (Goodwin,
Goodwin, & Gary, 1983) and alcohol abuse (Hutchinson, Tutchie, Gray, & Steinber,
1964; Horn, 1982, Kish & Cheney, 1982; Parker & Nober, 1980; Parsons, 1975)
have also been shown to be negatively related to intellectual functioning.
It has been speculated (Biren et. al., 1980; Horn, 1979; La Rue & Jarvik,
1982) that cardiovascular benefits derived from exercise may help to forestall
75
degenerative changes in the brain associated with normal aging. Craik and Byrd
(1982) proposed that brain capacity limitations are analogous to "mental energy" and
that physical energy is closely related to mental energy. They also propose that the
heightening of physical energy will result in the heightening of mental energy. People
who exercise regularly report more euphoric moods, lower stress and depression, and
better cognitive functioning than more sedentary people (Blumenthal et al., 1988).
Self-reported exercise was found to significantly predict individual differences in
reaction time, working memory, and Gf (Clarkson-Smith & Hartly, 1990).
Although it has been reported that maintained and vulnerable abilities are
positively related to self-reported exercise behavior (Clarkson-Smith & Hartly, 1989),
the evidence for whether or not exercise intervention affects vulnerable abilities in
experimental studies is mixed Some studies have shown that the introduction of an
exercise program helps to improve ability scores, others show that the exercise makes
little or no difference, and still other studies report that exercise actually results in a
decrease in ability Elsayed, Ismael, and Young (1980) reported that over a four-
month period an exercise intervention group and a non-exerciser group improved on a
retest measure of G f and were not significantly different in this improvement. In the
Powell and Pohndorf (1971) study, exercisers performed better on nonverbal
reasoning tasks than did non-exercisers, but the differences were not significant.
Dustman et al. (1984) found a significant Gf improvement for a non-exercising
control group but not for an exercise intervention group.
76
In sum, vulnerable abilities are thought to be associated with underdetermined
neuronal networks that are likely to be depleted by lifestyle factors that operate over
adulthood development. Maintained abilities are associated with overdetermined
networks and overdetermination may be related to lifestyle. Loss of physical health is
expected to be associated with neuronal loss and, thus, to loss o f vulnerable abilities
and to the aging decline of these abilities. Cardiovascular health may work to
forestall cognitive decline. Experimental studies designed to demonstrate this in the
short run have produced mixed results.
If self reports o f efforts to attain and maintain physical health are at least
monotonically related to actual maintenance of physical health, then the self report
measures should relate positively to maintenance o f abilities. A working hypothesis
of this study is that self report measures of involvement in exercise and physical
fitness produce mental energy and overall well-being that, in turn, positively affects
the cognitive performance of both vulnerable and maintained abilities.
Activity Level. Our theory o f lifestyle stipulates that people strive to realize
their potential. Individuals have beliefs about what they can do and what they can
become. They engage in activities directed at doing what they can do and becoming
what they can become. These strivings are embodied in their activity levels.
Folk wisdom promotes the hypothesis that staying active and involved can add
years and quality to life. An active lifestyle is often cited as the single most important
feature of "successful aging." According to Thomae (1980), centenarians and
77
professional people who have continued to be active and engaged into their seventies
and eighties are the "elite" of the aged population. M. Baltes (1988) and P. Baltes
(1987) described "selective optimization with compensation" as the prototypical
strategy for successful aging. This involves enriching life through aesthetic,
avocational activities that stimulate intellectual growth (Baltes & Baltes, 1990)
Similarly, the essential element of successful aging according to Featherman, Smith
and Peterson (1990) is activity (both physical and mental), which is required to
achieve "adaptive competence." In these theories it is assumed that humans have a
large reserve capacity for activity-a plasticity— that can be tapped to meet the
adaptive, behavioral, and cognitive requirements of old age.
Results from a number of studies add credence to theory that activity level
predicts maintenance of cognitive function in the elderly— the "use it or loose it"
hypothesis has found empirical support. In a study by Craik, Byrd, and Swanson
(1987), highly active elderly people of different levels of socioeconomic status,
recalled as much information in a short term memory experiment as did young college
students, while inactive elderly subjects performed much worse than the college
students In the Gribbin et al. (1980) longitudinal study, a socially active
environment, continued education, and occupational activity (both social and
physical) were positively related to the maintenance of intellectual performance.
DeCarlo (1974) found that recreational activity of a cognitive type was the best
predictor of intellectual functioning in old people.
Results from several studies suggest that an intellectually stimulating spouse,
exposure to stimulating environments, utilization o f cultural and educational
resources, and employment in complex jobs relate positively to intellectual
functioning (Miller, Slomezynski& Kohn, 1987; Schooler, 1987; Gruber & Schaie,
1986). In the Dutta, Schulenberg, and Lair (1986) study, retirement was found to be
associated with improvement in cognitive functioning for those retiring from
routinized jobs, but appeared to accelerate decrement in those retiring from complex
jobs. Arbuckle, et al., (1986) reported that older adults who maintain a lifestyle that
involves more intellectual and social stimulation perform better on memory tasks than
adults who do not maintain such a lifestyle This result held even after controlling for
confounds of health and education.
Activity level in general appears to be correlated with physical health and
mobility, which, in turn, is expected be enhance cognitive performance. The benefits
of active lifestyle also appear to extend into the realms of psychological health and
thus have an indirect effect on cognitive health as well. A working hypothesis of this
study is that self reports of involvement in non-sedentary activities will relate
positively to abilities, and involvement with sedentary activities will relate negatively.
It is expected that the relationship will be stronger for Gc than for Gf. This follows
from reasoning that activity involvement of the kind considered non-sedentary (as will
be defined operationally) is part of an acculturational process that enhances Gc and
distinguishes development of Gc from development of Gf (Horn, 1992).
79
Psychological Health. This domain includes strivings, thoughts, behaviors
and martialling o f motives to attain and maintain psychological health. This health,
like physical health, sets conditions for other kinds of realization of goals one strives
to attain through ones lifestyle. It is assumed that self reports indicating awareness of
need, and efforts, to attain psychological health will be positively related to actual
psychological health, which in turn will relate positively to cognitive abilities. To
understand relationships between psychological health and age related changes in
cognitive abilities, evidence linking aspects of psychological health to cognitive
performance must be explicated in the context of evidence for age differences in
psychological health constructs. This requires a statement of what is meant by
psychological health. The psychological health domain might be defined (in part) in
terms of dimensions that constitute the “Big Five” theory of personality (McCrea &
Costa, 1987), or the primary as well as secondary factors of personality (Cattell,
1957). There is evidence of metric invariance of the second-order factors of Cattell’s
questionnaire (Hofer, 1994). This is a plus Other evidence suggests, however, that
the means for these factors change very little over the adult period (Eysenck, 1981;
Hofer, 1994) If age differences in the changes of these personality constructs are not
significant, then such differences cannot be significantly related to cross-sectional age
differences in cognitive abilities such as Gf and Gc. Psychological health in the
present study should not be defined in terms of the personality dimensions of the Big
Five and 16PF
80
Psychological health concepts hypothesized to relate to cognitive abilities
involve one’s commitment to improve oneself, seek out social support, deal with
stress, and utilize resources. Such concepts also pertain to psychosocial adaptation,
satisfaction in close, interpersonal relationships, positive self-regard, tolerance for
ambiguity, and flexibility in new and unfamiliar situations. Personality dimensions that
are indicative of factors of the McCrea and Costa or Cattell systems may emerge from
analysis of items and scales designed to measure such a set psychological health
constructs. If so, the results will help indicate the variety of stimuli (items) that can
be used to identify the factors Also, such results will help fit the factors into theory
of psychological health that derives from an Adlerian conception of lifestyle.
Past research that shows a connection between psychological health
constructs and cognitive aging will be discussed in the following paragraphs.
Measures of intellectual abilities often have been included only rarely in
research on personality. When abilities have been included in such research, it has
generally been found that factors indicating neuroticism, emotional instability,
delinquency, paranoia, and, in general, poor adaptation and adjustment correlate
negatively at a low level (-.15 to -.30) with most measures of abilities, including Gf
and Gc. Self-report measures of self-sufficiency, expression of questioning attitude,
liking for thinking, conscientiousness, persistence, and reality checking, on the other
hand, generally correlate positively with measures of cognitive capabilities (Cattell,
1971, 1986; Horn, 1982).
In work specifically focused on adult development, Schaie (1984) found
measures of self-efficacy, self confidence, and being in stable marital-type relationship
were positively correlated with maintenance of intellectual functioning (including Gf
and Gc). In a study assessing the fluid abilities of a group of centenarians, Poon,
Martin, Clayton, Messner, Noble, and Johnson (1992) found that Gf was positively
related to a lack o f depression, good adaptation, and the ability to maintain the
management of everyday needs. Poor control of aggression and impulsivity (Kleban,
Lawton, Brody, & Moss, 1976) as well as a diagnosis of neurosis, were found to be
associated with poorer intellectual performance in the elderly (Nunn et. al., 1974).
Of the “Big five” personality traits, introversion-extroversion, and neuroticism
have been the most studied. Introversion has been found to be positively correlated
with better performance on memory ant information-processing measures. The better
performance of introverts is thought to result from a more intensive analytic style of
information processing (Eysenck, 1981; Stelmack, 1991). Neuroticism has been
associated with less accurate performance but faster reaction times on tasks requiring
well-learned responses (Arbuckle et al., 1986; Cattell, 1971; Costa, Fozard, McCrae,
& Bosse, 1976). Neuroticism, even in nonclinical samples, is thought to reflect a
heightened arousal level, producing a decreased accuracy of memory responses but
some increase in speed on simple reaction time measures (Gold & Arbuckle, 1987).
Cattell (1971) reasoned that the negative relationship between neuroticism and
82
cognitive functioning (both Gf and Gc) should be understood in terms of the inability
to mobilize and apply oneself effectively, not to lack of insight, per se.
In a study of 20-90 year-olds, Perlmutter and Nyquist (1990) found negative
correlations for both Gf and Gc with the number of self reported psychological
symptoms, including anxiety, fear, anger, paranoia, depression, and impulsivity.
These symptoms were also more prevalent in the older subjects than younger
subjects Although the authors did not report it, when age is partialled out of these
relationships, the correlations with Gf go to zero but remain significantly different
from zero for Gc. This suggests that the prevalence of these psychological symptoms
is negatively related to the aging enhancement of Gc but is not related to the aging
decline of Gf.
Satisfaction with social support has been shown to be related to reduced
short-term memory declines in adulthood (Arbuckle, Gold, Schwartzman &
Chaikelson, 1992) Social support has been found to buffer the effects of negative
stressors (Krause, 1986) Environmental stressors— especially negatives stressors— have
been found to be associated with intellectual decline (e.g., Amster & Krass, 1974).
Stress levels may be associated with poorer cognitive performance because
dealing with stress makes heavy demands on cognitive processing resources, thus
taking them away from maintenance and development of cognitive abilities. Social
support would be expected to mitigate the demands of stress on cognitive resources
83
and thus relate positively to cognitive performance. Greater social support may also
be indicative of greater social and intellectual stimulation.
A tolerance for ambiguity has been shown to be positively related to Gf, short
term memory, and concentration ability (Hooper, Hooper, & Colbert, 1984). In
Hooper, et al tolerance for ambiguity was defined as the willingness to accept
alternative states of affairs, and a capability to accept alternative interpretations or
alternate outcomes— feeling comfortable when faced by a complex social issue in
which opposed principles are intermingled Low tolerance was characterized by the
desire to have everything reduced to simple black and white terms The authors relate
development of Piagetian formal-operations to development of ambiguity tolerance.
A construct similar to ambiguity tolerance has been shown to be negatively related to
age (Blackburn, 1984)
Attitudinal flexibility was found to be significantly related to cognitive
flexibility in a longitudinal study (Schaie, Dutta, & Willis, 1991) o f cognitive aging.
Attitudinal flexibility reflected subjects’ self-reported tolerance for ambiguity,
unpredictability, and sudden changes in their daily lives. This personality factor
helped explain individual differences in the age related cognitive decline of inductive
reasoning tasks.
In sum, psychological health, broadly defined to encompass personality
constructs at a lower level than second and third-order factors of the “Big Five” or
16PF systems, includes variables indicative of social support, self efficacy, relationship
84
satisfaction, lack of depression, impulsivity control, positive adaptation, lack of
neuroticism, introversion, social support, and tolerance for ambiguity. These
variables have been shown to relate to cognition as well as to aging. They have
helped explain individual differences in cognitive abilities independently of
relationships to age, and they have been found to be related to age independently of
relationships to cognitive abilities. Psychological health variables have generally,
although not always, related more highly to maintained abilities than to vulnerable
abilities.
Personal Control. Theories of the self regulation and self-concept touch large
areas of social, personality, and developmental psychology. The ways in which the
self is perceived and monitored have been widely studied as essential aspects of
human emotive and cognitive functioning.
Self-esteem, according to James (1892), can be understood in terms of a ratio
of successes to aspirations. If we are able to achieve that to which we aspire, then
our sense of self-esteem is enhanced. When we fail to achieve what we have
envisioned for ourselves, our self-esteem suffers.
James reasoned that the concept of self is domain specific. That is, our self
esteem is vulnerable only in areas of our lives where we have specific aspirations.
Lack of competence in domains deemed unimportant to the self and will not adversely
affect self-esteem.
Cooley (1902), in his theory of the "looking glass self, suggested that the self
is a social construction, created by casting one's gaze into the social mirror to
ascertain the opinions of significant others toward the self. Others become the
mirrors into which we gaze for information about the self, and these opinions, the
reflected appraisals of others, are incorporated into the self. From this perspective, if
others hold the self in high regard, one's own sense of self-esteem will be high.
Conversely, if others have little regard for the self, one will form similar opinions
resulting in low self-esteem.
Thus, James and Cooley have put forth two very different theories about self
regulation and self-concept For James, self-esteem is the product of one's own
evaluation of one's competence in specific domains For Cooley, self-esteem is the
mirroring of the opinions that others hold about oneself. Both theories may be
basically correct within individuals or for explanations of individual differences.
Within individuals, for example, one may conform to the mirror theory in respect to
some areas of function— say, in regards to physical abilities— but conform to the
Jamesian theory in other areas— say, in regards to moral behavior. Between
individuals there may be some people who very much adopt their self image from the
reflection they get from others, whereas other people are relatively obtuse about the
views others may have of them and remain steadfast to views developed within an
egocentric evaluation framework. Indeed, the existing evidence suggests that both
the James and the Cooley kinds of theories are in broad respects correct. That is,
86
both systems operate within each individual and fluctuate over specific domains,
perspectives, and/or periods of development. In her studies of children and
adolescents, Harter (1985, 1987, 1990) concluded that both the James and Cooley
formulations, taken together, provide a powerful explanation for individual
differences in levels of self-esteem. In these studies, levels of social support were
shown to be correlated with levels of self-esteem. Subjects reported higher self
esteem in areas of particular interest.
Present self-perception may be predictive of future behavior (Bandura, 1989;
Hooker, 1992). Bandura has written about the role that self-regulatory processes
play in implementation of goals. "By cognitive representation in the present,
conceived future events are converted into current motivator and regulators of
behavior" (Bandura, 1989; p. 729). People who believe strongly in their capabilities
are persistent in their efforts to achieve goals. Outcome expectancy, in Bandura's
theory, is linked to self-efficacy in that one's subjective probability that an outcome is
attainable depends on the degree of confidence in one's own personal agency to effect
that outcome.
It has been suggested that forces outside the self also may influence goals and
their outcomes (Heckhousen, 1986; Rotter, 1966) Generally, the thought is that for
some people there is belief that other people, social factors, Deity, and a variety of
other forces may determine one’s behavior. For other people, such factors are not
important and only oneself sets the compass for travel into one’s own future.
87
Thus, there may be more than one locus of causality from which expectancies
arise. Beliefs about personal efficacy and external circumstances, including social
support, will influence goal attainment as elements of personal control.
The Unidimensional Theory. Much of the early work on locus of control was
based on a theory of a single dimension In particular. Rotter's (1966) theory
postulated that locus of control was a global or generalized trait and was expected to
be stable over time and consistent across situations. That theory specified that people
either believed that they controlled their lives (internal control), or that their lives
were controlled by forces beyond the self (external control). Measures were designed
to represent this theory. With such measures it was demonstrated that averages
changed with age in adulthood. But this evidence turned out to be equivocal Some
studies found an increase in internality with age (Seigler & Gatz, 1985); other studies
found the opposite. The drift over most studies suggested that older persons were
more likely to report that external forces control one's life (Lachman, 1985)
The Multidimensional Theory Lachman (1986) attributed the conflict in the
evidence to insufficiency in locus o f control measures (i.e., the Rotter I-E scale) She
argued that locus of control is multidimensional and that a unidimensional contrast
between "internal" and "external" hides changes along different dimensions. Older
people often acknowledge the importance of outside influences (i.e., other people or
luck) on some aspects of their behavior but still believe that what they do matters in
many walks of life.
88
Brandtstadter and Baltes-Gotz (1990) have shown that older adults are more
heterogeneous than younger adults in their attributions of personal control. Older
persons more frequently report that there are multiple forces in control of their lives
and view the dichotomy of internal versus external locus of control as limiting.
Uncontrollable events and life changes accumulate as we grow older. Very likely, a
more sophisticated belief system becomes necessary to cope under such
circumstances (Baltes, 1973, Brim & Ryff, 1980; Seligman & Elder, 1986)
A review of factor analytic studies of the Rotter scale provides further
evidence for a multidimensional theory of personal control. Although results based
on analyses of multiple choice items are similar to those generated by forced-choice
(ipsative) scales (Gatz & Good, 1978), the factors in either case have not been shown
to be even configurally invariant. Gurin, Gurin, and Morrison (1978) reported that
some items deal with the control of the person, others deal with areas of impersonal
control (e g , political systems), and still others deal with philosophical ideologies.
Gurin and Brim (1984) found factorial distinctions between personal efficacy and
environmental responsiveness Seigler and Gatz (1985) reported factors for
philosophical and experiential control. Tyler and Gatz (1979) found that the items of
the Rotter I-E scale yielded different factor structures for males and females In
general, these studies demonstrated only that beliefs about personal control are
multidimensional, but they don’t yield a consistent indication of constructs the
dimensions may represent In response to the mounting evidence against the
89
consistency and generality of personal control, Rotter (1975) acknowledged that one's
personal control may shift depending on situational factors and endorsed the use of
domain-specific assessments. He maintained that individuals may indeed possess a
generalized locus of control and that such an assessment would remain meaningful in
that persons who showed a generalized internal locus of control would be more likely
to show an internal control across a variety of situations and domains. Several
domain-specific personal control theories emerged.
A three dimensional theory of personal control was proposed by Levinson
(1974). Measurements for these dimensions were derived from the Rotter I-E scale.
The measures were designed to represent Internal Control, Control by Chance, and
Control by Powerful Others In the Levinson theory all three of these personal
control dimensions operate within individuals. That is, persons may believe they are
in control of certain outcomes, but, at the same time, acknowledge the operation of
chance and powerful others in their lives. The measures thus may be correlated.
Correlations between Internal control and the two other scales consistently have been
found to be positive, but not significantly different from zero in samples as large as
200. The correlation between the Chance and Powerful Others measures typically has
been positive and significant.
A number of studies have shown that personal control may change over time
and/or vary across situations. For example, Lefcourt (1982) reported that an
individual's control beliefs can be modified in response to accidental happenings (e.g.
90
draft lottery), natural events (e.g., aging, life-events), or through deliberate
interventions (e.g., clinical or educational programs). One may take responsibility for
successful outcomes, but relinquish responsibility for failures (Crandall, Katkovsky, &
Crandall, 1965) One may believe he or she is in complete control of things
intellectually, but have little confidence in their control over health issues.
Wallston and Wallston (1981) constructed the Multidimensional Health Locus
of Control scale, modeled after the Levinson (1974) scale, but aimed at assessing
beliefs about the sources of control over the prevention and cure of illness. They
found three factors of health locus of control including Internal (e.g., “If I take care of
myself, I can avoid illness ”), Chance (e.g., “No matter what I do, if I am going to get
sick, I will get sick ”), and Powerful Others (e.g., “Regarding my health, I can only do
what my doctor tells me to do”).
A three factor locus of control scale, modeled after the three factors of
Levison (1974), was developed to assess personal control specific to intellectual
capabilities (Lachman, Baltes, Nesselroade, & Willis, 1982; Lachman, 1986). The
Internal Control scale in this device measures the extent to which one believes that
they can control their performance in everyday cognitive tasks (e.g., “I could
remember important phone numbers if I practiced them”). The Chance Control scale
assesses the degree to which intellectual decline in later life is seen as inevitable (e.g.,
“My crossword puzzle skills will go downhill, even if I keep doing puzzles”). The
Powerful Others Control scale assesses the extent to which one sees other people as
91
better able to do things and sees oneself as dependent on others to solve cognitive
problems (e.g., “I would have to ask a salesperson how much I would save with a
20% discount”)
The Development of Perceived Control. Of particular interest to the study of
aging is whether or not one's sense of personal control changes over the lifespan. The
point of contention in the evaluation of any developmental research revolves around
the potential methodological problems encountered by both cross-sectional and
longitudinal designs The majority of personal control and aging studies have relied
heavily on unidimensional, generalized personal control assessments and have been
cross-sectional by design. The results of these studies vary considerably. As
summarized by Gatz and Karel (1993), Kogan, (1990), Lachman (1986), and Seigler
and Gatz (1985), results of several studies are different for several reasons. Studies
comparing college-aged, middle aged, and older adults have reported that older adults
demonstrate greater generalized internality, i.e., score higher, on average, on the
Internal Control scale and lower on the Chance and Powerful Other scales (Lachman,
1985; Seigler & Gatz, 1985; Staats, 1974). Other studies have reported age
decreases in generalized internality in adults over the age of 60 (e.g., Cicirelli, 1980).
Still other studies have reported no age differences through middle and old age
(Nehrke, Hulicka, & Morganti, 1980; Reker, Peacock & Wong, 1987). Gatz and
Karel (1993) reported cross-sectional results indicating a decrease in internality from
middle age to old age. Lachman (1986) reported that younger adults show more
92
internality on the domain specific intellectual control scale and older adults score
higher on the intellectual chance and powerful others scales.
It is plausible that an increase in externality that accompanies age may be due
to the effect of the socialization of a particular generation (i.e., an age cohort effect).
Older adults who grew up in the Depression era and witnessed WWII may believe
that they have little control over the circumstances surrounding them. To conclude
from cross-sectional data that increase in externality is a developmental phenomena
might well be inaccurate It may also be possible that history influences men and
women differently and that age-by-sex-by-cohort interaction effects are operating
within cross-sectional samples (Gatz & Karel, 1993).
Results from the Gatz and Karel (1993) study suggest that the contradictory
cross-sectional results in studies of personal control and aging may be attributable to
cohort effects, gender effects, and/or possible cohort X gender effects (e.g.
grandmothers were consistently more external than grandfathers). There were
problems of assessment in both the Lachman (1985) and the Gatz and Karel (1993)
cross-sequential studies in that a measure of personal control consisting of only three
items was used in each. Although each of the assessments used slightly different
items, results from each of these studies must be interpreted with care as the reliability
of measurement and construct validity may have been sacrificed
Although longitudinal studies of personal control have their own problems
(e g , select samples, retest effects), the results from a few repeated measures studies
93
have shown some consistency in interindividual stability and mean decrease in
internality over time when subjects were in their mid-forties to when they were in
their late-sixties (Siegler & Gatz, 1985) and for subjects from age 60 through age 89
(Lachman, 1983). However, in a cross-sequential study of multiple cohorts examined
over a 4-year period (Lachman, 1985), no significant mean changes in average
personal efficacy was found, although the oldest age groups consistently reported the
highest levels of internality
Lachman (1985) reported that higher levels of internality were significantly
related to lower aspirations or ambitions Lachman concluded that these results
suggest that efficacy is lower in the early middle years when individuals are more
likely to experience an overload in roles and demands across family and work careers
At the other end of the life-span, however, when the role crunch eases, efficacy may
increase because life tasks and goals are fewer and, indeed, more manageable” (p
218). This theory is consistent with the James' notion that efficacy may be manifested
in the ratio of the number of successes to the number of pretensions. Older persons
have fewer pretensions and, therefore, fewer opportunities to see those pretensions
fail.
The developmental pattern of personal control is thus an unresolved issue
Cross-sectional, longitudinal, or cross-sequential results do not yet indicate a
consistent pattern. Although the developmental evidence is not clear, it is clear that
the phenomena are multidimensional, and that separate domains must be considered,
94
and that different measures in this domain relate in different ways to different aspects
of personality and cognition.
Personal Control Related to Psychopathology. To what extent are personal
control age differences and/or changes maladaptive'1 Several studies have suggested
that lack of actual control and low belief in internal control are associated with
negative outcomes in old age such as increased morbidity and reduced behavioral
autonomy (Langer, 1983; Baltes & Baltes, 1986) People with a strong sense of
personal control may be more likely to take the actions necessary to remedy
undesirable situations and effect change. On the other hand, it has been suggested
that the ability to modify one's attributions of responsibility when faced with
uncontrollable events may also be adaptive (Schultz, Heckhausen & Locher, 1991).
To adequately understand the potential adaptive nature of beliefs about
control, it may be necessary to study a very comprehensive set of belief systems— that
is, systems that go beyond simply ‘ internal,” “Powerful Others” and “Chance.” For
example, belief in God, an external force, may have different developmental correlates
than belief that a powerful government controls our lives. It has been suggested that
religious persons, who place the responsibility of many life events in the hands of
God, are less psychologically competent than those who do not have such a belief
system.
Pargament and his colleagues (1988, 1990) have examined the difference in
coping styles and psychosocial competence between three groups of individuals: (1)
95
those who report a deferring problem solving style where the individual is passive and
responsibility for dealing with problems is deferred entirely to God; (2) those who
report a self-directing approach to problem solving where God is not considered
responsible for problems or their resolutions; (3) those who report a collaborative
approach to problem solving where the individual and God are though to have an
active, shared role in problem solving. Results indicated that the collaborative style
was indicative of greater coping, personal efficacy, self esteem, tolerance for
ambiguity, and problem solving skills than that of the deferring style. Interestingly,
the collaborative style was found to be more adaptive in terms of a tolerance for
ambiguity and problem solving skills than was the self-directing style It is suggested
by these results that the interaction of self-efficacious beliefs with a belief in God as a
collaborative partner is nnore adaptive than self-efficacy or belief in God alone.
Two points should be taken from the research done by Pargament. First, the
Pargament results make it evident that there are external belief systems that are not
well represented by the simple concepts of Powerful Other and Chance measures of
previous studies. The results suggest that personal control involves many belief
systems, including systems of different beliefs about God. It is possible that some
individuals believe that nothing or no one is in control. The Pargament study also
suggests that more study is needed of how internal and external belief systems interact
to facilitate sound psychological functioning and adaptive coping.
Evidence for an Eight Factor Theory. Building on previous research, as just
reviewed, Noll and Horn (1990) put together a number o f items to assess different
aspects of beliefs about control and analyzed the relations among responses to these
items in a sample of 242 college students. The aim was to identify how people
conceptualize control in their lives. A factor solution indicated eight reliable and
interpretable dimensions. The factors indicated statistically independent (though
correlated) patterns of belief. These can be described briefly as follows (All eight
factors had alpha reliabilities above .70): 1) Personal God: Belief that there is a God
who is in control of one’s life and the lives of others. 2) Luck: Belief that chance, or
fortune, controls much of ones own life and the lives o f others. 3) Superstition.
Belief that supernatural powers, omens, horoscopes, etc., represent determinants of
what happens in life. 4) Powerful Others: Similar to Levinson's construct, belief that
other people in powerful positions influence one's life. 5) Self Control of Health:
Similar to the Wallston and Wallston construct, belief that to a considerable extent
one determines one's own health. 6) Nihilism/Chaos: Belief that there is no particular
force is in control of individual lives. Life is not governed by anything identifiable. 7)
Self Efficacy: Similar to Rotter's Internal Locus of Control, belief that for the most
part one can control one's own existence. 8) Science: Belief in Science as a way of
"knowing" there is a basis for controlling major aspects of one’s own life.
The eight personal control factors were found to be metrically invariant (Horn
& McArdle, 1992) across gender (Noll, 1993), that is, the same factor pattern
97
represented the same factors in both groups. There were significant sex differences in
the means of several factors Females scored significantly higher on the God and
Powerful Others factors; males scored significantly higher on the Nihilism/Chaos
factor.
Perceived Control and Cognitive Functioning. Several theorists have
suggested that low perceived self efficacy regarding intellectual capacity can have
detrimental effects on cognitive functioning, including poor performance on tasks
assessing intelligence, memory, and problem solving (Bandura, 1989; Hertzog, Dixon,
& Hultsch, 1990) In the few studies comparing cognitive capabilities and personal
control the Lachman, Baltes, Nesselroade, and Willis (1982) Personality in
Intellectual-Aging Contexts (PIC) instrument has been used to assess beliefs about
intellectual controls
Using this device in a 5-year longitudinal study of adults in their 70’s Lachman
& Leff (1989) demonstrated that mean for generalized control did not change
significantly, but there was change in the direction of increased dependency on others
to carry out cognitive tasks (Powerful Others intellectual control scale). Low scores
on a Letter Series task (a measure of Gf) accounted for a significant amount of the
increase in Powerful Other scores. Higher levels of education were associated with
decreases in internal intellectual control
98
Cornelius and Caspi (1986) found that the best predictor of psychometric
intelligence (measures of Gf) was the Powerful Others dimension of the PIC: The
higher the score on this dimension, the lower the Gf scores.
Grover and Hertzog (1991) demonstrated that scores on the internal (INT)
dimension of the PIC scale significantly declined with age, and both the powerful
others (POW) and chance (CHA) dimensions increased from age 20 to 85 in 2 cross-
sectional samples. The INT scale was positively correlated with all intellectual
abilities tested and the POW and CHA scales were negatively related to cognitive
performance on all tests. Results indicated that younger subjects were significantly
more internal on the Perceptual Speed (speed ability), Induction (fluid ability), and
Spatial Visualization (visual ability) tasks than older adults. Thus, vulnerable abilities,
those abilities which have been shown to decline with age such as fluid ability,
visualization, and speed, are highly related to external control beliefs as measured by
an intellectual domain-specific control scale.
It appears from the results of previous research that a belief in one's
intellectual efficacy facilitates sound cognitive functioning and that external
intellectual control beliefs may be maladaptive in the area of cognitive functioning.
However, it is not clear whether the decreased ability levels of older adults cause a
decreased sense of intellectual efficacy or if a decrease in intellectual efficacy causes a
decrease in intellectual abilities. Probably both are true. Longitudinal studies that
have been designed to help understand this cause and effect relationship have been
99
plagued by confounding test-retest effects, unreliability of measures, and low sample
sizes Experience with a cognitive task will cause changes in performance, and beliefs
about controls— i.e., enduring self-efficacy beliefs about sound performance, and
external efficacy beliefs about low performance (Grover and Hertzog, 1991; Hertzog,
Dixon, & Hultsch, 1990; Lachman, Steinberg, & Trotter, 1987). What is clear,
however, is that personal control beliefs are multidimensional, domain specific, and
fluctuate with the onset of positive and negative life events.
In sum, personal control beliefs are a multivariate phenomena that, to some
extent, are specific to particular domains o f performance— cognitive, social, physical
Such beliefs change with the onset of positive or negative life events. It is not clear
whether older individuals are more external than younger adults-in general or in
specific domains It is not certain whether or not people change with age and
experience toward attributing causes to external factors, but it is logical that they do
and these findings tend to support this logic. It is notjdear whether increases in
externality with age are attributable to cohort influences, and/or other factors. It
appears that there is a broad, perhaps general, factor of belief in external (versus
internal) control and that this general externality is correlated with low cognitive
functioning in old age-i.e., in samples of people over the age of 70 years. The
evidence suggests, too, and the domain specific intellectual internality is associated
with sound cognitive functioning in samples of both young and old subjects. It is
debatable, particularly for older subjects, whether or not externality (in general)
100
should be viewed as psychologically maladaptive. Adopting certain external belief
systems may facilitate coping, particularly in situations where the loss of personal
control is inevitable.
Expectations (Hypotheses)
As was noted earlier, theory stipulates that loss in cognitive functioning is due
to a loss of neurons, but when there is substantial neuronal overdetermination there
can be notable neural loss without corresponding loss of ability. Maintained abilities
(crystallized knowledge and long-term retrieval) are hypothesized to be abilities that
are neurologically overdetermined. Vulnerable abilities (fluid reasoning, short-term
apprehension retrieval, concentration, cognitive speed, visualization, etc ) are thought
to be neurologically underdetermined and notably weakened with even relatively small
amounts of neuronal loss. Losses in underdetermined networks may not be detected
because changes are relatively small. Detection requires high reliability of
measurement, large sample sizes, and/or extended periods over which (irreversible)
decrements can occur These conditions often have been lacking in prior research,
particularly longitudinal research (Horn & Donaldson, 1980) Neural losses
associated with explicit brain damage are accompanied by notable behavior and
intellectual changes (Horn, 1968, 1979). In case studies of brain damaged patients,
maintained ability functioning often “springs” back to very nearly pre-injury levels,
while losses o f vulnerable abilities persist. Such evidence supports theory that
vulnerable abilities, particularly fluid reasoning abilities (and processes) are largely
101
tied to underdetermined neurological functions and that maintained abilities,
particularly crystallized knowledge, are largely tied to experiential, acculturational, or
learned overdetermined neurological functions (Carroll, 1993; Cattell, 1963; Eysenck,
1982; Horn, 1988; Horn & Hofer, 1992; Horn & Noll, 1994; McArdle, Horn, &
Goldsmith, 1984).
It is expected that different aspects o f lifestyle will be related to G f and Gc in
different ways. Lifestyle variables are expected to be indicative of different
influences-differing environmental stimulation, physiological well-being, motivational
structures, etc. Lifestyles that are indicative o f overall physical health are expected to
be positively related to maintenance of vulnerable abilities because healthy lifestyles
maintain the integrity of underdetermined neuronal networks. The more exposure
there is to neurological trauma, the greater the deficit expected for vulnerable
abilities. Lifestyles that involve immersion in culture are expected to be positively
related to enhancements of maintained abilities. The more stimulating the
environment, the greater the acculturational potential the higher maintained abilities
are expected to be.
Although it has been assumed that individual differences in cognitive
functioning can, to some extent, be understood in terms o f individual differences in
psychoneurological functioning, direct links are difficult to forge and much theory is
based on extrapolation from nonhuman animal studies, clinical case studies, and
plausibility Fodor (1983) for example, has suggested that the brain is organized into
independent “modular” systems. Carroll, (1985) has suggested that such systems
might correspond to broad factors of cognitive abilities. Extant evidence for such a
theory is sparse. Part of the problem with evidence that is has been based on a
measurement model that does not represent modality theory-namely the “g” factor
model that makes no distinction between vulnerable and maintained abilities (Eysenck,
1982, 1988, Eysenck & Bartlett, 1985, Hendrickson, 1982). In these studies,
overdetermined and underdetermined abilities (of theory) are confounded, so the
extent to which these different classes of abilities might represent different modalities
cannot be studied with any precision. The few studies that have been guided by
multi-factor ability theories have provided support for hypotheses of overdetermined
and underdetermined abilities and for hypotheses that declines in underdetermined
abilities are particularly linked to loss of function in the hypothalamus, mammillary
bodies, fornix, and temporal lobe areas of the brain (see Horn, 1985 and Carroll, 1993
for a more complete review). The suggestion is, too, that low scorers have to “work
harder” to perform on tests of inductive reasoning, spatial ability, and attentional
processes than do “high scorers” (Haier, et a l, 1988).
Cattell (1963) has proposed an investment hypothesis specifying order of
development of Gf and Gc Gf is hypothesized to be an investment in the
development of Gc. Basically, the thought is that Gf represents inherited potential
that is, to greater and lesser degrees in different individuals, invested along with
experiences in producing the knowledge that becomes Gc. Thus, Gc is thought to
103
develop out of Gf as individual differences in environmental influences accumulate
While plausible, the theory has not been supported by such evidence thus far gathered
(albeit there is only a small amount of evidence available). Schmidt & Crano (1974)
tested a model specifying that Gf caused Gc but Gc did not cause Gf, consistently
within interpretation of investment theory. Initial analyses seemed to support the
theory, but critical evaluation of these analyses and subsequent analyses suggested
lack of support. Horn (1985) found that in samples of four-year old children Gf and
Gc were clearly distinguished and no more highly correlated than in samples of adults,
contrary to an hypothesis derived from investment theory that early in development
Gf and Gc would not be distinguished or would be highly correlated Other evidence
has been likewise equivocal or not supportive of investment theory For example, the
theory suggests that heritability is higher for Gf than for Gc, but results from several
studies of heritability (reviewed by Horn, 1985) suggest either no differences in the
heritabilities of Gf and Gc, or, if anything, higher heritability for Gc (McArdle, Horn,
& Goldsmith, 1984)
It is possible that both genetics (manifested in neurological functioning) and
environment (learning and acculturation) play a role in vulnerable and maintained
abilities. Investments may come from either Gc or Gf. Indeed, Gf and Gc might be
related through a kind of spiral of influences of one by the other. An initial
investment in Gc, for example, might stimulate enhancement of Gf which, in turn,
drives Gc development, which then affects Gf, and so on. Such a spiral would
104
operate throughout development, perhaps with different emphases at different times.
For example, in early childhood, there might be emphasis the development of Gc.
The knowledge accumulated in this period might then be invested in maintenance of
Gf in adulthood.
Individual differences in the interactive influences of different abilities, as
suggested by the possible influences of Gf and Gc on each other, will inevitably
emerge partly in consequence of individual differences in culture, resources,
economics, familial structure, etc. It is very difficult to separate such influences and
to distinguish those coming primarily from genetic factors and those coming from the
experiential environmental influences. In such a system, ability levels would be
biological and environmental conglomerates of all previous Gf-Gc interactions.
It is plausible that a lifestyle indicative of sound physiological health (good
dietary practices, high levels of exercise, low stress levels, little brain trauma) would
benefit Gf to a greater extent than Gc. But the knowledge developed in Gc can alert
one to the need for good health maintenance leading to one devising ways to
accomplish this thus promoting Gf. Thus, Gc can be an investment in the
maintenance of Gf. A healthy lifestyle might be indicative of both Gf and Gc, but for
different reasons Similarly, a lifestyle that includes heavy drinking behavior may be
manifested in neurological damage ultimately affecting Gf functioning, but heavy
drinking and alcohol abuse may also be indicative of poor education and a lack of
acculturation, and thus be associated primarily (negatively) with Gc.
105
There are many other reasons why lifestyle indicators can be related to both
Gf and Gc. As was pointed out in Study 1, the measurements of these abilities are
not fully separate and independent. Gc characteristics, such as a familiarity with tests
or the application of puzzle solving strategies learned in school, affect Gf
performance. Likewise, Gf reasoning can aid in solving problems that are
acculturational in nature Lifestyles that relate to Gc will have some correlation with
Gf because measures of Gf are not entirely operationally independent of Gc measures
and vice verse
Lifestyles that indicate motivation— being “up” for intellectual tasks— will boost
both G f and Gc performance. Measures of self esteem, lack of depression, and social
involvement are indicative of such motivation Thus, high scores on these indicators
will predict both Gf and Gc scores Subjects who endorse these lifestyles may be
more apt to try harder and work longer on cognitive tasks than those who do not.
In general, lifestyles that are expected to be indicative of the decline of
vulnerable abilities because they indicate neurological damage may also be related to
maintained abilities. Similarly, lifestyles that are expected to be positively related to
the preservation of maintained abilities may to some extent be related to Gf as well.
Methods
Subjects
The sample of 577 adults ranging in age from 22 to 92 that was described in
Study 1 was studied here.
106
Measures
The measures comprising the cognitive battery described in Study 1 were
utilized in the present study. The Power Letter Series (PLS) test was used to
measure Fluid Reasoning (Gf), the second Vocabulary Test 2 (VOC2) was selected to
estimate Crystallized Knowledge (Gc), the Memory for Paired Associates (MPA) test
was used to measure Short-term Apprehension Retrieval, the two slow tracing trials
were combined as a reliable measure of Concentration (CON) ability, and the
Speeded Letter Comparison (SLC) task was selected to estimate Cognitive Speed
As stated earlier, the present study did not propose to confirm a theory of
lifestyle Rather, particular lifestyle constructs were selected from the broad,
theoretically derived battery o f a larger study for use in the present study. Along with
numerous demographic information (age, sex, income, education level, marital status,
time in stable relationship, number of children, etc.), a manageable subset o f scales
was selected from the larger battery in attempt to represent broad domains of life
goals: those for attaining and maintaining Physical Health, those for self realization
through Activities, those for attaining and maintaining Psychological Health, and
those pertaining to Personal Control.
Lifestyle measures were developed from items designed to indicate particular
constructs. Factor analyses were conducted to test these design hypotheses. Where
hypotheses were supported, items were combined into linear composites. The
107
internal consistencies, means, sigmas, and intercorrelations were determined. In most
cases the composites contained 5-15 items Occasionally, when full, multiple-item
scales were either not available, or not appropriate, one or two-item measures were
included in the battery The items included in each measure are reported in Appendix
B
Physical Health Measures. To assess the extent to which subjects’ attend to
a healthy diet, questions pertaining to the frequency o f engagement in a variety of
healthy dietary practices were included a scale titled Attention to a Healthy Diet.
Subjects w ere asked to indicate the frequency with which they eat healthy foods
(fresh fruit and vegetables, low fat items, high fiber), consciously avoid unhealthy
foods (salt, fat, large amounts of red meat, caffeine), and attempt to eat balanced, low
calorie meals Subjects w ere allowed to check anywhere along a continuum to
indicate the frequency with which they engage in the identified behavior,
never yearly monthly weekly daily
Possible scores ranged from one to nine for each item included in the scale.
Subjects received a score o f one if they checked “never”, two if they checked
somewhere in-between -‘never” and “yearly”, a three if they checked “yearly”, and so
on The highest score o f nine was given if “daily” was checked. The scale was made
up of 11 items and possible scores ranged from 11 to 99. The internal consistency of
the composite was 87 The structure o f this scale was previously confirmed in two
pilot studies (Noll& Horn, 1990; 1994)
A simple linear composite was used to assess Smoking Behavior. In the first
item subjects indicated the frequency with which they presently smoke cigarettes
(never, yearly, monthly, weekly, daily) and the second item inquired whether, and for
how many years, the subject had been a regular smoker (never, 1-3 years, 4-9 years,
10-19 years, 20+ years). Again, subjects were allowed to check anywhere along the
continuum and possible scores ranged from 1-9 or each item. The two items were
combined and the resulting alpha reliability for the two-item indicator of Smoking
Behavior is .72.
Alcohol use and potential abuse were also assessed. Per Week Drinking
Frequency assesses how many days in a ‘typical’ week subjects drink alcohol (none,
less than once per week, 1 or 2 times per week, 3 or 4 times per week, 5 or more).
Subjects are also asked to indicate the Average Number o f Drinks Consumed on Each
Single Occasion (none, 1 , 2, 3 or 4, 5+). Next, subjects were asked to indicated how
many times, over their lifetime, they had Passed Out from Drinking (never, once,
twice, 3 or 4 times, 5+). A linear composite measure from the “Alcohol Use
Inventory” (Horn, Wanberg, & Foster, 1980) assessing the extent to which one drinks
alcohol to escape, get over feeling depressed, forget problems, or shut out worries
was assessed. This Use o f Alcohol to Escape measure had a reliability of .89.
Subjects were asked to answer in terms of their present drinking behavior or to think
retrospectively if they had been a drinker in the past, but had stopped drinking
presently
109
Activity Measures. The Information Seeking scale consists of 14 items
requiring subjects to estimate the frequency with which they engage in acculturational
activities including; attending museums, reading nonfiction (magazines, newspapers,
books), keeping up with current events (read, watch or listen to news, discuss events
with others, vote in elections), attending shows of the performing arts (dance, drama,
music), and watching television for acculturational purposes (PBS, documentaries,
news).
The Passive Entertainment scale measures the extent of one’s passive, as
opposed to active, entertainment activities. The person who scores high on this 7-
item scale watches a lot of TV, uses TV for entertainment rather than for learning
(watches sit-coms, dramas, movies), and will watch just about anything on TV
however poor the quality. The assumption (hypotheses) was that time spend in this
kind of TV watching reduced time spent in acculturational activities that would
promote Gc and mind-exercising activities that would maintain Gf.
The Negative Attitude Toward 'IV scale indicates one’s attitude toward
television, in general, and a critical evaluation of how TV is used and abused by
American society. A high score on this factor reflects a general disdain for television,
a feeling that Americans, particularly children, watch too much TV, and the opinion
that TV keeps people from doing more worthwhile, creative activities. It was
assumed that a negative, critical attitude about TV would (indirectly) indicate a
110
predisposition to read and engage in culturally enriching activities that would promote
Gc and exercise thinking that would maintain Gf.
The Community Involvement factor indicates the frequency of community-
oriented activities (clubs, neighborhood watch programs, PTA, church, civic or
political groups), the general concern for the well-being of the community, and
amount of investment (social and political) made in the community. This 10-item
scale has an alpha reliability of .78 The major hypothesis represented by this scale is
that individuals actively involved in the community will, in such involvement, exercise
the Gf functions and acquire and maintain the knowledge of Gc.
Psychological Health Measures. Self Esteem was assessed using the “What I
am Like” scale for adults (Harter, 1985). Here, subjects indicate the extent to which
they are happy with their lives, feel they are worthwhile, and see themselves as
intellectually and socially capable. The alpha reliability of this scale was .84.
Optimism is a sub-scale of the “Inventory of Psychosocial Balance” (Domino
& Alfonso, 1989). The 9 items included in this scale assess the mindset that life has
meaning and purpose, there are worthwhile people and activities, and that one is
inspired by life, new situations, and one’s own contribution and productivity. The
scale was found to have an alpha reliability of .73.
The Secure Attachment measure is a sub-scale of the “Revised Adult
Attachment Scale” (Collins & Read, 1990). High scores on this 6 -item scale indicate
I l l
that one is secure in interpersonal relationships, able to get close to others, and is
generally trusting. The common factor has an alpha reliability of .8 6 .
High scores on the Confide in Others scale indicate a willingness to utilize
available social support systems including; counseling, talking with friends/relatives,
seeking out confidants, and communicating feelings to others The measure had an
alpha o f 8 8
The ability to be flexible, non-rigid, and tolerant of misbehavior and untidiness
is assessed by the 11-item Tolerance fo r Disorder scale. The questions allow
respondents to indicate their belief that a home should be spotlessly clean, that
children should obey, stay clean, be orderly, listen to adults, and never talk back. The
set of items for this factor had an alpha reliability o f .89. A high score represents a
high flexibility or tolerance.
Personal Control Measures. Five of the eight factors from the Noll & Horn
(1990) personal control scale were assessed A Belief in God (12 items) allow
subjects to endorse a belief in a Judeo-Christian, personal God belief system. High
scores reflect a belief that God is all-knowing, in control of what happens in life, can
be accessed by each person individually, cares and comforts individuals, and provides
answers about the meaning of life. This scale’s internal consistency reliability is .94.
A Belief in Nihilism/Chaos (5 items) reflects subjects’ feelings that nothing or
no one is in control and that one is not accountable to others. The alpha reliability of
the scale is 77.
112
The Belief in Luck (13 items) scale is similar to the scale developed in the
Levinson (1985) work. Subjects endorse statements indicating belief that luck plays a
large role in the success and failure of life endeavors. Its alpha reliability was found
to be .77
The Belief in Powerful Others scale is also like Levinson’s (1985) powerful
others construct, but it contains additional, less ambiguous items. A high score on
this 7-item scale reflects a general sense that other people are responsible for
successes and failures— other people influence thinking and decisions, help one get
over being depressed, and can get in the way of successes. The scale has an alpha
reliability of 77
Belief in Superstition ( 8 items) is a scale that assesses one’s belief in omens,
charms, numerology, horoscopes, astrological signs, supernatural powers, and
psychics It has an alpha reliability of .77.
Analyses
The procedures for analyses of data in this research are in several respects
different from those most often used in behavioral science research. Description and
statement of the rationale for these procedures therefore is offered in the following
sections
Missing Data Imputation A within-factor, multiple imputation missing data
procedure was used to impute the values of items that were missing at random. The
multiple imputation, EM-algorithm procedure developed by Graham and Hofer
113
(1993) allows the estimation of values of missing items by predicting those scores
from items that are present. The program utilizes information from all variables to
predict missing values via multiple regression. However, estimates resulting from
such procedures can be biased if irrelevant information is used to estimate missingness
(McArdle, 1994). To avoid bias, variables that are validly related to a variable for
which there is missingness should be used to estimate missingness. Ideally, and in the
present study, variables defining the same common factor are used. If the missingness
can be estimated from items that load on the same factor as the missing variable, and
if the factor is highly reliable (has good internal consistency), then the bias is
somewhat controlled.
Missing values were imputed using only variables of the same common factor
under assumption that the missingness was at random or completely as random
(Graham & Hofer, 1993) The at random assumption implies that the missingness
was not produced by (selected; dependent upon) the variables used to estimate the
missingness The assumption of missing completely at random has this same
implication plus an implication that the missingness is not produced by the variable for
which missing values are to be imputed— the system for the missingness was not based
on the missing variable. One cannot know for sure that these assumptions were
absolutely warranted in the present data, but the fact that only variables of a common
factor were used to estimate the missing data, the fact that the amount of missingness
was small-there was never a case when the missingness exceeded 3% of the total
114
data~and the fact that missing responses appeared to occur only randomly, suggests
that it is reasonable to suppose that the assumption of randomness is warranted. The
structure of the factors, the internal consistency of the factors, or the correlations
among factors did not change significantly before and after imputation. Full data
were available for analyses with 577 subjects.
Measurement Invariance. Whenever it is important to ask whether or not two
groups differ significantly on a construct, it is important to know whether or not the
same construct is being measured in those groups (Horn & McArdle, 1992). If an
invariant factor pattern can be shown to exits for groups being compared, there is
evidence that, indeed, the same scales of measurement are being used across those
groups Metric invariance requires that the factor loadings be precisely proportional
(or equal) across groups That is, if ai, a2, a3 ,. .., am are the loadings on a factor in one
group, and bi, b2, b?, ,bm are the loadings in another group, then metric invariance
obtains if (and only if) ai=kbi, a2=kb2 , as=kb5 , ..., am=kbm , when k is a constant. A
finding that metric invariance obtains indicates that factor scores can be obtained in
the same way in different groups. If metric invariance does not obtain, the factor
scores in calculated in different groups are different. One gets apples in one group
and oranges in another Comparisons of the means, variances, and correlations for
such different variables are dubious. Metric invariance tests along with tests that the
factor variances and covariances and item uniquenesses are invariant were performed
on the lifestyle factors assessed.
115
Part and Partial Correlational Procedures. The part correlational
procedures described in Study 1 will be used to provide evidence bearing on the
principal hypotheses of this study. The procedure entails removing the variance of a
variable from another and assessing how the residual— the part left over— relates to a
third variable or to several other variables (Horn, et al., 1981). The procedure may be
described as controlling (but only mathematically, statistically, not experimentally) for
the influences of the variable that is parted out The procedure simulates
experimental control and makes sense only under conditions where experimental
control would make sense
Part correlational analysis should be distinguished from partial correlational
analysis, in which the variance associated with a particular variable is removed in both
of two other variables, and the residuals o f these two variables are correlated. This is
reasonable only if it makes sense to control for an influence in both of two variables.
As Horn (1965) reasoned, it is sensible to suppose that one could control the amount
of sugar in the blood by parting out a measure of blood sugar in analyses of
relationships between liver function and, say, a surgical technique or the age of the
patient. It probably would not be sensible to part out blood sugar from the surgical
technique or age, however, for these are not variables in which sugar could be
controlled experimentally Part correlational analysis is a way of simulating control in
variables and under conditions in which it is logical. It is misleading to use partialling
116
analysis to remove associations under conditions in which it is not logical to suppose
that the parted-out variable could be a control variable.
In the present study it often makes sense to control a lifestyle influence in
analysis of a cognitive variable, but not to suppose that the same influence could stop
time (age) or change gender. For example, it is reasonable to suppose that a lifestyle
variable such as one’s attention to a healthy diet is associated with physiological
health and such health operates to effect sound cognitive functioning: the better your
dietary habits, the better your neuronal networks and the better your associated
vulnerable abilities. To remove the variance o f dietary practices that is associated
with cognitive speed, for example, is to simulate control of nutrients, physiological
functioning, etc., in the development and expression of cognitive speed. One can
interpret a correlation between the resulting residual and age as representing what
would have been the developmental consequences of not having a healthy diet To
use partial correlation in this case would also be to simulate control of a healthy diet
in age (representing development).
One approach to the study of the development of cognitive abilities is to study
the cognitive processes known to be involved abilities. In Study 1 it was
demonstrated that short-term apprehension retrieval (SAR), concentration (CON),
and cognitive speediness (Gs) are processes of fluid ability and that SAR is also a
process of crystallized knowledge. Lifestyle variables that help us understand these
117
cognitive processes and their age relationships thus help us understand the
intelligences of which these processes are a part.
To provide information about these processes the residual part of SAR, CON,
or Gs after removal of lifestyle variance will be correlated with age The difference
between the age correlations before and after parting will be calculated to indicate the
relative magnitude o f the lifestyle variable effect. If the R2 before control is
significantly different form the R2 after control, then it can be concluded that the
control variable has a significant effect on the age relationship (see Study 1 for a
discussion of the significance test for this procedure).
This procedure can also be used to understand the residual part o f Gf that is
not associated with cognitive processes SAR, CON or Gs (and Gc measurement
overlap) and the residual part of Gc that is not associated with the cognitive process
SAR (and Gf measurement overlap). To accomplish this, new residual variables were
calculated: rGf— fluid reasoning void of process variable variance and Gc variance,
and rGc— crystallized knowledge void of SAR variance and G f variance.
Understanding the Portion o f Ability Variance that is not Associated with
Aging. It is possible that a lifestyle variable is correlated with cognition only because
it and the cognitive variable are correlated with age. It is also possible that a lifestyle
variable appears to be negatively related to cognition simply because the lifestyle
variable has a strong positive association with age, but, when age is taken into
118
account, via part or partialling procedures, the relationship between the lifestyle
variable and cognition is as positive
The procedures for doing this are in principle the same as those described in
Study 1 and for the part correlational procedures described above, but with a
cognitive variable as the dependent variable. If the R2 for equation 2.1 is significantly
different from the R2 for the equation 2.2, then it can be concluded that the lifestyle
variable accounts for a significant part of the variance of the cognitive variable that is
not explained by age (G = the ability criterion variable, A = age, and L = a lifestyle
variable).
G=bi A with R2 g.a [2 .1 ]
G=bi A + b2 L with R2g.al [2.2]
The focus in these kinds of analyses is not on development (age variance is
removed), but on the behavioral processes as they are involved in cognitive
functioning at any age. These analyses thus aim at understanding of the nature of
cognitive ability , as such.
Results
Factor Analyses and Measurement Invariance
The lifestyle measures chosen for assessment in this study consisted of 14S
items designed to measure 16 common factors. The items were intercorrelated, 16
factors were calculated in accordance with the criteria of a principal factor model, and
these factors were related by oblique Procrustes (Hurley & Cattell, 1962) to the
119
positions specified by the item construction hypotheses. A target factor structure
matrix, T, is specified in which hypothesized coefficients are nonzero and all other
coefficients are zero. A rotation of the data factor structure matrix, A, is selected
such that the sum of the squares of the difference between T and A— i.e., (T-A)2 ~is a
minimum. The factor intercorrelations are not constrained to be zero. Unlike
confirmatory factor model fitting, as by LISREL (Joreskog & Sorbom, 1994), the
Procrustes procedure does not require that all hyperplane loadings be exactly zero.
The procedure forces the sum of squares of positive and negative loadings to be zero,
but under such conditions hyperplane loadings are not, in general, zero.
The Procrustes model is diagnostic. It indicates the extent to which the
factors can be estimated in accordance with the target hypotheses. If substantial
hyperplane loadings result, and the correlations among factors are large, there is
indication that the target does not provide a good representation of the data. This
model, and its diagnostic potential, is particularly useful when measures from several
different domains are evaluated. The procedure provides evidence of whether or not
factors are distinct from one another (structurally independent).
The Procrustes solution for 16 factors among 145 items is provided in
Appendix C. The loadings hypothesized, a priori, to be salient are underlined. Any
hyperplane loadings above . 2 0 0 and any hyperplane loadings that are higher than
hypothesized salient loadings are shown in bold. It can be seen that there are very
few such loadings. The following cases required action, however. Two items in the
120
Information Seeking factor (factor 1) have salient loadings on two factors. These
items neither added to the internal consistency of other factors nor took away from
the internal consistency of the Information Seeking factor They were retained,
therefore, in the Information Seeking scale.
The item that loads the lowest on the Self Esteem factor (factor 8 ) also loads
on the Powerful Others factor. Also, there is one Optimism (factor 7) item that loads
on other factors. These items do not decrease the internal consistency of the factor
on which they were designed to load and do not add to the internal consistency of any
factor on which they were not hypothesized to load. Therefore, these items too, were
retained as indicators of the original a priori factors Retaining these items elevated
the factor intercorrelations slightly (at the fourth decimal place), but not significantly.
The factor intercorrelations are reported in Table 2.1. The salient items for
each factor, the Procrustes loadings for each of these items, and the factor internal
consistency reliabilities are reported in Appendix B. The factor intercorrelations are
sufficiently low— correlations always lower than internal consistencies— and hyperplane
loadings are sufficiently low to suggest that the 16 factors are reliable, and reliably
independent from one another.
To test hypotheses of metric invariance, and thus measurement invariance of
the factors, the sample of 577 subjects was split into three age groups as follows:
Gpl
Gp2
Gp3
Interval Age Span N Stage
22-45 24 150 Enhancement
46-68 23 225 Plateau
69-92 24 2 0 2 Decline
Table 2 1
Intercorrelations Among Lifestyle Factors (N=577)______________________________________________________________
1 2 3 4 5 6 7 8 9 10 11
1 Attention to Healthy Diet 1 0 0
2 .Use of Alcohol for Escape -.03 1 . 0 0
3. Information Seeking .25** .0 1 1 . 0 0
4. Passive Entertainment 03 .07 .15** 1 . 0 0
5. Negative Attitude Toward TV 1 0 * 0 1 -.07 -.1 2 ** 1 . 0 0
6 . Community Involvement 19* . 0 2 .32** .03 -.15 1 0 0
7. Optimism .2 2 ** -.1 1 ** .09* -.03 .1 0 * .1 2 ** 1 . 0 0
8 . Self Esteem .14** - 1 1 ** .13** -.05 .05 .18** .40** 1 . 0 0
9. Confide in Others .08* 05 1 1 ** 07 0 1 .1 0 *
23**
.04 1 0 0
10. Tolerance for Disorder . 0 2 . 0 2 .08* -.03 - . 0 2 -. 0 1 .2 0 ** 06 2 0 ** 1 . 0 0
11 Secure Attachment 1 1 ** -.13** .05 -.09* 0 1 .2 1 ** .25** .32** - 0 2 05 1.00
12 .Belief in God .1 2 ** - . 0 2 -.03 .07 -.04 .16** .06 -.04 .03 -.18** - . 0 1
13 Belief in Nihilism/Chaos -03 .13** - . 0 2 .07 -.06 - . 0 1 -.32** -.15** - . 0 1 -.13** -.17**
14 Belief in Luck -.09* .14** .04 1 1 ** -.07 .07 -.45** -.31** -.17** -.13** -.18**
1 S.Belif in Powerful Others -.07 .2 2 ** - . 0 1 .15** -. 0 1 .07
_ 42**
-.40 -.04 -.20** -.27**
16.Belief in Superstition .06 .14** .0 1 .09* -.06 .08* -.2 1 ** -.23** .06 0 1 -.1 0 *
1 2 13 14 15 16
12.Belief in God 1 0 0
13. Belief in Nihilism/Chaos -.18** 1 0 0
14.Belief in Luck -.07 .53** 1 0 0
15.Belif in Powerful Others .06 .43** .63** 1 . 0 0
16.Belief in Superstition .1 0 * .37** .42**
4 7 **
1 . 0 0
* = Significant correlation at p<05
** = Significant correlation at p< 0 1
122
The age groups were formed in a manner to ensure that the N for all groups was at
least 150— i.e., more than the number of items being analyzed. It was required, also,
that groups span approximately equal numbers of years. This made the age variance
approximately equal within groups, and thus provided a base against which to
compare variances for other variables For example, if age groups span equal age
intervals, and if it is found that the factor variances are different for these groups, then
it can be reasonably assumed that these difference are not due to different spans of
age— for a group spanning 1 0 years, the variances for variables related to age would
tend to show lower variances than a group for which the age span was 15 or 2 0 years
The age grouping roughly represents stages in the development of abilities in
adulthood That is, results from previous studies suggest that enhancement in Gc is
steady until around age 45 (end age for groupl); from age 45 to the late 60’s there is
a plateau, (the period of group 2); and some findings (Baltes & Schaie, 1977) suggest
that there might be some decline in Gc in the post-age-70 period (the period of group
3) Items were converted to standard scores (mean = 0, std = 1) across the entire
sample (N=577) Then separate covariance matrices were computed within the three
groups. Multiple-group modeling analyses (using the LISREL7 program; Joreskog
& Sorbom, 1988) were carried out to test for factor (measurement) invariance across
the three age groups.
To accommodate the program restriction and keep the problem tractable, the
analyses were done separately for three sets of lifestyle variables. The first set
123
included the physical health and activity factors (Diet, Use of Alcohol to Escape,
Negative Attitude Toward TV, Information Seeking, Passive Entertainment, and
Community Involvement). The second set included the psychological health factors
(Confide in Others, Tolerance for Disorder, Self Esteem, and Optimism). The third
set included factors from the personal control domain (God Belief, Nihilism/Chaos,
Luck, Superstition, and Powerful Others control). The Procrustes analyses has
demonstrated that these groups of factors are largely independent.
Table 2 .2 provides a summary of the test statistics of measurement invariance
namely; chi-square (x2 ), Goodness of fit index (GFI), degrees of freedom (DF), ratio
of chi-square to degrees of freedom (x2/DF), and root mean square of the residuals
(RMSR) The first row for each set in the table provides results for configural
invariance: the zeros (hyperplane loadings) are required to be invariant (zero) across
the three age groups, but the nonzero factor pattern coefficients, the factor variances
and covariances, and item uniqueness are allowed to vary across groups.
The first row under Alternative Models in each set provides the test statistics
for metric invariance: the coefficients in each column of the pattern matrix are
required to be proportional across all groups, but the factor variances, factor
intercorrelations (covariances), and item uniquenesses are allowed to vary. The last-
mentioned conditions are systematically constrained in the tests of remaining rows in
each set in the table.
124
Table 2.2. Measurement Invariance Tests fo r Lifestyle Factors
Set 1: Physical Health and Activities (56 Variables: 6 Factors)
BASE MODEL: GEI DF y2 /DF
Configural Invariance Test .89 6940 4401 1.58
ALTERNATIVE MODELS:
Metric Invariance--LY=IN
+Factor Variances Equal Across
Groups—
LY=IN, PS diagonal = across
groups
+Factor Covariances Equal Across
Groups-
LY=IN, PS=IN
+Item Uniquenesses Equal Across
Groups—
LY=IN, PS=IN, TE=IN
X2 A from
Base Model
119
145
307*’ '
1004**
DF A from
Base Model y2A\DFA
106
118
148
260
1.12
1.22
2.07
3.86
Set 2: Psychological Health 144 Variables: 5 Factors!
BASE MODEL: GFI
Configural Invariance Test .85
ALTERNATIVE MODELS:
Metric Invariance— LY=IN
+Factor Variances Equal Across
Groups-
LY=IN, PS diagonal = across
groups
+Factor Covariances Equal Across
Groups—
LY=IN, PS=IN
+Item Uniquenesses Equal Across
Groups-
LY=IN, PS=IN, TE=IN
X ?
5324
X 2 A from
Base Model
105
124
157*"
1028**
DF
2627
DF A from
Base Model
78
88
108
196
XVDF
1.98
v A\DFA
1.35
1.41
1.45
5.24
RMSR
.070
RMSR
.071
.084
.087
.091
RMSR
080
RMSR
081
.092
.092
.093
Table 2.2 continued.
Set 3: Personal Control 145 Variables: 5 Factors)
BASE MODEL: GFI x? DF y2 /DF
Configural Invariance Test .87 4973 2805 1.77
X 2 A from DF A from
ALTERNATIVE MODELS: Base Model Base Model v2A\DFA
Metric Invariance— LY=IN 8 8 80 1.10
+Factor Variances Equal Across 97 98 0.99
Groups—
LY=IN, PS diagonal = across
groups
+Factor Covariances Equal Across 139** 110 1.26
Groups—
LY=IN, PS=IN
+Item Uniquenesses Equal Across 268** 200 1.34
Groups—
LY=TN, PS=IN, TE=IN__________________________________________
** = X 2 significantly different form Base Model x2 given DF change at p<01
125
RMSR
.081
RMSR
.084
.086
.089
.089
126
It can be seen in Table 2.2 that within each of the three sets of factors the fits
for each requirement of configural and metric invariance are good. The chi-squares
are large, but mainly because the N is large. The ratio of chi-square to DF is less that
2.00 (a good rule o f thumb) and the RMSR in each test is small. A small RMSR
(under .1 0 ) means that the covariances among items have been adequately accounted
for by the proposed models.
When factor loadings are constrained to be invariant across groups (LY=IN),
the change in chi-square from the base model to this model is not sufficient (given the
change in degrees-of-freedom) to conclude that the metric model yields a significantly
worse fit than the configural model. This, plus the small RMSR, supports an
hypothesis that the factors are metrically invariant. In each set of factors, it can be
concluded that the same constructs measured in one age group are measured in the
other two groups. Any group differences associated with factors scores computed
from the items of these metrically invariance factors can be interpreted as differences
in the constructs themselves, not differences in the scale of measurement.
Models that require the factor variances to be equal across age groups (PS
diagonal=across groups) do not fit significantly worse than models where these
variances are allowed to be free. This supports an hypotheses that the factor
variances for subjects of different ages do not vary significantly. Thus, homogeneity
of variance across age is indicated.
127
The factor covariances (PS=IN) do vary significantly across groups. Models
in which the factor covariances are required to be equal across groups fit significantly
worse than models in which these covariances are allowed to be free. Since the factor
variances do not vary significantly, the test indicating significant differences between
covariances indicates that the factor intercorrelations vary significantly across groups
Although the same constructs are measured across age groups (under conditions of
homogeneity of variance), interrelationsips among the factors of lifestyle do differ
When item uniquenesses are constrained to be equal across groups (TE=IN),
the models fit significantly worse than if they are allowed to vary. This indicates that
although the factors are metrically invariant and meet a test of homogeneity of
variance, the items contribute different proportions of common variance across
groups Means and standard deviations are reported in Appendix A for all indicators
used in analyses for the three age groups.
Relationships Among Lifestyle and Cognitive Variables
Lifestyle variables that correlate with both age and a cognitive ability plausibly
will be related to the development o f that ability. Zero-order correlations between
cognitive variables and lifestyle indicators can be found in Table 2.3. But a lifestyle
variable that is correlated zero with age may also be related to the development of
that ability Such an effect can be seen analytically with part correlational analysis.
Table 2.3. Zero-order Correlations of Lifestyle Variables with Age, Memory for
Paired Associates (MPA), Concentration (CON), Speeded Letter Comparison (SLC),
residua! Fluid Reasoning (rGf), and residual Crystallized Knowledge (rGc)_______
Corr Corr. Corr. Corr. Corr. Corr.
with with with with with with
Age MPA CON SLC rGf rGc
Income -.34** .26** .35** .35** .07 -.03
Education -.08 . 1 0 * .04 .25** .16** .2 1 **
# o f Children .25** -.1 1 ** 0 2 -.16** -.05 - 0 1
# o f Years in Relationship .69** -.31** -.29** -.49** -.1 1 * .18**
Attention to Healthy Diet .28** -.09* -.16** -.05 -.09* 1 0 *
Smoking Behavior .1 1 *
_ 13**
-.04 -.1 2 ** - . 0 2 .03
Use of Alcohol for Escape -.07 - . 0 1 .05 .04 -.08* -.03
Per Week Drinking Frequency -.05 .09* .06 .04 -.03 .03
Ave. # of Drinks at one sitting -.1 2 * .03 .03 .04 -.04 -.07
Frequency Passing out Drunk
. 14**
. 0 1 .09* .05 -.05 -04
Information Seeking .25** -.04 - 1 1 ** -.06 - 07 .27**
passive Entertainment .07 -.03 0 1 - . 0 2 -.07 . 0 2
Negative Attitude Toward TV -.03 . 1 2 ** 06 .03 -.06 - . 0 1
Community Involvement .34** - 18**
- 1 1 **
-.17** - . 0 0 05
Optimism -.09* . 1 2 ** .1 2 ** .18** .0 1 .1 0 *
Self Esteem - . 0 2 .09* .05 .17** .06 1 1 *
Confide in Others -.27**
14**
.2 2 ** .2 1 ** -.03 -.04
Tolerance for Disorder -.13** . 1 0 * .19** .13** .07 .13**
Secure Attachment .18** -.04 .03 -.06 - . 1 0 .07
Belief in God .05 -.13** -04 -.07 -.07 -.14**
Belief in Nihilism/Chaos 01
_ 1 1 **
-.13** -.09* -.07 -.23**
Belief in Luck .23** -.2 0 ** -.17** -.31** -.04 -.06
Belief in Powerful Others .1 2 * -.16** -.09* - 19** -.04 -.17**
Belief in Superstition .0 2 -.07 - 1 1 ** -09* -.07 -.16**
* = Significant correlation or part correlation at p< 05
** = Significant correlation or part correlation at p<01
f = Part correlation coefficient--Age parted from lifestyle variable
The model in this case is:
where A=age, L=Lifestyle variable, G=ability variable, the r’s represent correlations,
and, in particular, r^a.D represents the part correlation between age and a cognitive
variable in which the part associated with lifestyle has been parted out If rAL=0, but
rG L is substantial, say 6, then as long as rA G is not zero then
That is, the age correlation with the ability is increased by control for the lifestyle
variable, thus indicating that the lifestyle variable affects the age relationship with the
ability This is not proof, of course, but it is evidence for a prima facia case for the
claim that the lifestyle variable affects the development of the ability.
This result can occur for both longitudinal and cross-sectional correlations,
although the interpretation can be different for these two conditions. If the data are
longitudinal, and L and G vary both between time (age) within persons and between
persons, then the overall correlations of L and G with A and each other are, in each
case, a combination of the within person over time correlation and the between
person correlation. Symbolically,
[2.4]
r AG w f A G + br.U , [2.5]
r.A i. »r.vi+br.\i. [2.6]
rG L — wrG ,i.+brG L
[2.7]
130
where w and b subscripts represent “within” and “between” respectively. The within
correlations can be different in magnitude and sign from the between correlations.
For this reason longitudinal results can be different from cross-sectional findings It
would be misleading to interpret cross-sectional findings as indicating longitudinal
findings in such a case (Baltes, Reece & Nesselroade, 1977).
If the within and between correlations are of the same sign, however, the
results from cross-sectional analyses are indicative of— although not isomorphic to—
results that would obtain in longitudinal study. This can be seen in equations 2.5
through 2 .7. If the between person correlations are all zero, so all the correlations
indicate covariation within persons over time, the part correlation of equation 2 .3
indicates that as age increases ability increases, that as a lifestyle practice increases
ability increases, that lifestyle is correlated zero with time (age), and that when the
lifestyle association with ability is controlled, the correlation between ability and age
(time) increases relative to when the lifestyle influence is not controlled. This analysis
indicates that lifestyle represents something that affects the display of the ability over
time The ability will increase more over time— through development— when this
lifestyle influence is small (or absent) than when it is large. This result provides a
basis for a developmental interpretation— for support or rejection of a developmental
hypothesis
A cross-sectional correlation can be precisely the same as a longitudinal
correlation if for each and every repeat measure in longitudinal analysis there is an
131
exactly corresponding replication of a person of one age with a person from another
age. In a general way one can infer that age differences indicate age change but only
if it is reasonable to infer that older persons are replications of younger persons. That
is, if it is reasonable to suppose that repeated sampling of the younger persons would
result in a sample of the older persons. This may seem impossible, but in fact it is the
same form of assumption that is made when we interpret any average as being a
reasonable representation of what is true for individuals. In such cases, we assume
people are replications of other people and this is the same assumptions we make
when we infer that the difference between older persons and younger persons
indicates a change— we assume that older persons are replications of younger persons
(Horn & Donaldson, 1967, 1980, 1992). In other words, if the sampling of people at
different ages corresponds to the sampling of each person at different times in
longitudinal study, the b rjk and correlations in equations 2.5 through 2.7 will be
precisely the same. Under these conditions, results from part correlational analysis
can provide a basis for a developmental interpretation.
One must be thoughtful and cautious when considering an assumption that a
sampling of people of different ages is equivalent to a sampling of persons at different
times corresponding to the different ages. It is not likely to be true in detail even as it
can be true in approximation. To the extent that one sample of humans is much like
another sample of humans as concerns variables under consideration, and to the
extent that people of different ages would not have had drastically different
132
experiences affecting the variables under consideration, the bijk and w rjk correlations in
equations 2 5 through 2.7 will be the same in sign and similar in magnitude.
It is important to remember that aging is only imperfectly operationalized in
terms of the number of years since birth. The passage of time is only space through
which the influences of aging operate The influences operating in one unit of age
may be quite different from the influences operating in another unit. Some organisms
age faster than others A dog aged seven years is said to be equivalent in maturity
and time to death as a human aged 49 years Humans may differ in this way if not in
this magnitude.
Different aspects of lifestyle accelerate aging. For example, over-exposure to
ultraviolet light, heavy smoking, or writing a dissertation may make one older than if
these lifestyle experiences were avoided. Such variables may better indicate an aging
process than does chronological age.
A variable's correlation with age can represent nothing more than the passage
of time, not a causal influence on another variable. For example, "Number of times
one has visited shoe stores" is probably not much of an indicator of (cause of)
crystallized knowledge even if it is found to correlate .20 with Gc. This is true
because it correlates very highly with age— and primarily represents only the passing
of time in living— which also correlates (about .30) with Gc. In this case if age is
parted from the lifestyle variable (under an assumption that both age and number of
visits to shoe stores represent passage of time), one can see that the part correlation
will be zero when T a l equals ^ iAg a, the ratio of the ability-lifestyle correlation to the
ability-age correlation. In the example, if visits to shoe stores correlates .20/ 30 = .60
with age, then the part correlation between ability and "visits," controlling for time, is
zero. In other words, if the ability correlates with lifestyle only to about the level it
correlates with age, there is indication that the lifestyle variable does not indicate
ability, but instead indicates only the passage of time that age also indicates.
Cautiously, then, and with questioning persistently in mind, the analyses and
the results of this study will be based on assumptions that between subject
correlations, including part and partial correlations, are of the same sign as, and
approach the magnitude of, correlations that would obtain for repeated-measures
longitudinal data gathering The possibility of age cohort influences will also be
considered in the interpretation of results.
In general, then, association of lifestyle variables with age and with abilities
should be evaluated in terms of the plausibility that the variables indicate processes of
aging, or processes of cognition, or both. Variables that are highly correlated with
age indicate a process of aging to a greater extent than they indicate a process of
cognition Variables that correlate highly with cognition but lowly with age may
indicate a process of cognition to a greater degree than they indicate a process of age.
Variables that correlate with both age and cognition are most likely to indicate both
134
aging and cognition. The results of the part correlational analyses in the following
section have been organized in accordance with these considerations.
Interpretation o f the Part Correlational Results. Results when lifestyle
variables are parted form Memory for Paired Associates (MPA), Concentration
(CON), Speeded Letter Comparison (SLC), residualized Fluid Reasoning (rGf), and
residualized Crystallized Knowledge (rGc) are presented in Tables 2.4 through 2 .8
respectively The part correlations for lifestyle variables that have nonzero
correlations with age and near zero correlations with cognitive abilities are listed first
in the tables. Part correlations for variables that correlate nonzero with cognitive
abilities and near zero with age are listed next. Part correlations or lifestyle variables
that have nonzero correlations with both age and abilities are presented last in the
tables. This organization assists in the interpretation of significant correlational shifts
due to parting. The first column of numbers in each table provides the zero-order
correlations with age— for the ability in the first row, for lifestyle variables in the other
rows. The second column provides the correlations for an ability with lifestyle
indicators. The third column contains differences between the residual-ability
correlations with age and the unresidualized (zero-order) ability correlation with age.
The last column contains the part correlation between the ability and the lifestyle
indicator after time (represented by age) has been parted out of the lifestyle variable.
Significant third-column differences (between age-ability and age-residual
ability correlations) in the first (set 1) section of each table probably reflect processes
135
of aging to a greater extent than they reflect processes of cognition and cognitive
development. This is particularly reasonable if the correlations between ability and
lifestyle is not larger than zero after control for time in the lifestyle variable (last
column of the tables).
Significant third-column differences in the second and third sections of the
table (sets 2 and 3) may reflect processes that are operating in the development of
cognitive abilities. This interpretation is particularly reasonable if in the last column
of the table the correlation between ability and lifestyle is significant after control for
time (age) in the lifestyle variable
Short-term Apprehension Retrieval. It can be seen in Table 2.4 that the
correlation between age and MPA alone is - 469 When Information Seeking is
removed from MPA, the correlation is - 454 The difference between these
correlations ( 015) is the amount of MPA decline that can be accounted for by
Information Seeking control. The zero-order correlation of Information Seeking with
age is relatively high (.25) and its zero-order correlation with MPA is only -.04.
These correlations suggest that Information Seeking represents a process of aging and
not as much a process of short-term memory The amounts of time and energy spent
in seeking and gathering information are larger for older persons than for younger
persons
136
Table 2.4. Lifestyle Variables that Account for the Age Related Decline o f Memory fo r
Paired Associates (MPA)
Zero- Zero-Order Difference Part
Order Correlation from Age Correlation
Correla with MPA Correlation between MPA
tion (-.469) and Lifestyle-
Model: with Age As a result Age removed
of Parting from Lifestyle
BASE MODEL:
MPA aione -.469 1 . 0 0
4.31 IQ units/ decade
decline
ALTERNATIVE
MODELS:
Lifestyle Variables Parted
from MPA
Information Seeking .25* -.04 .015f .08*
Secure Attachment .18* -.04 013f .04
Frequency Passing out Drunk -.14* .0 1 .0 1 2 * .07
Ave. # of Drinks at one sitting -.1 2 * .03 .0 1 2 * - . 0 2
Belief in God .05 -.13* . 0 0 1 -.1 0 *
Negative Attitude Toward TV -.03 .1 2 * . 0 0 1 . 1 0 *
Belief in Nihilism/Chaos .0 1 -.1 1 * .003 -.1 2 *
Education -.08 . 1 0 * .008 .06
Self Esteem - . 0 2 .09* .004 .09
Per Week Drinking Frequency -.05 .09* .003 .07
# of Years in Relationship .69* -.31* .206* - . 0 2
Income -.34* .26* .076* .1 2 *
Community Involvement .34* -.18* .055* - . 0 2
Attention to Healthy Diet 28* -.09* .030* .05
Confide in Others -.27* .14* .031* . 0 2
# of Children .25* - 1 1 * .031* .0 1
Belief in Luck .23* -.2 0 * .034* -.09*
Tolerance for Disorder -.13* .1 0 * .0 1 0 * .05
Belief in Powerful Others .1 2 * -.16* .009 -.1 0 *
Smoking Behavior .1 1 * -.13* .009 -.09*
Optimism -.09* .1 2 * .008 . 1 0 *
= Significant Correlation Change (Significant R2 change form Base Model R ) at
p<05
* = Significant Correlation at p<05
137
Thus, at first it appears that the age difference is not a developmental
difference But the fact that Information Seeking continues to correlate significantly
with MPA after age variance has been removed from it (see column 4 o f Table 2 .4)
suggests that Information Seeking is developmentally related to short-term memory.
The small effects are significantly different from zero in this sample of 577 adults As
such, they suggest that persons who developmentally increase in information seeking
over adulthood are persons who maintain short-term apprehension and retrieval— do
not decrease as much as others
This (cautious) interpretation for Information Seeking is not indicated for the
other three variables in the first set of Table 2.4— i.e., for Secure Attachments, number
of alcoholic drinks at each sitting, and number of times passed out from drinking
After age (time) is parted from these variables, the correlations with MPA sink to or
remain at levels that are not significantly different from zero. These variables thus
appear to be indicative of processes of aging, not processes that affect cognition
The variables in the second section of Table 2 .4 correlate significantly
(although lowly) with MPA, but near zero with age These variables could indicate
developmental effects, as in equation 2 4, but the correlations with MPA must be
larger than observed here to indicate evidence for this hypothesis (correlational shifts
are not significant). For example, the correlation between education and MPA would
need to be at least .43 (not. 10, as is) in order for the part correlation to be
significantly different from the zero-order age-MPA correlation. With smaller age-
138
lifestyle correlations, however, the correlations between lifestyle and ability can be
smaller and still indicate significance For example, if Belief in God were to correlate
as much as -.35 with MPA, the part correlations would be significantly different from
the zero-order age-MPA correlations. Although the lifestyle variables in this second
section appear to be highly correlated with MPA they are not high enough to indicate
significant part correlation shifts
The third section of Table 2.4 contains lifestyle variables that have significant
correlations with both age and MPA. In many cases there is indication of a part
correlations that can be given a developmental interpretation It is wise, however, to
consider these possible interpretations cautiously, for the correlations may indicate
only that the lifestyle variable represents the same passage of time that is represented
by age.
Consider the part correlation for number of years in a (marriage) relationship,
for example. This variable correlates .69 with age and may very well be indicative of
an age cohort effect-those married early in this century have had a longer time to be
in a relationship than those married late in this century. For this variable, an older
person is not merely a replication of a comparable younger person and, therefore, to
think about this particular age difference as an age change is dubious. Time in
relationship indicates time in living, which is age The part correlation between MPA
and the number of years in relationship is zero when time (measured by age) is parted
out (last column in the Table 2 .4). Thus, a cautious interpretation of this finding is
139
that it indicates the decline with age of short-term memory, but a more realistic
interpretation is that this relationship indicates an age cohort difference that describes
a process of aging.
Several of the significant differences of the third column in the third section of
the tables require such cautious interpretation. In Table 2.4, the part correlations
between MPA and lifestyle, when age is controlled in lifestyle, are close to zero and
not significantly different from zero for all the variables except Income, Belief in
Luck, Belief in Powerful Others, Smoking Behavior, and Optimism There is a
significant difference between the age-ability correlations and age-residual ability
correlations (third column) for Income, and Belief in Luck. These two pieces of
evidence suggest that these variables are related to the development of short-term
memory in adulthood. This difference not being significant for the last three variables
in the table suggests that these variables are not a part of the age decline of MPA.
The significant part correlation differences in column 3 of Table 2 .4 that are
also accompanied by significant part correlations in column 4 are for Income,
Information Seeking, and Belief in Luck— i.e., the less the report of believing in luck,
the less the decline in MPA For the other lifestyle variables there is either no
evidence that the variable affects the age relationship for MPA (the part correlation
difference is not significantly different from zero) or such relationship as there might
be is a relationship with time in living, per se, not with short-term memory.
140
Table 2.5. Lifestyle Variables that Account fo r the Age Related Decline o f
Concentration (CON)
Zero- Zero- Difference Part
Order Order form Age Correlation
Correla Correla Correlation between CON
tion tion (-497) and Lifestyle-
Model: with Age with CON As a result Age removed
of Parting from Lifestyle
BASE MODEL:
CON alone -.497 1 0 0
4.57 IQ units/ decade
decline
ALTERNATIVE
MODELS:
Lifestyle Variables Parted
from CON
# of Children .25* 0 2 .0 1 2 * .15*
Secure Attachment .18* .03 0 1 0 * .13*
Ave # of Drinks at one sitting -.1 2 * 03 .014* - . 0 2
Smoking Behavior . 1 1 * -0 4 .014* . 0 2
Belief in Nihilism/Chaos 0 1 - 13* 004 - 14*
Belief in Superstition .0 1 - 1 1 * . 0 0 1 -.1 0 *
# of Years in Relationship .69* -.29* .229* .03
Income -.34* .35* .088* .2 1 *
Community Involvement .34* - 1 1 * .035* .06
Attention to Healthy Diet .28* - 16* .047* - . 0 2
Confide in Others -.27* .2 2 * .044* .1 0 *
Information Seeking .25* -.1 1 * .027* . 0 1
Belief in Luck .23* -.17* .032* -.06
Frequency Passing out Drunk -.14* .09* .015* .04
Tolerance for Disorder -.13* 19* . 0 0 2 .14*
Belief in Powerful Others . 1 2 * -09* .008 - . 0 2
Optimism -.09* .1 2 * . 0 0 2 .09*
= Significant Correlation Change (Significant R change form Base Model R 2 ) at
p<05
* = Significant Correlation at p< 05
141
Concentration. Table 2 .5 provides the same kind of information as Table 2 .4,
but the Concentration (CON) component of Gf is considered. It can be seen in this
table that the lifestyle variables that pass the two tests indicating a developmental
relationship— significance o f the difference between the part correlation, lifestyle
removed, and the zero-order correlation of CON with age, and significant correlation
o f lifestyle with CON after control for time in living— are Income, Secure Attachment,
Confide in Others, Optimism, and Number of Children.
Cognitive Speed. In Table 2.6 results comparable to those for MPA and CON
are provided fore SLC, the marker for cognitive speed. In this case the lifestyle
indicators that pass the two tests for indicating a developmental relationship are
Income, Information Seeking, and Attention to Healthy Diet.
Residual Fluid Reasoning. The results of Table 2 .7 are similar in form to
those in Tables 2.4-2.6, but the cognitive variable analyzed is residual fluid ability
(rGf)~i.e., Gf, estimated with PLS, from which the parts associated with MPA, CON,
SLC, and Gc have been removed.
The zero-order correlation of rGf with age is -. 15, indicating that there is an
element of fluid reasoning that declines with age after all the (linear) decline that can
be accounted for by cognitive speed, short-term memory and concentration has been
removed The only lifestyle variable that passes the two tests to suggest that it may
account for some of the remainder of this decline is Frequency of Passing Out Drunk.
142
Table 2.6. Lifestyle Variables that Account for the Age Related Decline o f Speeded
Letter Comparison (SLC)
Zero-Order Zero- Difference Part
Correlation Order from Age Correlation
with Age Correla Correlation between SLC
tion (-.629) and Lifestyle-
Model: with SLC As a result Age removed
of Parting from Lifestyle
BASE MODEL:
SLC alone -.629 1 . 0 0
5.77 IQ units/ decade
decline
ALTERNATIVE
MODELS:
Lifestyle Variables Parted
from SLC
Attention to Healthy Diet .28* -.05 019f .13*
Information Seeking .25* -.06 .018* .1 0 *
Secure Attachment 18* -.06 .0 1 1* .05
Frequency Passing out Drunk -.14* 05 .014* . 1 0
Ave. # of Drinks at one sitting -.1 2 * .04 .009* -.03
Education -.08 .25* .0 0 1 .2 0 *
Self Esteem . 0 2 .17* . 0 0 1 .17*
Belief in Nihilism/Chaos 0 1 -.09* . 0 0 2 -09*
Belief in Superstition 0 1 -.09* . 0 0 1 -.09*
# of Years in Relationship .69* -.49* .275* -.04
Income -.34* .35* .074* .16*
Community Involvement .34* -.17* .053* .05
Confide in Others -.27* .2 1 * .044* .06
# of Children .25* -.16* .039* .0 1
Belief in Luck .23* -.31* .006 -.17*
Tolerance for Disorder -.13* 13* .006 .06
Belief in Powerful Others .1 2 * -.19* .007 -.15*
Smoking Behavior .1 1 * -.1 2 * .007 -.05
Optimism -.09* .18* .006 .15*
= Significant Correlation Change (Significant R/ change form Base Model ) at
p<05
* = Significant Correlation at p< 05
143
Table 2.7. Lifestyle Variables that Account for the Age Related Decline o f Residualized
Fluid Ability (rGf)— Fluid Ability with Short-term Apprehension Retrieval,
Concentration, Clerical Speed, and Crystallized Ability Variance Removed
Model:
Zero-Order
Correlation
with Age
Zero-
Order
Correla
tion
with rGf
Difference
from Age
Correlation
(-151)
As a result
of Parting
Part
Correlation
between rGf
and Lifestyle-
Age removed
from Lifestyle
BASE MODEL:
rGf alone -.151 1 . 0 0
1.39 IQ units/ decade
decline
ALTERNATIVE
MODELS:
Lifestyle Variables Parted
from rGf
Income -.34* .07 0 2 0 f .0 2
Community Involvement .34* - 0 1 .011* .05
Confide in Others -.27* -.03 0 1 lf -.07
Information Seeking .25* -.07
4 - •
0 0
©
-.03
# of Children .25* -.05
* - ■
0 0
o
-.0 1
Belief in Luck .23* -.04 . 0 1 2 - 0 1
Secure Attachment 18* - . 0 1 - 0 1 1 t .0 2
Frequency Passing out Drunk -.14* -.08 .009* -.08*
Tolerance for Disorder - 13* .07 .0 1 1 * .05
Ave. # of Drinks at one sitting - 1 2 * -04 .009* -.06
Belief in Powerful Others .1 2 * -.04 .0 1 1+ - . 0 2
Smoking Behavior .1 1 * -.03 .0 1 1 * - 0 1
Optimism -.09* .01 .003 .01
Education -08 .16* .008 15*
Use of Alcohol to Escape -.07 -.1 0 * .004 -.1 0 *
# of Years in Relationship .69* - 1 1 * 048* . 0 2
Attention to Healthy Diet .28* -.09* .026* -.05
= Significant Correlation Change (Significant R2 change form Base Model R ) at
p< 05
* = Significant Correlation at p<05
144
Table 2.8. Lifestyle Variables that Account for the Age Related Decline o f Residualized
Crystallized Ability (rGc)— Crystallized Ability with Short-term Apprehension Retrieval,
and Fluid Ability Variance Removed
Model:
Zero-Order
Correlation
with Age
Zero-
Order
Correla
tion
with rGc
Difference
from Age
Correlation
(.310)
As a result
of Parting
Part
Correlation
between rGc
and Lifestyle-
Age removed
from Lifestyle
BASE MODEL:
rGc alone
2.85 IQ units/ decade
enhancement
ALTERNATIVE
MODELS:
Lifestyle Variables Parted
from rGc
.310
1.000
Income -.34* -.03
- 0 1 0 f .07
Community Involvement .34* 05 -.018f -06
Confide in Others -.27* -.04 -.0 1 2 * .04
# of Children .25* -.01 -.009* -.09*
Belief in Luck .23*
- . 0 6
+.014* -.13*
Secure Attachment 18* 07 -.009* .01
Frequency Passing out Drunk -.14* 01 -.0 1 1 * . 0 0
Ave. # of Drinks at one sitting - 1 2 * -.07
-.0 1 1 * -.04
Smoking Behavior .1 1 * .03 -.011* - . 0 0
Belief in Nihilism/Chaos .0 1
-.2 2 *
-.007 -.2 2 *
Education -.08 .2 0 *
+.030* .23*
Belief in Superstition . 0 2 - 1 6*
-.006 -.16*
Belief in God .05
-.14*
+.0 1 0 * -.15*
Self Esteem - 0 2 U*
- .0 0 1 .1 1 *
# of Years in Relationship 69*
1 8*
-.179* .01
Information Seeking .25* .27*
-.057* .19*
Tolerance for Disorder -.13* 13* + 028* .17*
Belief in Powerful Others .1 2 *
-.17*
+.025* -.2 0 *
Optimism -09*
1 0*
+.0 2 1 * .1 2 *
Attention to Healthy Diet .28* 1 0*
-.024* .01
= Significant Correlation Change (Significant R2 change form Base Model R2) at
p<05
* = Significant Correlation at p<05
145
Residual Crystallized Knowledge. The results of Table 2 .8 are of the same
form as those of Table 2 7 except the cognitive ability analyzed in this case is rGc, the
part of Gc, estimated by VOC2, that remains after MPA and G f (estimated with PLS)
have been removed. The correlation with age for this measure of crystallized
knowledge is .31. The lifestyle variables that pass the two tests indicating that they
relate to the adulthood improvement of rGc are of two kinds: (1) those that relate
negatively, suggesting that people who are low in these indicators of lifestyle are the
ones most likely to improve in Gc, and (2 ) those that relate positively, suggesting that
those reporting this lifestyle indicator are the ones most likely to enhance rGc over
adulthood. The negative indicators are Belief in Luck, Belief in Powerful Others,
Belief in God, and Number of Children. The positive indicators are Education,
Information Seeking, Tolerance for Disorder, and Optimism.
Non-Overlap Multiple Variable Relations fo r Ability Components
Table 2.9 includes part correlations of lifestyle indicators with abilities MPA,
CON, and SLC when age is held constant. Table 2.10 contains the same analyses for
rG f and rGc
Table 2.9. Correlations o f Lifestyle Variables with Age, Memory fo r Paired
Associates (MPA), Concentration (CON), and Speeded Letter Comparison (SLC)--
Part Correlations where Age has been Removedfrom Lifestyle Variables Also Shown
Corr Corr. Part Corr. Part* Corr. Part*
with with Corr. with Corr. with Corr
Age MPA MPA CON CON SLC SLC
Income -.34** .26** .1 2 ** .35** .2 1 ** .35** .16**
Education -.08 .1 0 * .06 .04 .0 1 .25** .2 0 **
# of Children .25** - 1 1 ** .01 . 0 2 .15** -.16** 0 1
# of Years in Relationship .69** -.31** - . 0 2 -.29** .03
_ 49**
-.04
Attention to Healthy Diet
*
*
00
-.09* .05 -.16** - . 0 2 -.05 .13**
Smoking Behavior .1 1 * -.13** -.09* -.04 . 0 2 -.1 2 ** -.05
Use of Alcohol for Escape -.07 - . 0 1 -.05 .05 .0 1 .04 - . 0 2
Per Week Drinking Frequency -.05 .09* .07 .06 .04 .04 .0 1
Ave. # of Drinks at one sitting -.1 2 * .03 - . 0 2 .03 - . 0 2 .04 -.03
Frequency Passing out Drunk -.14** 0 1 - . 0 2 .09* .04 .05 - . 0 1
Information Seeking .25** -.04 .08* -.1 1 ** .0 1 -.06 .1 0 *
Passive Entertainment .07 -.03 0 1 .0 1 .05 - . 0 2 .03
Negative Attitude Toward TV -.03 .1 2 ** 1 0 * .06 .04 .03 .0 1
Community Involvement .34** -.18** - . 0 2 -.1 1 ** .06
. 17**
.05
Optimism -.09* .1 2 ** .1 0 * .1 2 ** .09* .18** .15**
Self Esteem - . 0 2 .09* .09* .05 .05 .17** .17**
Confide in Others -.27**
14**
.0 2 .2 2 ** .1 0 * .2 1 ** .06
Tolerance for Disorder -.13** .1 0 * .05 19** .14** .13** .06
Secure Attachment .18** -04 .04 .03 .13** -.06 .05
Belief in God .05
13**
-.1 0 * -.04 - . 0 2 -.07 -.04
Belief in Nihilism/Chaos .01 -.1 1 ** -.1 2 ** - 13**
. 1 4 **
-.09* -.09*
Belief in Luck .23** -.2 0 ** -.09* -.17** -.06 -.31** -.17**
Belief in Powerful Others .1 2 * -.16** -.1 0 * -.09* - . 0 2
_ 19**
-.15**
Belief in Superstition .0 2 -.07 -.07
_ 1 1 **
-.1 0 * -.09* -.09*
* = Significant correlation or part correlation at p<05
** = Significant correlation or part correlation at p<01
f = Part correlation coefficient— Age parted from lifestyle variable
147
T able 2.10. Correlalions o f Lifestyle Variables with Age, Residualized Fluid A bility
(rGf), and Residualized Crystallized Ability (rGc)— Part Correlations where Age has
been Removedfrom Lifestyle Variables Also Shown
Corr Corr. PartT Corr. Part*
with with Corr. with Corr
Age rGf rGf rGc rGc
Income -.34** .07 . 0 2 -.03 .07
Education -.08 .16** .15** .2 1 ** .23**
# of Children .25** -.05 - 0 1 - . 0 1 -09*
# of Years in Relationship 69** - 1 1 * . 0 2 .18** 0 1
Attention to Healthy Diet
28**
-09* -.05 .1 0 * 0 1
Smoking Behavior .1 1 * - . 0 2 - .0 1 .03 - . 0 0
Use of Alcohol for Escape -.07
#
00
o
1
-.1 0 * -.03 - . 0 0
Per Week Drinking Frequency -.05 -.03 -.03 .03 04
Ave # of Drinks at one sitting -.1 2 * -.04 -.06 -.07 -.04
Frequency Passing out Drunk
. 1 4 **
-.05 -.08* -.04 . 0 0
Information Seeking .25** -.07 -.03 .27**
19**
Passive Entertainment .07 -.07 -06 . 0 2 - . 0 1
Negative Attitude Toward TV -03 -06 -.06 - . 0 1 - 0 1
Community Involvement .34** - . 0 0 .05 .05 -06
Optimism -.09* .0 1 .01 .1 0 * .1 2 **
Self Esteem - . 0 2 .06 .06 .1 1 * 1 1 *
Confide in Others -.27** -.03 -.07 -.04 04
Tolerance for Disorder -.13** .07 .05 .13** .17**
Secure Attachment
18**
- . 1 0 . 0 2 .07 .0 1
Belief in God .05 -.07 -.07 -.14**
_ jg**
Belief in Nihilism/Chaos .0 1 -.07 -.07 -.23** -.2 2 **
Belief in Luck .23** -.04 - . 0 1 -.06 -.13**
Belief in Powerful Others .1 2 * -.04 - . 0 2 -.17** -.2 0 **
Belief in Superstition . 0 2 -.07 -.07 -.16** - 16**
* = Significant correlation or part correlation at p<05
** = Significant correlation or part correlation at p<01
f = Part correlation coefficient--Age parted from lifestyle variable
rGf=Power Letter Series with the variance of Memory for Paired Associates,
Concentration, Speeded Letter Compassion, and Vocabulary test 2 removed
rGc=Vocabulary test 2 with the variance of Memory for Paired Associates and Power
Letter Series removed
148
A multiple regression (MR) prediction of the components of an ability from
lifestyle variables puts these indicators in competition and removes any overlap the
predictors may involve. If Belief in God and Belief in Powerful Others overlap in the
thinking of respondents or represent the same kinds of influences operating on an
ability, then this overlap will be counted only once, in one of the two variables, in an
MR analysis. If the overlap is the only element of the two variables that is related to
the ability, then only the predictor that most reliably measures this element will be
among the significant indicators identified with MR analysis. In this sense MR
analysis provides evidence of the best indicators of the ability.
If age (also regarded as time of living) is entered as one of the predictors in
this kind MR analysis, then any overlap the lifestyle indicators have with this time in
living variable is removed If that lifestyle variables then relates to the ability variable,
it is only because it indicates something other than time in living that is related to the
ability.
MR analysis with age partialled thus provides different information than is
provided by the two-test part correlational analyses summarized in previous sections.
MR analysis indicates relationships that are not developmental (insofar as variations in
age represent variations in development), whereas the two-test part correlation
analyses provide evidence of development.
149
Table 2.11 Regression Models that Account fo r the Most, Non-Redundant Variance
o f Memory for Paired Associates (MPA), Concentration (CON), and Speeded Letter
Comparison (SLC), Residual Fluid Ability (rGf), and Residual Crystallized
Knowledge (rGc) Above and Beyond Age
DV=MPA: R2 =.2710
DF Standardized
Variables in the Equation: Estimate
Intercept 1 0.007553
Age 1 -0.445927**
Income 1 0056509
Information Seeking 1 0083707*
Negative Attitude Toward TV 1 0.090132*
Optimism 1 0.028206
Smoking Behavior 1 -0073825*
Belief in God 1 -0 116867**
Belief in Nihilism/Chaos 1 -0.113029**
DV=CON: RJ=.3430
DF Standardized
Variables in the Equation: Estimate
Intercept 1 0.014649**
Age 1 -0 486541**
Income 1 0 166157**
# o f Children 1 0.132368**
Tolerance for Disorder 1 0109375**
Secure Attachment 1 0 067177*
Belief in Nihilism/Chaos 1 -0 058352
DV=SLC: R2 =.4661
DF Standardized
Variables in the Equation: Estimate
Intercept 1 0.006614**
Age 1 -0 584030**
Income 1 0.078599*
Education 1 0 135318**
Information Seeking 1 0.041614
Attention to Healthy Diet 1 0.039352
Optimism 1 0002403
Self Esteem 1 0.071206*
Belief in Luck 1 -0.118242*
150
Table 2.11 continued.
DV=rGf: R2 =.0530
DF
Variables in the Equation:
Intercept 1
Age 1
Education 1
Use of Alcohol for Escape 1
DV=rGc: R2 =.2584
DF
Variables in the Equation:
Intercept 1
Age 1
Education 1
Information Seeking 1
Tolerance for Disorder 1
Belief in God 1
Belief in Nihilism/Chaos 1
Standardized
Estimate
-0.006624**
-0 143925**
0.151152**
-0.098894*
Standardized
Estimate
0.008745**
0.311435**
0 149184**
0 153331**
0.096524*
-0.133230**
-0.211715**
* = t value for parameter significant at p<05
** = t value for parameter significant at p< 0 1
151
Results for MR analyses for lifestyle variables for which the zero-order
correlation with an ability component was significantly larger than zero are
summarized in Table 2 .11, separately for MPA, CON, SLC, rGf and rGc.
It can be seen in the table that for MPA, the significant predictors (in addition
to age) are Negative Attitude Toward TV and Information Seeking, and negatively,
Smoking Behavior, Belief in God, and Belief in Nihilism/Chaos. For CON, the
significant positive predictors are Income, Tolerance for Disorder, Secure Attachment
and Number of Children. Belief in Nihilism/Chaos is a negative predictor. SLC is
positively predicted by Income, Education, and Self Esteem. Belief in Luck is a
negative predictor. Only Education (positively) and Use of Alcohol to Escape
(negatively), are predictors of the residual in Gf that is left after control for MPA,
CON, SLC and Gc Residual Gc, on the other hand, is predicted by Education,
Information Seeking, Tolerance for Disorder and, negatively, Belief in God and Belief
in Nihilism/Chaos.
Discussion
The findings of this study indicate that fluid reasoning and crystallized
knowledge abilities, and the development of these abilities over adulthood, are
associated with features of lifestyle. Lifestyle indicators suggest plausible
explanations for part of the aging decline of vulnerable, fluid abilities and part of the
aging enhancement of maintained, crystallized abilities. Lifestyle indicators are also
related to individual differences in abilities independently of age.
152
The results of Study 1 reinforced a conclusion that fluid reasoning is
comprised of the cognitive processes of short-term apprehension retrieval,
concentration, and cognitive speed. The results of Study 1 also indicated that
crystallized knowledge involves a short-term memory process that is independent of
the comparable process involved in Gf
In Study 2 the processes of Gf and Gc were analyzed to provide a basis for
describing the vulnerable and maintained abilities Aspects of lifestyle that relate to
age differences in these processes point to factors involved in the age-related
differences in fluid reasoning and crystallized knowledge
The analyses of Study 1 isolated a portion of the variance and age-related
variance of fluid reasoning that was not accounted for by the three process variables
and crystallized knowledge— a fluid reasoning residual. Similarly, it was found that a
part of the variance and age-related variance of Gc remained after all that could be
accounted for by fluid reasoning and short-term memory had been removed In Study
2 analyses were directed at isolating aspects of lifestyle that would help illuminate and
define residual Gf and residual Gc
Lifestyle Indicators o f the Development o f Cognitive Abilities
The principal findings of Study 2 are summarized in Table 2.12. These
findings will be considered in discussion in the order in which they are listed in the
first part of this table ( 2 .1 2 a)
153
Table 2.12: Summary o f Findings
2.12a. Lifestyle Variables Passing Both Tests Indicating a Developmental
Relationship With An Ability Component
MPA CON SLC rGf rGc
Income X X X
Information Seeking X X X
Attention to Healthy Diet X
Secure Attachment X
Confide in Others X
Optimism X
Tolerance For Disorder X
Education X
Frequency Passing Out Drunk -X
Belief in Luck -X -X
Belief in God -X
Belief in Powerful Others -X
Number of Children X -X
2.12b. Lifestyle Predictors o f A bility Components with Time in Living (Age)
Partialled Out
MPA CON SLC rGf rGc
Negative Attitude Toward TV X
Information Seeking X X
Smoking Behavior -X
Belief in God -X -X
Belief in Nihilism/Chaos -X -X -X
Income X X
Tolerance For Disorder X X
Secure Attachment X
Number of Children X
Education X X X
Self Esteem X
Belief in Luck -X
Use of Alcohol to Escape -X
X = Lifestyle-Ability Component For Which a Positive Relationship is Indicated
-X = Lifestyle-Ability Component For Which a Negative Relationship is Indicated
154
Income significantly reduces the zero-order age correlations of the process
indicators and Gf and Gc, and thus is an indicator of reduction in amount of cognitive
decline with age associated with three processes of Gf: short-term memory,
concentration, and cognitive speed The correlations of income with these processes
remain significant when age is parted from income. But income is not such an
indicator of reduction in amount of decline with age for the residual of Gf after the
three processes and Gc are removed or for the residual of Gc after short-term
memory and Gf are removed
The multiple regression (MR) analyses in Table 2.12b indicate that Income is
associated with concentration and cognitive speed independently of the association of
any of these variables with age. In these analyses, too, Income is not associated with
residual Gf or residual Gc The zero-order correlations indicate that the more income
one has, the higher one scores on both Gf and Gc, as well as on the domponents
This is not a surprising finding. But if abstracted measures of Gf and Gc are obtained
by removing the components, income is not significantly related to the residuals.
Decline of reasoning ability associated with decline in income and with advancing age
thus is largely decline of the processes that are involved in short-term memory,
capacity for concentration, and cognitive speed
The "why" of these relationships is not made obvious by the findings. Income
may be a function of the abilities and processes, or vice versa, or these influences—
abilities and income— may be in dynamic interaction such that one causes the other
155
and, in turn, is caused by the other. Both decline in fluid ability and decline in the
cognitive process indicators may be reflections o f loss of structures and processes—
say neurological structures and processes-that support ability, and either result from,
or are brought on, by levels of income Both income levels and abilities may be
caused by a third variable that was not measured in the study. Examination o f such
hypotheses remains for further study.
Information Seeking accounts for some o f the aging decline of MPA and SLC
and some the aging increase in rGc, and it continues to correlate significantly with
MpA, SLC and rGc after its association with age, per se, has been statistically
controlled Thus, the suggestion is that persons who developmentally increase in
information seeking over adulthood are persons who tend to maintain short-term
memory and cognitive speed— do not decrease as much as others— and tend to add to
their crystallized knowledge The findings for the MPA and SLC processes suggests
that the stimulation o f Information Seeking exercises these processes. As noted
earlier, there is a relationship between MPA and Gc and MPA can be viewed as a
process of Gc. This relationship is due to the fact that there is a certain amount of
Long-term Retrieval (Glr) involved in sound Gc functioning and short-term
apprehension retrieval, as measured by MPA, is not independent of Glr. When the
residual part of Gc (rGc)-Gc void o f MPA and Gf— is added to the “best” model in
Table 2 11, the contribution that Information Seeking can make to explaining the
variance of MPA goes to zero. Information Seeking can be viewed as representing
156
the crystallized component that is operating in MPA performance. The finding for
rGc suggests that persons with Information Seeking attitudes tend to expand their
base of acculturational knowledge
Attention to Healthy Diet meets the two-test criteria for indication of a
developmental interpretation of the aging decline of cognitive speediness— this
variable accounts for some of the aging decline of speediness and continues to related
to speediness when age is removed. The suggestion is that those people who report
that they are interested in maintaining a healthy diet are the people who, over the
course of adulthood, tend to ward off some of the factors that reduce the speediness
that is involved in the aging decline of fluid reasoning.
Secure Attachment, Confide in others, Number o f Children, and, negatively,
Belief in Luck, all meet the two-test criteria and thus point to developmental
relationships for concentration (CON), but not for the other cognitive processes of
G f Secure Attachment and Confide in Others were hypothesized to be indicators of
psychological health. Low Belief in Luck is, in accordance with hypothesis, an
indicator of self control, and thus also a measure of psychological health and goal
directedness. Perhaps the abstracted number of children variable represents this kind
of influence, also These same variables, plus Tolerance for Disorder, show up in the
MR analyses as related to CON independently of age.
Sperling & Berman (1994) found that measures of psychological health were
related secure interpersonal attachments. In other studies it was found that,
157
compared to adults who report secure interpersonal attachments, persons who report
that their interpersonal attachments are insecure also report being more disturbed by
separation (Reite& Boccia, 1994), less adaptive (Antonucci, 1976), having greater
relationship instability (Antonucci & Akiyama, 1987), heightened depression (Paykel
1983), and low self efficacy (Bandura, 1986)
Amster and Krass (1986) found that social support, as obtained through
ability to confide in others, buffered the effects of negative stressors As discussed in
the introduction to the present study, negative life stressors have been linked to poor
cognitive performance.
These findings from prior research, coupled with the findings in the present
study, thus suggest that those who become able to confide in others and move,
developmentally, toward more secure attachments in interpersonal relationships will
tend to avoid the stresses and depresssions that can accompany disruptive
interpersonal relations. Those who become more self confident, serene, and secure
are not as likely as others to decline in the capacity for concentration that supports
fluid reasoning.
The variable Number of Children in the context of the relationship to
concentration that is seen in this study may indicate that the process of raising
children involves learning techniques for inhibiting distracting stimuli. That such
inhibition of distraction is related to aging and cognition has been shown by McDowd
and Fillion (1989). That it may stem from raising children is another matter, of
158
course, but the present results suggest that further study might support this
hypothesis.
The part of fluid ability that is not accounted for by the processes of short
term memory, concentration, cognitive speed, or Gc is a black hole. It might
represent a capacity for perceiving relationships, as such, or an ability to infer, or any
of a number of other processes that have been said to be indicative of human
intelligence.
In any case, the results of the two-test analyses suggest that, developmentally,
this factor is negatively affected by use of alcohol to the point of passing out. Such
drinking reflects anoxia in the central nervous system, and neurons die from lack of
oxygen The less passing out from drinking one reports, the smaller the decline of
rGf. In the MR analysis, Use of Alcohol to Escape relates to rGf independently of
age and level of education. This alcohol use factor account for the same rGf variance
as does the use of alcohol to the point of passing out, but it is a more reliable
indicator of alcohol abuse
Three major hypotheses regarding the localization of brain damage associated
with chronic abusive use of alcohol (and Korsakoffs syndrome) have been explored:
( 1) that abusive use of alcohol selectively damages the right cerebral hemisphere
(Jones & Parsons, 1972), (2) that such use selectively damages the frontal-limbic
diencephalic system (Tarter, 1975), (3) that such use produces diffuse damage
throughout the brain. Comparisons of control groups with groups comprised of
159
persons hospitalized for severe, chronic, abusive use of alcohol have indicated
enlargement of both lateral ventricles, the third ventricle, and the cerebral sulci on
both sides of the brain Such findings are consistent with an hypothesis of diffuse
damage But this evidence is far from definitive. The issue of localization of brain
damage associated with use of alcohol is not yet settled (Goldman, 1983).
Chronic, abusive use of alcohol is often associated with poor nutrition, vitamin
B deficiency, liver damage, and physical trauma, all of which can result in loss of
central nervous system functioning and thus decline of cognitive abilities.
Heavy alcohol consumption may significantly alter cerebral blood flow
(Altura, Altura & Gebrewold, 1983; Berglund & Risberg, 1980). Moreover, blood
that is diluted with alcohol is oxygen deprived. Capillary constriction also
accompanies large doses of alcohol Together these effects of alcohol can produce
anoxia, cell loss, and thus brain damage. Passing out from drinking is a consequence
of anoxia.
Early studies likened chronic, abusive use of alcohol to premature aging— the
cognitive deficits associated with abusive use of alcohol were said to mirror the
deficits associated with aging (Courville, 1955, Wechsler, 1941). Results from recent
studies question this hypothesis, however suggesting that young and old abusive users
(alcoholics) perform similarly and that the impairments in each case are not the same
as those that most prominently accompany aging (Ryan & Butters, 1980, Glosser,
Butters, & Kaplan, 1977)
160
Kapur & Butters (1977) report that visual-spatial and encoding-organization
deficits are major consequences of anoxia associated with alcohol use. Visual-spatial
ability (Gv) has been shown to be a vulnerable ability— after mid-adulthood, it declines
with age (Horn, 1972; 1985). It is also correlated with G f functioning. In fact, Gv
and Gf have not been adequately separated in some studies and indicators of the two
are often referred to interchangeably. There is thus some suggestion the rGf residual
of Gf that here was found to be related to indications of abusive use of alcohol is the
component Kapur and Butters found to related to anoxia associated with alcohol use.
The residual component of Gc is also a black hole. It is not the reasoning of
Gf and thus not the processes of concentration, cognitive speed, and short-term
memory of which Gf is comprised. Further, it is not the part of short-term memory
that is independent of Gf. But we have no cognitive variables to part out to point to
the processes of this residual.
Whatever this residual is, it is here found to be positively related over the life
course to influences associated with the extent of one's formal education and one’s
self-proclaimed commitment to seeking information, tolerance for disorder, optimism,
and disdain for belief that outside forces (luck, God, and powerful others) control
one's life. These influences also relate to the part of residual Gc that is independent of
age
Together these results suggest that the people who retain and improve their
crystallized knowledge over the course of adulthood are people who acquire belief in
161
their own capabilities and retain optimism and tolerance in the face of the imperfect
conditions of life— factors that help them retain desire to improve their base of
knowledge and understanding. In contrast people who become rigid, pessimistic and
convinced that outside forces control their lives are not likely to engage in the
a^culturational activities that maintain and enhance crystallized knowledge. Such
rigid and pessimistic views and beliefs may be indicative of what has been described as
a disengagement process that can accompany aging.
Lifestyle Variables that Explain Age-Independent Cognitive Abilities
Several lifestyle indicators were found to related to cognition above and
beyond age suggesting that there are components of lifestyle that are important to
cognitive functioning at any age. Several of these variables also accounted for
significant amounts of the age-related ability differences and have been discussed
above. Those variables that did not account for age-related differences, but relate to
cognitive functioning at all ages are discussed in this section.
Subjects of all ages who performed well on the MPA task also report having a
Negative Attitude Toward TV Although it is recognized that attitude and behavior
are not always perfectly correlated, people who score high on this attitudinal factor
will most likely watch less TV and be less tolerant of poor quality TV than those who
score low on this factor. These people report that they would do something other
than watch TV in their free time. Watching less TV and seeking out quality
programming are indicative of a lifestyle that is more intellectually active— there is
162
more reading, game playing, engagement in culture, socializing, etc. A greater level
of intellectual and cognitive stimulation has been linked to sound short-term memory
functioning in several studies (Arbuckle, Gold & Andres, 1986; Craik, Byrd &
Swanson, 1985, DeCarlo, 1974, Gribbin et al., 1980; Schaie, 1983) An intellectually
engaged lifestyle could provide practice for memory tasks, confidence that one can do
well, and the motivation to be sufficiently “up” for the testing session.
Subjects who report being regular smokers in the present and/or having
smoked for many years of their adult lives scored relatively low on MPA even after
age and rGc are removed. While there is some evidence suggesting that nicotine acts
as a stimulant that increases neural firing and actually heightens memory performance
(Pomerleau, 1995), it is most probable that the present indicator of smoking behavior
reflects a “high risk” lifestyle rather than an index for nicotine saturation. Heavy
smoking is often accompanied by heavy alcohol consumption, poor lung capacity and
cardiovascular integrity, poor circulation, poor diet, and poor overall health in
general Self-reported poor health has been implicated in many studies as the culprit
for poor intellectual performance including short-term memory performance
(Clarkson-Smith & Hartley, 1990; Elias, Elias, & Elias, 1990; Perlmutter & Nyquist,
1990, Seigler, 1989; Seigler & Costa, 1983) These result suggests that there may be
a physiological component to short-term apprehension retrieval-an underdetermined
network of neurons that is affected by the accumulation of trauma to the brain.
163
Self Esteem was found to be a positive predictor of SLC functioning. It is
particularly important to be motivated and “up” for speeded tests. This Self Esteem
measure is suggestive of an overall self efficacy or an internalized personal control
belief system. A strong sense of self, a goal oriented efficacy, and confidence in one’s
abilities are involved in the motivation to work quickly and accurately throughout
cognitive assessment. High Education is a positive predictor of rGf and the use of
alcohol for escape purposes is a negative predictor of age-independent rGf variance.
Possible Age Cohort Explanations
There are lifestyle variables that account for cognitive decline which may be
indicative of age cohort influences— variables that operate in historical time affecting
one age group and not others It can be argued that all of the lifestyle variables
measured in this cross-sectional study are indicative of age cohort phenomena and
that no true age differences have been detected. This begs the question of
interpretation, however. To some extent observed age differences probably always
represent age cohort differences and age cohort differences probably always represent
age differences and it is difficult to separate the two (Botwinick & Arenberg, 1976;
Elder & Caspi, 1990; Horn & Donaldson, 1980; 1992; Horn, McArdle & Mason,
1980). This is not to say that results should always be interpreted as indicating age
differences without considering plausible age cohort explanations. Age cohort
interpretations do not lend themselves to inferences about development, but merely
describe the sample in terms of its diverse historical characteristics.
164
It is possible that there are certain age cohort differences operating in the
present sample. For example, results indicated that one’s interest and ability to
Confide in Others accounts for the decline of CON. The fact that older subjects do
not readily confide in others, seek out emotional help, and utilize community
resources (counseling, etc.) may reflect the fact that this aspects of psychological
health maintenance is a relatively new idea (historically). Older subjects may be of the
mind-set that one needs to solve emotional problems without the intervention of
outsiders and the idea of seeking “help” may seem unreasonable or unattainable.
Younger subjects, on the other hand, are more used to the idea of communicating
deep thoughts to others and relying on outside resources in times of emotional need.
The same could be said for the result that Tolerance for Disorder accounts for
rGc enhancement. Older subjects who raised children in a very controlled, strict
atmosphere may read the items of the Tolerance for Disorder scale as indicative of
self discipline or self regulation. Younger subjects who learned alternative child
rearing methods might interpret these same items as being indicative of harsh rigidity
and inflexibility.
A developmental interpretation of these lifestyle indicators is also plausible It
may be that as one ages, the interest or the need to confide in another during times of
emotional stress is lessened due to successful adaptation and utilization of more
sophisticated coping mechanisms. As one ages, it is possible that a tolerance for
disorder and a flexible attitude toward child rearing are diminished. Both
165
interpretations must be considered since it is difficult to separate developmental and
age cohort influences in these data
“Semi-Real ” Ability Differences. It is possible that there are variables that
can explain observed cognitive declines but have not been measured in this study.
Flynn (1984) analyzed the change in IQ scores (as measured by the Stanford-Binet
and the Wechsler scales) from 1932 to 1978 for various standardization samples of
white Americans. The sample (N=7,500) ranged in age from 2 to 48. Care was taken
to make the sample as representative of the larger population as possible by
eliminating gifted and retarded subjects. In order to curtail the retest effects of
practice, the time between testing was standardized to one year or greater. He
reported a rise in mean IQ of 13.8 points in this 46 year period, or, an average rise of
3 IQ points per decade Flynn attributes this gain in generational IQ to increased test
sophistication and enhanced educational achievement. Test sophistication is the
notion that Americans are getting more practice in taking standardized tests. If this
were true, the gain in IQ observed would be “semi-real.” Generations are not getting
smarter, they are just benefiting from improved performance tools in the same way as
athletes benefit from lighter running shoes; although performances on the clock really
do improve, no one would claim that the athletes are better runners per se today than
before the advent of such a shoe. Enhanced educational achievement is the
explanation that Flynn clings to most strongly. Not only do cognitive functions
benefit from education, but the educational elite will be more superior in terms o f a
166
whole range of other environmental variables, such as SES, nutrition, child care, and
test sophistication in particular.
In the same study, however, Flynn reports a decline in SAT verbal scores from
1963 to 1981 of 37.5 points in the general population (taking into account the larger
standard deviations in more recent years due to the broadening of the candidate
samples). An ETS advisory panel attributed this decline to societal traits such as less
demanding textbooks, less demanding school standards in general, rates of student
absenteeism commonly running above 15%, the erosion o f the nuclear family, and the
advent of television. Flynn concludes that the lower SAT verbal scores pose serious
problems to the explanations offered for the rise in IQ scores during the same period
in history. The causal explanations of increased test sophistication and educational
achievement are not sufficient if we are to believe that the rise in IQ scores from 1963
to 1981 is real. Flynn concludes that an historical understanding of the United States
that goes back to the turn o f the century is needed if we are to assert causal
explanations for the rise in IQ and the loss o f SAT verbal scores observed in
standardization samples and in the general population. There may be certain “semi-
real” ability differences operating in the present sample.
Summary and Conclusions
Income, Information Seeking, Attention to Healthy Diet, Secure Attachment,
Confiding in Others, a lack external belief systems, and a large Number o f Children
are lifestyle indicators that plausibly account for the aging decline of vulnerable
167
cognitive processes that have been shown to be involved in fluid ability functioning.
Low incidences of alcohol abuse resulting in anoxia results in less aging decline of
residual fluid ability than does a high frequency of such abuse. This suggests that the
part of fluid ability that cannot be accounted for by short-term memory,
concentration, cognitive speediness, and crystallized knowledge may be closely tied to
biological functioning The age-related enhancement of residual crystallized
knowledge is positively related to Information Seeking, Optimism, Tolerance for
Disorder, Education, and negatively related to external belief systems. The part of
crystallized knowledge that cannot be explained by fluid reasoning or short-term
memory is chiefly related to cultural immersion, goal directedness, and engagement.
Age-independent individual differences in short-term apprehension retrieval
can be explained by low Smoking Behavior and a Negative Attitude Toward TV
Concentration ability variance can be accounted for by a Tolerance for Disorder and a
lack of external belief systems. Cognitive speediness is related to a high Education
levels, a high level of Self Esteem and a lack of external belief systems. Education is
also related to residual Gf and Gc at all ages.
Because cross-sectional data were analyzed, strong causal inferences have not
been made It is not clear whether aspects of lifestyle cause intellectual functioning or
whether intellectual functioning promotes certain lifestyles. In order for causal
relations to be totally unbiased, reasonable time lags must occur (Heise, 1975; James,
Muliak & Brett, 1982; Reichardt, 1983, Strotz&Wold, 1960). When only cross
168
sectional data are available, Gollob and Reichardt (1985, 1987, 1990) have proposed
that plausible causal relationships be explored via latent longitudinal structural
models. Such exploration does not necessarily lead to causal inferences, but
facilitates the design and interpretation of future, longitudinal studies of possible
cause and effect relationships Appendix D includes a discussion of the application of
latent longitudinal models to the present data set
This latent longitudinal analysis served only to demonstrate some basic
measurement principles. In order for cause and effect relationships to be unbiased
there must be adequate time lags, measures must have adequate internal consistencies,
there must be a non-trivial relationship between the cause and effect variables, and the
autoregressive effect on the outcome variable must be small enough to accommodate
an exogenous cause— autoregressive effects (e.g., test-retest correlations) o f 1.0
indicate zero change (see Appendix D for a more complete discussion of these
analyses)
Because the subjects comprising this sample were mainly white, middle class,
and highly educated, the generalizability of the results reported may be limited.
Future studies of cognitive aging should include subjects representing many different
cultures, socioeconomic backgrounds, and acculturation levels. Based on the findings
of this study, cognitive assessments should include the purest measures possible of
fluid ability (Gf) and crystallized knowledge (Gc). In order to understand Gf and Gc
more thoroughly, assessments should also include measures of cognitive processes
such as short-term apprehension retrieval, concentration, cognitive speediness, and
visualization. Measures, when appropriate, must be given under power conditions
(untimed) and older adults must be allowed to utilize compensatory strategies.
Studies of cognitive aging should include lifestyle assessments consisting of measures
of attitudinal flexibility, detection of external and internal belief systems, broad
indicators of psychological health, indicators of cultural immersion, physical health
measures, and assessments of brain trauma resulting from anoxia.
170
References
Adler, A. (1964). Problems of Neurosis (1929V New York: Harper Torchbooks.
Adler, A. (1968). The practice and theory of individual psychology. Totowa, NJ:
Littlefield, Adams, & Co.
Altura, B. M., Altura, B. T., & Geberwold, A. (1983). Alcohol-induced spasms of
cerebral blood vessels: Relation to cerebrovascular accidents and sudden death.
Science. 220. 331-333
Anderson, J. E. (1939). The limitations of infant and preschool tests in the
measurement of intelligence. Journal of Psychology. 8, 351-379.
Amster, L.E, & Krass, H.H. (1974). The relationship between life crises and mental
deterioration in old age. International Journal of Aging and Human
Development. 5, 51-55.
Antonucci, T. C. (Ed.) (1976) Attachment: A life-span concept [Special Issue].
Human Development. 19
Antonucci, T. C. & Akiyama, H. (1987) Social networks in adult life and a
preliminary examination of the convoy model. Journal of Gerontology. 42, 519-
527.
Arbuckle, T. Y., Gold, D., & Andres, D. (1986) Cognitive functioning of older
people in relation to social and personality variables. Journal of Psychology and
Aging. I, 55-62
Arbuckle, T. Y , Gold, D., Chaikelson, A S, Schwartzman, A. & Andres, D. (1992)
Engagement, negative affect and introversion as correlates of change in
intellectual abilities over adulthood. Paper presented at the Cognitive Aging
conference. Atlanta, November.
Baltes, M.M. (1988). The etiology and maintenance of dependency in the elderly:
Three phases of operant research. Behavior Therapy. 19. 301-319.
Baltes, P.B. (1973). Life-span models of psychological aging: A white elephant?
Gerontologist. 13. 475-512.
Baltes, P.B. (1987). Theoretical propositions of life-span developmental psychology:
On the dynamics between growth and decline. Developmental Psychology. 23,
611-626
171
Baltes, P.B. & Baltes, M.M. (1990). Successful aging: Perspectives from the
behavioral sciences. Cambridge: Cambridge University Press.
Baltes, M.M., & Baltes, P.B (1986) The Psychology of control beliefs and aging.
Hillsdale, NJ: Lawrence Erlbaum Associates.
Baltes, P B & Schaie, K.W (1973). Prototypical paradigms and questions in life
span research on development and aging The Gerontologist. 15. 458-467.
Baltes, P. B., Reese, H. W., & Nesselroade, J R. (1977) Life-span developmental
psychology: Introduction to research methods. Belmont, CA: Wadsworth.
Bandura, A (1986). Social foundations of thought and action. Englewood Cliffs,
NJ: Prenice-Hall.
Bandura, A (1989) Regulation of cognitive processes through perceived self-
efficacy Developmental Psychology. 25, 729-735
Bell, R Q (1953) Convergence: An accelerated longitudinal approach. Child
Development. 24. 145-152
Berglund, M & Risberg, J. (1980). Reversibility in alcohol dementia. In H
Begleiter (Ed ), Biological Effects of Alcohol. New York: Plenum Press.
Birren, J E. & Schaie, K.W (1977). (Eds) Handbook of the psychology of aging.
New York Van Nostrand Reinhold.
Birren, J.E , Woods, A.M., & Williams, M.V. (1980). Behavioral slowing with age:
Causes, organization, and consequences In L.W Poon (Ed ), Aging in the
1980s: Psychological issues (pp 293-3081 Washington, DC: American
Psychological Association.
Blackburn, J. (1984). The influence of personality, curriculum, and memory
correlates on formal reasoning in young adults and elderly persons. Journal of
Gerontology. 39, 207-209.
Blumenthal, J A., Emery, C.F., Cox, D R., Walsh, M.A., Kuhn, C M., Williams, R.B.,
and Williams, R.S. (1988). Exercise training in healthy Type A middle-aged
men: Effects on behavioral and cardiovascular responses. Psychosomatic
medicine. 50, 418-433.
172
Botwinik, J. (1977). Intellectual abilities. In J.E. Birren & K.W. Schaie (Eds ),
Handbook of the psychology of aging (pp. 58-65V New York: Van Nostrand
Reinhold
Botwinik, J.A. & Arenberg, D. (1976). Disparate time spans in sequential studies of
aging. Experimental Aging Research. 2, 55-61.
Botwinik, J & Storant, M. (1974). Memory related functions and age. Springfield
MA: Thomas.
Brandtstadter, J. & Baltes-Gotz, B (1990) Personal control over development and
quality of life perspectives in adulthood. In P. Baltes and M. Baltes (Eds ),
Successful aging: Perspectives from the behavioral sciences. 1990. (p. 197-
224). Cambridge: Cambridge University Press
Brimm, O.G., Jr. & RyfT, C D. (1980). On the properties of life events. In P.B.
Baltes & O G. Brim, Jr (Eds ), Life-span development and behavior. 3, 368-
398. New York: Academic Press.
Broadbent, D E (1966). The well ordered mind. American Educational Research
Journal. 3, 281-295
Burt, C (1909) Experimental tests of general intelligence. British Journal of
Psychology. 3, 94-177.
Burt, C (1940). The factors of the mind: An introduction to factor-analvsis in
psychology London: University London Press. [New York: Macmillan, 1941]
Carroll, J B (1987) Jensen's mental chronometry: Some comments and questions. In
S. Modgil & C. Modgil (Eds ), Arthur Jensen: Consensus and controversy (pp.
297-307V New York: Falmer Press.
Carroll, J B (1989) Factor analysis since Spearman: Where do we stand: What do
we know: in R. Kanfer, P.L. Ackerman & R. Cudeck (Eds ), Abilities.
motivation, and methodology: The Minnesota symposium on learning and
individual differences (pp. 43-67). New Jersey: Lawrence Erlbaum Associates.
Carroll, J. B. (1993). Human cognitive abilities: A survey of factor-analvtic studies.
New York: Cambridge University Press.
Carroll, J B. (1985). Exploratory factor analysis: A tutorial. In D. K. Detterman
(Ed ), Current topics in human intelligence. Vol. 1 (pp. 25-58). Norwood, NJ:
Ablex.
173
Case, R. & Globerson, T (1974). Field independence and central computing space.
Child Development. 45, 772-778.
Cattell. R.B. (1943). The measurement o f adult intelligence. Psychological Bulletin.
40, 153-193
Cattell, R.B. (1957). Personality and motivation structure and measurement
Yonders-on Hudson, New York: World Book, 871-880.
Cattell, R.B. (1963). Theory of fluid and crystallized intelligence: A critical
experiment. Journal of Educational Psychology. 54. 1-22.
Cattell, R.B. (1965). The scientific analysis of personality. Baltimore, MD: Penguin.
Cattell, R.B. (1971) Abilities: Their structure, growth and action. Boston:
Houghton-Mifflin.
Cicirelli, V.G. (1980) Relationship of family background variables to locus of control
in the elderly. Journal of Gerontology. 35. 108-114.
Clarkson-Smith, L., & Hartly, A.A. (1989). Relationships between physical exercise
and cognitive abilities in older adults. Psychology and Aging. 4, 183-189
Clarkson-Smith, L., & Hartly, A.A. (1990). Structural Equation Models of
Relationships between exercise and cognitive abilities. Psychology and Aging.
5, 437-446
Cliff N. (1987). Analyzing multivariate data. New York: Harcourt, Brace,
Jovanovich, Publishers.
Cohen, J., & Cohen, P. (1983). Applied multiple regression/correlation analysis for
the behavioral sciences. Hillsdale, NJ: Erlbaum.
Colemann, J.C. (1964). Abnormal Psychology and Modern Life Chicago: Scott,
Foresman, p, 664.
Collins, N L., & Read, S. J. (1990). Adult attachment, working models, and
relationship quality in dating couples. Journal of Personality and Social
Psychology. 58, 644-663.
Cooley, C H. (1902). Human nature and the social order. New York: Charles
Scribner's Sons.
174
Cornelius, S. & Caspi, A. (1986) Self-perceptions of intellectual control and
aging. Educational Gerontology. 12, 345-357
Costa, P. T., Fozard, J L., McRae, R. R., & Bosse, R. (1976). Relations of age and
personality dimensions to cognitive factors. Journal of Gerontology. 31. 663-
669
Courville, C. G (1955). The effects of alcohol on the nervous system of man. Los
Angeles: San Lucas Press.
Craik, F.I.M. (1977). Age differences in human memory. In J.E. Birren & K.W.
Schaie (Eds ), Handbook of the psychology of aging Princeton, NJ: D Van
Nostrand.
Craik, F.I.M. & Bird, M. (1982) Aging and cognitive deficits: The role of attentional
resources. In F.I M. Craik & S. Trehub (Eds ), Aging and cognitive processes
(pp. 191-211). New York: Plenum.
Craik, F.I.M., Bird, M., & Swanson, J.M. (1987). Patterns of memory loss in
three elderly samples. Psychology and Aging. 2, 79-86.
Craik, F I M & Lockhart, R. S. (1972). Levels of processing: A framework for
memory research Journal of Verbal Learning and Behavior. 11. 671-684.
Crandall, V.C., Katkovsky, W., & Crandall, V.J. (1965). Children's beliefs in their
own control of reinforcements in intellectual-academic achievement situations
Child Development. 36. 91-109.
Decarlo, T.J (1074). Recreation participation patterns and successful aging. Journal
of Gerontology. 29. 416-427.
Domino, G. & Affonso, D. (1988). The inventory of psychosocial balance. Available
from George Domino, Department of Psychology, University of Arizona,
Tucson, AZ 85721.
Doppelt, J.E. & Wallace, W L. (1955). Standardization of the Wechsler Adult
Intelligence Scale for older persons. Journal of Abnormal and Social
Psychology. 51. 312-330
Dustman, R.E., Ruhling, R.O., Russell, E.M., Sheerer, D.E., Bonekat, H.W.,
Shigeoka, J.W., Wood, J.W., & Bradford, D. C. (1984). Aerobic exercise
training and improved neuropsychological function of older individuals.
Neurobiologv of Aging. 5, 35-42.
175
Dutta, R., Schulenberg, J.E., & Lair, T.J. (1986). The effect of job characteristics
on cognitive abilities and intellectual flexibility. Paper presented at the annual
meeting of the Eastern Psychological Association, New York, April.
Eisdorfer, C., Busse, E.W., & Cohen, L.D. (1959). The WAIS-R performance of an
aged sample: The relationship between verbal and performance IQs. Journal of
Gerontology. 14. 197-201.
Elder, G.H. & Caspi, A (1990). Studying lives in a changing society: Sociological
and personological explorations. In A.I. Rabin, R.A. Zucker, R.A. Emmons, &
S Frank (Eds ), Studying Persons and Lives. New York: Springer.
Elias, M R., Elias, J.W., & Elias, P.K. (1990) Biological and health influences on
behavior. In J.E. & W.K. Schaie (Eds ), Handbook of the psychology of aging
(pp. 79-102). New York: Academic Press.
El Koussey, A. A H. (1935). The visual perception of space. British Journal of
Psychology Monograph Supplement. 7, No. 20.
Elsayed, M. Ismael, A.H., & Young, R.J (1980). Intellectual differences of adult
men related to age and physical fitness before and after an exercise program.
Journal of Gerontology. 35, 383-387.
Eysenck, H. J. (1981). The general features of the model. In H. J. Eysenck (Ed ), A
model for personality development. New York: Springer-Verlag.
Eysenck, H.J. (1982). A model for intelligence. New York: Springer.
Featherman, D.L , Smith, J., & Peterson, J.G. (1990). Successful aging in a post
retired society. In P. Baltes and M. Baltes (Eds ), Successful aging:
Perspectives from the behavioral sciences. 1990. (p.50-93). Cambridge:
Cambridge University Press.
Flynn, J R. (1984). The mean IQ of Americans: Massive gains 1932-1978.
Psychological Bulletin. 95, 29-51.
Fodor, J. A. (1983). The modularity of mind: An essay on faculty psychology.
Cambridge, MA: MIT Press.
Galton, F. (1883). Inquiries into human Faculty and its development. London:
Macmillan.
176
Gatz, M & Good, P R. (1978) An analysis of the effects of the forced-choice
format of Rotter's Internal-External scale. Journal of Clinical Psychology. 34,
381-385
Gatz, M. & Karel, M.J. (1993). Individual change in perceived control over twenty
years. International Journal of Behavioral Development.
Gatz, M., Siegler, I.C., George, L.K., & Tyler, F.B. (1986). Attributional components
of locus of control: Longitudinal, retrospective, and contemporaneous analyses.
In M.M Baltes & P.B. Baltes (E ds), The Psychology of control beliefs and
aging Hillsdale, NJ: Lawrence Erlbaum Associates.
Glosser, G., Butters, N., & Kaplan, E. (1977). Visuoperceptual processes in brain
damaged patients on the digit symbol substitution test. International Journal of
Neuroscience. 7, 59-66.
Gold, D P , & Arbuckle, T.Y (1990). Interactions between personality and cognition
and their implications for theories of aging. In E. A. Lovelace (Ed ), Aging and
cognition Mental processes, self awareness and interventions, (pp 351-378).
Amsterdam: Elsevier
Goldman, M.S. (October, 1983) Cognitive impairment in chronic alcoholics: Some
cause for optimism. American Psychologist. 1045-1054.
Gollob, H.F., & Reichardt, C S (1987). Taking account of time lags in causal
models. Child Development. 58, 80-92.
Goodwin, J.S., Goodwin, J.M., & Garry, P.J (1983). Association between nutritional
status and cognitive functioning in a healthy elderly population Journal of the
American Medical Association. 249. 2917-2921.
Graham, J. & Hofer, S.M. (1993). Covariance estimation with missing values: An
application of the EM algorithm. EMCOV v2.2. University of Southern
California
Green, L.W. (1984). Modifying and developing health behavior. Annual Review of
Public Health. 5,215-236.
Gribbin, K , Schaie, K.W., & Parham, I. A. (1980). Complexity of lifestyle and
maintenance of intellectual abilities. Journal of Social Issues. 36, 47-61.
177
Grover, D R., & Hertzog, C. (1991). Relationships between intellectual control
beliefs and psychometric intelligence in adulthood. Journal of Gerontology. 46,
109-115.
Gruber, A.L., & Schaie, K.W (1986). Longitudinal-sequential studies of marital
assortativitv. Paper presented at the annual meeting of the Gerontological
Society of America, Chicago November.
Gurin, P & Brim, O.G., Jr. (1984). Change in self in adulthood: The example o f
sense of control. In P.B. Baltes, O G. Brim, Jr. (Eds ), Life-span development
and behavior (Vol 6, pp 281-334). New York: Academic Press
Gurin, P., Gurin, G., & Morrison, B.M (1978). Personal and ideological aspects of
internal and external control. Social Psychology. 41, 275-296.
Gustafifson, J.E. (1984). A unifying model for the structure of intellectual abilities.
Intelligence. 8. 179-203
Haier, R J., Siegel, B. V Jr., Nuechterlein, K H , Hazlett, E., Wu, J , Paek, J.,
Browning, H. L., & Buchsbaum, M. S (1988) Cortical glucose metabolic rate
correlates of abstract reasoning and attention studied with positron emission
tomography Intelligence. 12. 199-217
Hakstian, A.R. & Cattell, R.B. (1978). Higher stratum ability structure on a basis of
twenty primary abilities Journal of Educational Psychology. 70, 657-659.
Harris, R. & Frankel, L J (1977). Guide to physical fitness after 50. New York:
Plenum.
Harter, S. (1985). The Social Support Scale for Children and Adolescents Manual.
University of Denver, CO.
Harter, S. (1987). The determinants and mediational role of global self-worth in
children. In N. Eisenberg (Ed ), Contemporary issues in developmental
psychology (p. 219-242) New York: John Wiley & Sons.
Harter, S. (1990). Processes underlying adolescent self-concept formation In R
Montemayor, G.R. Adams, & T.P. Gullotta (Eds ), From childhood to
adolescence: A transitional period? Newbury Park, CA: Sage Publications.
Hebb, D O (1942). The effects 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.
178
Hebb, DO. (1949). The organization of Behavior. New York: Wiley.
Heckhausen, H. (1986). Achievement and motivation through the life span. In A.B.,
Sorenson, F.E. Weinert, & L.R. Sherrod (Eds ), Human development and the
lifecourse: Multidisciplinary perspectives (pp. 445-466). Hilsdale, NJ: Lawrence
Erlbaum.
Heise, D R. (1975). Causal analysis. New York: Wiley.
Hendrickson, A.E. (1982). The biological basis of intelligence. Part I: Theory. In
H. J Eysenck (Ed ), A model for intelligence (pp. 151-196). Berlin: Springer.
Hertzog, C., Fisk, A.D. & Cooper, B.P. (1994). Aging and individual differences in
the development of skilled memory search performance. Paper presented to
Cognitive Aging Conference, Atlanta, April 8th.
Hertzog, C., Dixon, R., & Hultch, D. (1990) Relationships between metamemory,
predictions, and memory task performance in adults. Psychology and Aging 5.
215-227.
Hertzog, C., Schaie, K W , & Gribbin, K. (1978). Cardiovascular disease and
changes in intellectual functioning from middle to old age. Journal of
Gerontology. 33, 872-883.
Hofer, S.M. (1994). On the structure of personality and the relationship of
personality to fluid and crvstalized intelligence in adulthood. Unpublished
doctoral dissertation, University of Southern California.
Hooker, K. (1992). Possible selves and perceived health in older adults and college
students. Journal of Gerontology. 47, 85-95.
Hooper, F H., Hooper, J O , & Colbert, K.K. (1984). Personality and memory
correlates of intellectual functioning: Young adulthood to old age. In J. A.
Meacham (Ed.l. Contributions to human development (Vol. 11, pp. 1-111).
New York: Karger.
Horn, J.L, (1965). Fluid and crystallized intelligence: A factor analytic and
developmental study o f structure among primary mental abilities. Unpublished
doctoral dissertation, University of Illinois.
Horn, J.L (1967). On subjectivity in factor analysis. Educational and Psychological
measurement. 27, 811-820.
179
Horn, J.L. (1968). Organization of abilities and the development of intelligence
Psychological Review. 75. 242-259
Horn, J L. (1978). Human ability systems. In P B. Baltes (Ed ), Life-span
development and behavior. Vol. 1 (pp 211-256V New York: Academic.
Horn, J.L. (1979). The rise and fall of human abilities. Journal of Research and
Development in Education. 12 (21. 59-77
Horn, J.L. (1982). The aging of human abilities. In B.B. Wolman (Ed.1. Handbook
of Developmental Psychology. New York: Prentice Hall.
Horn, J.L (1985) Remodeling old models of intelligence. In B.B. Wolman (Ed ),
Handbook of Intelligence. New York: Prentice Hall
Horn, J.L (1988). Thinking about human abilities. In J R. Nesselroade (Ed ),
Handbook of multivariate psychology (pp 645-685V New York: Academic
Press.
Horn, J.L. & Bramble, W J. (1967). Second order ability structure revealed in right
and wrong scores. Journal of Educational Psychology. 58. 115-122.
Horn, J.L. & Cattell, R.B. (1966), Refinement and test of the theory of fluid and
crystallized intelligence. Journal of Educational Psychology. 57. 253-270.
Horn, J.L. & Cattell, R.B (1967). Age differences in fluid and crystallized
intelligence. Acta Psvcholoeica. 26. 107-129
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.
Horn, J.L, & Donaldson, G. (1980). Cognitive development in adulthood. In O.G
Brim & J. Lagan (Eds ), Constancy and change in human development (pp.445-
529). Cambridge, MA: Harvard University Press.
Horn, J.L. & Donaldon, G. (1992). Age, Cohort, and time developmental muddles:
Easy in practice, hard in theory. Experimental Aging Research. 18(41. 213-222.
Horn, J.L. Donaldson, G , & Engstrom, R. (1981). Application, memory, and fluid
intelligence decline in adulthood. Research on Aging. 3, 23-84.
180
Horn, J.L. & Hofer, S.M. (1992). Major abilities and development in the adult
period. In R. Sternberg & C. Berg (Eds ), Intellectual development (pp. 44-91).
New York: Cambridge University Press.
Horn, J.L. & McArdle, J.J. (1980). Perspectives on mathematical/statistical model
building (MASMOB), in research on aging. In L.W. Poon (Ed ), Aging in the
1980's Washington, D C.: American Psychological Association, 503- 541.
Horn, J.L. & McArdle, J.J (1992) A practical and theoretical guide to measurement
invariance in aging research. Experimental Aging Research. 18, 117-144.
Horn, J.L. & Noll, J.G. (1994). A system for understanding cognitive capabilities: A
theory and the evidence on which it is based. In K.K. Detterman (Ed ), Current
Topics in Human Intelligence. Vol. 4.
Horn, J.L. & Stankov, L. (1982) Auditory and visual factors of intelligence
Intelligence. 6. 165-185
Horn, J L , Wanberg, K W , & Foster, F.M (1983). Guide to the Alcohol Use
Inventory Minneapolis, MN: National Computer Services, Inc.
Hurley, J.L & Cattell, R B. (1962). The Procrustes program, producing direct
rotation to test an hypothesized factor structure. Behavioral Science. 7, 258-
262.
Hutchinson, H., Tutchie, M., Gray, K., & Steinberg, D. (1964). A study of the
effects of alcohol on mental functions Canadian Psychiatric Association
Journal. 9, 33-42.
James, L.R., Muliak, S. A., & Brett, J.M. (1982V Causal analysis: Assumptions,
models, and data. Beverly Hills, CA: Sage.
James, W. (1892). Psychology: The briefer course. New York: Henry Holt & Co
Jensen, A.R. (1980). Bias in mental testing. New York: Free Press
Jones, B M. & Parsons, O A. (1972). Specific versus generalized deficits of
abstracting ability in chronic alcoholics. Archives of General Psychiatry. 26,
380-384.
Joreskog, K.G. & Sorbom, D. (1988). LISREL 7: A Guide to the program and
applications. Mooresville, IN: Scientific Software.
181
Kahn, R.L. & Antonucci, T.C. (1980) Convoys over the life course: Attachment,
roles, and social support. In P Baltes & G. O. Brim, Jr. (Eds ), Life-span
development and behavior (Vol. 3, pp. 253-286). New York: Academic Press.
Kaiser, H.F (1958). The variamx criterion for analytic rotation in factor analysis.
Psychology. 23, 187-200.
Kapur, N. & Butters, N. (1977). An analysis of visuoperceptive deficits in alcoholic
Korsakoff s and long-term alcoholics. Journal of Studies on Alcohol. 38. 2025-
2035.
Kaufman. A S. (1979). Cerebral specialization and intelligence testing New York:
Wiley
Kelly, T.L. (1928). Crossroads in the mind of man: A study of differentiable mental
abilities. Stanford, CA: Stanford University Press.
Kish, G.B. & Cheney, T.M (1969) Impaired abilities in alcoholism: Measured by
the General Aptitude Test battery. Quarterly Journal of Alcohol Studies. 30.
384-388.
Kleban, M.H., Lawton, M.P., Brody, E.M., & Moss, M. (1076). Behavioral
observations of the mentally impaired aged: Those who decline and those who
do not. Journal of Gerontology. 31. 333-339
Kogan, N (1990). Personality and aging. In J.E. Birren & K.W. Schaie (Eds ),
Handbook of the psychology of aging. San Diego: Academic Press.
Krause, N. (1986) Social support, stress and well-being among older adults. Journal
of Gerontology. 41, 512-519.
Labouvie-Vief, G. & Schell, D A. (1982). Learning and memory in later life: A
developmental view. In B. Wolman & G. Strieker (Eds ), Handbook of
developmental psychology (pp. 828-846). Englewood Cliffs, NJ: Prentice-Hall.
Lachman, M E. (1983). Perceptions of intellectual aging: Antecedent or consequence
of intellectual functioning: Developmental Psychology. 19. 482-498.
Lachman, M E (1985). Personal efficacy in middle and old age: Differential and
normative patterns of change. In G.H. Elder, Jr. (Ed ), Life-course dynamics:
Trajectories and transitions. 1968-1980 (p. 188-213). Ithaca, NY: Cornell
University Press.
182
Lachman, M E (1986) Locus of control in aging research: A case for
multidimensional and domain-specific assessment. Psychology and Aging. 1, 34-
40.
Lachman, M E., Baltes, P B., Nesselroade, J R., & Willis, S.L. (1982). Examination
of personality-ability relationships in the elderly: The role of contextual
(interface) assessment mode. Journal of Research in Personality. 16, 485-501.
Lachman, M E., Steinberg, E., & Trotter, S. (1987). Effects of control beliefs and
attributions on memory self-assessments and performance. Psychology and
Aging, 2, 266-271.
Lachman, M E. & Leff R. (1989) Perceived control and intellectual functioning in
the elderly: a 5-year longitudinal study. Developmental Psychology. 25(51. 722-
728.
Langer, E.J. (1983). The psychology of control. Beverly Hills: Sage Publication
LaRue, A., & Jarvik, L. (1982). Old age and biobehavioral changes. In B.B
Wolman (Ed ), Handbook of Developmental Psychology (pp. 791-806).
Englewood Cliffs, NJ: Prentice- Hall.
Lefcourt, H.M. (1982). Locus of control: Current trends in theory and research (2nd
ed ). Hillsdale, NJ: Lawrence Erlbaum Associates
Levinson, H. (1974). Activism and powerful others: Distinctions within the concept
of internal-external control. Journal of Personality Assessment. 38, 377-383
Liben, L.S. (1977). Memory from a cognitive-developmental perspectives In
Overton & Gallagher (Eds ), Knowledge and Development. Vol. 1. Advances in
Research andTheorv (pp. 149-203) New York: Plenum.
McArdle, J.J. (1984). A dynamic and structural equation model of WAIS-R abilities.
Presented at the annual meeting of the Society of Multivariate Experimental
Psychologists, Evanston, IL.
McArdle, J.J. & Horn, J.L. (1983). Validation bv systems modeling of WAIS-R
abilities. National Institute of Aging
McArdle, J.J., Horn, J.L , & Goldsmith, H.H. (1984). A structural equation model
of the genetic components of Gf and Gc.
183
McArdle, J.J., Goldsmith, H.H., & Horn, J.L. (1981).Genetic structural equation
models of fluid and crystallized intelligence. Behavior Genetics. 60. 607-620
McDowd, J.M. & Fillion, D.L. (1992). Aging, selective attention, and inhibitory
processes: A psychophysiological approach. Psychology and Aging. 7, 65-71.
McRae, R.R. & Costa, P.T Jr. (1987). Validation of the five-factor model of
personality across instruments and observers. Journal of Personality and Social
Psychology. 52, 81-90.
Miller, J., Slomezynsk, L.M., & Kohn, M L. (1987). Continuity of learning
generalization through the life span: The effect of job on men's intellectual
process in the United States and Poland. In C. Schooler & K.W. Schaie (Eds ),
Cognitive functioning and social structure over the life course (pp, 176-202).
New York: Ablex.
Nehrke, M.F., Hulicka, I.H., & Morganti, J. (1980). Age differences in life
satisfaction, locus of control, and self-concept International Journal of Aging
and Human Development. 11. 25-33.
Neimark, E D. (1982) Adolescent thought: Transition to formal operations. In
Wolman (Ed ), Handbook of Developmental Psychology (pp 486-502).
Englewood Cliffs, NJ: Prentice-Hall.
Noll, J.G (1993). Gender invariance of several personal control dimensions. Paper
presented at the American Psychological Society conference, Chicago, IL.
Noll, J G. & Horn, J.L. (1990). A new look at personal control: A comprehensive
theory of belief systems. Unpublished scale.
Nunn, C., Bergmann, K., Britton, P.G., Foster, E.M., Hall, E.H., and Kay, D.W.K.,
(1974). Intelligence and neurosis in old age. British Journal of Psychiatry. 124r
446-452.
Palmore, E.B. (1985). How to live longer and like it. Journal of Applied
Gerontology. 4, 1-8.
Pargament, K.I., Ensign, D.S., Falgout, K , Olsen, H , Reilly, B , Van Haitsma, K , &
Warren, R. (1990). God help me: (I): Religious coping efforts as predictors of
the outcomes to significant negative life events. American Journal of
Community Psychology. 18, 793-825.
184
Pargament, K.I., Kennell, J , Hathaway, W., Grevengoed, N., Newman, J., &
Jones, W. (1988). Religion and the problem- solving process: Three styles of
coping. Journal for the scientific study of religion. 27. 90-104.
Parker, E.S., & Nobel, E P (1980). Alcohol and the aging process in social drinkers
Journal of Studies on Alcohol. 4J, 170-178.
Parsons, O.A. (1975). Brain damage in alcoholics: Altered states of unconscious. In
M M Gross (Ed ) Alcoholic intoxication and withdrawal. New York: Plenum.
Paykel, E.S. (1983). Recent life events and depression. In J. Angst (Ed ), The origin
of depression: Current concepts and approaches (pp 91-106) Berlin:
Springer-Verlag.
Perlmutter, M., & Nyquist, L. (1990). Relationships between self-reported physical
and mental health and intelligence performance across adulthood Psychological
Sciences. 45(4). 145-55
Pomerleau, O F. (1995). Individual differences in sensitivity to nicotine: Indications
for genetic research on nicotine dependence. Behavior Genetics. 25, 161-178.
Poon, L W , Martin, P., Clayton, G.M., Messner, S., Noble, C.A., & Johnson, M A
(1992). The influences of cognitive resources on adaptation and old age. Int'l
Journal of Aging and Human Development. 34 31-46
Pope, C.R. (1982). Life-styles, health status and medical care utilization. Medical
Care, 20, 406-413
Powell, R.R, & Pohndorf, R.H. (1971). Comparison of adult exercisers and non-
exercisers on fluid intelligence and selected physiological variables. Research
Quarterly. 42, 70-77.
Reichardt, C.S. (1983, October). Assessing cause. Paper presented at the
Evaluation Network/Evaluation Research Society, Chicago.
Reite, M. & Boccia, M L. (1994). Physiological aspects of adult attachment. In M.
B. Sperling & W. H. Berman (Eds ), Attachment in adults: Clinical and
developmental perspectives (pp 98-127Y New York: Guilford Press.
Reker, G.T., Peacock, E.J., & Wong, P.T.P. (1987). Meaning and purpose in life and
well-being: A life-span perspective. Journal of Gerontology. 42, 44-49.
185
Rimoldi, H.J.A (1948). Study of some factors related to intelligence
Psvchometrika. 13, 27-46.
Rotter, J.B. (1966). Generalized expectancies for internal vs. external control of
reinforcement Psychological Monographs. 80.
Rotter, J.B. (1975). Some problems and misconceptions related to the construct of
internal control of reinforcement. Journal of Consulting and Clinical
Psychology. 43. 56-67.
Ryan, C. & Butters, N. (1980). Learning and memory impairments in young and old
alcoholics: Evidence for the premature-aging hypothesis. Alcoholism. 4, 288-
293
Salthouse, T.A. (1985). Speed of behavior and its implications for cognition. In J E.
Birren & K.W Schaie (Eds ), Handbook of the psychology of aging (2nd ed .,
pp. 400-426). New York: Van Nostrand Reinhold.
Schaie, K.W. (1979). The primary mental abilities in adulthood: An exploration in the
development of psychometric intelligence. In P.B. Baltes & O.G. Brim, Jr.
(Eds ), Life- span development and behavior, vol 2 (pp 67-115V New York:
Academic
Schaie, K.W. (1983) The Seattle longitudinal study: A twenty- one year exploration
of psychometric intelligence in adulthood. In K.W. Schaie (Ed). Longitudinal
studies of adult psychological development, (pp. 64-135) New York: Guilford
Schaie, K.W (1989). Perceptual speed in adulthood: Cross- sectional and
longitudinal studies Psychology and Aging. 4(4), 443-453
Schaie, K W. (1990). Intellectual development in adulthood. In J. Birren & K.W.
Schaie (Eds ), Handbook of the Psychology of Aging, (pp. 291-309) San
Diego: Academic Press.
Schaie, K.W , & Baltes, P.B. (1977). Some faith helps see the forest: A final
comment on the Horn and Donaldson myth of the Baltes-Schaie position on
adult intelligence. American Psychologist. 32. 1118-1120.
Schaie, K. W., Dutta, R , & Willis, S.L. (1991) Relationship between rigidity-
flexibility and cognitive abilities in adulthood. Psychology and Aging. 6, 371 -
383
186
Schmidt, F.L. & Crano, W D. (1974). A test of the theory of fluid and crystallized
intelligence in middle- and low-socioeconomic-status children: A cross-lagged
panel analysis. Journal of Educational Psychology. 66. 255-261.
Schooler, C. (1987). Cognitive effects of complex environments during the life span:
A review and theory. In C. Schooler & K.W. Schaie (Eds ), Cognitive
functioning and social structure over the life course (pp, 24-49). New York:
Ablex
Schultz, R , Heckhausen, J., & Locher, J.L (1991). Adult development, control, and
adaptive functioning. Journal of Social Issues. 47. 177-196.
Seligman, M E.P., & Elder, G.H., Jr , (1986) Learned helplessness and Life-span
development. In A.B. Sorensen, F.E. Weinert, & L. Sherrod (Eds ), Human
development and life course: Multidisciplinary perspectives, (p 377-428).
Hillsdale, NJ: Lawrence Erlbaum.
Siegler, I.C (1989). Developmental health psychology. In M L. Stroandt & G.R.
Vandenbos (Eds ), The adult years: Continuity and change. Washington, DC:
American Psychological Association, 119-142.
Siegler, I.C., & Gatz, M. (1985) Age patterns in locus of control. InE. Palmore,
E.W. Busse, G L. Maddox, J.B Nowlin, and I.C. Siegler (Eds ), Normal aging
III. (p. 259- 267). Druham, NC: Duke University Press.
Siegler, R.S. & Costa, P T. Jr. (1983) How knowledge influences learning.
American Scientist. 71. 631-638.
Spearman, C. (1904). "General Intelligence," objectively determined and measured.
American Journal of Psychology. 15, 201-293.
Sperling, M B. & Berman, W.H (1994) Attachment in adults: Clinical and
developmental perspectives. New York: Guilford Press.
Staats, S. (1974). Internal versus external locus of control for three age groups.
International Journal of Aging and Human Development. 5, 7-10.
Stankov, L. (1988). Single tests, completing tasks and their relationship to the broad
factors of intelligence Personality and Individual Differences. 9, 25-33.
Stankov, L ., Roberts, R. & Spilsbury, G. (1994). Attention and speed of test-taking
in intelligence and aging. Personality & Individual Differences. 17, 273-284.
187
Stelmack, R.M. (1991). Advances in personality theory and research. Journal of
Psychiatry & Neuroscience. 16. 131-138.
Strotz, R H. & Wold, H.O.A. (1960). Recursive versus nonrecursive systems: An
attempt at synthesis Econometrica. 28. 417-427.
Swinton, S. S., Sipple, T., Hooper, F.H. & Hougum, C. (1972). The role of short
term memory in the development of logical operations Technical Report No.
445, Eric Doc No. 156-708 (Wisconsin Research and Development Center for
Cognitive Learning, Madison.)
Tarter, R E. (1975). Psychological deficit in chronic alcoholics: A review.
International Journal of Addiction. j0, 327-368.
Thomae, H (1980) Cognitive theory of personality and adjustment to aging In J.E.
Birren, & R.B. Sloane The Handbook of mental health and aging (p 295-
309) Prentice-Hall, Inc.: Englewood Cliffs, NJ.
Thomson, G.H. (1916). A hierarchy without a general factor. British Journal of
Psychology. 8, 271-281.
Thurstone, L.L (1938). Primary mental abilities. Psychometric Monographs. No. 1.
Thurstone, L L (1847). Multiple factor analysis: A development and explanation of
The Vector of the Mind Chicago, IL: University of Chicago Press.
Trabasso, T. (1977). The role of memory as a system in making transitive inferences.
In Kail & Hagen (Eds ), Perspectives on the Development of Memory and
Cognition (pp 333-336), Hillsdale, NJ: Erlbaum.
Tryon, CM . (1933). On the nature of “speed” and its relation to other variables.
Journal of General Psychology. 8, 198-216.
Tyler, F B ., & Gatz, M. (1979). A constuctivist analysis of the Rotter I-E scale.
Journal of Personality. 47. 11-35
Undheim, J O (1987). The hierarchical organization of cognitive abilities: Restoring
general intelligence through the use of linear structural relations (LISREL).
Multivariate Behavioral Research. 22. 149-171.
Wallston, K.A., & Wallston, B.S. (1981). Health locus of control scales. In H.M.
Lefourt (Ed ), Research with the locus of control construct. Vol. 1 .
Assessment Methods 1pp. 189-243) New York: Academic Press.
188
Wechsler, D. (1941). The effect of alcohol on mental activity. Quarterly Journal of
Studies on Alcohol, 2, 479-485
Woodcock, R.W. (1990). Theoretical foundations of the WJ-R measures of cognitive
ability. Journal of Psvcho-Educational Assessment.
189
Appendix A : Summary Statistics for all Cognitive Variables and Lifestyle Indicators
used in Analyses
This appendix contains a table of summary statistics for all cognitive ability
variables for several age groupings. The appendix also contains a table of summary
statistics for lifestyle indicators used in Study 2 for several age groupings.
Table A -l. Means and Standard Deviations fo r Standardized (mean 0, std= I)
Cognitive Variables Across Several Different Age Groups
22-35
N=61
X(std)
36-45
N=90
X (std)
46-55
N=101
X (std)
56-65
N= 8 8
X(std)
66-75
N=150
X(std)
76-92
N=87
X (std)
Power
Letter
Series
73(96) 53(1.02) 24( 89) -03(96) -43(81) -56(78)
Common
Analogies
.07(91) .36(97) .34( 80) -.02(1.03)
34(1.03)
-.22(98)
Esoteric
Analogies
-.59(1.09) .15(1.03) 31(92) - 10(1.03) -.12(92) .19(90)
Vocabulary
Test 1
-.74( 89) -04(90) .07( 80) -21(89) .06(1.11) .57(93)
Vocabulary
Test 2
- 55( 99) 00(1.07) 16(94) -.17(88) .0 1 ( 1 0 1 ) 35(93)
Memory for
Paired
Associates
.77(1.07) .38(1.07) 30(1.08) .11(80) - 38( 61) -.72(77)
Slow
Tracing:
centimeters
6 8 ( 82) .41(86) 38( 89) .20(72) - 36( 8 6 ) -.92(97)
Speeded
Cross-out
81(1.03) 42(94) .30( 79) - 15(63) -.43(84) -.77(88)
Speeded
Letter
Comparison
95(80) .67(91) 43( 90) -.11(60) -.51(67) -.84(90)
Note: Cognitive variables standardized (mean=0, std=l) across the entire sample (N=577)
before age groups were formed
Table A-2. Means and Standard Deviations o f Variables Used in Analyses fo r Three
Age Groupings—Variables standardized (mean=0, std 1) over entire sample
(N=577)
22-45 N= 150 46-68 N=225 69-92 N = = 2 0 2
Mean Std Mean Std Mean Std
Age 36.78 6.50 57.78 7.36 75.54 4.94
Income .34 1 08 .25 .97 -.55 .69
Education . 1 2 8 6 01 1.05 - 1 0 1 0 2
# of Children -.54 .82 .30 .99 .06 .97
# of Years in Relationship -91 .42 -.04 .83 84 8 6
Attention to Healthy Diet -.40 1.04 .04 . 8 6 .26 1 . 0 2
Smoking Behavior -.25 95 .13 1.03 .04 .97
Use of Alcohol for Escape .08 1 05 .01 .96 -.07 1 . 0 1
Per Week Drinking Frequency - . 0 1 1 09 .1 2 1 . 0 2 - . 1 2 89
Ave. # of Drinks at one sitting 13 1 18 .0 2 .96 - . 1 2 8 8
Frequency Passing out Drunk . 2 2 1 . 0 2 .04 .99 - . 2 1 .95
Information Seeking -.43 1 07 . 1 0 .95 .2 1 .91
Passive Entertainment -.17 1 03 09 95 . 0 2 1 . 0 2
Negative Attitude Toward TV 0 1 1 1 0 .0 2 1.05 -03 8 6
Community Involvement -.42 1 . 0 0 - .1 1 93 .44 .90
Optimism .07 1.05 .1 0 1 . 0 2 - 16 92
Self Esteem - . 0 2 1.06 .06 1 . 0 2 -.05 93
Confide in Others .33 98 .04 1 . 0 0 -2 9 .94
Tolerance for Disorder .1 1 96 .09 1 . 0 2 - 18 .99
Secure Attachment -29 1.06 .07 1 . 0 2 14 89
Belief in God -.05 1.08 . 1 0 1 .0 1 .03 93
Belief in Nihilism/Chaos .05 97 -.13 .91 1 0 1.03
Belief in Luck -.26 96 -.13 .92 .34 1 .0 1
Belief in Powerful Others -.05 1 03 -.16 .91 . 2 2 1.09
Belief in Superstition - . 0 2 1 03 -.07 .89 .09 1 09
Power Letter Series (PLS) .61 1 0 0 .06 .94 -.53 .76
Vocabulary Test 2(VOC2) - . 2 2 1 07 .03 .91 .13 1 .0 1
Memory for Paired Associates .53 1 08 .13 .94 -.54 .70
(MPA)
Concentration (CON) .52 85 .23 .81 -6 4 96
Speeded Letter Comparison .78 8 8 .08 .81 -.67 80
192
Appendix B: Lifestyle Factors Used in Analyses
This appendix contains lifestyle factors used in analyses. The items that load
on each factor, the factor loadings, and the alpha reliability of each factor are included
for each domain of study
193
Physical Health Domain
Attention to Healthy Diet
___________X Alnha=.87
.79 I am careful to eat foods that are known to be healthful
.77 At daily meals, I strive to eat a low fat diet
.75 Throughout the week, I am careful to eat a balanced diet
.69 In my daily meals, I am careful to eat a diet that is high in fiber
.64 I substitute low fat items for high fat items in my diet
.60 At daily meals, I restrict the calories I eat day
53 I select foods that are low in salt and leave out salt from foods
.45 I am careful to restrict (to small amounts) the caffeine I drink
.37 I eat fresh vegetables
35 (-)I eat lamb, beef, or pork (including ham & bacon)
.34 I eat fresh fruit
Use of Alcohol for Escape
___________X Alpha=.89
.80 How often do you drink to get over feeling depressed?
77 How often do you drink because things pile up?
.76 How often do you drink to shut out the worries of the world?
70 How often do you drink to forget?
Activity Domain
Information Seeking
X Alnha=.72
65 I watch PBS drama, arts or information programming on TV
65 I seek out documentaries and similar nonfiction programs on TV
.62 I view documentary movies or TV shows
41 I read in (i.e parts of) weekly-issued or monthly-issued
magazines
.39 I read in (i.e., parts of) one or more daily newspapers
.39 I read nonfiction books and magazines to gain new information
.38 I discuss current events or news stories with others
.31 I read as a hobby
27 I watch TV news
.24 I listen to the radio to keep up with world and national events
23 I attend museums, botanical gardens, nature centers, etc.?
19 I go to a show of the performing arts (drama, music, dance)?
16 I follow sports
12 I vote in elections
194
Passive Entertainment
__________ X Alnha=.68
.58 How many hours o f TV do you watch in a typical week?
.58 I use TV more for entertainment than for information
51 I watch at least one weekly drama (e.g., Hill Street Blues) on
TV
.44 I watch at least one situation comedy on TV
.41 I consult a TV guide in advance to decide what to watch rather
than just flipping through the channels
.30 I find I watch just about anything (however poor the quality) on
TV
28 On average, how many movies do you watch per month—
theater, video
Negative Attitude Toward TV
__________ X Alnha=.91
. 8 6 Americans, in general, spend too much time watching TV.
.82 TV keeps children from engaging in more educational games and
activities.
.81 Watching too much TV can result in a loss of creativity and
innovative thinking.
77 The amount of TV children watch should be limited
77 Too many people use TV as their chief source of entertainment.
71 I think too many people use TV as a baby-sitter for their children
. 6 8 I believe TV keeps people from doing more worthwhile
activities
53 The quality of programming that children are exposed to should
be monitored.
.49 The quality of most TV programming is poor.
.39 If I have free time, I would do something other than watch TV.
195
Community Involvement.
__________ X Alnha=.78
67 I get actively involved in solving problems in my community
60 I am involved in motivating or organizing people in my
community (i.e. neighborhood watch programs, homeowners
assn., etc.)
.58 How many clubs or local organizations do you belong to?
56 I attend civic meetings (PTA, tenant or neighborhood group,
church group, local political group, etc.)
.50 I attend club or organization meetings
45 I assist charities or groups that provide help to people in my
town or community
.41 My neighbors approach me for assistance
.38 I have a good talk with people in my neighborhood
23 How similar do you feel you are to the people in your
neighborhood?
.21 How long have you lived in your present neighborhood?
Psychological Health Domain
Optimism
Response patterns goes from strongly agree Howl to strongly disagree thiehl
__________ X Alpha=.73
.69 I find little sense in living.
.47 If I had the courage I would end my life.
.47 I am of no use to anyone.
.43 (-)There are many things I enjoy in life.
36 When one is old it makes no sense to start new hobbies or
activities.
.33 I have never met anyone whom I truly admired
.23 With all of our technology, there is no need for anyone to work
very hard.
16 When I am intimate with another person I lose sense of who and
where I am.
.09 I think that certain groups of people or races in the world are
inferior to others.
196
Self Esteem
A . Alpha=.84
.62 (-)Some adults don’t like the kind of person they are
57 (-)Some adults are not very happy being the way they are
.55 (-)Some adults don’t feel like they are just as smart as other
adults
.55 Some adults are satisfied with themselves
.53 Some adults are not disappointed with themselves
49 Some adults never question whether they are a worthwhile
person
.47 (-)Some adults don’t feel that they are intelligent
40 (-)Some adults don’t like the way they are leading their lives
39 Some adults feel that they are very intellectually capable
27 When some adults don't understand something, it doesn’t make
them feel stupid
Confide in Others
A . Alpha=.88
.79 When something is troubling me I talk to someone about it
.78 I communicate my feelings and deeper thoughts to another
person
.77 I seek out those people with whom I can share my innermost
thoughts
.72 I let others know when I feel down
6 6 I confide in another (friend/relative) when I have problems
6 6 Everyone needs someone whom they can confide in
.54 It is a good idea to stay in touch with your feelings
.29 I have attended group, marital, family, or personal counseling
Tolerance for Disorder
Response patterns goes from strongly agree flow! to strongly disagree (high!
A . Aloha=.89
.70 A home should be spotlessly clean
.69 Children should never disobey.
.63 Children should stay clean at all times.
63 In company with adults, children should be quiet and listen.
61 In the home there should be a place for everything and
everything should always be in its place.
60 Children should never cause trouble.
59 A child should never talk back.
.57 In adult company children can be seen, but should not be not
heard
.53 People should never complain or cause trouble in public
.53 A child should keep his or her toys and clothes neat and orderly.
.36 People should always be neat and clean.
197
Secure Attachment style
Response patterns goes from strongly agree How) to strongly disagree (hiehl
___________X Alpha= . 8 6
.78 In relationships, I often wonder whether my partner really cares
about me.
.74 In relationships, I often worry that my partner will not want to
stay with me.
.67 In relationships, I often worry that my partner does not really
love me.
.65 When I show my feelings for people, I'm afraid they will not feel
the same about me.
55 I want to get close to people, but I worry about being hurt by
others.
.44 I find that others are reluctant to get as close as I would like.
Personal Control Domain
Belief in God
___________ X Alpha=.94
85 I can have a personal relationship with God.
.82 God can be approached directly by all believers.
.82 God knows all of my thoughts.
.82 My prayers are acts of personal communication with God.
82 God cares for each individual person.
.81 I get answers for many of my questions about the meaning of life
from my belief in God.
81 Although a person may not believe it, every person has direct
access to God.
.81 It is possible to have a one-to-one relationship with God.
.80 God comforts me during troubled times in my life.
.71 I believe in a higher power who looks out for the good of
humanity.
.69 My religion directs a great many of my thoughts.
. 59 If there were no God, there would be no adequate standard of
what is right and wrong.
198
Belief in Nihilism/Chaos
__________ X ____Alpha=.77
69 No one can say what is right or wrong, so you might as well do
what feels good.
. 6 8 No one can say what is right or wrong, so you might as well do
what feels good.
.35 The world is a chaotic, hostile and competitive place so I might
as well take what I can and not worry about being an angel.
.35 I'm not accountable to anyone because no one can really tell me
how I should behave.
.32 Life is a grab bag; one is well advised to grab what one can,
enjoy the moment and not worry much about consequences
Belief in Luck
__________ X Alnha=.88
.56 People who never get sick are just plain lucky
.52 Finding one's place in life is mostly a matter of luck.
50 People who never get sick are just plain lucky.
41 More often the good things in my life have happened out of
chance.
.39 Most of the disappointing things in my life have contained a
large element of chance.
.37 When trying to solve a problem, luck is as important as effort
36 The outcomes of our lives are mainly a matter of random
chance.
.35 Unlucky breaks often ruin even the best of plans.
32 Success in my work has been due to how lucky I've been.
.31 Although we are prone to think life has direction, living is mainly
a matter of moving through time in accordance with the whims
of change.
.30 Whether you recover from illness is mostly a matter of chance.
.28 If I have bad habits, it is because I've been unfortunate.
25 It is only chance that determines what facts we forget or
remember.
199
Powerful Others
__________ X Alpha=.77
.41 I prefer that others do much of my thinking and deciding for me.
.41 If important people were to decide they didn't like me, I
probably wouldn't make many friends.
3 5 When something bad happens to you, you try and remember
what immoral act you did to deserve it
.35 I feel that what happens in my life is mostly determined by
powerful people.
32 If something good happens to you, you try to remember what
good deed you might have done to deserve it
.30 When I get down or depressed, its usually because of things
other people have done to me
.24 I can't get over being depressed unless there is someone around
to cheer me up.
Superstition
_________ X Alpha=.77
. 54 Psychics can help us find people and objects that are lost.
. 51 Horoscopes help rne know what is going to happen to me in the
future.
.49 I think there are supernatural powers, ESP for one, that affect
our lives in ways \ve can't control yet.
.48 People who scoff at psychics, astrology, and omens just don't
realize how true these things are.
.44 I feel better about my relations with other people when I know
our astrological (birth) signs are compatible.
3 5 Omens (a black cat, Friday the 13th) indicate how your day will
turn out.
33 Lucky charms can make the difference between success and
failure.
.23 If a hotel assigned you to room 13 or a room on the 13th floor,
this would make you feel uneasy
Table C -l. Procrustes Reference Vector Structure fo r 145 Variables and 16 Factors
1 2 3 4 5
1 . . 645 .. 093 .. 020 .. O il -.0 1 0
2 . .645 .096 -..022 -.. 034 - . 0 0 3
3. .624 .043 ..030 -.. 012 - .0 4 1
4. .406 . 157 .. 013 .. 036 .031
5. .392 .014 .. 018 .. 089 .023
6 . .387 .206 -..002 -,.011 -.0 2 0
7. .377 . 085 .002 -,. 069 . 062
8 . .310 .017 -,. 026 . 103 . 102
9. .267 .281 - .091 -,. 039 . I l l
1 0 . .235 . 041 .. 002 . 036 - . 0 0 6
11. .231 -,.103 -,. 018 . 155 .004
1 2 . .188 - .053 -..028 .244 - . 0 0 8
13. .161 .064 - .099 .113 - . 0 0 9
14. . 124 . 106 . 013 .190 . 153
1 . . 123 .581 -,.119 - .036 . 023
2 . - . 1 7 5 .579 .075 -,.014 -.0 1 2
3. .059 .512 - .055 - .009 - .0 3 2
4. - . 0 0 7 .435 - .010 . 013 - .0 0 7
5. .227 .406 . 008 .074 . 147
6 . -.0 1 2 .300 - . 042 .031 - . 1 4 8
7. . 020 .284 -..006 -..065 - .0 3 2
1 . - .0 0 1 - .003 .857 - .017 - . 0 0 6
2 . -.0 1 1 .017 . 815 . 043 -.0 0 2
3. - . 0 2 5 - . 058 .805 . 049 .001
4. - . 0 5 6 .007 . 772 .009 .005
5. .037 - .005 .769 -,.046 . 050
6 . .009 . 097 .707 -,.009 - . 0 3 0
7. - . 0 8 6 .055 .681 .000 - . 0 2 8
8 . - . 0 3 6 .088 .531 .041 -.0 0 1
9. .048 - . 099 .492 - .101 . 100
10 . - . 0 1 5 - .232 .389 - .020 . 081
6 7 8 9 10 11
062 ..004 .. 030 - . 0 3 8 ..047 .. 027
037 -..032 -.. 012 - . 0 3 4 -..017 -., 007
013 .. 016 .. 036 .060 -.. 041 ..048
085 ..010 -.. 015 - . 0 1 4 .. 069 ..063
058 ..055 -.. 058 - . 1 0 4 ..053 -.. 127
077 ..018 , .045 - . 0 4 0 ., 016 , .092
O il , .077 ,.087 . 162 ..007 -,.006
025 -..004 -.. 076 - . 0 1 4 .. 038 . 014
007 -,. 032 .073 - . 0 7 6 -,.070 -,. 036
057 -,.013 , .067 .037 -,. 105 , . 065
067 -,.076 -,. 021 . 110 , . 108 .059
045 -,.054 .003 - . 0 1 9 .065 .123
044 .042 .094 - . 0 3 5 - .016 .056
010 .075 - . 028 - . 1 1 3 .127 - . 110
032 . 032 . 013 - . 0 1 8 -..077 - .044
045 -,. 028 , . O il .025 -,. 030 .002
096 -,.053 .061 . 081 .042 .102
077 - .072 .025 . 073 . 023 .121
069 .062 - . 105 - . 0 8 1 - .021 - . 071
074 .002 -,.028 .089 -..043 . 086
015 -,. O il .007 .080 -,. 030 .098
007 .002 .037 .000 , .019 .040
020 -,.054 -,.021 - . 0 4 5 - .007 .008
006 - .008 .009 - .0 1 0 - .029 - .013
002 .045 . 028 . 037 .002 - . 050
036 . 004 . 066 - . 0 0 4 -,.030 .045
029 -,.025 -..035 - . 0 3 9 .034 - .012
008 ,.002 -,.044 - . 0 0 9 -..023 -..040
112 .030 , .026 .032 -,.023 .001
108 - .007 - .024 . 000 . 033 .013
098 - .043 .011 .058 - .023 - .038
12 13
053 -.. 002
009 -,.013
047 , .032
057 -,.016
007 , .052
038 -,. 108
094 . 060
014 -,. 086
007 . 038
022 , . 006
103 -..059
094 .018
023 .111
019 . 004
042 . 017
002 .020
002 - .075
105 - . 056
124 . 015
076 - . 052
041 .086
001 - .016
026 .035
034 - . 015
035 . 029
013 -..027
004 .030
026 - .018
084 - .020
026 - . 040
035 . 077
15 16
- . 0 2 5 ..062
- . 0 1 7 -.. 004
. 034 .. 018
.016 -.. 041
.011 -.. 162
- . 0 6 1 .047
.080 -,.035
- . 1 3 8 -.. 047
. 048 . 008
- . 0 1 6 .068
. 015 , . 067
-.0 1 1 .029
- .0 2 1 -,.055
.008 -,.115
- . 0 5 1 . 048
- . 0 6 8 , . 017
.075 .078
.072 - .001
- . 0 7 2 .006
. 117 . 031
- . 0 8 9 . 046
.081 .054
.017 .020
.078 .004
- . 0 2 8 -,. 042
. 050 , .040
.004 -..002
. 139 .006
.021 -,.001
.024 -.. 003
.000 -,. 060
14
025
067
031
020
022
052
056
068
049
004
012
046
002
085
032
048
022
005
022
043
045
026
040
016
004
027
003
053
039
028
040
Table C-l (continued). Procrustes Reference Vector Structure fo r 145 Variables and 16 Factors
1 2 3 4 5 6 7 8
1 . .025 . 062 .. 048 .670 -..040 . 005 . 023 , . 054
2 . . 059 . 020 , . 016 .604 -,. 084 . 000 - . 0 2 4 , . 016
3. . 034 .038 -.. 026 . 581 . 058 .008 - . 0 8 1 . 022
4. - . 0 4 3 .035 -,. 037 . 555 ,.065 . 006 .029 -,. 025
5. . 012 .069 -..066 .500 -.. 028 . 021 . 031 , . 089
6 . .006 - .005 .002 .446 -,. 094 .041 . 029 .056
7. .230 - .136 .026 .405 - .039 - .008 .060 - .008
8 . .208 - . 048 . 067 .379 - .014 - . 041 - . 0 2 8 - . 078
9. .090 .021 -,.095 .233 . 149 . 083 .016 - . 054
10. . 097 . 061 .001 .208 . 072 . 014 .028 - . 045
1 . .037 - . 074 .045 - .008 .788 . 007 . 026 . 010
2 . - . 0 3 7 - .006 .007 - .055 .771 .063 - . 0 7 6 .008
3. - .0 1 0 - .046 .035 .014 .748 .029 - . 0 1 5 .077
4 . .159 - . 072 .043 - . 009 .693 - .051 .003 - .032
5. . 047 .088 - . 022 . 047 . 636 . 031 - . 0 4 5 - .012
6 . - . 0 2 5 - .014 - .020 . 047 . 600 .079 -.0 0 2 .045
7. .032 . 013 . 019 . 010 ..531 -,. O il - . 0 4 8 -..008
8 . - . 0 3 0 .043 .070 - .022 .448 - . 054 - . 0 6 9 - .060
9. .071 .076 .004 .031 .366 - .070 - . 0 2 6 - .021
1 0 . - . 0 8 9 - .083 - .005 - .079 .354 - .024 -.0 0 2 .021
11. . 152 .040 . 010 . 064 .336 - .033 - . 0 4 0 . 060
1 . .034 .032 .025 .002 .003 . 804 - . 0 4 8 . 002
2 . . 001 , . 049 ..004 . 045 .. 023 . 766 - . 0 1 3 -,. 037
3. .023 .087 .020 .044 .026 .755 - . 0 2 3 .007
4. - . 0 0 9 - .006 - .009 - .008 - .026 .702 .009 - .023
1 . . 006 . 046 -,. 004 .062 -..008 -,. 043 . 694 . 021
2 . .057 .001 .010 .027 - .058 - . 051 .470 - . 07 9
3. - . 0 6 4 . 014 041 . 041 055 . 023 .4 6 7 - . 114
4. - . 0 0 3 .011 .022 - .067 - .056 .000 .428 - .120
5. - .0 0 2 - .042 - .013 .014 .013 - .095 .364 .082
9 10 11 12 13 14 15 16
070 - . 0 6 3 ..034 -,.080 -.. 005 - . 0 4 5 -.. 019 .. 048
041 - . 0 6 1 . 045 - . 088 , .048 - .0 6 2 .025 ,. 063
134 . 003 . 035 . 029 ,.021 . 066 .001 -,. 004
086 .029 -,. 012 .134 -,. 133 - . 0 4 6 .095 -,.017
020 - . 0 4 0 -,. 028 .027 -.. 006 . 035 . 003 . 033
037 .027 . 014 .073 -.. 057 - . 0 4 7 . 074 . 054
080 - . 0 2 4 -,.083 . 021 , . 074 - . 0 2 9 .008 .000
071 - . 0 5 7 . 159 .057 , . 047 - . 0 0 7 .140 - .010
094 .022 - .070 - . 006 - . 039 . 011 .089 - . 066
210 .046 - . 095 .017 . 071 - . 0 2 7 .055 - .098
016 - .0 6 7 . 004 . 031 . 025 - . 0 3 0 - . 020 . 026
038 - .0 3 7 .012 - .041 - .019 .047 - .002 .000
041 - . 0 2 3 - .021 - .055 - . 031 .011 .016 - . 040
004 - . 0 3 6 - . 014 .027 - . 010 - . 0 5 4 . 025 . 022
003 -.0 2 0 .018 - . 022 . 022 .006 - . 106 . 075
008 -.0 2 0 . 021 .020 - . 041 .040 . 056 . 023
028 - . 0 1 6 . 013 . 116 . 087 - . 0 2 9 - . 073 .107
053 - . 0 0 6 .078 .059 .095 - . 0 3 6 .020 - .038
089 .117 - .004 - .051 - .031 -.1 0 2 .107 - .045
063 - . 0 8 2 .031 - .009 - .004 .004 - . 083 .126
034 . 103 . 039 .044 -..027 - . 0 0 8 .134 - . 053
007 - . 0 3 1 .022 - .063 -..045 .010 .098 . 020
O il .065 . 045 . 028 .. 055 - . 0 0 3 .045 -,. 021
014 - . 0 1 7 -,.028 .027 . 033 - . 0 2 3 .015 - .045
053 .020 - .038 .013 .077 - . 0 5 2 .046 - .010
011 -.0 0 2 . 022 .008 -.. 020 - . 0 5 4 . 032 , . 079
036 -.1 2 1 .006 - .057 . 021 - . 0 5 2 - .027 . 071
00 6 . 0 0 5 013 . 002 . O i l . 030 065 - . 010
016 -.0 1 0 -,.015 .025 -,.015 - . 0 3 6 .048 -,.047
072 - . 0 3 4 -..065 .051 ,.122 .031 .058 -,.019
to
o
!
Table C-l (continued). Procrustes Reference Vector Structure fo r 145 Variables and 16 Factors
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16
6 . - . 0 6 2 .090 - .063 -,.086 -,. 138 . 003 , . 327 .052 . 146 .023 ..047 . 055 .131 - .005 - .0 0 1 - .057
7. -.0 0 2 .031 -,.009 .041 -,.036 -., 016 , .228 .035 -.. 037 . 006 .. 005 , . 029 . 058 .073 . 041 .087
8 . .109 - .013 .002 .044 -..041 - .007 .158 - . 0 1 9 . 028 .008 .080 - .054 - . 004 .117 . 120 .017
9. .045 - .041 .064 .017 - .100 .069 .092 - . 0 5 2 - .070 -.1 1 3 - .028 - .033 .105 .046 .150 - .053
1 . -.0 0 1 .007 - .052 - .051 .025 - . 059 . 034 . 616 . 028 - . 0 1 5 -,.058 . O il - . 003 .017 - . 0 3 8 .078
2 . - . 0 0 6 -..004 - . 070 . 014 .021 - .066 -.. 052 .566 . 014 .017 -,. 031 . 009 . 046 - . 051 .069 - . 002
3. .031 - .031 .032 .004 - .019 .051 .066 .546 .013 - . 0 1 4 .073 - .075 .000 .073 -.2 4 5 - .062
4. - . 0 4 5 , .025 - .005 .032 ..024 -,.046 -,.065 .546 -,. 045 .010 -.. 031 , . 016 .053 - .076 . 032 . 103
5. - . 0 0 4 - .022 .022 . 053 .032 - .022 - .042 .530 , . 006 .000 -,. 021 .006 .040 - .052 . 036 . 001
6 . . 101 .059 .068 - .019 - . 036 .011 - . 013 .491 - .044 .028 -,.012 .047 .054 - .004 - .1 1 2 .004
7. .060 .028 .071 - .012 - .008 .087 . 074 .466 .042 - . 0 1 9 .043 - .038 - .070 .051 -.240 - .057
8 . - . 0 3 3 -..055 - . O il . 130 . 048 - .034 -,. 065 .399 -,. 083 .014 -,. 029 .006 -.. 074 . 003 .097 - . 072
9. .089 .055 - .009 - . 114 .000 .044 - . 101 . 392 . 037 - . 0 2 8 -,.012 -,.097 -,. 100 . 164 -.274 . O il
1 0 . .013 .029 .022 - .012 .016 - .043 - .019 .270 - .021 .011 - .120 - .005 .074 .026 -.269 .023
1 . - . 0 3 3 . 096 .032 - . 001 - . 0 0 3 - .034 - .022 - . 024 .794 - . 0 2 5 - . 046 - .002 . 009 - . 019 .040 - . 015
2 . . 012 . 018 - . 0 0 8 - . 012 - . 0 1 7 - .041 -..018 . 049 .781 - .0 2 2 - . 030 - . 008 . 027 . 013 .017 - . 088
3. - . 0 0 9 .001 - . 0 4 9 .011 .011 - .005 - .037 - .007 .774 .009 .004 .014 .008 .007 .076 .027
4. .003 .006 - . 0 4 4 .088 .034 .005 -,.023 -,.091 .716 .010 -,.041 - .043 .007 - .049 .096 -..021
5. .064 .048 -.0 2 1 .063 - . 1 0 3 .024 .015 - .010 . 660 - . 0 4 5 - . 009 .013 .047 - .035 .026 - . 016
6 . - . 0 3 3 .065 .015 . 021 . 037 .008 -..030 . 004 .656 .034 - .036 . 006 . 055 .084 .008 -..086
7. - . 0 1 8 - .004 .070 - .048 .055 -,.032 -..089 .062 .536 . 050 .061 .018 .050 .022 - .049 -,. 046
8 . - . 0 6 6 - .012 .024 - . 097 . 076 ..012 ..032 -,. 044 .287 . 185 ..042 -,. 021 - .063 -,.069 -,. 016 ..176
1 . - .0 1 1 .081 .032 .043 -,.041 .049 .040 .047 - .010 .698 - .053 .000 - .062 .054 .023 .021
2 . .091 - .097 -,.026 .044 - .004 .015 .010 - .015 - .004 .687 .030 - .014 .043 .010 .060 .021
3. - . 0 0 7 .024 .081 - .047 - .046 - .010 .011 . 006 .055 .634 - .048 .052 - .030 .063 - . 014 - . 0 6 7
4. -.0 0 2 .003 -.. 042 . 031 .030 - .026 . 025 . 045 .075 .630 - .015 -,.001 .100 - .003 - . 105 - . 0 2 5
5. .005 .058 -,.002 - .047 . 023 .005 - .050 - .015 - . 070 .610 - .033 -..041 - .041 - .050 . 101 .040
6 . .060 - .053 . 001 .069 .042 .029 -..030 - . 025 .036 . 596 .081 . 026 . 001 .006 -,. 069 .001
7. .038 - .071 -,.015 . 091 .040 .013 - .054 - .019 . 046 . 592 .020 -.. 038 .040 .082 -,. 117 .051
8 . - . 0 4 6 .001 - .052 .017 .022 - .047 - .069 - .014 .070 .565 .005 .016 .113 - .049 - .078 - . 0 2 5
9. .043 - .059 -..029 - . 078 .002 - .053 . 032 . 003 . 092 .533 - .043 -,.028 - .038 - .040 .021 . 064
10. .022 .087 -,.017 - .079 -..058 - .016 - .080 - .027 - .035 .525 .054 -..033 - .040 - .072 , . 159 . 098
1 1 . .041 -,.116 .015 -,.070 -,. 055 .089 -.. 035 .019 -,. 021 .364 -..027 .020 -..027 -,. 025 .069 . 163
202
Table C-l (continued). Procrustes Reference Vector Structure for 145 Variables and 16 Factors
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15
1 . . 000 .. 025 ..009 -.. 033 .061 -..019 -..011 -..020 -,.087 -.. 014 .783 .016 .001 -,. 026 , . 115
2 . .041 .018 -..037 -,.020 .048 .013 ,.069 ,004 , . 000 . .006 '.741 .010 .051 - .033 .064
3. . 006 .031 -,. 042 .013 .017 - .038 - .024 - .001 - .073 - .009 .669 .050 .075 . 034 . 001
4 . . I l l . 082 . 018 - . 071 - . 0 3 7 - . 003 . 001 - .043 - . 009 . 041 .651 .019 .089 - .043 . 044
5. .040 .005 -..010 -.. 173 . 059 . 023 .058 -,. 024 . 106 - . 005 . 548 - . 0 2 8 - .047 - . 004 . 146
6 . - . 0 1 4 . 097 ..015 , . 030 .011 -,.077 -,.046 . 067 .007 -.. 044 .436 -.0 2 2 - . 001 . 009 . 129
1 . .016 - .008 ,.027 .004 - . 0 2 8 .008 -,.006 .033 - .005 .011 . 030 .848 - .007 - .025 - .017
2 . .017 - .043 - .001 - .018 - . 0 4 6 .039 - .022 - .014 - .014 .018 - .014 .819 - . 019 .021 - .057
3. - . 0 3 4 - . 029 .024 - . 001 - .0 2 0 . 017 . 035 - . 025 .002 - . 010 .014 .816 - . 006 .020 - . 017
4 . . 004 . 022 .012 - . 003 . 038 . 038 . 044 . 031 . 012 - .021 - . 021 . 815 . 036 .012 - . 035
5. - . 0 0 3 - .027 -..013 -,.024 . 023 .013 .048 -,. 007 . 064 .044 - . 033 . 815 . 022 .016 - . 040
6 . .015 - .020 . 006 .044 .023 - .020 - .010 .030 - .024 - . 014 .036 .813 - .027 .021 .030
7. - . 0 2 6 - .020 .029 - .030 .025 - .005 - .033 - . 032 .013 .022 . 014 .812 - . 017 .003 - .020
8 . -.0 0 2 - . 057 .071 - . 043 - . 0 1 6 .003 - . 012 . 043 . 004 . 021 .025 .809 - .036 - .016 - . 012
9. . 048 .016 -,. 001 . 053 . 004 . 034 .036 -,.016 - . 026 .013 - . 058 .797 .027 - .014 - . 037
10. - . 0 3 2 .008 -,. 058 . 007 . 018 .011 - .011 -.. 045 . 040 . 007 . 019 .706 - .044 - .015 . 007
11. .002 .000 -,.072 .081 .043 - .095 -..021 -,.015 - .056 - .048 . 045 .688 - . 126 . 013 .086
1 2 . - . 0 0 5 .017 -,.068 . 064 .034 - .038 - .032 - .067 - .031 - .109 - . 015 .586 - .043 .081 .057
1 . . 000 . 052 . 048 .030 .015 - . 002 .013 - .098 .007 - .012 .066 - . 0 3 3 .685 .015 - .085
2 . - . 0 2 3 - . 005 . 033 .041 -.0 1 2 . 000 .017 - . 056 . 025 .045 . 044 . 033 .683 .048 - .085
3. .005 - .081 .015 . 004 - .0 1 0 . 062 . 149 . 085 . 087 .020 . 032 - . 0 9 0 . 346 .086 . 131
4. - .0 8 4 .056 -,.027 -..034 . 163 . 062 . 151 . 060 .014 .072 - .005 - . 1 2 7 .346 .055 -,. 063
5. .127 - .061 .034 - .023 - . 0 9 8 .012 - .038 .023 .007 - .074 .027 - . 0 4 0 .319 .134 .042
16
049
047
055
064
028
008
028
005
029
018
026
003
029
047
017
047
012
061
016
039
057
084
104
N>
O
U>
Table C-l (continued). Procrustes Reference Vector Structure fo r 145 Variables and 16 Factors
1 2 3 4 5 6 7
1 . - . 0 1 5 .024 . 050 .024 .027 .056 - . 048
2 . - . 042 -,.014 -,.021 .025 -..005 -..055 . 099
3. . 018 .008 , . 039 , .014 -,.059 - .020 - . 029
4 . .026 .016 -,.085 .103 .041 - .007 - .062
5. . 123 .020 .059 - .012 - . 112 . 090 .016
6 . - . 0 5 3 .019 -..043 - . 007 .020 - . 077 .011
7. .045 . 024 .. 025 .041 -..043 .047 - . 003
8 . . 046 . 169 .078 . 005 ,.009 .005 . 054
9. .021 - .048 -..032 - .033 .049 - .059 - .034
10 . - . 0 2 4 - .028 - .030 - . 093 .035 - . 009 . 028
11 . .034 - .098 - .045 - .062 - . O il .014 . 034
12 . - . 0 2 7 - .071 - . 051 .002 - .020 .037 . 054
13. -.0 2 2 -,.027 -,. 071 .067 . 013 . 060 .052
1 . - . 0 6 2 .018 - .042 .054 .070 - . 0 2 7 -,.007
2 . .043 - . 104 .071 .068 . 040 . 049 . 106
3. - . 0 4 7 - . 021 . 009 . 024 . 000 . 039 - . 020
4 . . 002 .026 . 081 .067 - . 027 . 024 . 113
5. .022 .012 - .003 .053 . 001 .024 - .094
6 . - . 0 1 4 .052 .117 .009 -,.044 .038 ..092
7. .011 .005 .071 . 009 .011 .035 . 086
1 . . 024 .023 - .001 - .029 -,.024 .029 -,. 014
2 . - . 0 3 5 . 125 .028 -,.026 -.. 027 - . 0 6 2 -.. 078
3. .020 - .028 - . 036 .002 .105 .038 . 069
4. - . 0 6 8 .067 - .029 .002 .092 .010 -,.053
5. .036 .097 .013 .039 - . 077 - . 0 6 1 .022
6 . - . 0 8 0 - . 049 . 002 .019 . 017 - . 0 6 2 -.. 039
7 . - . 0 2 4 - .022 .030 - . 001 . 023 . 022 . 059
8 . -.0 1 0 .028 .012 - .003 . I l l .008 . .093
9 10 11 12 13 14 15 16
- . 0 5 2 . 022 . 026 , .004 - . 052 .562 .061 - . 179
-.011 . 019 .008 -.. 022 .013 . 515 .124 - . 018
- . 0 0 3 .007 - .004 -,.022 - .134 . 500 .103 - . 124
.013 - .047 - . 037 , . 020 . 027 .406 . 092 .083
.011 - .003 - . 103 . 084 . 043 . 389 .044 .028
. 025 . 014 . 045 . 063 . 049 .374 .215 . 124
-.0 0 1 . 022 ,. 006 -.. 025 , . 145 .364 . 174 .045
.003 - .009 . 039 ..050 -,.066 .351 .078 .014
- . 0 2 5 - .032 - .037 , . 001 .021 . 323 .147 . 119
. 010 . 012 .046 -.. Oil .300 .311 .025 - . 014
- . 0 2 8 - .040 - . 093 . 062 . 090 . 300 . 129 . 026
- . 0 0 7 - .053 .003 -,.047 - . 009 .278 . 131 .110
-.0 1 1 .055 - . 002 , . 046 .115 .253 . 192 . 077
- . 0 1 5 .052 .041 - .072 - .093 .101 .412 .115
.031 .022 .023 - .045 .019 . 093 .411 .055
. 006 - . 058 . 039 . 115 . 028 . 061 . 352 .251
- . 0 1 3 . 009 .087 - .071 - . 004 . 095 . 348 . 073
- . 0 1 7 - .117 .070 . 046 .036 . 116 .317 .198
. 108 .040 .108 -,.014 - .057 .152 .301 . 056
. 059 - . 022 .026 -.. 007 . 023 . 134 .241 . 112
.019 - .005 - .094 -,.089 .012 - .006 .219 .539
. Oil .054 - . 113 . 088 . 005 - . 028 - . 010 . 509
- . 0 2 9 .020 - .021 -,. Oil . 033 . 009 .278 .485
- . 0 7 1 .050 - .052 -,.002 .117 .063 - .010 .479
- . 0 1 6 .098 - .024 .105 - . 008 .031 - .089 .443
. 034 . 038 .019 -,. 016 .114 . 059 .231 . 345
- . 043 .034 - .033 - . 067 . 042 . 024 .384 . 334
.011 .066 .035 .064 - .043 .067 .312 .230
8
020
114
010
080
006
049
015
019
053
026
028
045
074
214
065
094
026
038
000
132
054
038
058
020
030
017
009
009
204
u > ro
I
C-l (continued). Procrustes Reference Vector Structure for 145 Variables and 16 Factors
I n f o r m a t i o n S e e k in g 7 = O ptim ism 13 = B e l i e f i n N i h i l i s m / C h a o s
P a s s i v e E n t e r t a i n m e n t 8 = S e l f E steem 14 = B e l i e f i n Luck
N e g a t i v e A t t i t u d e Toward TV 9 = C o n fid e i n O th e rs 15 = B e l i e f i n P o w e rfu l O th e r s
Community In v o lv e m e n t 10 = T o l e r a n c e f o r D i s o r d e r 16 = B e l i e f i n S u p e r s t i t i o n
A t t e n t i o n t o H e a l t h y D ie t 11 = S e c u re A tta c h m e n t
Use o f A lc o h o l f o r E scap e 12 = B e l i e f i n God
205
206
Appendix D: Causal Inferences
This appendix includes a detailed discussion of the latent time lag model
proposed by Gollob and Reichardt (1987). This model is used to estimate the
likelihood o f a causal interpretation o f lifestyle variables influencing cognition and to
demonstrate some basic measurement principles.
There are three major principles of causal effects that need to hold true if
causal estimates are to be interpreted as unbiased. The first principle is that causes
take time to exert their effects-values of a variable can be caused only by values of
prior variables (Heise, 1975; James, Muliak & Brett, 1982; Reichardt, 1983; Strotz &
Wold, 1960). True cause always takes time to occur. Secondly, a variable can be
caused by prior values o f the same variable Effects of this kind are called
autoregressive effects. A causal model that omits an autoregressive effect is implicitly
assumed to have an autoregressive effect of zero. The third principle is that effect
sizes can vary as a function of the length of time lag between a cause and the time for
which its effect is assessed. That is, different time lags typically have different effect
sizes For example, the effects of taking an aspirin on reducing headache pain may be
zero a few minutes after ingestion, may be substantial after 30 minutes, may be near
maximum after 2 hours, and may be zero again after 24 hours. It follows that
understanding an effect size depends in part on knowing the time lag and knowing
something about the expected effect. It is often assumed that the effect one variable
has on another occurs at only one unique time lag and other possible time lag effects
207
are ignored. While it is true that for any given substantive issue some time lags are
more important to study than others, it should not be assumed that one time lag is
correct and that all others are erroneous In the aspirin example, drastically different
conclusions could be reached depending in which time lag is studied. Therefore, it
has been suggested (Gollob & Reichardt, 1987) that the effect of many different time
lags be studied together in order to get the “range” of possible causal parameter
values. This “range” should be reported as the expected confidence interval for
causal effects
In cross-sectional designs, it is difficult to specify how values of variables are
related to prior values of variables, to allow autoregressive effects, and to study a
range of time lags on causal estimates Cross-sectional models that assert causal
relationships at one point in time, omitting these three principles, report extremely
biased estimates of effects However, Gollob and Reichardt (1985; 1987; 1990) and
Reichardt and Gollob (1984) have proposed fitting a “latent longitudinal model” when
only cross-sectional data are available These models are similar to traditional
longitudinal models except that latent (unobserved) variables are used in place of
observed variables at Time 1. Figure D. 1 depicts the use of this model when
estimating the effect of a lifestyle on a cognitive ability variable. In the Figure the
measured variables at Time 2 are enclosed by boxes, and the latent variables at Time 1
are enclosed by ellipses The size of the cross-variable effect is denoted by P, the
sizes of the autoregressive effects are denoted by < J > ‘s, the sizes of the effects of the
Figure D . 1
Latent Longitudinal Model (Gollob & Reichardt, 1987)
208
Ability
Time 1
Ability'
Time 2
Lifestyle
.Time 2 >
Lifestyle
.Time 1 >
Observed
Lifestyle
4*= Observed Correlation between Lifestyle Variable and Cognitive Ability Variable
O = Autoregressive Effects
P = Causal Effect: Lifestyle Variable at Time 1 Causes Cognitive Ability' Variable at Time 2
a = Alpha Reliability (Internal Consistency) of Measure
5 = Error of Measurement (1- a)
209
disturbances (error terms) are denoted by 8’s, and the size of the correlation at Time
1 is denoted by i(/. This latent longitudinal model is not “identified.” This means that
the data do not provide enough information to obtain unique estimates of the model
parameters. To identify this model the values of some of the model parameters must
be fixed. These parameters can be fixed by using a combination of substantive theory
and empirical results from other studies Once the latent longitudinal model is
identified, it can be fit to cross-sectional data and causal estimates will be less biased.
One way to ensure that the model is identified is to fix the values of the
autoregressive effects Since choosing plausible values for these parameters can be
difficult, it has been suggested that these parameters be fixed to estimates o f the test-
retest correlations rather than directly fixing the autoregressive effects themselves
This is possible because the autoregressive effects can be expressed as functions of
test-retest correlations (Gollob & Reichardt, 1987). If test-retest correlations are not
known, a great deal of uncertainty exists about the values of these parameters. In this
case, it is suggested that a variety of different sets of test-retest correlations be
imposed This way, a set of estimated effect sizes will arise and the true effects size
will most likely be contained in this interval. Reichardt and Gollob (1984) have
recommended that a variety of models be fit in an effort to bracket the size of the true
effects. These brackets provide a realistic assessment of the uncertainty concerning
the values of the causal effects
210
There are several conditions that must be present in order for cognitive
changes to show up as such on measurement instruments (Horn & Donaldson,
1980; 1992): 1) Sample sizes must be large enough to detect the effect. If the time
lag is small, 2 or 3 years, the sample size must be very large (10,000 and up) in order
for change to be detected. Changes in cognition that occur over larger time lags, 7-
10 years, can be detected with smaller sample sizes (500-1000). 2) Tests must be
reliable (internally consistent) 3) Practice effects must be minimal. 4) Measures
must be purely vulnerable, or purely maintained and should not contain large amounts
of both abilities simultaneously. Mixture measures requiring both vulnerable and
maintained abilities will only serve to mask change— the vulnerable abilities involved
will be declining, the maintained abilities involved will be stable or increasing. Given
these guidelines, a latent time lag of at least 10 years should be assumed for latent
longitudinal causal models The model also allows for measurement error to be
modeled— alpha reliabilities of measures can be taken into account.
Perfect test-retest correlations (i.e., 10) are indicative o f zero change— the
variance of the Time 2 variable can be completely accounted for by the variance of the
Time 1 variable Test-retest correlations that are high (.80 or .90) indicate little
change. If little change is operating, then only a small amount of the variance o f the
Time 2 variable is left over to be explained by exogenous causal variables. For causal
variables have an effect on Time 2 variables, the Time 2 variable must contain
variance that cannot be accounted for fully by its measurement at Time 1 . In Figure
211
D. 1 the parameter must be less than perfect in order for the causal parameter pi,2
to be non-trivial.
Table D-l contains causal estimates (Pi,2 parameters) o f lifestyle variables at
Time 1 causing cognitive ability variables at Time 2. For each model the < J> i,3
parameter (test-retest correlations for cognition at Time 1 and cognition at Time 2)
was fixed at a value within a range of plausible test-retest correlations that have been
reported in past research of vulnerable ability change (5 levels of test-retest
correlations were tested; .50, .60, .70, .80,. 90)). The 0 2 4 parameter was fixed at a
value within a range of plausible test-retest correaltions for lifestyle variables. Two
levels of test-retest correlations were chosen for each lifestyle variable. Ten models
(5X2) were run for each lifestyle variable that was plausibly hypothesized to cause
ability changes. The 4* parameter for each model was fixed at the value of the part
correlation between the lifestyle variable and the cognitive variable of interest (age
parted from the lifestyle variable). The a (internal consistency) parameters were fixed
at the value of the actual internal consistencies obtained in the present study. The
only parameter estimated in the models was the Pi,2 causal parameter
Table D-l illustrates some basic measurement principles. In order for cause
and effect relationships to be demonstrated there must be adequate time lags so that
change can occur, measurement must take place with little error, there must be a non
trivial relationship between the cause and effect variables at any age, and the
autoregressive effect of the variable that is being caused must be small enough to
212
accommodate the exogenous cause There are very few non-trivial Pi,2 relationships
for cognitive sub-process variables in these data. A Tolerance for Disorder was
shown to “cause” CON, external belief systems “caused” SLC. There were several
confidence intervals that included non-trivial P1.2 values for lifestyle variables at Time
1 predicting residual Gc (rGc) at Time2. A high level of Information Seeking, a high
Tolerance for Disorder, a lack of external belief systems all yielded confidence
intervals that contain one or more non-trivial P1.2 parameter estimates when causes of
rGc are explored. Again, causal inferences are not asserted here. This analysis does
nothing more than illustrate some basic measurement principles.
213
Table D-1 . Cross-lag CausaI Parameters (Pi,2) when Autoregressive Effects (0 i},
0 i 4 ) are Fixed to a Range o f Plausible Test-retest Correlations— Pi 1 Parameters for
5 Values o f 0 u and 2 Values o f 0 i 4 are Shown
Model: Tolerance for Disorder at Time 1 causes
0»i.3 O 2.4
CON at Time 2 .90 JO
¥ = 1 4 50 .16** .15**
a for Tolerance for Disorder at Time 2= 89 .60 .13 .12
a for CON at Time 2= 96 .70 1 0 . 1 0
5 for Tolerance for Disorder at Time 2= 1 1 80 .07 05
5 for CON at Time 2=04 .90 05 .03
Model: Smoking at Time 1 causes MPA at Time 2
O 1.3 O 2.4
¥ = -.09 70 JO
a for Smoking at Time 2=1.0 .50 -.09 -.08
a for MPA at Time 2= 71 60 -08 -.07
5 for Smoking at Time 2=0 .70 -07 -.05
6 for MPA at Time 2= 29 .80 -.06 -05
90 -.04 -03
Model: Belief in Luck at Time 1 causes MPA at Time 2
^ 1 * 2 .4
*
1 1
1
O
O
JO 70
a for Belief in Luck at Time 2= 8 8 .50 -.09 -.08
a for MPA at Time 2= 71 60 -.08 -.07
6 for Belief in Luck at Time 2= 12 .70 -05 -.06
5 for MPA at Time 2=29 .80 -04 -.04
90 -.03 -.02
Model: Belief in Luck at Time 1 causes SLC at Time 2 0 ,3 O 2.4
V P= - 17 JO 70
a for Belief in Luck at Time 2= 8 8 .50 -.15** -.14
a for SLC at Time 2 = 71 60 -.14 -.13
5 for Belief in Luck at Time 2= 12
.70 -13 -11
8 for SLC at Time 2=29 80 -11 -09
.90 - . 1 0 -.08
214
Table D -l (continued).
Model: Information Seeking at Time 1 causes
$ 1 , 3 $2.4
rGc at Time 2 J*Q JO
¥=.20 .50 .20** .20**
a for Information Seeking at Time 2= 72 .60 .15 15
a for rGc at Time 2= 77 .70 .13 12
8 for Information Seeking at Time 2= 28 .80 .10 09
5 for rGc at Time 2= 23 .90 .08 07
Model: Tolerance for Disorder at Time 1 causes
$1.3 $2,4
rGc at Time 2 JO J10
¥=.20 .50 .22**
22**
a for Tolerance for Disorder at Time 2= 89 .60 .21** .20**
a for rGc at Time 2= 77 .70 .17* 16
5 for Tolerance for Disorder at Time 2= 1 1 .80 .14 .13
5 for rGc at Time 2= 23 .90 .12 11
Model: Belief in Luck at Time 1 causes rGc at Time 2
$1.3 $2.4
¥ = -17 J50 JO
a for Belief in Luck at Time 2= 88 .50 -.15** -14
a for rGc at Time 2= 77 .60 - 14 -13
5 for Belief in Luck at Time 2= 12 .70 -.13 -11
6 for rGc at Time 2=23 .80 -.11
i
o
.90 - 10 -.08
215
Table D-l (continued)
Model: Belief in Powerful Others at Time 1 causes
< 1 * 1 ,3 <1*2.4
rGc at Time 2 JO J)0
¥ = - 23 .50 -.22** -.22**
a for Belief in Powerful Others at Time 2= 85 60 -.17* -.16
a for rGc at Time 2= 77 .70 -.12 -.10
8 for Belief in Powerful Others at Time 2= 15 .80 -.10 -.08
8 for rGc at Time 2= 23 .90 -.09 -.07
¥ = Observed Correlation between Lifestyle Variable and Cognitive Ability Variable
< 1 * 1 .3 < 1 * 2 .4 = Autoregressive Effects+
Pi,2 = Causal Effect: Lifestyle Variable at Time 1 Causes Cognitive Ability Variable at
Time 2*
x l = Alpha Reliability (Internal Consistency) of Measure1
8 = Error of Measurement (1 -a)f
* = Pi,2 parameter significant at p< 05
** = Pi,2 parameter significant at p< 01
* See Figure D-1
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
PDF
00001.tif
Asset Metadata
Core Title
00001.tif
Tag
OAI-PMH Harvest
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-oUC11257217
Unique identifier
UC11257217
Legacy Identifier
9621623