Module 2
Research Methods in Aging and Physical Changes
a. Variables in Developmental Research
A variable is a characteristic that ‘‘varies’’ from individual to individual.
Behavioral scientists attempt to understand why some people are higher on a
particular variable while others may be lower on the same variable. The variable on
which people are observed to differ is the dependent variable in research designs, and
is often referred to as the outcome variable. The independent variable explains or
‘‘causes’’ the range of scores on the dependent variable. Although developmental
psychologists classify age as an independent variable, age is truly not ‘‘independent’’
because its value cannot be controlled or manipulated, as you will read more about
shortly. To overcome this challenge, methodologists have devised a range of statistical
techniques designed to handle the complex nature of age.
An experimental design, often considered the gold standard of research,
involves the manipulation of an independent variable followed by the measurement of
scores on the dependent variable. Respondents are randomly assigned to treatment
and control groups and then measured on the dependent variable. It is assumed that
people vary on the dependent measure because they were exposed to different levels
of the independent variable. Clinical trials typically compare the effectiveness of a
drug on one group compared to another, and can provide compelling experimental
evidence regarding cause and effect.
The use of experimental designs in the study of adult development presents a
problem given that age is not a true independent variable. People cannot be randomly
assigned to an age group. This means that you can never state with certainty that
aging ‘‘caused’’ people to receive certain scores on a dependent variable of interest.
What you can establish is whether different age groups varied in their responses to
different levels of the independent variable. If a manipulation produces a different
effect in a younger age group than in an older one, you may infer that the treatment
has some relationship to age. For example, consider an experiment in which special
instructions are given for a memory test to a younger and an older age group. If the
performance significantly improved across both age groups but to a greater degree for
the older adults, you could deduce that age is related to sensitivity to the type of
manipulation involved in these instructions. This is an example of an interaction
between variables—the manipulation affects both groups, but the effect is stronger for
one group than for the other. Important to remember in this situation is that random
assignment to age group is impossible. The study of aging therefore presents a
methodological challenge. Even in the above example, you are not able to determine
that age ‘‘caused’’ differences in responses to the treatment groups.
Instead, studies of aging use the quasiexperimental design, the term used to
describe the process of comparing groups on predetermined characteristics. You
cannot conclude that the predetermined characteristic caused the variations in the
dependent variable, but are able to use the results to describe the differences between
groups. Consider a study involving comparisons in levels of happiness between older
adults living in the community and older adults in assisted living facilities. Since
random assignment did not occur, you cannot conclude which living situation caused
higher levels of happiness in one group over the other. Other factors may account for
the differences in happiness rather than living situation. What you can do is attempt to
rule out other alternatives, such as levels of physical functioning, which may account
for differences in happiness. Once you feel that other explanations have been ruled
out, and if the differences are repeatedly demonstrated, you can make the cautious
inference that living situation had something to do with variations in people’s
happiness scores. Similarly, when comparing age groups on performance on a
variable of interest, such as on memory performance for lists of numbers, you cannot
conclude that age caused older adults to perform differently than the comparison age
groups.
As you will see later, ruling out alternative explanations is of tremendous
importance not only when diverse age groups are compared, but also when examining
other characteristics associated with age. This principle is crucial in ensuring the
validity and reliability of research findings, as it helps to isolate the specific factors
that contribute to observed differences and prevents the misattribution of causality.
When comparing diverse age groups, it is essential to consider a multitude of
potential confounding variables that might influence the results. For instance,
socioeconomic status, education level, cultural background, and health conditions can
all vary significantly across different age cohorts and impact the outcomes being
studied. By systematically ruling out these alternative explanations, researchers can
more accurately identify the genuine effects of age itself, rather than conflating age-
related changes with other factors.
Moreover, the importance of ruling out alternative explanations extends
beyond simple age comparisons to encompass a broader range of characteristics
associated with aging. These characteristics might include cognitive abilities, physical
health, social interactions, emotional well-being, and lifestyle choices, among others.
Each of these factors can be influenced by a host of external variables, making it
imperative to account for and control these variables to draw valid conclusions.
For example, suppose a study finds that older adults tend to have lower
cognitive performance compared to younger adults. Without ruling out alternative
explanations, one might prematurely conclude that cognitive decline is an inevitable
consequence of aging. However, other factors such as differences in educational
opportunities, levels of physical activity, nutrition, and chronic health conditions
could also contribute to this observed difference. By carefully controlling for these
variables, researchers can better understand the true nature of cognitive aging and
distinguish between age-related changes and those driven by other influences.
Similarly, when examining physical health across age groups, it is crucial to
consider factors such as access to healthcare, lifestyle habits, and environmental
exposures. For instance, older adults may have different dietary patterns or levels of
physical activity compared to younger individuals, which can affect health outcomes
independently of age. Ruling out these alternative explanations helps to clarify the
specific impacts of aging on physical health and identify effective interventions to
promote healthy aging.
The principle of ruling out alternative explanations is also vital in
understanding the social and emotional aspects of aging. Social networks, levels of
social support, and cultural attitudes towards aging can vary widely and influence the
well-being of older adults. For instance, a study might find that older adults in one
community report higher levels of life satisfaction compared to those in another
community. To accurately interpret this finding, researchers must consider factors
such as the availability of social services, community engagement opportunities, and
cultural norms surrounding aging in each community. By accounting for these
variables, the study can more accurately determine the role of social and
environmental factors in shaping the well-being of older adults.
Furthermore, ruling out alternative explanations is essential in longitudinal
studies that track changes within individuals over time. As people age, they
experience various life events and transitions, such as retirement, loss of loved ones,
and changes in health status, all of which can influence the outcomes being measured.
By identifying and controlling for these life events, researchers can better isolate the
effects of aging itself and provide more precise insights into the aging process.
In conclusion, ruling out alternative explanations is of paramount importance
in research involving diverse age groups and characteristics associated with aging.
This practice ensures that findings are valid and reliable, allowing for a clearer
understanding of the true effects of aging. By carefully considering and controlling
for potential confounding variables, researchers can draw more accurate conclusions,
develop effective interventions, and ultimately contribute to a deeper and more
nuanced understanding of the aging process. As we delve further into this topic, the
importance of this principle will become increasingly evident, highlighting its critical
role in advancing the field of aging research.
b. Descriptive Research Designs
As you just learned, studies of aging are by definition quasi-experimental and
thus do not allow cause and effect conclusions. A secondary issue is that age is
entangled with other associated variables. For example, the period of history in which
people were born can influence their performance. Designs attempting to rule out
these possible influences have been developed, as we will discuss shortly. First,
however, we will cover the traditional research designs used to study aging that do not
account for these differences. We use the term ‘‘descriptive’’ to refer to studies in
which age alone is the variable of interest. For instance, if you hear a news story
stating that older adults are more likely to have a certain health problem than younger
adults, you may jump to the conclusion that age ‘‘causes’’ this particular health
problem. However, chances are that the study compared older and younger adults at
one point in time. It is possible that the older people had the health problem because
of something that happened to them in their youth. Perhaps the older people were
exposed to poorer nutritional practices than their younger counterparts; therefore the
findings reflect disparities in early life experiences rather than age.
This example highlights the types of issues facing developmental researchers.
Many studies in the psychology of adult development and aging involve the use of
simple designs in which people of different ages are compared on the variable of age.
Researchers are able to do little except state that age differences exist. Although
descriptive research designs lack the ability to determine cause and effect,
understanding differences between age groups still serves a useful purpose. More
complex designs are then needed to examine factors other than age.
Three factors jointly influence the individual’s performance on a given
psychological measure at any point in life: age, cohort, and time of measurements. As
you will see, these variables are highly related to each other, and untangling them
presents many challenges requiring both creativity and scientific rigor. Age is
measured chronologically, and is most generally quantified in years. The older a
person is, the more calendar years that person has experienced. There may or may not
be a direct connection between these calendar years and the changes going on within
the person. Developmental psychologists use it as convenient shorthand but
understand that age is an imperfect index.
Social aging refers to exposure to changes over time in the society of which
the individual is a part and is indexed by two measures: the year of a person’s birth
and the time at which measurements are taken. The year (or period) of birth is referred
to as cohort. Conceptually, the term ‘‘cohort’’ represents the more familiar term
‘‘generation’’ in that it refers to the group of people who were born during (and hence
lived through) some of the same social influences. For example, Americans in the
cohort who were in college during the Vietnam War–era shared certain experiences
(such as serving in the armed forces, taking part in protests, and/or living at the height
of the Cold War). The particular nature of these experiences were specific to this
period of history and although wars and protests have gone on during different eras,
the experiences in the 1960s had their own distinct features. Time of measurement is
the year or period in which testing occurs.
Like cohort, its meaning goes beyond the calendar in that it indicates the social
and historical influences on the individual at the point when data are collected. For
instance, adults tested now are more proficient at using computers than were adults
tested in the 1980s, when personal computers were far less available or accessible. A
measure of development that depends on being able to use a computer is influenced
by the year in which the study is conducted. The connection between time of
measurement and cohort creates difficulties when investigators attempt to separate
these indices of social and historical context. Time of measurement is inherently
linked to cohort; people of the same age are also part of the same cohort, because time
of measurement minus cohort year equals age. This fundamental problem means that
unless other controls are imposed, researchers never know whether results are due to
aging or to the exposure of participants to historical change.
In a longitudinal study, people are followed repeatedly over time. By
observing and studying people as they age, researchers aim to determine whether
participants have changed over time as the result of the aging process. You have very
likely experienced the equivalent of a longitudinal study in watching what happens to
you and the people you know as they get older. Consider your own experience from
the start of kindergarten to high school graduation. Many of you may have attended
school with the same set of students throughout your early educational experiences,
and watched your friends and peers grow and develop along with you. Perhaps you
marveled at how some friends remained the same over the years while others were not
at all like they were when they were younger. Your best friend in first grade became
your enemy in high school. Peers you thought you would never associate with in 7th
grade became close friends and confidantes in 11th grade.
Now that you are in college, your relationships with your childhood friends
will change even more with the passage of time. You may be attendants at their
weddings or even find romance with past loves. Fast forward to your 10-year reunion
(or if you have already attended yours, reflect back), when you are reminded of the
person you were during your adolescent development. Many of your former
classmates will look and act the same, and you may remark that the years have done
little to change these people. However, other classmates may be difficult or
impossible to recognize. The head cheerleader, considered flighty and vain during
high school, may have entered the Peace Corps in an attempt to make the world a
better place. As you catch up with her, you are struck by how sincere and down to
earth she has become—quite a different person from her teenage self. The
characteristic nerd has evolved into a shrewd and successful businessman, who no
longer resembles the awkward and shy teenager you remembered.
As you interact with your former classmates, you develop your own
hypotheses regarding reasons for the changes you have witnessed. One possibility is
that they have not changed so much as has your understanding of who they are.
Maybe you were too harsh to judge the cheerleader or nerd and it is only now that you
can see them as the complex and multifaceted people they were all along. Of course,
you may not have to wait until your 10-year reunion, as social networking sites such
as MySpace and Facebook make following (or ‘‘Facestalking’’) the development of
past classmates as easy as turning on your computer. Fascination with the forces that
drive people to change over time is often what attracts most researchers to the field of
adult development and aging. When researchers are lucky enough to have longitudinal
data, they often feel that they have gained true insight into the very essence of change
over time in human behavior. However, similar to descriptive research designs,
longitudinal studies are not without their limitations.
Theoretically, researchers are unable to state with certainty that the changes
observed over time are a result of the person’s own aging or the result of the changing
environment in which the person functions. The individual cannot be removed from
the environment to see what would happen if he or she had lived in a different time or
place. It is simply not possible to know if people are inherently changing or whether
they alter due to the circumstances in which they are aging. This is the key limitation
to longitudinal studies: the inability to differentiate between personal and historical
time. In addition to the theoretical problems, practical problems also plague
longitudinal research. The most prominent concern is the length of time longitudinal
research takes to conduct. In order to be of greatest value to the study of adult
development and aging, longitudinal studies should generally span at least a decade or
more. As a result, they can be quite costly to conduct. Furthermore, the results are
often not available for many years, creating problems for researchers who do not have
the luxury or patience to see the outcomes. In the many scenarios, the original
investigator may not even live long enough to see the results come to fruition.
Participants in a longitudinal study are also likely to be lost over time, a
problem referred to as selective attrition. The loss of participants from the original
sample creates a host of practical and theoretical problems. Practically, as the number
of participants diminishes, it becomes increasingly difficult to complete statistical
analyses on the data. Power calculations are performed to determine how large a
sample should be to detect differences between groups (such as sex, social class, or
race). If the sample size is too small, the study may lack the potential to answer the
original research questions. The loss of participants also hampers the investigator’s
ability to draw inferences from the sample to the population as a whole. The people
who disappear from the sample do so for a variety of reasons such as poor health or
death, lack of motivation, or an inability to continue in the study because they have
moved from the area or are otherwise unreachable. The study ‘‘survivors’’ may be
higher on all or some of these factors. Conclusions made about the survivors may
therefore not apply to the general population.
Consequently, longitudinal data may be skewed given that different types of
people are present in the samples across succeeding test occasions. Researchers
studying the phenomenon of attrition worry that the survivors still present for testing
many years into the study were healthier, more motivated, or in other ways different
from the people who are no longer in the study. The researcher may erroneously
conclude that participants in the sample ‘‘improved’’ when in reality, the sicker and
less motivated are simply gone from the sample. To address the problem of attrition,
longitudinal researchers typically conduct analyses to determine whether the pattern
of participant dropout was random, or whether it reflected a systematic bias that kept
the healthier and more motivated participants in the sample. Such a technique is
referred to as nonrandom sampling, and means that successive samples are
increasingly unlike the populations they were intended to represent. Various statistical
techniques are employed to determine whether non-random sampling has occurred,
and if so, whether relationships between the variables are affected. Without such
procedures, the study’s results become difficult to interpret.
A variety of methods can also address the practical problems encountered in
longitudinal research. Researchers are motivated to retain their samples for as long as
possible. A vigilant monitoring and maintenance of databases containing contact
information is critical. Large-scale longitudinal studies are often housed in established
institutions or agencies that have clerical assistance available to the investigator.
Administrative personnel whose job it is to retain study participants can maintain the
‘‘care and feeding’’ of the sample in between testing occasions. The research staff
may send out greeting cards for holidays and birthdays, or update participants on
study progress via newsletters and e-mails. Many longitudinal studies use Web sites to
allow participants to engage interactively. Creating a personal touch encourages
respondents to continue participation, and also enables investigators to keep better
track of moves, deaths, or other issues. Other approaches to tackling attrition include
the use of a simulated longitudinal design, in which several cohorts are followed up
over a 5- or 10-year period. For example, a study could include people in their 30s,
40s, and 50s who are followed up over a 10-year period. A second study may recruit
people in their 50s, 60s, and 70s and follow them up 10 years later as well. Results
from both studies can be combined to produce a simulated longitudinal design that
includes participants from the ages of 30 to 70 and took 20 instead of 40 years to
complete. Sophisticated statistical methods are used to analyze and interpret results
from these designs so that they simulate the effect of a much longer-term study.
Practice effects are another complication in longitudinal studies. Participants
who complete tests on multiple testing occasions may become better at answering the
questions. If, for example, the particular test measures intelligence, participants may
purposely learn the answers in between testing occasions. Similarly, in studies
examining personality, a participant may suspect or find out the meaning of a
response that implies something unfavorable. On the next test occasion, the
respondent may be less likely to endorse that statement. Researchers often use
alternate forms of tests on different testing occasions to avoid such problems. From
the investigator’s point of view, a far more serious dilemma is posed by the nature of
the tests themselves which, over time, may become outdated. The cutting edge theory
developed in the 1980s may have since been refuted, but the researcher is still left
with measures based on that theory and not the newer one. One way to address this
problem is to reanalyze or rescale the test scores to correspond to the newer theory, if
possible. This was the strategy used in studies of personality development by
researchers at the Institute for Human Development in Berkeley, who began a study of
child development in the late 1920s, a project that continued far longer than initially
planned. By the time participants were in their adult years, the original measures were
no longer theoretically relevant. The researchers rescored the data using newer
theoretical and empirical frameworks that were also more age-appropriate.
Despite their flaws, longitudinal studies have the potential to add invaluable
data on psychological changes in adulthood and old age. Furthermore, as data
accumulate from multiple investigations concerning related variables, it is possible to
overcome the limitations of any one particular study to gain as accurate a depiction of
change across adulthood as possible. Even though one study may have its problems,
convergence across several investigations allows researchers to feel greater
confidence when findings are similar from one study to the next.
The alternative to following people over time is to compare groups of different
ages at one point in time. The cross-sectional design is by far the more frequently
used research method in the field of developmental science in general, but particularly
in research on aging. The goal of cross-sectional research on adult development and
aging is to describe age differences. However, a key aim is being able to draw
conclusions about changes associated with the aging process that cause these
differences. To ensure that they are correct in their assumption, researchers attempt to
control for cohort differences that could potentially obscure or exaggerate the effects
of age. They can best achieve this control by selecting samples comparable in
important factors such as amount of education and social class. Even if they cannot
achieve this level of control, they can ensure that differences relevant to the main
purpose of the study are kept to a minimum. For instance, in a study of aging and
verbal memory, it would be important for researchers to ensure that the age groups
being compared have similar vocabulary or verbal comprehension skills. Like the
longitudinal design, the cross-sectional design is only applicable to one historical
period.
Age differences obtained in a cross-sectional study are specific to the cohorts
of people compared. For example, people born in the 1950s and tested in the 2000s at
the age of 50 may be higher on an attribute than people born in the 1970s and tested in
the 1990s at the age of 20 because of social factors specific to each of these cohorts. A
similar difference between 20- and 40-year-olds may not be encountered among
people compared in a study conducted in the 1980s, when the samples were born 10
years earlier. For example, although commonsense wisdom regards young adults as
less conservative than middle-aged adults, it is possible that middle-agers who lived
through the 1960s are less conservative than young people growing up in the 1980s.
The same difference between age groups may not reveal itself if the study is
conducted in a different historical era. The graph is from an investigation comparing
the relative amounts of white matter in the brains of young, middle-aged, and older
adult samples (Brickman et al., 2006). The higher the amount of white matter, the
better the individual’s brain functions. From, it would seem that because the younger
age groups have more white matter, their brains are functioning better compared to
the older adult brains. However, the story is not so simple. The age groups were not
comparable in education, with the older adults having the fewest years of formal
schooling. The cohort difference in education makes it impossible to know whether it
was age or education that most strongly influenced the white matter of these three age
groups.
Selective survival, the bane of longitudinal investigators, also serves to
challenge cross-sectional researchers. Study participants, by definition, are survivors
of their respective age groups. Thus, they may represent a healthier or luckier group
of people than those in their cohort who did not live as long. Perhaps they are the ones
who are more cautious, smarter, and genetically hardier and so were able to avoid the
many diseases that could have caused their death prior to old age. As a result, older
adults in a cross-sectional study may look different from the younger ones because the
two groups are drawn from two different populations—those who die young (but are
still represented in the young adult group) and those who survive to be old. The
sampling of young adults drawn exclusively from a college population, a common
technique in psychological research, may not be representative of the younger cohort
either. Any time you read about a study comparing younger and older adult samples,
keep in mind the demographic makeup of the younger sample and not just the
comparison between samples of different ages.
The problem of assigning participants to specific age groups presents
additional difficulties with crosssectional research. If researchers wish to compare
‘‘young adults’’ with older adults, should they restrict themselves to the sort of typical
18–22 year range or should they allow the age range to expand to the mid- or even
late 20s? As it turns out, the age range of older adult samples is rarely defined as
narrowly as is the range of those in college. Often, researchers have to settle for an
age range for the older group that is larger than is desirable. In some studies, the range
is as large as 20 to 30 years (or more). Some studies define the ‘‘older’’ sample as all
respondents over the age of 50 or 60 and, having done so, fail to look for any possible
age differences within the older sample. By the time all is said and done, age
differences in the older sample may be as great, if not greater than, differences
between the older and younger samples.
A related problem to determining acceptable age ranges is the question of how
to divide the adult age range when selecting samples. Is it better to divide samples of
people in cross-sectional studies into decades and then examine age differences
continuously across the adult years? Or perhaps is it better to compare people at the
two extremes of the adult span? Researchers increasingly include middleaged samples
along with the younger and older adults rather than compare only those at the two
extremes of the age distribution. The inclusion of three age groups creates a more
justifiable basis for ‘‘connecting the dots’’ between their scores on measures of
psychological functioning across the adult years. We raise these problems not
necessarily because we have all the perfect answers but to sensitize you to the need to
look closely at study samples when you are examining research on aging. After this
course, when you come across an article in the newspaper about a study that claims to
have discovered an important truth about the aging process, be sure to read between
the lines to determine how carefully the samples were selected.
Another area of concern to researchers conducting cross-sectional studies is
the need to be sensitive to how different age groups will react to the test materials
administered in research settings. In studies of memory, for example, there is a risk
that the older adults will find some of the measures challenging and perhaps
intimidating because they are not used to having their abilities evaluated in a formal
setting such as the psychology lab. Young adults are far more comfortable with test
situations either because they are currently or were recently in school where they were
frequently tested. To an older adult, particularly one who is sensitive to memory loss,
anxiety about the situation rather than actual performance can result in decreased
scores, a topic we address further below. Task equivalence also applies to the way
different cohorts react to measures of personality and social attitudes. The same item
may have very different meanings to people of different generations or people of
different educational or cultural backgrounds. A common problem in personality and
mental health research is that a measure of, for example, depression, may have been
tested on a young adult sample but not on an older sample. Items on such a scale
concerning physical changes, such as alterations in sleep patterns presumably related
to depression, may in fact reflect normal age-related differences and not differences in
depression. Older adults will therefore receive a higher score on the depression scale
by virtue of changes in their sleep patterns alone, not because they are actually
suffering from depression.
These problems aside, cross-sectional studies are relatively quick and
inexpensive compared to longitudinal studies. Another advantage of crosssectional
studies is that the latest and most up-to-date technology can be implemented, whether
in the biomedical area or in the psychological and social domains. If a new tool or
technique comes out one year, it can be tested cross-sectionally the next. Researchers
are not tied to obsolete methods that were in use some 30 or 40 years ago. The best
cross-sectional studies, though never able to permit causal inferences about aging,
employ a variety of controls to ensure that differences other than age are kept to a
minimum and that the ages selected for study span across the adult years. Most
researchers regard their cross-sectional findings as tentative descriptions of the effects
of aging on the function of interest. They are aware of the importance of having their
findings replicated and verified through studies employing a longitudinal element.
c. Sequential Research Designs
We have probably convinced you by now that the perfect study on aging is
virtually impossible to conduct. Age can never be a true independent variable because
it cannot be manipulated. Furthermore, age is inherently linked with time, and so
personal aging can never be separated from social aging. However, considerable
progress in some areas of research has been made through the application of
sequential designs. These designs consist of different combinations of the variables
age, cohort, and time of measurement. Simply put, a sequential design involves a
‘‘sequence’’ of studies, such as a cross-sectional study carried out twice (two
sequences) over a span of 10 years. The sequential nature of these designs is what
makes them superior to the truly descriptive designs conducted on one sample,
followed over time (longitudinal design) or on different-aged samples, tested on one
occasion (cross-sectional design). Not only do sequential studies automatically
provide an element of replication, but when they are carried out as intended, statistical
analyses can permit remarkably strong inferences to be drawn about the effect of age
as distinct from cohort or time of measurement.
One of the most influential articles to be published in the field of adult
development and aging was the landmark work by psychologist K. Warner Schaie
(1965) in which he outlined what would later be called the Most Efficient Design, a
set of three designs manipulating the variables of age, cohort, and time of
measurement. It is ‘‘most efficient’’ because it enables the most amount of
information to be condensed into the most inclusive data framework. Researchers
organize their data by constructing a table that combines year of birth (cohort) with
year of testing (time of measurement). The three designs that make up the Most
Efficient Design and the respective factors they include are the time-sequential design
(age by time of measurement), the cohort-sequential design (cohort by age), and the
cross-sequential design (cohort by time of measurement). When all three designs are
analyzed, they make it theoretically possible for the researcher to obtain separate
statistical estimates of the effects of each of the three factors. Schaie and his
collaborators have employed such techniques to the Seattle Longitudinal Study, a
large-scale study of intelligence. Thus, depending on the pattern of significant effects,
the researcher may be able to draw conclusions about the relative influences of
personal and historical aging on test performance. For example, if age effects are
significant in the time-sequential and cohort-sequential designs and there are no
cohort or time effects in the cross-sequential design, then ‘‘true’’ aging effects may
exist. Another scenario involves significant effects of time of measurement in both the
time-sequential and cross-sequential designs, patterns that suggest it was the time in
history rather than the age of the participants that most influenced the pattern of
scores.
Similarly, if the cohort factor is significant in the two research designs in
which it is used, while significant age effects or time of measurement effects are not
observed, then the researcher might turn their attention to early childhood
environmental factors in these samples. This nuanced approach recognizes that cohort
effects can reveal important generational influences that differ from pure age-related
changes or the specific timing of data collection.
A cohort effect refers to the impact of being born and growing up during a
particular time period, with its unique cultural, social, and historical contexts. When
researchers observe significant cohort effects, it suggests that the shared experiences
of individuals within a cohort have influenced their development and current
outcomes. These experiences can include exposure to specific educational practices,
economic conditions, societal norms, and technological advancements, all of which
can shape a cohort in distinctive ways.
For instance, if two separate studies both find that a particular cohort exhibits
distinct behaviors, attitudes, or health outcomes compared to other cohorts, and these
differences cannot be attributed to age or the timing of the measurements, the
researcher might hypothesize that early childhood environmental factors played a
crucial role. These factors could encompass a wide range of influences, such as the
quality of early childhood education, levels of parental involvement, exposure to early
childhood nutrition and healthcare, and the broader socio-economic environment
during formative years.
By investigating these early environmental factors, researchers can gain a
deeper understanding of how formative experiences contribute to long-term outcomes.
For example, a cohort that grew up during a period of economic prosperity with
widespread access to quality education and healthcare might display higher levels of
cognitive functioning and better health outcomes in adulthood compared to cohorts
that did not have similar advantages. Conversely, a cohort that experienced significant
economic hardship or limited access to education and healthcare during childhood
might show different developmental trajectories, which could be misinterpreted if
only age or time of measurement effects were considered.
In exploring these early childhood factors, researchers might conduct in-depth
analyses of historical records, educational policies, public health initiatives, and
socio-economic conditions prevalent during the early years of the cohort under study.
They might also examine longitudinal data that tracks individuals from childhood into
adulthood to identify specific early-life conditions that correlate with later-life
outcomes. This comprehensive approach allows researchers to piece together a
detailed picture of how early environments shape developmental pathways over the
lifespan.
Moreover, understanding the role of early childhood environmental factors can
inform public policy and intervention strategies. If significant cohort effects are linked
to early childhood conditions, it underscores the importance of investing in early
childhood education, healthcare, and socio-economic support. Policymakers can use
this evidence to advocate for programs that provide a strong foundation for children's
development, thereby promoting better long-term outcomes for future generations.
In conclusion, when significant cohort effects are observed in the absence of
significant age or time of measurement effects, researchers are prompted to consider
the influence of early childhood environmental factors. This shift in focus allows for a
more nuanced understanding of developmental processes and highlights the long-
lasting impact of early life conditions. By examining these factors, researchers can
uncover the specific experiences that shape different cohorts and use this knowledge
to inform policies and practices that support healthy development across the lifespan.
d. Correlational Designs
An alternative approach to describing group differences using the quasi-
experimental design is the correlational design, in which relationships are observed
among variables as they exist in the world. The researcher makes no attempt to divide
participants into groups or to manipulate variables.
Comparisons of age groups or groups based on divisions such as year of birth
or time of measurement are useful for many research questions in the field of
gerontology. However, often this approach is neither the most efficient nor the most
informative. By grouping people into categories, researchers lose a great deal of
information that could be preserved if they used actual age in years. The variable of
age is a continuous variable, meaning that it does not have natural cutoff points as
does a categorical variable such as gender. There may be a difference between people
of 42 and people of 45 years of age, but when they are all grouped in the ‘‘40-year-
old’’ category, this distinction is obscured. In the correlational design, age can be
treated as a continuous variable and it is therefore unnecessary or even desirable to
put people into arbitrarily defined groups. The relationship between age and another
variable is expressed through the statistic known as the correlation (represented by the
letter r) whose value can range from +1.0 to −1.0. A significant positive correlation
indicates that the two variables are positively related so that when the value of one
variable increases, the other one does as well. A significant negative correlation
indicates that the two variables are negatively related so that when one increases in
value the other one decreases. A correlation of zero indicates no relationship between
the variables. In a correlational study, the researcher makes no assumptions about
what caused what—there are no ‘‘independent’’ or ‘‘dependent’’ variables. A
correlation between two variables means simply that the two variables are related, but
like the proverbial chicken and egg, the researcher cannot say which came first.
Let’s consider as an example the relationship between age and response speed.
The correlation between these two variables is most often positive; that is, when age
increases, response speed does as well (keep in mind that a higher response speed
indicates slower performance). When interpreting this relationship, the researcher may
be tempted to conclude that age ‘‘caused’’ the increase in response speed. However,
this conclusion is not justified because age was not experimentally manipulated. In a
correlational study, there are no independent or dependent variables. Therefore, the
possibility that variable A accounts for variable B is equal to the possibility that B
accounts for A. It may be tempting to assume that response speed does not cause
increasing age; rather, age causes increasing response times. However, since an
experiment was not conducted, the possibility that increased response speed caused
aging cannot be ruled out. Here is another example to clarify this point. There is a
correlation between certain Type A personality characteristics and cardiovascular
disease. People with Type A personality, who are hard-driving, competitive, often
hostile, and impatient, are more likely to have heart disease. Does the personality type
cause heart disease (which might seem most logical), or does heart disease cause
people to have personality problems? Because the design is correlational, neither
possibility can be excluded.
If you are a psychology student, you have probably incorporated the mantra
‘‘correlation does not equal causation’’ into your everyday language. This phrase
refers in part to the fact that correlations do not allow researchers to state that one
variable was the direct cause of another. However, the other implication of this phrase
refers to the possibility that a third but unmeasured variable accounts for the apparent
relationship between the two observed variables. In the case of the relationship
between age and response speed, this third variable might be ‘‘number of functioning
brain cells.’’ Age may be related to number of brain cells, and number of brain cells
may be related to response speed. The apparent correlation between age and response
speed might disappear entirely when number of brain cells is measured and factored
into the relationship.
When examining data from a typical correlational study, it is important to keep
in mind that causation cannot be inferred from correlation and to be on the lookout for
competing hypotheses related to unmeasured variables. While this is no less true of
gerontology than of other disciplines, because arguments related to age are often so
compelling, it is particularly easy to fall into the trap. Correlational studies contain a
wealth of information despite their inability to determine cause and effect. The value
of the correlation itself provides a useful basis for calculating the strength of the
relationship. Furthermore, it is possible to manipulate a larger number of variables at
one time than is generally true in studies involving group comparisons. Advanced
correlational methods have become increasingly available in the past 20 years that
allow researchers to navigate the difficulties involved in causality with traditional
correlational methods, and we will turn to those next.
In contrast to simple correlational designs, which involve determining the
statistical relationship between two variables (called a bivariate relationship), a
multivariate correlational design involves the analysis of relationships among more
than two variable. The researcher using a multivariate design can simultaneously
evaluate the effects of many potentially important factors, rather than being restricted
to the study of two variables, which can lead to overlooking an important third (or
fourth) variable. Multivariate correlational methods also enable researchers to test
models in which a set of variables is used to ‘‘predict’’ scores on another variable. In
multiple regression analysis, the predictor variables are regarded as equivalent to the
independent variables, and the variable that is predicted is regarded as equivalent to a
dependent variable. Although the design is still correlational in that the experimenter
does not manipulate an independent variable, the statistics involved enable
investigators to suggest and test inferences about cause–effect relationships.
A variant of multiple regression is logistic regression, in which researchers test
the likelihood of an individual receiving a score on a discrete yesno variable. For
example, a group of investigators may want to test the probability that a person will
receive a diagnosis of cardiovascular disease or not, depending on whether the person
has one of several risk factors. Logistic regression is often used to determine whether
non-random sampling has occurred, as discussed earlier with subject attrition in
longitudinal studies. Using the yes-no variables of ‘‘survivor’’ and ‘‘dropout,’’
researchers can attest to whether differences between survivors and dropouts are due
to chance or to other factors. Multivariate correlational designs have the potential to
test complex models and so are increasingly being used in research on adult
development and aging in which there are so many problems with age as a variable. In
structural equation modeling (SEM), researchers develop hypotheses regarding the
relations among observed (measures) and latent (underlying) variables or factors
(Hoyle, 1995).
SEM serves purposes similar to multiple regression, in a more powerful way,
taking into account the possibilities that there are complex relationships among the
variables and factors of interest. A multivariate method recently introduced to
developmental research is hierarchical linear modeling (HLM) (Raudenbush & Bryk,
2002). In this method, also referred to as multilevel modeling, individual patterns of
change are examined over time rather than simply comparing mean scores of people
at different ages. Such a technique is particularly important in longitudinal research,
because not every participant exhibits the same patterns of change over time. While
some individuals may increase, others may decrease, and some may not change at all.
Solely looking at overall mean scores fails to capture this individual variation. In
HLM, individual patterns can be explored statistically in addition to examining
whether particular variables affect some individuals more than others.
e. Types of Research Methods
Data on adult development and aging can be captured using a variety of data
collection strategies, or research methods. Each method has advantages and
disadvantages that are important to consider according to the particular field of study,
the nature of the sample, available resources, and desired applications.
The majority of information about physical and cognitive changes associated
with the aging process comes from laboratory studies, in which participants are tested
in a systematic fashion using standardized procedures. The laboratory method is
considered the most objective way of collecting data because each participant is
exposed to the same treatment, using the same equipment and the same data recording
procedures. For example, in a study of memory, participants may be asked to recall a
set of items presented on a computer. At a later point, they may be asked at to recall as
many of those items as possible using some type of automated response system. There
are obvious advantages to the laboratory study. The objective and systematic way in
which data are recorded provides the investigator assurance that the results are due to
the variables being studied rather than to extraneous factors. For instance, in the
memory study, all participants would be presented with the recall items
systematically, in a way that does not depend on the voice inflections of the
researcher, the quality of the visual stimuli, or the amount of time used to present the
items.
A limitation of the laboratory study is the inability to apply the stimuli
presented to real-life experiences of most adults. It is possible that the older person
feels uncomfortable when tested in an impersonal and possibly intimidating manner
using unfamiliar equipment. Consequently, the findings may underestimate the
individual’s abilities in everyday life, and may not generalize to real-world scenarios.
There are often instances in which researchers wish to explore a phenomenon
of interest in an open-ended fashion. The investigation of social influences on adult
development such as, for example, personal relationships, may demand the researcher
use a method that captures potentially relevant factors within a broad spectrum of
possible influences. Qualitative methods allow for the exploration of such complex
relationships outside the narrow restrictions and assumptions of quantitative methods.
In other cases, researchers may be working in an area in which conventional methods
are neither practical nor appropriate for the problem under investigation. Qualitative
methods are also used in the analysis of life history information, which is likely to be
highly varied from person to person and not easily translated into numbers. The main
advantage for using qualitative methods is that they provide researchers with
alternative ways to test their hypotheses. The qualitative method can be adapted in a
flexible manner to the nature of the problem at hand.
In archival research, investigators use existing resources that contain data
relevant to a question about aging. The archives might consist of a governmental data
bank, or the records kept by an institution, school, or employer. Another source of
archival data is newspaper or magazine reports. An advantage of archival research is
that the information is readily accessible, especially given the growth of Web-based
data sets including those of the U.S Census. Data files can be downloaded directly
from the Internet, or publications can be accessed using portable document files
(PDFs) that are easily read and searched. Disadvantages are that the researcher does
not necessarily have control over the form of the data. For instance, a governmental
agency may keep records of employment by age that do not include information on
specific occupations of interest to the researcher. Another disadvantage is that the
material may not be systematically collected or recorded. Newspaper or school
records, for example, may have information that is biased or incomplete.
Researchers rely on the survey method to gain information about a sample that
can then be generalized to a larger population. Surveys are typically short and easily
administered with simple rating scales to use for answers. For instance, surveys are
given to poll voters on who they will be casting their ballots for in upcoming
elections. Occasionally, more intensive surveys may be given to gain in-depth
knowledge about aging and its relationship to health behaviors, health risks, and
symptoms. The U.S. Census was collected through survey methodology. However, it
is considered archival in that it has extensive historical records going back to the year
1750 when the first U.S. Census was conducted. Surveys have the advantage of
providing data that allow the researcher to gain insight into the behavior of more
people than it would be possible to study in the laboratory or other testing site. They
can be administered over the telephone or, increasingly, via the Web. Interview-based
surveys given by trained administrators provide knowledge that is easily coded and
analyzed while still providing comprehensive information about the behavior in
question. Typically, however, surveys tend to be short with questions that are subject
to bias by respondents, who may attempt to provide a favorable impression to the
researcher. Consequently, although the data may be generalizable to a large
population, the quality of the data itself may be limited.
When researchers want to provide an in-depth analysis of particular
individuals, they use the case report, which summarizes the findings from multiple
sources for those individuals. Data may be integrated from interviews, psychological
tests, observations, archival records, or even journal and diary entries. The focus of
the case report is on the characteristics of the individual and what has influenced his
or her development and life experiences. Personal narratives may also be obtained
with this method, in which individuals describe their lives as they have experienced
them along with their ideas about why their lives have evolved in a given manner.
Although the case report has the benefit of providing insights into the lives of
individuals as they change over time, it relies heavily on clinical judgments by the
researcher. Therefore, for a case report to provide valuable information, a high level
of expertise is required so that the findings are presented in a manner that balances the
objective facts with the subjective analysis of the researcher.
A less formal research method is a focus group, which is a meeting of a group
of respondents oriented around a particular topic of interest. In a focus group, an
investigator attempts to identify important themes in the discussion and keep the
conversation oriented to these themes. The goal is to develop concrete research
questions to pursue in subsequent studies. For example, attitudestoward mental health
providers by older adults may be assessed by a focus group in which participants 65
and older share their concerns and experiences with counselors and therapists. An
advantage of the focus group is that issues can be identified through a focus group
prior to conducting a more systematic investigation. This approach, often considered a
pilot study, is particularly useful when little preexisting research on the topic is
available. An obvious disadvantage isthatthe method is not particularly systematic,
and the data cannot readily be analyzed or systematically interpreted.
In the observational method, researchers draw conclusions about behavior
through careful and systematic examination in particular settings. Recordings may be
made using videotapes or behavioral records. In one type of observational method
known as participant-observation, the researcher participates in the activities of the
respondents. For example, a researcher may wish to find out about the behavior of
staff in a nursing home. The researcher may spend several days living with people in
the nursing home. The researcher’s subjective experiences would become part of the
‘‘data.’’ There are elaborate procedures available for creating behavioral records in
which the researcher precisely defines the behavior to be observed (the number of
particular acts) and specifies the times during which records will be made. This
procedure may be used to determine whether an intervention is having its intended
effects. If an investigator is testing a method to reduce aggressive behavior in people
with Alzheimer’s disease, behavioral records could be made before and after the
intervention is introduced. After observing the effects of the intervention, the
method’s effectiveness could be determined by a return to baseline condition to assess
whether the aggressive behavior increases without the intervention.
f. Measurement Issues in Adult Development and Aging
Research designs, no matter how cleverly engineered, are unable to yield
worthwhile results if the methods used to collect the data are flawed. Researchers in
adult development and aging, like all other scientists, must concern themselves with
the quality of the instruments they use to capture data. The task is made more difficult
because the instruments must be usable with people who are likely to vary in ability,
educational background, and sophistication with research instruments. Earlier we
pointed out the problems involved in comparing older and younger adults on
measures used in cross-sectional studies. Here we will look specifically at some of the
ways developmental researchers can ensure that their measures are equivalent across
age groups.
The first measurement issue to consider is that of reliability. A measure is
reliable if it yields consistent results every time it is used. The importance of
reliability is highlighted by considering the analogy of measurements used in cooking.
If your tablespoon were unreliable (say, it was made of floppy plastic), the amount of
ingredients added would vary with every use. Your cookies may be hard and crunchy
one time and soft and gooey the next. A psychological test must also provide similar
scores upon repeated administration. This principle is one of the first qualities
psychologists look for in a measure—its ability to provide consistent scores.
Reliability can be assessed by test-retest reliability, which is determined by giving the
test on two occasions to assess whether respondents receive similar scores across both
administrations. Another form of reliability relates to internal consistency, which
indicates whether respondents answer similarly on comparable items. The second
criterion used to evaluate a test is validity, meaning that the test measures what it is
supposed to measure. A test of intelligence should measure intelligence, not how good
your vision is. Returning to the example in the kitchen, if a tablespoon were marked
‘‘teaspoon,’’ it would not be measuring what it is supposed to measure, and your
baked products would be ruined. Tablespoons are fairly easy to assess for validity, but
unfortunately psychological tests present a far greater challenge. For this reason,
validity is a much more difficult quality to capture than reliability.
The concept of validity varies depending on the intended use of the measure.
Content validity provides an indication of whether a test designed to assess factual
material accurately measures that material. Your next exam may very well include
questions testing how well you have understood. Criterion validity indicates whether a
test score accurately predicts performance on an indicator measure, as would be used
in a test of vocational ability that claims to predict success on the job. Finally,
construct validity is used to assess the extent to which a measure intended to assess a
psychological construct is able to do so. Construct validity is difficult to establish and
requires two types of evidence: convergent and divergent validity. Convergent validity
is needed to determine that the measure relates to other measures that are of
theoretical similarity. A test of intelligence should have a positive relationship to
another test of intelligence that has been well-validated. Divergent validity
demonstrates that the measure does not relate to other measures that have no
theoretical relationship to it. A test of intelligence should not be correlated with a test
of personality, unless the personality test assesses some aspect of intelligence.
Although psychologists are generally aware of the need to establish the
reliability and validity of measures used in both research and practical settings, less
attention is focused on psychometric properties when used in gerontological research.
Measures whose reliability and validity were established on young adult samples are
often used inappropriately without testing their applicability to samples of adults of
varying ages. The process can become quite complicated. If Form A of a measure is
found to be psychometrically sound with college students but only Form B has
adequate reliability and validity for older adults, the researcher is faced with the
prospect of having to use different forms of a test within the same study. Nevertheless,
sensitivity to measurement issues is crucial if conclusions drawn from the research are
to have value.
g. Ethical Issues in Research
All scientists who engage in research with humans or other animals must take
precautions to protect the rights of their participants. In extreme cases, such as in
medical research, the life of an individual may be at stake if the individual is
subjected to risky procedures. Research in which respondents are tested or put
through stressful experimental manipulations also requires that standard protocols are
followed. Recognizing the importance of these considerations, the American
Psychological Association developed a comprehensive set of guidelines for
psychologists that includes the appropriate treatment of human participants in
research.
Researchers must present a potential respondent as full a disclosure as possible
of the risks and benefits of becoming involved in any research project. When the
individual is a minor child or an adult who is not able to make independent decisions,
the researcher is obligated to inform the individual’s legal guardian about the nature
of the study. Having provided information about the study, the researcher must then
obtain a legal signature indicating that the participant understands the risks and
benefits involved in the study. At this point the researcher is able to obtain the full
informed consent of the respondent or the respondent’s legal representative. When the
individual is an animal, the researcher is similarly bound to ensure that the animal is
not mishandled or subjected to unnecessary harm, although different protective
procedures are followed.
Research participants are also entitled to know what the study was about after
completion in a process called debriefing. If you have ever participated in research,
you may have been curious during the course of the study to know what was being
tested. In some cases you might have been surprised to find out about the ‘‘real’’
purpose of the study. Perhaps you were told that you were going to be asked to fill out
a series of questionnaires in a quiet laboratory room. In the middle of the
questionnaire, you hear a loud noise in the hallway, followed by a man screaming.
Although you may think the point of the study is to gather information about you, the
goal of the research was to assess your response to the sudden commotion. After the
experiment is over, the researcher is obligated to tell you the truth about the purpose
of the study. You may be embarrassed if in fact you did not get up to help, but you at
least had a right to know that you were being tested on this attribute. The debriefing
might make you feel better because you would realize that your response of not
helping reflected the experimental manipulation rather than a malevolent personality
attribute you possess.
As this example illustrates, research participants may learn information about
themselves that is potentially upsetting or damaging. In fact, ethical guidelines for
research in psychology dictate that the researcher not only provides feedback but also
must be ready to suggest support or counseling for people who become distressed
while involved in the experiment. Respondents are also entitled to withdraw from a
study without risk of penalty should they choose to do so. The experimenter should
not coerce them into completing the study, and even if they decide to discontinue
participation, they should still receive whatever reimbursement was initially
promised. If they are students in a class or clients receiving services (such as hospital
patients), they should not fear having their grades lowered or services withheld from
them. Finally, research participants are entitled to know what will happen to their
data. In all cases, the data must be kept confidential, meaning that only the research
team will have access to the information provided by the participants. The other
condition usually attached to the data is that of anonymity. Participants are guaranteed
that their names will not be associated with their responses. The condition of
anonymity obviously cannot be kept if the study is a longitudinal one because the
researchers must maintain access to names for follow-up purposes. In this case, the
condition of confidentiality applies, and the researchers are obligated to ensure that all
records are kept private and secure.
These ethical standards are enforced in all institutions receiving federal or
local funding for research through Institutional Review Boards (IRBs), which review
all proposed studies to be carried out at that institution or by anyone employed by that
institution. These reviews ensure that the rights of research participants are adequately
protected according to the criteria discussed above. In addition, the American
Psychological Association’s ethical guidelines ensure that studies conducted
specifically in the field of psychology meet predetermined criteria for protection of
human and animal subjects. An important development in the area of protection of
human participants was the implementation in April 2003 of national standards within
the United States to protect the privacy of personal health information.
