11-Hofer-45484.

qxd 2/26/2008 12:00 PM Page 181

11
DIMENSIONS OF COGNITIVE AGING
Executive Function and Verbal Fluency

SUSAN KEMPER AND JOAN M. MCDOWD

A
persistent issue in cognitive aging is have been implicated as contributing to a wide
whether all aspects of cognition are range of linguistic and cognitive abilities; for
equivalently affected by developmental example, verbal fluency involves not only
processes or whether some are differentially semantic knowledge of lexical items and the
spared from developmental decline (Zelinski & ability to search semantic memory using phono-
Lewis, 2003). This argument has been nowhere logical or categorical rules but also “executive”
more contentious than in debates concerning the skills required to track prior responses and block
effects of aging and age-associated diseases such intrusions from other semantic categories.
as Alzheimer’s disease (AD) and Parkinson’s Complicating this question has been the problem
disease (PD) on language use. Linguistic tasks of defining and measuring EF.
are commonly used to assess cognitive status Measures of EF have proliferated over the
and neuropsychological impairments; they past half-century. In their review of executive
include tests of verbal fluency, vocabulary, and function, Royall et al. (2002, Table 11.3) listed
prose comprehension and recall. A related 46 studies conducted between 1983 and 2001
concern has been whether these tests measure that involve factor analyses of EF. These studies
discrete, autonomous linguistic abilities employed 34 different measures of EF and
(Fodor, 1982; Waters & Caplan, 1996) or com- report factor structures ranging from one to four
posite abilities that draw on multiple cognitive independent EF factors. This summary indicates
domains, including working memory and execu- that there is little consensus regarding appropri-
tive function (Just & Carpenter, 1992). In partic- ate measures of EF or about the underlying
ular, impairments of executive function (EF) structure of EF. Some of this heterogeneity

AUTHORS’ NOTE: We thank Laura Berman, Pat Laubinger, Kelly Lyons, Alicia MacKay, and Kim Metcalf for
their assistance with data collection. This research was supported in part by grants from the National Institutes of
Health to the University of Kansas through the Mental Retardation and Developmental Disabilities Research
Center (Grant P30 HD-002528); and the Center for Biobehavioral Neurosciences in Communication Disorders
(Grant P30 DC-005803); as well as by Grants RO1 AG06319, K04 AG000443, P30 AG10182, RO1 AG09952,
RO1 AG18892, and RO1 AG025906 from the National Institute on Aging. Its contents are solely the responsibility
of the authors and do not necessarily represent the official views of the National Institutes of Health.

181
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 182

182– • –DIMENSIONS OF COGNITIVE AGING

comes from the use of different populations “best” neuropsychological measure of EF based
(e.g., people with schizophrenia, children, col- on factor loadings and the Stroop task as the best
lege students, older adults, individuals with AD) measure of inhibition, a keeping track task as the
both within and between studies. For example, best measure of updating, and a paired-associate
Kanne, Balota, and Storandt (1998) found that learning task combined with keeping track task
people with AD had a different EF factor struc- as the best measure of time sharing. However,
ture than did the elderly control participants. In their results indicated that the various neuropsy-
that study, the normal older adults showed a chological measures were not very highly related
single-factor structure for EF, whereas the to one another (i.e., had low convergent validity)
patients with AD showed a three-factor structure and were fairly highly related to other variables,
with mental control, verbal memory, and visuo- particularly fluid intelligence, indicating low dis-
spatial factors. In contrast, a study involving criminant validity. The same pattern of results
only people with AD produced a single factor was reported for the cognitive process variables:
structure (Loewenstein et al., 2001). Likely rel- little convergent or discriminant validity for inhi-
evant is the fact that Loewenstein et al. (2001) bition, updating, or time-sharing abilities. All
administered only three measures of EF, and were fairly highly correlated with fluid intelli-
Kanne et al. administered five; only two tasks gence. The authors concluded that individual dif-
overlapped in the two studies. A further contrast ferences in measures of EF may in fact reflect
is provided by a study healthy aging adults by differences in much broader abilities, such as
Royall, Chiodo, and Polk (2003); they reported fluid intelligence.
a three-factor structure: (1) Procedural Control, Miyake and colleagues (Friedman & Miyake,
(2) Abstraction, and (3) Attention Switching. 2004; Miyake, Emerson, & Friedman, 2000;
Together, these studies point to important ques- Miyake et al., 2000) have addressed similar
tions that remain to be answered: What is the questions but take a somewhat different
cognitive structure underlying the construct of approach and reach different conclusions. For
EF? Does the structure vary with age or clinical example, Miyake, Friedman, et al. (2000) noted
population? What are the best measures of EF? that descriptions of task requirements underlying
Two recent studies highlight the continuing the same EF measure may vary widely, and
need to resolve these issues. Salthouse, perhaps rightly so because such measures are
Atkinson, and Berish (2003), noting the com- typically complex and may require several cog-
plexity and breadth of notions of EF, undertook nitive processes to carry out successfully. At the
an examination of the construct validity of EF in same time, different EF measures may require
a sample of 261 adults ranging in age from 18 to different cognitive processes for good perfor-
84 years. Their approach was to examine con- mance. To address these issues, Miyake,
vergent and discriminant validity among a set of Friedman, et al. (2000) reported a study address-
neuropsychological and cognitive tasks typically ing “the unity and diversity of executive func-
associated with EF and a set of psychometric tions” (p. 49) using confirmatory factor analysis
tasks. The neuropsychological tasks in their and structural equation modeling. Like the
battery included the Wisconsin Card Sorting Salthouse et al. study, Miyake, Friedman, et al.
Test; the Connections test (based on the Trail- identified a set of EF tasks used in neuropsycho-
Making Test), the Tower of Hanoi, verbal flu- logical studies: the Wisconsin Card Sorting Test,
ency, and figural fluency. Three additional sets of the Tower of Hanoi, a random number genera-
tasks designed to assess component (cognitive) tion test, the operation span task, and a dual task.
processes of EF included measures of time They also identified a set of tasks designed to
sharing, updating, and inhibitory processes. measure three subtypes of EF: (1) shifting,
Psychometric measures included indices of ver- (2) updating, and (3) inhibition. To address the
bal ability, fluid intelligence, episodic memory, question of unity and/or diversity of functions,
and perceptual speed. A series of structural equa- they first conducted confirmatory factor analyses
tion analyses were then conducted to look at using data from 137 college students performing
the relations among these sets of variables. the EF subtype measures. They found that a
Salthouse et al. identified verbal fluency as the three-factor solution fit the data better than any
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 183

Executive Function and Verbal Fluency– • –183

of the one- or two-factor solutions, indicating A second issue concerns the task impurity
that there are separable dimensions of EF. They problem (e.g., Miyake, Friedman, et al., 2000).
identified their letter memory and keeping track Many measures of EF are relatively complex,
tasks as the best measure of updating, the involve a variety of component processes that
plus/minus and number/letter tasks as the best may or may not be part of EF, and are sometimes
measures of shifting, and the antisaccade and also used to assess other cognitive processes. One
Stroop tasks as the best measures of inhibition. example of such a task is verbal fluency. Verbal
In addition to these analyses, Miyake, Friedman, fluency tasks, sometimes termed generative nam-
et al. (2000) also tested a series of structural ing, typically require the person to generate as
equation models to examine the contribution of many words as possible meeting a criteria in a set
these separable factors to the more complex neu- amount of time. Many variations of the standard
ropsychological EF tasks. Using structural equa- letter fluency task have been developed including
tion modeling, they reported that the shifting generating words beginning with a target letter,
factor is most relevant for Wisconsin Card words belonging to a semantic category, items
Sorting Test performance, the inhibition factor is occurring in a supermarket, and so on. Counts
most relevant for the Tower of Hanoi, and that of valid responses and various types of errors
both inhibition and updating are relevant for the (e.g., perseverations and intrusions) are typically
random number generation task. The authors assessed.
concluded from this study that the three EF func- First introduced by Borkowski, Benton, and
tions they measured (updating, shifting, and Spreen (1967) and Benton (1968), measures of
inhibition) are “clearly distinguishable” and that verbal fluency are frequently part of batteries
each plays a different role in more complex EF designed to assess EF in adults and children and
measures, such as the Wisconsin Card Sorting diagnosis mild cognitive impairment and
Test and the Tower of Hanoi. dementia due to AD, PD, HIV/AIDs, or other
Given these conflicting conclusions, there is a disorders; verbal fluency measures are also used
clear need to resolve this controversy regarding to assess semantic memory or word knowledge
the structure of EF. Furthermore, neither of these and the impact of strokes and other closed head
studies directly addresses how aging affects EF, injuries, developmental disorders, and clinical
and they leave unresolved how best to assess EF conditions on semantic processes.
in clinical populations. The Miyake, Friedman, Verbal fluency was traditionally assumed to
et al. (2000) study involved only young adults, correspond to more general notions of discourse
and the question remains whether the same pat- fluency, although there is no widely agreed-on
tern of results would be obtained with a sample definition or measure of discourse fluency.
of older adults. Indeed, in a partial replication of Discourse fluency is commonly assumed to
Miyake, Friedman, et al.’s study with adults 20 involve word retrieval, sentence formulation,
to 81 years of age, Fisk and Sharp (2004) found and articulation processes and to be subject to
support for four factors: three corresponding to lapses of attention, memory limitations, and
Miyake, Friedman, et al.’s updating, inhibition, motor and articulatory control problems.
and shifting components and a fourth, word Discourse is marked by many types of dysfluen-
fluency factor. Salthouse et al.’s (2003) study cies: interjections, filled and unfilled pauses, and
involved both young and old adults, but the age sentence fragments. Fillers, defined as speech
groups were not analyzed separately, and, based serving to fill gaps in the speech flow, include
on the arguments of Hofer and Sliwinski (2001) both lexical and nonlexical fillers. Fillers may
and Hofer, Flaherty, and Hoffman (2007), there serve pragmatic and discourse functions (Fox
is some reason to predict that the pattern of Tree, 1995) or reflect word finding problems or
results might be different for the two groups. other breakdowns in semantic retrieval, syntac-
Finally, neither study examined EF in clinical tic planning, or sentence production. Nonlexical
populations. We can only echo Rabbitt’s (1997) fillers, such as “uh,” “umm,” “duh,” and so on,
lament: “Life would be simpler if there were are also common (Bortfield, Leon, Bloom,
generally agreed paradigmatic ‘executive’ tasks Schober, & Brennan, 2000; Brennan & Schober,
available for rigorous empirical analysis” (p. 8). 2001; Ferber, 1991) and are generally considered
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 184

184– • –DIMENSIONS OF COGNITIVE AGING

to reflect problems in sentence and discourse increment with every additional word recalled.
planning. According to Mayr and Kliegl’s reasoning, the
Some forms of dysfluency, such as unfilled constant (intercept) of this function represents
pauses; circumlocutions; “empty speech,” such executive functions that support or enable seman-
as pronouns lacking clear referents; and substi- tic retrieval processes—such as systematic
tution errors (substituting he for she) during searching and avoiding perseverations and non-
spontaneous speech have also been noted in category intrusions. The slope parameter repre-
older adults and may reflect age-related impair- sents semantic retrieval processes.
ments in accessing and retrieving lexical infor- They tested their view with a fluency task
mation (Obler, 1980; Ulatowska, Cannito, that included a manipulation of semantic
Hayashi, & Fleming, 1985). Burke, Worthley, retrieval difficulty by varying the frequency and
and Martin (1988) observed that word finding familiarity of the categories and a manipulation
problems are common in the speech of older of demands on EF by including both normal and
adults and often result in tip-of-the-tongue expe- “switching” versions of fluency tasks. This
riences. Burke and her colleagues (Burke & switching manipulation was motivated by a sug-
Laver, 1990; Burke, MacKay, Worthley, & gestion (Troyer, Moscovitch, & Winocur, 1997)
Wade, 1991) suggest that aging affects the abil- that executive processes are particularly relevant
ity to retrieve complete phonological informa- in memory retrieval tasks that require switching
tion about words, resulting in the retrieval of between semantic clusters. Thus, participants
partial phonological information characteristic were asked to produce category exemplars,
of tip-of-the-tongue experiences. either blocked by category in the typical way
Kemper and Sumner (2001) compared mea- (e.g., animal, animal, animal, etc.) or alternating
sures of verbal ability, including initial letter and between two categories (e.g., animal, tool, ani-
category fluency measures, obtained from a mal, tool, animal, tool, etc.). They hypothesized
group of young adults, 18 to 27 years of age, and that the category difficulty manipulation should
a group of older adults, 63 to 88 years of age. For affect the slope parameter, and the switching
older adults, Kemper and Sumner found that ini- manipulations should affect the constant para-
tial letter fluency and category fluency were meter. They also hypothesized that age differ-
related to other measures of processing effi- ences in the constant parameter would indicate
ciency, such as reading rate. Processing effi- age differences in EF in fluency tasks, whereas
ciency appeared to impose general limitations on age differences in the slope parameter would
task performance by older adults, affecting how indicate age differences in semantic processes.
efficiently they can search their mental lexicon Mayr and Kliegl (2000) observed no age dif-
for words with the appropriate initial letter and ferences for the no-switch condition in the slope
how efficiently they can search their memory for parameter, although the difficulty manipulation
answers to comprehension questions. In con- did affect slope as expected, but equally so for
trast, young adults’ performance on the fluency young and old adults. In contrast, there was a
tasks appeared to be constrained by their knowl- significant age difference in the constant para-
edge of lexical items as measured by vocabulary meter, assumed to reflect EF. “This pattern [sup-
tests. Hence, verbal fluency tasks may measure ports] the assumption of no age effects in
semantic knowledge in young adults but pro- semantic processing (i.e., the slope parameter),
cessing efficiency or EF in older adults. but age sensitivity in non-semantic, executive
Mayr and Kliegl (2000) recently developed an processes” (Mayr & Kliegl, 2000, p. 36). These
approach to disentangle the semantic and execu- findings illustrate the importance of separating
tive components of fluency performance. In their component processes in order to identify the
task, fluency performance is audio-recorded. ones underlying group differences in task perfor-
Interword response intervals are computed and mance. In the case of older adults in Mayr and
modeled as a function of retrieval position: tn = c Kliegl’s study, EF appears to be responsible for
+ s × n, where tn is the time between the recall of the group differences in fluency performance.
word n – 1 and word n, c is a constant represent- It is interesting that the switch condition
ing EF, and s is the slope representing the time did not produce larger age differences in the
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 185

Executive Function and Verbal Fluency– • –185

constant parameter than the no-switch condi- participants to name exemplars of unfamiliar
tion. Mayr and Kliegl (2000) concluded that EF categories (e.g., insects, writing utensils, fabrics,
impairment specific to switching is not the fluids), (3) an easy switching version requiring
source of age differences in fluency perfor- participants to name exemplars of two categories
mance. They suggested that updating and set (e.g., birds alternating with body parts), and (4) a
maintenance may be more likely candidates for hard switching version requiring participants to
age-impaired EF functions. name exemplars of two categories (e.g., writing
In summary, Mayr and Kliegl’s (2000) utensils and fluids). Instructions and procedures
approach and findings illustrate two important followed those of Mayr and Kliegl with the excep-
points. First, there are significant interpretive tion that categories were presented on cue cards
problems when task components are poorly placed before the participants rather than on a
understood, and task performance could be computer monitor. All responses were digitally
affected by any of them. Second, separate analy- recorded for later analysis.
sis of task components is very useful in identify- The results are summarized in Figures 11.1
ing specific deficits. In the case of normal aging, and 11.2. The critical dependent measure is the
fluency task performance deficits appear to be interword response interval, timed from the off-
due more to EF deficits than to semantic memory set of one response to the onset of the next
differences. A somewhat different approach to response. Slopes and constant parameters from
disentangling semantic function and EF in verbal a linear regression analysis of these data are also
fluency has been taken by Troyer and her col- presented.
leagues (Troyer et al., 1997; Troyer, Moscovitch, Mayr and Kliegl (2000) argued that the slope
Winocur, Alexander, & Suss, 1998; Troyer, of these functions represents semantic search;
Moscovitch, Winocur, Leach, & Freedman, we found that the slopes increased with task dif-
1998; but see Mayr, 2002; Troster et al., 1998). ficulty and that the slopes were greater for the
To demonstrate the feasibility of Mayr and participants with PD than for the healthy older
Kliegl’s (2000) approach to the analysis of flu- adults, suggesting that PD does affect the effi-
ency tasks, we have conducted a series of small ciency of semantic processing. Mayr and Kliegl
pilot studies using available participant pools also argued that the constants represent nonse-
from the Parkinson’s Center and the Center for mantic EF; our results show that the constants
Brain Aging at the University of Kansas increased with task difficulty and were greater
Medical Center. Our first study was a small pilot for the participants with PD than the group of
study with 10 healthy older adults (mean age: healthy older adults, suggesting executive func-
73.4 years; mean education: 15.2 years) and 10 tions also contribute to group differences in flu-
older adults with Parkinson’s disease (mean age: ency performance.
72.8 years; mean education: 16.1 years). All To further explore this approach, we con-
participants with PD were taking Carbidopa- ducted a second pilot study with 30 healthy
Levodopa; all were 5 to 6 years postinitial diag- older adults and 30 older adults with PD. This
nosis and had Hoehn and Yahr (1967) scores of study was designed to test the validity of Mayr
1 or 2. None of the participants had a history of and Kliegl’s (2000) interpretation of slope and
heart disease, cancer, neurological disorders, or constant parameters. In this pilot, we regressed
alcoholism. All were native speakers of English. the participants’ slopes and constants from the
None were taking anticholinergics, antidepres- category fluency tests onto measures of EF and
sants, or anxiolytics. There were 7 women and verbal ability. The selection of tests was dictated
3 men in the group of healthy older adults and by individuals currently assessed by the
6 women and 4 men in the group with PD. Parkinson’s Center; they do not provide an ideal
All participants were given four versions of a test of our hypotheses. We hypothesized that the
category verbal fluency test, modeled after those constant parameter should be associated with
of Mayr and Kliegl (2000). They were: (1) an easy EF; that is, participants who score poorly on EF
version requiring participants to name exemplars tests should have higher constants, if the con-
of familiar categories (e.g., birds, clothes, body stants reflect executive function. From the tests
parts, and colors), (2) a hard version requiring available, we selected the Wisconsin Card
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 186

186– • –DIMENSIONS OF COGNITIVE AGING

No-Switch Semantic Categories

20

15
Interword Interval(s)

10

0
1 2 3 4 5 6 7 8 9
Retrieval Position

Older Adults - Easy Older Adults With PD - Easy

Older Adults - Hard Older Adults With PD - Hard

Figure 11.1 Interword Response Times for Easy and Hard Verbal Fluency Tests Administered to Healthy
Older Adults and Older Adults With Parkinson’s Disease: No-Switch Condition

Switch Semantic Categories

20

15
Interword Interval(s)

10

0
1 2 3 4 5 6 7 8 9
Retrieval Position

Older Adults - Easy Older Adults with PD - Easy

Older Adults - Hard Older Adults with PD - Hard

Figure 11.2 Interword Response Times for Easy and Hard Verbal Fluency Tests Administered to Healthy
Older Adults and Older Adults With Parkinson’s Disease: Switching Condition
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 187

Executive Function and Verbal Fluency– • –187

Sorting test (Heaton et al., 1993) and the FAS We then defined a first-order factor for EF by
fluency test (Spreen & Benton, 1977) as the loading scores for the FAS and Wisconsin Card
measures of EF. In addition, the slope parameter Sorting tests onto a single factor. Two sets of
might be expected to be related to vocabulary regression analyses were then performed. In a
scores, which typically is indexed by verbal test of our hypothesis that EF should be related
ability. Scores on the Shipley Institute of Living to the constant parameter, we regressed the
Scale vocabulary subtest (Shipley & Zachary, slopes and constants for the switching conditions
1940), as well as years of formal education, onto the EF factor scores in Step 1 and then
were available for use as measures of verbal entered the verbal ability covariates (Shipley
ability. In addition, a backward digit span test vocabulary score and years of education) in Step 2.
had been administered, and we considered this In Step 3, we entered the working memory mea-
to be a measure of working memory. These sure (backward digit span) and age.
hypotheses were tested in a group of 20 older The results are summarized in Table 11.3.
adults and 8 older adults with PD. Mean perfor- The results of Step 1 indicate that individual
mance levels on demographic, verbal ability, variation in EF accounts for 6% to 18% (for
and EF variables are shown in Table 11.1. healthy older adults) and 42% to 58% (for adults
Participants were tested on the easy and hard with PD) of the variance in the constants for the
category fluency tests, described above, in both switching conditions but little (< .01–.04) of the
no-switching and switching conditions. The variance in the slopes. Adding verbal ability
results are summarized in Table 11.2. Consistent to the regression models in Step 2 accounts
with our previous findings, we found slope dif- for little additional variance in the constants
ferences between easy and hard categories, (.02% – .05% increase in R2) but adds to the pre-
reflecting ease of semantic retrieval; further- diction of individual variation in the slopes; ver-
more, slopes for healthy older adults were lower bal ability accounts for a 10% to 60% increase
than those for individuals with PD, suggesting in R2. Adding the remaining covariates in Step 3
that PD does affect the efficiency of semantic provided no additional increase in the fit of the
retrieval. regression equations. We also reversed the order
We also found that the constants were greater of entry, entering the verbal ability measures in
for the switch conditions than for the no-switch Step 1 and the EF measures in Step 2, with sim-
conditions and greater for individuals with PD ilar results: The verbal ability measures
than for the healthy older adults, suggesting a accounted for 25% to 35% of the variance in the
breakdown of EF with PD. slopes for the switching trials in Step 1 but only

Table 11.1 Comparison of the Healthy Older Adults and Older Adults With Parkinson’s Disease Who
Participated in the First Pilot Study

Older adults with

Healthy older adults Parkinson’s disease

M SD M SD

Variable
Age 75.7 4.3 73.3 5.7
Education 17.2 3.2 14.6 2.3
Shipley Vocabulary 32.9 5.2 29.6 6.8
Backward digit span 6.9 2.1 7.0 1.2
FAS Verbal Fluency 36.4 14.5 38.2 12.8
Wisconsin Card Sorting
Categories 2.5 1.5 1.3 1.5
Total errors 21.2 10.4 29.5 13.1
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 188

188– • –DIMENSIONS OF COGNITIVE AGING

Table 11.2 Comparison of the Easy and Hard Verbal Fluency Tests Administered to Healthy Older Adults
and Older Adults With Parkinson’s Disease and Alzheimer’s Disease

Slopes Constants

No switch Switch No switch Switch

Easy Hard Easy Hard Easy Hard Easy Hard

Participant
Healthy older adults 0.69 1.95 0.43 1.36 –1.00 –0.16 0.46 0.74
Older adults with
Parkinson’s disease 1.16 2.39 0.17 2.56 –1.90 –0.41 0.43 2.74
Older adults with
Alzheimer’s disease 1.74 4.12 2.10 5.54 1.76 2.22 1.95 3.88

Table 11.3 Results of the Regression Analysis (Total R2) of the Slopes and Constants Obtained From the
Easy and Hard Switching Fluency Tests

Healthy older Older adults with

adults Parkinson’s disease

Slope Constant Slope Constant

Step
Step 1: Executive function
Easy categories .06 .06 .02 .42
Hard categories .14 .18 0.8 .58
Step 2: Verbal ability
Easy categories .27 .10 .03 .52
Hard categories .41 .22 .13 .61

0% to 8% of the variance in the constants, and given in our second pilot study; however, we
the EF measures accounted for 15% to 46% were constrained in our choice of covariates to
additional variance in the constants in Step 2 but those currently assessed by the Brain Aging
only 2% to 5% additional variance in the slopes. Center. The Wisconsin Card Sorting task was
We also conducted a third pilot study with 8 not administered; however, the Trail Making
AD patients recruited from the Brain Aging Test (Reitan & Wolfson, 1995) was adminis-
Center. All participants with AD were taking tered, so we used the proportional difference
Aricept, and 5 of the 8 were also taking score (Test A – Test B/Test A) as a measure of
Nemenda; all were judged to have mild to mod- EF (M = –0.68, SD = 0.55). The Boston Naming
erate AD. None of the participants had a history Test (Kaplan, Goodglass, & Weintraub, 2001;
of heart disease, cancer, other neurological dis- M = 11.63, SD = 16.69), rather than the Shipley
orders, or alcoholism. All were native speakers vocabulary test, was available as a measure of
of English. None were taking anticholinergics, verbal ability, along with education (M = 17.4,
antidepressants, or anxiolytics. We administered SD = 3.2). Backward digit span scores (M = 5.4,
the same easy and hard category fluency tests SD = 1.5) and age (M = 76.0, SD = 6.7), as well
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 189

Executive Function and Verbal Fluency– • –189

as an overall rating of dementia severity (Mini- the order of entry, entering the verbal ability mea-
Mental Status Examination [Folstein, Folstein, sures in Step 1 and the EF measures in Step 2,
& McHugh, 1975]; M = 24.5, SD = 5.3) were with similar results: The verbal ability measures
also available and entered in Step 3. accounted for much (49%–74%) of the variance
The slopes and constants from the category in the slopes for the switching trials in Step 1 but
fluency tests for the older adults with AD are only 10% to 15% of the variance in the constants,
also presented in Table 11.2. The slopes for and the EF measures accounted for 38% to 54%
older adults with AD were greater than those for additional variance in the constants in Step 2 but
healthy older adults as well as those for older only 3% to 16% additional variance in the slopes.
adults with PD, confirming our hypothesis that In comparing the results of our regression
AD affects the efficiency of semantic retrieval. analysis of fluency task performance by individ-
We also found an effect of AD on the constants, uals with AD and PD we note that the EF and
indicating that AD also affects the efficiency of verbal abilities covariates account for consider-
EF (see Ullman et al., 1997). ably more variance in the performance of the
The results of the regression analysis are sum- older adults with AD on the switch trials than for
marized in Table 11.4. In Step 1, we entered the the older adults with PD. Much of the variance in
EF measure, the proportional difference score the constants for the older adults with PD
from the Trail Making Test. In Step 2, we entered remains unexplained. It is likely that unmeasured
the verbal ability measures, the Boston Naming variance in the severity of PD may contribute to
Test and education. In Step 3, we entered age, the variance in the slopes on the switch trials.
backward digit span, and Mini-Mental State Together, this series of pilot studies sup-
Examination score. The results of Step 1 indicate ports the validity of Mayr and Kliegl’s (2000)
that individual variation in EF accounts for much analysis of the components of fluency tasks;
(48% – 75%) of the variance in the constants for although very preliminary, our findings indi-
the switching conditions but less (24%–28%) of cate a dissociation of verbal and executive
the variance in the slopes. Adding verbal ability functions in fluency task performance in terms
to the regression models in Step 2 accounts for of their relation to the slope and intercept para-
little additional variance in the constants but adds meters derived from the fluency task data. In
to the prediction of individual variation in the addition, these results show that this approach
slopes; verbal ability accounts for a 10% to 60% of decomposing EF tasks, in this case fluency
increase in R2. Adding the remaining covariates tasks, can lead to a better understanding of the
in Step 3 provided little additional increase in the component cognitive processes that affect per-
fit of the regression equations. We also reversed formance in those tasks. These results also
demonstrate how this approach will advance
our understanding of AD and PD by revealing
Table 11.4 Results of the Regression Analysis
(Total R2) of the Slopes and
how these age-associated diseases affect
Constants Obtained From semantic processes and EF.
the Easy and Hard Switching
Fluency Tests: Older Adults
With Alzheimer’s Disease CONCLUSION

Decomposing fluency tasks into their compo-

Step Slope Constant nent processes may provide an answer to
Rabbitt’s (1997) lament (see p. 183, this chapter):
Step 1: Executive function Life would be easier.
Easy categories .28 .48 Life would be easier because verbal fluency
Hard categories .24 .73 tasks may be sensitive tests of EF for use in clin-
Step 2: Verbal ability ical contexts as well as for population-based
Easy categories .38 .59
assessments of cognition. Impairments of EF
Hard categories .74 .75 have been linked to functional impairments
(Cahn-Weiner, Malloy, Boyle, Marran, & Salloway,
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 190

190– • –DIMENSIONS OF COGNITIVE AGING

2000; Carlson et al., 1999; Farias, Harrell, coupled with the automated analysis of inter-
Neumann, & Houtz, 2003; Grigsby, Kaye, word response times could help identify how
Baxter, Shetterly, & Hamman, 1998; Royall, population demographics and health status affect
Palmer, Chiodo, & Polk, 2004), although the lack verbal/semantic processes versus EF. Indeed,
of an easily administered “gold standard” for EF automatic speech recognition could add a new
has hindered research on this issue. dimension to existing test batteries by permitting
The traditional FAS verbal fluency test is automatic scoring of verbal responses as well as
commonly included with the Wisconsin Card the measurement of response times. The Center
Sorting Task and the Trail Making Test in neu- for Spoken Language Understanding makes
ropsychological test batteries. Although the available a toolkit of techniques to map speech
Wisconsin Card Sorting Task has been used input onto text output. Using these techniques, it
extensively, it is time consuming to administer is possible to approximate human speech recog-
and score, although computerized versions are nition under ideal circumstances with around
available. The Trail Making Test is also widely 90% accuracy; higher rates of accuracy can be
used, and it is easily administered. However, obtained by training the systems on the acoustic
both Wisconsin Card Sorting Task and Trail characteristics of individual speakers, groups of
Making Test lack convergent and discriminant speakers (e.g., elderly speakers) or models of
validity. No-switching and switching fluency accented speech. Further accuracy can be gained
tests could be easily automated using existing by limited word or response sequences and lexi-
voice recognition techniques, making these tests cal inventories and limited environmental or sys-
suitable for administration in clinical settings tem noise. These techniques are now capable of
using only an audio- or video-recorder. Our running in real time on desktop computers. It is
pilot studies suggest that the decomposition of further possible to develop add-ons that measure
interword response times, particularly on pause duration and interword response times
switching trials, may discriminate among the (Hosom, Shriberg, & Green, 2004). Tools such
effects of aging, PD, and AD on semantic and as the Coh-Mex system of Graesser, McNamara,
executive processes. Louwerse, and Chai (2004) are now available to
Life would also be easier because this perform extensive text-based analyses, statistical
approach provides a significant guidepost to the analyses of word frequencies, extract measures
future of cognitive aging studies. For far too long of coherence, syntactic complexity, and semantic
we have relied on cognitive and neuropsycho- content. These technologies are now available,
logical tests of dubious construct validity. We and their application could open up new
have failed to bring to bear the vast armature of approaches to clinical and population-based
contemporary cognitive psychology to the task assessments of health and cognition.
of decomposing these tasks into their component
processes. Parameterizing task performance—
REFERENCES
into slopes and intercepts, time–accuracy func-
tions, Rasch scaling, or other techniques—will Benton, A. L. (1968). Differential behavioral effects of
yield a new generation of tests that can provide frontal lobe disease. Neuropsychologia, 6, 53–60.
insights into the differential effects of aging and Borkowski, J., Benton, A., & Spreen, O. (1967). Word
clinical conditions; clarify neuroimaging find- fluency and brain damage. Neuropsychologia, 5,
ings regarding brain–behavior linkages; and 135–140.
advance our understanding of functional status, Bortfield, H., Leon, S. D., Bloom, J. E., Schober,
care needs, and rehabilitation outcomes. M. F., & Brennan, S. E. (2000). Disfluency rates
Life would be easier because parameterized in spontaneous speech: Effects of age, relation-
verbal fluency tests could add a new level of ship, topic, role, and gender. Memory and
sophistication to large, population-based studies Language, 44, 123–147.
of health and cognition. Currently, cognition Brennan, S. E., & Schober, M. F. (2001). How listen-
is assessed by tests such as immediate and ers compensate for disfluencies in spontaneous
delayed word recall, counting backward, identify- speech. Memory and Language, 45, 274–296.
ing words from definitions, and subtracting Burke, D. M., & Laver, G. D. (1990). Aging and word
7s. Switching and no-switching fluency tests retrieval: Selective age deficits in language. In
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 191

Executive Function and Verbal Fluency– • –191

E. A. Lovelace (Ed.), Aging and cognition: Mental A latent-variable analysis. Journal of Experimental
processes, self-awareness, and interventions Psychology: General, 133, 101–135.
(pp. 281–300). New York: Elsevier–North Holland. Graesser, A. C., McNamara, D. S., Louwerse, M. M.,
Burke, D. M., MacKay, D. G., Worthley, J. S., & Wade, & Chai, Z. (2004). Coh-Metrix: Analysis of text
E. (1991). On the tip of the tongue: What causes cohesion and language. Behavior Research
word finding failures in young and older adults. Methods, Instruments, and Computers, 36,
Journal of Memory and Language, 30, 542–579. 193–202.
Burke, D., Worthley, J., & Martin, J. (1988). I’ll never Grigsby, J., Kaye, K., Baxter, J., Shetterly, S., &
forget what’s-her-name: Aging and the tip of the Hamman, R. (1998). Executive cognitive abili-
tongue experience in everyday life. In M. M. ties and functional status among community-
Gruneberg, P. Morris, & R. N. Sykes (Eds.), dwelling older persons in the San Luis Valley
Practical aspects of memory: Current research Health and Aging Study. Journal of the
and theory (Vol. 2, pp. 113–118). Chichester, American Geriatrics Society, 46, 590–596.
UK: Wiley. Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G.,
Cahn-Weiner, D. A., Malloy, P. F., Boyle, P. A., & Curtis, G. (1993). Wisconsin Card Sorting Test
Marran, M., & Salloway, S. (2000). Prediction (WCST) manual, revised and expanded. Odessa,
of functional status from neuropsychological FL: Psychological Assessment Resources.
tests in community-dwelling elderly individu- Hoehn, M., & Yahr, M. (1967) Parkinsonism, onset,
als. The Clinical Neuropsychologist, 14, progression and mortality. Neurology, 17, 427–442.
187–195. Hofer, S. M., Flaherty, B. P., & Hoffman, L. A. (2007).
Carlson, M., Fried, L., Xue, Q., Bandeen-Roche, K., Cross-sectional analysis of time-dependent data:
Zeger, S., & Brandt, J. (1999). Association Mean-induced association in age-heterogeneous
between executive attention and physical func- samples and an alternative method based on
tioning performance in community-dwelling sequential narrow age-cohort samples.
older women. Journal of Gerontology: Social Multivariate Behavioral Research, 41, 165–187.
Sciences, 54B, 262–270. Hofer, S. M., & Sliwinski, M. J. (2001).
Farias, S., Harrell, E., Neumann, C., & Houtz, A. Understanding ageing: An evaluation of research
(2003). The relationship between neuropsycho- designs for assessing the interdependence of age-
logical performance and daily functioning in ing-related changes. Gerontology, 47, 341–352.
individuals with Alzheimer’s disease: Ecological Hosom, J. P., Shriberg, L., & Green, J. R. (2004).
validity of neuropsychological tests. Archives of Diagnostic assessment of childhood apraxia of
Clinical Neuropsychology, 18, 655–672. speech using automatic speech recognition
Ferber, R. (1991). Slip of the tongue or slip of the (ASR) methods. Journal of Medical Speech-
ear? On the perception and transcription of Language Pathology, 12, 167–171.
naturalistic slips of the tongue. Journal of Just, M. A., & Carpenter, P. A. (1992). A capacity
Psycholinguistic Research, 20, 105–122. theory of comprehension: Individual differences
Fisk, J. E., & Sharp, C. A. (2004). Age-related in working memory. Psychological Review, 99,
impairment in executive functioning: Updating, 122–149.
inhibition, shifting, and access. Journal of Kanne, S. M., Balota, D. A., & Storandt, M. (1998).
Clinical and Experimental Neuropsychology, 26, Relating anatomy to function in Alzheimer’s dis-
874–890. ease: Neuropsychological profiles predicting
Fodor, J. A. (1982). Modularity of mind. Cambridge: regional neuropathology 5 years later. Neurology,
MIT Press. 50, 979–985.
Folstein, M. F., Folstein, S. E., & McHugh, P. R. Kaplan, E. F. Goodglass, H., & Weintraub, S. (2001).
(1975). Mini-Mental State Examination. Lutz, The Boston Naming Test (2nd ed.). Philadelphia:
FL: Psychological Assessment Resources. Lippincott Williams, & Wilkins.
Fox Tree, J. E. (1995). The effects of false starts and Kemper, S., & Sumner, A. (2001). The structure of
repetitions on the processing of subsequent words verbal abilities in young and older adults.
in spontaneous speech. Journal of Memory and Psychology and Aging, 16, 312–322.
Language, 34, 709–738. Loewenstein, D. A., Ownby, R., Schram, L.,
Friedman, N. P., & Miyake, A. (2004). The relations Acevedo, A., Rubert, M., & Argueelles, T.
among inhibition and interference control functions: (2001). An evaluation of the NINCDA–ADRDA
11-Hofer-45484.qxd 2/26/2008 12:00 PM Page 192

192– • –DIMENSIONS OF COGNITIVE AGING

neuropsychological criteria for the assessment of mediator of age-related cognitive decline in

Alzheimer’s disease: A confirmatory factor analy- normal adults. Journal of Experimental Psychology:
sis of single versus multi-factor models. Journal General, 132, 566–594.
of Clinical and Experimental Neuropsychology, Shipley, W. C., & Zachary, R. A. (1940). Shipley
23, 274–284. Institute of Living Scale. Los Angeles: Western
Mayr, U. (2002). On the dissociation between cluster- Psychological Services.
ing and switching in verbal fluency: Comment Spreen, O., & Benton, A. L. (1977). Neurosensory
on Troyer, Moscovitch, Winocur, Alexander, and Center Comprehensive Examination for Aphasia.
Stuss. Neuropsychologia, 40, 562–566. Victoria, BC: Neuropsychology Laboratory,
Mayr, U., & Kliegl, R. (2000). Complex semantic University of Victoria.
processing in old age: Does it stay or does it go? Troster, A., Fields, J. A., Testa, J. A., Paul, R. H.,
Psychology and Aging, 15, 29–43. Blanko, C. R., Hames, K. A., et al. (1998).
Miyake, A., Emerson, M. J., & Friedman, N. P. (2000). Cortical and subcortical influences on clustering
Assessment of executive functions in clinical set- and switching in the performance of verbal flu-
tings: Problems and recommendations. Seminars ency tasks. Neuropsychologia, 36, 295–304.
in Speech and Language, 21, 169–183. Troyer, A. K., Moscovitch, M., & Winocur, G.
Miyake, A., Friedman, N. P., Emerson, M. J., Witzki, (1997). Clustering and switching as two compo-
A. H., Howerter, A., & Wager, T. D. (2000). The nents of verbal fluency: Evidence from younger
unity and diversity of executive functions and their and older healthy adults. Neuropsychology, 11,
contributions to complex “frontal lobe” tasks: A 138–146.
latent variable analysis. Cognitive Psychology, 41, Troyer, A. K., Moscovitch, M., Winocur, G.,
49–100. Alexander, M. P., & Suss, D. (1998). Clustering
Obler, L. K. (1980). Narrative discourse style in the and switching on verbal fluency: The effects of
elderly. In L. K. Obler & M. L. Albert (Eds.), focal, frontal- and temporal-lobe lesions.
Language and communication in the elderly Neuropsychologia, 36, 499–504.
(pp. 75–90). Lexington, MA: Heath. Troyer, A. K., Moscovitch, M., Winocur, G., Leach,
Rabbitt, P. (1997) Introduction: Methodologies and L., & Freedman, M. (1998). Clustering and
models in the study of executive function. In switching on verbal fluency tests in Alzheimer’s
P. Rabbitt (Ed.), Methodology of frontal and exec- and Parkinson’s disease. Journal of the
utive function (pp. 1–38). Hove, UK: Psychology International Neuropsychological Society, 4,
Press. 137–143.
Reitan, R. M., & Wolfson, D. (1995). Category Test Ulatowska, H. K., Cannito, M. P., Hayashi, M. M., &
and Trail Making Test as measures of frontal lobe Fleming, S. G. (1985). Language abilities in the
functions. Clinical Neuropsychologist, 9, 50–56. elderly. In H. K. Ulatowska (Ed.), The aging
Royall, D. R., Chiodo, L. K., & Polk, M. J. (2003). brain: Communication in the elderly (pp.
Executive dyscontrol in normal aging: 125–139). San Diego, CA: College-Hill.
Normative data, factor structure, and clinical Ullman, M. T., Corkin, S., Coppola, M., Hickok, G.,
correlates. Current Neurology and Neuroscience Growdon, J. H., Koroshetz, W. J., et al. (1997).
Reports, 3, 487–493. A neural dissociation within language: Evidence
Royall, D. R., Lauterbach, E. C., Cummings, J. L., that the mental dictionary is part of declarative
Reeve, A., Rummans, T. A., Kaufer, D. I., et al. memory, and that grammatical rules are
(2002). Executive control function: A review of processed by the procedural system. Journal of
its promise and challenges for clinical research. Cognitive Neuroscience, 9, 266–276.
Journal of Neuropsychiatry and Clinical Waters, G. S., & Caplan, D. (1996). The capacity
Neuroscience, 14, 377–405. theory of sentence comprehension: Critique of
Royall, D. R., Palmer, R., Chiodo, L., & Polk, M. Just and Carpenter (1992). Psychological
(2004). Declining executive control in normal Review, 103, 761–772.
aging predicts change in functional status: The Zelinski, E. M., & Lewis, K. L. (2003). Adult age dif-
Freedom House study. Journal of the American ferences in multiple cognitive functions:
Geriatrics Society, 52, 346–352. Differentiation, dedifferentiation, or process-
Salthouse, T. A., Atkinson, T. M., & Berish, D. E. specific change? Psychology and Aging, 18,
(2003). Executive functioning as a potential 727–745.