A MODEL OF EXECUTIVE FUNCTIONS IN VERY OLD COMMUNITY

DWELLERS: EVIDENCE FROM THE SYDNEY OLDER PERSONS STUDY

Olivier Piguet1,2, David A. Grayson3, Robyn L. Tate4, Hayley P. Bennett1,2, Tanya C. Lye1, Helen Creasey1,
William S. Brooks1,2 and G. Anthony Broe2,5
(1Centre for Education and Research on Ageing of the University of Sydney at Concord Repatriation General Hospital,
Sydney, NSW; 2Prince of Wales Medical Research Institute and the University of New South Wales, Sydney, NSW;
3School of Psychology, the University of Sydney, Sydney, NSW; 4Rehabilitation Studies Unit, Department of Medicine,
the University of Sydney at Royal Rehabilitation Centre Sydney, Sydney, NSW; 5Prince of Wales Hospital, Randwick,
NSW, Australia)

ABSTRACT

Executive functions (EF) are generally described as showing greater sensitivity to ageing compared to other cognitive
domains. Numerous pitfalls exist in the measurement of EF due to loose definitions and lack of agreement on these concepts
and uncertainty about the constructs being measured. To this date, the validity of EF constructs has not been examined in
the old-old population. Performance of 122 randomly selected community dwellers aged between 81 and 97 years on nine
EF tasks (seven of which commonly used in clinical practice) was examined. Factor analytic procedures using structural
equation modelling (SEM) failed to satisfactorily explain the data according to four a priori models, the first two models
reflecting two major constructs commonly found in current models of EF (“set” and “switch”), the last two reflecting task
requirements. The best measure for each task was extracted using statistically driven analyses and further SEM revealed an
orthogonal two-factor model which provided a good fit of the data, explaining between 8% and 25% of the total variance.
This model can be interpreted in terms of reactive and spontaneous flexibility as proposed by Eslinger and Grattan (1993),
with the first factor reflecting internally driven strategies and the second environment dependent strategies. Furthermore,
these findings also suggest that: (a) unique tasks of EF may not be applicable to all age groups due to individual experience
and changes in strategies; and (b) current clinical instruments may be inadequate to measure very specific aspects of the
complex construct of EF.

Key words: executive function, ageing, measurement, structural equation modelling

INTRODUCTION neurologically intact individuals (Crawford et al.,

2000; Rabbitt and Lowe, 2000; Wecker et al.,
Studies on executive functions (EF) in ageing 2000).
commonly report a poorer EF performance with Knowledge about EF is hampered by an
advancing age compared to younger age groups. absence of clearly defined abilities or processes
These changes include decline in concept formation encompassed by this construct (e.g., “judgment”,
and abstraction, decrease in mental flexibility, “will”, “self-awareness”, “creativity”, “error
increased time required to shift mental sets, to utilization”, “inhibition” or “planning”) and by
adapt to novel situations or to solve new or poorly defined measurement criteria. Despite this
unusual problems (Albert et al., 1990; Daigneault apparent lack of cohesion, EF are generally viewed
et al., 1992; Daigneault and Braun, 1993; Levine et as the overarching mechanisms responsible for the
al., 1995; Mittenberg et al., 1989; Parkin and control of cognition (Phillips, 1997). They are very
Lawrence, 1994; Robbins et al., 1998; Shimamura important in aspects of behaviour that require (1)
and Jurica, 1994; Stuss et al., 1996; West, 1996). setting up plans of actions (that may be composed
Deficits in EF seem to increase rapidly in the of subtasks) for the achievement of a pre-
eighth decade of life and to show greater age- determined goal; (2) the initiation and monitoring
related decline compared to other cognitive of the plans; and (3) the ability to shift and adapt
functions such as language, memory and perception plans in response to external or internal pressures
(Mittenberg et al., 1989; Tranel et al., 1994). These (Luria, 1973). Executive functions are expressed in
changes have been explained by a more rapid activities as varied as problem solving (Lezak,
ageing in structures thought to predominantly 1995), planning (Tranel et al., 1994), cognitive
mediate EF, namely the frontal lobes, compared to estimation (Shallice, 1988), prospective memory
other brain structures (Raz et al., 1997). However, (Burgess, 1997; Shimamura, 1995), and when
this “frontal aging hypothesis” has been challenged mental representation of tasks and goals are
(Greenwood, 2000). Recent studies focusing on required (Pennington et al., 1996). They are also
normal cognitive ageing showed evidence of present in situations requiring concept formation,
decline related to age on some (e.g., Stroop test, reasoning, cognitive flexibility (“spontaneous” and
Tower of London, Modified card sorting test) but “reactive” flexibility) (Grattan and Eslinger, 1991)
not other (e.g., verbal fluency, cognitive estimates, inhibition, working memory or resource allocation.
use of common objects) measures of EF in Most clinical EF tasks generally aim to test one of

Cortex, (2005) 41, 27-37

28 Olivier Piguet and Others

the following aspects of cognition: plan setting, of skewed score distributions. In another study,
shifting and monitoring, or manipulation of abstract Della Sala et al. (1998) assessed 48 adult frontal
concepts. These tasks tend to be: (a) novel, so as to lobe patients on five “traditional” frontal tests and
invoke higher levels of cognitive activation such as five non-frontal tests. Their results revealed greater
novel planning and goal setting; (b) of a certain positive correlations within the frontal tasks than
duration, to tap into the monitoring aspect of between the frontal and non-frontal tasks. Factor
behaviour and on-line manipulation of information; analyses conducted on the five frontal tests, and on
and (c) effortful, in order to engage cognitive the complete set only yielded single factor
power fully and assess the levels of monitoring and solutions indicating that the two groups of tasks
inhibition processes. Complexity is not necessary, could not be differentiated on statistical grounds.
being a relative construct which may be handled by The authors interpreted their results to indicate a
other cognitive systems (Stuss and Alexander, necessary frontal involvement in all tests and that
2000). Some degree of unpleasantness may be the difference was one of degree and not of kind.
present as EF tasks are designed to be effortful and To this date, factor analytic studies of EF tasks
challenging (Denckla, 1996; Phillips, 1997). in elderly people are lacking. In this age group, in
Unlike other cognitive abilities, EF are addition to brain structural changes that may (West,
independent of direct external or internal sensory 1996) or may not (Greenwood, 2000) affect the
inputs. Because of their primary role as coordinator frontal lobes more predominantly, affecting EF
of purposeful actions, the correspondence between performance, other factors need to be taken into
the observed behaviours and the underlying account. First, both inter- and intra-individual
cognitive processes is uncertain (Burgess, 1997). cognitive variability increase with age (Christensen
Because of this reliance on other cognitive systems et al., 1999), possibly reflecting such structural
for their expression, pure measures of EF probably changes and expressed via the use of different
do not exist. As such, descriptors of task demands approaches in dealing with cognitive demands.
(e.g., “concept shifting” or “inhibition”) may be Second, the presence of diffuse neurodegenerative
useful labels but do not necessarily imply disorders associated with age, such as dementia
corresponding different or discrete underlying further compounds this cognitive heterogeneity.
cognitive processes. Furthermore, there is evidence These disorders are accompanied by cognitive
that EF tasks that appear, on the surface, to require changes in systems not necessarily directly
similar cognitive demands do not necessarily result implicated in the resolution of EF type problems
in a consistent performance (Lowe and Rabbitt, but relied upon for their appropriate execution
1997). Conversely, a disruption in an underlying (Collette et al., 1999; Goldberg and Bilder, Jr.,
cognitive process may be exhibited in various 1987; Lafleche and Albert, 1995).
fashions (Kimberg and Farah, 1993). In addition, Thus, the evidence presented above indicates
EF tasks tap processes incidental to their main that EF measurement varies, becoming increasingly
purpose and the likelihood of measurement error is complex with advancing age. In clinics involving
therefore greater than for tasks measuring other elderly patients, who not uncommonly present with
cognitive functions (Pennington et al., 1996). As various simultaneously occurring pathological
Della Sala and colleagues (1998) indicated, EF are processes, such changes need be considered and
necessary but not sufficient for the adequate their possible impact on EF task performance taken
performance of EF tasks. Despite the existing into account. Further, research of this complex and
confusion and lack of agreement in the literature ever evolving construct and its possible
about these terms, clinicians have embraced this fractionation is necessary so that the cognitive
nomenclature with enthusiasm, probably because of abilities supporting somewhat intangible constructs
what appears to be their obvious face validity. can be better identified. In an attempt to address
Few studies have investigated the empirical these issues, the aim of this study was to identify
evidence and validity of the constructs underlying the structure underlying the pattern of performance
EF tasks (Boone et al., 1998; Della Sala et al., on a range of EF tasks in a group of community
1998; Kafer and Hunter, 1997). Kafer and Hunter dwellers aged 81 years and over.
(1997) examined four tests purported to measure
planning and problem solving in a sample of 130
neurologically intact adults aged 17 to 55 years. METHODS
Structural equation modelling failed to validate any
of three hypothesised models and was further Participants
unable to extract an adequate model that would
account for the data. The authors concluded that Participants were recruited from the 6-year
their findings questioned the existence of the review of the Sydney Older Persons Study (SOPS)
“planning” and “problem solving” constructs (Creasey et al., 2001; Waite et al., 2001). SOPS is
underlying the tasks used. They postulated that the a longitudinal study following a randomly selected
tests used were possibly not sensitive measures of representative sample of 630 community dwellers
these constructs in the general population because from the inner west of Sydney, Australia, aged 75
Executive functions in ageing 29

TABLE I
Demographic Characteristics for the Study Sample (N = 122)

Variable N %

Gender (M/F) 64/58 52/48

Mean SD Range
Age 85.5 3.1 81-97
Education (Years) 10.4 2.1 6-19
Estimated full scale IQ 108.7 10.0 83-128
MMSE total score 26.7 2.6 18-30
Boston naming test; total score 46.7 7.4 28-60
WMS-R Logical memory; % forgetting 33.8 26.5 – 15-100
Judgment of line orientation 16.8 2.5 8-20
Note. MMSE = Mini Mental State Examination; WMS-R = Wechsler Memory Scale-Revised.

years and over at the time of recruitment in 1991- emphasis of the study was on a community sample,
1993. All the SOPS participants capable of giving these participants were not excluded in order to
informed consent independently, or with the maintain a more accurate reflection of the cognitive
assistance of a relative if necessary, were included functioning of this age group.
in the present study. Participants were excluded
when they were deemed too cognitively or Procedure and Materials
physically impaired, to prevent potential distress
from high level cognitive testing or MRI Testing occurred in two sessions one month
procedures included in the study protocol (not apart on average, the first one at the participant’s
presented here). In order to maintain the broadest home and the second at the local university
possible panorama of cognition in this age teaching hospital. Tests of EF were administered as
segment, no other a priori medical exclusion part of a larger battery of cognitive tests, the order
criteria were applied. This study received ethics remaining constant across individuals. The aim was
approval from the Central Sydney Area Health to select tasks that were commonly used in clinical
Service. practice and would tap into different theoretical
Of the 299 remaining SOPS participants constructs of EF. Further, with our elderly sample
potentially available, 167 agreed to participate in in mind, preference was given for EF tasks that
this study (56%). One participant died before were of short duration and that would minimise the
testing could be completed and one participant with possible impact of sensory deficits common in this
a Mini Mental State Examination (MMSE; Folstein age group. For these reasons, popular tests such as
et al., 1975) score of 13 points was excluded; 14 the Wisconsin Card Sorting Test (Heaton et al.,
participants changed their mind and an additional 7 1993) or the Stroop test (Golden, 1978) were
subsequently declined because of poor health. By omitted, being deemed too complex, too long or
the time of the study completion, another 6 susceptible to peripheral sensory deficits. The
participants were excluded because of continuing following tests were administered: (1) WAIS-R
poor health. Three participants lived too far from Similarities subtest (Wechsler, 1981); (2) New
the testing centre and could not be seen. Finally, tower of London (New TOL) (Hanes, 1997); (3)
because of the MRI strong magnetic fields, 8 Semantic fluency (animal category) (Spreen and
participants with a cardiac pacemaker had to be Strauss, 1991); (4) Phonemic fluency (FAS)
excluded and another 5 declined to participate as (Benton and Hamsher, 1983); (5) Ruff figural
they were claustrophobic. Thus, 122 individuals fluency task (RFFT) (Ruff, 1988); (6) Oral trail
(64 men and 58 women) took part in the current making test-part B (Oral Trails) (Kaye et al.,
study. The sample’s demographic characteristics are 1990); and (7) the WAIS-R-Digit span backward
presented in Table I as well as descriptive statistics subtest (DSB) (Wechsler, 1981).
for some measures of the integrity of the memory Portions of two computerised tests measuring
(Wechsler Memory Scale-Revised Logical Memory working memory integrity were also included in
subtest; Wechsler, 1987), visuo-perceptual (subset the battery: the N-Back Colour 1- and 2-Back
of Judgment of Line Orientation; Benton et al., (Cohen et al., 1993) and the California
1983) and language systems (Boston Naming Test; Computerized Assessment Package (CalCAP)
Kaplan et al., 1983). These findings indicate that abbreviated version (Miller, 1996). For the N-Back,
the sample’s cognitive test performance (in terms the aim is to respond as quickly as possible
of posterior brain functions) is comparable to other whenever a coloured circle briefly presented on a
samples of similar age and education (Ivnik et al., computer screen is identical to the stimulus
1996; Neils et al., 1995). Examination of the range presented immediately before (in the 1-Back
of MMSE scores reveals 3 participants with a score condition) or to the one presented before the last
of below 21 points (18, 19 and 19 points). As the one (in the 2-Back condition). Stimuli were
30 Olivier Piguet and Others

TABLE II
Description of the Measures of Executive Functions and Four A Priori Groupings

Task Measure Set Switch Mode

1) WAIS-R Similarities Total words Y N non-C
2) New tower of London Highest level achieved Y Y non-C
Planning time item 3 Y N non-C
Execution time item 3 N Y non-C
3) Semantic fluency Switching score N Y non-C
Clustering score Y N non-C
4) Phonemic fluency Switching score N Y non-C
Clustering score Y N non-C
5) Ruff figural fluency test Total designs Y Y non-C
Total repetitions N Y non-C
6) Oral trails-part B Total time Y Y non-C
Efficiency score N Y non-C
7) WAIS-R Digit span backward Longest span Y Y non-C
8) N-Back
Colour 1-Back Reaction time Y N C
Efficiency N Y C
Colour 2-Back Reaction time Y N C
Efficiency N Y C
9) CalCAP
Simple reaction time Reaction time Y N C
Complex reaction time Reaction time Y N C
Efficiency N Y C
Serial pattern matching 1 Reaction time Y N C
Efficiency N Y C
Serial pattern matching 2 Reaction time Y N C
Efficiency N Y C
Note. Y = belongs to model; N = does not belong to model (overlap is possible between set and switch models); C = computerised administration; non-C =
non-computerised administration. Tasks: CalCAP = California computerized assessment package; WAIS-R = Wechsler Adult Intelligence Scale-Revised.

presented in a pseudo random order on the centre the centre of a 12-inch colour screen. The
of a 12-inch colour screen. There were 38 different instructions were very similar for the four tasks.
items comprising 37 possible pairs including 8 Participants were asked to respond as quickly as
target pairs (9 pairs for the Colour 2-Back task). possible to certain types of configurations by
Stimulus presentation was 1000 msec with an inter pressing the space bar on the keyboard and were
stimulus interval (ISI) of 500 msec. Sitting at a asked to ignore all the other (irrelevant) stimuli.
distance of approximately 60-70 cm from the Each task was preceded by a maximum of 4
screen, participants were instructed to respond to practice trials. The mean RT for the correct
the targets as quickly as possible by pressing with responses was recorded for all the tasks. An
their dominant hand on a keypad connected to the efficiency measure accounting for the number of
external port. For each task, the reaction time was false positive responses in the total number of
recorded as well as an efficiency rate which was a responses was also recorded except for the single
composite measure of accuracy and false positives. RT task where this measure was not applicable.
The abbreviated CalCAP comprised four tasks The tests and measures collected are presented in
varying in complexity. First was a simple reaction Table II.
time (RT) task which required participants to
respond to any stimulus as quickly as possible. Planned Statistical Analyses
Fifteen stimuli were presented with a random ISI
ranging from 1000 to 5000 msec. The second task, In order to examine the pattern of relations
a choice RT task, required a response only to the within the EF variables, four a priori models using
digit “7”. This task was composed of 100 trials different groupings of measures were tested with
with 15 target stimuli presented in a pseudo structural equation procedures. These groupings
random order. Stimulus duration was 100 msec were: (1) “set” which included measures assumed to
with an ISI of 800 msec. The third task was the tap into cognitive abilities such as strategies,
“Serial Pattern Matching 1”, a task requiring a planning, problem solving and reasoning skills; (2)
response to the presentation of two identical digits “switch” which included measures assumed to tap
in sequence (i.e. one immediately after the other). into aspects of EF such as flexibility, self-
The procedure was identical to the previous task monitoring and set shifting; (3) “mode” which was
except that it contained 20 target stimuli. The final taking into account test administration
task “Serial Pattern Matching 2” required (computerised or not); and (4) “task” which
participants to respond when they saw two digits in examined grouping of measures by tasks. Allocation
sequence and in increasing order (e.g. “2” followed of measures to “set” and “switch” was based on task
by “3”). Number of trials, number of target requirements, face validity and clinical judgment.
stimulus were identical to the previous task. The For each model, maximum likelihood parameter
CalCAP stimuli (single digits) were presented on estimates were obtained using AMOS 3.6.1
Executive functions in ageing 31

TABLE III
Descriptive Statistics for the EF Measures

Task Measure Mean SD Range # of MVs

WAIS-R Similarities: Total score 12.0 7.0 0-27 0
New tower of London: Highest level achieved 4.2 1.6 1-9 7
Planning time item 3 8.2 7.0 1-35 7
Execution time item 3 22.5 24.0 3-174 7
Semantic fluency: Switching score 7.7 3.1 1-18 1
Clustering score 0.71 0.56 0-3.67 1
Total score 13.2 4.3 3-31 1
Phonemic fluency: Switching score 25.8 11.0 0-52 1
Clustering score 0.12 0.13 0-0.55 1
Total score 30.0 12.5 2-62 1
Ruff figural fluency test: Total designs 37.6 13.5 12-77 11
Total repetitions 8.1 9.0 0-66 11
Oral trails-part B: Total time 56.7 26.6 15-108 4
Efficiency score 87.0 20.0 0-100 4
WAIS-R Digit span backward: Longest span 4.1 0.9 2-7 26
Colour 1-Back: Reaction time 0.76 0.23 0.40-1.49 18
Efficiency 63.40 37.70 – 87-100 18
Colour 2-Back: Reaction time 1.05 0.36 0.43-3.03 18
Efficiency 11.00 39.82 – 187-87 18
CalCAP simple RT: Reaction time 525.49 264.54 289-2447 10
CalCAP complex RT: Reaction time 554.40 81.37 413-831 12
Efficiency 0.86 0.15 0.20-1 12
CalCAP serial pattern matching 1: Reaction time 567.69 79.06 388-855 17
Efficiency 0.56 0.23 – 0.05-1 17
CalCAP serial pattern matching 2: Reaction time 584.99 95.01 236-866 27
Efficiency 0.42 0.18 0.04-0.80 27
Note. CalCAP = California computerized assessment package; WAIS-R = Wechsler Adult Intelligence Scale-Revised; MV = missing value.

(Arbuckle, 1997). Currently, in the domain of model fitting “set” was examined first. This was
structural equation modelling, a large number of followed by a second model adding “switch” as a
indices are available to test the goodness of fit of a second factor before the third model, adding
model. (See Byrne, 2001 for a good introduction “mode”, was examined. Finally, a “task” model
text on these increasingly popular methods). In this was examined independently. These models did not
study, goodness of fit was established using the include the total number of words for the two
goodness of fit index (GFI), the adjusted GFI fluency tasks as they were deemed to reflect a
(Arbuckle, 1997) and the root mean square error composite measure of the number and size of word
approximation (RMSEA) (Browne and Cudeck, clusters already represented in the data set
1992). A GFI in excess of .90 indicates a good fit of (switching and clustering scores respectively).
the model onto the data (Marsh et al., 1988); an None of the a priori models was able to
adjusted GFI will always be lower than the GFI, provide an acceptable fit of the interrelations
imposing a penalty related to the number of among the set of variables (see Table IV). The
parameters being estimated, and values in excess of initial single factor (“set”) model gave rise to a
.90 indicate very good fits; whilst a RMSEA value GFI of .552. The GFI improved only marginally
of less than 0.04 is indicative of a close fit of the after the introduction of the “switch” and “mode”
model (Browne and Cudeck, 1992). The models factor (.686 and .722 respectively). The “task”
with their respective groupings are presented in model, positing 9 factors underlying the nine tests
Table II. The data, correlations and SDs, used for was unable to converge and gave rise to an
these analyses are presented in the Appendix, with inadmissible solution because of the presence of
missing values (MVs) imputed. For all the EF negative variance or correlations larger than 1.00
measures, existing missing values were inferred among some variables.
from the scores from the remaining variables within Given the unsuccessful modelling of measures
the set using a multiple regression procedure. In according to theoretical neuropsychological
other words, the obtained score, in replacement of constructs, the components from each test were
the MV, was a composite score reflecting the pattern examined individually. It appeared that certain
of associations among predictors (i.e., the other measures exhibited partial logical constraints. For
variables) with the predictor having the MV. example, the switching and mean cluster size scores
from the Semantic fluency task had a negative
correlation (r = – .50) and were “competing” in
RESULTS that high scores on one component could only be
achieved at the expense of the other. The same
The descriptive statistics for the EF measures phenomenon was observed to a lesser degree for
are presented in Table III. Within the structural the RFFT between the number of repetitions and
equation modelling procedures, a single factor the number of unique designs (r = – .14).
32 Olivier Piguet and Others

TABLE IV
Results of the Confirmatory Factor Analyses for Four A Priori Models

Model Admissible χ2 df GFI AGFI RMSEA

“Set” Yes 1093.494 262 .552 .487 0.162
“Set” & “Switch” Yes 711.228 247 .686 .619 0.125
“Set”, “Switch” & “Mode” Yes 595.997 234 .722 .643 0.113
“Task” No 610.154 222 .725 .628 0.120
Note. GFI = goodness of fit; AGFI = adjusted goodness of fit; RMSEA = root mean square error of approximation.

Given these constraints, a statistically driven the two verbal fluency measures) and a correlated
analysis of the EF measures was conducted to uniqueness parameter between these residuals
clarify the underlying structure within each test (yielding r = .42) was added to the model. This
with the goal of extracting one meaningful measure improved the fit of the two-factor model
per task. For each task, this was achieved by first substantially, giving rise to a GFI value of .970, an
identifying and removing existing constraints. adjusted GFI of .925 and a RMSEA value of 0.00
Unnecessary measures, as determined by the results (χ2 = 16.553; df = 18; p = .554)1. This model was
of the analyses performed on each task were also then “reified” (i.e., an oblique factor rotation
removed. As such, to resolve the competition chosen) by defining a 2-factor model with the first
between the switching and clustering measures factor loading exclusively on N-Back and not on
from the two verbal fluency tasks, these were DSB and, conversely, the second factor loading
replaced with their respective global measure (i.e., exclusively on DSB but not on N-Back.
total number of words). Single indicators for the Constraining the latent Factor1-2 correlation from
WAIS-R DSB and Similarities subtests were this model (r = – .221; Critical Ratio = – 1.126) to
retained. For the seven CalCAP measures, no 0 resulted in no loss of information: GFI = .968,
logical constraints, such as the ones described adjusted GFI = .923 and RMSEA = 0.00 (χ2 =
above, were present. Maximum likelihood factor 17.854; df = 19; p = .532). This indicated that the
analytic investigations within this set of measures correlation between the two factors indicated by N-
revealed two factors with an eigenvalue greater Back and DSB tasks was well within chance
than 1. Confirmatory factor analysis of this two- bounds (i.e., not significantly different from 0) and
factor solution using AMOS 3.6.1 (Arbuckle, 1997) strongly supported a two-orthogonal-factor model
produced a GFI of .969 (χ2 = 14.837; df = 8) but as acceptable for the nine tasks. Finally, the
yielded a correlation of .968 between the two loadings from FAS and Similarities to Factor-1 (the
factors, these being in essence indistinguishable. factor indicated by N-Back) were not significant, so
The CalCAP measures were therefore summarised a final model was run with these set to zero: GFI =
into one composite score composed of their z-score .966, adjusted GFI = .926 and RMSEA = 0.00 (χ2 =
weighted total. A similar procedure was applied to 19.260; df = 21; p = .568). The estimates from this
the N-Back measures. As two factors could not be final model are shown in Table V.
fitted to four indicators, these measures were Inspection of the relative contribution of each
summarised with their average z-score weighted factor to the explained variance for each EF
total into one composite score. Following the same measure allowed interpretation in terms of three
procedure, a composite score was constructed from distinct groups: (a) a group of measures showing
the time and performance measures from the Oral loadings on Factor 2 in excess of 85% (DSB,
Trails task as they seemed to tap a common Similarities, FAS and RFFT); (b) a group of
construct conceptually. For the New TOL test, the measures with approximately equal relative
highest level of complexity successfully completed contributions from Factor 1 and 2 to the explained
was retained. Finally for the RFFT, a composite variance (Semantic fluency, Oral Trails, CalCAP
measure was constructed taking into account the and New TOL); and finally (3) N-Back, loading
number of errors in the total performance. exclusively on Factor 1. The intercorrelations
The factorial structure of this set of nine among the nine EF composite measures are
variables was investigated with AMOS 3.6.1 presented in Table VI with the tasks grouped
(Arbuckle, 1997). An initial exploratory model according to their respective loadings. The overall
fitted with one general factor gave rise to a GFI of average correlation was .35, however, within and
.905, an adjusted GFI of .842 and a RMSEA value between group average correlations were more
of 0.094 (χ2 = 55.89; df = 27; p = .001). The varied with a positive correlation coefficient of
introduction of a second factor saw the exploratory .446 among the measures loading primarily on
model improve significantly with a GFI of .943, an
adjusted GFI of .864 and a RMSEA of 0.078 (χ2 =
33.107; df = 19; p = .023); increment χ2 = 22.783 1This over association between these two measures cannot be ignored. The
introduction of the correlated uniqueness parameter can be seen as a third
on 8 df; p = .004. The highest modification index minor common factor for these two tasks only, whilst remaining
between residuals was ascertained (12.447 between uncorrelated with the other two factors.
Executive functions in ageing 33

TABLE V
Final Two-factor Model for the Nine EF Measures

Variance Ratio Relative contribution to

Factor 1 Factor 2
explained (%) (F1/F2) explained variance
Factor 1 Factor 2
Colour N-Back 0.499 0 24 Infinite 1.00 0.00
DSB 0 0.614 23 0 0.00 1.00
Similarities 0 4.519 18 0 0.00 1.00
FAS 0 7.892 16 0 0.00 1.00
RFFT 3.342 10.048 25 0.33 0.10 0.90
Semantic fluency 1.098 2.226 12 0.49 0.20 0.80
Oral Trails 0.384 0.488 23 0.79 0.38 0.62
CalCAP 0.275 0.359 13 0.77 0.37 0.63
New TOL 0.654 0.484 8 1.35 0.65 0.35
Note. CalCAP = California computerized assessment package; DSB = Digit span backward; FAS = Phonemic fluency; New TOL = New tower of London;
RFFT = Ruff figural fluency test.

TABLE VI
Correlations (× 100) and SDs for the Nine EF Measures Grouped according to their Respective Factor Loadings

Loading Factor 1 Factor 2 Mixed

Measure SDs 1 2 3 4 5 6 7 8 9
Factor 1 1. Colour N-Back 0.71 100
Factor 2 2. DSB 0.88 10 100
3. Similarities 6.94 9 43 100
4. FAS 12.41 12 42 43 100
5. RFFT 15.13 26 55 43 43 100
Mixed 6. Semantic fluency 4.29 24 33 35 61 48 100
7. Oral Trails 0.92 36 40 41 41 40 42 100
8. CalCAP 0.77 31 29 38 36 40 35 47 100
9. New TOL 1.54 31 29 15 24 35 35 39 29 100
Note. All correlations larger than .15 significant at the .005 level.
CalCAP = California computerized assessment package; DSB = Digit span backward; FAS = Phonemic fluency; New TOL = New tower of London; RFFT =
Ruff figural fluency test.

Factor 2 (DSB, Similarities, FAS and RFFT) and and shifting of plans of actions failed to explain
.379 among those loading equally on both factors the data adequately. The presence of logical
(i.e. “Mixed”; Semantic fluency, Oral Trails, constraints between measures of the same task
CalCAP and New TOL). Among the groups, the (yielding negative correlations as “method” effects)
correlation between N-Back (i.e., Factor 1 such as switching and clustering scores in Semantic
measure) and the Factor 2 measures was much fluency made it impossible to examine the exact
lower (average r = .141) whilst the correlations role of some components and may have
between the “Mixed” group and Factor 1 and 2 compounded these findings. Our statistically driven
groups were .307 and .366 respectively. approach, devoid of any a priori theory, was an
attempt to clarify the existing underlying constructs
within each test, to reduce the existing dependency
DISCUSSION among some measures and to shed light on the
relations among these tasks. From these
Bearing in mind the smallness of our sample investigations, the associations among these
and the large number of EF tests examined, in this measures can be explained by two orthogonal
sample composed of elderly community dwellers, factors with three distinct groups emerging from
none of the postulated EF models was found to these analyses. Factor 1 was defined such that one
yield even approximate fit of the covariance data. measure, N-Back, loads exclusively on it. Factor 2
Somewhat surprisingly, the EF measures did not fit was defined such that DSB loaded exclusively on
the groupings according to existing models of EF it, and we found that three other measures
cognitive constructs thought to reflect the more essentially load only on this factor as well.
basic (core) cognitive requirements underlying Surprisingly, these two factors are in essence
tasks of EF. Thus, the “set” grouping supposedly uncorrelated (this feature of the data can directly
requiring means-goal analysis, plan of actions, be seen from the near zero correlations between N-
strategy application, activation of past experience Back and the four Factor-2 indicator tests in Table
and appropriate activation of schemas, plans or 6). Finally, the variance on the last four measures
scripts; or the “switch” grouping supposedly is explained by contributions from both Factors 1
requiring on-line maintenance of activation, and 2. The variance explained in these measures by
discrepancy analysis between “now and end goal”, the two common factors was low (8%-25%),
34 Olivier Piguet and Others

indicating that the tests probably are somewhat results suggest that, in a very old sample, these two
idiosyncratic to a major extent. tasks require quite different processes for
One interpretation emerging from the results of appropriate performance. Within a working
the factor analytic investigations fit Eslinger and memory and age difference framework, it is
Grattan’s (1993) theoretical framework of EF. The possible that N-Back and DSB load on a common
measures that predominantly load on Factor 2 factor but not when other aspects of EF are
(FAS, RFFT, Similarities and DSB) are all considered.
characterised by commonly encountered stimuli in Several methodological aspects of this study
the context of unusual demands and manipulations. deserve comment. First, the total variance explained
An optimal performance is reliant on the generation by the two factors remains small on all EF measures
of a diversity of responses and creative approaches (between 8 and 25%). This indicates the presence of
that are all internally driven. Such tasks require uncorrelated specific factors, as well as noise,
divergent thinking or, in the words of Eslinger and associated with the task and suggests that this group
Grattan (1993): spontaneous flexibility. In contrast, of tasks is not strongly related. Nevertheless, within
N-Back, which is loading exclusively on Factor 1, the shared portion of the variance, the processes can
is a task that is unusual from the outset and requires be explained along the lines of two underlying
unusual responses. In this instance, self-generated factors reflecting internal/external or
strategies cannot be applied for appropriate divergent/convergent thinking processes. Second, it
performance and correct responses are only may be argued that these findings may not be
achieved in response to stimuli triggered by the applicable to other tests of EF such as the WCST or
environment. This task requires participants to shift the Stroop but this needs to be determined.
response set in relation to external cues and relies Nevertheless, the tasks were selected on the basis of
on convergent thinking or reactive flexibility (i.e., sound criteria: to be commonly used in clinical
the “readiness to freely shift cognition and practice; to be relevant in the context of the
behaviour according to the particular demands and assessment of very old participants; and to map onto
context of a situation”; Eslinger and Grattan, 1993, previously described underlying cognitive
p. 18). Whilst a verbal, internally driven, strategy constructs within the EF domain. Thirdly, the
may be used, the task requires a constant update of contribution of certain components, such as
information within working memory which can switching and clustering, could not be examined
only be achieved successfully by responding to the fully. Although both may be associated with EF,
environment. The tasks that belong to the “Mixed” because of task requirements, these cannot be
group (i.e., loading on both factors) necessitate measured simultaneously. When they arise from the
concurrent activation of on-line maintenance of same task, there is a risk of obtaining findings that
information (spontaneous flexibility) within a given are influenced not by underlying EF constructs but
framework of activation (reactive flexibility). For by task constraints thus distorting the model. Had
example, Semantic fluency requires the production the “switching” and “clustering” scores been
of words within a single semantic category and is collected on separate occasions, it is predicted that
therefore subject to clear external boundaries. the negative correlation between the two measures
Spontaneous flexibility remains however a would have been attenuated or would have even
necessary component reflected in the choice of become positive. Finally, the discrepancy between
strategies (e.g., clusters and switches among these findings and other studies may be related to
subcategories of animals). In this group, the the nature of our sample. This community sample
difference in factor loadings would reflect the comprises randomly selected “old-old” participants,
strength of the existing framework (e.g., less for some of whom may present with some level of
Semantic fluency and greater for New TOL). cognitive impairment. Whilst not homogeneous in
Alternative explanations to the data set were that respect (i.e., not composed of completely
unsatisfactory. These findings were not explained healthy subjects only), this sample more accurately
by task requirements, our results showing no reflects the cognitive functioning of this age group,
specific groupings for the measures by modality with its increasingly common within and between-
(such as language versus non-language based subject variability in test performance (Christensen
processes, motor or visuospatial abilities). et al., 1999). It is also more similar to the type of
Similarly, decline in speed of processing, which patients clinicians are likely to encounter in their
has been put forward as the primary underlying practice.
process accounting for changes in EF performance Beyond the uncertainty regarding the face
with ageing (Salthouse, 1994; Salthouse and validity of some of the existing tasks of EF (Burgess,
Babcock, 1991), did not contribute to the various 1997; Levine et al., 1998; Phillips, 1997; Rabbitt,
groupings of measures. Our findings are in contrast 1997), our findings underscore the difficulty in
with other studies. Kensinger et al. (2000) and investigating the measurement of EF, particularly in
Shimamura (2000) describe DSB and N-Back as the oldest segment of the population. Test
measuring the same cognitive construct, namely performance data on this age group remain scarce.
updating and manipulation of information. Our Perhaps the specific cognitive requirements for the
Executive functions in ageing 35

successful completion of tasks of EF vary depending CHRISTENSEN H, MACKINNON AJ, KORTEN AE, JORM AF,
HENDERSON AS and JACOMB P. Dispersion in cognitive ability
on the cognitive development of the participants. as a function of age: A longitudinal study of an elderly
Denckla (1996) has suggested that some EF tasks tap community sample. Aging, Neuropsychology and Cognition,
into different cognitive constructs in children and in 6: 214-228, 1999.
COHEN JD, FORMAN SD, BRAVER TS, CASEY BJ, SERVAN-
adults. It may be fallacious to entertain the thought SCHREIBER D and NOLL DC. Activation of the prefrontal
of a unitary EF task, applicable to all age groups, cortex in a nonspatial working memory task with functional
given that a novel task at a certain age may not be so MRI. Human Brain Mapping, 1: 293-304, 1993.
COLLETTE F, VAN DER LINDEN M and SALMON E. Executive
at a later stage of development due to continuing dysfunction in Alzheimer’s disease. Cortex, 35: 57-72, 1999.
exposure to similar situations. Consequently, it is CRAWFORD JR, BRYAN J, LUSZCZ MA, OBONSAWIN MC and
possible that how very old participants approach STEWART L. The executive decline hypothesis of cognitive
aging: Do executive deficits qualify as differential deficits and
tasks of EF differ from younger adult participants. do they mediate age-related memory decline? Aging,
At this point in time, as suggested by Kafer and Neuropsychology and Cognition, 7: 9-31, 2000.
CREASEY H, WAITE LM, GRAYSON DA, BENNETT HP, DENT O and
Hunter (1997), available instruments may not allow BROE GA. The impact of the neurodegenerative disorders on
clinicians to measure the many hypothesised facets ageing: An overview of the Sydney Older Persons Study.
of EF. Alternatively, it is also possible that measures Australasian Journal on Ageing, 20: 11-16, 2001.
DAIGNEAULT S and BRAUN CMJ. Working memory and the Self-
of EF are age specific and that different tasks may Ordered Pointing Task: Further evidence of early prefrontal
be required at different ages. decline in normal aging. Journal of Clinical and Experimental
The size and representativeness of the sample is Neuropsychology, 15: 881-895, 1993.
DAIGNEAULT S, BRAUN CMJ and WHITAKER HA. Early effects of
certainly a strength for a study involving normal aging on perseverative and non-perseverative
participants aged 81 years and over although the prefrontal measures. Developmental Neuropsychology, 8: 99-
number of participants may arguably be small by 114, 1992.
DELLA SALA S, GRAY C, SPINNLER H and TRIVELLI C. Frontal lobe
statistical standards. These results are presented as functioning in man: The riddle revisited. Archives of Clinical
a conceptual analysis in an attempt to raise Neuropsychology, 13: 663-682, 1998.
DENCKLA MB. A theory and model of executive function. A
awareness and discuss an important methodological neuropsychological perspective. In GR Lyon and NA
issue in the assessment of EF rather than to Krasnegor (Eds), Attention, Memory, and Executive Function.
provide a definitive answer to a very complex Baltimore: Paul H. Brooks Publishing, 1996, pp. 263-278.
ESLINGER PJ and GRATTAN LM. Frontal lobe and frontal-striatal
question. Nevertheless, the results of these analyses substrates for different forms of human cognitive flexibility.
conducted on the EF measures gave rise to a model Neuropsychologia, 31: 17-28, 1993.
of EF which is intuitive and fits previous findings FOLSTEIN MF, FOLSTEIN SE and MCHUGH PR. “Mini-mental state”:
A practical method for grading the cognitive state of patients
(Eslinger and Grattan, 1993). Obviously, this model for the clinician. Journal of Psychiatric Research, 12: 189-
will need to be examined and tested further. In any 198, 1975.
case, continuing research on EF is needed not only GOLDBERG E and BILDER RM, JR. The frontal lobes and
hierarchical organization of cognitive control. In E Perecman
to better understand the subtleties of human (Ed), The Frontal Lobes Revisited. New York: Lawrence
cognition but also to develop appropriate Erlbaum, 1987, pp. 159-187.
GOLDEN CJ. Stroop Color and Word Test. Chicago: Stoelting,
instruments that will improve the measurement of 1978.
specific cognitive constructs. GRATTAN LM and ESLINGER PJ. Frontal lobe damage in children
and adults: A comparative review. Developmental
Acknowledgement. This research was supported in part Neuropsychology, 7: 283-326, 1991.
by a grant from the National Health and Medical Research GREENWOOD PM. The frontal aging hypothesis evaluated. Journal
of the International Neuropsychological Society, 6: 705-726,
Council of Australia, the Ageing and Alzheimer’s Research 2000.
Foundation and an Infrastructure Stream C grant from the HANES KR. New Tower of London. Carlton, Australia: Vanguard,
Department of Health of New South Wales, Australia. 1997.
HEATON RK, CHELUNE GJ, TALLEY JL, KAY GG and CURTIS G.
Wisconsin Card Sorting Test (WCST) Manual Revised and
Expanded. Odessa,Fl: Psychological Assessment Resources,
REFERENCES 1993.
IVNIK RJ, MALEC JF, SMITH GE and TANGALOS EG.
ALBERT MS, WOLFE J and LAFLECHE G. Differences in abstraction Neuropsychological tests’ norms above age 55: COWAT,
ability with age. Psychology and Aging, 5: 94-100, 1990. BNT, MAE Token, WRAT-R Reading, AMNART, STROOP,
ARBUCKLE JL. Amos User’s Guide Version 3.6. Chicago: TMT, and JLO. The Clinical Neuropsychologist, 10: 262-278,
Smallwaters Corporation, 1997. 1996.
BENTON AL and HAMSHER K. Multilingual Aphasia Examination. KAFER KL and HUNTER M. On testing the face validity of
Iowa City: AJA, 1983. planning/problem-solving tasks in a normal population.
BENTON AL, HAMSHER KD, VARNEY NR and SPREEN O. Journal of the International Neuropsychological Society, 3:
Contributions to Neuropsychological Assessment. New York: 108-119, 1997.
Oxford University Press, 1983. KAPLAN EF, GOODGLASS J and WEINTRAIB S. The Boston Naming
BOONE KB, PONTÓN MO, GORSUCH RL, GONZÁLEZ JJ and MILLER Test. Philadelphia: Lea and Febiger, 1983.
BL. Factor analysis of four measures of prefrontal lobe KAYE K, GRIGSBY J, ROBBINS LJ and KORZUN B. Prediction of
functioning. Archives of Clinical Neuropsychology, 13: 585- independent functioning and behaviour problems in geriatric
595, 1998. patients. Journal of the American Geriatrics Society, 38:
BROWNE MW and CUDECK R. Alternative ways of assessing model 1304-1310, 1990 .
fit. Sociological Methods and Research, 21: 230-258, 1992. KENSINGER EA, MILLIGAN DK, LOCASCIO JJ and CORKIN S. The
BURGESS PW. Theory and methodology in executive function effect of Alzheimer’s disease and Parkinson’s disease on the
research. In P Rabbitt (Ed), Methodology of Frontal and components of working memory. In JH Growdon, RJ
Executive Function. Hove, Sussex: Psychology Press, 1997, Wurtman, S Corkin, and RM Nitsch (Eds), Molecular
pp. 81-116. Genetics of Dementia. Cambridge, MA: Center for Brain
BYRNE BM. Structural Equation Modeling. Basic Concepts, Sciences and Metabolism Charitable Trust, 2000, pp. 291-297.
Applications and Programming. London: Lawrence Erlbaum, KIMBERG DY and FARAH MJ. A unified account of cognitive
2001. impairment following frontal lobe damage: The role of
36 Olivier Piguet and Others

working memory in complex, organized behavior. Journal of ROBBINS TW, JAMES M, OWEN AM, SAHAKIAN BJ, LAWRENCE AD,
Experimental Psychology: General, 122: 411-428, 1993. MCINNES L and RABBITT PMA. A study of performance on tests
LAFLECHE G and ALBERT MS. Executive function deficits in mild from the CANTAB battery sensitive to frontal lobe dysfunction
Alzheimer’s Disease. Neuropsychology, 9: 313-320, 1995. in a large sample of normal volunteers: Implications for
LEVINE B, STUSS DT and MILBERG WP. Concept generation: theories of executive functioning and cognitive aging. Journal
Validation of a test of executive functioning in a normal aging of the International Neuropsychological Society, 4: 474-490,
population. Journal of Clinical and Experimental 1998.
Neuropsychology, 17: 740-758, 1995. RUFF RM. Ruff Figural Fluency Test Administration Manual. San
LEVINE B, STUSS DT, MILBERG WP, ALEXANDER MP, SCHWARTZ M Diego: Neuropsychological Resources, 1988.
and MACDONALD R. The effects of focal and diffuse brain SALTHOUSE TA. The aging of working memory. Neuropsychology,
damage on strategy application: Evidence from focal lesions, 8: 535-543, 1994.
traumatic brain injury and normal aging. Journal of the SALTHOUSE TA and BABCOCK RL. Decomposing adult age
International Neuropsychological Society, 4: 247-264, 1998. differences in working memory. Developmental Psychology,
LEZAK M. Neuropsychological Assessment. New York: Oxford 27: 763-776, 1991.
University Press, 1995. SHALLICE T. From neuropsychology to mental structure. New
LOWE C and RABBITT P. Cognitive models of ageing and frontal York: Cambridge University.Press, 1988.
lobe deficits. In Rabbitt P (Ed), Methodology of Frontal and SHIMAMURA AP. Memory and frontal lobe function. In MS
Executive Function. Hove, Sussex: Psychology Press, 1997, Gazzaniga (Ed), The Cognitive Neurosciences. Cambridge,
pp. 39-59. MA: MIT Press, 1995, pp. 803-813.
LURIA A. The Working Brain. New York: Basic Books, 1973. SHIMAMURA AP. The role of the prefrontal cortex in dynamic
MARSH HW, BALLA JR and MCDONALD RP. Goodness-of-fit filtering. Psychobiology, 28: 207-218, 2000.
indexes in confirmatory factor analysis: The effect of sample SHIMAMURA AP and JURICA PJ. Memory interference effects and
size. Psychological Bulletin, 103: 391-410, 1988. aging: Findings from a test of frontal lobe function.
MILLER EN. CalCAP California Computerized Assessment Neuropsychology, 8: 408-412, 1994.
Package Manual. Los Angeles: Norland Software, 1996. SPREEN O and STRAUSS E. A Compendium of Neuropsychological
MITTENBERG W, SEIDENBERG M, O’LEARY DS and DIGIULIO DV. Tests. New York: Oxford University Press, 1991.
Changes in cerebral functioning associated with normal aging. STUSS DT and ALEXANDER MP. Executive functions and the
Journal of Clinical and Experimental Neuropsychology, 11: frontal lobes: a conceptual view. Psychological Research, 63:
918-932, 1989. 289-298, 2000.
NEILS J, BARIS JM, CARTER C, DELL’AIRA AL, NORDLOH SJ, STUSS DT, CRAIK FIM, SAYER L and FRANCHI D. Comparison
WEILER E and WEISINGER B. Effects of age, education, and of older people and patients with frontal lesions: Evidence
living environment on Boston Naming Test performance. from word list learning. Psychology and Aging, 11: 387-395,
Journal of Speech and Hearing Research, 38: 1143-1149, 1996.
1995. TRANEL D ANDERSON SW and BENTON A. Development of the
PARKIN AJ and LAWRENCE A. A dissociation in the relation between concept of “executive function” and its relationship to the
memory tasks and frontal lobe tests in the normal elderly. frontal lobes. In F Boller and J Grafman (Eds), Handbook of
Neuropsychologia, 32: 1523-1532, 1994. Neuropsychology (vol 9). Amsterdam: Elsevier, 1994, pp. 125-
PENNINGTON BF, BENNETTO L, MCALEER O and ROBERTS RJJ. 148.
Executive functions and working memory: Theoretical and WAITE LM, BROE GA, GRAYSON DA and CREASEY H. The
measurement issues. In LG Reid and NA Krasnegor (Eds), incidence of dementia in an Australian community population:
Attention, Memory and Executive Function. Baltimore: Paul the Sydney Older Persons Study. International Journal of
H. Brookes Publishing Co., 1996, pp. 327-348. Geriatric Psychiatry, 16: 680-689, 2001.
PHILLIPS LH. Do “Frontal Tests” measure executive function? WECHSLER D. WAIS-R Manual. New York: The Psychological
Issues of assessment and evidence from fluency tests. In P Corporation, 1981.
Rabbitt (Ed), Methodology of Frontal and Executive Function. WECHSLER D. WMS-R Manual. New York: The Psychological
Hove, Sussex: Psychology Press, 1997, pp. 191-213. Corporation, 1987.
RABBITT P. Introduction: Methodologies and models in the study WECKER NS, KRAMER JH, WISNIEWSKI A, DELIS DC and KAPLAN
of executive function. In P Rabbitt (Ed), Methodology of E. Age effects on executive ability. Neuropsychology, 14: 409-
Frontal and Executive Function. Hove, Sussex: Psychology 414, 2000.
Press, 1997, pp. 1-38. WEST RL. An application of prefrontal cortex function theory to
RABBITT P and LOWE C. Patterns of cognitive ageing. cognitive aging. Psychological Bulletin, 120: 272-292, 1996.
Psychological Research, 63: 308-316, 2000.
RAZ N, GUNNING FM, HEAD D, DUPUIS JH, MCQUAIN J, BRIGGS
SD, LOKEN WJ, THORNTON AE and ACKER JD. Selective aging
of the human cerebral cortex observed in vivo: Differential
vulnerability of the prefrontal gray matter. Cerebral Cortex, 7: Olivier Piguet, Prince of Wales Medical Research Institute, Barker St, Randwick NSW
268-282, 1997. 2031, Australia. e-mail: [email protected]

(Received 14 June 2002; reviewed 30 September 2002; revised 2 December 2002; accepted 4 March 2003; Action Editor
Carlo Umiltà)
APPENDIX
Correlations (× 100) and SDs for the Initial 24 Component Scores (see Table II for test-domain labels)

Measure SD 1 2 2 2 3 3 4 4 5 5 6 6 7 8 8 8 8 9 9 9 9 9 9 9
1) Total words 6.97 100
2) Highest level achieved 1.55 15 100
2) Planning time item 3 7.01 13 6 100
2) Execution time item 3 23.98 22 44 12 100
3) Switching score 3.12 26 22 3 11 100
3) Clustering score 0.56 6 11 –3 8 – 50 100
4) Switching score 10.97 38 20 –4 13 39 7 100
4) Clustering score 0.13 21 20 5 8 12 19 9 100
5) Total designs 13.53 40 29 11 19 24 11 34 1 100
5) Total repetitions 9.00 12 16 –6 23 4 2 20 – 1 – 14 100
6) Total time 26.61 44 29 22 27 22 11 38 15 53 –1 100
6) Efficiency score 20.04 32 43 13 32 20 9 26 17 33 –5 69 100
7) Longest span 0.89 43 29 20 27 28 –5 43 –3 46 23 46 28 100
8) 1-Back Reaction time 0.23 –2 2 8 13 12 7 6 – 12 5 1 12 30 – 14 100
8) 1-Back Efficiency 37.70 0 28 29 37 17 –3 5 –7 24 –5 20 30 12 48 100
Executive functions in ageing

8) 2-Back Reaction time 0.36 7 27 15 27 6 10 21 0 14 –1 20 21 3 38 38 100

8) 2-Back Efficiency 39.82 22 32 23 24 11 –1 5 –3 48 –7 34 24 28 8 50 24 100
9) Simple Reaction time 264.54 25 10 –3 15 11 2 21 18 34 –9 33 15 20 16 13 10 19 100
9) Complex Reaction time 81.37 26 19 –3 20 21 1 22 5 34 – 10 31 24 1 29 29 35 14 36 100
9) Complex Efficiency 0.15 39 33 11 41 20 5 21 8 33 –2 35 22 18 15 26 34 18 46 71 100
9) Serial 1 Reaction time 79.06 –1 0 –1 – 12 17 2 12 –5 13 –7 11 –2 –2 14 4 6 –4 18 39 14 100
9) Serial 1 Efficiency 0.23 32 23 18 21 16 12 29 19 49 – 12 57 40 26 16 29 24 22 39 63 53 38 100
9) Serial 2 Reaction time 95.01 11 –2 –3 – 10 15 –3 15 13 18 –4 17 4 –2 12 –4 10 – 17 9 35 18 63 37 100
9) Serial 2 Efficiency 0.18 34 34 13 9 28 8 29 12 47 –8 48 36 25 17 17 20 25 17 48 39 33 51 40 100
37