What Cognitive Abilities Are Involved in Trail-Making Performance

NIH Public Access
Author Manuscript
Intelligence. Author manuscript; available in PMC 2012 January 1.
NIH-PA Author Manuscript
Published in final edited form as:

Intelligence. 2011 ; 39(4): 222232. doi:10.1016/j.intell.2011.03.001.
What cognitive abilities are involved in trail-making

performance?
Timothy A. Salthouse
Department of Psychology, University of Virginia, Charlottesville, VA 29904
Abstract
The cognitive abilities involved in the Connections (Salthouse, et al., 2000) version of the trail
making test were investigated by administering the test, along with a battery of cognitive tests and
tests of complex span and updating conceptualizations of working memory, to a sample of over
3,600 adults. The results indicate that this variant of the trail making test largely reflects individual
differences in speed and fluid cognitive abilities, with the relative contributions of the two abilities
varying according to particular measure of performance considered (e.g., difference, ratio,
residual). Relations of age on trail making performance were also examined. Although strong age
differences were evident in the Connections and working memory measures, with both sets of
variables there was nearly complete overlap of the age differences with individual differences in
speed and fluid cognitive abilities.
Keywords
neuropsychological assessment; meaning of tests; working memory; fluid ability; age differences
Trail making tests have been extensively used in neuropsychological assessment (e.g.,
Butler, Retzlaff & Vanderploeg, 1991; Rabin, Barr & Burton, 2005; Sellers & Nadler,
1992). Most variants of this test, which was apparently introduced in 1938 by Partington
(Partington & Leiter, 1949), have at least two conditions. In condition A the participant is to
draw lines to connect circled numbers in a numerical sequence (i.e., 1-2-3, etc.) as rapidly in
possible. In condition B the participant is to draw lines to connect circled numbers and
letters in an alternating numeric and alphabetic sequence (i.e., 1-A-2-B, etc.) as rapidly as
possible.
Although trail making tests are very simple, they have been hypothesized to reflect a wide
variety of cognitive processes including attention, visual search and scanning, sequencing
and shifting, psychomotor speed, abstraction, flexibility, ability to execute and modify a
plan of action, and ability to maintain two trains of thought simultaneously (for citations see
Lezak, Howieson & Loring, 2004; Salthouse & Fristoe, 1995, and Strauss, Sherman &
Spreen, 2006). A number of studies have examined the patterns of relations between trailmaking performance and other variables, with recent reviews by Sanchez-Cubillo, Perianez,
Adrover-Roig, Rodrigues-Sanchez, Rios-Lago, Tirapu and Barcelo (2009) and Misdraji and
Gass (2010). However, most of the studies had limitations such as weak statistical
2011 Elsevier Inc. All rights reserved.

Publisher's Disclaimer: This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our
customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of
the resulting proof before it is published in its final citable form. Please note that during the production process errors may be
discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.
Salthouse
Page 2
procedures, a restricted set of other variables, and relatively small samples, many of which
involved mixtures of different patient groups.
Correlational analyses are most informative when the predictor variables are latent
constructs rather than manifest (observed) variables, and when several predictors are
examined simultaneously. Because they focus on the systematic variance shared by different
observed variables representing the relevant ability, latent constructs minimize task-specific
influences and measurement error. Moreover, the availability of multiple predictors in the
same analysis allows unique (or direct) and shared (or indirect) relations to be distinguished,
instead of relying on an undifferentiated mixture of the two types of influences as is the case
with simple correlations. The current project therefore involved a relatively large sample of
participants to obtain precise estimates of the relations, and multiple measures of cognitive
abilities were acquired to examine relations at the construct level instead of the level of
individual variables. In addition, relations of four different cognitive abilities were examined
simultaneously to allow unique relations to be identified.
The target variables were derived from a variant of the trail making test known as the
Connections Test. The test, which was illustrated in Figure 1 of Salthouse, Toth, Daniels,
Parks, Pak, Wolbrette and Hocking (2000), consists of a set of pages containing 49 circles,
with either a number or a letter with in each circle. The task for the participant is to draw
lines as quickly as possible to connect the elements in sequence, with different pages
involving numeric, alphabetic, or alternating numeric and alphabetic sequences. A recent
confirmatory factor analysis by Atkinson and Ryan (2008) revealed that the measures from
the Connections Test had loadings on the same factors as measures from two other trail
making variants (i.e., the Delis-Kaplan Trail Making Test and the Comprehensive Trail
Making Test). Among the differences between the Connections Test and the standard trail
making test are: (a) irrelevant influences of visual search and hand movements are
minimized because successive target elements are in adjacent locations rather than scattered
around the page; (b) a counterbalanced presentation order of simple (A) and alternating (B)
trials eliminates a confounding in which the alternating condition always follows the simple
condition; (c) both letters and numbers are presented in the simple condition instead of only
numbers, which eliminates a confounding of type of material and condition; (d) multiple
trials are presented in each condition to increase the precision of the performance measures
and allow reliability to be assessed; and (e) administration is efficient because a limit of 20
seconds is allowed to work on each page instead of monitoring the time until all items are
completed.
The cognitive abilities involved in performance on this version of the trail making test were
investigated with the contextual analysis model portrayed in Figure 1. The term contextual
analysis refers to analyses in which the effects on a target variable are examined in the
context of effects on other variables (e.g., Salthouse, 2005; 2010a; Salthouse, Pink &
Tucker-Drob, 2008). The model is portrayed with age as a key predictor variable, but it is
important to recognize that the underlying logic is identical when other variables, such as
severity of a neuropathological condition, presence or absence of particular genes,
psychosocial status, etc., replace age in the model.
Two types of information are available from contextual analyses. One type concerns the
relations of the predictor variables to the target variable. When predictor variables
correspond to different cognitive abilities, the results are informative about the relative
influence of each ability on the target variable, independent of any age-related influences on
the predictor and target variables. That is, because relations of age are postulated to all
variables, the influences of age are effectively partialled out, such that the relations among
other variables correspond to what would be expected at the average age in the sample. In
Salthouse
Page 3
some cases there may be interest in whether the relations between the cognitive abilities and
the target variables differ as a function of age, and this question can be examined by
interaction analyses or by repeating the analyses with subsamples of participants within
narrow age ranges.
The second type of information available from contextual analysis concerns the relations of
age (or any individual difference variable of interest) on the target variable after taking
individual differences in the predictor variables into account. The rationale is that age
relations on the reference cognitive abilities are already well established, and therefore the
primary interest is in the magnitude of the relation of age on the target variable after these
influences have been taken into consideration. If there is little or no direct relation (i.e.,
independent of indirect relations through the cognitive abilities), then the results would be
consistent with an interpretation that most of the age-related influences on the target variable
are shared with influences on the predictor variables. In contrast, a discovery of a unique
relation of age on the target variable would imply that an explanation of the effects of age on
the predictor variables would not be sufficient to account for the effects of age on the target
variable. Information of this type is relevant to the question of how many distinct
dimensions are involved in cross-sectional age differences in cognitive functioning.
Although many different variables or abilities could be used as predictors in contextual

analyses, a considerable amount of research has been conducted with the reference variables
used in the current project. The variables represent four broad cognitive abilities which have
been found to have statistically independent age-related influences (e.g., Salthouse, 2004;
Salthouse & Ferrer-Caja, 2003). The Gf ability consists of three reasoning tests and three
spatial visualization tests which have been combined because the correlation between the
reasoning and spatial visualization constructs was .84 (Salthouse, 2005). These variables
have good reliability (i.e., the median coefficient alpha was .84) and validity (i.e.,
standardized factor loadings greater than .65), with similar patterns in different samples
(Salthouse, Pink & Tucker-Drob, 2008), across different age ranges (Salthouse, 2004), and
when age is partialled from all variables (Salthouse, et al., 2008).
Prior research with contextual analysis methods has revealed that few significant unique
relations of age on the target variables are evident when those variables are considered in the
context of age-related influences on the reference cognitive abilities. This is true for a
variety of memory variables (e.g., Salthouse & Siedlecki, 2007a; Salthouse, Siedlecki &
Krueger, 2006), including prospective memory (Salthouse, Berish & Siedlecki, 2004),
source memory (Siedlecki, Salthouse & Berish, 2005), and several measures of working
memory (Salthouse, Pink & Tucker-Drob, 2008). Contextual analyses have also revealed
very strong relations of fluid cognitive ability (Gf) to constructs hypothesized to represent
executive functioning (e.g., Salthouse, 2005; Salthouse, Atkinson & Berish, 2003; Salthouse
& Davis, 2006), and to assorted variables postulated to represent specific aspects of
executive functioning (e.g., Salthouse, 2005; 2010a; Salthouse, Atkinson & Berish, 2003;
Salthouse & Siedlecki, 2007b). These patterns suggest that Gf may reflect nearly the same
dimension of individual differences as executive functioning (also see Decker, Hill & Dean;
2007; Obonsawin, Crawford, Page, Chalmers, Cochrane & Low, 2002), and therefore to the
extent that trail making performance reflects executive functioning, a unique influence of Gf
on trail making measures would be predicted in the contextual analyses.
If trail making performance is found to be significantly related to Gf, there are at least two
reasons to ask whether the Gf relation is primarily attributable to relations shared with
working memory (WM). First, WM has been hypothesized to be involved in trail making
performance because of the need to keep track of encountered letters and numbers (e.g.,
Crowe, 1998; Sanchez-Cubillo, et al., 2009; Zakzanis, Mraz & Graham, 2005). And second,
Salthouse
Page 4
a considerable amount of research has documented strong correlations between Gf and WM

(e.g., Ackerman, Beier & Boyle, 2004; Blair, 2006; Colom, Rebollo, Palacios, JuanEspinosa, & Kyllonen, 2004; Salthouse, et al., 2008).
The role of WM in the relations of Gf on trail making performance was examined by adding
WM as another predictor in the contextual analysis. The rationale is that if one predictor
variable is primarily responsible for the variance in the target variable, then little influence
of the other variable would be expected when the predictors are examined simultaneously.
In other words, if individual differences in WM are responsible for the Gf relations, then
unique relations of WM with trail making performance should be found when WM is
considered simultaneously with Gf, and inclusion of WM in the analyses should reduce, or
possibly even eliminate, the relation of Gf with trail making performance.
An important issue to be considered when analyzing trail making performance is how to

represent the contrast between performance in the simple (A) and alternating (B) conditions.
A number of derived indices have been proposed (e.g., Corrigan & Hinkeldey, 1987; Drane,
Yspeh, Huthwaite & Klinger, 2002; Lange, Iverson, Zakrzewski, Ethel-King & Franzen,
2005; Misdraji & Gass, 2010), with the simple difference (BA) and ratio (B/A) indices the
most common. With moderately large sample sizes a potentially more informative index,
because it is empirically-based, is the residual derived from regression equations relating
performance in the B condition to performance in the A condition. That is, if the predicted B
performance is designated as B*, then B* = a + bA + error, and the residual is determined by
subtracting the predicted B from the observed B (i.e., B B*). Although a variety of
methods have been used to attempt to derive purer measures of the theoretically most
interesting processes, it is important to recognize that they each involve somewhat different
assumptions about the relations between the A and B variables, and thus they could have
different relations with cognitive abilities and with age.
To summarize, four major questions were investigated in the current project. First, what
cognitive abilities are associated with performance in the Connections version of the trail
making test? Second, is the pattern of cognitive ability relations similar for performance in
the simple (A) and alternating (B) versions, and for the different indices that can be used to
express the contrast between them? Third, what are the relative roles of Gf and WM on
performance in the simple (A) and alternating (B) conditions? And fourth, are there unique
age-related influences in these tasks? The relevant data were obtained from a large sample
of adults between 18 and 98 years of age who performed the Connections version of the trail
making test and a battery of cognitive ability tests. Working memory influences were
investigated in two subsamples from the larger sample in which data were also available on
different types of WM tasks.
Methods
Participants
The total sample consisted of 3,665 adults ranging from 18 to 98 years of age, and included
two partially overlapping subsamples who also performed WM tests. Participants were
recruited from advertisements and referrals from other participants. Various demographic
characteristics and their relations with age are presented in Table 1, where it can be noted
that most participants reported themselves to be in good to excellent health, and had
completed several years of college. The table also contains age-adjusted scaled scores for
four variables from a commercial test battery to assess representativeness of the sample. It
can be seen that the participants in this sample performed at higher levels than the nationally
representative normative samples, and that there were slightly higher scaled scores on the
digit symbol and logical memory variables at older ages.
Salthouse
Page 5
Procedure
The Connections test began with brief instructions indicating that the task consisted of using
a pencil to draw lines connecting circles in a specified sequence as rapidly as possible.
Different conditions within the test involved different sequences for connecting the circles,
with either numeric, alphabetic, or alternating numeric and alphabetic sequences. The eight
separately timed pages were presented in the order N (Numbers), L (Letters), NL (NumbersLetters), LN (Letters-Numbers), LN, NL, L, and N.
Participants were allowed to work on each test page for 20 seconds, and performance was
assessed in terms of the number of correct connections minus the number of incorrect
connections. Unlike some versions of trail making tests, errors were not pointed out by the
examiner during the test because of a concern that differences in how quickly the examiner
noted and reported the error, and how quickly the examinee responded to the error, would
introduce additional variability in the primary performance measure of number of correct
responses in the designated time. Furthermore, preliminary analyses of the errors revealed
that they were relatively infrequent, and thus were unlikely to distort the patterns based on
the number correct scores.
Higher scores correspond to better performance when performance is assessed in terms of

the number of items completed in a fixed period of time. Scores in the simpler A condition
will therefore usually be larger than those in the B condition, which means that the BA
difference will be negative and the B/A ratio will be less than 1, with values of the former
farther from 0, and of the latter farther from 1, indicating larger discrepancies between
performance in the simple and the alternating conditions. It is worth noting that performance
in the A and B versions could be expressed in units of seconds per item, in which case the B
values would typically be larger than the A values, and larger discrepancies between
conditions would be evident in BA differences greater than zero and B/A ratios greater
than 1.)
The 16 cognitive tests are briefly described in the appendix, and more details are provided in
other publications (e.g., Salthouse, 2005; Salthouse, Atkinson & Berish, 2003; Salthouse &
Ferrer-Caja, 2003; Salthouse, et al., 2008). The variables represent the following cognitive
abilities, fluid ability (Gf), verbal episodic memory, perceptual speed, and vocabulary.
A total of 830 participants performed the two complex span WM tasks illustrated in Figure 2
(see Conway, Kane, Bunting, Hambrick, Wilhelm & Engle, 2005, for more detailed
descriptions). Both tasks involved the presentation of information that was to be
remembered, either positions in a matrix or identities of letters, and intervening processing
requiring a decision. Processing in the Operation Span (OSpan) task involved answering
arithmetic problems, and processing in the Symmetry Span (SymSpan) task involved
making vertical symmetry decisions (i.e., was the display symmetric along the vertical
axis?). The number of to-be-remembered items increased across successive trials, and at the
end of each sequence within a given trial the participant attempted to recall the presented
items from a set of 12 letters (OSpan) or 16 locations (SymSpan) alternatives. Separate
measures of processing (accuracy on the processing component) and storage (number of
items recalled) were obtained in each span task.
A total of 1,056 participants performed two running memory tests designed to require
updating of continuously changing information (see Salthouse, et al., 2008, for more
detailed description of the tasks). As illustrated in Figure 3, one updating task involved
sequences of letters and the other involved sequences of matrix positions. The requirement
in both versions was to report the last 4 items from a sequence which varied unpredictably in
Salthouse
Page 6
length from 4 to 12 items. Performance in both tasks was assessed as the number of items
recalled in the correct order from the last four items in the lists.
Results
Figure 4 portrays the means, with standard errors, of the number of items correct in the four
types of trials in the Connections test as a function of age decade. It can be seen that, with
the exception of the youngest age, performance was slightly faster with numbers than with
letters, and was faster when the alternating sequence started with numbers than when it
started with letters, despite the fact that the two sequences are identical after the first item. In
all remaining analyses the N and L variables were combined for the simple (A) condition,
and the NL and LN variables were combined for the alternating (B) condition.
Strong age relations were evident in both the simple (A) and alternating (B) measures as the
correlations with age were .61 for A and .48 for B. Nonlinear age trends were also
examined, but the increment in R2 associated with the quadratic age term was only .01 for
both the A and B measures. Examination of Figure 4 reveals that the difference between the
mean number of correct items in the simple (A, solid lines) and alternating (B, dotted lines)
conditions was smaller at older ages. This pattern is also apparent in the cross-sectional age
slopes (i.e., number of items/year) which were more negative for performance in the A (.
28) condition than in the B (.14) condition.
There was no significant age relation on the residual (B.A) index (r = .00), and only a small
relation on the ratio (B/A) index (r = .07), but a large positive correlation of age with the B
A difference (r = .49). This latter correlation indicates that increased age was associated with
a smaller absolute discrepancy (i.e., the difference was less negative, and closer to 0)
between performance in the simple (A) and alternating (B) conditions.
The contextual analysis model in Figure 1 was next used to determine which cognitive
abilities were involved in this task, and the extent to which the cross-sectional age
differences in the simple and alternating conditions were independent of age differences in
other cognitive variables.1 The structural equation models were investigated with the
AMOS (Arbuckle, 2007) program including the full-information maximum likelihood
estimation algorithm to deal with missing data by using all of the available information in
the estimates. Standardized factor loadings and factor correlations for the complete sample
are contained in Table 2, and means (and standard errors) of the composite cognitive ability
constructs used in the contextual analyses are portrayed as a function of decade in Figure 5.
It should be noted that the age trends in Figure 5 are similar to those obtained from
nationally representative samples (cf., Salthouse, 2009; Figures 1.6 to 1.8 in Salthouse,
2010b).
The results of the contextual analyses in the total sample are presented in the top of Table 3.
Entries in the column labeled total age are simple correlations, and entries in the other
columns are standardized regression coefficients derived from a model like that portrayed in
Figure 1. Note that performance in the simple (A) condition was strongly related to speed,
1It is important to note that the contextual analysis results are not an artifact of the use of statistical procedures to control the variation
in the reference cognitive abilities because a similar reduction of the age relations was evident when the analyses were conducted with
subsamples within a narrow range of scores on the reference abilities, thereby relying on matching instead of statistical adjustment. To
illustrate, the correlation in the total sample between age and performance in the simple (A) condition was .61, and after statistical
control of the variation in both the speed and Gf composite scores it was .12. In the sample of participants within the middle 20% of
the distribution on both the speed and Gf composites (N = 135), the age correlation was .19. Although the two adjusted correlations
were quite comparable, this degree of matching required elimination of over 90% of the data, and thus it is much less efficient than
statistical control which is based on all of the data. (See Salthouse, 2010b, for another example of the advantage of statistical control
rather than matching).
Salthouse
Page 7
and was also related to Gf. In addition, there were unique relations of Gf on performance in
the alternating (B) condition after controlling influences through the simple condition. The
entries in the bottom of Table 3 indicate that the patterns were also similar in separate
analyses conducted on data from participants in each age decade. The only notable
differences across age groups were more negative total age relations for adults over 60 years
of age.
The next set of analyses was conducted with composite target variables formed by averaging
the z-scores for the variables in the simple (A) and alternating (B) conditions, and then
computing the difference, ratio, and residuals from the composite scores. Contextual
analysis results on these composite variables are reported in the top of Table 4.
It can be seen that there were strong relations of speed and a moderate relation of Gf on the
measure of performance in the simple (A) condition, and a slightly weaker speed relation but
a stronger Gf relation on the measure of performance in the alternating (B) condition. Of
particular interest is the absence of a Gf relation on the difference (BA) measure, which
was primarily related to speed. Higher speed was associated with a smaller B/A ratio
(reflecting a greater discrepancy between performance in the alternating and simple
conditions among people with high levels of speed), and higher Gf was associated with a
larger B/A ratio (reflecting a smaller discrepancy between performance in the B and A
conditions for people with high levels of Gf). A similar pattern was evident in the analysis
with the residuals representing deviations of B from the prediction based on A, as the
deviations were smaller (more negative) with higher speed, and were larger (more positive)
with higher Gf.
The results in the top of Table 4 indicate that inferences about relations of cognitive abilities
on Connections performance will vary depending on the particular measure of performance
examined. For example, a moderate to large relation of speed was apparent in all measures,
but there was no relation of Gf with the difference score (BA) measure. If Gf is assumed to
reflect nearly the same dimension of individual differences as executive functioning, these
results imply that the difference score eliminates the influence of executive functioning in
trail making performance as assessed in the Connections Test.
Relations of sex, self-rated health, and amount of education were also examined with each
of the measures of Connections performance. Although some correlations were significantly
different from 0 (e.g., .20 between health and performance in the simple (A) condition), none
of the relations was independent of individual differences in the reference abilities. That is,
when these variables replaced age in the model in Figure 1, their standardized regression
coefficients on the A and B target variables ranged from .02 to .03, and none was
significantly different from 0.
All analyses except that with the difference score measure revealed a moderate to large
relation of Connections performance with Gf, in addition to a unique relation of Gf on
performance in the alternating (B) condition after control of the variance in performance in
the simple (A) condition (cf. Table 3). Because some of the participants also performed WM
tasks, relations on the Connections scores could be examined with Gf and WM as
simultaneous predictors. As noted above, the rationale is that when both predictors are
considered simultaneously, most of the predictive relations can be inferred to overlap if only
one of the predictors is found to have unique relations. For example, if most of the Gf
relations on performance are attributable to aspects of WM, then when both are considered
simultaneously only WM might be expected to have significant relations with the measures
of trail making performance.
Salthouse
Page 8
The (storage) recall and (processing) error measures in each complex span task were
converted to z-scores, and the means and standard errors plotted by decade in Figure 6. It
can be seen that performance was significantly lower with increased age in each measure.
The age correlations were .35 for OSpan storage accuracy, .24 for OSpan processing
errors, .56 for SymSpan storage accuracy, and .22 for SymSpan processing errors.
The initial analysis with the complex span WM construct used it as the target variable in a
contextual analysis with the reference cognitive abilities as predictors. A latent complex
span construct was created with the two processing and two storage variables as indicators,
and loadings for this construct, which were all significantly (p<.01) different from zero, are
reported in the bottom of Table 2. The contextual analysis results in which the WM
construct was predicted from the reference cognitive abilities are presented in the bottom of
Table 4. It can be seen that the largest relation with the complex span WM construct was
with Gf ability. The absence of a unique age relation indicates that when the age variation in
the reference abilities was controlled, there were no significant relations of age on the WM
measures.
Next the contextual analysis was repeated on the two Connections measures after adding the
complex span WM construct as another predictor. The results of these analyses, presented in
the middle panel of Table 5, indicate that there was no significant relation of the complex
span WM construct on performance in this version of the trail making test when it was
examined in the context of other abilities. Furthermore, the pattern of reference cognitive
ability relations on the Connections trail making variables was very similar to that without
inclusion of the WM predictor (i.e., compare the values in the top two rows with those in the
fifth and sixth rows).
It is possible that relations of the Gf construct were stronger than those with the WM
construct because the Gf construct was based on six variables instead of only two, such that
its scope may have been broader than the WM construct. The contextual analysis was
therefore repeated with a narrower Gf construct based on only the Matrix Reasoning and
Letter Sets variables. As can be seen in the third and fourth, and seventh and eighth, rows of
Table 5, the results in this analysis were very similar to those from the analysis with the
broad Gf construct, and to the analysis without the WM construct.
Complex span tasks are not the only way to assess WM, and they may not be the most
sensitive method of assessing sequencing and updating processes which may be involved in
the Connections task. Because it is possible that somewhat different results would be found
with an updating conceptualization of WM, similar analyses were carried out with the
updating WM measures. As with the complex span measures, the number correct scores in
the letters and positions versions of the running memory tasks were converted to z-scores,
and the means and standard errors of these updating z-scores plotted in Figure 7 as a
function of age. It is apparent in the figure that there were nearly linear age trends in the
updating tasks, with age correlations of .27 for running memory letters and .39 for
running memory positions.
Contextual analysis results with the updating WM construct as the target variable are
reported in the bottom of Table 4. As with the complex span WM construct, there was a
strong Gf relation and no unique relation of age on WM, suggesting that this WM construct
also largely reflects the same dimension of age-related individual differences as the Gf
construct. The bottom of Table 5 contains the results of the contextual analyses of trail
making performance with the addition of the updating WM construct. Notice that there was
little relation of the updating WM construct on the Connections trail making measures
beyond Gf, and that the pattern of reference construct relations was similar with and without
Salthouse
Page 9
WM. Furthermore, this was also true in the analysis with a narrower Gf construct defined by
only two variables.
A final analysis involved a combined WM construct based on the four complex span
variables in addition to the two updating variables for the 234 participants with data on all
six variables. The results in the bottom of Table 4 indicate that, as with the separate WM
constructs, the combined WM construct was strongly related to Gf, and the results in Table 5
indicate that inclusion of combined WM construct had little effect on the coefficients for
other abilities, and that the broader WM construct was not significantly related to
Connections trail making performance when it was considered simultaneously with the other
cognitive ability predictors.
Discussion
The results of this project provide an unambiguous answer to the question of what cognitive
abilities are involved in the Connections version of the trail making test. That is, the
discovery of large relations of speed and Gf abilities in both conditions of the task implies
that these abilities contribute to successful trail making performance. The speed influence
likely occurs because of the requirement for rapid responding in both conditions. The Gf
relation may be due to the cognitive demands of maintaining the current sequence position
while searching for next element in the sequence, which are required in both the simple (A)
and alternating (B) versions of the task. Furthermore, the existence of a unique Gf relation in
the alternating version suggests that there are additional requirements related to fluid
abilities when there is a need to alternate between two sequences.
Trail making tests such as the Connections Test are very sensitive to individual differences,
and particularly to age-related differences. Surprisingly, however, the age relations in the
current project were greater in the simple (A) condition than in the more complex alternating
(B) condition. This may be because multiple processes are involved in both conditions of the
test, but age-related individual differences in processes specific to the more complex version
were small, and highly variable, relative to individual differences in processes common to
both versions.
Another important finding in the study is that the nature of the contrast between
performance in the simple (A) and alternating (B) conditions affects the relations with other
cognitive abilities and with age. For example, the discovery that the general pattern of
ability relations was similar in the simple (A) and alternating (B) conditions suggests that
there may be little advantage of using one versus the other as a single measure of trail
making performance. The availability of both measures obviously yields greater information
than any one measure, but both measures appear to assess a mixture of speed and Gf
abilities. Furthermore, the contextual analysis results indicate that the simple difference (B
A) primarily reflects speed, and thus it may not be sensitive to other influences of interest in
the test.
Contrasts expressed as ratios (B/A) or residuals (B.A) were found to have similar patterns. In
both cases there were weak direct age relations, but significant Gf and speed relations. The
ratio measure does not completely eliminate the influence of speed, but it does so to a much
greater extent than the simple difference, and for this reason it may be the preferable derived
index in trail making tasks when sample sizes are too small to permit meaningful regression
analyses to derive residuals.
Both WM conceptualizations had strong correlations with the Connections variables, but
because they are highly correlated with Gf, they need to be examined simultaneously with
Gf to determine whether they have unique, or incremental, prediction. That is, when two
Salthouse
Page 10
variables are both related to a criterion, the most relevant question is not necessarily their
total (i.e., unique plus shared) relations, but rather their unique contributions.

The analyses with WM and Gf as simultaneous predictors revealed no unique prediction of

WM, and this was true even when the Gf construct was defined narrowly based on only two
variables and the WM construct was defined broadly with variables from both the complex
span and updating conceptualizations of WM. There was also very little shift in the
magnitude of the Gf influence on Connections performance when the WM construct was
included in the model. These results are consistent with an interpretation that WM, at least
as operationalized with these particular complex span and updating tasks, may reflect nearly
the same dimension of individual differences as Gf, but from a narrower perspective and
consequently with more limited predictive power (also see Salthouse, et al., 2008, for a
similar interpretation). Many of the prior studies in which relations of WM were
investigated have not included other predictor variables in the analyses, and therefore any
variance that WM shared with other predictors was necessarily attributed to WM. However,
the availability of multiple predictors allows their relative contributions to be assessed, and
in the current case, even the narrow Gf construct appears to have broader predictive power
than WM. Similar findings were recently reported by Shamosh, DeYoung, Green, Reis,
Johnson, Conway, Engle, Braver and Gray (2008) as there was little unique contribution of
WM in the prediction of delay discounting performance when it was examined
simultaneously with a measure of general intelligence.
In conclusion, the current article has both substantive and methodological contributions. The
substantive results indicate that the Connections version of the trail making test primarily
reflects cognitive abilities of speed and fluid intelligence (Gf), that the pattern of influences
depends on the measures used to summarize trail making performance, and that there are no
relations of working memory with Connections trail making performance beyond those
shared with fluid ability. Among the methodological contributions are a set of procedures to
investigate the meaning of a neuropsychological or cognitive test by determining its unique
relations to established cognitive abilities, to estimate the magnitude of unique age-related
influences on a target variable, and to determine whether additional constructs, such as
working memory, represent the same dimensions of individual differences as established
cognitive ability constructs.
Supplementary Material
Refer to Web version on PubMed Central for supplementary material.
Acknowledgments
Preparation of this manuscript was supported by a grant from the National Institute on Aging (R37 024270). The
information in this manuscript has not been previously published, and the author has no conflicts of interest.
References
Ackerman PL, Beier ME, Boyle MO. Working memory and intelligence: The same or different
constructs? Psychological Bulletin. 2004; 131:3060. [PubMed: 15631550]
Arbuckle, JL. AMOS (Version 7) [Computer Program]. Chicago: SPSS; 2007.
Atkinson TM, Ryan JP. The use of variants of the trail making test in serial assessment: A construct
validity study. Journal of Psychoeducational Assessment. 2008; 26:4253.
Blair C. How similar are fluid cognition and general intelligence? A developmental neuroscience
perspective on fluid cognition as an aspect of human cognitive ability. Behavioral and Brain
Sciences. 2006; 29:109160. [PubMed: 16606477]
Salthouse
Page 11

Butler M, Retzlaff P, Vanderploeg R. Neuropsychological test usage. Professional Psychology:

Research and Practice. 1991; 6:510512.
Colom R, Rebollo I, Palacios A, Juan-Espinosa M, Kyllonen PC. Working memory is (almost)
perfectly predicted by g. Intelligence. 2004; 32:277296.
Conway ARA, Kane MJ, Bunting MF, Hambrick DZ, Wilhelm O, Engle RW. Working memory span
tasks: A methodological review and users guide. Psychonomic Bulletin & Review. 2005; 12:769
786. [PubMed: 16523997]
Corrigan JD, Hinkeldey NS. Relationships between parts A and B of the Trail Making Test. Journal of
Clinical Psychology. 1987; 43:402409. [PubMed: 3611374]
Crowe SF. The differential contribution of mental tracking, cognitive flexibility, visual search, and
motor speed to performance on Parts A and B of the Trail Making Test. Journal of Clinical
Psychology. 1998; 54:585591. [PubMed: 9696108]
Decker SL, Hill SK, Dean RS. Evidence of construct similarity in executive functions and fluid
reasoning abilities. International Journal of Neuroscience. 2007; 117:736748.
Drane DL, Yuspeh RL, Huthwaite JS, Klinger LK. Demographic characteristics and normative
observations for derived trail making indices. Neuropsychiatry, Neuropsychology and Behavioral
Neurology. 2002; 15:3943.
Kline, RB. Principles and practice of structural equation modeling. 2. New York: Guilford Press; 2005.
Lange RT, Iverson GL, Zakrzewski MJ, Ethel-King PE, Franzen MD. Interpreting the trail making test
following traumatic brain injury: Comparison of traditional time scores and derived indices.
Journal of Clinical and Experimental Neuropsychology. 2005; 27:897906. [PubMed: 16183622]
Lezak, MD.; Howieson, DB.; Loring, DW. Neuropsychological Assessment. 4. Oxford University
Press; New York, NY: 2004.
Misdraji EL, Gass CS. The Trail Making Test and its neurobehavioral components. Journal of Clinical
and Experimental Neuropsychology. 2010; 32:159163. [PubMed: 19459077]
Obonsawin MC, Crawford JR, Page J, Chalmers P, Cochrane R, Low G. Performance on tests of
frontal function reflect general intellectual ability. Neuropsychologia. 2002; 40:970977.
[PubMed: 11900749]
Parkington JE, Leiter RG. Partingtons Pathway Test. The Psychological Service Center Bulletin.
1949; 1:920.
Rabin LA, Barr WH, Burton LA. Assessment practices of clinical neuropsychologists in the United
States and Canada: A survey of INS, NAN, and APA Division 40 members. Archives of Clinical
Neuropsychology. 2005; 20:3365. [PubMed: 15620813]
Salthouse TA. Localizing age-related individual differences in a hierarchical structure. Intelligence.
2004; 32:541561.
Salthouse TA. Relations between cognitive abilities and measures of executive functioning.
Neuropsychology. 2005; 19:532545. [PubMed: 16060828]
Salthouse TA. Decomposing age correlations on neuropsychological and cognitive variables. Journal
of the International Neuropsychological Society. 2009; 15:650661. [PubMed: 19570312]
Salthouse TA. Is flanker-based inhibition related to age? Identifying specific influences of individual
differences on neurocognitive variables. Brain and Cognition. 2010a; 73:5161. [PubMed:
20303636]
Salthouse, TA. Major issues in cognitive aging. Oxford University Press; New York, NY: 2010b.
Salthouse TA, Atkinson TM, Berish DE. Executive functioning as a potential mediator of age-related
cognitive decline in normal adults. Journal of Experimental Psychology: General. 2003; 132:566
594. [PubMed: 14640849]
Salthouse TA, Berish DE, Siedlecki KL. Construct validity and age sensitivity of prospective memory.
Memory & Cognition. 2004; 32:11331148.
Salthouse TA, Davis HP. Organization of cognitive abilities and neuropsychological variables across
the lifespan. Developmental Review. 2006; 26:3154.
Salthouse TA, Ferrer-Caja E. What needs to be explained to account for age-related effects on multiple
cognitive variables? Psychology and Aging. 2003; 18:91110. [PubMed: 12641315]
Salthouse
Page 12

Salthouse TA, Fristoe NM. A process analysis of adult age effects on a computer-administered trail
making test. Neuropsychology. 1995; 9:518528.
Salthouse TA, Pink JE, Tucker-Drob EM. Contextual analysis of fluid intelligence. Intelligence. 2008;
36:464486. [PubMed: 19137074]
Salthouse TA, Siedlecki KL. An individual differences analysis of false recognition. American Journal
of Psychology. 2007a; 120:429458. [PubMed: 17892087]
Salthouse TA, Siedlecki KL. Efficiency of route selection as a function of adult age. Brain and
Cognition. 2007b; 63:279287. [PubMed: 17079064]
Salthouse TA, Siedlecki KL, Krueger LE. An individual differences analysis of memory control.
Journal of Memory and Language. 2006; 55:102125.
Salthouse TA, Toth J, Daniels K, Parks C, Pak R, Wolbrette M, Hocking K. Effects of aging on the
efficiency of task switching in a variant of the Trail Making Test. Neuropsychology. 2000;
14:102111. [PubMed: 10674802]
Sanchez-Cubillo I, Perianez JA, Adrover-Roig D, Rodriguez-Sanchez JM, Rios-Lago M, Tirapu J,
Barcelo F. Construct validity of the Trail Making Test: Role of task-switching, working memory,
inhibition/interference control, and visuomotor abilities. Journal of the International
Neuropsychological Society. 2009; 15:438450. [PubMed: 19402930]
Sellers AH, Nadler JD. A survey of current neuropsychological assessment procedures used for
different age groups. Psychotherapy in Private Practice. 1992; 11:4757.
Shamosh NA, DeYoung CG, Green AE, Reis DL, Johnson MR, Conway ARA, Engle RW, Braver TS,
Gray JR. Individual differences in delay discounting: Relation to intelligence, working memory,
and anterior prefrontal cortex. Psychological Science. 2008; 19:904911. [PubMed: 18947356]
Siedlecki KL, Salthouse TA, Berish DE. Is there anything special about the aging of source memory?
Psychology and Aging. 2005; 20:1932. [PubMed: 15769211]
Strauss, E.; Sherman, EMS.; Spreen, O. A compendium of neuropsychological tests: Administration,
norms, and commentary. 3. Oxford University Press; New York, NY: 2006.
Wechsler, D. Wechsler Adult Intelligence Scale. 3. San Antonio, TX: The Psychological Corporation;
1997a.
Wechsler, D. Wechsler Memory Scale: Third Edition. San Antonio, TX: Psychological Corporation;
1997b.
Zakzanis KK, Mraz R, Graham SJ. An fMRI study of the Trail Making Test. Neuropsychologia. 2005;
43:18781886. [PubMed: 16168730]
Appendix A
Description of variables
Fluid cognitive ability (Gf)
Matrix Reasoning: Determine which pattern best completes the missing cell in a matrix.
Shipley Abstraction: Determine the words or numbers that are the best continuation of a
sequence.
Letter Sets: Identify which of five groups of letters is different from the others.
Spatial Relations: Determine the correspondence between a 3-D figure and alternative 2-D
figures.
Paper Folding: Determine the pattern of holes that would result from a sequence of folds and
a punch through the folded paper.
Form Boards: Determine which combinations of shapes are needed to fill a larger shape.
Salthouse
Page 13
Memory
Logical Memory: Number of idea units recalled across three stories.
Word Recall: Number of words recalled across trials 1 to 4 of the same word list.
Paired Associates: Number of response terms recalled when presented with a stimulus item.
Speed
Digit Symbol: Use a code table to write the correct symbol below each digit.
Letter Comparison: Same/different comparison of pairs of letter strings.
Pattern Comparison: Same/different comparison of pairs of line patterns.
Vocabulary
WAIS Vocabulary: Provide definitions of words
WJ-R Picture Vocabulary: Name the pictured object
Antonym Vocabulary: Select the best antonym of the target word
Synonym Vocabulary: Select the best synonym of the target word

Salthouse
Page 14

Figure 1.
Illustration of the structural equation model used to investigate the relations of reference
cognitive abilities (predictors) on simple and alternating versions of trail making
performance. Circles correspond to latent constructs representing the variance common to
several observed (manifest) variables, which are portrayed as squares. Single-headed arrows
represent directed (regression) relations, and double-headed arrows represent correlations.
Arrows with numbers adjacent to them had their coefficients were fixed to 1, and arrows
without a source represent variances. The cognitive variables used to define the latent
predictor constructs are not portrayed in the figure, but are described in the appendix, with
the factor loadings presented in Table 2.

Salthouse
Page 15

Figure 2.
Illustration of the sequence of displays in the Symmetry Span and Operation Span tasks used
to assess working memory with complex span procedures.

Salthouse
Page 16

Figure 3.
Illustration of the sequence of displays in the two running memory tasks used to assess
working memory with updating procedures.

Salthouse
Page 17

Figure 4.
Mean number of items correct when connecting elements in numerical, alphabetic, and
alternating numeric and alphabetic order starting with either numbers or letters as a function
of age decades. Bars around the means are standard errors. Sample sizes in each decade
ranged from 220 to 830.

Salthouse
Page 18

Figure 5.
Mean z-scores as a function of age decade on four reference cognitive abilities in the
complete sample. Bars around the means are standard errors. Sample sizes in each decade

Salthouse
Page 19

Figure 6.
Mean z-scores as a function of age decade for storage and processing measures from the
complex span tasks. Bars around the means are standard errors. Sample sizes in each decade

Salthouse
Page 20

Figure 7.
Mean z-scores as a function of age decade for the running memory measures from the
updating tasks. Bars around the means are standard errors. Sample sizes in each decade

Salthouse
Page 21
Table 1
Sample characteristics
Sample 1
Mean (SD)
N
Age Correlation
3,665
Age
51.0 (18.5)
.65
.03
Years of Education
15.6 (2.7)
.17*
Self-Rated Health
2.2 (0.9)
.15*
Vocabulary
12.6 (3.0)
.04
Digit Symbol
11.3 (2.9)
.07*
Word Recall
12.2 (3.4)
.00
Logical Memory
11.8 (2.9)
.09*
Sex (% Female)
Scaled Scores
Sample 2 (Complex Span WM)

N
830
Age
49.6 (18.1)
Sex (% Female)
.66
.03
Years of Education
15.9 (2.6)
.24*
Self-Rated Health
1.7 (0.9)
.11*
Vocabulary
13.2 (2.7)
.05
Digit Symbol
11.7 (2.7)
.12*
Word Recall
12.8 (3.2)
.04
Logical Memory
12.2 (2.7)
.13*
Scaled Scores
Sample 3 (Updating WM)

N
1,056
Age
56.1 (16.5)
.66
.05
Years of Education
15.9 (2.7)
.18*
Self-Rated Health
2.1 (0.8)
.07
Vocabulary
12.2 (2.8)
.17*
Digit Symbol
11.8 (3.2)
.10*
Word Recall
12.1 (3.1)
.05
Logical Memory
11.9 (2.9)
.12*
Sex (% Female)
Scaled Scores
p<.01
Note: Samples 2 and 3 are partially overlapping subsets of Sample 1. Health was rated on a scale ranging from 1 for excellent to 5 for poor.
Scaled scores are age-adjusted scores from the Wechsler (1997a, 1997b) test batteries. In the nationally representative samples used to establish the
norms for these tests, the age-adjusted scaled scores were designed to have means of 10 and standard deviations of 3.
Salthouse
Page 22
Table 2
Standardized coefficients
Construct
Variable
Gf
Matrix Reasoning
.86
Shipley Abstraction
.83
Letter Sets
.69
Spatial Relations
.78
Paper Folding
.77
Form Boards
.70
Word Recall
.78
Paired Associates
.76
Logical Memory
.70
Digit Symbol
.85
Pattern Comparison
.77
Letter Comparison
.78
WAIS Vocabulary
.84
Picture Vocabulary
.81
Synonym Vocabulary
.88
Antonym Vocabulary
.83
Numbers 1
.87
Numbers 2
.90
Letters 1
.89
Letters 2
.89
Numbers-Letters 1
.78
Numbers-Letters 2
.71
Letters-Numbers 1
.68
Letters-Numbers 2
.65
OSpan Storage
.78
OSpan Processing
.51
SymSpan Storage
.65
SymSpan Processing
.50
Running Memory Letters
.72
Running Memory Positions
.84
Memory
Speed
Vocabulary
Coefficient
Connections
Simple (A)
Alternating (B)
Complex Span
Updating
Construct Correlations
Gf: Memory = .65; Speed = .56; Vocabulary = .74; Complex Span = .87; Updating = .80
Memory: Speed = .50; Vocabulary = .66; Complex Span = .59; Updating = .59
Speed: Vocabulary = .48; Complex Span = .54; Updating = .54
Vocabulary: Complex Span = .62; Updating = .56
Complex Span: Updating = .81.
.09*
.55*
Alternating (B)
.06
.00
Alternating (B)
.01
.14
Alternating (B)
.03
.07
Alternating (B)
.01
.03
Alternating (B)
.03
.24*
Alternating (B)
.00
.24*
Alternating (B)
.01
.13
.23*
.08
Simple (A)
Alternating (B)
8089 (N = 199)
.02
.23*
Simple (A)
7079 (N = 439)
.03
.22*
Simple (A)
6069 (N = 587)
.01
.07
Simple (A)
5059 (N = 829)
.05
.10
Simple (A)
4049 (N = 582)
.06
.11
Simple (A)
3039 (N = 341)
.08
.11*
Simple (A)
2029 (N = 667)
.03
.63*
Unique
Simple (A)
p<.01
Total
Total Sample (N = 3,665)
Age
.57*
NA
.74*
NA
.67*
NA
.77*
NA
.56*
NA
.66*
NA
.61*
NA
.73*
NA
Simple (A)
.17
.40*
.32*
.36*
.40*
.43*
.39*
.40*
.44*
.41*
.29*
.17
.32*
.30*
.35*
.31*
Gf
.16
.07
.02
.07
.06
.03
.01
.02
.03
.03
.14
.10
.01
.04
.01
.01
Mem
.05
.64*
.01
.61*
.02
.67*
.08
.60*
.09
.61*
.06
.68*
.15
.69*
.00
.69*
Speed
.17
.19
.09
.23*
.03
.12
.01
.11
.06
.13
.04
.03
.01
.07
.01
.09*
Vocab
.95
.94
.94
.95
.95
.95
.95
.95
CFI
.05
.05
.05
.05
.05
.05
.05
.06
RMSEA
Contextual analysis results based on Figure 1 for total sample and for each decade
Table 3
Salthouse
Page 23
Note: Total age refers to the simple correlation of age and the variable. Unique age is the standardized regression coefficient for age when the reference cognitive abilities are included as simultaneous
predictors of the target variable as portrayed in Figure 1. NA means that the parameter was not estimated because it was not applicable in the model. Entries in the Gf, Mem (Memory), Speed, and Vocab
(Vocabulary) columns are standardized regression coefficients obtained from a model similar to that in Figure 1. CFI is the Comparative Fit Index and RMSEA is the Root Mean Squared Error of
Approximation. CFI values greater than .92 and RMSEA values less than .10 are often considered to represent a good fit (Kline, 2005).
Salthouse
Page 24

.10*
.48*
Alternating (B)
.10*
.14*
.08
.15
.02
.49*
.07*
.00
.61*
.45*
.58*
Difference (BA)
Ratio (B/A)
Residual (B.A)
Complex Span
Updating
Combined
Gf
Mem
.81*
.79*
.80*
.08
.14
.06
.01
.15
.05
.13
.05
.04
.15*
.07
.08
.20*
.02
.01
.09*
.62*
.01
.08
Vocab
.04
.67*
Speed
.42*
.00
.00
Working Memory
.46*
.36*
.04
.52*
.30*
Connections
.92
.93
.92
.93
.93
.93
.93
.93
CFI
.06
.08
.07
.08
.08
.08
.08
.08
RMSEA
Note: Connections variables are composites formed by averaging z-scores from the relevant variables. Total age refers to the simple correlation of age and the variable. Unique age is the standardized
regression coefficient for age when the reference cognitive abilities are included as simultaneous predictors of the target variable as portrayed in Figure 1. Entries in the Gf, Mem (Memory), Speed, and
Vocab (Vocabulary) columns are standardized regression coefficients obtained from a model similar to that in Figure 1. With some variables the coefficient for speed is negative because higher speed is
associated with a smaller (less negative) difference, ratio, or residual (e.g., people with higher speed had smaller difference between performance in the alternating condition and performance in the simple
condition). CFI is the Comparative Fit Index and RMSEA is the Root Mean Squared Error of Approximation.
p<.01
.03
.61*
Simple (A)
.06
Unique
Total
Age
Contextual analysis results on composite variables in the Connections test and on latent constructs in working memory tasks
Table 4
Salthouse
Page 25
.09*
.55*
Alternating (B)
.12*
.04
.09*
.55*
.63*
.55*
Alternating (B)
Simple (A)
Alternating (B)
.12*
.06
.08*
.55*
.63*
.55*
Alternating (B)
Simple (A)
Alternating (B)
.14*
.55*
Alternating (B)
.46*
.34*
.35*
.31*
Gf
.63*
.21
.38*
.24
.70*
NA
.74*
NA
.54*
.56*
.30*
.44*
WM = Updating
.73*
NA
.73*
NA
.02
.00
.00
.02
.03
.02
.01
.01
.03
.02
.01
.01
Mem
.09*
.55*
Alternating (B)
.04
.13*
.63*
.55*
Simple (A)
Alternating (B)
Gf = Matrix Reasoning & Letter Sets
.03
.63*
Simple (A)
.72*
NA
.73*
NA
.66*
.35
.32*
.34*
.03
.02
.01
.00
.03
.67*
.01
.69*
.03
.68*
.02
.71*
.04
.67*
.00
.68*
.04
.67*
.00
.69*
Speed
WM = Combined Complex Span and Updating
.08
.63*
Simple (A)
.04
.63*
Simple (A)
.72*
NA
.73*
NA
No WM
Simple (A)
WM = Complex Span
.03
.63*
Simple (A)
p<.01
.03
.63*
Simple (A)
Unique
Total
Age
.05
.10
.01
.09
.04
.14
.02
.11*
.05
.08
.01
.08*
.03
.09*
.01
.09*
Vocab
.20
.10
.03
.03
.07
.22
.06
.16
.18
.13
.04
.10
NA
NA
NA
NA
WM
Contextual analysis results on Connections performance with the addition of working memory variables
Table 5
Salthouse
Page 26
Note: Total age refers to the simple correlation of age and the variable. Unique age is the standardized regression coefficient for age when the reference cognitive abilities are included as simultaneous
predictors of the target variable as portrayed in Figure 1. NA means that the parameter was not estimated because it was not applicable in the model. Entries in the Gf, Mem (Memory), Speed, and Vocab
(Vocabulary) columns are standardized regression coefficients. All CFI values were equal to or greater than .95 and all RMSEA values were less than .07.
Salthouse
Page 27

What Cognitive Abilities Are Involved in Trail-Making Performance

Uploaded by

Copyright:

Available Formats

What Cognitive Abilities Are Involved in Trail-Making Performance

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

What Cognitive Abilities Are Involved in Trail-Making Performance

Uploaded by

Copyright:

Available Formats

NIH Public Access

NIH-PA Author Manuscript

Published in final edited form as: