INTRODUCTION
Neuropsychological investigations of mild TBI patients have yielded
mixed results, the interpretation of which is complicated by
methodological inconsistencies, including the use of different
cognitive tests, diagnostic criteria, injury-assessment intervals,
patient groups (e.g., consecutive hospital admissions, symptomatic
referrals), types of injury (e.g., sports concussion, motor vehicle
accidents, mixed causes), and levels of injury severity (i.e., mild
only, mild and moderate, mild to severe). Although deficits have been
found in speed of information processing, memory, attention and
executive functions, both the clinical significance and the etiology of
these deficits have been questioned (Binder,
1997; Binder et al., 1997).
With respect to speed, a number of information processing tasks have
been found to discriminate between mild TBI patients and controls
(Comerford et al., 2002; Gronwall & Sampson, 1974; MacFlynn et al., 1984; van
Zomeren, 1981; Waterloo et al., 1997),
and between subgroups of mild TBI patients (Waterloo
et al., 1997). While information processing speed usually
returns to normal by 1 to 6 months post injury (Gronwall, 1989; MacFlynn et al.,
1984), long-term reductions in mental speed have been found in
some mild TBI patients (Leininger et al.,
1990).
Subjective accounts of memory problems are also common, with Rimel et
al. (1981) reporting that 59% of mild TBI
patients complained of memory problems. However, objective assessments
of memory have provided mixed findings. Whereas Gentilini et al. (1985), Levin et al. (1987), and Mathias and Coats (1999) found no evidence to indicate visual or
verbal memory problems at 1 month post injury, other studies have shown
significantly poorer working memory (Newcombe et
al., 1994) and delayed recall (Dikmen et al.,
1986) at similar post-injury intervals. Problems with verbal
learning (Leininger et al., 1990) and delayed
recall (Stuss et al., 1985) have additionally
been reported in chronically affected mild TBI patients at 3 and 5
months post injury, respectively.
Deficits in selective and sustained attention have also been
demonstrated following mild TBI (Bigler &
Snyder, 1995; Gentilini et al., 1989;
Mclean et al., 1983; Stuss
et al., 1989). However, studies of executive functioning have
reported inconsistent findings. For example, deficits in verbal and
non-verbal fluency have been reported by some (e.g., Fos et al., 1995; Levin et al.,
1991; Mathias & Coats, 1999) but
not others (Leininger et al., 1990).
Similarly, tests of mental flexibility have reportedly differentiated
between mild TBI patients and controls in some (Dikmen et al., 1986; Fos et al.,
1995), but not all, studies (e.g., Leininger
et al., 1990).
When a meta-analysis of mild TBI research was completed by Zakzanis
et al. (1999) in order to weight and compare
the research findings using a common yardstick, namely effect size
(Cohen's d), the largest effects (≥1.0, averaged over
studies) for individual measures were reportedly for the Controlled
Oral Word Association Test (mean d = 1.22), the short delay
trial of the Rey Auditory Verbal Learning Test (mean d =
1.09), semantic fluency (mean d = 1.08), and the immediate
(mean d = 1.06) and delayed (mean d = .99) trials of
the Logical Memory subtest of the Revised Wechsler Memory Scale. Medium
effect sizes (≥.5; Cohen, 1992) were also
reported for a large number of other commonly used tests. The effect
sizes for these five measures convert to an overlap between the test
scores of the mild TBI and control groups of between 38% and 45% (Zakzanis, 2001). Moreover, in order to overcome a
possible bias arising from the tendency to publish positive results,
Zakzanis et al. (1999) calculated the
fail-safe N for each of these findings. This statistic
estimates the number of studies, with non-significant group
differences, that would be required in order to reverse the conclusion
that persons who have had a mild TBI perform more poorly. The fail-safe
N's for these measures were 363, 108, 107, 210, and 196,
respectively. Thus, even if there were a bias in the studies that have
been published or that underwent meta-analysis, it would take an
extremely large number of studies with opposing findings to negate the
importance of these measures in detecting cognitive deficits following
mild TBI. Finally, when Zakzanis et al. (1999) combined measures that assessed particular
cognitive domains, the largest effects sizes were found for cognitive
flexibility and abstraction (mean d = .72), delayed recall
(mean d = .71), memory acquisition (mean d = .69),
and attention and concentration (mean d = .63), indicating
that neuropsychological assessments of mild TBI patients should target
these cognitive functions.
Turning to the etiology of these problems, there is increasing
evidence that even mild TBIs can cause brain damage, suggesting that
physiological factors may contribute to some of the cognitive problems
and post-concussional symptoms that are experienced by these patients
(Guo et al., 2000; Plassman et al., 2000). Specifically,
neuropathological studies have reported DAI in samples of
mild-to-moderate and mild-to-severe TBI patients (e.g., Oppenheimer, 1968; Povlishock,
1992; Povlishock et al., 1992; Strich, 1961). However, with the exception of
Blumbergs et al. (1994; 1995), very few studies have focused on only
mild TBI. The Blumbergs et al. study examined the distribution of DAI
in 6 mild TBI patients who died of other causes, and found evidence of
damage to the CC (particularly the posterior one-third: G. Scott,
personal communication, February, 1997) and fornices in all cases,
suggesting that these pathways are most vulnerable to damage as a
consequence of TBI. The frontal and temporal lobes also showed evidence
of axonal injury. However, it is not known whether this damage was
extensive enough to disrupt cognitive functioning.
DAI, focal cortical lesions (particularly in the frontal and temporal
lobes) and whole brain atrophy have also been reported using magnetic
resonance imaging (MRI) with groups of mild, mild-to-moderate, and
mild-to-severe TBI patients (Bigler, 2001,
2003; Gale et al.,
1995; Gentry et al., 1988; Levin et al., 1987, 1991;
MacKenzie et al., 2002; Mittl et al., 1994). In addition, DAI in the corpus
callosum (CC), especially the posterior portion, and the fornices has
been reported using MRI (Gale et al., 1993,
1995; Gentry et al.,
1988; Tate & Bigler, 2000), with
some studies finding a relationship between the amount of damage and
injury severity (Gale et al., 1995; Gentry et al., 1988).
Although imaging has provided evidence of brain damage following mild
TBI, these techniques almost certainly underestimate the amount of
damage as both MRI and CT fail to detect pathology that is evident at
post-mortem, particularly non-hemorrhagic axonal injury (Jones et al., 1997). Indeed, mild TBIs do not
usually lead to abnormalities that are detectable on routine clinical
CT or MRI (Bigler & Snyder, 1995) due to
the limited resolution of current imaging techniques (Bigler, 2003), although significant progress has
been made (Garnett et al., 2000a, 2000b; Lewine et al., in press
a, in press b; McGowan et al., 2000; Sinson et
al., 2001). Thus, an absence of visible pathology does not
equate to an absence of abnormality, as there may be structural damage
that falls below the current threshold of detection (Bigler, 2001; Bigler &
Snyder, 1995). Given that injury severity is linearly related to
atrophy following moderate to severe injuries (Bigler, 2003), white matter pathology may also be
present following mild TBI, although to a lesser degree and with
lesions smaller than one mm3, which is currently the lower
limit of detection. If white matter integrity is diminished, this
disruption could affect speed of neural transmission and information
processing.
Normal information processing is dependent on intact neural
structures and functional pathways that subserve a particular cognitive
ability. Measures of information processing speed may therefore be the
most sensitive way of detecting any abnormalities within these systems
following mild TBI, particularly in the absence of any visible
pathology on neuroimaging. In mild TBI, the frontotemporal and callosal
regions of the brain and their functional pathways are most susceptible
to injury (Bigler, 2001; Blumbergs et al.,
1994, 1995).
Although information processing speed and callosal function are often
not systematically assessed in clinical settings, the functional
consequences of diffuse damage affecting the cerebral hemispheres and
corpus callosum can be operationalized and measured using speed of
information processing tasks developed in the cognitive psychology
literature.
The CC is the major fiber tract that connects the cerebral
hemispheres (Clarke & Zaidel, 1994), with
the anterior cerebral hemispheres being connected through the anterior
portion of the CC (rostrum, genu, anterior mid-body) and the posterior
areas being connected by the posterior CC (posterior mid-body,
splenium; de Lacoste et al., 1985; Pandya et al., 1971; Witelson,
1989). Although research examining the cognitive sequelae to
callosal damage following TBI is limited, there are reports of callosal
disconnection following severe TBI (Levander &
Sonesson, 1998; Rubens et al., 1977;
Vuilleumier & Assal, 1995), with patients
displaying gross deficits in their ability to transfer and integrate
information between hemispheres (e.g., apraxia, finger localization,
dichotic listening), consistent with the deficits shown by patients who
have had commissurotomies (e.g., Benavidez et al.,
1999; Bentin et al., 1984; Damasio et al., 1980; Geffen et
al., 1985). However, there do not appear to be any studies that
have used information processing measures in order to detect more
subtle deficits that may arise from less severe damage. Information
processing measures have, however, frequently been used to investigate
callosal functioning in normal samples, providing an experimental
paradigm for the examination of callosal functioning following mild TBI
(e.g., Banich, 1995; Banich & Karol, 1992; Banich
& Shenker, 1994; Hoptman & Davidson,
1994).
The present study was therefore designed to examine cognitive
performance following mild TBI, with an emphasis on the cognitive
functions that are mediated by those areas of the brain that are
susceptible to diffuse damage following TBI. To this end, the inter-
and intra-hemispheric processing of visual and tactile information
(i.e., reaction time; RT), selective and sustained attention, visual
and verbal fluency, and immediate and delayed verbal memory of a group
of mild TBI patients was assessed one month post-injury. It was
predicted that the mild TBI group would perform more poorly on these
cognitive tasks, that the mild TBI group would show reduced speeds of
information processing, particularly for the more difficult tasks, and
that tasks requiring the transfer of information across the CC would be
more affected in the mild TBI group than tasks requiring
intra-hemispheric processing.
METHODS
Research Participants
The TBI group consisted of 40 participants (8 females, 32 males),
aged between 18 and 60 years1
The
age range of participants was restricted to an upper limit of 60 years in
order to avoid the confounding effect caused by the normal slowing of reaction
times in people over 60 years of age (Stuss et al.,
1989).
(
M = 32.4,
SD = 12.7) who had sustained a mild TBI due to a
motor vehicle accident, assault, fall, or accident involving a blow to the
head, and who were initially attended to by ambulance paramedics. Mild TBIs
were defined as a Glasgow Coma Scale score (GCS) of 13 to 15 immediately
after their injury (
M = 14.7,
SD = .53), as measured
by the paramedics, and a loss of consciousness not exceeding 20 min.
The majority of patients (70%) had a GCS of 15 (27% had GCS = 14, 3%
had GCS = 13). Eight people (20%) reported no loss of consciousness, 18
(44%) were unconscious for less than 1 min, 9 (23%) were unconscious
for 1 to 5 min, and 5 (13%) were unconscious for more than 5 but less
than 20 min. All TBI participants were taken to the accident and
emergency department of a major hospital for examination, with 17 (43%)
being hospitalized (10 for observation, 6 for observation and treatment
of minor orthopedic injuries, 1 for a soft tissue injury to the
abdomen). Twelve patients underwent CT scanning and 2 had an MRI as
part of their routine clinical investigation, with no abnormalities
being visible on any of these scans. The mild TBI group had completed
an average of 12.4 years of education (
SD = 2.3) and had a
mean estimated premorbid IQ of 102.5 (
SD = 10.5). Eleven (28%)
were considering litigation at the time of their assessment. Scores on
the Impact of Events Scale (
Horowitz et al.,
1979) indicated that this group was experiencing only mild
levels of injury-related psychological distress at the time of their
assessment (
M = 22.8,
SD = 16.7, possible range:
0–65).
A control group from the general community (friends of the mild TBI
group and members of community groups), consisting of 40 participants
who had not previously sustained a head injury or experienced a loss of
consciousness, were individually matched, as closely as possible, to
the mild TBI group on the basis of age (M = 32.4, SD
= 12.7), gender (8 females, 32 males), education (M = 12.7,
SD = 2.1), and intelligence (M = 104.2, SD =
8.8). Participants were excluded from either group if they had any
history of neurological or psychiatric problems, if English was their
second language, or if they had any physical problem that would prevent
them from completing the RT tasks (e.g., sensory or motor problems).
Measures
Background information was collected using the Philadelphia
Head Injury Questionnaire (Curry et al.,
1991). The Alcohol Use Disorders Identification Test (AUDIT;
Saunders et al., 1993) was used to asses
alcohol consumption in order to ensure that the two groups
were comparable (maximum score: 40). The National Adult Reading Test
(NART; Nelson & Willison, 1991) was used
to estimate premorbid intelligence and the Impact of Events
Scale (IES; Horowitz et al., 1979) was used
to assess injury-related psychological distress in the mild
TBI group (15 items, response options: not at all,
rarely, sometimes, often, score range:
0–65).
The Visual Elevator, Telephone Search, and Telephone Search while
Counting subtests of the Test of Everyday Attention (TEA; Robertson et al., 1994) were used to assess
attention. The Visual Elevator task measures selective
attention and cognitive flexibility, requiring participants to
calculate which floor a person was on using a series of visually
presented elevators. Accuracy (range: 0–10) and timing (average
time to switch attention) scores were calculated for this measure. The
Telephone Search task measured selective attention, and required
participants to locate and circle symbols in a simulated telephone
directory, with the score being the average time taken to detect a
target. Finally, the Telephone Search While Counting subtest measured
sustained and divided attention by requiring participants to search a
simulated telephone directory for symbols, while also counting a string
of tones. Scores from this and the previous test were combined to
calculate a dual task decrement score.
Verbal fluency was assessed using the Controlled Oral Word
Association test (COWA; Spreen & Strauss,
1998), which requires participants to generate words beginning
with the letters F, A and S within three 1-min periods. The total
number of correct words generated for the three letters was calculated.
Non-verbal fluency was assessed using the Ruff Figural Fluency
Test (RFFT; Evans et al., 1985), which
consists of five sets of 40 squares, with each square containing a
five-dot matrix. Participants were given 1 min per set and instructed
to join two or more dots in order to make as many unique patterns as
possible, with the score being the total number of unique designs.
The Rey Auditory Verbal Learning Test (RAVLT; Schmidt, 1996; Spreen &
Strauss, 1998) was used to assess verbal memory (i.e.,
learning, delayed recall, recognition memory). The RAVLT involves the
verbal presentation of a list of 15 words over five trials, with
participants recalling as many words as possible after each trial. This
is followed by the single presentation and recall of a second list,
after which participants are required to recall the first list, both
immediately and after a 20 min delay, in addition to recognizing these
words from a printed list of 50 words. The total number of words
recalled in Trials 1–5, the immediate and delayed recall trials,
and the recognition trial, were calculated.
Information processing speed was assessed using visual and
tactile RT tasks requiring either a compatible response
(intra-hemispheric processing) or an incompatible response
(inter-hemispheric processing). These tasks were designed to satisfy
the methodological requirements for tasks that are designed to assess
interhemispheric processing, outlined by Banich and Shenker (1994). Each RT task commenced with 10 practice
trials, followed by 64 experimental trials, with equal numbers of
stimuli being presented to each side. Participants were instructed to
respond as quickly and as accurately as possible. The mean RT (and
SD) and number of correct responses for each individual on
each task were calculated, with the data from both hands being
combined. In RT paradigms, unusually fast or slow responses are
generally thought to reflect anticipatory responses (RTs of less than
20 ms) or lapses in attention (RTs greater than 2 SDs above an
individual's mean RT) that should be removed from the data. For
this reason, outliers were identified separately for each person on
each task using these criteria and excluded from further statistical
analysis. On average, 1–2 responses were removed using this
procedure, with no group differences in the number of outliers that
were removed.
The visual RT tasks were based on the assumption that, when focused
on a central fixation point, the input from the left visual field (left
of fixation) is transmitted to the right visual cortex, while
information in the right visual field is transmitted to the left visual
cortex. Four RT tasks, consisting of either a two- or four-choice
stimulus-response format and requiring either a compatible or
incompatible response, were used. The stimuli were presented on a 26-cm
color computer monitor for a duration of 150 ms (less than the time
taken to move the eyes from central fixation to a stimulus; Banich, 1997), with an interstimulus interval of
500 ms. Participants were seated 40 cm from screen and instructed to
focus on the fixation point which appeared on the screen 500 ms prior
to the stimulus onset. The experimenter, who sat behind the computer
screen, monitored central fixation. For the 2-choice tasks,
participants responded to one of two squares (2 cm × 2 cm each)
that appeared 6 cm on both sides of the central fixation point. A
compatible response involved pressing a key with the index finger of
the hand on the same side as the stimulus (solid white square). An
incompatible response involved making a response with the index finger
of the hand on the side opposite to that of the stimulus. For the
four-choice tasks, participants were required to respond to one of four
squares (2 cm × 2 cm each), with two on each side of the fixation
point. The compatible 4-choice task required participants to respond,
using the index (inner square) or middle (outer square) finger of the
hand on the same side as the stimulus. For the incompatible task,
participants were required to use the same finger as for the compatible
task but with the hand on the side opposite to that of the stimulus.
Four tactile RT tasks, involving compatible and incompatible
eight-choice and eight-choice, two-sequence vibro-tactile tasks, were
developed from the finger localization tasks used by Geffen et al.
(1985). The vibro-tactile keys consisted of
two sets of four keys arranged in a semi-circle to fit the shape of
each hand, excluding the thumb. Each key (1.5 cm in diameter) had a
vibrating stylus (3 mm) in the center that delivered the stimulus. The
key travel required to register a response was .6 mm and the pressure
required was 240 g. The vibrating stimuli were driven up and down
through the center of the response keys by modified miniature solenoids
and oscillated at a frequency of 108 Hz. For the eight-choice task,
participants were required to place the four fingers of both hands on
the response keys. Each stimulus involved a key vibrating for 250 ms
followed by an inter-stimulus interval of 500 ms, with each finger
being stimulated an equal number of times. Whereas participants were
required to respond to the vibrating key by pressing that same key for
the compatible task, a response with the equivalent finger of the
opposite hand was required for the incompatible task. As with visual
information, tactile information is transferred to the contralateral
hemisphere, which controls motor responses made on the same side as the
tactile stimuli. Thus, compatible responses required intra-hemispheric
processing and incompatible responses required the inter-hemispheric
transfer of information. Finally, for the eight-choice, two-sequence
tactile tasks, the stimuli consisted of two vibrating keys, one after
the other, with a 10 ms interval between the two. Participants had to
respond by pressing the two stimulus keys in the correct order for the
compatible task and by pressing the equivalent keys with the opposite
hand in the same order for the incompatible task. RTs for this task
measured the time between the onset of the first stimulus and
completion of the second response.
Procedure
Mild TBI patients were identified through ambulance records on a
prospective basis (not on the basis of symptomatology) and assessed
approximately 4 weeks post injury (M = 26.3 days, SD
= 6.1) during a single test session lasting 1½ to 2 hr.
Approximately 30% of those identified were either not contactable
(incomplete/incorrect address, not contactable by phone) or
declined to take part in the study. While this may have introduced a
selection bias, it is not possible to determine whether it favored
individuals with a better or worse outcome. Participants completed the
background questionnaire, followed by the measures of attention (TEA),
memory (RAVLT), visual and tactile RT, fluency (COWA, RFFT), premorbid
IQ (NART), and injury-related stress (IES). With the exception of
estimated premorbid IQ, raw scores were analyzed. All participants were
administered the Rey 15-item memory test to assess symptom validity
(Lezak, 1995; Schretlen et
al., 1991). The majority of TBI participants (97.5%) scored 11
or more on this measure, with only 1 participant (2.5%), who was not
involved in litigation, scoring 9. All control participants scored 11
or more. The two groups did not differ significantly in terms of their
Rey 15 scores (mean and frequency) and Rey 15 performance was related
to IQ (r = .44, p = .000). Thus, although the Rey
15-item test is one of the least sensitive measures of symptom
validity, these findings, combined with the normal RAVLT recognition
performance of this group (refer to Results) and the fact that this
assessment was undertaken entirely for research purposes (the results
were not available for the purposes of litigation), reduce the
likelihood of disingenuous performance by the mild TBI group.
RESULTS
Comparability of Matched Samples
The mild TBI and Control groups were compared in order to determine
whether they had been successfully matched. One-way ANOVAs revealed
that the two groups did not differ in terms of their age
[F(1,79) = .00, p = .993], alcohol usage
[i.e., AUDIT scores: F(1,72) = 1.257, p =
.266], years of education [F(1,79) = .425,
p = .516], or estimated premorbid IQ
[F(1,79) = .574, p = .451], indicating that
the two groups were comparable and consequently alleviating the need to
use any of these variables as covariates.
Cognitive Performance
Standard neuropsychological tests
The performance of the mild TBI and Control groups on the
neuropsychological tests was compared using one-way ANOVAs (refer to
Table 1 for means and standard deviations).
With respect to attention, the mild TBI groups differed from
the control group on the TEA Visual Elevator accuracy
[F(1,79) = 4.6, p = .035] and timing
[F(1,79) = 5.18, p = .026] scores, as well
as the time to detect targets in the Telephone Search task
[F(1,79) = 6.43, p = .013], with the mild
TBI group being less accurate and slower to switch attention when cued
to do so, while also taking longer to select relevant from irrelevant
target stimuli. In contrast, there was no difference between groups on
the divided attention task (TEA Telephone Search While Counting task),
indicating that the mild TBI group was not more adversely affected by
the need to divide their attention between two competing tasks than the
control group.
Means and standard deviations for the background measures and
neuropsychological tests
A comparison of the fluency scores of the mild TBI and
control groups yielded a significant difference on the test of
non-verbal or design fluency [F(1,79) = 9.96, p
= .002] but not for the test of verbal fluency (p =
.051). Finally, when the two groups were compared in terms of their
verbal memory performance, they were found to differ in terms
of their learning on the RAVLT [total of Trials 1–5:
F(1,79) = 5.74, p = .019], and on their
immediate [F(1,79) = 8.13, p = .006] and
delayed [F(1,79) = 7.76, p = .007] recall
of the word list, with the mild TBI group recalling significantly fewer
words on all of these measures. As expected, there was no difference
between the recognition memory performance of the mild TBI and control
groups.
Reaction time tasks
Three-way ANOVAs with repeated measures were used to assess
differences between the mean RTs of the mild TBI and control groups.
The visual and tactile data were analyzed separately. Task difficulty
(visual tasks: two- or four-choice; tactile tasks: eight-choice or
eight-choice, two-sequence) and compatibility (compatible or
incompatible stimulus–response format) were the within-subjects
factors, while group (mild TBI or control) was the between-subjects
factor. A summary of the mean visual and tactile RTs is given in Table 2, together with the results of t
tests comparing groups on individual tests, effect sizes, and
percentage overlap between groups.
Means (SDs), significance test results, effects sizes, and
percentage of overlap in the groups, for the visual and tactile RT
tasks
A repeated measures ANOVA comparing the mean RTs of the mild TBI and
control groups on the visual RT tasks found a significant
effect for group (refer to Tables 2
and 3 for means and ANOVA results). Thus,
mild TBI participants were slower to respond to these tasks than the
controls. Not unexpectedly, there were also highly significant main
effects for difficulty and compatibility, reflecting the fact that both
groups were slower on the more difficult four-choice task and were
slower to respond when the task required an incompatible response.
Consistent with our predictions, the interaction effects between Group
× Difficulty, and Group × Compatibility were significant,
indicating that the mild TBI group performed disproportionately slower
on the more difficult tasks and on the tasks requiring the
inter-hemispheric transfer of information, when compared to the control
group. These effects are illustrated in Figure
1a. The Difficulty × Compatibility interaction was also
significant, indicating that both groups were slower to make an
incompatible response to the more difficult visual RT task.
Results of the three-way ANOVAs performed on the visual and tactile
mean RT data
Mean RTs for the (a) visual and (b) tactile tasks, illustrating the
effects of Group and, in the case of the visual tasks, the Group
× Difficulty and Group × Compatibility interaction
effects.
A repeated measures ANOVA performed on the tactile RT data
revealed a significant group effect for the tactile RT tasks (refer to
Tables 2 and 3 for means and ANOVA results), with mild TBI
participants being slower to respond. The main effects for difficulty
and compatibility were also highly significant, indicating that both
groups were slower on the more difficult tactile task (eight-choice,
two-sequence task) and on tactile tasks requiring an incompatible
response. Unlike the visual RT tasks, the interaction effects between
Group × Difficulty, and Group × Compatibility were not
significant (see Figure 1b).
Both speed and accuracy can potentially be affected by TBI, making it
desirable to assess the influence of the independent variables (group,
difficulty, compatibility) on both speed (RTs) and accuracy (error
rates). However, the analysis of errors rates is problematic if RTs and
error rates are correlated with one another because participants may be
differentially trading off speed for accuracy in response to the
different tasks. Pearson correlation coefficients between the RTs and
error scores for each of the visual and tactile RT tasks, calculated
separately for the mild TBI and control groups, did not yield any
significant correlations. Two 3-way repeated measures ANOVAs (Group
× Difficulty × Compatibility) were, therefore, performed on
the error rates for the visual and tactile tasks. Both ANOVAs failed to
find significant effects for group, Group × Difficulty, and Group
× Compatibility for the visual and tactile tasks, suggesting that
mild TBI participants did not make more errors overall, or in response
to the more difficult tasks or tasks requiring an incompatible response
(refer to Table 2 for Ms and
SDs). Thus, accuracy of performance on the visual and tactile
RT tasks was not adversely affected in the mild TBI group.
Finally, effect sizes (Cohen's d) for the standard
cognitive tests and the RT tasks, together with the percentage overlap
between groups, were calculated using the procedures described by
Zakzanis (2001). Effect sizes for the
cognitive tests that differed significantly between groups ranged
between .43 and .71, with figural fluency having the greatest effect
size (i.e., largest deficit; refer to Table
1). Small effect sizes were found for the two-choice visual RT
tasks but there were medium to high-medium effect sizes for the
four-choice tasks (Cohen, 1992) (refer to
Table 2). Effect sizes for the tactile RT
tasks all fell within the medium range (.49–.66), as defined by
Cohen (1992). Interestingly, the mild TBI
group showed the greatest deficit on the four-choice compatible visual
RT task and the eight-choice, two-sequence compatible tactile RT task,
with approximately 57% of overlap in the scores of the TBI and control
groups.
DISCUSSION
The present study found that when mild TBI participants were assessed
approximately 4 weeks after their injury, they showed evidence of
problems in the speed and accuracy with which they were able to switch
attention, the speed with which they were able to select relevant from
irrelevant information, their non-verbal or design fluency, their
initial learning of verbal information, and their immediate and delayed
recall of verbal information. Moreover, effect size calculations
indicated that design fluency and delayed memory were most impaired in
this group.
In the case of attentional switching, the TBI group was slower and
less accurate than their non-injured peers. Speed and accuracy scores
were not significantly correlated (r = .19, p = .25),
suggesting that, in addition to reduced processing speed, the mild TBI
group demonstrated problems in the redeployment of their attention. The
mild TBI group was also slower when required to selectively focus on
stimuli embedded amongst other irrelevant stimuli. In contrast, divided
attention appeared to be intact. However, non-verbal fluency, the
initial learning of verbal material, and the immediate and delayed
recall of verbal information, were all impaired. The fact that the mild
TBI group showed normal recognition memory, suggests that the poorer
verbal recall of this group may reflect subtle problems in verbal
retrieval. However, once encoded, the information remains accessible,
provided external cues were given to assist retrieval.
The finding that mild TBI participants showed deficits in selective
and sustained attention is consistent with a number of studies that
have reported problems with attention (e.g., Gentilini et al., 1985, 1989; Mclean et al.,
1983; Stuss et al., 1989). The poorer
verbal memory of the mild TBI group also confirms the findings of
previous studies undertaken with mild TBI patients (e.g., Dikmen et al., 1986; Stuss et
al., 1985). Finally, the failure to find deficits in the verbal
fluency skills of the mild TBI group is consistent with the findings of
Leininger et al. (1990) but not with those of
Fos et al. (1995), Levin et al. (1991), and Mathias and Coats (1999). Non-verbal fluency, on the other hand, was
clearly compromised in the mild TBI group; a finding supported by Levin
et al. (1987, 1991)
in the acute stages after a mild TBI (around the time of the injury)
but not in the more chronic phases (i.e., 1–3 months
post injury). Importantly, the effects observed here accord very
well with those reported by Zakzanis et al. (1999) who noted, in their meta-analysis of mild TBI
research, that the domains with the greatest observed deficits are
cognitive flexibility (assessed here by the RFFT, COWA), followed by
delayed recall (RAVLT short and long delayed recall), memory
acquisition (RAVLT Trials 1–5), attention and concentration (TEA
subtests) and verbal ability (not assessed). With the exception of the
COWA, our effects sizes are similarly ranked and are consistent in size
with the mean values that are reported for these domains.
It could be argued that some caution should be exercised when
interpreting these findings, as multiple statistical comparisons were
made without correcting for the increased likelihood of making a Type I
error. More stringent levels of significance were not adopted because
it was thought that the consequences of making a Type II error were
equally serious (i.e., falsely concluding that mild TBI patients do not
have cognitive problems when they do). It is therefore important to
consider the internal consistency of the results. In this study, seven
of the nine predicted differences were significant, indicating a
coherent pattern of findings. Moreover, the effect sizes, which were
largely above .5 (medium effect size), enable the results to be
evaluated independently of statistical significance.
In addition to these tests, this study used RT tasks in order to
determine whether there were information processing deficits,
consistent with what would be predicted to occur as a consequence of
diffuse damage or disruption. Specifically, callosal transfer was
assessed by comparing intra- with inter-hemispheric processing, with
visual and tactile RT tasks being chosen in order to target the
posterior CC, as there is evidence to suggest this region is vulnerable
to diffuse damage following mild TBI (Blumbergs et al., 1994, 1995; Gale et al.,
1993, 1995; Gentry et al., 1988).
As predicted, the mild TBI group was slower on the visual tasks,
particularly when task difficulty was increased, as well as being more
impaired on the tasks requiring the transfer of information between
hemispheres. The results from the tactile RT tasks also provided some
evidence of a general slowing following mild TBI but this group was not
more adversely affected by the difficult or incompatible tasks.
Finally, there was no evidence to suggest that the TBI group
experienced problems in the accuracy with which they were able to
perform these tasks. Thus, it appears that RTs might provide a more
useful index of the functional consequences of mild TBI. Inaccurate
finger localization has, however, been found following commissurotomy
(e.g., Geffen et al., 1985) and severe TBI
(e.g., Levander & Sonesson, 1998),
suggesting that error scores may be more useful when there is evidence
of focal damage or more extensive damage to the CC. Although slower RTs
have previously been reported for TBI patients (e.g., MacFlynn et al., 1984; van
Zomeren, 1981; Waterloo et al., 1997),
there does not appear to have been any research that has attempted to
target inter-hemispheric transfer using RT tasks. Similarly, RT tasks
have not been widely used to target the functional consequences of the
diffuse damage and dysfunction following TBI, despite evidence that
they may be more sensitive to the effects of diffuse damage than
standard neuropsychological tests (Waterloo et al.,
1997).
The effect size calculations (Zakzanis,
2001) for the RT tasks showed that the group differences were
all medium or greater in size, which equates to 0.5 standard deviations
or more difference between the groups on these measures (Cohen, 1992). Moreover, these effects were
comparable to those obtained for the standard cognitive tests.
Interestingly, the largest effect size was for one of the visual RT
tasks (four-choice compatible).
The failure to find significant results suggestive of problems with
the inter-hemispheric transfer of tactile information may be
attributable to a number of factors. Firstly, it is possible that mild
injuries do not cause sufficient damage or disruption to interfere with
the inter-hemispheric transfer of tactile information. Blumbergs et al.
(1994, 1995) only
documented the distribution, and not the extent, of diffuse damage
following mild TBI. Given that the amount of DAI to the
CC has been found to be related to injury severity (Gale et al., 1995; Gentry et
al., 1988), it may be that only moderate and/or severe TBI
patients demonstrate deficits in the inter-hemispheric transfer of
tactile information. Alternatively, it may be that our measures were
not sensitive enough. Finally, the mild TBI group may have sustained
reversible axonal injury (Gentleman et al.,
1995; Sherriff et al., 1994) or
transient biochemical changes. This raises the additional possibility
that the deficits found in the mild TBI group may partially or
completely resolve over time. Indeed, this may explain why studies
often report a higher frequency of problems in the early stages after a
mild TBI than in the later stages (Kibby & Long,
1996; Lishman, 1988; Van der Naalt et al., 1999). It remains to be seen
whether measures designed to target the effects of diffuse damage are
better able to predict which mild TBI patients will experience
long-term residual problems than traditional neuropsychological
tests.
Finally, there were a number of problems associated with the RT tasks
that need to be addressed before undertaking further research. For
example, some participants admitted responding to the non-stimulus
(i.e., the outline of a blank square rather than the solid square) for
the incompatible visual RT tasks, turning the incompatible task into a
compatible task for these participants. This could be overcome by
keeping the same stimulus but having the outline of the squares present
at all times. Participants also reported having difficulty in
responding to the tactile stimuli, especially the eight-choice,
two-sequence tactile RT task, due to problems with the dexterity and
sensitivity of the little finger. A simpler task, using fewer stimuli
and longer inter-stimulus intervals, may better discriminate between
groups. Lastly, while the incompatible RT tasks were designed to target
inter-hemispheric transfer, it is also conceivable that slower
performance reflected compromised executive functioning as this
condition required participants to inhibit a more automatic response
(response compatible with the stimulus) and make a more deliberate
response (response not compatible with the stimulus).
In addition, there are a number of cautionary notes that should be
considered when interpreting the data. Although the data (Rey 15-item
and RAVLT recognition memory performance) suggests that disingenuous
performance was not a problem in this TBI sample, the Rey 15-item is
one of the least sensitive screening measures, and so exaggerated
symptomatology cannot be completely ruled out. However, the fact that
testing was completed entirely for research purposes and that the one
person who scored 9 was not involved in litigation, reduces some of the
motivation to perform sub-optimally. The fact that 30% of the mild TBI
patients identified for this study were not contactable or declined to
participate is also worthy of consideration. Although recruitment rates
are rarely provided in mild TBI research, higher participation rates
are unlikely given the nature of the injury and the age group involved.
Moreover, it is not possible to determine whether this results in a
sample that has a better or worse outcome than that of the
non-participants. Finally, the fact that healthy controls rather than
medical controls were used in this study may increase the size of the
group differences. While healthy controls were chosen because most test
norms are based on healthy samples and it was our intention to
determine whether there were cognitive differences that would be
detectable in a clinical situation, this research needs to be
replicated with medical controls in order to determine the effects of
general injuries, and any associated stressors, on cognitive
functioning. If injury-related psychological factors are contributing
to the current results, group differences are likely to be less using
medical controls.
In conclusion, the present study suggests that in the early stages
after a mild TBI, patients experienced problems with selective
attention (speed and accuracy), non-verbal fluency, the initial
learning and free recall of verbal information, the speed with which
they were able to process visual and tactile information, and with
visual tasks requiring the inter-hemispheric transfer of information.
These deficits, including slower performance on two attention tasks,
fewer designs on the non-verbal fluency task (also a timed task),
slower initial encoding on a verbal memory task, and slower visual and
tactile RT scores, are consistent with what would be expected to occur
as a result of disruptions to integrated white matter pathways. In
addition, deficits in the visual RT tasks requiring the
inter-hemispheric transfer of information may reflect damage or
disruption to callosal pathways.
The possibility that the current battery of measures, especially the
RT measures, provide functional assessments of diffuse damage or
dysfunction now needs to be evaluated by replicating this study with
the addition of MRI or other types of neuroimaging. Significant
advancements in detecting subtle white matter abnormalities have
recently been made (Adelson et al., 2001;
Arfanakis et al., 2002; Hofman et al., 2002) and should be investigated in
the context of RT and mild TBI. Additional research, using samples of
moderate and severe TBI patients, medical controls, and RT tasks that
incorporate the above-mentioned modifications, is also needed in order
to extend our understanding of the functional consequences of diffuse
damage caused by TBIs. Moreover, research is needed to determine the
permanency of these deficits and the extent to which RT measures
predict long-term outcome.
ACKNOWLEDGMENTS
The authors would like to acknowledge the Medical Director, Dr. Hugh
Grantham, and the staff of the South Australian Ambulance Service, for
their assistance with this project. This study was partly supported by
grants from the Faculty of Health Sciences at the University of
Adelaide (grant to J. Mathias) and the Ira Fulton Foundation (grant to
E. Bigler). Preliminary findings from this study were presented at the
conference Neuropsychological Assessment for Research and Clinical
Practice, London, October 1999.
