Journal of Clinical and Experimental Neuropsychology, 2014

Vol. 36, No. 10, 1066–1075, http://dx.doi.org/10.1080/13803395.2014.971718

Cognitive correlates of under-ambiguity and under-risk

decision making in high-functioning patients with
relapsing remitting multiple sclerosis
Martina Gaia Cogo1, Stefania Rota2, Maria Letizia Fusco1, Cristina Mapelli3,
Francesca Ferri3, Ildebrando Marco Appollonio3, and Valeria Isella3
1
Neuroimmunology Clinic, Department of Neurology, S. Gerardo Hospital, University of Milan
Bicocca, Monza, Italy
2
Department of Neurology, Saronno Hospital, Saronno, Italy
3
Neuropsychology Section, Department of Neurology, S. Gerardo Hospital, University of Milan
Bicocca, Monza, Italy

(Received 23 April 2014; accepted 29 September 2014)

Introduction: Impairment of decision making in relapsing remitting multiple sclerosis is still controversial, and its
neuropsychological correlates have never been explored thoroughly, especially in patients with minimal physical
and cognitive deficits. In the present study we investigated the cognitive underpinnings of decision making under
ambiguous and explicit conditions in patients with very mild relapsing remitting multiple sclerosis, using a dice
and a card gambling game. Method: The study sample included 60 patients and 35 healthy subjects. In the Game
of Dice Task, winning and losing probabilities are obvious to the subject, while in the Iowa Gambling Task they
are initially ambiguous and have to be gradually identified. Performance at the two tasks was correlated with
scores obtained at tests investigating cognitive processing speed, memory, language and executive functions.
Results: Patients’ performance did not differ from that of controls at either gambling task. There was only a trend
for them to be significantly slower than healthy subjects in progressively recognizing advantageous decks in the
Iowa Gambling Task. While the Game of Dice was unrelated to neuropsychological tests, predictors of perfor-
mance at the Iowa task were Letter Fluency and the Symbol Digit Modalities Test for the initial, under-
ambiguity, trials and the Wisconsin Card Sorting Test for the last, purely under-risk, trials. Conclusions: Our
results suggest that high-functioning patients with relapsing remitting multiple sclerosis are substantially capable
of making advantageous decisions, even if they may be slower in processing options and shifting strategy when
selection criteria are not explicit.

Keywords: Iowa Gambling Task; Game of Dice Task; Decision making; Multiple sclerosis; Executive functions.

Multiple sclerosis (MS) is a disabling disorder typi-
cally affecting individuals in their early and middle
adulthood. Neuroimaging studies have highlighted
the frequent involvement of frontal lobes and cir-
cuits in MS (Filippi & Rocca, 2010), leading to
deficits of executive abilities in these patients
(Chiaravalloti & De Luca, 2008). One such ability
is decision making. At its most basic level, decision
making is the capacity to select an option from a
number of alternatives through the contrast of
risks and benefits. In subjects with MS, and espe-
cially in those who are socially and professionally

The research was conducted at the Department of Neurology, S. Gerardo Hospital, University of Milan Bicocca, Via Pergolesi 33,
20900, Monza (MB), Italy.
We would like to thank Maura Frigo and Margherita Gardinetti for their expert help with the neurological assessment of multiple
sclerosis patients.
Funding: None. Conflicts of interest: None.
Address correspondence to: Valeria Isella, Neuropsychology Section, Department of Neurology, S. Gerardo Hospital, University of
Milan Bicocca, Via Pergolesi 33, 20900, Monza (MB), Italy (E‑mail: valeria.isella@unimib.it).

© 2014 Taylor & Francis


active, the inability to make proficient personal
and career choices might have a very heavy
impact. In addition, these patients are confronted
with complex health-related issues concerning
diagnostic procedures and treatments that they
need to be able to manage competently.
One of the most classical tools for measuring
decision ability is Bechara et al.’s Iowa Gambling
Task (IGT; Bechara, Damasio, Damasio, &
Anderson, 1994), a card game mimicking real-life
decision processes. In the IGT, 100 cards have to
be chosen from four decks, each associated with a
different display of losses and gains of facsimile
money. Displays are not made obvious to the sub-
ject, who should gradually identify nonrisky decks
and shift to them in order to increase the profit.
Another popular gambling task, the Game of Dice
Task (GDT; Brand et al., 2005), is a virtual dice
game in which the subject has to guess which
number, or combination of numbers, contains the
number that will be thrown next.
This test has been proposed as a measure of
decision making in a different context from the
IGT, because information on winning and losing
probabilities and on amount of gains and losses
associated with selection options are explicitly pro-
vided at each trial. This decision-making scenario
has been defined by Brand, Labudda, and
Markowitsch (2006) as under risk. The entire pre-
frontal cortex would be engaged in this decision
condition: The orbital and ventromedial regions
would mediate the influence of somatic markers,
while the dorsolateral areas would allow a rational
analysis of the scenario, which requires flexibility
and monitoring abilities. In the alternative decision
context, called under ambiguity (Brand et al.,
2006), rewards and penalties associated with
options are not made explicit and must be identi-
fied through an efficient feedback processing that
would rely on the activity of orbital and ventrome-
dial regions of the prefrontal cortex. According to
Brand et al. (2006), the initial phase of the IGT
would take place in this second type of scenario,
while the final one would evolve under risk.
Both the GDT and the IGT have been widely
used to assess decision making in a variety of
psychiatric and neurological disorders, and beha-
vioral and neuroimaging studies generally support
their correlation with frontal lobe functioning
(Brand et al., 2004, 2006; Buelow & Suhr, 2009;
Delazer, Sinz, Zamarian, & Benke, 2007). For
what concerns MS, a very recent study by Farez,
Crivelli, Leiguarda, and Correale (2014) adminis-
tered both tests to 27 subjects with mild relapsing
remitting (RR) MS and 27 healthy controls.
Patients showed a poorer performance on the
DECISION MAKING IN MULTIPLE SCLEROSIS 1067
GDT and on the final IGT trials, supporting an
impairment of the ability to decide advantageously
in risky conditions. Literature dealing with the
IGT in MS is richer. The most numerous sample
of patients was analyzed by Simioni et al. (2008),
who assessed over hundred, minimally impaired,
subjects with RR MS and found no difference in
total IGT score in comparison with a normal con-
trol group. The same authors also carried out a
longitudinal study on 70 patients, who were tested
with the IGT at the baseline and two years later on
average (Simioni et al., 2009). Performance at the
test was shown to be declined at follow-up. A
smaller pool of 20 MS individuals was studied by
Kleeberg et al. (2004), with an opposite result:
Patients showed impaired learning of the winning
strategy across the task compared with healthy
subjects. In line with these data, Nagy et al.
(2006) found an impairment in 21 MS cases both
at the original IGT and at an alternative version of
the test whereby advantageous decks are character-
ized by immediate large punishment and larger
future reward (instead of immediate small reward
and smaller future punishment). The authors con-
cluded that decision-making deficits would be due
to difficulties in evaluating long-term outcomes in
MS, rather than to increased sensitivity to reward.
Finally, Roca et al. (2008) showed a poor perfor-
mance at the IGT in 12 individuals with MS.
Interestingly, in their patients Roca et al. (2008)
also correlated behavioral data with the severity of
white matter damage in frontosubcortical fiber tracts
measured with diffusion tensor imaging, but found
no significant relationship. Neuropsychological stu-
dies have also failed to show a correlation between
performance at the IGT and frontal dysfunction in
MS (Farez et al., 2014; Nagy et al., 2006; Simioni
et al., 2008). Only Kleeberg et al. (2004) found a
significant relationship with Part A of the Trail
Making Test (TMT). These data are in disagreement
with evidence in favor of a frontal-executive sub-
strate for the IGT gathered in other healthy and
clinical populations (Buelow & Suhr, 2009) and
remain unexplained. A possible account might be
that no published study has ever explored thoroughly
the cognitive underpinnings of performance at IGT,
and GDT, in patients with RR MS. This was the
main aim of our work. In particular, we hypothesized
that, taking into account multiple neuropsychologi-
cal measures, a significant relationship would emerge
between decision-making competency and executive
functions. In addition, we wanted to further verify
the presence of decision-making impairment in MS
in a large sample with very mild neurological dis-
ability and no, or minor, cognitive deficits, extend-
ing the investigation to decision under risk with the
1068 COGO ET AL.

GDT and with a separate analysis of the first
(under-ambiguity) and last (under-risk) sections
of the IGT.
METHOD
Subjects
Participants were selected from the neuroimmunol-
ogy clinic of S. Gerardo Hospital, Monza, Italy,
between January 2010 and January 2013. Inclusion
criteria were a diagnosis of possible or definite MS
according to McDonald et al.’s (2001) criteria,
with a relapsing remitting course, and an
Expanded Disability Status Score (EDSS;
Kurtzke, 1983) ≤3.5. Subjects were excluded if
they obtained an abnormal (i.e., below the cutoff
provided by norms) age- and education-adjusted
score at more than two neuropsychological tests
included in the study protocol. Other exclusion
criteria were a current or past history of patholo-
gical gambling, substance abuse, or major psychia-
tric disorders, including severe depression defined
by a score >30 at the Beck Depression Inventory
(BDI; Beck, Ward, Mendelson, Mock, & Erbaugh,
1961). Normal controls were volunteers from the
local community, who were comparable to patients
for sociodemographic features and intelligence
quotient (IQ) as measured with the Italian version
of the Wechsler Adult Intelligence Scale–Revised
(WAIS–R; Orsini & Laicardi, 1997). They were
also screened for gambling history. All participants
were informed of the study aim and design and
signed a consent approved by the hospital Ethics
Committee according to the Declaration of
Helsinki (1961). The identity of patients and con-
trols was kept secret through identification codes
and prevention of unauthorized access to the
records.
After the exclusion of seven patients with more
than two abnormal tests at the general neuropsy-
chological assessment, the final RR MS group
included 60 subjects, 20 (33.3%) men and 40
women, with a mean age of 34.0 years (SD =
7.7), a mean education of 13.1 years (SD = 3.1),
and a mean total IQ of 114.0 (SD = 9.6). Healthy
controls were 35 subjects, 11 (31.4%) men, and 24
women, with a mean age of 32.5 years (SD = 8.6),
a mean education of 12.5 years (SD = 2.5), and a
mean total IQ of 116.3 (SD = 6.2). There was no
statistically significant difference between the two
groups for sociodemographic and psychometric
features.
Neurological and neuropsychological characteris-
tics of patients are shown in Table 1. Overall, they
were neurologically and cognitively minimally
impaired. Fifty out of 60 (83.3%) had an EDSS
score ≤2.0, and 37 (61.7%) obtained a normal score
at all neuropsychological tests. Only five patients
(8.3%) were depressed at the BDI (four mildly, one
moderately).

TABLE 1
Neurological and neuropsychological features of multiple sclerosis patients

Variables Mean ± SD Min–max Cutoff for normality

Disease duration (months) 40.8 ± 45.0 2–264 —

Expanded Disability Status Scale
Mean 1.4 ± 1.0 0–3.5 —
Median 1.5 —
Number of relapses 1.5 ± 0.9 0–4 —
Rey Auditory Verbal Learning Test
Immediate Recall 40.3 ± 8.7 13.5–55.8 ≥28.5
Delayed Recall 10.7 ± 2.9 3.0–15.0 ≥4.7
Symbol Digit Modalities Test 54.1 ± 11.7 33.0–82.4 ≥37.9
Letter Fluency 31.9 ± 10.2 10.0–55.0 ≥17
Category Fluency 41.9 ± 8.7 26.0–61.0 ≥25
Wisconsin Card Sorting Test
Global scorea 44.5 ± 30.0 11.2–128 ≤90.5
Perseverative errors 15.7 ± 8.1 2.9–44.1 ≤42.6
Nonperseverative errors 14.6 ± 10.7 2.6–57.6 ≤29.9
Set maintenance 1.0 ± 1.4 0–5 ≤3
Raven Colored Progressive Matrices 31.7 ± 2.7 25.5–36.0 ≥19
Beck Depression Inventory 7.3 ± 6.2 0–25 ≥14

Notes. All test scores are adjusted for age, sex, and education.
a
Overall index of performance, which estimates the number of cards used by the subject in excess of the minimum necessary to
achieve the six categories; it is computed by subtracting from the number of administered trials the number of categories completed,
multiplied by 10 (i.e., by the number of correct responses required for each category; Laiacona et al., 2000).

General neuropsychological assessment
The general neuropsychological battery was focused
on executive functions, including the Symbol Digit
Modalities Test (SDMT) from the WAIS–R
(Wechsler, 1981), Letter Fluency (Novelli et al.,
1986), Raven’s Colored Progressive Matrices
(RCPM; Basso, Capitani, & Laiacona, 1987), and
the Wisconsin Card Sorting Test (WCST; Laiacona,
Inzaghi, De Tanti, & Capitani, 2000). In addition, we
evaluated episodic memory, with the Rey Auditory
Verbal Learning Test (RAVLT; Caltagirone et al.,
1995), and semantic memory and language, with
Category Fluency (Novelli et al., 1986).
Patients underwent global cognitive assessment
and the two gambling tasks at least six weeks from
a relapse or from corticosteroid treatment, usually
in a couple of sessions.
Decision-making tasks
The IGT was used in a computerized version that was
administered according to the procedure suggested by
Bechara et al. (1994). In brief, the subject sits in front
of the computer screen where the images of four decks
of cards (A to D) are shown and is informed that the
aim of the task is to earn as much (virtual) money as
possible, starting from an initial capital of 2000 €. The
“player” is then asked to select one card from any of
the four decks, using the mouse. After every choice, he
is informed of the exact amount of money lost or
gained. The initial cards from Decks A and B provide
high money gains, while subsequent cards are asso-
ciated with heavy losses, resulting in a long-term
financial disadvantage. On the contrary, cards from
Decks C and D provide lower initial wins than cards
from the previous two decks, but are also associated
with lower later debits, so that their final outcome is
advantageous. After the initial, exploratory trials, the
expected behavior is for the subject to learn from
positive and negative visual and auditory feedbacks,
identify the nonrisky decks, and shift to them. The
total number of trials is 100. As an index of perfor-
mance we computed the net number of advantageous
choices, obtained by subtracting cards selected from
Decks A and B from cards selected from Decks C and
D. This parameter was obtained (a) for all 100 trials,
(b) for five consecutive blocks, each including 20
trials, and (c) for the initial, under-ambiguity, trials
and the final, under-risk, trials (Brand et al., 2006).
The cut point between the under-ambiguity and
under-risk sections of the test is not fixed (Brand
et al., 2006); we derived it empirically, based on the
profile of our patients’ actual performance. We also
calculated the so-called “learning index,”
DECISION MAKING IN MULTIPLE SCLEROSIS 1069
corresponding to the difference between the mean
score (advantageous – disadvantageous choices) at
the first two blocks and the mean score at the last
three blocks.
The GDT (Brand et al., 2005) is also a computer-
ized gambling task. The subject is provided with a
starting capital of (fictitious) 1000 € and is asked to
“throw” a virtual dice shown on the computer screen
with a mouse click, for a total of 18 trials. Before
each throw, the player has to guess which number
will appear and will be rewarded with money for
every correct answer. He or she is given the opportu-
nity to make only one guess—that is, provide a single
number—or to increase the chances to guess cor-
rectly by indicating two, three, or four numbers.
The single number option is rewarded with a larger
amount of money, while multiple number combina-
tions are associated with smaller gains. Winning and
losing probabilities are visualized on the computer
screen for the whole task duration, and visual and
auditory feedbacks are provided after each throw.
Measures of performance are the number of riskiest
choices (i.e., the single number option) and a net
score obtained subtracting risky decisions (i.e.,
choices from the single number and the two-numbers
set) from less risky decisions (i.e., choices from the
three- and four-number sets).
Statistical analysis
Statistical analysis was conducted with PASW statis-
tics 21.0 (SPSS, Inc., 2009, Chicago, IL, www.spss.
com). Sociodemographic characteristics of patients
and controls were compared by means of Student’s t
test or chi-square analysis, as appropriate. Student’s t
test was used for intergroup comparison of scores at
the two gambling tasks. For the GDT, we compared
net score and single-number score. For the IGT, we
compared learning index and net number of advan-
tageous choices for all 100 trials, for the first 60
(under-ambiguity score) and for the last 40 (under-
risk score). Progression of advantageous choices
throughout the IGT was also analyzed, using
repeated measures analysis of variance (ANOVA),
with net number of advantageous choices at subse-
quent blocks of cards as within-subjects factor and
group as between-subjects factor. This analysis was
carried out for all the five 20-trial blocks and for the
under-ambiguity and under-risk sections of the test.
Level of significance was set at p < .05. Pearson’s r or
Spearman’s rho coefficients were calculated to assess
correlations for normally and non-normally distrib-
uted data, respectively; correlations were measured
in the MS group, between GDT score and IGT
parameters (total score, score for Trials 1–60 and
1070 COGO ET AL.

61–100, and learning index) and the following neu-
ropsychological measures: SDMT, RAVLT, Letter
and Semantic Fluency, RCPM, WCST. Bonferroni
correction for multiple tests was applied, yielding a
level of significance of p < .001. Linear regression
was performed with gambling tasks scores as depen-
dent variable and cognitive measures as predictors.
RESULTS
Comparison of scores obtained by patients and con-
trols at the two gambling tasks is shown in Table 2.
There was no statistically significant difference in
the mean net scores (whose distribution in the two
groups is shown in Figure 1, left) and single number
scores at the GDT, nor in any of the IGT para-
meters: mean net number of advantageous choices
in all 100 trials (Figure 1, right) and in each 20-trial
block, and mean learning index.
In a repeated measures ANOVA performed on the
IGT (Figure 2), there was a significant main effect of
block, F(4, 372) = 11.845, p = .000, η2p = .113, due to
a progressively higher number of advantageous
choices. There was no significant main effect of
group, nor a Block × Group interaction. At

TABLE 2
Comparison of mean scores obtained at the Iowa Gambling Task and at the Game of Dice Task by patients with multiple sclerosis
and healthy controls

Patients (n = 60) Controls (n = 35)

Task Mean ± SD Mean ± SD t

Game of Dice Task

Net score (nonrisky – risky choices) 9.5 ± 5.2 8.1 ± 4.3 1.467
Number of choices of single number 1.9 ± 0.9 1.3 ± 0.6 1.004
Iowa Gambling Task (net number of advantageous choices)
All 100 trials 6.4 ± 21.2 12.9 ± 20.8 –1.470
Trials 1–20 –1.10 ± 5.4 –2.63 ± 5.9 1.256
Trials 21–40 0.33 ± 5.5 2.46 ± 5.8 –1.747
Trials 41–60 0.83 ± 7.1 3.66 ± 6.7 –1.907
Trials 61–80 3.57 ± 8.4 5.03 ± 9.1 –0.778
Trials 81–100 2.73 ± 8.7 4.40 ± 8.3 –0.926
Trials 1–60 0.07 ± 12.8 3.49 ± 12.7 –1.258
Trials 61–100 6.30 ± 13.4 9.43 ± 14.3 –1.053
Learning index 3.07 ± 5.7 3.75 ± 6.9 –0.499

Figure 1. Scatterplot of net score at the Game of Dice Task (left) and total score (net number of advantageous choices in 100 trials) at
the Iowa Gambling Task (right) obtained by multiple sclerosis patients and normal controls.
 DECISION MAKING IN MULTIPLE SCLEROSIS 1071

Group interaction, F(2, 186) = 4.177, p = .017, η2p =

Figure 2. Comparison between multiple sclerosis patients and nor-
mal controls for mean net number of advantageous choices [(C + D)
– (A + B)] at the Iowa Gambling Task. The first 60 trials correspond
to decision making under ambiguity, the last 40 to decision making
under risk.
inspection, the performance curve of MS patients
showed a clear-cut increase in steepness starting
from around the 60th trial. Therefore we performed
two repeated measures ANOVAs for the first three
blocks of cards (Trials 1–20, 21–40, and 41–60), as an
index of under-ambiguity decision making, and for
the last two blocks (Trials 61–80 and 81–100), as an
index of under-risk decision making. For Cards 1–60
there were a significant main effect of block, F(2,
186) = 14.394, p = .000, η2p = .134, again for a
gradual increase in advantageous choices, no signifi-
cant main effect of group, and a significant Block ×
.043: Compared with normal controls, MS patients
chose a significantly lower number of advantageous
cards in the second block, F(1, 93) = 5.796, p = .018,
η2p = .059. For Cards 61–100 there were no signifi-
cant main effects of block or group, nor a significant
Block × Group interaction. Means comparison
between patients and controls was also performed
for scores at the first 60 trials and at the last 40, but
no statistically significant difference emerged
(Table 2).
Correlation with neuropsychological tests was
carried out in the MS group for the two GDT
scores, and for all 100 trials, Trials 1–60 and 61–
100, and the learning index of the IGT. Results are
shown in Table 3.
The GDT did not correlate significantly with
any of the variables.
After Bonferroni correction, the net number of
advantageous choices at IGT 100 trials correlated
significantly with letter fluency and with the
WCST global score. The subscore from Trials 1–
60 correlated with letter fluency and SDMT, and
the one from Trials 61–100 correlated with the
WCST global score. The learning index did not
correlate significantly with any neuropsychological
variable. Correlation analysis was also performed
between the GDT and IGT scores, but we found
no statistically significant relationship (data not
shown).
Neuropsychological tests showing significant
correlation coefficients with the IGT—that is,
the WCST global score, letter fluency, and the
SDMT—were entered into a linear regression
analysis as possible independent predictors of per-
formance at IGT Trials 1–60 or 61–100. For the

TABLE 3
Results of correlation analysis between scores at the two gambling tasks and cognitive measures, in the group of patients with
multiple sclerosis

Game of Dice Task Iowa Gambling Task

Variables Net score Single number All trials Trials 1–60 Trials 61–100 Learning index

Symbol Digit Modalities Test –.06 .01 .36 .38* .20 .07
RAVLT
Immediate Recall –.29 .10 .21 .18 .10 .11
Delayed Recall –.31 .15 .28 .31 .14 .14
Letter Fluency –.11 .08 .39* .42* .22 .14
Category Fluency .13 –.03 .35 .33 .24 .13
Wisconsin Card Sorting Test
Global score .04 .00 –.39* –.21 –.41* –.27
Perseverative errors .09 –.05 –.25 –.15 –.27 –.21
Nonperseverative errors .04 –.01 –.20 –.09 –.17 .00
Set maintenance –.01 .01 –.28 –.18 –.31 –.26
RCPM .04 –.01 .19 .24 .08 .07

Notes. RAVLT = Rey Auditory Verbal Learning Test; RCPM = Raven Colored Progressive Matrices.
*p < .001.
1072 COGO ET AL.
first blocks, two predictors explained 46% of the
variance (R2 = .21), F(2, 59) = 7.658, p = .001:
letter fluency (β = 0.31, p = .016) and the SDMT
(β = 0.26, p = .04). For the last blocks, the WCST
global score (β = –0.30, p = .02) explained 30% of
the variance (R2 = .09), F(1, 59) = 5.673, p = .021.
DISCUSSION
In the present study we investigated cognitive cor-
relates of decision making under ambiguity and
under risk, as measured with the GDT and the
IGT, in a large sample of minimally impaired
individuals with RR MS, compared with healthy
controls matched for sociodemographic features
and IQ score.
For the IGT, correlation and regression analyses
showed that the first 60 and last 40 trials seem to
engage distinct cognitive functions. Initially, in the
stage of advantageous decks identification, perfor-
mance was correlated with letter fluency and the
SDMT, suggesting that it could rely on the ability
to elaborate a strategy and quickly process card–
outcome associations. In the subsequent blocks,
when maintenance of the profitable tactic was
required for a positive outcome, a correlation
emerged with the WCST global score, supporting
an influence on performance of the ability to not
perseverate on disadvantageous selections and
“maintain the set” of advantageous ones. These
findings fit well with the under ambiguity–under
risk distinction proposed by Brand et al. (2006)
and with their suggestion that, unlike the early
sections of the IGT, the final ones might tap into
the risky, rather than the ambiguous, decision con-
text. The first 60 choices, mostly performed under
ambiguity, were associated with the ability to read-
ily identify associations between cards and gains/
losses, while the last 40, carried out purely under
risk, were correlated with shifting and set mainte-
nance abilities. A recent review on cognitive corre-
lates of the IGT in healthy and clinical populations
demonstrated that the test is mostly unrelated with
executive functions (Toplak, Sorge, Benoit, West,
& Stanovich, 2010), but in studies where an asso-
ciation was detected, it was usually with card sort-
ing tests or with the TMT. Further, in a sample of
healthy individuals, Brand, Recknor, Grabenhorst,
and Bechara (2007) were able to demonstrate a
significant relationship between the last blocks of
IGT and the WCST. Our study is in line with these
reports, having detected the expected association
between the IGT and executive dysfunction in MS
that past studies had not highlighted (Farez et al.,
2014; Nagy et al., 2006; Roca et al., 2008; Simioni
et al., 2008). The significant correlation between
the first part of the IGT and letter fluency and
SDMT has also an intriguing clinical implication:
A 60-card version of the test might be used for a
sensitive neuropsychological screening of mini-
mally compromised patients with MS. As a matter
of fact, reduced phonological fluency and cognitive
slowing are the earliest neuropsychological deficits
in MS (Chiaravalloti & De Luca, 2008; Henry &
Beatty, 2006).
We would have expected an association with the
WCST also for the GDT, which is considered a
task of decision under risk and has been shown to
correlate with measures of categorization, set shift-
ing, and attention in neurological patients and
normal individuals (Brand et al., 2004, 2005,
2006; Schiebener, Zamarian, Delazer, & Brand,
2011). Furthermore, a very recent study by
Muhlert et al. (2014) investigated grey and white
matter magnetic resonance imaging (MRI)
abnormalities associated with impairment in a dif-
ferent test of decision under risk, the Cambridge
Gambling Task (CGT; Rogers et al., 1999), in
patients with MS. They did demonstrate a signifi-
cant correlation with frontal, prefrontal, and ante-
rior cingulate cortex. On the other hand, the only
other published study dealing with the GDT in MS
(Farez et al., 2014) had results that were similar to
ours, as they also failed to find an association with
executive functions. The cognitive underpinnings
of risky behavior in MS are therefore still an
open issue that needs further investigation with
neuroimaging and neuropsychological techniques.
In our MS patients, the ability to follow success-
ful rules in the IGT, once they had been figured
out, seemed to be preserved, and their overall per-
formance was in fact overlapping with the control
group. Accordingly, they also performed well at
the GDT, which had explicit rules and outcomes.
Conversely, strategy planning and cognitive pro-
cessing speed appeared to be less efficient, affecting
the IGT interim outcome. Healthy subjects worked
out the advantageous decks quite quickly and
stuck to them steadily to the end of the test.
Patients also showed a positive performance
curve, but their slope was significantly less steep
in the first 60 (under-ambiguity) trials, indicating a
slower understanding of the winning approach to
the task. These results (MS patients’ good perfor-
mance at the GDT and bad performance at the
initial rather than final trials of the IGT) are com-
pletely at odds with the findings of Farez et al.’s
study (2014). A possible explanation for this dis-
crepancy might be their smaller clinical sample size
(27 vs. 60 patients). Moreover, they did not con-
sider cognitive impairment as an exclusion

criterion, and their MS group was in fact impaired
at selective reminding and spatial recall tests and
the Paced Auditory Serial Addition Test (PASAT),
which were significantly correlated with gambling
performance. Finally, their control group was par-
ticularly brilliant at the IGT. If we compare mean
scores obtained at each IGT block, our patients
appear to be actually even more impaired then
Farez et al.’s patients. On the other hand, their
controls made a much higher number of advanta-
geous choices than our controls on average, espe-
cially in the last blocks. Unlike in our study, Farez
et al.’s healthy participants were not matched with
patients for IQ. Even if data on the influence of
intelligence on the IGT are inconsistent (Toplak
et al., 2010), we cannot rule out that a lower IQ
might have interfered with their patients’ perfor-
mance at the test. The same might hold true for the
GDT as well. Additionally, for this test a metho-
dological element might also have played a role in
intergroup differences. While we considered only
choices of the single number option as disadvanta-
geous decisions, Farez et al. also took into account
pairs, which led to a less strict and therefore more
sensitive score of risky behavior (a net score is also
mentioned by the authors in the Method section,
but not reported in the results). The study by
Muhlert et al. (2014) also showed impaired deci-
sion under risk using the CGT. In this case, the
study population was very large (>100 subjects),
but not comparable with our sample, as it included
patients with RR, primary progressive and second-
ary progressive MS, with and without cognitive
deficits, and with various levels of physical disabil-
ity. Other three reports on IGT in MS are also
inconsistent with our data (Kleeberg et al., 2004;
Nagy et al., 2006; Roca et al., 2008). All three
studies showed an excess of disadvantageous selec-
tions at the IGT in patients compared with con-
trols, in particular in the last blocks of cards.
Again, the main reason for these discrepant find-
ings might be the smaller samples’ size. Even if
small samples are less likely than large ones to
yield significant findings, due to lower statistical
power, they may lead to artefactual positive find-
ings, being more prone to the inclusion of idiosyn-
cratic cases. Besides, in two of these reports
patients seemed cognitively more compromised
(Nagy et al., 2006; Roca et al., 2008), and the
same cannot be ruled out for the third study
(Kleeberg et al., 2004), in which the neuropsycho-
logical assessment was very limited. In this last
publication, patients were also particularly anxious
and had a longer disease duration and a slightly
worse EDSS score. There is, however, an exception
in the literature, as Simioni et al.’s (2008) results
DECISION MAKING IN MULTIPLE SCLEROSIS 1073
are in agreement with our findings. First of all,
they also found no significant intergroup differ-
ences for IGT mean total score and learning
index in their large sample of high-functioning
patients. Secondly, even if they did not analyze
performance across blocks, the boxplots represent-
ing the evolution of selections in subjects with
definite MS (Figure 1 in their paper) show a pat-
tern similar to our patients’ curve, with a delayed
shift to advantageous decks.
One limitation of the present study is the lack of
measures of emotional state. Emotions are a crucial
component of making decisions (see Damasio,
1996, and Dunn, Dalgleish, & Lawrence, 2006, for
the “somatic marker hypothesis” on the role of
emotional reactivity in decision making). Their rele-
vance in decision competency in patients with RR
MS has been demonstrated by Kleeberg et al.
(2004), who showed that lower skin conductance
responses during IGT execution were associated
with poorer performance. For these reasons, our
findings on the cognitive processes underlying deci-
sion making in MS cannot be considered conclusive
until interactions between cognition and emotions
in decision processes are considered. A second lim-
itation concerns the criterion that we used for defin-
ing the shift to under-risk decision making—that is,
the increase in advantageous choices. Decision
under risk requires insight into task contingencies
by definition (Brand et al., 2006), but an advanta-
geous behavior in the IGT does not necessarily
imply such an insight. Test authors themselves
noticed that some healthy individuals and patients
tend to “say the right thing but do the wrong thing”
in performing the task (Bechara, Damasio, &
Damasio, 2000). Probably this dissociation occurs
only in a minority of subjects, but our findings
would better be verified using an index of passage
to the under-risk condition that kept into account
conscious knowledge of rules and outcomes.
Another limitation is the difference in the size of
our clinical and control groups, which might have
resulted in a decline in statistical power. We doubt,
though, that this might have had a crucial impact
on our final results, given the major overlap of
scores obtained by patients and normal controls at
the two gambling tasks (see Figure 1). Finally,
another caveat of our study is that we did not take
into account possible differences in the reliability of
the two measures of decision making. Reliability
influences the sensitivity of a test and the strength
of its relationship with other variables. We cannot
exclude that the negativity of intergroup compari-
sons and correlation analysis for the GDT might at
least partly be due to a lower reliability of this task
than the IGT.
1074 COGO ET AL.
In spite of these limitations, we believe that our
study has contributed to increasing knowledge on the
cognitive mechanisms underlying decision processes
in MS. Moreover, we have demonstrated that deci-
sion making is substantially intact in high-function-
ing patients with RR MS. More precisely, these
subjects might be slow at identifying the most profit-
able options when they are ambiguous, due to diffi-
culties in building a strategy and establishing
associations. Nevertheless, they are subsequently
capable of making advantageous decisions consis-
tently, thanks to a preserved ability of set
maintenance.
REFERENCES
Basso, A., Capitani, E., & Laiacona, M. (1987). Raven’s
Coloured Progressive Matrices: Normative values on
305 adult normal controls. Functional Neurology, 2,
189–194.
Bechara, A., Damasio, A. R., & Damasio, H. (2000).
Emotion, decision making and the orbitofrontal cor-
tex. Cerebral Cortex, 10, 295–307.
Bechara, A., Damasio, A. R., Damasio, H., &
Anderson, S. W. (1994). Insensitivity to future con-
sequences following damage to human prefrontal cor-
tex. Cognition, 50, 7–15.
Beck, A. T., Ward, C. H., Mendelson, M., Mock, J., &
Erbaugh, J. (1961). An inventory for measuring
depression. Archives of General Psychiatry, 4, 561–571.
Brand, M., Fujiwara, E., Borsutzky, S., Kalbe, E.,
Kessler, J., & Markowitsch, H. J. (2005). Decision-
making deficits of Korsakoff patients in a new gam-
bling task with explicit rules: Associations with execu-
tive functions. Neuropsychology, 19, 267–277.
Brand, M., Labudda, K., Kalbe, E., Hilker, R.,
Emmans, D., Fuchs, G., … Markowitsch, H. J.
(2004). Decision-making impairments in patients
with Parkinson’s disease. Behavioural Neurology, 15,
77–85.
Brand, M., Labudda, K., & Markowitsch, H. J. (2006).
Neuropsychological correlates of decision-making in
ambiguous and risky situations. Neural Networks, 19,
1266–1276.
Brand, M., Recknor, E. C., Grabenhorst, F., &
Bechara, A. (2007). Decisions under ambiguity and
decisions under risk: Correlations with executive
functions and comparisons of two different gam-
bling tasks with implicit and explicit rules. Journal
of Clinical and Experimental Neuropsychology, 29,
86–99.
Buelow, M. T., & Suhr, J. A. (2009). Construct validity
of the Iowa Gambling Task. Neuropsychology
Review, 19, 102–114.
Caltagirone, C., Gainotti, G., Carlesimo, G. A., Parnetti,
L., Fadda, Gallassi, R., … Nocentini, U. (1995).
Batteria per la valutazione del deterioramento mentale
(Parte I): Descrizione di uno strumento di diagnosi neu-
ropsicologica [Battery for the evaluation of mental dete-
rioration (Part I): Description of an instrument for the
neuropsychological diagnosis]. Archivio di Psicologia,
Neurologia e Psichiatria, 56, 461–470.
Chiaravalloti, N. D., & De Luca, J. (2008). Cognitive
impairment in multiple sclerosis. The Lancet
Neurology, 7, 1139–1151.
Damasio, A. R. (1996). The somatic marker hypothesis
and the possible functions of the prefrontal cortex.
Philosophical Transactions of the Royal Society B:
Biological Sciences, 351, 1413–1420.
Delazer, M., Sinz, H., Zamarian, L., & Benke, T. (2007).
Decision-making with explicit and stable rules in mild
Alzheimer’s disease. Neuropsychologia, 45, 1632–
1641.
Dunn, B. D., Dalgleish, T., & Lawrence, A. D. (2006).
The somatic marker hypothesis: A critical evaluation.
Neuroscience & Biobehavioural Reviews, 30, 239–271.
Farez, M. F., Crivelli, L., Leiguarda, R., & Correale, J.
(2014). Decision-making impairment in patients with
multiple sclerosis: A case-control study. BMJ Open, 4
(7), e004918. doi:10.1136/bmjopen-2014-004918
Filippi, M., & Rocca, A. M. (2010). MRI and cognition
in multiple sclerosis. Neurological Sciences, 31, S231–
S234.
Henry, J. D., & Beatty, W. W. (2006). Verbal fluency
deficits in multiple sclerosis. Neuropsychologia, 44,
1166–1174.
Kleeberg, J., Bruggimann, L., Annoni, J. M., van Melle,
G., Bogousslavsky, J., & Schluep, M. (2004). Altered
decision-making in multiple sclerosis: A sign of
impaired emotional reactivity? Annals of Neurology,
56, 787–795.
Kurtzke, J. (1983). Rating neurologic impairment in
multiple sclerosis: An expanded disability status
scale (EDSS). Neurology, 33, 1444–1452.
Laiacona, M., Inzaghi, M. G., De Tanti, A., & Capitani,
E. (2000). Wisconsin Card Sorting Test: A new global
score, with Italian norms, and its relationship with the
Weigl sorting test. Neurological Sciences, 21, 279–291.
McDonald, W. I., Compston, A., Edan, G., Goodkin, D.,
Hartung, H. P., Lubin, F. D., … Wolinsky, J. S. (2001).
Recommended diagnostic criteria for multiple sclerosis:
Guidelines from the international panel on the diagnosis
of multiple sclerosis. Annals of Neurology, 50, 121–127.
Muhlert, N., Sethi, V., Cipolotti, L., Haroon, H.,
Parker, G. J., Yousry, T., … Chard, D. (2014). The
grey matter correlates of impaired decision-making in
multiple sclerosis. Journal of Neurology,
Neurosurgery and Psychiatry. doi:10.1136/jnnp-2014-
308169
Nagy, H., Bencsik, K., Rajda, C., Benedek, K., Beniczky,
S., Kéri, S., … Vécsei, L. (2006). The effects of reward
and punishment contingencies on decision-making in
multiple sclerosis. Journal of the International
Neuropsychological Society, 12, 559–565.
Novelli, G., Papagno, C., Capitani, E., Laiacona, M.,
Vallar, G., & Cappa, S. F. (1986). Tre test clinici di
ricerca e produzione lessicale. Taratura su soggetti
normali [Three clinical tests of lexical search and
production. Standardisation in healthy subjects].
Archivio di Psicologia, Neurologia e Psichiatria, 4,
477–506.
Orsini, A., & Laicardi, C. (1997). Wechsler Adult
Intelligence Scale–Revised. Contributo alla taratura
italiana [Wechsler Adult Intelligence Scale-Revised.
Contribution to the Italian standardization].
Florence: Giunti Organizzazioni Speciali.
Roca, M., Torralva, T., Meli, F., Fiol, M., Calcagno,
M., Carpintiero, S., … Correale, J. (2008). Cognitive
deficits in multiple sclerosis correlate with changes in

fronto-subcortical tracts. Multiple Sclerosis, 14, 364–
369.
Rogers, R. D., Owen, A. M., Middleton, H. C.,
Williams, E. J., Pickard, J. D., & Sahakian, B. J.
(1999). Choosing between small, likely rewards and
large, unlikely rewards activates inferior and orbital
prefrontal cortex. Journal of Neuroscience, 19, 9029–
9038.
Schiebener, J., Zamarian, L., Delazer, M., & Brand, M.
(2011). Executive functions, categorization of prob-
abilities, and learning from feedback: What does
really matter for decision making under explicit risk
conditions? Journal of Clinical and Experimental
Neuropsychology, 33, 1025–1039.
Simioni, S., Ruffiex, C., Kleeberg, J., Bruggimann, L.,
Annoni, J. M., & Schluep, M. (2008). Preserved
DECISION MAKING IN MULTIPLE SCLEROSIS 1075
decision making ability in early multiple sclerosis.
Journal of Neurology, 255, 1762–1769.
Simioni, S., Ruffieux, C., Kleeberg, J., Bruggimann, L.,
du Pasquier, R. A., Annoni, J. M., … Schluep, M.
(2009). Progressive decline of decision-making perfor-
mances during multiple sclerosis. Journal of the
International Neuropsychological Society, 15, 291–295.
Toplak, M. E., Sorge, G. B., Benoit, A., West, R. F., &
Stanovich, K. E. (2010). Decision-making and cogni-
tive abilities: A review of associations between Iowa
Gambling Task performance, executive functions,
and intelligence. Clinical Psychology Review, 30,
562–581.
Wechsler, D. (1981). Manual for the Wechsler Adult
Intelligence Scale-Revised New York, NY:
Psychological Corporation.