Professional Documents
Culture Documents
www.elsevier.com/locate/b&c
Abstract
Through the use of several tests of cognition we have documented sex differences in young children, adolescents, and adults on
tasks that rely on the integrity of the orbital prefrontal cortex. In children under three years of age, males performed with signif-
icantly fewer errors than did females on tests of object reversals. No significant sex differences were found in older children, despite
the use of a more challenging object reversal task. Sex differences were also found in adolescents and adults on the Iowa Gambling
Task. On this decision-making task, in contrast to males, females appear to be responding to different elements of the task. Dis-
cussion of the implications for these findings is presented.
Ó 2004 Elsevier Inc. All rights reserved.
0278-2626/$ - see front matter Ó 2004 Elsevier Inc. All rights reserved.
doi:10.1016/S0278-2626(03)00279-3
W.H. Overman / Brain and Cognition 55 (2004) 134–147 135
job!’’) also follows a correct response. After reaching the performance of normal infant females until much
criterion performance (90% correct responses for two later in life (Clark & Goldman-Rakic, 1989; Goldman,
consecutive test days) the subject learns a second dis- 1971). (d) In contrast, on the concurrent discrimination
crimination with two new stimuli. These first two dis- task, normal infant females out-perform infant males
crimination problems serve as control tasks for general (Bachevalier, Hagger, & Bercu, 1989); however, or-
learning ability. On the first test session after attaining chiectomized infant males perform as well as normal
criterion on the second discrimination problem, the re- females and better than normal males. (e) Neonatal
ward contingencies for that problem are reversed with- ablations of area TE in the inferior temporal cortex
out warning. Thus, the previously negative stimulus impair concurrent discrimination performance in nor-
becomes positive, and the previous positive stimulus mal females but not in infant males (Bachevalier,
becomes negative. The subject learns this reversed dis- Brickson, Hagger, & Mishkin, 1990). (f) In males, levels
crimination to criterion whereupon the reward contin- of circulating testosterone are inversely correlated with
gencies are reversed again, and so on, for as many performance on the concurrent discrimination task
reversals as the experiment calls for. The subjectÕs score (Hagger, Bachevalier, & Bercu, 1987).
for this task is the number of errors to criterion for each Taken together, these data strongly indicate that, in
reversal and across all reversals. A high number of er- infant males. Perinatal testosterone accelerates the
rors is interpreted as the failure to inhibit a prepotent functional maturation of orbital prefrontal cortex and
response (e.g., the learned response to the initial dis- slows the functional maturation of area TE in the inferior
crimination) and/or the lack of behavioral flexibility. temporal lobe.
2.2. Concurrent discrimination 2.4. Results with children on object reversal and concur-
rent discrimination
In this task, on each test day, the subject is consec-
utively presented with a number of object pairs. One We have replicated, in children, the behavioral com-
member of each pair conceals food (Cheerio or Froot ponents of the double dissociation found in monkeys. We
Loop) and one does not conceal food. In addition, a tested children in the WGTA using almost exactly the
social reinforcer, (e.g., ‘‘Good job!’’) is used as a rein- same testing procedures as used with monkeys, i.e., a
forcer. The same series of object pairs is presented once limited number of daily trials for many consecutive days,
each test day (24-h retention interval). The positive and food and social reward, no verbal task instructions. The
negative objects within each pair and the serial order of data can be found in Overman et al., 1996, 1997. The re-
the pairs remain constant across daily sessions, but the sults were the following: (a) Male and female children
left–right position of the baited and unbaited objects are were equal in initial discrimination learning. (b) Males
randomized daily. Testing continues, day after day, until under the age of 29 months were superior to age-matched
the subject attains a criterion of 90% correct choices for females in reversal learning following a pattern almost
two consecutive test days. exactly like that found in monkeys. With the addition of
more subjects we have recently discovered that the male
2.3. Sex specificity, neural bases, and the role of gonadal superiority exists through 36 months of age (Overman &
hormones Godin, unpublished data). (c) Females under the age of 36
months were superior to age-matched males in learning a
There is a convincing double dissociation of perfor- concurrent discrimination task. (d) There were no sig-
mance on these two tasks showing that two neural sys- nificant sex differences on either task in older children or
tems develop at different rates in male and female infant adults. (e) In addition to slower learning relative to males,
monkeys. Specifically, infant males perform better than about 20% of the females under 29 months showed hyper-
females on the object reversal task, and infant females emotional behaviors during reversal training.
perform better than males on the concurrent discrimi- As shown in Fig. 1, we found that male and female
nation task. The evidence can be summarized as follows: children under the age of 30 months display significant
(a) Seventy-five day-old male monkeys are superior to differences in learning object reversals almost exactly as
age-matched females on object reversal tasks (Goldman, seen in infant monkeys. Children older than 29 months
Crawford, Stokes, Galkin, & Rosvold, 1974). (b) While and adults showed no sex differences in performance
normal infant males out-perform infant females on the (lower graph).
object reversal task, perinatally androgenized infant fe- Fig. 2 shown the results of testing young monkeys
males perform as well as normal males and better than and children on the concurrent discrimination. In both
normal females (Clark & Goldman-Rakic, 1989). (c) groups of youngest subjects, females out-performed
Early ablations of orbital prefrontal cortex impair object aged-matched males by a factor of nearly two. Older
reversal performance of normal infant males and an- children and monkeys and adults of both species showed
drogenized infant females but such lesions do not impair no sex differences. For infants of both species, the
136 W.H. Overman / Brain and Cognition 55 (2004) 134–147
(19 females and 14 males) ranging in age from 11 to 12 responded in two consecutive trials at 87% correct (two
years, and 64 college age adults (14 females and 24 blocks of 13/15 correct responses). During this dis-
males). Originally we had proposed to test children as crimination and the remaining test trials, the following
young as two years of age but we quickly discovered that reinforcement system was introduced. After each set of 7
the computerized reversal task was too demanding in correct responses, a 3-s ‘‘colored light show’’ appeared
terms of length of test sessions, response requirements, on the screen rather than the happy face. At each light
and comprehension of the reward system. Consequently, show the child was allowed to choose a colorful sticker
only children three years of age and above were tested. and was instructed to move one bead on a colorful
abacus. After a row of 10 beads was moved the child
3.1.1. Procedure was allowed to choose a small toy on the table (small
During testing each child sat in front of a stand-alone dinosaur, bracelet, animal, yo-yo, etc.). At the end of the
computer screen that was run by a separate lap top test session, the child was allowed to choose a larger toy
computer in front of the experimenter. This arrange- (toy motorcycle, car, necklace, ring, change purse, etc).
ment prevented the subjects from being distracted by the Incorrect responses were followed by a 2 s black screen.
keyboard. Most children were capable of using a mouse
to point and click on the stimuli, but the program was 3.2. Teaching the second discrimination
designed to teach mouse use to the few inexperienced
children. After attaining criterion on the triangle–square dis-
crimination, the subject was shown a pentagon (+) vs. a
3.1.2. Methods circle ()). The same reinforcement contingencies were
Each child was tested separately in one session that used as above. Criterion for this discrimination was
lasted from 30 to 60 min depending on the complexity of again 87% responses for two consecutive blocks of 15
the reversal task and the ability of the subject. The child trials.
was seated in front of the computer screen and, the ex-
perimenter explained that, in this ‘‘game,’’ stickers and 3.2.1. Reversal learning
prizes could be obtained (stickers and prizes were out After attaining criterion on the pentagon-circle dis-
of reach but were visible on the right hand side of the crimination, without warning, the pentagon became the
table). negative stimulus and the circle became the positive
stimulus. The percentage of the reversals could be pre
3.1.3. Pretraining: Teaching use of the mouse programmed so that the density of reversal was 100, 93,
On the first trial a 200 200 200 black-outlined trian- 86, 80%, etc. A 93% reversal consisted of 14 trials in
gle appeared on left or right half of the computer screen which the S+ was correct and one (disconfirming) trial
while the opposite half screen was black. The experi- in which S) was correct. The 86 and 80% reversals had 2
menter said ‘‘watch what happens when I click on the and 3 disconfirming trials, respectively, in each block of
triangle’’ and then clicked on it. This was followed by a 15 trails. Table 1 shows trials of full and partial rever-
yellow ‘‘happy face’’ filling most of the screen and a sals. For any partial reversal, the subject learned to re-
simultaneous ‘‘tinkle’’ of music. The face lasted 1.5 s, spond to the stimulus that was correct on most of the
disappeared and was followed by a 1.5-s message saying trials. Performance criterion was always 87% correct
‘‘GET READY.’’ Then the triangle appeared on the responses for two consecutive blocks of trials. Correct
right or left half of the screen. The child was handed the responses on disconfirming trials were always counted as
mouse and asked, ‘‘Can you do what I just did?’’ Eight ‘‘correct’’ by the computer even though the subject was
such trials occurred with the triangle appearing ran- not reinforced for that response. After attaining crite-
domly on the left or right of the screen. On each suc- rion on the first reversal, the reinforcement contingen-
cessive trial the darker half of the screen progressively cies switched without warning so that now the pentagon
turned lighter so that on trial 8 the entire screen was was again positive and the circle was again negative.
white. Three such reversals were given to each subject.
Table 1
Average number of training sessions and average errors through criterion for discrimination and reversal learning in the Wisconsin General Testing
Apparatus and on a computer
Group # sessions Disc. 1 Disc. 2 Rev. 1 Rev. 2 Rev. 3
WGTA 13.1 6.3 2.1 7.6 2.8 3.4
Computer 1.0 0.6 2.0 7.1 3.5 3.5
There were no disconfirming trials on these reversal tasks.
Table 2
Average errors to criterion on discriminations and reversals for 8 groups of participants
Group Discrimination 1 Discrimination 2 Reversal 1 Reversal 2 Reversal 3
WGTA preschool 6.3 2.1 7.6 2.8 3.4
Computer preschool 0 disco. 0.6 2.0 7.1 3.5 3.5
Computer preschool 1 disco. 0.9 1.8 7.78 6.6 7.0
Computer preschool 2 disco. 1.1 1.1 13.4 11.2 9.3
Computer adoles. 2 disco. 0.03 0.2 19.1 13.6 15.2
Computer adoles. 3 disco. 0.08 0.2 56.7 40.7 29.5
Computer adult 2 disco. 0.2 0.2 10.6 7.9 8.8
Computer adult 3 disco. 0.15 0.0 22.3 16.1 10.1
‘‘Disco.’’ ¼ disconfirming trials.
second discrimination. Thirty-three of these (18 males were tested in a single session, which was rarely more
and 15 females) were trained in the WGTA at 15 trials a than 1 h in duration. Stated another way, children in the
day, day after day, until attainment of criterion on each WGTA group experienced distributed practice while
phase of the task as reported earlier (Overman et al., children in the computer group experienced massed
1996). Children in this group received food reward practice.
(Froot Loops) and social rewards (‘‘good boy/girl’’) for Despite the difference in practice time, learning per-
each correct response. formances were remarkably similar (Table 2 and Fig. 3).
As shown in Table 2 the two groups of children who The only exception to this was that, because of the fade-
learned the traditional reversal task (no disconfirming in procedure, children in the computer group learned the
trials) differed vastly in the amount of time they par- first discrimination significantly faster than those in the
ticipated in the tasks. Children in the WGTA group WGTA group, p < :05. There were no other significant
required and average of 13.1 individual test sessions for differences between the tasks. The second discrimination
completion of the task. Since each test session took was learned equally fast by both groups (2.1 vs. 2.0,
about 20 min each day, each child in the WGTA group p > :05). The reversals were learned equally fast in both
spent a total of approximately 4.4 h with the experi- groups (average total errors across all reversals 13.8 vs.
menter. In contrast, children in the computer group 14.1, p > :05). Both groups demonstrated the classic
Fig. 3. Mean errors to criterion on ‘‘full’’ (100%) and ‘‘partial’’ object reversal tasks for male and female children, adolescents, and young adults.
Asterisks indicate significant differences between reversal conditions (i.e., number of disconfirming trials).
W.H. Overman / Brain and Cognition 55 (2004) 134–147 139
reversal learning pattern with high numbers of errors on vious task. Still, there were no significant differences
the first reversal followed by decreasing errors thereaf- between males (35.9 average total errors) and female (29
ter. The equivalent performance in the WGTA and on average total errors) on this task.
the computer means that results from the more efficient
computerized task can confidently be compared with 4.4. Performance by adolescents: Computer partial
results from animals and humans who were tested in the reversal with 2 disconfirming trials
WGTA.
A group of 32 7th grade students (19 boys and 13
4.2. Computerized partial reversal learning in children: 1 girls mean age of 12 years) were administered the com-
disconfirming trial puterized partial reversal task with 2 disconfirming tri-
als. As incentives for driving to the university for testing
Another group of 79 preschool children (42 males parents received a $15 coupon for groceries and for
and 37 females) were tested on two discriminations as participation children were given a tee-shirt and cou-
described above and three reversals of the second dis- pons to local stores.
crimination; however, the reversal was only a 93% re- Fig. 3 shows the total errors across three reversals
versal (one disconfirming trial randomly occurring for adolescent males and females. There were no sig-
among each block of 15 trials). Criterion requirements nificant sex differences with or without inclusion of
were the same as above: two consecutive blocks of 88% outlying subjects. One interesting result was that ado-
correct responses at each phase of learning. There were lescents made significantly more errors on the 2 dis-
no differences in learning the first two discriminations confirming-reversal problem than did preschool
compared with children learning a regular ‘‘full’’ rever- children. This was true whether or not the two outlying
sal on the computer. This is because the discrimination adolescent subjects (one male and one female) were
tasks were exactly the same in both groups. included in the analysis.
Fig. 3 shows the performance of male and female Because of the spontaneous comments made by
children on the computer task with one disconfirming children and adolescents, we speculate that adolescent
trial. The partial nature of this task (presence of the participants differed in that they were trying more than
disconfirming trial) rendered it significantly more dif- children to figure out complex (and non-existent) ‘‘pat-
ficult than the regular reversal task, p < :05. The dif- terns’’ for the deliveries of reinforced and non-rein-
ference was about 7 errors for females and 5 errors forced trials. In other words, the children were better at
for males. Thus, our effort to make a reversal task probability matching than were adolescents. This is
more challenging by making it a partial reversal was somewhat similar to reports of animals being better at
successful. However, there were no significant sex dif- probability matching than adult humans (Wolford,
ferences with males learning three reversals in an av- Miller, & Gazzaniga, 2002). Perhaps intellectual devel-
erage of 19.9 errors and females in an average of 22.9 opment allows the older participant to generate erro-
errors. neous pattern strategies of which younger participants
are not yet capable. In addition the adolescent subjects
4.3. Computerized partial reversal learning in children: 2 appeared to be very frustrated by the task and less so
disconfirming trials than adults on the same task who made significantly
fewer errors as shown below.
Another group of 62 preschool children (45 males
and 17 females) were tested on two discriminations as 4.5. Adolescents: Computer partial reversal with 3
described above and three reversals of the second dis- disconfirming trials
crimination; however, the reversal was only a 87% re-
versal (2 disconfirming trial randomly occurring among In an effort to make the partial reversal task even
each block of 15 trials). Criterion requirements were two more challenging 11 adolescents (5 males and 6 females)
consecutive blocks of 88% correct responses at each were given the test with 3 disconfirming trials in every
phase of learning. There were no differences in learning block of 15 trials. As shown in Fig. 3, this problem was
the first two discriminations compared with children exceedingly difficult with approximately a 100% increase
learning a regular reversal or a reversal with one dis- in errors as compared to the task with 2 disconfirming
confirming trial on the computer. trials. Although the participants completed all three
Fig. 3 shows the total errors on reversals 1, 2, and 3 reversals they became frustrated and upset during the
for male and female on the computer task 2 discon- session which required a minimum of 90 min to com-
firming trials. The additional disconfirming trial ren- pete. Since there was no evidence for sex differences on
dered the task significantly more difficult than the task this task and since the task was obviously unpleasant for
with one disconfirming trial or the task with no dis- the participants we stopped testing adolescents with 3
confirming trials (p’s < :05) of 8.1 errors from the pre- disconfirming trials.
140 W.H. Overman / Brain and Cognition 55 (2004) 134–147
4.6. Adults: Computer partial reversal with 2 and 3 radic loss of reward (disadvantageous decks). Cards in
disconfirming trials the other two decks, e.g., red and green decks, are as-
sociated with low reward but with even lower sporadic
Thirty-two adult college students (16 males and 16 losses (advantageous decks). Consistent choice of ad-
females) were testing on partial reversal tasks with 2 vantageous cards (red and green) will result in low but
disconfirming trials, and an additional 32 adults (8 long-term gain, whereas consistent choice of disadvan-
males and 24 females) were tested with 3 disconfirming tageous cards (yellow and blue) will result in overall loss
trials. The results are shown in Fig. 3. There were no sex of money. Over the course of 100–200 trials, normal
differences under either condition (p > :05); however, the control participants gradually formulate the strategy of
3 disconfirming trial condition was significantly more picking from the low-paying (advantageous) decks,
difficult than the 2 disconfirming trial condition (average which result in gain in the long run. In contrast, patients
of 48 errors vs. 27 errors). Adults made significantly with damage centered on the ventromedial prefrontal
fewer errors than did adolescents in both conditions cortex do the opposite, and select more cards from the
(p’s < :05) high-paying (disadvantageous) decks, which results in
loss in the long run (Bechara et al., 1994, 1997.) How-
4.7. Summary of partial reversal testing ever, the same patients perform normally on standard
tests of executive function such as the Wisconsin Card
It is evident that, for preschool children, computer- Sorting Test (WCST) Bechara et al. (1994, 1997).
ized discrimination, and reversal learning is comparable
to learning using the WGTA apparatus. There is no 5.2. Orbital prefrontal cortex and decision-making
reason to believe that this is not the case for adolescents
and adults. The reversal task was clearly rendered more The role of the orbital PFC in decision-making is not
difficult for all ages of subjects by adding disconfirming precisely understood at this time. Several operations
trials. However, partial reversal manipulation did not have been proposed. Damasio and his colleagues have
result in significant differences between males or females written extensively about the ‘‘somatic markers’’ which
at any age. Our failure to find sex differences among are emotional feelings that gradually come to be con-
young was puzzling because, as shown below, we were nected, by learning, ‘‘to predicted outcomes of certain
documenting sex differences among adults on another scenarios’’ (Damasio, 1994, p. 174). These emotional
task known to be sensitive to orbital prefrontal damage. ‘‘markers,’’ as defined by Skin Conductance Responses
(SCRs), are thought to act as an early warning system
which guides decisional behavior before the subject is
5. Cognitive sex differences among adults: The Iowa consciously aware of the precise rule. Unlike normal
Gambling Task controls, patients with orbital prefrontal damage do not
generate anticipatory SCRs during the course of playing
One of the reoccurring themes of this Special Issue the IGT (Bechara et al., 1994). The lack of SCRÕs is
concerns cognitive changes that follow damage to the thought to underlie disadvantageous decision-making in
orbital prefrontal cortex. One of the most common such patients (see Bechara, this volume). It should be
changes involves impairments in the quality of personal noted that, recently, the somatic marker hypothesis has
decision-making abilities. Patients with orbital damage come under scrutiny (Tomb, Hauser, Deldin, &
demonstrate relative insensitivity to future consequences Caramazza, 2002).
even when their decisions are against their personal in- In contrast to the somatic marker hypothesis, Rolls
terest; however, these patients maintain relatively nor- (1996) has suggested that the orbital PFC mediates de-
mal intellectual functioning in other realms (Bechara, cision-making by providing action selection mechanisms
Damasio, Damasio, & Anderson, 1994; see also Bec- with information about the reinforcing properties of
hara, this issue). unconditioned and conditioned stimuli. The precise
distinctions and overlaps between these two hypotheses
5.1. The Iowa Gambling Task are not critical for this present research.
Bechara et al. (1994, & see this issue) developed the 5.3. Populations that perform poorly on these decision-
‘‘Iowa Gambling Task’’ (IGT) that detects deficits in making tasks
personal decision-making. This task simulates real-life
decision-making in the manner in which it presents an The following patient populations have been shown
uncertainty of reinforcers and punishers. The task re- to perform at sub-normal levels on the IGT (impaired
quires the participant to choose cards from four decks. performance is defined by the continued choice of dis-
Cards in two of the decks, e.g., yellow and blue decks, advantageous $100 cards): (1) patients with damage
are associated with high reward but even higher spo- centered on the ventromedial prefrontal cortex (Bechara
W.H. Overman / Brain and Cognition 55 (2004) 134–147 141
et al., 1994, 1997); (2) polysubstance abusers (stimulants feedback (correct or incorrect) as well as auditory
or opiods) (Grant, Contoreggi, & London, 2000); (3) feedback (high tone for correct and low tone for incor-
cocaine abusers vs. abstinent cocaine abusers and non- rect). Learning difficulty is increased by the inclusion of
abusers (Bartzokis, Lu, Beckson, & Rapaport, 2000) prearranged trials that are ‘‘disconfirming’’ (e.g., they
[Patients in these studies performed normally on the give erroneous feedback about correct or incorrect
Wisconsin Card Sorting Test]; (4) violent offenders vs. choices). Neither disconfirming nor equal-probability
non-violent offenders (Fishbein, 2000); (5) pathological trials are included in scoring.
gamblers (Cavedini, Riboldi, Keller, DÕAnnucci, & The participants were divided into 6 groups defined
Bellodi, 2002). by age and hormonal status: young males (mean age ¼
Patients with amygdala damage also perform poorly 19.1 years), older males (mean age ¼ 59.4 years), young
on the IGT. However, unlike patients with damage to females with low hormones (menstruating, mean age ¼
the orbital PFC, those with amygdala damage do not 19.8 years), young females with high hormones (mid-
produce SCRs even after receiving a reinforcer or pun- luteal, mean age ¼ 22.4), older postmenopausal women
isher (neither group produced anticipatory SCRs) either taking estrogen replacement therapy (mean age ¼
(Bechara, Damasio, Damasio, & Lee, 1999). This sug- 54.5 years) or not taking ERT (mean age ¼ 62. 7 years).
gests that the two regions play different roles in decision- Hormonal status was defined via Radio Immuno Assay
making: amygdala damage has indirect consequence on of dried blood spots (Shirtcliff, Reavis, Overman, &
attaching affective qualities to stimuli where as the VM Granger, 2001). Each subject was tested individually.
region is not essential for the generation of affective Each subject completed two preliminary questionnaires
attributes to (i.e., in response to) emotional stimuli and then the two tests of probability learning, the IGT
(Bechara et al., 1999). and the CWT. Half of the participants in each group
took the Iowa Task first and half took the California
5.4. Sex differences in performance on the IGT Task first. On the IGT, participants were given 150 trials
and every 10th trial they were asked if they knew the
Over the past four years we have tested over 700 non- operative rule (that the $50 red and green decks gained
substance-abusing participants (age range 12–65 years) money in the long run and the $100 yellow and blue
on the IGT (plus control tasks) and have discovered the decks lost money in the long run).
following (described in detail below). (1) Performance Results showed no significant task differences be-
on the IGT improves with age, from 12 years of age tween any of the four hormonal groups of women and
until adulthood. (2) Gonadal hormone status does not they were collapsed into a single female group. Simi-
affect performance in adults. (3) Across age, males larly, there were no task differences between the young
outperform females. (4) Young male adultsÕ perfor- and older men, so they were grouped together. While
mance significantly improves when they have prior ex- there were no sex differences on the California Weather
perience with another decision-making task (the Task, there were interesting sex differences on the IGT,
California Weather Task). (5) By using a novel error as shown in Fig. 4.
analysis, we have found that females demonstrate dif- Males chose significantly more advantageous cards
ferent choices from males due to what we believe are ($50 cards) than did females on the 2nd and 3rd block of
emotional responses. 50 trials (p < :002). By the 3rd block males chose $50
cards on 79% of the trials while females chose them on
5.4.1. Sex differences among adults 68% of the trials. Furthermore, more males stated the
In a recent study we tested 191 young and older adults correct rule at some point during the task than did fe-
on the IGT and the California Weather Task (CWT) males (63.5 and 38%, respectively) (p < :009). And
(Reavis & Overman, 2001). The latter task is a slowly males stated the decision-making rule significantly ear-
acquired probabilistic habit task on which Parkinsons lier in training (75th trial on average) than did females
and Huntingtons patients are significantly impaired (97th trial on average; p < :009; Reavis & Overman,
(Knowlton, Mangels, & Squire, 1996). In the California 2001). It is interesting to note that Kerr and Zelazo (this
Weather Task, participants are shown up to four dis- issue) reports that 3- to 4-year-old boys out perform
tinctly patterned cards on a computer screen and grad- age-matched girls on a child version of a decision-
ually learn which combinations of pattern predict ‘‘rain’’ making task.
or ‘‘sunshine.’’ A particular card pattern is associated
with the outcome of ‘‘rain’’ 75, 57, 43 or 24% of the time 5.4.2. Task order effect: Does practice help?
and thus 24, 43, 57, or 75% of the time with the outcome One of the most interesting findings in this study was
of ‘‘sunshine.’’ Participants are exposed to any combi- an effect of task order. While there was no overall sig-
nation of cards on each trial, and they gradually learn nificant difference between younger and older males in
which cards and combinations are probabilistically related choice of $50 cards (70 vs. 64%, respectively), younger
to a given outcome. The computer provides written men were clearly superior on the task relative to older
142 W.H. Overman / Brain and Cognition 55 (2004) 134–147
older ones (e.g., 10th, 11th, 12th grades, and adults) the long run. However, neither the two disadvantageous
chose $50 cards about 70% of the time. As shown in decks (Yellow and Blue) nor the two advantageous
Fig. 6, an overall significant sex difference was found in decks (Red and Green) are equal in the absolute number
the ANOVA, but clearly there were no sex differences of penalties or ‘‘)’’ (minus) signs. The participant wins
among 6th graders, 10th graders or adults. We have no some money (but not necessarily as net win) on every
explanation for the odd appearance of the data points card, either $50 or $100 and, thus, there is a ‘‘+’’ (plus)
for 10th graders other than a vagary of sampling. sign on every card in the task. However, losses occur less
The lack of sex differences among this sample adults often, so only some of the $50 and $100 cards contain a
seemingly conflicts with our finding of a male advantage ‘‘)’’ (minus) sign.
in the earlier study (Reavis & Overman, 2001). In that Consequently, the cards can be classified as those
study, task order was counter balanced and young adult which contain only ‘‘+’’ signs vs. those that also contain
males who had the IGT second scored significantly a ‘‘)’’ sign. By this definition, both the disadvantageous
higher than those who had this task first. We speculate and the advantageous decks contain ‘‘+’’ cards and ‘‘)’’
that the previous finding of sex differences may have cards. Table 3 shows a summary of each colored deck
been driven by this order effect that so strongly affected with respect to monetary wins and losses.
males. This hypothesis was investigated though the use This table can be summarized as follows: (1) Every
of a new type of error analysis as presented in the next card contains a ‘‘+’’ sign.
section. (2). Two decks could be considered ‘‘+’’ (plus) decks
because they contain 10 pluses and only one minus per
5.4.4. New error analysis for the IGT 10 trials; however, one of these is a $50 deck (Red) and
Heretofore, researchers who have administered the one is a $ 100 deck (Yellow). (3). Two decks could be
IGT measured the percentage of advantageous ($50) considered as ‘‘)’’ (minus) decks (relatively speaking)
cards chosen in blocks of trials (or the number of $50 because they contain 10 pluses and 5 minuses per 10
cards minus the number of $100 cards). In other words, trials; however, one is a $50 deck (Green) and one is a
the only error was considered to be choice of disad- $100 deck (Blue).
vantageous ($100) cards. But there are other types of In contrast to males, females choose the ‘‘+’’ decks
errors that can be made on this task. In our data set with rather independently of their status of disadvantageous
normal adolescents and young adults we have devised a ($100) or advantageous ($50) (Figs. 7A and B).
new error analysis that reveals very interesting differ- Results of all females (grade 6 through adults) are
ences between males and females. shown in the upper graph, results of all males (grade 6
The new error analysis looks at choice of cards as a through adults) are shown in the lower graph. As shown
function of the absolute number of minus ‘‘)’’ signs and in Fig. 9, both males and females quickly learn to avoid
plus ‘‘+’’ signs (as well as the absolute net outcome of the disadvantageous blue deck presumably because it
the cards). Two of the decks (Yellow $100 and Blue $100 has five large penalties per 10 trials. Both males and
decks) are disadvantageous because the penalties mon- females quickly learn to choose the advantageous red
etarily outweigh the rewards in the long run. Two of the deck, presumably because it has only one small penalty
decks (Red $50 and Green $50 decks) are advantageous per 10 trials. The sexes differ on choices of the other two
in that the rewards monetarily outweigh the penalties in decks. Males learn to choose the advantageous ($50)
Green decks and avoid the disadvantageous ($100) yel-
low deck that has one large loss per 10 trials. In contrast,
females choose the disadvantageous ($100) Yellow card
fairly constantly for the duration of the task.
Table 3
Summary of each colored deck with respect to monetary wins and losses, and number of pluses and minuses
RED (+) Green (+) Yellow ()) Blue ())
Amount won or lost per 10 trials +$250 +$250 )$250 )$250
Number of plus cards per 10 trials 9 5 9 5
Number of minus cards per 10 trials 1 5 1 5
Number of break-even cards per 10 trials 0 3–5 0 0
outcome while females appear to be more sensitive to The first loss did not occur until the 9th card in that
the immediate presence or absence of ‘‘+’’ or ‘‘).’’ deck. Perhaps, the females formed strong preferences for
During testing we often observed that a female would the ‘‘winning’’ yellow cards and were slower to change
verbally respond negatively (e.g., ‘‘oh, no’’) after to other decks compared to males. This hypothesis
choosing an advantageous Green card with the value of would fit with the slow reversal behavior observed in
+ $50 and ) $25. Given such reactions and the simple infant females (Overman et al., 1996).
math involved, the sex difference does not appear to Regardless of the cause, it is clear that male and fe-
reflect femalesÕ inability to perform a simple calculation. malesÕ decision-making is being guided by different
Rather, it appears as if the female was reacting emo- factors. We are currently investigating the possible fac-
tionally negatively to the presence of the ‘‘)’’ sign. These tors underlying this difference.
data may indicate that, on this task, femalesÕ decision-
making is guided by avoidance of negativity (‘‘)’’ signs) 5.5.1. What is being learned by males when a probabilistic
more than is malesÕ decision-making, which is guided task is given prior to the IGT?
more by long-term outcome. Recall that in our previous study, young adult males
Another possibility is that females may be less effi- showed significantly improved performance (selecting
cient in reversing associations related to the yellow deck. more advantageous $50 cards) after having the probabi-
In this deck, the first 8 cards were $100 ‘‘wining’’ cards. listic California Weather Task (Reavis & Overman, 2001).
W.H. Overman / Brain and Cognition 55 (2004) 134–147 145
To better understand the nature of this improvement, when the IGT was second in the test series, men selected
we have reanalyzed those data in terms of choice of significantly more advantageous Red and Green Cards.
color (i.e., plus ‘‘+’’ or minus ‘‘)’’ cards). The results are It is obvious that ‘‘something’’ was imparted to the
shown, first for young adult females who had the IGT men by having the California Weather task first. We do
first or second, and next for the young adult males. not believe that this was an increase in the ability to
Fig. 8 shows the selection of different cards by women perform calculations, because the Weather Task does
who had the IGT first (upper graph) or second, after the not require mathematical calculations.
California Weather Task (lower graph). A comparison Two possibilities come to mind: one based on ‘‘cool’’
of these graphs reveals that there were little changes brain function and one ‘‘hot’’ brain function, to use cur-
between groups in the avoidance of the disadvantageous rent parlance of frontal lobe function. The ‘‘cool’’ possi-
$100 Blue card (with 5 penalties per 10 trials) or in the bility is that the first task imparted some non-emotional
preference for the disadvantageous $100 Yellow card cognitive ability, e.g., probability estimation, that en-
(with one large penalty per 10 trials). When the IGT was hanced performance on the second task. Given the nature
second there appeared to be some increase in the selec- of the California Weather Task, males may have learned
tion of the advantageous $50 Green card (with 5 minus to judge the long-term probabilities of net gain or loss
signs per 10 trials) whereas preference remained about associated with the different type of cards. Females did
the same for the advantageous $50 Red Card (with one not appear to benefit in the same manner from having
small penalty per 10 trials). experience with the California Weather Task, again per-
In contrast, as shown in Fig. 9, men showed quite haps because they were being more strongly guided by the
different card selection when the IGT followed the immediate presence of plus ‘‘+’’ or minuses ‘‘).’’ This
probabilistic Weather Task. brings us to the possible ‘‘hot’’ or more emotional ex-
The largest change was seen in the decrease of choice planation that may apply differentially to males and fe-
of the disadvantageous $100 Yellow card (with one large males. The California Weather Task is difficult and
penalty per 10 trials). Aversion was apparent even in the frustrating, due to the presence of disconfirming trials. [It
first block of 50 trials and extremely apparent in the last is also noteworthy to know that all experimenters were
block of trial, during which both the disadvantageous females.] Perhaps the difficulty of the California Weather
Yellow and Blue cards were rarely chosen. In addition, Task (in the presence of a female experimenter) aroused
anxiety/emotion in the males. And perhaps such anxiety/
emotion somehow primed the orbital prefrontal system in
such a way as to improve performance in males. This may
not have occurred in females due to the facts that their
experimenter was female, and perhaps less threatening,
and that their affective behavior appeared to be centered
around avoidance of minus ‘‘)’’ cards.
This second speculation is somewhat bolstered by
other reports at this Special Issue. For example, Bec-
hara, this volume, reports that simultaneous stimulation
of the vagus nerve improves performance on the IGT,
presumably by enhancing some emotional state. Seguin
(this issue) reports that anxiety can influence perfor-
mance on the IGT. Blair (this issue) reports that PET
scans of the lateral orbital prefrontal cortex is influenced
by viewing photographs of angry faces.
In summary, it is clear that for males, prior experi-
ence with the California Weather Task somehow en-
hanced performance on subsequent IGT. This may not
be a trivial finding. The final explanation of this phe-
nomenon may significantly expand our understanding
of the orbital prefrontal system.
differences in the functions of orbital prefrontal cortex. related to the well-documented numerical difference in
Thus, positron emission tomography (PET) study re- male and females who abuse various substances?
veals a significant difference in the metabolic rate of
glucose utilization in the orbital prefrontal cortex of 19-
to 32-year-old normal men and women (Andreason,
Zametkin, Guo, Baldwin, & Cohen, 1993). During a References
mentally ‘‘idling’’ state, normal adult men have higher
relative metabolism (as shown by PET) than women in Andreason, P. J., Zametkin, A. J., Guo, A. C., Baldwin, P., & Cohen,
R. M. (1993). Gender-related differences in regional cerebral
several brain regions including the orbital prefrontal glucose metabolism in normal volunteers. Psychiatry Research,
cortex, and the temporal–limbic system (hippocampus, 51, 175–183.
amygdala, and temporal lobe) (Gur et al., 1995). In Bachevalier, J., Brickson, M., Hagger, C., & Mishkin, M. (1990). Age
contrast, women have higher metabolism in the middle and sex differences in the effects of selective temporal lobe lesions
and posterior cingulate gyrus (Gur et al., 1995). Results on the formation of visual discrimination habits in rhesus monkeys
(Macca mulatta). Behavioral Neuroscience, 104, 885–889.
from Magnetic Resonance Imaging (MRI) investiga- Bachevalier, J., Hagger, C., & Bercu, B. (1989). Gender differences in
tions show that sex differences in the frontal lobe, and, visual habit formation in 3-month-old rhesus monkeys. Develop-
to a lesser extent, in the temporal lobes, continue into mental Psychobiology, 22, 585–599.
advanced age. Hemispheric volumes obtained from Bartzokis, G., Lu, P. H., Beckson, M., Rapaport, R., et al. (2000).
MRIÕs of young (18–40 years of age) and older (42–80 Abstinence from cocaine reduces high-risk response on a gambling
task. Neuropsychopharmacology, 22, 102–103.
years of age) adults reveal a sexually dimorphic, region- Bechara, A., Damasio, A. R., Damasio, H., & Anderson, S. W. (1994).
specific, age-related reduction in the volume of frontal Insensitivity to future consequences following damage to human
and temporal regions (Cowell et al., 1994). The brain pre-frontal cortex. Cognition, 50, 7–15.
volume reductions are greater in men than in women. Bechara, A., Damasio, H., Damasio, A. R., & Lee, G. P. (1999).
The authors propose that these region-specific brain Different contributions of the human amygdala and ventromedial
prefrontal cortex to decision making. Journal of Neuroscience, 19,
dimorphisms have origins in early neuroendocrinologi- 5437–5473.
cal development (Cowell, Allen, Zalatimo, & Dennen- Bechara, A., Damasio, H., Tranel, D., & Damasio, A. R. (1997).
berg, 1992). Deciding advantageously before knowing the advantageous strat-
egy. Science, 275, 1293–1295.
Cavedini, P., Riboldi, G., Keller, R., DÕAnnucci, A., & Bellodi, L.
(2002). Frontal lobe dysfunction in pathological gamblers. Biolog-
7. Summary and implications ical Psychiatry, 51, 334–341.
Clark, A. S., & Goldman-Rakic, P. (1989). Gonadal hormones
Our lab has documented sex differences in the be- influence the emergence of cortical function in non-human prima-
havior of young children, adolescents, and adults on tes. Behavioral Neuroscience, 103, 1287–1295.
tasks that rely on the integrity of the orbital prefrontal Corbier, P., Edwards, D. A., & Roffi, J. (1992). The neonatal
testosterone surge: A comparative study. Archives Internationales
cortex. Others in this Special Issue have at least some de Physiologie, de Biochimie et de Biophysique, 100, 127–131.
data to support our findings. For the most part, studies Costa, P., & McCrae, R. R. (1992). NEO PI-R revised NEO personality
presented in this issue, did not include analyses for sex inventory. Florida: Psychological Assessment Resources.
differences. Hopefully, these behavioral and imaging Cowell, P. E., Allen, L. S., Zalatimo, N. S., & Dennenberg, V. H.
data will be analyzed for sex differences. (1992). A developmental study of sex and age interactions in the
human corpus callosum. Developmental Brain Research, 66, 187–
Evidence is accumulating that regions of the frontal 192.
cortex, including orbital prefrontal, may be anatomically Cowell, P. E., Turetsky, B. I., Gur, R., Grossman, R. I., Shtasel, D. L.,
and functionally different for males and females. It is & Gur, R. (1994). Sex differences in aging of the human frontal and
unclear at this time how such sex differences may be temporal lobes. Journal of Neuroscience, 14(8), 4748–4755.
manifest in behaviors other than laboratory tasks. Recent Damasio, A. R. (1995). DescantesÕ error. New York. G.P. PutnamÕs
Sons.
evidence indicates that performance on the IGT may be Fishbein, E. (2000). Neuropsychology function, drug use, and violence:
related to several brain areas outside the orbital regions, A conceptual framework. Criminal Justice and Behavior, 27, 139–
e.g., dorsolateral and dorsomedial prefrontal cortices 159.
(Manes et al., 2002). Consequently, the behavioral sex Goldman, P. S. (1971). Functional development of the prefrontal
differences reported in the present data may be related to cortex in early life and the problem of neuronal plasticity.
Experimental Neurology, 32, 366–387.
differential functioning of one or more of these brain Goldman, P. S., Crawford, H. T., Stokes, L. P., Galkin, T., & Rosvold,
systems. H. E. (1974). Sex-dependent behavioral effect of cerebral cortical
Many questions remain: Are such brain differences lesions in the developing rhesus monkey. Science, 186, 540–542.
related to sex differences in emotional behaviors? Are Grant, S., Contoreggi, C., & London, E. D. (2000). Drug abusers show
they related to sex differences in broader real-life deci- impaired performance in a laboratory test of decision-making.
Neuropsychologia, 38, 1180–1187.
sion-making processes? Are the sex differences more or Gur, R. C., Mozley, L., Mozley, P. D., Resnick, S., Karp, J., Alavi, A.,
less ‘‘wired’’ at birth or are they conditioned over the life Arnold, S., & Gur, R. E. (1995). Sex differences in regional cerebral
span? Are sex differences in orbital prefrontal cortex glucose metabolism during a resting state. Science, 267, 528–531.
W.H. Overman / Brain and Cognition 55 (2004) 134–147 147
Hagger, C., Bachevalier, J., & Bercu, B. B. (1987). Sexual dimorphism Overman, W. H., & Godin, M. (unpublished data). Male superiority
in the development of habit formation; effects of perinatal steroidal on object reversal extends to 34 months of age in children
gonadal hormones. Neuroscience, 22(Suppl. S520). (unpublished data).
Knowlton, B. J., Mangels, J. A., & Squire, L. R. (1996). A neostriatal Polich, J., Armor, D., & Braiker, H. (1981). The course of alcoholism:
habit learning system in humans. Science, 273, 1399–1402. Four years after treatment. New York: Wiley.
Manes, F., Sahakian, B., Clark, L., Rogers, R., Antoun, N., Aitken, Reavis, R., & Overman, W. H. (2001). Adult sex differences on a
M., & Robbins, T. (2002). Decision-making processes following decision-making task previously shown to depend on the orbital
damage to the prefrontal cortex. Brain, 125, 624–639. prefrontal cortex. Behavioral Neuroscience, 115, 196–206.
Overman, W. H., & Bachevalier, J. (1999). Inferences about the Rolls, E. T. (1996). The orbital prefrontal cortex. Philosophical
functional development of neural systems in children via the Transactions of the Royal Society of London Biological Sciences,
application of animal tests of cognition. In C. A. Nelson & M. 351, 1433–1444.
Luciana (Eds.), Handbook of developmental cognitive neuroscience Shirtcliff, E. A., Reavis, R., Overman, W. H., & Granger, D. A. (2001).
(pp. 337–357). Cambridge, MA: MIT Press. Measurement of gonadal hormones in dried blood spots versus
Overman, W. H., Bachevalier, J., Schuhmann, E., & Ryan, P. (1996). serum: Verification of menstrual cycle phase. Hormones and
Cognitive gender differences in very young children parallel Behavior, 39, 258–266.
biologically based cognitive gender differences in monkeys. Behav- Stahl, F., Gotz, F., Poppe, I., Amendt, P., & Dorner, G. (1978). Pre-
ioral Neuroscience, 110, 673–684. and early postnatal testosterone levels in rat and human. In G.
Overman, W. H., Bachevalier, J., Schuhmann, E., & Ryan, P. M. Dorner & M. Kawakami (Eds.), Hormones and brain development
(1997). Sexually dimorphic brain-behavior development: A com- (pp. 99–109). New York: Elsevier/North-Holland Biomedical
parative perspective. In N. A. Krasnegor, G. Reid Lyon, & P. S. Press.
Goldman-Rakic (Eds.), Development of the prefrontal cortex: Tomb, I., Hauser, M., Deldin, P., & Caramazza, A. (2002). Do somatic
Evolution, neurobiology, and behavior (pp. 109–124). Baltimore: markers mediate decisions on the gambling task? Nature Neuro-
Paul H Brooks. science, 5, 3–4.
Overman, W. H., Frassrand, K., Ansel, S., Trawalter, S., Bies, B., & Wolford, G., Miller, M. B., & Gazzaniga, M. (2002). The left
Redmond, A. (2003). Performance on the Iowa decision-making hemisphereÕs role in hypothesis formation, Journal of Neuroscience,
task by adolescents and young adults (Under review). 20, RC64, 1–4 (Online publication).