This article was downloaded by: [130.132.173.

39]
On: 24 December 2014, At: 11:17
Publisher: Routledge
Informa Ltd Registered in England and Wales Registered Number: 1072954 Registered office: Mortimer
House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Clinical and Experimental

Neuropsychology
Publication details, including instructions for authors and subscription information:
http://www.tandfonline.com/loi/ncen20

Attentional set shifting in autism spectrum

disorder: Differentiating between the role of
perseveration, learned irrelevance, and novelty
processing
a

Joseph H. R. Maes , Paul A. T. M. Eling , Elke Wezenberg , Constance Th. W. M.

Vissers

a b

& Cees C. Kan

Donders Institute for Brain, Cognition and Behaviour , Radboud University

Nijmegen , Nijmegen, The Netherlands
b

Department of Psychiatry , Radboud University Nijmegen Medical Centre ,

Nijmegen, The Netherlands
Published online: 05 Aug 2010.

To cite this article: Joseph H. R. Maes , Paul A. T. M. Eling , Elke Wezenberg , Constance Th. W. M. Vissers & Cees C.
Kan (2011) Attentional set shifting in autism spectrum disorder: Differentiating between the role of perseveration,
learned irrelevance, and novelty processing, Journal of Clinical and Experimental Neuropsychology, 33:2, 210-217,
DOI: 10.1080/13803395.2010.501327
To link to this article: http://dx.doi.org/10.1080/13803395.2010.501327

PLEASE SCROLL DOWN FOR ARTICLE

Taylor & Francis makes every effort to ensure the accuracy of all the information (the Content)
contained in the publications on our platform. However, Taylor & Francis, our agents, and our licensors
make no representations or warranties whatsoever as to the accuracy, completeness, or suitability
for any purpose of the Content. Any opinions and views expressed in this publication are the opinions
and views of the authors, and are not the views of or endorsed by Taylor & Francis. The accuracy of
the Content should not be relied upon and should be independently verified with primary sources of
information. Taylor and Francis shall not be liable for any losses, actions, claims, proceedings, demands,
costs, expenses, damages, and other liabilities whatsoever or howsoever caused arising directly or
indirectly in connection with, in relation to or arising out of the use of the Content.
This article may be used for research, teaching, and private study purposes. Any substantial or
systematic reproduction, redistribution, reselling, loan, sub-licensing, systematic supply, or distribution
in any form to anyone is expressly forbidden. Terms & Conditions of access and use can be found at
http://www.tandfonline.com/page/terms-and-conditions

JOURNAL OF CLINICAL AND EXPERIMENTAL NEUROPSYCHOLOGY

2011, 33 (2), 210217

NCEN

Attentional set shifting in autism spectrum disorder:

Differentiating between the role of perseveration,
learned irrelevance, and novelty processing

Attentional Set Shifting In Asd

Joseph H. R. Maes,1 Paul A. T. M. Eling,1 Elke Wezenberg,2

Constance Th. W. M. Vissers,1,2 and Cees C. Kan2
1Donders

Downloaded by [] at 11:17 24 December 2014

Institute for Brain, Cognition and Behaviour, Radboud University Nijmegen,

Nijmegen, The Netherlands
2Department of Psychiatry, Radboud University Nijmegen Medical Centre, Nijmegen, The
Netherlands

Autism spectrum disorders (ASD) are associated with impaired attentional set shifting, which may reflect
enhanced perseverative responding, enhanced learned irrelevance, and/or reduced novelty processing. We assessed
the contribution of these potential error sources in ASD adults. A total of 17 ASD and 19 matched comparison
individuals first solved a discrimination learning task. Thereafter, the participants faced three types of attentional
shift, specifically designed to isolate the effect of the three possible error sources. ASD participants made more
errors than comparison individuals in a shift implying a choice between a novel relevant stimulus attribute and a
familiar attribute that was previously relevant but now irrelevant. However, they made fewer errors in a shift
involving a choice between a novel irrelevant attribute and a familiar, previously irrelevant but now relevant
attribute. The results in combination suggest that the performance difference, at least in the present shift task, is
caused by reduced novelty processing in ASD participants.
Keywords: Autism spectrum disorder; Attentional set shifting; Learned irrelevance; Perseveration; Novelty processing.

Autism spectrum disorders (ASD) are pervasive developmental disorders, characterized by aberrant social
interaction and communication, repetitive behaviors,
and insistence on sameness (e.g., Frith & Happ, 2005).
It has been proposed that the core deficit involves cognitive dysfunctions, especially those related to frontallobe-mediated executive functions. Recognizing that the
concept of executive function is an umbrella term, Hill
(2004a, 2004b), in evaluating the theory of executive dysfunction in autism, focused on studies on planning, mental flexibility, and inhibition. It was concluded that
deficits in planning and mental flexibility have been
demonstrated, but that the issue is still far from resolved.
A specific problem appears to be that impaired performance on tests for executive functioning may be caused by
other cognitive deficits. Another review of studies using
a wide variety of executive-functioning tests indicated

that the perseverative response style in adults with

autism, observed in several studies, does not appear to
be related to underlying difficulties in response inhibition, nor can it be attributed to problems in working
memory (Russo et al., 2007). Yet another recent review
specifically focused on cognitive flexibility, taking various testsfor example, the Wisconsin Card Sorting Test
(WCST, Heaton, Chelune, Talley, Kay, & Curtiss, 1993)
and the Cambridge Neuropsychological Test Automated
Battery (CANTAB)and experimental procedures into
account and also looking at various subgroups (Geurts,
Corbett, & Solomon, 2009). It appeared that no consistent evidence for cognitive flexibility deficits in autism
can be found. The authors argued that, to advance the
field, experimental measures must evolve to reflect mechanistic models of flexibility deficits. Finally, using a taskswitching paradigm, with which effects of switching and

Since January 2010 C. Th. W. M. Vissers has been at Centre of Excellence for Neuropsychiatry, Vincent van Gogh Institute for
Psychiatry, Venray, The Netherlands. We thank Leontine Kock and Hanneke Bertens for their help in collecting the data.
Address correspondence to J. H. R. Maes, Radboud University Nijmegen, Donders Institute for Brain, Cognition and Behaviour,
Centre for Cognition, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands (E-mail: r.maes@donders.ru.nl).

2010 Psychology Press, an imprint of the Taylor & Francis Group, an Informa business
http://www.psypress.com/jcen
DOI: 10.1080/13803395.2010.501327

Downloaded by [] at 11:17 24 December 2014

ATTENTIONAL SET SHIFTING IN ASD

restarting and general task performance can be examined separately, Poljac et al. (2010) observed a switching
deficit for children with dyslexia, but not for children
with ASD. Taken together, the evidence for a deficit in
cognitive flexibility is equivocal, and there is a great need
for tasks enabling a more fine-grained decomposition of
processes involved in switching.
The aim of the present experiment was to further
unravel in ASD impaired cognitive flexibility or attentional set shifting, using a novel shift-learning task. A
majority of studies using standard shifting tasks, such as
the WCST, found that ASD participants have more difficulty finding the new rule than do matched comparison
groups (see Geurts et al., 2009, for a review). A common
interpretation of this difficulty is that it reflects perseverative responding: the inability to disengage attention
from, or inhibit responding to, the previously relevant
stimulus attribute that has become irrelevant. However,
continued choice of the wrong attribute after a shift may
just as well be caused by an impaired ability to redirect
ones attention to a previously irrelevant (but currently
relevant) stimulus dimension, a tendency reflecting a
process termed learned irrelevance (Mackintosh, 1975).
In fact, the results of two of our previous studies suggest
that, at least in healthy participants, learned irrelevance
plays a more dominant role than perseveration (Maes,
Damen, & Eling, 2004; Maes, Vich, & Eling, 2006). In
these studies, different types of shift were created in different groups of participants using a two-choice discrimination learning task with multidimensional stimuli.
These shifts enabled the assessment of the separate contribution of perseveration and learned irrelevance. Specifically, in a perseveration (P) group, the previously
relevant stimulus attribute became irrelevant, whereas a
novel stimulus attribute was introduced as now-relevant
attribute. The learned irrelevance mechanism could not
play a role, because the previous irrelevant attribute was
no longer present. Instead, in a learned irrelevance
(LI) group, the previous irrelevant stimulus attribute
became relevant, whereas a novel stimulus attribute
became the irrelevant attribute. Because the former relevant attribute was no longer present, perseverative
responding was impossible. Finally, in a standard shift
group, the previous relevant attribute became irrelevant,
and vice versa, which allowed (at least in principle) both
perseveration and learned irrelevance to affect responding (P+LI shift group). The number of incorrect
choices in the latter group did not differ significantly
from that in the LI group, both of which displayed
more errors than the P group. These results are indicative of learned irrelevance, rather than perseveration,
being the determinant of postshift errors in a standard
shift condition.
To the best of our knowledge, data on learned irrelevance in autism are scarce, and the available data are inconclusive. Two studies that used an attentional-set-shifting
procedure that enabled separating perseveration and
learned irrelevance provided mixed results. Wong,
Maybery, Bishop, Maley, and Hallmayer (2006) examined
set shifting in parents and siblings of ASD individuals
and found that fathers from ASD individuals displayed

211

more learned irrelevance errors, but not perseveration

errors, than did fathers from control persons. Turner
(1997) reported that, relative to control individuals,
learning-disabled ASD participants displayed impaired
performance on a perseveration shift, but not a learned
irrelevance shift, whereas high-functioning ASD individuals did not differ from controls on either type of
shift.
Still another attentional process potentially involved
in attentional set shifting is the processing of novel stimuli.
For example, some shift conditions, such as those created in the P and LI conditions outlined above, imply a
choice between a stimulus that was also present in a previous phasethat is, a familiar stimulusand a novel
stimulus. The number of errors displayed in each of
these conditions may be related to the responsivity that
the participant displays to novel stimuli in general. Specifically, in a P condition, a tendency to not pay much
attention to novel stimuli works against switching to the
current relevant stimulus (which is a novel stimulus). In
contrast, in a LI condition, such a tendency works in
favor of selecting the postshift relevant dimension, which
was previously irrelevant, but familiar.
Notably, children with ASD have been shown to display a reduced responsivity to novel stimulifor example,
as measured by electrophysiological responses (e.g.,
Courchesne, Lincoln, Kilman, & Galambos, 1985;
Goodwin et al., 2006; Orekhova et al., 2009; Van Engeland,
Roelofs, Verbaten, & Slangen, 1991). This, in turn, may
also be related to low novelty seeking, a personality trait
associated with autism (e.g., Anckarster et al., 2006).
Therefore, it is conceivable that the pattern of responding
in attentional-set-shifting tasks is profoundly affected by
a novelty-processing deficit.
In the (very) large majority of prior studies, the cognitive
processes related to perseveration, learned irrelevance,
and impaired novelty processing are either intermixed or
not examined in a counterbalanced manner with respect
to the order in which they are assessed. The present
study aims at improving this examination, by assessing
in a balanced design the relative contribution of perseveration, learned irrelevance, and novelty processing to
the performance of ASD participants in attentional set
shifting. To this end, we used an adapted version of a
task described in Maes and Eling (2009), in which ASD
participants and healthy comparison participants first
learned a discrimination task, followed by P, LI, and
P+LI shifts (order randomly determined). A number of
hypotheses can be formulated, based on different
assumptions regarding the main cognitive deficit underlying set-shifting performance in ASD (see Table 1 for an
overview). If impaired novelty processing is the dominant process affecting performance in this task, we must
expect more errors in the P shift for the ASD participants
than for the comparison group. As outlined above,
reduced responding to novelty implies sticking to the
now irrelevant, but familiar, stimulus attribute. At the
same time, we must then also expect fewer errors in the
LI shift for the ASD than for the comparison group
because a reduced responsivity to the novel stimulus
attribute implies attention for the now relevant and

212

MAES ET AL.

METHOD

TABLE 1
Overview of assumptions and corresponding hypotheses
about number of errors in the different types of shift
Assumption concerning
affected process in ASD

Participants
The participants consisted of 23 ASD and 20 healthy
comparison (COMP) individuals. A total of 6 ASD participants and 1 COMP participant were unable to solve
the preshift training phase of the discrimination learning
task within 80 trials. The data from these participants
were excluded from further analysis, because a meaningful interpretation of the shift data is only possible for
participants that were able to solve the initial, preshift
task. The diagnoses of the participants in the ASD group
had been made clinically, according to the DSMIV
(Diagnostic and Statistical Manual of Mental Disorders
Fourth Edition; American Psychiatric Association, 1994),
at the Department of Psychiatry of the Radboud University Nijmegen Medical Centre, The Netherlands, by
means of a thorough psychiatric examination, including
a collateral diagnostic developmental interview (in most
cases with the parents). A Dutch version of the Autism
Spectrum Quotient (AQ; Baron-Cohen, Wheelwright,
Skinner, Martin, & Clubley, 2001; see below) was always
administered during the diagnostic procedure. The ASD
group consisted of 7 individuals with Aspergers disorder,
5 with PDD-NOS (pervasive developmental disorder
not otherwise specified), and 5 with autistic disorder (the
subtypes were determined on the basis of the DSMIV
criteria and psychiatric examination). All ASD participants lived in the Nijmegen region, whereas the matched
COMP participants were selected across The Netherlands,
partly among visitors of the Radboud University
Nijmegen Medical Centre. After being informed about
the basic testing procedures, all individuals agreed to
take part in the study and received a small financial
reward for their participation. The participants were
individually tested in a quiet room, either in the psychiatric
outpatient department or in the home environment.
Demographic and clinical data for each group are shown
in Table 2 (see below for more information on the
neuropsychological test performance).

Hypotheses

Impaired novelty processing

P shift: ASD > COMP

LI shift: ASD < COMP
P+LI shift: ASD = COMP

Increased perseverative
tendencies

P shift: ASD > COMP

LI shift: ASD = COMP
P+LI shift: ASD = COMP

Decreased or increased
learned irrelevance

P shift: ASD = COMP

LI shift: ASD > or <

COMP
P+LI shift: ASD > or <
COMP

Downloaded by [] at 11:17 24 December 2014

Note. ASD = autism spectrum disorder individuals. COMP

= comparison participants. P = perseveration. LI = learned
irrelevance.

familiar attribute. The P+LI condition does not imply a

novel stimulus attribute. Hence, in this case we may
expect no group performance difference. Instead, if ASD
participants mainly differ from the comparison group
in the strength of perseverative tendencies, we may
expect the former participants to make more errors in
the P shift than the latter, with no difference in LI-shift
performance. Moreover, at least in healthy participants, learned irrelevance, rather than perseveration, is
the dominant process affecting performance in the
combined P+LI shift. If this also holds for ASD participants, we must then expect no difference between
groups with respect to this type of shift. Finally, if the
major difference between the participant groups concerns the learned irrelevance process, given the lack of
conclusive data on this issue we may expect either
fewer or more errors in the LI and the P+LI shift conditions in the ASD than in the comparison participants, whereas there will be no difference in the P
condition.

TABLE 2
Demographic and clinical data
COMP (n = 19: 15M/4F)

Age
Education level
AQtot*
Raven
NLV
WCST N categories
WCST % P errors

SD

Minmax

38.4
3.1
11.7
11.0
89.4
4.8
11.7

14.8
0.9
5.4
1.4
8.0
1.9
5.9

2258
24
021
712
66100
06
424

ASD (n = 17: 13M/4F)

M

SD

Minmax

45.2
2.9
34.3
11.0
90.2
4.9
11.4

13.1
1.1
8.7
1.2
7.8
1.8
7.6

2166
14
2047
912
8098
16
527

Note. ASD = autism spectrum disorder individuals. COMP = comparison participants. M = male; F = female. AQ = Autism Spectrum
Quotient. NLV = Dutch version of the National Adult Reading Test. WCST = Wisconsin Card Sorting Test; N categories = number of
finished categories; %P = percentage of perseverative errors. In the ASD group, n = 9 for the Raven and NLV scores. Education
level: 1 = primary school; 4 = university. This scale (14) corresponds to approximately 9, 11, 15, and 18 years of education, respectively.
*p < .001.

ATTENTIONAL SET SHIFTING IN ASD

gradually expose the participant to the four exemplars of

the task-relevant stimulus attribute, to maximize the chance
of ultimately solving the preshift Stage 3 (see below).

Materials and procedure

Discrimination learning task
All participants performed a discrimination learning
task (DLT), which was programmed using E-prime software (Psychology Software Tools, Inc., Pittsburg, USA),
run on a laptop with a 15-inch monitor. The task was a
slightly adapted version of the task used by Maes and
Eling (2009). The task consisted of several stages. Prior
to initiating Stage 1, the participant, seated in front of
the computer screen, read the following instructions
(translated from Dutch):

Downloaded by [] at 11:17 24 December 2014

213

In this experiment, you will repeatedly see two figures on

the screen. One of these figures is correct; the other is
incorrect. There are rules that determine which figure is
correct. Your task is to discover these rules as quickly as
possible. Press Z if you think that the left figure is correct;
press M if you think that the right figure is correct.
Try to make your choice as quickly as possible. After
each choice, the computer will tell you whether your
choice was correct or incorrect. The rules will change
during the experiment. The rules will be very simple.
Next, you will see an example in which the filled figure is
the correct one. Press Z to choose the left figure; press
M to choose the right figure.

Next, two practice trials were presented. A picture consisting of a black square (left) and an open circle (right)
on Trial 1 and a picture of a black square (right) and an
open triangle (left) on Trial 2 were presented. After the
participant had made a response to each of these trials
and had received feedback about the correctness of the
responses, the following text appeared:
Now the real experiment will start. Please try to find the
rules as quickly as possible. The rules will be as simple as
in the example. Keep in mind that the rules will change
during the test.

The following stages (Stages 16) were presented sequentially without interruption. Stages 1 and 2 were presented to

Stage 1: Preshift Learning 1. Each trial consisted of

the presentation of two stimuli. One stimulus was presented on the left and one on the right side of the screen.
Each stimulus consisted of 14 identical geometrical figures that are described in detail in Maes et al. (2004).
Stimuli varied on three attributes: color, shape, and
quantity. We used the following colors: purple, yellow,
green, and blue. For shapes we used triangle, hexagon,
square, and circle, and quantity varied from 1 to 4 elements. Within this and subsequent stages, two attributes
varied. One attribute was the task-relevant and the other
the task-irrelevant attribute, while the third attribute was
held constant, and stimuli did not vary on this attribute.
For example, all stimuli were presented in blue if the
attribute color was held constant. Figure 1 shows
examples of stimulus pairs used in the different phases of
the experiment (Figure 1 focuses on Stages 36, but similar stimulus pairs were used in Stages 1 and 2).
If color was task relevant, purple and yellow indicated
a correct response (i.e., had to be selected); for shape, triangle and hexagon indicated a correct response, whereas
2 and 3 were correct if quantity was task relevant. From
here on, we refer to a correct exemplar to indicate the
exemplar that had to be chosen in order to receive positive feedback and to an incorrect exemplar to refer to the
exemplar that would yield negative feedback if chosen.
Within each trial, one of the stimuli contained a correct
exemplar of the relevant attribute, whereas the other
stimulus contained an incorrect exemplar of the relevant
attribute. The leftright position of the correct and
incorrect exemplars was counterbalanced across trials.
Each stimulus also contained an exemplar of the taskirrelevant attribute, the specific exemplar of which was
counterbalanced across trials. Counterbalancing the specific identity of task-relevant and task-irrelevant stimulus
attributes in this and the subsequent stages of the experiment, as well as counterbalancing the order in which the

Shift 1

Shift 2

Shift 3

Stage 3: Pre-shift

Stage 4: P-shift

Stage 6: P+LI-shift

Rel: Shape
Irrel.: Quantity

Rel: Colour
Irrel.: Shape

Stage 5: LI-shift
Rel: Shape
Irrel.: Quantity

Rel: Quantity
Irrel.: Shape

Figure 1. Example of stimuli presented during Phases 36. Within each phase, stimuli varied in two of three stimulus attributes (color,
shape, or quantity), with one attribute being relevant for solving the task (Rel.) and the other one being irrelevant (Irrel.). The correct
(to-be-chosen) stimulus of each pair is indicated by a plus sign. Spatial location (left or right) of the correct stimulus was counterbalanced within each phase. The nature of relevant and irrelevant attributes and the order of the different types of shift (P, LI, and P+LI)
were counterbalanced using different task versions. To view a color version of this figure, please see the online issue of the Journal.

214

MAES ET AL.

different types of shift were presented (Stages 46, see

below), implied 36 task versions. Each participant was
randomly assigned to 1 task version. In Stage 1 (Preshift
Learning 1), two stimulus attributes were varied: two
exemplars of the relevant attribute (one correct and the
other incorrect) and four exemplars of the irrelevant
attribute. Stage 2 commenced after the participant had
made eight successive correct choices.

Downloaded by [] at 11:17 24 December 2014

Stage 2: Preshift Learning 2. In this stage, the other

two exemplars of the relevant stimulus dimension (again
one correct and one incorrect), together with the four
exemplars of the irrelevant dimension, were presented.
All further details were as in Stage 1. Stage 3 started
after a correct response on eight successive trials.
Stage 3: Preshift Learning 3. The final part of the
preshift phase consisted of a random mix of the trial
types presented during Stages 1 and 2. Thus, all four
exemplars of the relevant and irrelevant exemplars were
used. The next stage started after eight successive correct
responses.
Stage 4: Shift 1. A shift was introduced in this stage
(Shift 1). The shift was a P, LI, or P+LI shift. In the case
of a P shift, the previously relevant stimulus attribute
became irrelevant, and the attribute that had been constant in the preshift phase became relevant. If the shift
concerned a LI shift, the previous irrelevant attribute
became relevant, and the previous constant attribute
became the irrelevant attribute. Finally, in case of a
P+LI shift, the previous relevant attribute became irrelevant and vice versa. This phase, as well as each of the
next two phases, was terminated after the participant
had made eight consecutive correct responses or after 80
trials, whichever came first.
Stages 5 and 6: Shifts 2 and 3. A different type of shift
was introduced in each of the next two stages of the
experiment, with the immediately preceding stage as the
basis for defining the type of shift. For example (see also
Figure 1), if in a specific version of the task, Shift 1 was a
P shift, Shift 2 a LI shift, and Shift 3 a P+LI shift, the
stimulus attribute that had been irrelevant in Shift 1
became relevant in Shift 2 and again irrelevant in Shift 3.
Likewise, in this example, the relevant attribute from
Shift 1 was held constant in Shifts 2 and 3.
The dependent measure for each participant consisted
of the number of errors in Stage 3 (preshift), and the Pshift, LI-shift, and P+LI-shift stages.
Neuropsychological tests
The discrimination learning task described above was
part of a larger test battery, which included a Dutch version of the AQ and a computerized version of the WCST
(Heaton & Goldin, 2003). Initially, the NLV, a Dutch
version of the National Adult Reading Test (Schmand,
Lindeboom, & Harskamp, 1992), and Set I from the
Raven Advanced Progressive Matrices Test (APM;
Raven, Raven, & Court, 1998) were not part of the

standard diagnostic procedure and were added to the

battery at a later stage of the study. For this reason, the
NLV and APM data, which were used as intelligence
indices, were only available for 9 of the 17 ASD participants. However, in both the COMP and ASD groups
there was a strong and significant correlation between
education level and NLV score, Spearmans = .64, and
= .82, respectively, ps < .01, suggesting that education
level can be used as a reliable index of intelligence. Given
the clear absence of a between-group education level difference, these results suggest the absence of a significant
difference in general intelligence. Moreover, scores on
the descriptive characteristics and main performance
variables did not significantly differ between ASD individuals for which NLV and AMP scores were available
and ASD participants for which these data were lacking.
From the AQ questionnaires, we report the AQ total
score (an overall score measuring frequency of autistic
behavior, maximum: 50). From the WCST measures, we
only report the number of finished categories and the
percentage of perseverative errors. As performance
index on the NLV and APM, we used the raw score and
the number of correct items, respectively.
Data analysis
The primary analysis consisted of a Group (ASD vs.
COMP participants) Shift Type (preshift, P, LI, or
P+LI) repeated measures analysis of variance (ANOVA)
on the number of errors (trials with an incorrect
response), followed by simple main effect analyses. In
the latter analyses, we used the error term and degrees of
freedom based on the overall analysis, according to Satterthwaites approximation (Winer, 1971). Further analyses determined Spearmans rho correlation among the
percentage of WCST perseveration errors and the
number of P-shift, LI-shift, and P+LI-shift errors. The
level of significance was set at p < .05 in all statistical tests.
RESULTS
Demographic and clinical data for each group are shown
in Table 2. The two groups did not differ on any of the
variables shown in this table, ANOVA, Fs < 2.11, except
on the AQ total score, F(1, 36) = 89.34, p < .001. The
ASD diagnosis was only partly based on the AQ, and 3
participants with this diagnosis had a score below 26
(namely, 20, 23, and 25), which is suggested to be a valid
cutoff (Woodbury-Smith, Robinson, Wheelwright, &
Baron-Cohen, 2005). However, importantly, removing
the data of these 3 participants from the analysis did not
result in any changes in the pattern of main results
reported in this study.
Figure 2 displays, separately for COMP and ASD participants, the mean number of errors during the preshift
phase (Stage 3) and during each of the three types of shift.
ANOVA on the error frequencies revealed a significant
shift type effect, F(3, 102) = 7.89, p < .001, and a significant Group Shift Type interaction, F(3, 102) = 2.76,
p < .05. The group main effect was not significant, F < 1.
Subsequent main effect analyses revealed that the groups

ATTENTIONAL SET SHIFTING IN ASD

Mean Errors (+SEM)

25

COMP
ASD

20
15
10
5
0

Downloaded by [] at 11:17 24 December 2014

Pre

P-Shift

LI-Shift

P+LI-Shift

Figure 2. Groups mean number of errors (+ SEM) during the

preshift phase and during each of the three types of shift introduced after this phase. ASD = autism spectrum disorder individuals. COMP = comparison participants. P = perseveration;
LI = learned irrelevance. Note that, relative to the preshift
phase of the task, the ASD participants made significantly more
errors during the P shift, but not during the LI shift. Instead,
again relative to the preshift phase, the comparison participants
made significantly more errors during the LI shift, but not
during the P shift. Both groups displayed more errors during
the P+LI shift than during the preshift phase.

differed significantly in the number of P-shift errors,

F(1, 128) = 5.30, p < .05, but not in the number of errors
during any of the other stages, Fs < 1.89. However, relative
to the number of preshift errors, the groups clearly differed in their response on both the P and LI shifts. That
is, for COMP participants the number of errors during
the LI shift was larger than that during the preshift
phase, F(1, 102) = 11.39, p < .01, whereas there was no
difference between the number of errors in the P-shift
and preshift phases, F(1, 102) = 3.79. Instead, the ASD
group displayed the reverse pattern of errors: more
errors during the P-shift than the preshift phase, F(1, 102) =
12.51, p < .01, and no difference between the LI-shift
and preshift phases, F(1, 102) = 1.03. Both groups made
more errors during the P+LI-shift phase than during the
preshift phase, Fs(1, 102) > 7.10, ps < .01.
The percentage of perseveration errors made during
the WCST was significantly and positively correlated
with the number of P-shift errors during the DLT, = .34,
p < .05, but not significantly correlated with the number
of LI-shift or P+LI-shift errors, = .05 and = .19,
respectively. The number of P+LI-shift errors was positively correlated with the number of LI-shift errors, = .35,
p < .05, but not with the number of P-shift errors, = .10.
DISCUSSION
For some time now, studies have attempted to demonstrate that a deficit in executive functioning is a core feature of autism. Various reviews have shown that results
are inconclusive (Geurts et al., 2009; Hill, 2004a, 2004b;
Russo et al., 2007). While there may be a problem in set
shifting, the underlying mechanisms are not clear, and
Geurts et al. (2009) have suggested that other paradigms
need to be developed in order to analyze these problems

215

in more detail. The aim of the current study was performed in this framework. Specifically, we used a paradigm that allowed us to differentiate between various
factors that may prevent patients with autism to flexibly
shift attentional set: perseveration, learned irrelevance,
and reduced novelty processing.
Relative to their preshift performance, the COMP
group had more difficulty solving the LI shift than the P
shift, whereas the reverse held for the participants in the
ASD group. Both COMP and ASD groups made more
errors during the combined, P+LI, shift than during the
preshift phase. The collective results can be most parsimoniously explained by assuming that, given a choice
between a relatively novel stimulus attribute and a familiar attribute, the ASD participants (regardless of specific
subclass) were more inclined to choose the familiar
attribute than the COMP participants (see first assumption
and corresponding hypotheses in Table 1). This tendency
is favorable for performing the LI shift, in which the
familiar attribute is task relevant, but hinders adequate
performance of the P shift, in which the more familiar
attribute is irrelevant.
The number of errors during the P+LI shift was correlated with the number of LI-shift, but not P-shift, errors,
suggesting that the main process involved in the P+LI
shift was learned irrelevance, a conclusion that is consistent with our previous studies using between- and withinsubject versions of our discrimination learning task
(Maes et al., 2004; Maes et al., 2009; Maes et al., 2006).
Accordingly, the absence of a group difference in errors
during the P+LI shift (also relative to preshift responding: both groups showing more P+LI-shift than preshift
errors) suggests the absence of a difference in learned
irrelevance. Therefore, the group difference regarding
performance during the LI shift can be explained by
assuming a reduced novelty processing in the ASD
group. This same tendency to avoid novel attributes
hampers the ASD participants in the P shift, which suffices
to explain the group difference in P-shift performance.
Therefore, the data also do not demand the presumption
of a between-group difference in perseveration.
The percentage of WCST perseveration errors was
significantly (but relatively weakly) correlated with the
P-shift errors during the DLT, suggesting at least one
common underlying process. According to our interpretation of the joint data described above, this process
might be linked to novelty processing. At the same time,
there was no significant correlation between WCST perseveration and P+LI-shift performance. At first sight,
this is surprising, because the WCST shifts are conceptually identical to the P+LI shift in the discrimination
learning task: Both types of shift imply that a previously
relevant stimulus attribute becomes relevant, and vice
versa. However, given the correlation differences, some
feature of the WCST shifts must make these shifts more
similar to the P than to the P+LI shift in the DLT. Possibly, this is related to the fact that the WCST implies the
presence of 64 different stimuli (target cards), which are
unique with respect to attribute exemplars (but not with
respect to stimulus attributes). Instead, in the P+LI shift
the stimuli remained identical to those presented during

Downloaded by [] at 11:17 24 December 2014

216

MAES ET AL.

the immediately preceding phase. These differences

imply a larger novelty of the stimuli in the WCST than
in the P+LI shift, but both these tasks still imply a
stronger stimulus familiarity than the P shift, in which a
complete stimulus attribute, not present in the immediately preceding phase, was suddenly introduced. On this
account, the P-shift condition is more sensitive for differentiating between ASD and COMP participants than is
both the WCST perseveration measure and the P+LI shift
in the DLT task, which accords with the empirical results.
The majority of studies using the WCST found
impaired performance in ASD individuals relative to
matched comparison participants (Geurts et al., 2009).
The fact that we did not find a difference using this task
may be due to specific characteristics of our ASD sample.
Our sample consisted of ASD adults with normal (verbal) intelligence, and at least some studies suggest that,
compared to matched controls, this population does not
show impaired WCST performance (e.g., Hill & Bird,
2006; Liss et al., 2001; but see, e.g., Ambery, Russell,
Perry, Morris, & Murphy, 2006; Lopez, Lincoln,
Ozonoff, & Lai, 2005). Moreover, the use of a computerized test version (as in our study) may have decreased
further the chance of finding differences on a number of
WCST performance measures (e.g., Ozonoff, 1995;
Pascualvaca, Fantie, Papageorgiou, & Mirsky, 1998).
However, our results suggest that differences might even
be found in this type of subgroup when using tasks enabling a more fine-grained analysis of cognitive component processes. Conversely, our results also suggest that
in previous studies that did find WCST performance differences, these differences might reflect cognitive processes other than those typically assumed to be involved
(such as perseveration).
So far, we have explained the data in terms of differences in novelty processing, claiming that ASD participants tend to avoid novelty or show reduced novelty
processing. However, we cannot exclude the possibility
that the difference between participant groups involves a
difference in preference for familiarity. Both enhanced
novelty avoidance and enhanced familiarity preference
would result in more errors in the perseveration shift
condition and fewer errors in the learned irrelevance
shift condition. Possibly, responses to novelty and familiarity reflect two separate processes, an assumption for
which there is some support based on neurophysiological studies (e.g., Xiang & Brown, 1998). Moreover, Gustafsson and Paplinski (2004) used an artificial neural
network approach to investigate different theories on the
development of autism. Specifically, using simulations
within a self-organizing network, they examined to what
extent attention-shift impairments, novelty avoidance,
and/or familiarity preference can best model impairments in the categorization of two-dimensional stimuli
that are characteristic of autism. The authors concluded
that familiarity preference could best account for the
development of abnormal networks, showing autisticlike learning characteristics. However, clearly much
more experimental and clinical work is required to assess
which of the two possibilities, novelty avoidance or
familiarity preference, is the more important one.

Irrespective of these speculations, the present results

suggest that reduced novelty processing or enhanced
familiarity preference (a phenomenon implying learning
about stimuli), rather than the more commonly suggested attentional-shifting impairment or perseverativeresponding mechanisms, may profoundly affect the performance of ASD participants in tasks aimed at measuring attentional set shifting. More generally, the present
results underscore the necessity to carefully examine
component processes underlying task performance in
executive function tasks and to develop tasks that can be
used for this purpose. Failure to do so may seriously
undermine the validity of claims about the presence
or absence of impaired cognitive processes in patient
populations that are made on the basis of test results.
Original manuscript received 19 February 2010
Revised manuscript accepted 7 June 2010
First published online 5 August 2010

REFERENCES
Ambery, F. Z., Russell, A. J., Perry, K., Morris, R., & Murphy,
D. G. M. (2006). Neuropsychological functioning in adults
with Asperger syndrome. Autism, 10, 551.
American Psychiatric Association. (1994). Diagnostic and statistical manual of mental disorders (4th ed.). Washington,
DC: Author.
Anckarster, H., Stahlberg, O., Larson, T., Hakansson, C.,
Jutblad, S.-B., Niklasson, L., et al. (2006). The impact of
ADHD and autism spectrum disorders on temperament,
character, and personality development. American Journal
of Psychiatry, 163(7), 12391244.
Baron-Cohen, S., Wheelwright, S., Skinner, R., Martin, J., &
Clubley, E. (2001). The Autism Spectrum Quotient (AQ):
Evidence from Asperger syndrome/high-functioning autism,
males and females, scientists and mathematicians. Journal of
Autism and Developmental Disorders, 31(1), 517.
Courchesne, E., Lincoln, A. J., Kilman, B. A., & Galambos, R.
(1985). Event-related brain potential correlates of the processing of novel visual and auditory information in autism.
Journal of Autism and Developmental Disorders, 15(1), 5576.
Frith, U., & Happ, F. (2005). Autism spectrum disorder. Current Biology, 15(19), 8690.
Geurts, H. M., Corbett, B., & Solomon, M. (2009). The paradox of cognitive flexibility in autism. Trends in Cognitive
Sciences, 13(2), 7482.
Goodwin, M. S., Groden, J., Velicer, W. F., Lipsitt, L. P.,
Baron, M. G., Hofmann, S. G., et al. (2006). Cardiovascular
arousal in individuals with autism. Focus on Autism and
Other Developmental Disabilities, 21(2), 100123.
Gustafsson, L., & Paplinski, A. P. (2004). Self-organization of
an artificial neural network subjected to attention shift
impairments and familiarity preference, characteristics studied in autism. Journal of Autism and Developmental
Disorders, 34(2), 189198.
Heaton, R. K., Chelune, G. J., Talley, J. L., Kay, G. G., & Curtiss, G.
(1993). Wisconsin Card Sorting Test manual: Revised and
expanded. Odessa, FL: Psychological Assessment Resources.
Heaton, R. K., & Goldin, J. N. (2003). Wisconsin Card Sorting
Test: Computer Version 4 Research Edition. Lutz, FL: Psychological Assessment Resources.
Hill, E. L. (2004a). Evaluating the theory of executive dysfunction in autism. Developmental Review, 24(2), 189233.
Hill, E. L. (2004b). Executive dysfunction in autism. Trends in
Cognitive Sciences, 8(1), 2632.
Hill, E. L., & Bird, C. M. (2006). Executive processes in
Asperger syndrome: Patterns of performance in a multiple
case series. Neuropsychologia, 44, 28222835.

Downloaded by [] at 11:17 24 December 2014

ATTENTIONAL SET SHIFTING IN ASD

Liss, M., Fein, D., Allen, D., Dunn, M., Feinstein, C., Morris,
R., et al. (2001). Executive functioning in high-functioning
children with autism. Journal of Child Psychology and
Psychiatry, 42(2), 261270.
Lopez, B. R., Lincoln, A. J., Ozonoff, S., & Lai, Z. (2005).
Examining the relationship between executive functions and
restricted, repetitive symptoms of autistic disorder. Journal
of Autism and Developmental Disorders, 35(4), 445460.
Mackintosh, N. J. (1975). A theory of attention: Variations in
the associability with reinforcement. Psychological Review,
82(4), 276298.
Maes, J. H. R., Damen, M. D. C., & Eling, P. A. T. M. (2004).
More learned irrelevance than perseveration errors in rule
shifting in healthy subjects. Brain and Cognition, 54(3),
201211.
Maes, J. H. R., & Eling, P. A. T. M. (2009). Do learned irrelevance and perseveration play a role during discrimination
learning? Learning and Motivation, 40(3), 274283.
Maes, J. H. R., Vich, J., & Eling, P. A. T. M. (2006). Learned
irrelevance and response perseveration in a total change
dimensional shift task. Brain and Cognition, 62(1), 7479.
Orekhova, E. V., Stroganova, T. A., Prokofiev, A. O., Nygren,
G., Gillberg, C., & Elam, M. (2009). The right hemisphere
fails to respond to temporal novelty in autism: Evidence from
and ERP study. Clinical Neurophysiology, 120(3), 520529.
Ozonoff, S. (1995). Reliability and validity of the Wisconsin
Card Sorting Test in studies of autism. Neuropsychology,
9(4), 491500.
Pascualvaca, D. M., Fantie, B. D., Papageorgiou, M., & Mirsky,
A. F. (1998). Attentional capacities in children with autism:
Is there a general deficit in shifting focus? Journal of Autism
and Developmental Disorders, 28(6), 467478.
Poljac, E., Simon, S., Ringlever, L., Kalcik, D., Groen, W. B.,
Buitelaar, J. K., et al. (2010). Impaired task switching performance in children with dyslexia but not in children with

217

autism. Quarterly Journal of Experimental Psychology,

63(2), 401416.
Raven, J., Raven, J. C., & Court, J. H. (1998). Raven manual
Section 4: Advanced progressive matrices. Oxford, UK:
Oxford Psychologists Press.
Russo, N., Flanagam, T., Iarocci, G., Berringer, D., Zelazo, P. D.,
& Burack, J. A. (2007). Deconstructing executive deficits
among persons with autism: Implications for cognitive
neuroscience. Brain and Cognition, 65(1), 7786.
Schmand, B., Lindeboom, J., & Harskamp, F. (1992). NLV:
Nederlandse Leestest voor volwassenen: Handleiding [NLV:
Dutch Reading Test for Adults: Manual]. Lisse, The Netherlands: Swets & Zeitlinger.
Turner, M. (1997). Towards an executive dysfunction account
of repetitive behavior in autism. In J. Russell (Ed.), Autism
as an executive disorder (pp. 57100). New York: Oxford
University Press.
Van Engeland, H., Roelofs, J. W., Verbaten, M. N., & Slangen, J. L.
(1991). Abnormal electrodermal reactivity to novel visual stimuli in autistic children. Psychiatry Research, 38(1), 2738.
Winer, B. J. (1971). Statistical principles in experimental design
(2nd ed.). New York: McGraw-Hill.
Wong, D., Maybery, M., Bishop, D. V. M., Maley, A., &
Hallmayer, J. (2006). Profiles of executive function in parents
and siblings of individuals with autism spectrum disorders.
Genes, Brain and Behavior, 5(8), 561576.
Woodbury-Smith, M. R., Robinson, J., Wheelwright, S., &
Baron-Cohen, S. (2005). Screening adults for Asperger syndrome using the AQ: A preliminary study of its diagnostic
validity in clinical practice. Journal of Autism and Developmental Disorders, 35(3), 331335.
Xiang, J.-Z., & Brown, M. W. (1998). Differential neuronal
encoding of novelty, familiarity and recency in regions of
the anterior temporal lobe. Neuropharmacology, 37(45),
657676.