Content uploaded by Emilia Ivanova Barakova
Author content
All content in this area was uploaded by Emilia Ivanova Barakova on Apr 05, 2016
Content may be subject to copyright.
ORIGINAL PAPER
Improving Collaborative Play Between Children with Autism
Spectrum Disorders and Their Siblings: The Effectiveness
of a Robot-Mediated Intervention Based on Lego
Ò
Therapy
Bibi Huskens
•
Annemiek Palmen
•
Marije Van der Werff
•
Tino Lourens
•
Emilia Barakova
Published online: 28 November 2014
Ó Springer Science+Business Media New York 2014
Abstract The aim of the study was to investigate the
effectiveness of a brief robot-mediated intervention based
on Lego
Ò
therapy on improving collaborative behaviors
(i.e., interaction initiations, responses, and play together)
between children with ASD and their siblings during play
sessions, in a therapeutic setting. A concurrent multiple
baseline design across three child–sibling pairs was in
effect. The robot-intervention resulted in no statistically
significant changes in collaborative behaviors of the chil-
dren with ASD. Despite limited effectiveness of the
intervention, this study provides several practical implica-
tions and directions for future research.
Keywords ASD Children Robot-intervention Lego
Ò
therapy Collaborative play
Introduction
Children with autism spectrum disorders (ASD) show
impairments in social reciprocity, eye contact, shared
interests and enjoyment, and interpreting social cues
(Weiss and Harris 2001). These social impairments affect
their interactions with other children. For example, during
free play, they show more parallel play than collaborative
play compared to typical developing children (Bauminger
et al. 2008) and during games and social activities they
show problems in initiating and maintaining interactions
with peers (Bauminger et al. 2003). Given these charac-
teristics, interventions are needed to improve the collabo-
rative skills of children with ASD and to practice working
and negotiating with peers (Ben-Sasson et al. 2013).
Lego
Ò
therapy is an intervention that aims at improving
skills to initiate and maintain interactions and is based on
collaborative Lego
Ò
play (LeGoff 2004; Owens et al.
2008). Specific target skills are verbal and non-verbal
communication (e.g., self-initiated interactions), turn-tak-
ing, sharing, reciprocity, and collaborative problem solv-
ing. The results of several studies indicate that Lego
Ò
therapy is a promising intervention in improving the initi-
ations and duration of social contact with peers in children
with ASD (LeGoff 2004; LeGoff and Sherman 2006;
Owens et al. 2008). Until now, research on the effective-
ness of Lego
Ò
therapy mainly focused on therapy groups in
which only children with ASD participated. No Lego
Ò
therapy studies are conducted in which social skills were
trained in interactions between children with ASD and
typical developing peers or siblings.
Robotic interventions are also used to improve the
interaction skills of children with ASD. A robot can model
social behavior, respond to a child, or mediate social
behavior between children (Scassellati et al. 2012). Studies
Electronic supplementary material The online version of this
article (doi:10.1007/s10803-014-2326-0) contains supplementary
material, which is available to authorized users.
B. Huskens (&) A. Palmen
Dr. Leo Kannerhuis, Center for Autism, P.O. Box 62,
6865 ZH Doorwerth, The Netherlands
e-mail: b.huskens@leokannerhuis.nl
A. Palmen M. Van der Werff
Department of Special Education, Radboud University
Nijmegen, P.O. Box 9104, 6500 HE Nijmegen, The Netherlands
T. Lourens
TiViPe, Kanaaldijk ZW 11, 5706 LD Helmond, The Netherlands
E. Barakova
Eindhoven University of Technology, P.O. Box 513,
5600 MB Eindhoven, The Netherlands
123
J Autism Dev Disord (2015) 45:3746–3755
DOI 10.1007/s10803-014-2326-0
involving robots have focused on a range of social target
behaviors such as imitation (e.g., Duquette et al. 2008),
basic social interaction skills (e.g., Robins et al. 2004a) and
joint attention (e.g., Robins et al. 2004b). In their review on
the clinical use of robots for children with ASD, Diehl et al.
(2012) concluded that, it is difficult to draw firm conclu-
sions on effectiveness, because most studies are only
exploratory and have methodological limitations. Further-
more, as most studies on robotic interventions in children
with ASD involve qualitative reports, there is a need for
studies providing quantitative measures (Scassellati et al.
2012). Recently, Huskens et al. (2013) investigated the
differential effectiveness of an intervention that was con-
ducted by a robot and a human trainer. A concurrent
multiple baseline design and quantitative measures were
used and it could be concluded that the robot-intervention
was effective in increasing the self-initiated questions of
children with ASD. For future research, Huskens et al.
(2013) suggested to deploy robots as mediators to enhance
the social interaction between a child with ASD and others
(e.g., parents, peers, siblings) and to assess the social
validity of robot-mediated interventions.
Parents of children with ASD often report difficulties in
play between their children (Ferraioli et al. 2012). Interac-
tions between a child with ASD and his/her sibling are often
more negative than the interactions between two typically
developing siblings. In improving the interactions between
children with ASD and their siblings, several types of
interventions have been used. In the study of Baker (2000),
for example, improvements in interactions with siblings were
the result of teaching children with ASD play interactions
based on their thematic ritualistic behavior. Also, sibling
mediated interventions have been used in improving inter-
actions (e.g., Tsao and Odom 2006; Walton and Ingersoll
2012). In such interventions, siblings of children with ASD
are taught strategies in promoting social interactions of their
brother or sister with ASD. For example, Walton and In-
gersoll (2012) used a sibling-implemented reciprocal imita-
tion training in improving imitation and joint engagement in
four boys with ASD, aged 45–57 months (3;9–4;9 years).
Although all of the children with ASD showed some
improvements during treatment, skill gains were found to be
inconsistent across children.
Given the upcoming evidence for both robotic inter-
ventions and Lego
Ò
therapy in improving social skills of
children with ASD, these two principles are combined in
the current study. The aim of our study was to investigate
the effectiveness of a brief robot-mediated intervention
based on Lego
Ò
therapy between children with ASD and
their siblings, during play sessions. We hypothesized that
the combined intervention would improve the collaborative
behaviors (i.e., initiations, responses and play together) of
the children with ASD.
Methods
Participants
Three pairs of children participated in the study; a pair
consisted of a child diagnosed with ASD and his/her sibling.
Inclusion criteria for the children with ASD were: (a) age
5–13 years, (b) a full-scale IQ above 80, (c) an ASD diag-
nosis according to the DSM-IV criteria (American Psychi-
atric Association 2000) (clinical judgment) and (d) not
participating in other interventions on social play skills or
peer/sibling interaction during the intervention period.
Inclusion criteria for the siblings were: (a) a typical
developing (TD) child who lives at home, (b) age
5–13 years, having a maximum age difference of 5 years
with their brother or sister with ASD (to reduce possible
influences of developmental age differences), and (c) vis-
iting a regular elementary or regular secondary school.
Table 1 displays the characteristics of the included
participants. Pair 1 consisted of two twin brothers. For all
children the Social Communication Questionnaire (SCQ:
Rutter et al. 2003; Dutch translation: Warreyn et al. 2004)
was filled in by parents. The SCQ-score of Chris (pair 2;
score 14) did not meet the ASD cut-off score of 15.
However, his diagnosis was confirmed by a psychiatrist
according to the DSM-IV criteria. Participation of the
children was on a voluntary basis (i.e., no compensation)
and informed consent was obtained from their parents. The
study was approved by the Ethics Committee of the Faculty
of Social Sciences of the Radboud University Nijmegen.
Setting and Materials
The sessions were conducted in a small meeting room at the
Dr. Leo Kannerhuis, a treatment facility for individuals with
Autism Spectrum Disorders. Lego
Ò
was used as play
material for the children and consisted of 16 Lego
Ò
con-
structions. During baseline, assignment cards (with a written
assignment and a photograph of the Lego
Ò
construction) and
Table 1 Child characteristics
Pair Child Age
a
Gender Diagnosis SCQ-score
b
1 Brett 10;1 Male ASD, ADHD 24
c
Alex 10;1 Male – 1
2 Chris 09;1 Male Asperger Syndrome 14
Debby 11;5 Female – 8
3 Eric 05;7 Male ASD 18
c
Felicia 07;4 Female – 1
a
Years; months
b
Social Communication Questionnaire Total score
c
Above cut-off score (C15) for ASD
J Autism Dev Disord (2015) 45:3746–3755 3747
123
a general rule board were used. The assignment cards indi-
cated what the children had to build and the general rule
board consisted of six general play rules. During the inter-
vention sessions, a NAO robot from Aldebaran robotics was
placed in front of the children (just within reach) on the
table. This robot was 57 cm tall, had a neutral color, a
‘simple’ face (only mouth and eyes), a syntactic Dutch
female voice, and it could move its arms, legs, and fingers. In
addition, the robot had a microphone, a speaker, digital
cameras, and touch sensors. The speech of the robot was pre-
programmed. To control the robot, a laptop was used by the
trainer. TiViPE—a visual programming environment—was
used for programming the robot (Barakova et al. 2013).
Above that, an additional game rule board with four play
rules, two role cards, four little instruction booklets with
building steps, and three pictures of the Lego
Ò
construction
for sessions four and five, were used in the intervention. The
instruction booklets consisted of the building steps for the
Lego
Ò
constructions and the building steps indicated which
Lego
Ò
bricks the children needed and how they had to build
the Lego
Ò
construction. Finally, to record all sessions, a
digital video camera on a tripod was used. This camera was
placed in the corner of the rooms.
Design
A concurrent multiple baseline design across child–sibling
pairs was used to investigate the effectiveness of the
intervention. By using a concurrent multiple baseline
design across at least three pairs, the results are controlled
for alternative explanations as maturation and history
(Horner et al. 2005). The pairs were randomly assigned to
the different baseline lengths of three, four and five ses-
sions. Intervention and post-intervention were in effect for
respectively five and three sessions.
Data Collection and Response Categories
All sessions were videotaped. The videos were observed
and coded in a randomized order. In coding the behavioral
categories a continuous 10 s partial interval recording
system was used. Although each session lasted 30 min,
only 10 min were used for data collection for practical
reasons (i.e., one observation took 1.5–2 h to complete),
establishing 60 intervals per session. The recording of
baseline and post-intervention sessions started after the
introduction of the session by the trainer and the registra-
tion of intervention sessions started after the introduction
of the session by the robot.
Data was collected on collaborative behaviors, consisting
of the following behavioral categories: (a) Interaction ini-
tiations, consisting of questions, statements and instructions
directed to the TD sibling, (b) responses, consisting of
adequate responses to a question and adequate responses to
an instruction of the TD sibling, and (c) play together,
consisting of manipulating materials together with the TD
sibling in order to achieve a common goal. The definitions
of the behavioral categories are presented in Supplementary
material 1. Data was collected for each ASD child sepa-
rately, except for ‘play together’ that was collected for each
pair. When a behavior was present during an interval (e.g.,
ASD child directed a question to the TD sibling), a plus (?)
was recorded on the datasheet. When a behavior category
was absent during an interval, a minus (-) was recorded.
Reliability of Recording
In order to remain naı
¨
ve of the intervention phase all
10-min videos were coded in a randomized order. The third
author (i.e., primary observer) trained an independent,
naive secondary observer (i.e., a research assistant) in
recording the behavioral categories. The secondary obser-
ver was blind to the goal of the study. Interobserver
agreement (IOA) was assessed on an interval-by-interval
basis. Agreement was defined as both observers identifying
the same behavior categories as absent or present during an
interval. Disagreement was defined as both observers
identifying different behavior categories as absent or
present during an interval. To determine IOA, prevalence-
adjusted and bias-adjusted kappa (PABAK; Byrt et al.
1993) was calculated. Before starting data collection, both
observers practiced by observing independently the videos
of a comparable unpublished pilot study. Instructions and
recording were rehearsed until PABAK was above 0.80 on
two consecutive recording sessions. The primary and sec-
ondary observer independently recorded 33 % of all ses-
sions, equally divided across conditions and pairs.
Observations started after participants finished all phases of
the intervention.
Mean overall PABAK was 0.91 (SD = 0.06, range
0.81–0.99) indicating excellent agreement between the two
observers (Cicchetti et al. 2006). Mean PABAK for the
behavior categories were also excellent (i.e., interaction
initiations M = 0.92, SD = 0.04; Responses M = 0.94,
SD = 0.05; Play together M = 0.85, SD = 0.15).
Dependent Measures
For each child with ASD, the percentage of occurrence of
each separate behavior category was calculated by dividing
the number of pluses for that behavior category by the total
number of registered intervals (i.e., 60), multiplied by
100 %. It was hypothesized that the percentages of
occurrence of all behavior categories would increase dur-
ing intervention.
3748 J Autism Dev Disord (2015) 45:3746–3755
123
Procedures
A detailed description of the procedures are presented in
Supplementary material 2.
Baseline and Post-intervention
The baseline consisted of three to five 30-min sessions and
the post-intervention consisted of three 30-min sessions;
sessions were implemented once a week. During these
conditions, the children received an assignment card and
had to collaborate with each other during Lego
Ò
play for
25-min. The trainer did not provide any additional
instructions and did not help the children in building the
Lego
Ò
construction.
Intervention
The intervention consisted of five 30-min sessions, imple-
mented once a week. Sessions were leaded by the robot
instead of the trainer. The trainer was present during all
sessions to control the robot with the laptop and to assist
the robot when needed. In the session, the robot first
introduced itself by telling that it would help the children to
play together with Lego
Ò
during five sessions. After this,
the robot explained the role of the trainer as an assistant in
providing help and materials. Then, the robot explained the
roles of the children during the Lego
Ò
play: one of the
children would be the guide and one of the children would
be the builder. The guide had to describe the instructions
from an instruction booklet (task analysis) and the builder
had to collect the Lego
Ò
bricks and had to put them
together. The robot told the children that, in this way, they
had to collaborate building the Lego
Ò
construction. While
the children were working, the robot reinforced and
prompted them. For example, when a child performed a
role of the other child, when a child asked questions about
the roles, when a child showed off-task behavior, or when a
child did not do anything. The possible prompts for the
guide were: (a) ‘Guide, explain to the builder which bricks
he needs’, (b) ‘Guide, explain to the builder what he has to
do in this step’, (c) ‘Guide, wait until the builder is ready’,
and (d) ‘Guide, can you help the builder?’. The possible
prompts for the builder were: (a) ‘Builder, wait for the
instruction of the guide’, (b) ‘Builder, listen carefully to the
guide’, (c) ‘Builder, look for the bricks you need’,
(d) ‘Builder, put the bricks together, like the guide
explained’, and (e) ‘Builder, can you help the guide?’ A
possible prompt for both children was: ‘Boys, can you help
each other?’ The robot also reacted in cases of questions,
rule violations, and other problems.
During the five intervention sessions, 19 rule violations
occurred (49 Brett, 59 Chris, and 109 Eric) in which the
robot had to ask the assistant for help. Most violations
related to the rule to build things together (169). All vio-
lations could be resolved by the assistant. Other problems
occurred during the fourth intervention session, in which
Brett as well as Chris were aggressive to their TD sibling
(respectively hitting and yelling). The session of Brett and
his TD sibling was interrupted for 5 min, but could be
resumed. The aggression of Brett and Chris resulted in
resistance in playing further by the TD siblings for 1 and
4 min respectively.
Social Validity
After the second post-intervention session, social validity
questionnaires were filled in by all children and their par-
ents to evaluate the acceptability of the procedures and the
effectiveness of the intervention. All statements were rated
on a 5-points Likert scale ranging from 1 (not at all) to 5
(very much). All children completed the questionnaire
independently, with the exception of Eric, because he could
not read. His mother helped him by reading out the state-
ments. In the last question of the questionnaire parents and
children gave the intervention a score ranging from 1 to 10.
Treatment Integrity
Data on treatment integrity was collected by an indepen-
dent observer (a research assistant) for 33 % of all sessions
equally divided across conditions and pairs. Treatment
integrity was calculated per session and based on the ratio
of number of executed components and the number of
planned components (more specific, the number of events a
procedural component was emitted as planned, divided by
the number of opportunities to emit that component, mul-
tiplied by 100 %).
The mean percentage of treatment integrity during
baseline was 94 % (SD = 3.44, range 90–98 %). During
intervention, the mean percentage of treatment integrity
was 98 % (SD = 1.52, range 97–100 %). Finally, during
post-intervention, the mean percentage of treatment integ-
rity was 93 % (SD = 6.19, range 82–96 %).
Data Analysis
First, data analysis involved the calculation of mean per-
centages of the behavior categories across conditions and
visual inspection of the data. Second, to determine the
effect size of the intervention, Tau-U was calculated. Tau-
U can be used in single case research and examines the
percentage of non-overlap of the data between conditions
(Parker et al. 2011a). Additionally, Tau-U controls for a
positive baseline trend (Parker et al. 2011b). Tau-U, the
standard deviations of Tau-U and the p values were
J Autism Dev Disord (2015) 45:3746–3755 3749
123
calculated across conditions for each pair. To calculate
Tau-U, Single Case Research, a web based calculator for
single case research analysis, was used (Vannest et al.
2011).
Results
Collaborative Behaviors
Interactions Initiations
Figure 1 presents the interaction initiations for all ASD
participants across conditions. Visual inspection reveals an
inconsistent pattern for all participants across conditions.
Compared to baseline, all ASD participants showed more
initiations during the ‘guide’ intervention sessions; however,
during the ‘builder’ intervention sessions no changes were
found compared to baseline. During the first intervention
session all participants were builder as well as guide as two
constructions were made. All participants increased the
percentage interaction initiations during intervention (Brett:
baseline M = 7.78, SD = 6.73, intervention M = 17.00,
SD = 17.73; Chris: baseline M = 15.00, SD = 9.13, inter-
vention M = 37.33, SD = 36.87; Eric: baseline M = 9.00,
SD = 3.84, intervention M = 11.67, SD = 7.64). As
expected from visual inspection, Tau-U analyses revealed no
statistical significant changes for all children across
conditions.
Responses
Figure 2 shows the responses of the ASD children across
conditions. Visual inspection showed an increase in respon-
ses of Brett and Chris across sessions. Compared to baseline,
Brett and Chris responded both at a higher level during
intervention (Brett: baseline M = 2.78, SD = 2.55, inter-
vention M = 15.00, SD = 7.07; Chris: baseline M = 14.17,
SD = 7.39, intervention M = 33.33, SD = 14.09); how-
ever, no changes on level or trend were visible for Eric
(baseline M = 20.83, SD = 13.36, intervention M = 15.00,
SD = 12.69). An inconsistent pattern is shown for all par-
ticipants across conditions. Tau-U analysis revealed no sta-
tistical significant changes across conditions for all children.
Play Together
Figure 3 presents the percentages of ‘play together’ for all
participants across conditions. Visual inspection revealed a
decreasefor Eric and his TDsibling.Percentagesvaried across
conditions for all pairs. Mean percentages of ‘play together’
decreased for all pairs during intervention, (Brett: baseline
M = 21.67, SD = 10.14, intervention M = 10.00,
SD = 6.66; Chris: baseline M = 15.42, SD = 13.70, inter-
vention M = 7.00, SD = 10.37; Eric: baseline M = 21.33,
SD = 9.75, intervention M = 5.67, SD = 4.35). Tau-U
analysis indicated that the change for Eric and his TD sibling
was statistically significant (Tau-U =-0.96, 90 % CI -1.00
to -0.33). During post-intervention, mean percentages of
Brett and Chris were about the same as during intervention.
Social Validity
Parents reported that both the children with ASD and their
siblings more enjoyed the robot sessions (M = 4.3, range
3–4) than the sessions without the robot (M = 3.3, range
3–5). Parents were mixed positive about the effectiveness
of the training on improving the collaborative behaviors of
their children (M = 3.3, range 2–5). Finally, the parents
rated the robot-mediated training with a 7 (M = 7, range
5–9).
The children with ASD reported the sessions without the
robot as a little bit more enjoyable (M = 3.7, range 3–4)
than the sessions with the robot (M = 3.3, range 2–5),
while the TD siblings enjoyed the sessions with the robot
more (M = 4.7, range 4–5) than the sessions without the
robot (M = 3.3, range 2–4). The children with ASD
reported that they learned to play together following
intervention (M = 4.3, range 3–5). The TD siblings were
also positive, however they rated this item lower than the
children with ASD (M = 3.7, range 3–4). The components
of the intervention were rated as acceptable by both the
children with ASD and the siblings, M = 3.1 (range 1–5)
and 3.7 (range 1–5), respectively. In addition, all children
liked to play with the Lego
Ò
(M = 4.0, range 3–5 for both
children with ASD and their TD siblings) and liked to play
with their siblings, respectively M = 3.8 (range 2.5–5) and
M = 3.7 (range 3–4). The children had individual prefer-
ences to be builder or guide (ranges 1–5). The children
with ASD reported the robot as exciting (M = 4.0, range
2–5), while the TD siblings reported it as less exciting
(M = 2.2, range 1–4). However, both groups reported that
they wished that they could have more training sessions
with the robot, respectively M = 3.3 (range 2–5) and
M = 4.3 (range 3–5). Finally, all TD siblings and two
children with ASD rated the training as positive (respec-
tively M = 8.7, range 8–10; M = 7, range 6–8). Eric did
not rate the training, because he did not understand this
question.
Discussion
The aim of the current study was to investigate the effec-
tiveness of a brief robot-mediated intervention based on
Lego
Ò
therapy on increasing collaborative behaviors of
3750 J Autism Dev Disord (2015) 45:3746–3755
123
children with ASD duringplay sessions with their TD siblings.
Although no statistically significant changes in interaction
initiations, responses and play together for the children with
ASD were found, the robot-intervention revealed for two out
of three pairs an increase in responses across sessions, as well
as an increase in interaction initiations during the ‘guide’
sessions. It may be concluded that the robot-mediated Lego
Ò
therapy was not effective in improving collaborative behav-
iors of children with ASD, although visual analysis revealed
some possible positive effects.
0
5
10
15
20
25
30
35
40
45
50
123 45678 910111213
Percentages
Bre
0
10
20
30
40
50
60
70
80
90
1 2 3 4 5 6 7 8 9 10 11 12 13
Percdentages
Chris
0
10
20
30
40
50
60
12345 678910 111213
Percentages
Eric
*
*
*
*
*
*
Baseline
Intervenon
Post-Intervenon
Sessions
Fig. 1 Percentages of interaction initiations of ASD participants across conditions
J Autism Dev Disord (2015) 45:3746–3755 3751
123
The results of the current study are partially consistent
with Owens et al. (2008) in that no significant changes in
the percentage interaction initiations between baseline and
the Lego
Ò
therapy intervention were found. In contrast,
LeGoff (2004) and LeGoff and Sherman (2006) found
increases in initiations with peers. There were several
differences between the procedures of Lego
Ò
therapy in
prior studies and in the current study, which may explain
the inconsistent findings. These issues are discussed below.
In the current study, a robot was used. Despite the
suggested benefits of the use of robots in interventions for
children with ASD (Dautenhahn and Werry 2004; Diehl
0
10
20
30
40
123 45678 910111213
Bre
0
10
20
30
40
50
60
1234 56789 10111213
Chris
0
10
20
30
40
50
1 2 3 4 5 6 7 8 9 10 111213
Sessions
Eric
Baseline
Intervenon
Post-Intervenon
Percentages Percentages Percentages
Fig. 2 Percentages of responses of ASD participants across conditions
3752 J Autism Dev Disord (2015) 45:3746–3755
123
et al. 2012), the use of a robot also induced some limita-
tions. For example, the behavioral repertoire of the robot is
limited, indicating limited prompt levels and reinforcement
options. In the study of Huskens et al. (2013) a least-to-
most prompt hierarchy was used with four prompt levels
(i.e., open-question prompt, waiting prompt, tell-prompt,
and fill-in prompt), all directed to only one target behavior
(i.e., self-initiated questions). However, the current study
focused on a broad range of target behaviors, related to the
Lego
Ò
roles (e.g., builder has to give instructions on which
0
10
20
30
40
123 45678 910111213
Bre
0
10
20
30
40
50
1234 56789 10111213
Chris
0
10
20
30
40
50
1 2 3 4 5 6 7 8 9 10 11 12 13
Sessions
Eric
Baseline
Intervenon
Post-Intervenon
Percentages
Percentages
Percentages
Fig. 3 Percentages of play together of ASD participants across conditions
J Autism Dev Disord (2015) 45:3746–3755 3753
123
Lego
Ò
bricks are needed, and on how to put the Lego
Ò
bricks together). As a consequence, prompts needed to be
directed to all of these specific target behaviors. As the
robot only had a limited behavioral repertoire, no differ-
entiation in prompt levels (e.g., least-to-most prompting)
could be made for each target behavior. For example, in
case of repeated incorrect responses, only the same prompt
could be used directed to that specific behavior and no
adaptations in prompt levels could be made by the robot.
Therefore, the possibilities to adapt the robot’s behaviors to
the individual prompt needs and preferences of the children
were limited. It is recommended to develop the robot by
increasing the variability in prompt levels (e.g., least-to-
most hierarchy) to respond to the children’s individual
needs and specific target behaviors. Some researchers start
to address this problem. Greczek et al. (2014), for example,
developed a computational model of graded cueing feed-
back, but the framework needs to be expanded to more
complex interactions in the future. For now, it is recom-
mended to use robots in interventions with only one or two
specific target behaviors and not during interventions with
a broader range of target behaviors, as for example Lego
Ò
therapy. Another important aspect concerns the fact that a
technical assistant had to program the robot and had to be
present during the sessions to set up the robot, next to the
trainer. Simplifying programming the robot, could enable
therapists to adapt the robot’s behavioral repertoire without
the help of a technical assistant. This could ameliorate the
use of robots in interventions for children with ASD.
The procedures used in the current study differed from
the procedures of Lego
Ò
therapy in prior studies regarding
(a) the intensity of the intervention, that is compared to the
studies of LeGoff (2004), LeGoff and Sherman (2006) and
Owens et al. (2008), the intensity of the current interven-
tion can be rated as low, which may have contributed to the
limited results. (b) The opportunities to practice the dif-
ferent roles and skills, that is compared to the study of
Owens et al. (2008), where children switched roles after a
certain amount of time or instruction steps and when they
demonstrated mastery of the skills of one of the roles,
children in the current study had less opportunity to master
the roles. (c) The moment of transition to more complex
Lego
Ò
constructions differed, that is in the current study,
no behavioral criteria were used and the moments of
transitions were pre-determined, whereas in the study of
Owens et al. (2008), transitions to more complex Lego
Ò
constructions were made when the children were able to
build simple and quick models. (d) Baseline observations
did not occur in unstructured play situations as in the study
of Owens et al. (2008) and the general rules of Lego
Ò
therapy were already used during baseline. The use of the
Lego
Ò
rules during baseline may have offered a certain
amount of structure, which may have elicited children to
show more collaborative behaviors than in unstructured
play situations. In the current study different materials and
types of verbal instructions were used as antecedent stim-
ulus between baseline/post-intervention sessions and
intervention sessions. It is recommended to keep these
aspects constant between conditions.
The current study was the first study investigating the
effectiveness of Lego
Ò
therapy for children with ASD and
their TD siblings. As interactions between children with
ASD and their TD siblings are often found to be more
negative than between two TD siblings (Ferraioli et al.
2012) more and longer intervention sessions may also be
necessary to break the negative interaction patterns and to
improve results.
Results of social validity indicate that both the children
with ASD and the TD siblings reported improvements in
‘play together’, while such improvements were not found
according to the behavioral measures. This subjective
perception of improvement may be caused by a placebo
effect, by which participants report improvements after
receiving an intervention, while real improvements are
lacking (Linde et al. 2011). Another possiblity is that the
children used a different definition of ‘play together’.
Most studies on the effectiveness of robotic interven-
tions in persons with ASD showed methodological limi-
tations, decreasing their internal validity (Diehl et al.
2012). In addition, one of the major shortcomings of robot
studies is the lack of quantitative measures (Scassellati
et al. 2012). In the current study: (a) a single-subject design
(i.e., concurrent multiple baseline design across child–
sibling pairs) was used, providing control for alternative
explanations as maturation and history, (b) adequate
treatment integrity (M = 95 %) and interobserver agree-
ment scores (overall PABAK = 0.91) were found,
(c) dependent variables were quantified and operational-
ized in a transparent way, and (d) sufficient information
was provided for replication. By taking in account these
methodological characteristics, the current study substan-
tially contributes to the research on the effectiveness of
robotic interventions for children with ASD.
The current study was the first study that investigated
the effectiveness of robot-mediated Lego
Ò
therapy on
collaborative behaviors of children with ASD and their TD
siblings. To improve the results, it is recommended in
future studies to extend the intervention period with more
sessions, to increase the duration of each session, to use the
same materials and instructions across conditions, to switch
roles more often, and to establish behavioral criteria to
indicate when a child is ready for a transition to longer and
more complex Lego
Ò
constructions. In order to get a more
realistic impression of the target behaviors, it is also rec-
ommended to conduct baseline observations in unstruc-
tured play situations.
3754 J Autism Dev Disord (2015) 45:3746–3755
123
As long as robots cannot be programmed in a way that
their behaviors could be easily adapted to children’s indi-
vidual abilities and needs, it may be concluded that robots
may better be used in interventions that target one specific
behavior than in interventions that target a broad range of
target behaviors such as Lego
Ò
therapy.
Acknowledgments The authors gratefully acknowledge the support
of the Innovation-Oriented Research Program ‘Integral Product Cre-
ation and Realization (IOP IPCR)’ of the Dutch Ministry of Economic
Affairs, Agriculture and Innovation, The Hague. The authors thank
the children and their parents for their participation. We also would
like to thank Rianne Verschuur and Margreet Weide for their assis-
tance during preparation and data collection. Finally, we would like to
thank Terence Nelson for his assistance with the robot in the inter-
vention sessions.
References
American Psychiatric Association. (2000). Diagnostic and statistical
manual of mental disorders (4th edn., text rev.; DSM IV-TR).
Washington, DC: American Psychiatric Association.
Baker, M. J. (2000). Incorporating the thematic ritualistic behaviors
of children with autism into games: Increasing social play
interactions with siblings. Journal of Positive Behavior Inter-
ventions, 2, 66–84.
Barakova, E. I., Gillesen, J. C. C., Huskens, B. E. B. M., & Lourens,
T. (2013). End-user programming architecture facilitates the
uptake of robots in social therapies. Robotic Autonomic Systems,
61, 704–713.
Bauminger, N., Shulman, C., & Agam, G. (2003). Peer interaction
and loneliness in high-functioning children with autism. Journal
of Autism and Developmental Disorders, 33, 489–507.
Bauminger, N., Solomon, M., Aviezer, A., Heung, K., Brown, J., &
Rogers, S. J. (2008). Friendships in high-functioning children
with autism spectrum disorder: Mixed and non-mixed dyads.
Journal of Autism and Developmental Disorders, 38,
1211–1229.
Ben-Sasson, A., Lamash, L., & Gal, E. (2013). To enforce or not to
enforce? The use of collaborative interfaces to promote social
skills in children with high-functioning autism spectrum disor-
der. Autism, 17, 608–622.
Byrt, T., Bishop, J., & Carlin, J. B. (1993). Bias, prevalence and
kappa. Journal of Clinical Epidemiology, 46, 423–429.
Cicchetti, D., Bronen, R., Spencer, S., Haut, S., Berg, A., Oliver, P., &
Tyrer, P. (2006). Rating scales, scales of measurement, issues of
reliability: Resolving some critical issues for clinicians and
researchers. The Journal of Nervous and Mental Disease, 194,
557–564.
Dautenhahn, K., & Werry, I. (2004). Towards interactive robots in
autism therapy. Pragmatics & Cognition, 12, 1–35.
Diehl, J. J., Schmitt, L. M., Villano, M., & Crowell, C. R. (2012). The
clinical use of robots for individuals with autism spectrum
disorders: A critical review. Research in Autism Spectrum
Disorders, 6, 249–262.
Duquette, A., Michaud, F., & Mercier, H. (2008). Exploring the use of
a mobile robot as an imitation agent with children with low-
functioning autism. Autonomous Robots, 24, 147–157.
Ferraioli, S. J., Hansford, A., & Harris, S. L. (2012). Benefits of
including siblings in the treatment of autism spectrum disorders.
Cognitive and Behavioral Practice, 19, 413–422.
Greczek, J., Kaszubski, E., Atrash, A., & Mataric¸, M. J. (2014).
Graded cueing feedback in robot-mediated imitation practice for
children with autism spectrum disorders. In 23rd IEEE sympo-
sium on robot and human interaction communication (RO-
MAN’14). Edinburgh, UK, Aug 25–29, 2014. http://robotics.ucs.
edu/publications/media/uploads/pubs/Ro-MAN-14-greczek.pdf.
Accessed November 14, 2014.
Horner, R. H., Carr, E. G., Halle, J., McGee, G., Odom, S., & Wolery,
M. (2005). The use of single-subject research to identify
evidence-based practice in special education. Exceptional Chil-
dren, 71, 165–179.
Huskens, B. E. B. M., Verschuur, R., Gillesen, J. C. C., Didden, R., &
Barakova, E. I. (2013). Promoting question-asking in school-
aged children with autism spectrum disorders: Effectiveness of a
robot intervention compared to a human-trainer intervention.
Developmental Neurorehabilitation, 16, 345–356.
LeGoff, D. B. (2004). Use of LEGOÓ as a therapeutic medium for
improving social competence. Journal of Autism and Develop-
mental Disorders, 34, 557–571.
LeGoff, D. B., & Sherman, M. (2006). Long-term outcome of social
skills intervention based on interactive LEGOÓ play. Autism, 10,
317–329.
Linde, K., Fa
¨
ssler, M., & Meissner, K. (2011). Placebo interventions,
placebo effects and clinical practice. Philosophical Transactions
of the Royal Society B, 366, 1905–1912.
Owens, G., Granader, Y., Humphrey, A., & Baron-Cohen, S. (2008).
LEGO
Ò
therapy and the social use of language programme: An
evaluation of two social skills interventions for children with
high functioning autism and Asperger syndrome. Journal of
Autism and Developmental Disorders, 38, 1944–1957.
Parker, R. I., Vannest, K. J., & Davis, J. L. (2011a). Effect size in
single-case research: A review of nine nonoverlap techniques.
Behavior Modification, 35, 303–322.
Parker, R. I., Vannest, K. J., Davis, J. L., & Sauber, S. B. (2011b).
Combining nonoverlap and trend for single-case research: Tau-
U. Behavior Therapy, 42, 284–299.
Robins, B., Dautenhahn, K., Te Boekhorst, R., & Billard, A. (2004a).
Effects of repeated exposure to a humanoid robot on children
with autism. In S. Keates, J. Clarkson, P. Langdon, & P.
Robinson (Eds.), Designing a more inclusive world (pp.
225–236). London: Springer.
Robins, B., Dickerson, P., Stribling, P., & Dautenhahn, K. (2004b).
Robot-mediated joint attention in children with autism: A case
study in robot–human interaction. Interaction Studies, 5,
161–198.
Rutter, M., Bailey, A., & Lord, C. (2003). The Social Communication
Questionnaire (SCQ). Los Angeles: Western Psychological
Services.
Scassellati, B., Admoni, H., & Mataric, M. (2012). Robots for use in
autism research. Annual Review of Biomedical Engineering, 14,
275–294.
Tsao, L., & Odom, S. L. (2006). Sibling-mediated social interaction
intervention for young children with autism. Topics in Early
Childhood Special Education, 26, 106–123.
Vannest, K. J., Parker, R. I., & Gonen, O. (2011). Single Case
Research: Web based calculators for SRC analysis (version 1.0)
[Web-based application]. College Station, TX: Texas A&M
University. Retrieved Tuesday July 23, 2013. Available from
singlecaseresearch.org.
Walton, K. M., & Ingersoll, B. R. (2012). Evaluation of a sibling-
mediated intervention for young children with autism. Journal of
Positive Behavior Interventions, 14, 241–253.
Warreyn, P., Raymaekers, R., & Roeyers, H. (2004). SCQ. Handle-
iding Vragenlijst Sociale Communicatie. Destelbergen: SIG vzw.
Weiss, M. J., & Harris, S. L. (2001). Teaching social skills to people
with autism. Behavior Modification, 25, 785–802.
J Autism Dev Disord (2015) 45:3746–3755 3755
123
