Effects of a Video-Based Intervention on Emotion Recognition for Children With Autism Who Have Limited Speech in China

Participants

Four children were included in this study based on the following criteria. The participant (1) met the diagnosis criteria for autism spectrum disorder, with the diagnosis based on current assessment and including the results of the Childhood Autism Rating Scale (CARS) or equivalent tests (e.g., Gilliam Autism Rating Scale-Second Edition). (2) The participant was 4–12 years old. (3) The participant had limited speech (e.g., receptive language skills equivalent to at least a 30-month level, as determined by the Peabody Picture Vocabulary Test (PPVT), and the participant’s speech was less than 30% intelligible under the semantic context-unfamiliar listener’s condition of the Index of Augment Speech Comprehensibility for Children (I-ASCC)).

Recruitment of Participants

Once the primary investigator obtained the Institutional Review Board’s approval for the study, approval was sought from the autism research center at Guizhou Education University. The primary researcher then sought permission from the teachers of the children and consent from the children’s parents. A total of four participants, Yu, Han, Ben, and Wong (pseudonyms), participated in this study.

Yu was six years and six months old when the study began. At the age of three, he was diagnosed with ASD using the Chinese version of the CARS, with a score of 36.5, suggesting severe autism. In the mornings, he received early intervention services and support at the autism research center, which included 2 hr of applied behavior analysis (ABA) therapy with individualized programs in gross motor imitation, matching objects and pictures, receptive object recognition, and following two-step commands. In the afternoons, he was educated in a private elementary school. During the weekends, he stayed home with his mother, who supported his independent skills. His mother helped Yu conduct emotion recognition training, and only happy emotions could be recognized. In addition, his mother reported that Yu is fond of watching cartoon videos, especially Peppa Pig (a British animated series) and Hikarian (a Japanese animated series). According to the PPVT and I-ASCC report, his receptive language skills are equivalent to a 30-month level and 27%, respectively. His special education teachers reported that Yu has difficulty with emotion recognition.

Han was a boy four years and seven months old who was diagnosed with ASD at age two using the Chinese version of the CARS, with a score of 34, suggesting moderate autism. During this study, Han received individual training for 3 hr per day, which included gross motor imitation and speech therapy. According to his teachers’ report, he has difficulties in speech and matching tasks. Han’s speech skills comprised only a few vocalizations, and his communication consisted of two-to three-word utterances. According to the PPVT, Han’s receptive language skills are equivalent to at least a 36-month level, whereas his I-ASCC is 24%. His grandmother reported that he likes to watch animation videos and news and to listen to the radio. Ultraman (a Japanese television series) is his favorite, and he often imitates the actions of this cartoon figure. Teachers reported that Han fails to correctly recognize emotions.

Ben was seven years and three months of age. He was diagnosed with autism at three years old using the Chinese version of the CARS, with a score of 36, indicating severe autism. He was in first grade and receiving special education services, including academic and social skills instruction. Ben also received ABA therapy at the autism research center in the evening, and the training included gross motor imitation, matching objects, and following two-step directions. Ben did not follow verbal directions unless visual cues were provided. Direct observations suggested that Ben did not have an echoic repertoire but could imitate two-syllable approximations, which were recognized by familiar people only when specific establishing operations occurred. In terms of speech, Ben’s receptive language skills are equivalent to at least a 42-month level, whereas his I-ASCC is 27%. His father reported that Ben likes watching TV programs, including social news, entertainment channels, and sports channels. In addition, teachers reported that Ben cannot recognize emotions in social cues and the matching emotion picture task.

Wong was five and two months of age. He was diagnosed with autism at four years of age using the Chinese version of the CARS, with a score of 34, indicating moderate autism. He received ABA therapy at the autism research center in the morning, which included gross motor imitation, matching objects and pictures, receptive object identification, and following three-step directions. He attended preschool in the afternoon. His receptive language skills are equivalent to a 30-month level according to the PPVT, and he exhibited 29% in the I-ASCC. His parents reported that he likes cartoons and listens to nursery rhymes. Wong could ask his mother to turn on the TV when he wanted to watch it. His special education teachers reported that Wong has difficulty with emotion recognition.

Settings and Materials

The participants were recruited from an autism research center located in a metropolitan city in southwestern China. The experiment was conducted in Mandarin. The individual therapy room was 3 by 4 m in size and contained a table and two chairs for one-on-one therapy.

A laptop computer and digital camera were prepared for the experiment. We used The Transporters as our main experimental material. The Transporters consists of fifteen five-minute episodes, and a key emotion or mental state is included in each episode. In the study, we selected six basic emotions (i.e., happy, angry, afraid, sad, surprised, and disgusted) to investigate the effect because typically developing children between 2 and 7 years of age recognize and understand these emotions. Additionally, to adapt the material to the Chinese language context, the Chinese version of The Transporters was supplemented as follows: (1) A new text script was created according to the animation. (2) The voiceover was modified to make it conform to the Chinese language context. (3) Preschool and primary school teachers were asked to modify the text script to make it more appropriate for children. (4) A new video was dubbed by a professional who qualified as an excellent speaker of Mandarin (e.g., 3A level) to synchronize the new script with the original animation.

Experimental Design

A multiple baseline across participants was employed to examine the effect of VBI on emotion recognition in children with ASD who have limited speech in China. Data for each participant were collected and analyzed separately to understand the unique changes in that individual. A multiple baseline across participants design features a staggered introduction of the independent variable to participants in sequential order (Gast & Ledford, 2009). One participant receives intervention while the others remain in the baseline condition. If the introduction of the treatment is followed by a change in the dependent variable on each baseline, this represents strong evidence of a treatment effect (Bailey & Burch, 2017; Ledford, Lane, & Gast, 2018). In addition, the choice of a multiple baseline design over a reversal design was based on ethical concerns given the ability of emotion recognition ability by participants during the intervention phase.

Data Processing

A single subject design relies heavily on visual inspection. For each participant’s graphed data, the experimenter examined the level, trend, variability, overlap, immediacy of effects, and consistency of data patterns between phases (Kennedy, 2005). The level of participant data was determined to be stable when 80% of the data fell within a 25% range of the median for a given phase. Trends in the data were established using the split-middle method technique within and across phases (Gast & Ledford, 2014). To further verify the results, this study also used the statistical procedure Tau-U. Tau-U is a method for measuring data nonoverlap between two phases (Parker, Vannest, Davis, & Sauber, 2011). It is a “distribution free” nonparametric technique, with a statistical power of 91%–95% of linear regression when data conform to parametric assumptions (Brossart, Laird, & Armstrong, 2018). Tau-U follows the “S” sampling distribution. Therefore, p-values and confidence intervals are available. According to parker et al., (Parker et al., 2011), Tau-U is defined as follows

T a u − U = S P − S A m n = T a u − m − 1 2 n t A

The interpretation of Tau-U effect sizes is as follows. A numerical score between 0.93–1 is considered a large effect, 0.66–0.92 a medium effect, and 0–0.65 a small effect (Parker et al., 2011).

Dependent Variable and Independent Variable

The dependent variable was the percentage of correct emotion recognition. The percentage of correct emotion recognition was calculated by dividing the total number of matches by the number of correct matching pictures. A response was counted as accurate if a participant responded independently and matched a picture to a situation within 5 s. An incorrect response was defined as providing an incorrect or no response within 5 s after the experimenter’s request or if the participant had provided no responses after completing the video two times. The independent variable in this study was The Transporter as an intervention tool.

Preference Assessment

Prior to each intervention session, the experimenter conducted a multiple stimulus without replacement (MSWO) assessment (DeLeon & Iwata, 1996) to identify each participant’s preferred items for that training session. Specifically, 10 to 12 preferred items from a pool of previously identified preferred items were randomly selected and presented. Each participant was allowed to select one of the items at a time and interact with it for 30 s to 1 min. The selected item was not presented again in the subsequent trials. The assessment ended when the participant refused to select any items or until no items were left. The assessment session ended within 10 min for one participant. The first five frequently selected items were considered highly preferred and were used for that training session.

Baseline

The baseline condition simulated participants’ daily routines in their natural environment. During the baseline probe, no VBI was presented. The experimenter selected six basic facial emotion pictures that were taken from The Transporters, provided a simple scene that occurred in The Transporters for each facial emotion, and then asked each participant to match the facial emotion picture within 5 s. For matching afraid emotions, for example, the experimenter said to a child, “Sally is stranded above the viaduct with a storm brewing. How does Sally feel?” The participant would choose one of Sally’s facial expression pictures from three varying facial expression pictures (e.g., angry, afraid, and happy). A correct response was defined as the participant choosing a facial expression within 5 s. An incorrect response was defined as the participant choosing a wrong picture or providing no response after the experimenter initiated the question. Each participant’s baseline data had to be stable (e.g., 80% of the data falling within 25% of the median with a zero-accelerating trend) (Gast & Ledford, 2014).

Intervention

The VBI intervention condition consisted of a 5-min video followed by a 5-min training session. The experimenter placed a laptop computer on the table and told each participant “We are going to learn happy emotions today.” The experimenter then played the video and highlighted each key emotion after each episode was completed (e.g., The happy emotions is over today). After the video, each participant was asked to respond to questions (e.g., Which two pictures have the same expression?) Notably, this activity was differentiated over three levels by the number of characters offered and by the number of facial expressions offered.

During level 1, use two Transporters characters expressing the same facial expression (e.g., see Figure 1). During level 2, use one Transporters character expressing two different facial expressions (e.g., see Figure 2). During level 3, use three Transporters characters expressing two different facial expressions (e.g., see Figure 3). This level only focused on facial expressions, with no emphasis on the characters. The mastery criterion for each level was at least 80% accurate matching of pictures during the intervention sessions for three consecutive sessions. Once participants with ASD achieved the mastery criterion, they would not receive any highly preferred items first, and then, the experimenter allowed them to engage in activities unrelated to emotion recognition.OPEN IN VIEWER

OPEN IN VIEWER

Figure 2.

Matching the angry emotion picture in level 2.

OPEN IN VIEWER

Figure 3.

Matching the afraid emotion picture in level 3.

Generalization

The generalization phase was conducted 7 and 15 days after the intervention phase. Two tasks were presented in the generalization phase. The first task was conducted after 7 days, and participants had to match familiar situations taken from The Transporters to the facial expressions of characters familiar from the series (e.g., see Figure 4). The second task was conducted after 15 days, and its aim was to test generalization to facial expressions that are not attached to vehicles. In this task, the participants had to match novel situations with novel expressions using a selection of human non-Transporters faces (e.g., see Figure 5).OPEN IN VIEWER

OPEN IN VIEWER

Figure 5.

Gary obtained first prize in the competition.

Social Validity

After the intervention was completed, the experimenter conducted a teacher/parent survey regarding acceptability, feasibility, and satisfaction with the VBI. The questionnaire had eight items. Each item was rated on a 5-point Likert-type scale (1 = “strongly disagree” to 5 = “strongly agree”). In addition, the experimenter conducted a comprehensive interview with the teachers and parents to compare the effects before and after the interventions.

Interobserver Agreement

To assess interobserver agreement (IOA), another observer (a PhD student in special education) was presented with the results of the study. The experimenter randomly selected 30% of the video recording clips during all phases for each participant. The IOA was calculated by dividing the number of agreements by the number of agreements plus disagreements and then multiplying by 100. The agreement on the percentage of VBI for Yu averaged 96%, with a range from 94% to 98%, 94% for Han, with a range from 92% to 96%, 97% for Ben, with a range from 95% to 98%, and 96% for Wong, with a range from 97% to 98%.

Procedural Fidelity

The procedural fidelity of this study was evaluated by an adapted procedural fidelity checklist developed by LaCava. The fidelity checklist had nine steps that were divided into three sections: planning, intervention, and session analysis. The researcher used the checklist to confirm the steps required for procedural fidelity with the intervention. A trained observer completed the procedural fidelity checklist for steps five to seven. The researcher and the observer marked the steps that were implemented and provided a score of one for each step. The procedural fidelity data were collected during 50% of VBI sessions across the participants. The procedural fidelity percentage was calculated by dividing the number of correct steps by the total steps and multiplying by 100. The data on procedural fidelity ranged from 97% to 99%, with an average of 98% across participants.

Social Validity

The average ratings were 4.67 (SD = 0.56) for the acceptability of the intervention, 4.6 (SD = 0.49) for feasibility, and 4.23 (SD = 0.81) for satisfaction, which indicates good social validity using the VBI for emotion recognition. In addition, teachers and parents in the open interview suggested that there was an improvement in emotion recognition, and they hoped to continue the intervention. For example, Yu’s mother stated, “He can say it was happy and point out the cartoon’s emotions when he watched the cartoon at home.” “Ben can quickly match emotion pictures compared to his first time training and increased compliance in the classroom. He becomes more interested in facial expressions.” Ben’s teacher said. Han’s father stated, “He can control his emotions, particularly when his preferred items were taken away, which was a huge change.” Wong’s father stated, “My son was truly excited to go to the therapy room and watch cartoons, which made us happy.”

Limitations and Directions for Future Research

The results and implications of this study’s findings must be explained relative to several limitations. First, the study did not develop recognition of complex emotions in the intervention, as previous research found that the majority of children with ASD first find difficulty with basic emotion recognition, followed by complex emotions (Lacava, Golan, Baron-Cohen, & Smith Myles, 2007; Rice, Wall, Fogel, & Shic, 2015). Thus, future studies should seek to replicate and extend this intervention with complex emotions. An additional limitation is that the study only involved matching facial expressions. We recommend that future researchers conduct more facial expression tasks, such as sorting facial expressions and odd-one-out facial expression tasks, to examine the intervention’s effects on emotion understanding for children with ASD. Furthermore, future research may consider comparing varying demographic variables. A systematic review found that only two studies have discussed the effect of VBI on adults with ASD, but even fewer studies have examined the effect of VBI on adults with ASD who have limited speech (Lee, Lam, Tsang, Yuen, & Ng, 2018).