Efficacy of a Facial Emotion Training Program for Children and Adolescents with Autism Spectrum Disorders

Abstract

Twenty-five high-functioning, verbal children and adolescents with autism spectrum disorders (ASD; age range 8–15 years) who demonstrated a facial emotion recognition deficit were block randomized to an active intervention (n = 12) or waitlist control (n = 13) group. The intervention was a modification of a commercially-available, computerized, dynamic facial emotion training tool, the MiX by Humintell©. Modifications were introduced to address the special learning needs of individuals with ASD and to address limitations in current emotion recognition programs. Modifications included: coach-assistance, a combination of didactic instruction for seven basic emotions, scaffold instruction which included repeated practice with increased presentation speeds, guided attention to relevant facial cues, and imitation of expressions. Training occurred twice each week for 45–60 min across an average of six sessions. Outcome measures were administered prior to and immediately after treatment, as well as after a delay period of 4–6 weeks. Outcome measures included (a) direct assessment of facial emotion recognition, (b) emotion self-expression, and (c) generalization through emotion awareness in videos and stories, use of emotion words, and self-, parent-, and teacher-report on social functioning questionnaires. The facial emotion training program enabled children and adolescents with ASD to more accurately and quickly identify feelings in facial expressions with stimuli from both the training tool and generalization measures and demonstrate improved self-expression of facial emotion.

Appendix

Training Protocol

Please note the following information about the implementation of the MiX™ with the coaching and imitation components is available in the published study by Russo-Ponsaran et al. (2014). The protocol was designed by a doctorate-level researcher, a clinical psychologist, and a bachelor-level researcher. It is well-documented in a written instruction manual. Since the training was manualized, coaches followed a standardized format for each session and all verbal prompts were scripted, ensuring that the same content and quantity of information was being conveyed.

Coaches included the original bachelor-level researcher who assisted with development and a master’s-level researcher. Both coaches exceeded “expert” levels of facial emotion recognition per the pre-existing criteria provided by the MiX (≥90 % at 1/5 s). Coaches tracked participants’ progress at each session using a standard form that documented the children’s previous session’s rate and accuracy scores, current session progress, an emotion-review plan, and the rate to be used for the forthcoming session.

Training consisted of 1-h sessions twice per week for up to 8 weeks (16 sessions). During the first session, the child was asked to define each of seven emotions (joy, sadness, fear, anger, contempt, surprise, and disgust). If the child was unable to define an emotion, the coach provided a standardized, age-appropriate definition for the emotion from the training manual (e.g., “sad” was further described as “unhappy, gloomy, or down”, while “contempt” was further described as “intense dislike”). Each of the children then indicated understanding after the definition was provided and then training began. With the exception of the emotion word review conducted at the first training session, each training session followed the same format: didactic instruction with imitation exercises, practice test, and competency test.

At the end of every training session, the child would complete a competency test. Presentation speed was set manually and the child would either move to a faster or slower rate at the next training session based on a scaffold design (Fig. 2). All children began training with the MiX presentation speed set to ½ s, which is the slowest micro-expression rate available. The child advanced to the next faster presentation speed at the subsequent session if 80 % or greater accuracy was achieved on the competency test’s overall percent correct. If, instead, the child scored below 80 % at a competency test, a slower presentation rate would be employed during the next session. The child was considered to have mastered the scaffold training protocol once reaching a set criterion of 80 % correct at 1/5 s on the MiX post-test on three consecutive sessions. This criterion was consistent with the MiX program definition of proficiency (Matsumoto, personal communication, November 16, 2012). Therefore, we adopted the same criteria for evidence of proficiency in our study. Three consecutive sessions were used in order to rule out spurious results. To quantify immediate effects of treatment, at the third and final consecutive session at which the child reached criterion for completion, the outcome assessment battery of direct and indirect facial emotion and general social functioning were re-administered.

Didactic Instruction

Sessions began with the MiX video-based didactic instruction about how each emotion is expressed with the face. The instructional videos included audiovisual instruction with a voice over describing facial components while videos of actors demonstrated the expressions. Emotion-specific videos were first played without interruption at a speed automatically set within the MiX. Then, the video was re-started. As the voice-over described each part of the face, the coach would hold up a customized screen cover that showed only the part of the face being described to focus attention on key facial features. For example, when describing the eyes, a screen cover that blocked all but the eyes was used. After describing each component, the video would be replayed and the entire face was unveiled.

Imitation

Immediately after each facial component was described, children were prompted to imitate that part of the face with their own faces. During imitations, the child was able to see his or her own face on the computer screen through a web camera. Imitations were executed for eyes (3 practices), mouth (3 practices), and then the whole face (5 practices) for each emotion, with the exception of fear and surprise. For these emotions, since there is additional information within eyebrow movement, imitation of eyes (3 practices) and eyebrows (3 practices) were practiced independently. Although the child could request a break at any time, a formal break was offered after all emotion video instruction and imitation components were completed.

Repeated Practice

Following instruction and imitation exercises, the child completed a practice test (42 items) at the designated session speed. Practice test items of faces varied by gender and ethnicity. Practice items spanning each of the seven different emotions were randomly presented. A child could indicate a response either by verbalizing the name of the emotion s/he thought was being expressed or by pointing to an emotion label from those listed on the computer screen. During practice items, the coach would then click on selected response, and the child was immediately notified if s/he correctly identified the emotion for each item. The child was given the opportunity to repeat each missed item once. If the child did not correctly identify the emotion on the second attempt, the coach would label the correct emotion and accompanying facial movement for the child according to the scripted instruction from the instruction manual.

Competency Testing

Once the didactic component and practice items were completed, the child took a competency test (28 items) from the MiX. The competency test was conducted similarly to the practice items, with the exception that the child would not receive item-level feedback and was not allowed a second chance after a selection was made. At the conclusion of the competency test, the child would receive a percent correct for overall performance and for each emotion at the designated presentation speed.

Subsequent Sessions

At the start of each subsequent session, the coach reviewed any individual emotions for which the child scored <50 % on the competency test from the previous session. During the review, the child would watch and listen to the entire “Review” video from the MiX for that particular emotion(s). The session would then continue with the standard didactic instruction videos and imitation exercises as previously described. If a child scored at least 50 % accuracy on all emotions on the MiX competency test, then training followed the standard didactic instructions and practice, without the use of additional review videos. The administration of MiX internal competency tests at varying speeds during each session allowed for ongoing monitoring of each child’s emotion recognition accuracy. Given the adaptive format, the total number of training sessions varied according to each child’s progression through the program based on both speed and accuracy on the MiX competency test.