Using LENA Automated Analysis to Monitor the Language Experience of Children During Therapy, Preschool and with Primary Caregivers

Background: Correlations have been reported between early language environments and children’s cognitive, social and emotional development (e.g., Hart and Risley, 1995), and interest in the causative influence of maternal responsiveness and turn-taking on the early development of typically-developing (TD) children is growing (e.g., Laundry, Smith & Swank, 2006).  Hart & Risley investigated the home language environment of TD infants and toddlers longitudinally for three years, completing in-home hour-long audio recordings and transcribing the interactions.  They reported positive correlations between the number of words children were exposed to before age 4 and their IQ and academic success through elementary school.  More recently, Landry and colleagues reported results of experimental studies with premature and TD infants using a short, intensive intervention focusing on maternal responsiveness and turn-taking.  Treatment group children experienced more adult-child interactions and showed significant elevations in communicative and social development compared to controls.

            There is surprisingly little early language environment research focusing on preschool and non-primary caregivers.  The paucity of such studies is likely due to logistical difficulties in collecting and transcribing naturalistic interactions through audio/video technology.  This issue is addressed by the LENA (Language Environment Analysis) framework, which comprises a lightweight audio recorder that children wear plus automated computer-processing tools.  This approach utilizes speech-recognition algorithms to segment the audio stream and generate estimates of 1) adult words spoken near the child, and 2) turn-taking interactions between caregiver and child (alternations between adult and child segments bounded by 5 seconds of silence/non-speech).

Objectives: The LENA framework was used to examine the language environments of children with ASD and TD children during and outside of preschool and therapy times.

Methods: Participants were 74 children with ASD between 24-48 months of age and 44 age-matched TD peers drawn from the LENA Research Foundation’s natural language corpus.  Participants recorded continuously throughout the day, and parents completed session diaries indicating specific times children attended therapy or preschool.  Rates for adult word counts and turn-taking were computed for therapy hours, preschool hours, and typical hours (i.e., when the child was not attending preschool nor in therapy with a professional).  Analyses include 831 recording sessions; 9,972 audio hours.

Results: For the ASD sample, adult word count was significantly higher during therapy (t(45)=5.78,p<.01) than during typical hours.  Likewise, ASD children engaged in significantly more turns during therapy than during typical hours (t(45)=4.83,p<.01). Comparing groups during preschool hours, ASD adult word count was significantly higher than TD adult word count (t(68)=2.84,p<.01), and ASD turns were significantly higher than TD turns (t(68)=6.32,p<.01).  During typical hours, ASD children engaged in fewer turns than TD children (t(106)=2.95,p<.01).

Conclusions: Children with ASD experienced more adult talk and greater vocal engagement while in therapy with a professional compared to more typical hours.  ASD children experienced a more enhanced language environment during preschool compared to TD children, while turn-taking during typical hours was significantly lower for children with ASD.  These results demonstrate the feasibility of automated analysis for monitoring language environments provided by parents, preschool teachers and therapists working with children with ASD.