• PDF Slides

           The purpose of this proposal is to describe research being conducted using a breakthrough new measurement instrument designed to provide parents and clinicians with automatically generated feedback about the language learning environment and development of infants and toddlers.  Below we describe this new measurement system along with the normative information collected thus far, followed by a report of pilot data for three children at risk for autism spectrum disorder (ASD). At the time of the conference we will also present the results from a larger study being conducted with young children identified with ASD.

Background

           Researchers and clinicians know very little about the natural language environment of young children with ASD and how this compares to typically developing children.  Professionals working with children at risk for language and developmental disabilities have traditionally been restricted by limitations in measurement technologies.  Typically, information about a child’s language environment and development has been obtained through parent report questionnaires sometimes aided by brief periods of professional observation in a clinic or home environment.  Clinical evaluations typically involve the administration of standard assessments designed to provide comprehensive information about the child’s overall development. These evaluations may provide a valid snapshot of the child’s general skill level.  However, they generally provide little or no information about the child’s actual communication performance and virtually no valid information on the child’s language learning environment.  Indeed, we know very little about the typical daily language learning environments of young children with ASD.  Are these environments similar to those of typically developing children, or do they differ in ways that may reflect the child’s ASD and may further complicate development?  These are fundamentally important questions that have significant implications for both evaluation and optimal early intervention. However, until now it has not been possible to answer these questions except in the most limited ways due to the limitations of environmental measurement systems, which have not advanced significantly since the invention of portable tape recorders and video cameras.    

Autism spectrum disorders provide a unique challenge for professional evaluation.  Social initiation and response behaviors are important considerations in the identification of ASD, but any child’s natural emotional and social reaction to a stranger in a clinical setting can complicate evaluation.  Further, although many treatment programs are designed to modify parent behavior, the efficacy of intervention programs in the home is most often measured by behavior in the clinic; it has been impossible for the clinician to know what is happening in the home when no observer is present.  Given the limitations of current evaluation methods, the need for a measurement device that can be used in the home without a clinical observer is apparent.

The LENA System

           Recent remarkable advancements in speech identification technology have made development of the LENA (Language Environment Analysis) System possible.  LENA is a language monitoring and feedback system designed to provide researchers, clinicians, and parents with automatically generated information about the in-home language environment of infants and toddlers; clinical observation is not necessary.    

The LENA System includes the LENA digital language processor (DLP) that children 2-36 months of age wear in the pocket of custom-made clothing. It records everything the child says and hears over a continuous 16-hour day. The audio data is transferred to a computer and analyzed by the LENA software. Parents and clinicians can access automatically generated feedback reports to view objective information about their child’s language environment and development.

The Adult Word Count (AWC) report provides an estimate of the number of adult words the child hears, and the Conversational Turns (CT) report estimates the number of conversational interactions the child engages in with an adult.  LENA also identifies and analyzes child speech, filtering out vegetative sounds, cries and other fixed signals, then provides information about how often the child is talking or vocalizing.  Child vocalizations are defined as child speech of any length surrounded by 300 ms of silence.  A child vocalization can be a word, a sentence of any length or a babble of any length.   The Child Vocalizations (CV) report provides information about the frequency with which the child vocalizes.  The LENA reports permit AWC, CT and CV estimates to be viewed as hourly, daily, or monthly totals. Daily AWC and CT percentile ranking estimates based on a normative database are reported in the LENA software. Observational reliability evaluations comparing human transcriptions to LENA generated data have shown LENA measurements to be reliable and valid.

Normative information

<>In January, 2006, we began the Infoture Natural Language Study, an ongoing data collection effort designed to provide information about the language environment of infants and toddlers using LENA.  Data collected during this study contributes to normative information for use with LENA and provides new insight into the language environment of children that has never before been accessible. 

<>The normative phase of the Infoture Natural Language Study was conducted between January 2006 – June 2006 (see Gilkerson and Richards 2007 for a detailed overview).  A total of 314 normally developing infants and toddlers from monolingual English-speaking households were recruited from the Denver-metro area.  The normative phase participant pool was representative of the US Census with respect to mother’s attained education (e.g. 24% college degree, 29% some college, 24% high school diploma, 23% high school drop outs).  Participants were sent LENA DLPs in the mail and provided day-long audio recordings of their natural language environment once a month.  Parents did not receive feedback from the recordings sessions.  Participant children were evaluated by a speech language pathologist using a battery of standard assessments in order to obtain information about their language abilities (e.g. Preschool Language Scale-4 (PLS-4) (Zimmerman, Steiner and Pond, 2002), Receptive-Expressive Emergent Language Test (REEL) (Bzoch, League & Brown, 2003)).  The resulting normative database contains over 18,000 hours of spontaneous speech data and can be referenced for information about the average number of adult words, conversational turns and child vocalizations occurring in the language environment of typically developing children 2-36 months of age.

ASD Pilot Study

In June 2007 we began a 6-month pilot study using LENA to compare the language environment of normally developing children to children diagnosed with significant language delays.  Thirty language delayed children 2-36 months of age were recruited from hospitals and clinics in the Denver-metro area.  Participants agreed to provide day-long audio recordings of their child’s natural language environment with the LENA DLP once a month for 6 months.  During the second month of the study, participants recorded on three different days within a seven day period; these sessions are referred to as “triple recording sessions”.  For each recording session, parents simply turned on the LENA DLP at the beginning of the day, put the DLP in the pocket of their child’s clothing, removed it at the end of the day and sent the DLP back to us for analysis.  Importantly, parents did not receive feedback about their recordings.  During the first month of the study, participants were also evaluated by a certified speech language pathologist using the PLS-4 and the REEL.

Speech language pathologists identified three study participants as at risk for ASD.  We have collected 6 day-long recordings for each of these participants over the past 4 months (once each month, plus two additional during the “triple” recording session).  Each 16 hour audio recording was automatically analyzed using the LENA software. 

We compared the language environment and development of the 3 at-risk for ASD children with the language environment and development of the 314 typically developing children from the normative phase of the Infoture Natural Language Study.  Table 1 displays the LENA estimates and percentile ranking for Adult Word Counts (AWC), Conversational Turns (CT) and Child Vocalizations (CV) for the at-risk children compared with the average (50^th percentile) for each measurement derived from our normative sample.  Note that because CTs and CVs increase significantly with age, the percentiles are age dependent.  AWCs do not increase with age so the 50^th percentile is consistent across the ages.

Table 1.  LENA measurements for children at risk for ASD compared to a normative sample.

		Age		AWC Average (%)			CT Average (%)			CV Average (%)			Language %
ID		(M)		ASD	Norm		ASD	Norm		ASD	Norm		PLS	REEL
5		29		12598 (62)	11270		375 (21)	577		1915 (25)	2631		1	1
13		34		6155 (11)	11270		419 (24)	618		2407 (38)	2766		1	1
19		28		5822 (10)	11270		213 (8)	568		1109 (8)	2602		1	1

Table 1 indicates that the average number of adult words is well below the mean (10^th - 11^th percentile) for two of the ASD children.  However, the AWC estimate for one of the ASD children is above the mean (62^nd percentile).  This result is not surprising, since these children are in speech therapy and the parents have been educated on the importance of talk, we would not necessarily expect their AWC estimates to be low.  Table 1 also shows that the ASD participants are engaging in conversational turns with an adult at a rate that is significantly below the mean (8^th-24^th percentiles) compared to age-matched peers.  Finally, results indicate that the ASD participants are below the mean (8^th-38^th percentile) with respect to how often they vocalize compared to other children their age.  Thus, results suggest that the automatic LENA measurements are consistent with what we might expect with children at risk for ASD.

LENA Autism Study

           The results from the pilot study suggest that the LENA System can be used to provide researchers and clinicians with important new information about the language environment and development of children with ASD compared to typically developing children.  To confirm this, we are currently conducting a larger study targeting 25 children who have been diagnosed with ASD.  Following on the results from the pilot study, the LENA Autism Study will investigate whether children diagnosed with ASD are significantly below the norm with respect to the frequency with which they vocalize and the frequency with which they engage in conversational turns with adults.  We will also see whether parents whose children have been diagnosed with ASD produce word counts similar to parents of typically developing children. 

LENA Autism Study participants will provide day-long audio recordings once per week for 6 weeks and will complete standard assessments including the MacArthur Communicative Development Inventory (Fenson et al., 2007) and the Child Development Inventory (Harold, 1992).  The LENA Autism study is currently in the recruitment phase and the first stages of data collection.  Study results along with the pilot data described above will be reported at the Autism Society of America 2008 conference.

Conclusion

The evaluation capabilities for behavioral scientists and clinicians have been greatly limited by the enormous difficulties and expense involved in measuring language performance in typical environments.  Indeed, prior to the recent development of the LENA System, professionals have been unable to unobtrusively and objectively obtain information about a child’s natural language environment.  The LENA System addresses these problems by providing automatically generated estimates of the number of adult words the child hears each day, as well as the number of conversational turns he/she engages in and the number of child vocalizations produced each day.  Importantly, the LENA System provides normative comparison information in the form of percentile ranking.  Thus, it allows us to compare the relative richness of the language environment and development of children with autism compared to typically developing children.            Pilot data suggests that the potential of the LENA System is far-reaching.  Results show that children at risk for ASD are significantly below the mean with respect to child vocalization frequency and conversational turns.  This result is promising, and suggests that the LENA System could aid in the early detection of ASD and, most importantly, the intensive early treatment of ASD. LENA may allow clinicians to easily and accurately measure and monitor intervention effects over long periods of time and to evaluate and modify intervention strategies. The LENA Autism Study will test this prediction with 25 children diagnosed with ASD.  These results will be presented at the conference.