The Language ENvironment Analysis (LENA) System in Toddlers With Early Indicators of Autism: Test-Retest Reliability and Convergent Validity With Clinical Language Assessments.

The toddlers were young, a mean age of 25.87 months, and many had very little spoken language, so a lot of them scored at or near the bottom of the language tests. The authors say this reduced the spread of scores and may have made the links between LENA and the language measures look weaker than they really are. The two recordings used to test stability were a mean of 2.68 days apart. This says nothing about whether these measures hold steady over months, which is the timescale most therapists work on. LENA counts sounds and cannot judge words or grammar, so some recorded 'conversational turns' are probably a child and an adult vocalizing in sequence by chance rather than a real back-and-forth. The recordings were taken at the start of a larger trial, from children already enrolled in early intervention, so this sample may not look like the toddlers you see in other settings. Everything here is measured at one time point. Nothing in the study shows that more conversational turns cause better language. The reliability analysis rested on 81 children and the vocal development measures on fewer still, so those figures carry some uncertainty.

The authors declare no conflicts of interest. The work was funded by the National Institute of Mental Health.

Do not read this as evidence that recording a child's day with this device helps their language, or that adult word count or child vocalization count tell you anything about a toddler's language ability. Neither of those two counts related to any clinical language measure in this sample, and the paper is testing how well a research tool measures, not testing a treatment.