Accuracy of the Language Environment Analyses (LENATM) system for estimating child and adult speech in laboratory settings.
Virginia A Marchman, Adriana Weisleder, Nereyda Hurtado and 1 others
PMID 32690113WHAT IT FOUND
LENA accurately ranked which caregivers talked more, but overestimated their absolute word counts by about 25%.
It poorly estimated child vocalizations, undercounting them at 1;7 and failing to capture developmental growth. Use it for relative adult speech, not absolute numbers.
Key findings
01LENA adult word counts correlated strongly with transcriber counts but significantly overestimated the absolute number of words, with an average error just above 25%.
02LENA child vocalization counts were less accurate, showing large error rates and failing to capture the true increase in vocalizations between ages 1;7 and 2;2.
03LENA was not validated for conversational turns because the study transcripts lacked the timing information required for this analysis.
STILL TO COME
How it was doneWhat they foundWhat it means for SLPs
Read the rest of this summary
You get three full summaries a month, free, and we do not ask for a card. Search, the TL;DRs and your library stay unlimited either way.
What it does not show
The study only included Spanish-speaking families, whereas LENA algorithms are trained primarily on English, which may affect generalizability. Recordings were short (20 minutes) compared to the recommended one-hour minimum, which may have inflated error rates. Caregivers were almost exclusively adult females, so results may not apply to male caregivers. The study did not validate conversational turns, a key metric often used in clinical assessments.
Declared interests
Funding was provided by the National Institutes of Health (Extramural and Non-U.S. Gov't grants). No specific conflicts of interest regarding the LENA Foundation or commercial entities are detailed in the provided text.
The easy way to misread this
Do not use LENA's absolute word counts for clinical baselines or progress monitoring. The system overestimated adult words by roughly 25% and significantly underestimated child vocalizations, meaning absolute numbers reported by LENA do not match actual speech production.