4. When to Lean on LENA Measures?

When to Lean on LENA Measures?

Posted by Dr Jill Lany on 01/07/2025 in Blogs
Add Comment

If you've ever wondered how much the language a baby hears affects how they learn to talk, you're not alone. This topic has been of central interest to language development researchers for well over 50 years. In the past decade, there’s been a resurgence of interest, largely due to new technology that makes it easier to capture children's language environments.

Compact wearable technologies – like voice and video recorders, physical activity monitors, and heartbeat sensors – now allow for unobtrusive collection of rich, continuous data (deBarbaro, 2019). These sensors store information on embedded chips, giving researchers and users access to huge volumes of data. Wearables have transformed many aspects of daily life: embedded in smartphones and watches, they track everything from steps to sleep.

The LENA®: A Window Into Children’s Language Environments

One such device, the  LENA® (short for Language ENvironment Analysis) is popular among language development researchers (Laudańska et al., 2025). The LENA is a small audio recording device designed to be worn by a child – typically in a vest pocket – and parents report that it is easy to use (McElwain et al., 2024). A single LENA recording device can capture up to 16 hours of data across one or two full days, far exceeding the short sessions typical in labs.

Importantly, the LENA system includes tools for automatically processing recordings (Greenwood et al., 2011). For example, it can tell the difference between human speech and electronic sounds, label them, and estimate how much speech a child hears during the recording period (Cristia et al., 2020).

Does the LENA Predict Language Outcomes?

A recent meta-analysis – a study that combines data from many others – reported that automated LENA measures do predict language development, though the strength of this relationship is modest (Wang et al., 2020). For example, more adult talk and more conversational turns are related to better language development during infancy and childhood.

However, I recently came across a new preprint suggesting that LENA word count measures captured during the first year of life do not predict preschool-aged language outcomes (Egan-Daily et al., 2025). This result stood out for a few reasons.

First, as just mentioned, a meta-analysis of 17 studies indicates that LENA data is generally predictive of language development (Wang et al., 2020). Second, a study from the same longitudinal sample used in the preprint found that LENA-derived measures relate to how quickly babies can process language in real time (Bergelson & Aslin, 2017). Many studies – including my own and other LuCiD researchers’– have found that real-time processing skills predict vocabulary growth, likely because they reflect core language learning processes (Lany, 2018; Lany et al., 2018a, 2018b; Peter et al., 2019). Third, my research team recently published findings that LENA measures taken during the same age window as in the preprint do predict  infants’ real-time processing skills (Homer et al., 2025).

Together, this work suggests that infants’ experience with adult speech, especially in conversations, plays a key role in language learning.

Reconciling Conflicting Findings

So how should we interpret these seemingly conflicting findings about the utility of LENA data? 

The preprint in question had some limitations. For example, the sample was small and not very diverse, making replication and extension critical. This is why meta-analysis is so valuable – it allows us to assess patterns across studies and contexts. And the recent meta-analysis suggests that LENA measures are predictive of language outcomes (Wang et al., 2020). 

However, it may be that LENA measures are better at predicting short-term language growth, rather than outcomes several years later, as the preprint tried to do.

Still, this prompted me to ask: What is the LENA best suited for? 

LENA’s Strengths

A key strength of LENA is its ability to record large amounts of natural language input. It captures far more data than would be possible in a lab setting.

Its portability is another big advantage. A LENA device travels easily with a child into homes and other spaces. Recording at home also reduces the need for families to visit labs, making studies more inclusive, and enabling settings generally inaccessible to researchers. One compelling example comes from NICU research, where LENA recordings revealed very low levels of speech exposure for preterm infants (Caskey et al., 2011).

There’s also solid evidence that LENA’s estimates of the volume of adult child speech are largely valid. While it tends to underestimate these, LENA metrics still align closely with hand-transcribed data (Cristia et al., 2020). This can save researchers enormous time and effort.

Beyond convenience, the LENA enables standardized comparisons across labs and contexts. It also makes it easier to sample key moments from long recordings, like the most talkative times, for more detailed analysis (Cristia et al., 2020).

The LENA is also being used to evaluate interventions, including those that  address parent beliefs about language learning or assess interaction quality in child care settings (Cunha et al., 2024; Joseph et al., 2022;  Suskind et al., 2016).

LENA’s Limitations

Still, the LENA has limitations.

It misses much of what happens outside its hearing range. And even audible speech can be hard to interpret without context. For example, it’s often unclear whether speech is directed at the child, another caregiver, a sibling, or even a pet. Turn-taking estimates can be unreliable (Cristia et al.,, 2020; Ferjan Ramírez et al., 2021), and LENA’s guesses about who is talking—adult, child, male, female—can be wrong (Cristia et al., 2021; Lehet et al., 2021).

In addition, the quality of LENA recordings may not always be sufficient for all fine-grained acoustic analysis. Nonetheless, my research group recently used the recordings to determine whether prosodic cues that support language learning are detectible in more natural contexts outside the lab – and we found that they are (Wang et al., 2024)!

Cost is another issue. LENA systems can cost thousands of dollars, making them hard to access for researchers outside well-funded institution. Open-access alternatives are under development (Laudańska et al., 2025), but they’re still in early stages.

Emerging Exciting Uses for the LENA

Despite these limitations, the LENA has untapped potential.

Beyond measuring children’s language input, the LENA has been used to assess children’s exposure to background noise (Simon et al., 2022; Suarez-Rivera et al., 2024) and music (Hippe et al., 2024).

Because of the volume and contextual diversity of the data, LENA is especially suited to capturing patterns that emerge over multiple timescales—from seconds and minutes to hours, days, or weeks (e.g., Warlaumont et al., 2022).

And researchers are now combining LENA with other wearables, such as head-mounted cameras, to gain insight into what children see while they hear language. One recent study even combined LENA with smartphone data, showing that parents talk less when using their phones (Mikhelson et al, 2024).

Conclusion

The LENA system offers valuable standardized tools for studying early language environments. While it has clear limitations – particularly in terms of cost, context sensitivity, and speaker identification – it remains a useful option for many research questions. 

Ongoing work is needed to identify best practices and determine when and how to lean on the LENA most effectively.

 

References

Author

Leave a Comment

* Indicates fields are required