Early Literacy Research Library (ELRL) - Article

Randomised Controlled Trial of Maternal Infant-Directed Reading Among Hospitalized Preterm Infants

Mayne, J., McGowan, E.C., Chiem, A., Nwanne, O., Tucker, R., Vohr, B.R. (2022) Randomised Controlled Trial of Maternal Infant-Directed Reading Among Hospitalised Preterm Infants. Acta Paediatrica, 111(10), 1921-1932. ,

Access: Institutional Access

Publication year


study description

Prospective randomised controlled trial.

core topic(s)

Early Literacy , Shared Reading

Population Characteristics

Infant/Newborn , Neonatal/NICU , Pregnancy/Postpartum

Exposures, Outcomes, Other

Brain/Neurocognitive , Child Behaviors and Skills , Child Development (general) , Clinic-Based Programs and Interventions , Language and Literacy Development , Parent Behaviors and Skills , Parent-Child Relationships/Interactions


A randomised trial to study the impact of a maternal-driven, infant-directed reading intervention on preterm infant language compared with matched controls.


Maternal infant-directed reading intervention.

outcomes evaluated

Infant-directed language.


Woman & Infants Hospital of Rhode Island's Neonatal Intensive Care Unit.


Infants born at 22-32 weeks in Women & Infants Neonatal Intensive Care were gestationally stratified to a reading intervention (n = 33) or standard care (n = 34). At 32-, 34- and 36-weeks postmenstrual age, 16-h language recordings were obtained in the hospital. Bivariate group comparisons and regressions adjusting for gestational age and multiples were run to predict word counts and conversational turns. Longitudinal analyses were conducted by negative binomial models containing intervention, randomised gestation group, recording number (1-3), an intervention × recording number interaction term and multiple birth adjustment by generalised estimating equations.

sample size

n=33 (dyads, intervention); n=34 (dyads, control); n=67 (dyads, total)


Measure of Language Characteristics: Language Environment Analysis (LENA) measures adult word counts, infant vocalizations, conversational turns, meaningful language, silence, background noise, and electronic noise measure from audio recordings.


In adjusted analyses, by 36-weeks postmenstrual age, infants in the reading group had twice the number of conversational turns as infants receiving standard care (Rate ratio 1.98, 95% CI 1.33-2.93, p < 0.05). In longitudinal analyses, only infants in the reading group had a significant increase in the conversational turns between 32- and 36-weeks postmenstrual age (Rate ratio 2.45, 95% CI 1.45-4.14, p < 0.05).


A maternal infant-directed reading curriculum in the hospital demonstrated a positive impact on interactive conversations by 36-weeks postmenstrual age.


A limitation of the study was the desired sample size was not attained as study enrolment ceased due to the beginning of the COVID-19 pandemic, with 67 infants completing our study instead of 70. A second limitation of our study was that the PI was not blinded to randomisation status. Although there is the possibility of unconscious bias, the study protocol was followed, as seen in Figure 2, without additional interactions of the PI with participants. Despite the varying language input of nurses, therapists and physicians, prior research demonstrates that parents are the primary contributors to the infant's language environment.2Counts, which are the primary outcomes of this trial, were derived by the LENA software, which runs a signal-processing algorithm that has been trained on a large sample of manually annotated natural language recordings and uses acoustical features to es-timate speech components.13 As a ‘word pedometer’, the LENA device plays an important role in commending and encouraging mothers that ongoing learning is occurring, as they may not notice the daily progress. The LENA feedback is intended to incentivise mothers and support continued language input, be it reading or talking. Although LENA has been used in several NICU studies, there is no normative dataset for preterm vocalisations. One study in Sweden found the agreement of LENA and human coders re-garding preterm child vocalisations and conversational turns was only fair in a small sample size of 14 infants.30 In addition, the presence of the LENA device may induce an unintended Hawthorne effect. Finally, only English proficient families were enrolled due to the educational component of our study. Further study would be important to explore language development of preterm infants in multilingual families.