Teaching by Listening: The Importance of Adult-Child Conversations To Language Development
Teaching by Listening: The Importance of Adult-Child Conversations To Language Development
Teaching by Listening: The Importance of Adult-Child Conversations To Language Development
quisition among infants and toddlers. Address correspondence to Frederick J. Zimmerman, PhD,
University of California, Los Angeles, Department of Health
METHODS: Two hundred seventy-five families of children aged 2 to 48 Services, School of Public Health, Los Angeles, California. E-mail:
months who were representative of the US census were enrolled in a fredzimmerman@ucla.edu
cross-sectional study of the home language environment and child PEDIATRICS (ISSN Numbers: Print, 0031-4005; Online, 1098-4275).
language development (phase 1). Of these, a representative sample of Copyright © 2009 by the American Academy of Pediatrics
71 families continued for a longitudinal assessment over 18 months FINANCIAL DISCLOSURE: Drs Gilkerson, Xu, Gray, Yapanel, and
(phase 2). In the cross-sectional sample, language development scores Mr Richards are employed by the LENA Foundation a non-profit
organization that developed the data-collection product. To the
were regressed on adult word count, television viewing, and adult-child extent that this article will raise awareness of their product,
conversations, controlling for socioeconomic attributes. In the longitu- they may benefit from its publication. However, the specific
results of the article do not create a conflict of interest for
dinal sample, phase 2 language development scores were regressed them. Drs Zimmerman and Christakis have no material interest
on phase 1 language development, as well as phase 1 adult word count, and no financial relationships relevant to this article to disclose.
television viewing, and adult-child conversations, controlling for socio-
economic attributes.
RESULTS: In fully adjusted regressions, the effects of adult word count
were significant when included alone but were partially mediated by
adult-child conversations. Television viewing when included alone was
significant and negative but was fully mediated by the inclusion of
adult-child conversations. Adult-child conversations were significant
when included alone and retained both significance and magnitude
when adult word count and television exposure were included.
CONCLUSIONS: Television exposure is not independently associated
with child language development when adult-child conversations are
controlled. Adult-child conversations are robustly associated with
healthy language development. Parents should be encouraged not
merely to provide language input to their children through reading or
storytelling, but also to engage their children in two-sided conversa-
tions. Pediatrics 2009;124:342–349
342 ZIMMERMAN et al
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014
ARTICLES
The amount of language input a child quent child language development Of the 364 participants invited to par-
receives before age 3 is significantly among children who were between ticipate, 334 (92%) participants com-
and strongly associated with subse- the ages of 2 and 36 months at enroll- pleted consent forms and were en-
quent language acquisition and cogni- ment. We hypothesized that adult rolled. During the course of the
tive development.1–8 Because of this re- speech would have a beneficial effect original 6-month study period, 13 par-
lationship, pediatricians and others on child language development, but ticipants dropped out or moved away,
are encouraged to advise parents to that this beneficial effect would be and final assessments could not be
provide as much language input to partially mediated by an increase in scheduled with an additional 8 fami-
young children as possible, through adult-child conversations. lies. Five participants were excluded
reading, storytelling, and simple nar- because their recordings were all ⬍12
ration of daily events. Although this ad- METHODS hours in duration; 5 were excluded be-
vice is undoubtedly sound, it may not This research relies on data collected cause their Preschool Language Scale
place enough emphasis on the child’s in children’s own environments: (PLS) data were judged by raters to
role in language-based exchanges. homes, playgrounds, schools, and any- have poor validity; and 28 were ex-
The framing of advice offered to par- where else children use or hear lan- cluded because their PLS assessments
ents may differ depending on how guage. Data were collected for 12-hour were not completed during phase 1.
adult speech is understood to foster periods 1 day a month for 6 months (or These exclusions left a sample of 275
child language development. If adult over 18 months in the longitudinal (82% of those enrolled). The final sam-
speech input is presented as intrinsi- sample) by using a new product called ple is highly representative of the US
cally valuable, because it serves as a LENA (Language Environment Analysis census data with regard to maternal
model for language that children intu- [LENA Foundation, Boulder, CO]). The education.
itively copy, then parents can conclude hardware component is a small digital For a randomly chosen day, parents
that the more adult speech the better, recorder called a digital language pro- were instructed to begin recording
even if some of this adult speech cessor, which fits into a pocket on a when their child woke up in the morn-
comes through television or videos. special vest worn by the child. The dig- ing and to continue without interrup-
Many parents have drawn exactly such ital language processor weighs 2 oz tion until bedtime. Parents were given
conclusions.9,10 On the other hand, if and can hold 16 hours of digitally re- the option of deleting all data on the
the primary value of adult speech is to corded sound.14 The software compo- device if they felt that anything re-
potentiate child speech as part of a nent consists of a digital sound ana- corded during that day compromised
trial-and-error, experiential process of lyzer that produces estimates of the their privacy. Only 1 parent chose to
language acquisition, then adult child’s exposure to adult speech, child exercise this option. Each child con-
speech is valuable inasmuch as it fos- speech, and television during the re- tributed an average of 4.7 recording
ters child speech, and either adult cording period. sessions throughout the 6-month
speech or electronic stimulus that period.
crowds out child speech may be Study Sample Participants’ language capacity was
counterproductive. Parents of children aged 2 to 48 formally assessed by a speech lan-
Heavy television viewing in the early months were invited to participate guage pathologist by using the Pre-
childhood years has been shown in through advertising in local newspa- school Language Scale, Fourth Edition
previous studies to be associated with pers and direct-mail solicitation. From (PLS-4).15 These assessments oc-
poor development of language and among those expressing an interest, a curred throughout the 6-month study
reading and math skills, although the sample was invited to participate, period, with each child assessed an av-
reasons for this adverse association stratified on maternal education and erage of 2.3 times.
are still not well understood.11–13 the age of the child. Because the em- A subset of 80 families, selected to be
This analysis uses a large, unique phasis of the study was on language representative of the entire sample
new data set of parent and child lan- development, by design all but 15 of the stratified by age groups, was recruited
guage use collected naturalistically children were aged 2 to 36 months. to participate in an 18-month-long con-
in the home environment to test the Children whose parents reported a de- tinuation with the same protocol as de-
independent contributions of adult velopmental or language delay were scribed above. This subsample pro-
language input, child language use, excluded. Only children in English- vides the opportunity to conduct
and television viewing on subse- speaking households were eligible. longitudinal analyses and is referred
344 ZIMMERMAN et al
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014
ARTICLES
near proximity. To facilitate interpret- Within each set of regressions, we There is considerable sample varia-
ability, we rescaled this variable in tested whether the effects of adult tion around these means, and in par-
thousands. Conversational turns esti- word count and television were medi- ticular the variation is higher for con-
mates were rescaled in hundreds. ated through conversational turns. Ac- versational turns (for which the
cordingly, we conducted 5 regressions coefficient of variation is 54%) than for
Statistical Methods within each sample set: 1 each in adult speech (coefficient of variation:
We conducted both cross-sectional which the 3 main predictors were sep- 34%).
and longitudinal regressions. In both arately included; 1 in which all 3 main
The independent effects of all 3 main
predictors were simultaneously in-
sets of regressions, we regressed predictors on language development
cluded; and 1 regression of conversa-
PLS-4 scores on the recorded mea- are significant in the cross-sectional
tional turns on adult word count and
sures of adult word count, conversa- regressions fully adjusted for social
television.
tional turns, and television, controlling and demographic characteristics
for sociodemographic variables, in- RESULTS (Table 3). Each 1000-word increase in
cluding the child’s age, gender, and Table 2 presents descriptive statistics the adult word count is associated
race, the mother’s and father’s educa- for the cross-sectional sample. Chil- with a 0.44 increase in the PLS-normed
tion level, and household income. Fa- dren hear an average of some 13 000 score (95% confidence interval [CI]:
ther’s education was imputed where words spoken to them by adults each 0.09 – 0.79). On average, each hour of
missing, and a missing-value flag was day. Children participate in some 400 television viewing per day to which the
included in the regressions. Because adult-child conversational turns a day. child is exposed is associated with a
the main predictor variables as well as
the outcome variables were measured
at more than 1 time period for many TABLE 2 Sample Descriptive Statistics
participants, we averaged the mea- Variable Cross-Sectional Sample Longitudinal Sample
(N ⫽ 275) (N ⫽ 71)
sures over the number of recording
PLS-4 standardized score (phase 1), mean (SD) 108 (13) 105 (13)
sessions and included the number of PLS-4 standardized score (phase 2), mean (SD) 105 (16)
sessions used to create these aver- Adult word count (1000s/d), mean (SD) 12.8 (4.4) 13.2 (4.7)
Conversational turns (phase 1) (100s/d), mean (SD) 4.1 (2.2) 3.3 (1.9)
ages as covariates. To appropriately
Conversational turns (phase 2) (100s/d), mean (SD) 4.7 (1.9)
weight the different observations for Television exposure (foreground) (h/d), mean (SD) 1.0 (0.6) 1.2 (0.7)
the different number of sessions that Child demographics
Girl, % 51 58
contributed to the child averages,
Black, % 3 4
we used analytical weights, with the Latino, nonblack, % 8 8
number of sessions as the weighting Other nonwhite race/ethnicity, % 7 7
Age, mo (phase 1), mean (SD) 21.2 (10.8) 13.7 (10.0)
variable.
Age, mo (phase 2) 27.9 (11.2)
In the cross-sectional regressions, Household income
⬍$20 000 annually, % 20 23
we regressed PLS-4 scores in phase 1
$20 000–$40 000 annually, % 31 35
on the contemporaneously recorded $40 000–$60 000 annually, % 19 17
measures of adult word count, con- $60 000–$100 000 annually, % 19 15
⬎$100 000 annually, % 11 10
versational turns, and television. In
Mother’s education
the longitudinal regressions, we re- ⬍ High school, % 21 18
gressed phase 2 PLS-4 scores on High school graduate, % 24 20
Some college, % 29 37
phase 1 adult word count, conversa-
College degree or higher, % 26 25
tional turns, and television, also con- Father’s education
trolling for phase 1 PLS-4 scores. ⬍ High school, % 20 17
High school graduate, % 16 25
Although the longitudinal analysis is Some college, % 21 24
the stronger study design, it also has College degree or higher, % 24 28
Father’s education data unknown, % 19 6
substantially fewer participants than
No. of observation sessions over which PLS was 2.3 (1.2) 2.4 (0.7)
the cross-sectional analysis. Results measured, mean (SD)
from both sets of regressions are No. of observation sessions over which predictors 4.7 (1.3) 6.0 (1.3)
were measured, mean (SD)
reported.
b P ⬍ .01.
2.68 decrease in the language score this regression, unlike the cross- that children have with adult care-
(95% CI: ⫺5.25 to 0.11). Each hundred sectional one, the effect of adult word givers is robustly and positively as-
conversational turns per day is associ- count is not fully mediated by conver- sociated with scores on a well-
ated with a 1.92 increase in the lan- sational turns, but retains its signifi- validated measure of child language
guage score (95% CI: (1.12–2.73). cance and much of its magnitude. Tele- development in a variety of model
When all 3 of the main predictors are vision viewing is not a significant specifications.
simultaneously included in the regres- predictor of language development in The longitudinal results reported here,
sion, adult word count and television any of the longitudinal regressions. although emerging from a smaller
exposure are no longer significant, Conversational turns is a robust pre- sample than the cross-sectional re-
suggesting complete mediation by dictor in the longitudinal model (coef- sults, are particularly compelling, be-
conversational turns, which retains its ficient: 3.33 [95% CI: 1.58 –5.07]), and cause they control for the child’s base-
statistical significance with a slightly retains its significance and much of its line language development. This model
higher magnitude (coefficient: 2.08 magnitude in the fully saturated model teases apart the separate effects of a
[95% CI: 0.97–3.19]). In the regression (coefficient: 2.18 [95% CI: 0.18 – 4.18]). child’s own ability and a child’s conver-
of conversational turns, adult word Adult word count predicts subsequent sations with others on subsequent lan-
count has a significant and positive as- conversational turns (coefficient: 0.17 guage development.
sociation (coefficient 0.28 [95% CI: [95% CI: .06 –.29]). Phase 1 language These results are consistent with a
0.24 – 0.32]), whereas television expo- ability is not a significant or meaning- number of causal interpretations. It
sure has a significant and negative as- ful predictor of phase 2 conversational might be that children who are more
sociation (coefficient: ⫺0.51 [95% CI: turns. advanced with their language abilities
⫺0.79 to 0.23]). are better at initiating or prolonging
A similar pattern obtains in the longi- DISCUSSION conversations. It could also be that
tudinal regressions, in which phase 1 The analyses presented here collec- there is a common but unobserved
PLS-4 scores are controlled (Table 4). tively make a strong case for the im- variable that causes certain children
Adult word count is a significant pre- portance of adult-child conversations to have high language scores and to
dictor of subsequent PLS-4 scores (co- to early child language development. engage in a lot of conversation. The sig-
efficient: 1.34 [95% CI: 0.59 –2.10]). In The number of conversational turns nificance of conversational turns in
TABLE 4 Longitudinal Regression of Phase 2 PLS-4 Scores on Phase 1 PLS-4, Adult Word Count, Conversational Turns, and Television Exposure, and of
Conversational Turns on Adult Word Count and Television Exposure
Dependent Variable PLS-4 (Phase 2), PLS-4 (Phase 2), PLS-4 (Phase 2), PLS-4 (Phase 2), CT (Phase 2),
Coefficient (95% CI) Coefficient (95% CI) Coefficient (95% CI) Coefficient (95% CI) Coefficient (95% CI)
Predictors
Phase 1 PLS-4 0.46 (0.24 to 0.68)a 0.46 (0.21 to 0.70)a 0.32 (0.10 to 0.54)a 0.38 (0.16 to 0.61)a 0.03 (⫺0.01 to 0.06)
Adult word count (1000s/d) 1.34 (0.59 to 2.10)a 0.82 (0.05 to 1.60)b 0.17 (0.06 to 0.29)a
TV exposure (foreground) (h/d) 1.45 (⫺4.71 to 7.61) ⫺0.38 (⫺5.90 to 5.14) 0.14 (⫺0.66 to 0.94)
Conversational turns (100s/d) 3.33 (1.58 to 5.07)a 2.18 (0.18 to 4.18)b
R2 0.68 0.60 0.69 0.71 0.56
Results adjusted for child’s age, gender, race/ethnicity, mother’s and father’s education, household income, and number of recording sessions (N ⫽ 71).
a P ⬍ .01.
b P ⬍ .05.
346 ZIMMERMAN et al
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014
ARTICLES
predicting phase 2 language scores own have been shown to produce im- interactions, including emotional tone,
even when phase 1 language ability is provements in children’s home liter- questions and responses, whether
controlled in a longitudinal regression acy environment at very low cost.30–32 adult speech is directed to the key
militates against such interpretations, Yet the most effective reading is dia- child or merely overheard by him or
as does the nonsignificance of phase 1 logic reading, which involves explicitly her, and other qualitative differences
language ability in predicting phase 2 soliciting language use by the among dyads. Recent research sug-
conversational turns. child.33–36 More generally, parents gests that children are able to learn
Finally, it could be that the child-adult should be taught that although adult vocabulary by overhearing adult
conversations are themselves causing speech is valuable, an equally impor- speech, and this intriguing finding, too,
language development. Such a causal tant goal should be to get kids talking suggests important new avenues for
conclusion would be consistent with as much as possible. this research.42 More work needs to be
several strands recently introduced Television viewing before age 2 has performed to refine the machine-
in the empirical and theoretical litera- previously been associated with de- based measurement of television ex-
tures on child language acquisition. lays in language development and posure, an important component of
Parents are most efficient at promot- poor reading skills.11,13,37 Television the early childhood environment.10,13
ing child language development when may adversely affect language devel- CONCLUSIONS
they calibrate their own speech to be opment by limiting opportunities for This research provides strong, albeit
just challenging enough for the child, parental language input38 or by limit- not absolute, evidence from naturalis-
neither so simplistic that the child ing opportunities for child speech. Re- tic observations that adult-child con-
learns nothing from the parent’s model, cent research has shown that al- versations are an essential component
nor so sophisticated that the child is though preschoolers can learn of child language development. Parents
bewildered. This just-challenging zone vocabulary from television, when used should continue to be encouraged to
has been termed the “zone of proximal without parental interaction, televi- provide speech input to their children by
development.”25–27 Because maintain- sion is an ineffective medium for incul- talking to them, reading them books, and
ing adult speech in the zone of proxi- cating language skills among infants by telling them stories. At the same time,
mal development depends on the and toddlers.20,39–41 When conversa- it should be made clear to parents that
adult’s being in touch with the child’s tional turns are included in the regres- an important goal of this talk is to elicit
rapidly changing abilities, frequent ex- sions, television drops from signifi- talk from the child. Reading and story-
posure to child language through cance, suggesting that the adverse telling should be punctuated by ques-
adult-child conversations may help effects of television exposure, if any, tions and exchanges, and it may be
keep the adult’s own speech in the would operate by reducing opportuni- appropriate to counsel parents to en-
zone of proximal development.28 Child ties for adult-child interactions. courage parent-child conversations.
language development has also been This research leaves many important Parents should strive to read and talk
shown to benefit from active correc- questions unanswered. We were not with children and not merely to them.
tion of errors by adult speakers.29 able to separately identify the effects Parent-child interactions are best when
More conversations mean more op- of child language use and conversa- they are a two-way street.
portunities for mistakes and, there- tional turns, which were highly corre-
fore, corrections. In addition, more lated with each other (correlation co-
ACKNOWLEDGMENTS
conversations may afford the child efficient: 0.84). It will be useful in future Dr Zimmerman’s time on this project
more opportunity to practice and con- research to tease out the independent was supported in part by National In-
solidate newly acquired language. Fi- contribution of these 2 similar factors. stitute of Mental Health grant
nally, more conversations may indi- It was beyond the scope of this study to 5K01MH64461-5. Data collection and
cate more adult responsiveness to the identify all the factors that promote cleaning were paid for by The LENA
child’s communication in general. adult-child conversations, but this is Foundation.
These results have meaningful impli- clearly a high priority for future re- We gratefully acknowledge Terrance
cations. Much of the advice given to search. The novel technology used in Paul for conceiving of the LENA system
parents has focused on the value of this research clearly has tremendous and for personally funding and direct-
reading to children as a way of facili- promise but is not yet able to distin- ing its development as well as the de-
tating adult language input. Indeed, guish more nuanced, but undoubtedly velopment of the LENA Natural Lan-
reading-promotion programs on their very important, elements of child-adult guage Corpus.
348 ZIMMERMAN et al
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014
ARTICLES
28. Tamis-LeMonda CS, Bornstein MH, Baumwell L. Maternal responsiveness and children’s achieve-
ment of language milestones. Child Dev. 2001;72(3):748 –767
29. Chouinard MM, Clark EV. Adult reformulations of child errors as negative evidence. J Child Lang.
2003;30(3):637– 669
30. Mendelsohn AL, Mogilner LN, Dreyer BP, et al. The impact of a clinic-based literacy intervention on
language development in inner-city preschool children. Pediatrics. 2001;107(1):130 –134
31. Weitzman CC, Roy L, Walls T, Tomlin R. More evidence for reach out and read: a home-based study.
Pediatrics. 2004;113(5):1248 –1253
32. Sharif I, Rieber S, Ozuah PO, Reiber S. Exposure to reach out and read and vocabulary outcomes in
inner city preschoolers. J Natl Med Assoc. 2002;94(3):171–177
33. Arnold DH. Accelerating language development through picture book reading: replication and
extension to a videotape training format. J Educ Psychol. 1994;86(2):235–243
34. Sénéchal M. The differential effect of storybook reading on preschoolers’ acquisition of expres-
sive and receptive vocabulary. J Child Lang. 1997;24(1):123–138
35. Hargrave AC, Sénéchal M. A book reading intervention with preschool children who have limited
vocabularies: the benefits of regular reading and dialogic reading. Early Child Res Q. 2000;15(1):
75–90
36. Whitehurst GJ, Lonigan CJ. Child development and emergent literacy. Child Dev. 1998;69(3):
848 – 872
37. Wright JC, Huston AC, Murphy KC, et al. The relations of early television viewing to school readiness
and vocabulary of children from low-income families: the early window project. Child Dev. 2001;
72(5):1347–1366
38. Tanimura M, Okuma K, Kyoshima K. Television viewing, reduced parental utterance, and delayed
speech development in infants and young children. Arch Pediatr Adolesc Med. 2007;161(6):
618 – 619
39. Krcmar M, Grela B, Lin K. Can toddlers learn vocabulary from television? An experimental ap-
proach. Media Psychol. 2007;10(1):41– 63
40. Kuhl PK, Tsao FM, Liu HM. Foreign-language experience in infancy: effects of short-term exposure
and social interaction on phonetic learning. Proc Natl Acad Sci U S A. 2003;100(15):9096 –9101
41. Barr R, Muentener P, Garcia A. Age-related changes in deferred imitation from television by 6- to
18-month-olds. Dev Sci. 2007;10(6):910 –921
42. Akhtar N, Jipson J, Callanan MA. Learning words through overhearing. Child Dev. 2001;72(2):
416 – 430
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014
Teaching by Listening: The Importance of Adult-Child Conversations to
Language Development
Frederick J. Zimmerman, Jill Gilkerson, Jeffrey A. Richards, Dimitri A. Christakis,
Dongxin Xu, Sharmistha Gray and Umit Yapanel
Pediatrics 2009;124;342
DOI: 10.1542/peds.2008-2267
The online version of this article, along with updated information and services, is
located on the World Wide Web at:
http://pediatrics.aappublications.org/content/124/1/342.full.html
Downloaded from pediatrics.aappublications.org at UCSF Kalmanovitz Library & CKM on December 5, 2014