Long-Term Effects of Early-Life Otitis

Media on Language Development

Anne Zumach
Ellen Gerrits
Michelene Chenault
Lucien Anteunis
Maastricht University Medical Centre,
Maastricht, the Netherlands
Purpose: The aim of the present study was to examine the long-term consequences of
early-life otitis media (OM) and the associated hearing loss ( HL) on language skills
of school-aged children.
Method: In a prospective study, the middle-ear status of 65 Dutch healthy-born
children was documented every 3 months during their first 2 years of life; language
comprehension and production were evaluated at 27 months and again at 7 years.
Results: The positive relation that was found between OM-related HL and language
development at 27 months could no longer be discerned at school age. Accordingly,
parent-reported HL between 2 and 7 years had no effect on scores at school age.
Conclusion: The present study shows that negative consequences of early-life OM or
the underlying HL on language comprehension and production appear to be resolved
by the age of 7. It also shows that parent-reported HL between 2 and 7 years is not
related to language skills at school age.
KEY WORDS: otitis media, language disorders, long-term results, hearing loss,
developmental outcomes

L
anguage development is generally assumed to have its sensitive
period during early childhood (Bailey, Bruer, Symons, & Lichtman,
2001; Knudsen, 2004; Werker & Tees, 2005), particularly during
the first 2 years of life when basic language skills are acquired, forming
the basis for further language development. Therefore, any reduction
in the quality of language input during this period could result in a set-
back that may never be compensated.
The present study addresses the long-term effects of early-life otitis
media (OM) and consequent hearing loss ( HL) on language development.
Recurrent OM during early life means not only a reduction in auditory
output but also fluctuations in the sound of speech during OM-positive
and OM-negative periods. A possible consequence is that the fluctuation
of sound input makes it harder to imprint speech patterns, which is a
crucial component of language learning.
Prospective studies addressing the long-term effects of early-life OM
on language development report contradictory findings. In a group of 147
children at age 7, Teele, Klein, Chase, Menyuk, and Rosner (1990) found
significant positive correlations between longer periods with OM in the
first 3 years of life and poorer articulation and language-related skills
such as usage of morphologic markers. However, Roberts, Burchinal, Davis,
and Collier (1991) and Paradise and Feldman (2007) found no relation-
ship between early-life OM and later language comprehension and pro-
duction abilities.
Prospective longitudinal studies with language assessments at dif-
ferent ages show that negative effects of OM on early language may be

34 Journal of Speech, Language, and Hearing Research • Vol. 53 • 34–43 • February 2010 • D American Speech-Language-Hearing Association

Downloaded from jslhr.asha.org on June 11, 2010

resolved by school age. Roberts, Burchinal, and Zeisel
(2002) found a relation between recurrent OM occur-
rences and HL in the first 2 years and expressive lan-
guage at age 5, but that this had been resolved by age 7.
Likewise, Schilder et al. (1993) found a relationship be-
tween OM incidents between 2 and 4 years and language
development in preschool children, but this relationship
had disappeared by ages 7–8.
In a meta-analysis conducted by Roberts, Rosenfeld,
and Zeisel (2004), the authors found either no associa-
tions or just very small negative associations of OM and
underlying HL with later language development. These
authors, as well as Roberts, Hunter, and colleagues (2004)
and Roberts, Rosenfeld, and Zeisel (2004), have been
critical of study designs that do not adjust for poten-
tially confounding variables such as low socioeconomic
status, at-risk birth, and neurological problems, which
may dilute estimates of the negative contribution of OM
and its underlying HL.
In most of these studies, OM—rather than the un-
derlying HL—is estimated, whereas it is hearing loss as
a result of OM that results in periods of fluctuating
hearing ability, which in turn may affect language de-
velopment during this sensitive period. Furthermore,
OM and underlying HL should be estimated in the first
2 years of life, as it is during that time period that the
highest incidence of OM occurs (e.g., Rovers, Schilder,
Zielhuis, & Rosenfeld, 2004).
In the original study group of the presently studied
children (Anteunis & Engel, 2000; Anteunis, Engel, et al.
2000), a detrimental effect of OM and its underlying fluc-
tuating HL on language development at 27 months was
found. Other prospective studies have also found that
language skills between 1 and 3 years were negatively
affected by preceding incidents of OM (Friel-Patti &
Finitzo, 1990; Paradise et al., 2000; Shriberg et al., 2000;
Vernon-Feagans, Emanuel, & Blood, 1997). This conforms
with findings of studies mentioned previously, although it
is in contrast with a study by Wright and colleagues (1988)
that did not find this effect at age 2.
In the present study, we considered repeated mea-
surements of OM and HL during the first 2 years of life
and related them to toddler and school-age language
skills in children. The existence of potential confounders
as suggested by Roberts, Hunter, and colleagues (2004)
has been addressed, and only healthy-born children have
been considered. Vernon-Feagans, Miccio, and Yont (2003)
presented a contextual transactional model of OM and de-
velopmental outcomes, whereby child characteristics and
the quality of the home were put forth as possibly im-
portant moderators. Therefore, characteristics such as
gender, being breastfed, cognitive development, and the
educational level of the parents are considered as po-
tential moderators in the present study.
Downloaded from jslhr.asha.org on June 11, 2010
Method
Participants
The study sample consisted of 65 children, 34 of
whom were boys and 31 of whom were girls. This sample
is a subset of a larger prospective study in which the
middle-ear status of 250 children was examined every
3 months during their first 2 years of life in the Maastricht
Otitis Media With Effusion Study (MOMES) at the Uni-
versity Hospital Maastricht ( Maastricht, the Nether-
lands) between October 1989 and April 1995 (Anteunis
& Engel, 2000). The original study group was composed
of 100 high-risk–born neonates and 150 healthy-born
children (Engel et al., 2000).
The original cohort of 150 healthy-born children had
been recruited from the General Delivery Outpatient
Department. All children from this original cohort who
had attended at least seven of nine scheduled ear, nose,
and throad ( ENT) consults between birth and 24 months
were invited to participate in the present study.
The 65 participating children were all healthy-born
children. They completed an additional assessment at
7 years (age range = 6.6–7.10 years; M = 7.3; SD = 0.4). All
of them were born in the southern part of the Nether-
lands ( Limburg), and their middle-ear status had been
followed from birth up to 24 months. At 27 months, lan-
guage comprehension had been tested in 53 of the par-
ticipating children, and language production had been
tested in 44 of the participating children. The children
had normal hearing and no current OM indication at the
time of language assessment. Two children had a slight
conductive HL during the first examination at age 7 and,
therefore, were rescheduled for a new appointment, dur-
ing which time their middle-ear status and hearing were
normal. The number of months the child had been breast-
fed was registered, and for the present study, this was
dichotomized into whether the child had any breast-
feeding. The parents’ educational levels at birth had
been categorized according to the Standard Onderwijs
Indeling (SOI) scale (Standaard Onderwijs Indeling van
het Centraal Bureau voor de Statistiek, 1999), whereby
both parents’ scores were summed obtaining a scale rang-
ing from 4 to 14. For the present study, this sum was
dichotomized so that 4–9 was low (vocational training,
high school, or less) and 10–14 was high ( higher voca-
tional training or higher education), with 50.8% of the
cases falling into the lower group.
Procedure
Evaluation at 0–27 months. Every 3 months between
birth and 24 months, the children underwent first otoscopy,
then hearing assessment, followed by tympanometry. At
27 months, only tympanometry was conducted.
Zumach et al.: Early Otitis Media and Language Development 35
 An experienced otologist and audiologist inspected
the external ear canal and the eardrum using a hand-held
otoscope ( Beta 200, Heine Optotechnik, D-82211). In un-
determined cases, an otomicroscope (Carl Zeiss, OPMI 9)
was used. The otoscopic findings were described and cat-
egorized in terms of normal, indefinite, or OM.
Tympanograms for both ears were recorded on a
Grason-Stadler 1723 middle-ear analyzer and were char-
acterized according to a modified classification (Engel,
Anteunis, Chenault, & Marres, 2000) according to Jerger
(1970). To standardize the diagnosis of the middle ear,
otoscopic and tympanometric findings were combined
in the MOMES diagnostic algorithm (Anteunis, Engel,
Hendriks, & Manni, 1999; Engel, Anteunis, Volovics,
Hendriks, & Marres, 1999), a modification of the one rec-
ommended by Cantekin (1983).
The hearing assessment method applied was de-
pendent on the age of the child. Up to the age of ap-
proximately 9 months, the infants were subjected to a
nonreinforced behavioral procedure ( BOA), described by
Biesalski (1971). After this age, the children underwent
a conditioned orientation reflex procedure (CORA), de-
scribed by Suzuki and Ogiba (1960). In both BOA and
CORA, narrow band noise (center frequencies of 0.5, 1, 2,
4, and 8 kHz and 30 dB HL octave slopes) delivered by
a clinical audiometer ( Maden OB 822) was presented.
Minimum response levels for both ears were obtained
with a 10-dB HL down /5-dB HL up search and were
defined in both methods as the faintest level at which the
participant responded consistently (at least two out of
three times) in the presence of an auditory stimulus. The
behavioral threshold was defined as the lowest intensity
level at which a behavioral response occurred in relation
to an auditory stimulus, agreed to by two observers. Be-
cause of differences in signal, sound field, and test protocol,
normative data for this setting were obtained by using the
data of the healthy-born children (from the original set of
150 children). The average threshold of all OM-free chil-
dren at each of the nine visits for the frequencies of 0.5, 1, 2,
4, and 8 kHz served as the reference threshold of 0 dB HL
for the corresponding age. The hearing threshold of the
children represents the difference between the behavioral
threshold and the reference level of OM-free children. In
this way, the obtained behavioral thresholds were cor-
rected and enabled comparable hearing thresholds for
visits at different ages (Anteunis et al., 2000; Zumach,
Gerrits, Chenault, & Anteunis, 2009).
At 27 months, spontaneous expressive language
was assessed by recording spontaneous language sam-
ples during an unstructured parent–child interaction
setting with a standard set of toys. The utterances were
analyzed according to the Grammatical Analysis of De-
velopmental Language Disorders (GRAMAT) framework
by Bol and Kuiken (1989) by a speech therapist. This analy-
sis is a Dutch adaptation of the descriptive morphosyntactic
36 Journal of Speech, Language, and Hearing Research • Vol. 53 • 34–43 • February 2010
Downloaded from jslhr.asha.org on June 11, 2010
framework by Crystal, Fletcher, and Garmen (1976). The
mean length of utterance ( MLU) and the mean length of
the five longest utterances ( MLUL) in words were com-
puted. Reliability of the GRAMAT transcription and
analysis was established by three independent research-
ers from the University of Amsterdam ( Department of
Linguistics and Literature) on a sample of 10 parent–
child interactions. The percentage of agreement, calcu-
lated using Cohen’s kappa (k), varied between 0.60 and
0.95. k is a conservative measure; therefore, these findings
suggest acceptable interrater agreement of the analysis.
Verbal comprehension was tested with the Dutch adap-
tation of the Reynell Developmental Language Scales
(Reynell, 1977; Schaerlaekens, Zink, & van Ommeslaege,
1993) by Bomers and Mugge (1982).
At 27 months, the general development was as-
sessed by one of two psychology assistants blinded to
the child’s history. This assessment included the Bayley
Scales of Infant Development (Bayley, 1969; van der
Meulen & Smrkovský, 1984), whereby cognitive develop-
ment was estimated by the nonverbal Bayley Mental
Development Index ( MDI).
Evaluation at school age. At age 7, otoscopy and a
tympanometric examination, conforming to the same
procedure used in the first 2 years, were completed by an
otolaryngologist who was blinded to the child’s OM and
hearing loss history.
Prior to this examination, retrospective data per-
taining to the period between 27 months and 7 years
were obtained via a questionnaire, which was completed
by the parents at home. The questionnaire regarded
whether there had been ENT consultations, placement
of tympanostomy tubes, speech therapy, or hearing prob-
lems during the age intervals of 0–2 years, 2–4 years,
4–6 years, and 6 years to age of testing.
Two subtests from the Dutch Language Tests for
Children (Van Bon, 1982) were used to test the children’s
general language ability: the Word Forms Production
( WFP) test and the Concealed Meaning (CM) test. WFP
is a test of morphological expressive language skills and
addresses knowledge of irregular word forms. The CM
test is receptive in nature, focusing on the child’s under-
standing of the nonexplicit content of sentences.
Of the 65 children assessed at school age, there
were 53 children for whom the cognitive development and
Reynell scores were available and 44 children for whom the
MLU and MLUL were available, resulting in a relatively
small sample size.
Statistical Analysis
Data analysis employed the statistical package SPSS
15.0. Descriptive statistics, correlation coefficients, and
regression coefficients for the various language scores
were obtained. Raw language scores were converted into
standard scores (scale from 1 to 10, with 5 as the mean).
To examine whether the present smaller subset is
representative for the larger sample of Anteunis, Engel,
and colleagues (2000), the correlation coefficients of the
smaller sample were compared with a similar analysis
in the larger core group of 124 children. In that study,
Anteunis, Engel, and colleagues (2000) examined which
factors influence language development at 27 months.
They found, in children that were either healthy born
(n = 76) or high-risk born (n = 48), that HL resulting from
OM has a negative effect on language production at
27 months in infants. Correlation coefficients that were
found for language production and comprehension scores
in relation to OM and HL in the two analyses were similar
to those in the present study.
The statistical analysis to investigate the effect of
OM and the underlying HL on language development
was conducted in two phases. First, language scores at
27 months ( Reynell and MLU) were examined relative
to the middle-ear status with consideration for possible
moderating factors. These factors were analyzed in the
dichotomized form. Thereafter, the two outcome scores
at 7 years—namely, WFP and CM tests—were examined
as dependent variables employing multivariate analysis
of variance (MANOVA).
Language scores at 27 months and number of OM
incidents and mean hearing loss in the first 2 years
were considered as explanatory variables. These last
two variables were also considered in a dichotomized
form whereby the median was the cutoff point to increase
power. The five language outcome scores were examined
relative to background characteristics (gender, breast-
feeding, and education level of parents) by applying t tests.
Holm’s step-down procedure (Holm, 1979) was applied
to correct for the possibility of a type I error due to mul-
tiple testing. Thereafter, multiple regression analysis
with list-wise deletion (the only cases that were included
were those that had valid values for the variables in the
model) was used to determine whether these background
characteristics had a moderating role in the relation be-
tween OM and HL and the language scores.
Results
Descriptive statistics of the number of OM incidents,
HL at each visit, the average overall visits in early life,
and language outcome scores at 27 months and 7 years are
presented in Table 1. As suggested by Vernon-Feagans
(2003), the number of OM incidents and average HL
were dichotomized, their median values being 3.5 and

Table 1. Tested variables at 0–27 months and at school age.

Variable N M Minimum Maximum SD

0–27 months
Number of OM incidents 65 3.8 0 9 2.2
HL (dB) visit
0 months 53 –1.5 –20.3 13.8 6.5
3 months 64 2.4 –13.5 38.4 10.6
6 months 64 6.7 –11.4 31.5 9.2
9 months 65 3.0 –9.1 24.6 8.1
12 months 64 0.1 –12.3 27.7 7.2
15 months 64 0.9 –10.4 18.4 6.6
18 months 64 0.4 –7.5 11.4 4.9
21 months 64 0.6 –9.4 20.7 5.6
24 months 63 2.0 –7.2 35.2 7.0
Average HL overall visits 63 1.6 –3.8 12.12 3.4
Cognitive development (MDI) 21 months 53 110.5 75 149 16.0
Language production (MLU) 27 months 44 2.3 1.3 3.7 0.7
Language production (MLUL) 27 months 44 4.8 1.6 8.0 1.7
Language comprehension (Reynell) 27 months 53 5.9 1 9.5 2
School age
Age 65 7.3 6.7 7.9 0.4
Language production (WFP) 65 4.9 1.8 9.5 1.6
Language comprehension (CM) 64 5.7 1.5 9.3 1.9

Note. OM = otitis media; HL = hearing loss; MDI = Bayley Mental Development Index; MLU = mean length of utterance;
MLUL = mean length of the five longest utterances; Reynell = Reynell Developmental Language Scales; WFP = Word Forms Production
test; CM = Concealed Meaning test.

Zumach et al.: Early Otitis Media and Language Development 37

Downloaded from jslhr.asha.org on June 11, 2010

1.1, respectively. All language scores had a normal dis-
tribution. Table 2 presents the language scores accord-
ing to gender, breastfeeding, parents’ education level,
dichotomized number of incidents of OM and HL, and
whether independent t tests resulted in statistically sig-
nificant different outcomes for the five language scores.
The correlation coefficients among the five language
scores, number of OM incidents, and HL are presented
in Table 3.
Moderating Factors in the Relation
Between Early-Life OM History and HL
and Language Skills at 27 Months
In Table 2, it can be seen that breastfeeding was
positively related to MLU at 27 months, whereas there
was no significant relation between any of the language
scores and gender. Furthermore, higher parental edu-
cation levels had a positive effect on all language scores
at early age and at school age. After correcting for mul-
tiple comparisons according to Holm’s step-down proce-
dure (Holm, 1979), Reynell and MLU scores were still,
and in this order, statistically significantly higher for the
more educated parent group.
Only MLU appeared to be somewhat affected by
dichotomized hearing loss, as more hearing loss resulted
in lower MLU. Upon adjusting for cognitive develop-
ment, the standardized coefficients for HL in predicting
MLU dropped from –0.372 to –0.298 ( p = .049), whereas
the standardized coefficient for cognitive development was
0.409 (p = .008), with the explained variance increasing
from 8.3% to 23.1%. Thus, even after adjusting for cog-
nitive development, which is a strong predictor, HL was
still a statistically significant predictor of MLU. Upon
adjusting for parents’ education, the standardized coef-
ficient for HL in predicting MLU dropped from –0.289 to
–0.239 (p = .107), whereas the coefficient for parents’ edu-
cation was 0.311 ( p = .038), with the explained variance
increasing from 8.3% to 17.8%. Inclusion of the moderat-
ing factor of the parents weakens the effect of hearing
loss on MLU.
Moderating Factors in the Relation
Between Early Life OM and HL
and Language Skills at 7 Years
Breastfeeding and gender were not related to any
language score at age 7. Higher parent’s education did
have a positive effect on both CM and WFP. After cor-
recting for multiple comparisons according to Holm’s
step-down procedure (Holm, 1979), CM scores were still
statistically significantly higher for the more educated
parent group.
Average HL in the first 2 years of life did not appear
to be related to language scores at age 7, whereas the
dichotomized OM history was weakly related to lan-
guage production at 7 years. The relation between WFP
and the dichotomized OM variable resulted in a 0.734
(standardized beta = –.229, p = .068) lower score for the
group with more OM. Adjusting for parents’ education,
this difference decreased to 0.627 (standardized beta =
.196), with the coefficient for parents’ education being

Table 2. Child characteristics relative to language scores.

27 months of age 7 years of age

a a a b
Variable N MLU MLUL Reynell WFPc CMc

Gender
Male 28 2.2 4.8 0.4 5.3 6.1
Female 27 2.3 4.9 0.3 4.7 5.4
Breastfed
No 36 2.2 4.4 0.3 4.8 5.5
Yes 28 2.4* 5.4 0.3 5.1 6.0
Parents’ educational level
Low 33 2.0* 4.2* 0.1* 4.5* 5.1*
High 32 2.5 5.4 0.6 5.4 6.4
Otitis media incidents
0–3.5 33 2.4 4.9 0.3 5.2 5.7
> 3.5 30 2.2 4.8 0.4 4.6 5.8
Hearing loss
< 1.1 dB 31 2.5 5.2 0.3 4.9 5.8
> 1.1 dB 32 2.1 4.5 0.5 4.9 5.7

*p < .05.
a
Lengths of sentences. bStandard normal scores. cStandardized scores (range: 1–10).

38 Journal of Speech, Language, and Hearing Research • Vol. 53 • 34–43 • February 2010

Downloaded from jslhr.asha.org on June 11, 2010

 Table 3. Pearson correlation coefficients between language scores and cognitive development at 27 months and language
scores at 7 years.

27 months of age 7 years of age

Variable dB HL OM MLU MLUL Reynell MDI WFP CM

dB HL 1 0.488a –0.247 –0.220 0.170 –.174 0.027 0.015

N 63 63 43 43 52 52 63 62
OM incidents 0.488a 1 –0.102 –0.056 0.057 –0.062 –0.103 0.088
N 63 64 44 44 53 53 64 63
MLU –0.247 –0.102 1 0.864a 0.450a 0.421b 0.510a 0.254
N 43 44 44 44 43 38 44 44
MLUL –0.220 –0.056 0.864a 1 0.357c 0.472a 0.478a 0.263
N 43 44 44 44 43 38 44 44
Reynell 0.170 0.057 0.450a 0.357c 1 0.431a 0.374b 0.468a
N 52 53 43 43 53 38 53 53
MDI –.174 –0.062 0.421b 0.472a 0.431a 1 0.149 0.087
N 52 53 38 45 45 53 53 53
WFP 0.027 –0.103 0.510a 0.478a 0.374b 0.149 1 0.373a
N 63 64 44 44 53 53 65 64
CM 0.015 0.088 0.254 0.263 0.468a 0.087 0.373a 1
N 62 63 44 44 53 53 64 64

a
p < .005. bp < .01. cp < .05.

0.862 (standardized beta = .269), and the explained
variance increased from 5.3% to 12.4%.
Language Scores at 27 Months
and 7 Years
In Table 3, it can be seen that MLU, MLUL, and
Reynell correlate weakly with WFP scores at 7 years:
.510 ( p < .001), .478 ( p = .001), and .374 ( p = .006), re-
spectively. Moreover, a positive correlation between lan-
guage comprehension skills at 27 months ( Reynell) and
at 7 years (CM) was found (.468, p < .001).
Children having parents with more education had
statistically significantly higher scores for all five
language tests (see Table 2). Parent education appeared
to be a strong moderator in the relation between com-
prehension at 27 months and at 7 years. The correlation
coefficient of the Reynell scores was then .424 ( p = .001)
with children with more educated parents having on aver-
age higher language production scores (r = .83, p = .087).
Parents’ education did not appear to have a moderating
effect on the relation between language production scores
at 27 months and 7 years.
Relation Between Language
Production and Comprehension
and Combined Effects
Language production and comprehension mea-
sured at 7 years are weakly but statistically significantly
correlated (r1 = .373, p < .002). Applying MANOVA to
these two language scores, it appears that the MLU and
Reynell at 27 months are both highly significant con-
tributors ( Pillai’s trace = 0.008 and 0.003, respectively),
with MLU contributing significantly only to the WFP score
(b2 = 1.09), t(42) = 3.235, p = .002, and the Reynell con-
tributing only significantly to the score on the CM test
(b = 1.52), t(42) = 3.635, p = .001. Upon including parents’
educational level as a covariate, the coefficient of MLU
for WFP shifted slightly downward (b = 0.98, p = .007),
and the coefficient of Reynell for CM was slightly lower
at 1.23 ( p = .005). Parents’ education appeared to be
related to WFP only with a higher score, on average (b =
1.3), t(1) = 1.242, p = .003. When considering the effect of
HL, the estimate for HL at –0.734 was almost significant
for WFP, t(1) = –1.738, p = .09, whereas HL appeared to
have no effect on CM. When parents’ education was in-
cluded in this model, the estimates for HL were not sig-
nificant, whereas a higher parental education resulted
in a 1.478 higher CM score, t(1) = 2.685, p = .011.
Parent-Reported Variables Between
0 and 7 Years and Language Skills
at School Age
For each age interval (0–2 years, 2–4 years, 4–6 years,
and 6–8 years), the effects of parent-reported hearing
1
r = correlation coefficient.
2
b = regression coefficient.

Zumach et al.: Early Otitis Media and Language Development 39

Downloaded from jslhr.asha.org on June 11, 2010

problems, ENT consults, insertion of tympanostomy tubes,
speech therapy, speech development between birth and
7 years, and hearing problems were examined with in-
dependent t tests. Children who had received tubes had,
on average, a 0.882 lower WFP score than the rest of the
studied group, t(63) = 2.149, p = .035. Children having
had an ENT consult between 3 and 7 years had, on
average, 0.758 lower WFP scores, t(63) =1.781, p = .080.
Children who had received speech therapy between age
3 and 7 years had 0.4 lower WFP scores, t(63) = 3.189,
p = .002. According to the parents’ report, 15% of the
children had HL between 0 and 2 years, 26% had HL
between 2 and 4 years, 20% had HL between 4 and
6 years, and 12% had HL from 6 years to the time of
evaluation. Fifty-four percent of the children had no HL
in any of the intervals, 30% had HL during one period,
8% had HL during two and three periods, and 2% of the
children had HL in all four time intervals. Parent-
reported HL in any of the four time intervals appeared to
have no effect on WFP and CM.
Discussion
The present study could not demonstrate a direct
effect of either OM or HL in the first 2 years of life on
either language comprehension or production skills at
school age. Negative effects of HL on language produc-
tion at 27 months, and of OM at 7 years, were only mar-
ginally significant.
The role of the actual HL at early life has often been
disregarded in investigating long-term effects on lan-
guage development. As periods of OM can occur without
auditory reduction, the underlying hearing loss might
be a better predictor of language problems. This is con-
firmed in the present study, as we found that the number
of OM incidents was not positively related to language
scores at early life, whereas HL was found to be related.
This is in line with the findings of other earlier studies
(e.g., Anteunis & Engel, 2000; Roberts, Hunter, et al.,
2004).
Current models of development, as outlined by Vernon-
Feagans et al. (2003) and Gravel (2003), place OM as just
one of many factors that influence child development. HL
as a mediating factor and possible moderating factors such
as breastfeeding and parent education, as well as gender
and cognitive development, were included to understand
the influence of other causal variables on development
and under which conditions OM may affect development.
We found that both breastfeeding and gender of the child
appeared to have a weak relation or no relation to language
scores at early life and school age, with neither variable
affecting language scores significantly. However, cog-
nitive development was a strong predictor for early lan-
guage scores. Likewise, there was a relation between
40 Journal of Speech, Language, and Hearing Research • Vol. 53 • 34–43 • February 2010
Downloaded from jslhr.asha.org on June 11, 2010
language skills and the level of parent’s education: It
appeared that children of higher educated parents were
less affected by HL in their language development. Thus,
even after adjusting for cognitive development, which is
a strong predictor, HL was still a statistically significant
predictor of MLU.
All children in the present study were healthy born.
Therefore, the influence of OM and the underlying HL
may be clearer to estimate than in children for whom
multiple risk factors obscure the contribution of HL to
later problems (Roberts et al., 2002).
Information about HL in the time interval between
2 and 7 years was obtained through a parent question-
naire. We found no associations between parent-reported
HL between birth and 7 years and language skills at age
7. When conducting separate analyses for the intervals
between 0–2, 2–4, 4–6 and 6–8 years, neither HL in any
of the intervals nor the cumulative number of all age in-
tervals with HL appeared to be related to language skills.
This argues against the contention that relatively recent
HL has an effect on language scores at age 7.
Although no effect of parent-reported HL between 0
and 2 years could be found, we did find that clinically
measured HL was significantly related to the toddlers’
language skills at 27 months. This suggests that parent-
reported HL might be a rough and /or unreliable mea-
sure. Anteunis and colleagues (1999) suggest that it is
difficult for parents of infants to recognize occurrences of
OM. The findings of the present study imply that par-
ents may even be unable to report OM-related hearing
loss when the child is no longer an infant. Although OM
presence was examined prospectively by an otolaryn-
gologist every 3 months, the parents only reported once
for each time interval of 2 years whether hearing loss
was present or not. Previous studies have also argued
that parent-reported information often lacks reliability
(Hartley & Moore, 2005; Roberts et al., 1991; Rockette,
2003). First, periods of HL due to OM can be missed by
parents and, therefore, are not reported or are reported
only when HL has been detected by an ENT doctor.
Second, periods with HL in early life might be remem-
bered less precisely than periods that have occurred
more recently. Third, parents of children who are fre-
quently affected by OM might be more alert to the po-
tential accompanying HL than parents of children who
have just had moderate middle-ear problems. This jus-
tifies apprehension in the use of parent-reported data—
in particular, those addressing HL. In contrast, it has
been shown that parents remember doctor consultations
or treatment more readily than they do episodes of ear
disease (Daly, Lindgren, & Gieblink, 1994), which ar-
gues for the reliability of the information about ENT
consults, tympanostomy tube insertion, and speech ther-
apy. In the present study, children who had one or more
ENT consults had significantly lower language production
scores at age 7. This may only reflect the fact that children
with language problems are more likely to be referred to an
ENT doctor to determine whether their language prob-
lems are caused by HL (Hartley & Moore, 2005).
Children who underwent tympanostomy tube inser-
tion (n = 9) had, on average, a significantly lower lan-
guage production score, but there was no control group
as there was in the studies by Rovers et al. (2000) and
Paradise and Feldman (2007). Both of these studies
were unable to find a difference in language compre-
hension and production skills between groups that had
either early or delayed tube insertion. In the present
study, it is likely that the children received tympanos-
tomy tubes to improve language problems and, there-
fore, they scored lower than the rest of the group.
Children who had had speech therapy still had sig-
nificantly lower WFP scores, although none of the chil-
dren could be classified as having a language problem at
age 7. This shows that after treatment, language prob-
lems were perhaps resolved but not to the level seen in
children who never had language problems.
Because we found a significant correlation between
HL and language scores at 27 months but not between
HL and language scores at age 7, we examined the
correlation between early and late language scores. If
early and late language scores would not be correlated,
one might question whether the variation in language
scores at age 7 was too low to reflect differences in lan-
guage skills and, consequently, would hide any correlation
between early-life HL and language at age 7. However,
early and late language scores were significantly related,
which may be an argument for the reliability of the ana-
lyzed data and against the apprehension that language
tests at age 7 may have been not sensitive enough.
It has been suggested that long-term consequences
of fluctuating HL in early life might not become appar-
ent in general language development but are restricted
to lower-order speech perception processes (Grievink,
Peters, Van Bon, & Schilder, 1993; Groenen, Crul, Maassen,
& Van Bon, 1996). Consequences of OM-related hear-
ing loss have been shown in various speech perception
tasks—for example in speech recognition in noise (Schilder,
Snik, Straatman, Van den Broek, 1994; Zumach et al.,
2009) and phonetic coding (Mody, Schwartz et al., 1999).
These findings, together with the findings of the pres-
ent study, support the assumption that long-term con-
sequences of early-life OM-related hearing loss may
not apply for general language comprehension or pro-
duction but may be restricted to speech perception
processes.
A disadvantage of multiple testing might be that it
inflates the possibility of a type I error and thus weakens
the conclusions that can be made. At the same time, the
relatively small sample size may be the reason for not
Downloaded from jslhr.asha.org on June 11, 2010
finding a more statistically significant effect. The pre-
sent data set did generate parameter values that were
comparable to the data from the original study group of
Anteunis and Engel (2000), indicating that the conclu-
sions found here with 65 children could be considered
representative for the original sample of 124 children.
However, it is still possible that insufficient power
may have caused the lack of a significant relation be-
tween HL and language. To increase power, the consid-
ered variables were dichotomized. Mean hearing loss
was dichotomized at the small value of 1.1 dB HL, on
average, over the first 2 years of life, which is because a
much larger sample is required to obtain enough chil-
dren in the sample with a clinically relevant HL. Con-
sidering an average cutoff level of 8 dB resulted in just
3 children with “hearing loss” and having a 0.3 lower
WFP score, which is 1.5 times the standard deviation. A
substantially larger sample than ours, also including a
number of children with a moderate HL, would render
sufficient power.
Our results found a positive correlation between
language skills at age 7 and various variables—such as
parental education, tympanostomy tube insertion, and
speech therapy—at an early age. The relation between
late language scores and the average HL in the first
2 years and the number of OM incidents was not present
or very weak. However, there might be a relation be-
tween moderate early-life OM-related hearing loss and
later language scores, which possibly could have been
demonstrated if a larger sample similar to our own had
been obtained.
In Anteunis and Engel (2000), factors such as cog-
nitive development, gender, age, parents’ educational
level, breastfeeding, passive smoking, dialect speaking,
and perinatal risk factors were found to play an impor-
tant role in language development. Considering these
factors in analyzing the present data, the effects were
less pronounced. To increase power, moderating factors
were considered in the dichotomized form. This revealed
significant positive correlations of parents’ education on
all language scores at 27 months and school age. In con-
trast, neither OM nor HL in the first 2 years of life was
significantly related to either language comprehension
or production skills at school age.
Conclusion
In the present study, we addressed the question of
whether early-life OM and its underlying fluctuating HL
has long-term consequences on language development.
This question was unable to be confirmed by the out-
comes of our study. Although a fluctuating HL between
birth and 2 years had negative consequences on
Zumach et al.: Early Otitis Media and Language Development 41
language scores at 27 months, neither early-life OM nor
HL appeared to affect language scores of school-aged
children significantly. To diminish the possibility that
the lack of an effect was due to the relatively small sam-
ple size, power was improved through dichomization of
the considered variables. However, studies with larger
sample sizes are needed to confirm the findings.
References
Anteunis, L. J. C., & Engel, J. A. M. (2000). Maastricht
Otitis Media With Effusion Study: A prospective longitudinal
study in infants from 0 to 2 years. Maastricht, the Nether-
lands: Datawyse.
Anteunis, L. J. C., Engel, J. A. M., Chenault, M. N., & Manni,
J. J. (2000). Language outcome at the age of 27 months
and the influence of early life otitis media with effusion. In
L. J. C. Anteunis & J. A. M. Engel ( Eds.), Maastricht Otitis
Media with Effusion Study: A prospective longitudinal study
in infants from 0 to 2 years ( pp. 157–177). Maastricht, the
Netherlands: Datawyse.
Anteunis, L. J. C., Engel, J. A. M., Chenault, M. N., et al.
(2000). Maturational changes of behavioural auditory thresh-
old levels in healthy-born and in high-risk-born infants. In
L. J. C. Anteunis & J. A. M. Engel ( Eds.), Maastricht Otitis
Media with Effusion Study—A prospective longitudinal study
in infants from 0 to 2 years ( pp. 141–154). Maastricht, the
Netherlands: Datawyse.
Anteunis, L. J. C., Engel, J. A. M., Hendriks, J. J. T., &
Manni, J. J. (1999). A longitudinal study of the validity of
parental reporting in the detection of otitis media and related
hearing impairment in infancy. Audiology, 38, 75–82.
Bailey, D. B., Bruer, J. T., Symons, F. J., & Lichtman, J. W.
(2001). Critical thinking about critical periods. Baltimore:
Brookes.
Bayley, N. (1969). Bayley Scales of Infant Development.
San Antonio, TX: Pearson.
Biesalski, P. (1971). Technique et jugement de l’audiometrie
du nourrisson: La surdité du premier âge [Technique and as-
sessment of audiometry in infants: Congenital deafness].
Besan0on, France: Campanov.
Bol, G., & Kuiken, F. (1989). Handleiding GRAMAT. Een
methode voor het diagnosticeren en kwalificeren van
taalontwikkelingsstoornissen [Manual GRAMAT: A method
to diagnose and quantify language development disorders].
Nijmegen, the Netherlands: Berkhout.
Bomers, A. J. A. M., & Mugge, A. M. (1982). Reynell
taalontwikkelingstest [Reynell Developmental Language
Scales]. Nijmegen, the Netherlands: Berkhout.
Cantekin, E. I. (1983). Algorithm for diagnosis of otitis media
with effusion. Annals of Otology, Rhinology and Laryngol-
ogy, 92, 6.
Central Office for Statistics. (1999). Standaard Onderwijs
Indeling van het Centraal Bureau voor de Statistiek [Stan-
dard Educational Classification of the Central Office for
Statistics]. Voorburg/Heerlen, the Netherlands: CBS.
Crystal, D., Fletcher, P., & Garmen, M. (1976). The gram-
matical analysis of language disorders: A procedure for
assessment and remediation. New York: Elsevier-North
Holland.
42 Journal of Speech, Language, and Hearing Research • Vol. 53 • 34–43 • February 2010
Downloaded from jslhr.asha.org on June 11, 2010
Daly, K. A., Lindgren, B., & Gieblink, G. S. (1994). Validity
of parental report of a child’s medical history in otitis media
research. American Journal of Epidemiology, 139, 1116–1121.
Engel, J. A. M., Anteunis, L. J. C., Chenault, M. N., & Marres,
E. (2000). Otoscopic findings in relation to tympanometry
during infancy. European Archives of Oto-Rhino-Laryngology,
257(7), 366–371.
Engel, J. A. M., Anteunis, L. J. C., Volovics, A., Hendriks,
J. J. T., & Marres, E. H. M. A. (1999). Prevalence rates of
otitis media with effusion from 0 to 2 years of age: Healthy-
born versus high-risk born infants. International Journal
of Pediatric Otorhinolaryngology, 47, 243–251.
Friel-Patti, S., & Finitzo, T. (1990). Language learning in a
prospective study of otitis media with effusion in the first
two years of life. Journal of Speech, Language, and Hearing
Research, 33, 188–194.
Gravel, J. S. (2003). OME and hearing loss. In R. Rosenfeld &
C. Bluestone ( Eds.), Evidence-based otitis media (2nd ed.,
pp. 342–359). Hamilton, Ontario, Canada: BC Decker.
Grievink, E. H., Peters, S. A., Van Bon, W. H., & Schilder,
A. G. (1993). The effects of early bilateral otitis media with
effusion on language ability: A prospective cohort study.
Journal of Speech and Hearing Research, 36, 1004–1012.
Groenen, P., Crul, T., Maassen, B., & Van Bon, W. (1996).
Perception of voicing cues by children with early otitis media
with and without language impairment. Journal of Speech,
Language, and Hearing Research, 39, 43–54.
Hartley, D. E. H., & Moore, D. R. (2005). Effects of otitis
media with effusion on auditory temporal resolution. Inter-
national Journal of Pediatric Otorhinolaryngology, 69,
757–769.
Holm, S. (1979). A simple sequentially rejective multiple test
procedure. Scandinavian Journal of Statistics, 6, 65–70.
Jerger, J. (1970). Clinical experience with impedance audi-
ometry. Archives of Otolaryngology–Head & Neck Surgery,
92, 311–324.
Knudsen, E. I. (2004). Sensitive periods in the development of
the brain and behavior. The Journal of Cognitive Neurosci-
ence, 16, 1412–1425.
van der Meulen, B. F., & Smrkovský, M. (1984). Bayley
Ontwikkelingsschalen. [Bayley Scales of Infant Development].
Groningen, the Netherlands: University of Groningen.
Mody, M., Schwartz, R. G., Gravel, J. S., & Ruben, R. J.
(1999). Speech opinion and verbal memory in children with
and without histories of otitis media. Journal of Speech,
Language, and Hearing Research, 42, 1069–1079.
Paradise, J. L., Dollaghan, C. A., Campbell, T. F., Feldman,
H. M., Bernard, B. S., Colborn, D. K., et al. (2000). Lan-
guage, speech sound production, and cognition in three-
year-old children in relation to otitis media in their first
three years of life. Pediatrics, 105, 1119–1130.
Paradise, J. L., & Feldman, H. M. (2007). Tympanostomy
tubes and developmental outcomes at 9 to 11 years of age.
The New England Journal of Medicine, 365, 248–261.
Reynell, J. (1977). The Reynell Developmental Language
Scales. London: Windsor nferNelson.
Roberts, J. E., Burchinal, M. R., Davis, B. P., & Collier,
A. M. (1991). Otitis media in early childhood and later lan-
guage. Journal of Speech, Language, and Hearing Research,
34, 1158–1168.
Roberts, J. E., Burchinal, M. R., & Zeisel, S. A. (2002).
Otitis media in early childhood in relation to children’s
school-age language and academic skills. Pediatrics, 110,
696–706.
Roberts, J. E., Hunter, L., Gravel, J., Rosenfeld, R.,
Berman, S., Haggard, M., et al. (2004). Otitis media,
hearing loss, and language learning: Controversies and
current research. Journal of Developmental and Behavioral
Pediatrics, 25, 110–122.
Roberts, J. E., Rosenfeld, R. M., & Zeisel, S. A. (2004).
Otitis media and speech and language: A meta-analysis
of prospective studies. Pediatrics, 113, 238–248.
Rockette, H. E. (2003). Design considerations for clinical
studies. In R. M. Rosenfeld & C. D. Bluestone ( Eds.), Evidence-
based otitis media (2nd ed., pp. 28–39). Hamilton, Ontario,
Canada: B. C. Decker.
Rovers, M., Schilder, A. G., Zielhuis, G. A., & Rosenfeld,
R. M. (2004). Otitis media. Lancet, 363, 463–475.
Rovers, M., Straatman, H., Ingels, K., van der Wilt, G.,
Van den Broek, P., & Zielhuis, G. A. (2000). The effect of
ventilation tubes on language development in infants with
otitis media with effusion: A randomized trial. Pediatrics,
106, E42.
Schaerlaekens, A., Zink, I., & van Ommeslaege, K. (1993).
Reynell taalontwikkelingstest. [Reynell Developmental Lan-
guage Scales]. Nijmegen, the Netherlands: Berkhout.
Schilder, A. G. M., Snik, A. F. M., Straatman, H., & Van
den Broek, P. (1994). The effect of otitis media with effusion
at preschool age on some aspects of auditory perception at
school age. Ear and Hearing, 15, 224–231.
Schilder, A. G. M., van Manen, J. G., Zielhuis, G. A.,
Grievink, E. H., Peters, S. A. F., & Van den Broek, P.
(1993). Long-term effects of otitis media with effusion on
language, reading, and spelling. Clinical Otolaryngology
and Allied Sciences, 18(3), 234–241.
Shriberg, L. D., Flipsen, P., Thielke, H., Kwiatowski, J.,
Kertoy, M. K., Katcher, M. L., et al. (2000). Risk for speech
disorder associated with early recurrent otitis media with
effusion: Two retrospective studies. Journal of Speech,
Language, and Hearing Research, 43, 79–99.
Suzuki, T., & Ogiba, Y. (1960). A technique of pure-tone
audiometry for children under three years of age: Condi-
tioned orientation reflex (COR) audiometry. Revue de
laryngologie–otologie–rhinologie, 81, 33–54.
Downloaded from jslhr.asha.org on June 11, 2010
Teele, D. W., Klein, J. O., Chase, C., Menyuk, P., & Rosner,
B. A. (1990). Otitis media in infancy and intellectual ability,
school achievement, speech, and language at age 7 years.
The Journal of Infectious Diseases, 162, 685–694.
Van Bon, W. H. J. (1982). Taaltests voor Kinderen [Language
Test for Children]. Lisse, the Netherlands: Swets & Zeitlinger.
Vernon-Feagans, L., Emanuel, D. C., & Blood, I. (1997).
The effect of otitis media and quality of daycare on children’s
language development. Journal of Applied Developmental
Psychology, 18, 395–409.
Vernon-Feagans, L., Miccio, A. W., & Yont, K. M. (2003).
Speech, language, pragmatics, and attention. In R. M.
Rosenfeld & C. D. Bluestone ( Eds.), Evidence-based otitis
media (2nd ed., pp. 360–382). Hamilton, Ontario, Canada:
B. C. Decker.
Werker, J. F., & Tees, R. C. (2005). Speech perception as a
window for understanding plasticity and commitment in
language systems of the brain. Developmental Psychobiol-
ogy, 46, 233–251.
Windsor, J., Scott, C. M., & Street, C. K. (2000). Verb and
noun morphology in the spoken and written language of
children with language learning disabilities. Journal of
Speech, Language, and Hearing Research, 43, 1322–1336.
Wright, P. F., Sell, S. H., McConnell, K. B., Sitton, A. B.,
Thompson, J., Vaughn, W. K., & Bess, F. H. (1988).
Impact of recurrent otitis media on middle-ear function,
hearing, and language. Journal of Pediatrics, 113, 581–587.
Zumach, A., Gerrits, E., Chenault, M. N., & Anteunis, L.
J. C. (2009). Otitis media and speech-in-noise recognition
in school-aged children. Audiology and Neurotology, 14,
121–129.
Received December 3, 2008
Revision received April 1, 2009
Accepted July 17, 2009
DOI: 10.1044/1092-4388(2009/08-0250)
Contact author: Anne Zumach, Maastricht University
Medical Center, Department of Otorhinolaryngology and
Head & Neck Surgery, P. Debyelaan 25, Maastricht
6229HX, the Netherlands. E-mail: anne.zumach@mumc.nl.
Zumach et al.: Early Otitis Media and Language Development 43
Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.