The interplay between prenatal speech input and brain maturation in early language
development

By listening to speech, infants acquire perceptual skills needed to acquire
their native language (L1). The development of these skills starts at week 20
of pregnancy, when fetuses begin perceiving speech (Gervain, 2018). The
maternal abdomen filters acoustic information and only allows vowel sounds
(e.g. /ei/ and /i/ in *baby*) and variations in pitch, duration, and intensity
(*prosodic cues* for short) to reach the fetus. This prenatal experience
promotes the development of preferences for, and specific neural responses to
the mother*s voice and L1 as compared to unfamiliar voices and rhythmically
different languages (Beauchemin et al., 2011; Gervain, 2015, 2018; Kisilevsky,
Gilmour, Stutzman, Hains, & Brown, 2012; Nazzi, Floccia, & Bertoncini, 1998;
Vouloumanos & Werker, 2007). Neonates show language-specific neural responses
to native speech but not to non-speech, indicating speech-specific functional
neural organization (Gervain, 2018; Sato et al., 2012; Vannasing et al., 2016).
Thus, prenatal experience with speech creates memories of the mother*s voice
and L1, and promotes speech-specific neural organization needed for language
acquisition (Kisilevsky, 2016).

Although the role of prenatal speech input in neonatal speech perception is
becoming clearer, little is known about the consequences of atypical prenatal
experiences with speech on language development. Conditions such as preterm
birth (i.e. before week 37 of pregnancy), growth restriction, and maternal
gestational diabetes (GDM) are associated with altered prenatal responsiveness
to the mother*s voice and with deficits in long-term language development (
Kisilevsky, 2016). For example, compared to typically developing (TD) infants,
growth-restricted fetuses respond less strongly to the mother*s voice, do not
prefer their mother*s voice at birth, and display language deficits by 15
months (Kisilevsky, 2016). Preterms, who have had limited exposure to speech in
the womb, perform poorly in linguistic rhythm and stress pattern discrimination
(i.e. prosodic perception) postnatally, which may be causally related to later
language delays (Barre et al., 2011; Gonzalez-Gomez & Nazzi, 2012; Putnick et
al., 2017; van Noort-van der Spek et al., 2012). These findings suggest that
atypical prenatal experiences with speech affect early postnatal prosodic
processing, which in turn has negative consequences for language development.

An underlying cause for atypical prenatal experience with speech may be
protracted development of the brain (Kisilevsky, 2016). Brain immaturity
results in less efficient processing and storing of information, ultimately
resulting in reduced impact of prenatal experience with speech despite exposure
similar to that of unaffected fetuses. However, how brain immaturity reduces
the impact of prenatal speech experience on language development is unclear.
For example, brain immaturity may affect prenatal experience with speech
through altering the sensorineural threshold for prenatal input (i.e. fetal
perception), or through changes in the auditory recognition memory (Kisilevsky,
2016), which affects language development differently. A clear understanding of
how prenatal experience with speech interacts with brain immaturity is needed
to determine the role of prenatal speech input in language development.

In order to gain this understanding, the linguistic capacities of prematurely
born children may be studied. In Europe, between 8 and 9%, and worldwide
between 9 and 13% of all births occur prematurely (Blencowe et al., 2012). By
definition, the central nervous system is less developed at birth in preterm
neonates, compared to fullterm neonates. Previous research on the role of
prenatal speech input in language development has shown that this neural
immaturity can lead to language delay at a later stage (Barre et al., 2011;
Gonzalez-Gomez & Nazzi, 2012; Putnick et al., 2017; van Noort-van der Spek et
al., 2012). Also treatment in neonatal intensive care units (NICUs), which
exposes neonates to other adverse conditions (e.g. abnormal postnatal
stimulation and stress) may contribute to this delay (Vohr, 2014). However,
linguistic skills in prematurely born children have been investigated mostly at
a later age, often after the first year of life. Very few studies have focused
on language processing in preterm infants at birth. In order to understand the
effects of neural underdevelopment on prenatal speech perception, it is
necessary to examine the abilities of preterm neonates to recognize and process
linguistic input at birth. Because rapid development of the neural auditory
system takes place within the third trimester of pregnancy (Moore & Linthicum,
2007), these abilities may vary between preterm babies born at different
gestational ages. Therefore, a comparison between the language processing
capacities of these different groups can be made in order to gain insight into
prenatal linguistic development. In addition, the interaction between prenatal
language experience and brain immaturity may be studied by investigating
linguistic processing in full-term infants from mothers with GDM (GDM infants).
GDM occurs in 9-25% of pregnancies and has neurocognitive implications for the
children. GDM fetuses react less strongly to the mother*s voice than low-risk
controls, and habituate less to vibro-acoustic stimulation (Kisilevsky et al.,
2012), which indicates sensorineurological immaturity. Compared to controls,
GDM infants score lower on cognitive and language tests (Adane et al., 2016;
Battin et al., 2018), perform poorly in verbal communication, and are twice as
likely to have a language impairment in childhood (Dionne et al., 2008). Due to
the high prevalence of GDM and high risk for language impairments, many
children may experience language deficits, which may have significant negative
consequences for their emotional, academic, and social functioning later in
life. Early language interventions are thus strongly called for. However,
little is known about their early language development.

The present research project aims to determine how neural immaturity affect
early language development by examining specific auditory, speech, and language
processing skills in GDM neonates and preterm neonates. Specifically, general
auditory and speech-specific perceptual skills will be compared between term
GDM infants and matched controls, tested within 72 hours after birth.
Furthermore, language perception will be compared between preterm neonates born
at different gestational ages, tested within one week after birth.

GDM experiments (within 72 hours after birth):
One reason for GDM fetuses being unable to recognize their mother*s voice may
be that GDM delays fetal neural maturation, which may affect their ability to
process auditory input. We hypothesize that if the brain is unable to properly
process prenatal input, it is unable to use that information to prepare the
language network before birth. This is a critical, fundamental first step in
language acquisition. In this experiment, we aim to find out what auditory
input babies are able to process at birth, and how that is affected by prenatal
neural development and by prenatal input to speech.

To get a more detailed picture on how GDM babies process auditory stimuli, we
will measure neural activity as a result of auditory stimuli using
multi-channel EEG in event-related potentials (ERP) paradigms. This experiment
will take place within 72 hrs after birth. This was chosen to minimize the
amount of time the baby is exposed to language and speech postnatally, which
may impact ERPs. Due to practical and logistical reasons, this experiment will
take place within 72 hours after birth while mother and child are still in the
hospital. To examine the auditory neural maturation, we will use EEG in two
steps in one session.

First, we target domain-general auditory neural processing. Here, we test
whether GDM delays neural maturation of the auditory system. In particular, we
will examine whether the neonates can process quick successive non-speech
auditory stimuli (i.e. rapid auditory processing). The development of this
ability develops prenatally and does not require exposure to speech. Although
it develops independently of speech input, it is an essential skill for the
discriminating and learning of speech sounds, which starts as soon as a baby is
born. If a baby*s hearing is deficient in temporal resolution, this will
negatively affect language outcome. For example, rapid auditory processing not
only predicts 39-41% of the variance in later language outcome but it has also
shown to predict possible future language impairments such as developmental
language disorder (Benasich et al., 2002; Molfese & Molfese, 1997).

Second, we examine the outcome of prenatal development of speech-specific
auditory processing at birth. Fetal experience with speech promotes the
development of the basic architecture of the language neural network, which can
be tested when the babies are born (Gervain, 2015, 2018; Sato et al., 2012). To
test whether this neural network has started to develop, we will test whether
the neonatal brain distinguishes between speech and non-speech. The distinction
between domain-general auditory processing and speech-specific auditory
processing allows for a better discrimination between the effects of GDM on
general neural immaturity and the specific effects on the prenatal experience
with speech.Preterm experiment (within 1 week after birth):
If brain immaturity affects the ability to perceive and analyze prenatal
linguistic input in GDM fetuses, the question arises how healthy fetuses are
able to properly process the same acoustic information without any linguistic
experience, while their neural auditory systems are still developing.
Specifically, since it is mostly prosodic information that reaches the
intra-uterine environment of the fetus, it is unclear how and at what
developmental stage this type of language-specific information can be
recognized and learned. We hypothesize that fetuses have innate knowledge of
so-called *prosodic boundaries*, i.e. prosodically marked endings of phrases in
speech streams. In other words, any developing fetus is equipped with a priori
knowledge that enables it to recognize and process the type of linguistic
information it can hear within the maternal womb. If this is the case, then
preterm newborns who are born between 28 and 33 weeks gestational age should
have similar abilities in processing prosodic information at birth, despite the
neural immaturity associated with preterm birth. At this stage in prenatal
development, fetal hearing is established but language experience is at a
minimum. If preterm newborns are nevertheless able to process prosodic
information in speech at birth, it would suggest that this ability is driven by
innate linguistic mechanisms.

In order to test this hypothesis, neural responses to the presentation of
linguistic stimuli will be measured in two groups of healthy, prematurely born
infants, who are born at different gestational ages, using multi-channel EEG in
an ERP paradigm. Perceiving and processing prosodic boundaries is associated
with a specific ERP response, called the *Closure Positive Shift* (CPS)
(Steinhauer et al., 1999). These neural responses will be measured in one group
of infants born at 28, 29 or 30 weeks of gestation (in this study: *early
preterm neonates*), and in one group of infants born at 31, 32 or 33 weeks of
gestation (*late preterm neonates*). These different age groups are included
since it is not clear at what age CPS may be appear as a response to processing
prosodic boundaries. At 28-30 weeks, the infants will have very little
linguistic experience, but perhaps they may possess underdeveloped neural
mechanisms for demonstrating higher cognitive ERP components. At 31-33 weeks,
these neural mechanisms will be more developed. However, at this age, the
prosodic abilities of the infants may be influenced by their increased
experience with the ambient language, which would make it difficult to assess
the role of innate knowledge. To keep a balance between early neural
development on the one hand and linguistic experience on the other hand, these
two age groups are included. Testing will take place within a week after birth,
in order to minimize postnatal linguistic influence. Responses matching the CPS
component will be taken as evidence for innate prosodic processing principles,
which may guide prenatal linguistic development in typically developing
fetuses, but which may fail in GDM fetuses through impaired handling of
auditory stimuli.

Our main question is: How does GDM brain immaturity affect skills that are
critical for later language development?
To answer this question, we will assess two skills in GDM infants and compare
them to term controls in two independent (separate) experiments (Experiment 1
and 2, respectively), in correspondence with our first two primary objectives.

In addition, prenatal development of preterm neonates will be studied in one
independent experiment (Experiment 3) by assessing linguistic skills in two
groups of prematurely born neonates, in correspondence with our third primary
objective.

Primary Objectives:
1. to determine the differences between GDM and TD neonates in the general
maturation of auditory perception, as measured with multi-channel EEG in an ERP
paradigm.
2. to determine the differences between GDM and TD neonates in the maturation
of speech perception, as measured with multi-channel EEG in an ERP paradigm.
3. to determine the differences in the maturation of speech perception between
early preterm neonates, born at a gestational age of 28-30 weeks and late
preterm neonates, born at a gestational age of 31-33 weeks, as measured with
multi-channel EEG in an ERP paradigm.

This study is an observational cross-sectional study with multiple
(independent) outcomes.

The mother will fill in questionnaires and the child will partake in behavioral
and neuroimaging experiments.

Gestational diabetes has long-term adverse effects on the child*s cognitive
functioning, e.g. on their language abilities. We hypothesize that this is
related to prenatal neural immaturity and their experience with
speech/language. Considering that the first 1001 days of a child*s life from
conception is the most critical for their neurodevelopment due to heightened
neural plasticity, neural immaturity may change the fetal sensitivity to
auditory stimuli, which then affects to what extent the child learns from this
auditory stimulation. To have a better understanding what exactly is affected,
more knowledge is needed about their prenatal, perinatal, and postnatal
cognition and auditory maturation. The second and third objectives are related
to the increase of our understanding of how the prenatal environment affects
early language development in general. There is an increasing amount of
evidence that fetuses learn from the sounds they hear prenatally, but it is yet
unknown whether these experiences with sounds are required for early language
development.

There are no known risks associated with any of the experiments in this study.
In all the experiments, the child will be presented with auditory stimuli at a
volume that is safe for their hearing and thus will cause no harm. EEG is a
non-invasive method to measure brain activity. EEG requires the use of a
special hat which carries the sensors to measure brain activity. This hat is
especially made for neonates and is thoroughly tested for their sensitive heads
for this purpose. The sensors are made to only receive electrical information
that are produced internally by the brain. EEG therefore does not send off
electricity itself and is safe for usage. All researchers involved in this
study will be trained prior to their first contact with a study participant.

Although there are no risks, there is some burden that the parents/guardians
and child may experience. First, one or both parent(s)/guardian(s) will be
present during each of the experimental sessions, which they may consider to be
boring. They will also have to answer questions, e.g. on their socioeconomic
status, which they may be uncomfortable with. However, we will make sure that
it is clear to the parents/guardians that all provided information will be
confidential and protected. Second, the child will be tested twice in one
session in the fullterm experiments and once in the preterm experiment. They
may find this bothersome. However, the experiments have been designed in such a
way that the baby can be asleep; an activity that a newborn already does most
of the day. This allows for more comfort during the experiment.

We have also tried to minimize the burden that falls upon parents/guardians and
child by planning the testing moments when the mother and child are already at
the hospital or by testing at home. Therefore, they do not have to spend extra
time to come to the lab.

In light of the vast amount of knowledge we will gather from this study, we
believe that the relatively low burden that the parents/guardians and the child
face are negligible.