Speech learning in infancy

Speech learning before birth

It is now well known that babies learn about their native language well before they are even born. Scientists generally agree that it is language melody and rhythm what is learnt prenatally. We reasoned that it is physiologically and acoustically equally possible that unborn babies may learn about the individual speech sounds (i.e. phonemes) of their native language and went on to test that prediction.

In this project, we examine whether at birth, babies show signs of prenatal learning of native-language vowels (namely, the Czech a, e, and é) We have tested over a hundred newborns with electroencephalography (EEG), and recorded how sleeping, one to two days old babies process vowels and comparable nonspeech sounds.

We have found that the babies processed the vowel, speech stimuli more robustly than they processed the nonspeech stimuli. Moreover, for the vowels, their brains showed higher level (perhaps abstract) processing asymmetries which very much resemble those previously found in adults. Our findings are explainable in terms of prenatal category formation. However, alternative explanations in terms of speech-specific acoustic confounds in our stimulus material, or even some degree of innateness for speech processing, cannot be ruled out. We thus continue to investigate prenatal speech learning through computational simulations (project below) and, hopefully, future cross-linguistic comparisons.

Our manuscript reporting this newborn project is currently under review. You can have a look at a poster with interim analyses that we presented at the Boston University Conference on Language Development.

This project is done in close collaboration with Jan Kremláček (Faculty of Medicine in Hradec Králové). We’re grateful to Gábor Stefanics (University of Zurich) who shared his expertise on newborn EEG testing.


Tracing speech sound development accross the first year of life

The acquisition of the sound system of Czech has not yet been described. Such a gap in knowledge needs to be filled, because Czech is a rather unique language in that it contrasts speech sounds not only by their quality (making distinctions such as parník vs. perník, ‘steamboat’ vs. ‘gingerbread’) as all of the world’s languages do, but also by their length (making distinctions such as ráda vs. rada, ‘like 3rd sg.’ vs. ‘council’). In order to properly understand how humans learn their language, we need to know how a (Czech or Finnish but not Spanish or French) developing baby comes to realize than both types of contrasts are important in her language.

This project traces the trajectory of vowel quality and vowel length development across the infants’ first year of life. We test babies at 4, 6, 8, and 10 months of age, aiming to find out how the development of perceptual sensitivity to both types of contrasts unfolds.

This conference abstract reports on our findings to date.

This project is carried out together with Václav Jonáš Podlipský and Šárka Šimáčková from the Faculty of Arts at Palacký University in Olomouc.


Infants’ and children’s perception of accents

This project entails two experiments.

In the first one, we investigate whether infants learn speech sounds more effectively from a native-accented talker than from a non-native one. Babies between 4 and 10 months of age are played utterances spoken by talkers with typical, standard Czech accent and by talkers with atypical, foreignized accent. Preliminary data (reported in this poster) suggest that the youngest, 4-month old, babies prefer listening to the typical over the foreignized accent while older infants do not seem to have such preferences. The four-month olds are now being tested on whether they not only preferentially listen to native-accented talkers, but also if they more efficiently learn new speech sounds from them.
This experiment is done jointly with Václav Jonáš Podlipský and Šárka Šimáčková (Palacký University).

In the second experiment, we test whether pre-school children have preferences for peers speaking their local over a regional dialect, and whether any such preference is modulated by the child’s own dialectal experience? Infants, older children and adults are known to favor native-accented over foreign-accented speakers: people prefer playing with, or trust more, a person who has a native accent than with someone who has a foreign accent. In this project we investigate whether similar preferences exist in children’s perception of regional accents of their language. Pre-school children from two regions, central Bohemia and Silesia, where distinct varieties of Czech are spoken, will be tested on their social peer preferences. Children will view photos of other kids and will hear utterances spoken in their own and in the other dialect. They will be asked to indicate which of two simultaneously presented kids they would like to play with. The project will show whether regional accent has similar effects as those previously reported for foreign accents, and if that effect is modulated by a child’s (multi-)dialectal exposure.
This experiment is done as a student internship project within the Open Science programme. 


Projects