Infants as young as four months can abstract phonological features, a skill previously believed to develop significantly later, according to a new study published in the journal Developmental Science.
The traditional understanding of language development suggests that babies begin by responding universally to speech sounds and only later become attuned to the specific characteristics of their native language. This process, known as perceptual attunement, typically starts between six and ten months of age. Until now, phonological abstraction, or the ability to identify and generalise language features, was thought to occur only after this perceptual tuning process.
But researchers at Western Sydney University’s MARCS Institute for Brain, Behaviour and Development have challenged this long-held view. Their new findings suggest that infants between four and six months old already possess the ability to abstract phonological features. This marks the first direct experimental evidence of this ability at such a young age.
“For a long time, it was assumed that infants first become attuned to the sounds of their native language before they develop the ability to abstract phonological features,” said Eylem Altuntas, PhD, a researcher at Western Sydney University. “However, studies on internationally adopted adults suggested that even before six months of age, infants may already be forming phonological abstractions by generalising sound features rather than simply memorising individual sounds. Since this had not been directly tested in infants, we set out to investigate whether four- to six-month-olds can abstract the distinction between two mini-languages in which consonants differ in terms of place of articulation.”
The study involved exposing infants to artificial mini-languages made up of multisyllabic nonsense words. One set of these languages contained consonants articulated primarily with the lips, such as “b”, “v”, and “w”. The other set consisted of consonants formed with the tongue tip, like “d”, “z”, and “l”. These languages were paired visually with cartoon sea animals, with labial consonants associated with jellyfish images and coronal consonants associated with crabs.
After training with audio presentations of these mini-languages, the infants were shown silent videos of a speaker articulating consonants from each language. Even without hearing any sound, the infants spent significantly more time looking at the visual images correctly matched to the sounds they had previously learned. Their attention to the matching pairings demonstrated they had abstracted the differences in consonant articulation visually, relying solely on previously learned audio information.
“Our findings provide the first direct evidence that infants can abstract phonological features as early as four months, well before perceptual attunement occurs,” Altuntas explained. “During testing, they saw a silent video of a speaker producing labial and coronal words and looked significantly longer when the speech matched the image they had learned. Our findings suggest that infants as young as four months can abstract phonological features by detecting shared articulatory information across consonants. Crucially, they maintained this abstraction across a shift from audio training to video testing, demonstrating amodal phonological learning.”
These outcomes overturn traditional models such as the Word Recognition and Phonetic Structure Acquisition model and the PRIMIR framework. Both models propose infants must become perceptually attuned to their native language and familiar with numerous words before phonological feature abstraction can occur. Instead, this research indicates infants possess a sophisticated ability to identify speech features without needing extensive exposure or vocabulary, and they can transfer this knowledge across sensory modalities.
This discovery could have far-reaching implications for early education and speech therapy. Understanding that young infants can abstract phonological features at such an early age suggests potential benefits from introducing language interventions at earlier stages than currently practised. It also highlights the sophistication of infant cognition, suggesting that educational approaches might benefit from tapping into infants’ natural abilities sooner.
The MARCS Institute’s findings align with previous research involving individuals who were adopted into environments with different languages at very young ages. Such studies indicated these individuals retained some early-acquired phonological features from their birth languages, despite not using them subsequently. This latest study provides direct infant-based evidence supporting that early abstraction is not only possible but likely common among infants.
“Building on this work, we aim to investigate whether infants can abstract other phonological contrasts, such as voicing, at similarly young ages,” Altuntas added. “We are also interested in how early bilingual exposure might shape these abilities. Does growing up with two languages enhance or alter phonological abstraction? Additionally, we will examine how this skill develops at later ages, particularly after perceptual attunement.”
