Bite-block listening disrupts Japanese children’s English r/l but not native s/sh

Japanese children aged 8 to 12 who listened to English words with a silicone bite-block held between their teeth went on to produce both a native and a non-native speech contrast less distinctly, even though the block was removed before they spoke. That is the central finding of a new study published in Frontiers in Human Neuroscience, in which Fei Ting Woon, Amanda Seidl and F. Sayako Earle at the University of Delaware recorded 39 typically developing Japanese children producing English-like words under normal listening conditions and after brief articulatory perturbation. The work offers a rare within-child comparison of how well-established native sound categories and still-emerging second-language categories respond to the same temporary disruption of the mouth’s usual sensorimotor routine.

The team chose two contrasts that sit at very different points on each child’s developmental timeline. The native pair was /s/ versus /ʃ/, sounds that exist as separate phonemes in Japanese and are typically acquired between 4 and 5 years of age. The non-native pair was English /l/ versus /ɹ/, a contrast that Japanese lacks because the language has a single liquid phoneme realized as an alveolar tap. English-learning Japanese children famously struggle with this distinction, and prior work by Idemaru and Holt showed that the difficulty persists even after years of second-language exposure. The sample spanned an interesting policy transition in Japan: since 2020, English has been a required academic subject in grades 5 and 6 with 70 hours of instruction per year, while grades 3 and 4 receive only 35 hours per year as part of Foreign Language Activities.

Each child completed three tasks in a fixed order: a production task without any perturbation, a perception task on English R and L, and then a second production task in which the bite-block was worn during listening. In both production tasks, children heard a recorded model produced by an adult female native English speaker and repeated the word immediately afterward. The target words, such as soup, shoop, loop and roop, were constructed following English phonotactics and were considered novel for the participants, but because the task was repetition after an auditory model, children did not need lexical knowledge of the wordforms to perform it. Each target was elicited three consecutive times, yielding six recordings per speech sound per condition and 48 words per child overall. The bite-block itself was a food-grade silicone rectangle, grape- or orange-scented, held between the incisors. Participants removed it immediately after hearing the model and then produced the word, so the perturbation was confined to the listening phase.

Acoustic analysis focused on two measures. For the fricatives, the researchers extracted Center of Gravity, the spectral frequency measure that distinguishes the high-frequency noise of /s/ from the lower-frequency noise of /ʃ/, using the middle 40 milliseconds of each annotated segment. For the liquids, they used the difference between the third and second formants, F3 minus F2, a reliable correlate of the English /ɹ/ versus /l/ contrast, with /ɹ/ characterized by a low F3 that often approximates F2 and /l/ by a higher F3. The adult model recordings were confirmed to be acoustically distinct on both measures before testing began. Statistical models were linear mixed-effects models with by-participant random intercepts, and the analyses were preregistered.

Under normal listening conditions, children differentiated both contrasts. For the native pair, /s/ tokens showed a mean Center of Gravity of 7,969 Hz against 4,890 Hz for /ʃ/, a highly significant difference with the model accounting for substantial variance. Crucially, this native contrast did not vary with age: neither a main effect of age nor a Sound by Age interaction reached significance, and simple-effects checks confirmed the contrast was significant at all age levels tested. The authors interpret this as evidence that the /s/–/ʃ/ distinction remains stable from age 8 through age 12. For the non-native pair, the picture was more nuanced. Overall, /l/ tokens showed higher F3–F2 values than /ɹ/ tokens, a significant but modest effect. The continuous Sound by Age interaction did not reach statistical significance. However, simple-effects analysis revealed that the /l/–/ɹ/ contrast was not significant at younger ages, became significant at the mean age of about 10 years 3 months, and was significant and numerically larger at older ages. The authors describe this as a modest, age-related sharpening of the non-native contrast, consistent with a pattern in which older children differentiate /ɹ/ and /l/ more robustly than younger children at baseline.

The bite-block manipulation reduced acoustic distinctness for both contrasts at the group level. For the native pair, the mean Center of Gravity difference between /s/ and /ʃ/ fell from 2,947 Hz in the no-block condition to 2,460 Hz in the block condition, a reduction of roughly 487 Hz that was statistically significant. For the non-native pair, the formant contrast difference between /l/ and /ɹ/ was positive in the no-block condition at 147 Hz on average and near zero in the block condition, and the condition effect was significant. In both cases there was substantial individual variability: 24 of 39 children showed a reduction in the native contrast with the block while 15 showed an increase, and 25 of 39 showed a reduction in the non-native contrast while 14 showed an increase.

The developmental profiles of susceptibility diverged in a way the authors highlight as a key contribution. For the native contrast, the blocking effect was stable across the age range: neither the main effect of age nor the Condition by Age interaction was significant, indicating that susceptibility to perturbation during listening was comparable across 8 to 12 years and did not decrease with greater articulatory experience. For the non-native contrast, the Condition by Age interaction again did not reach significance, but simple-effects analysis showed the blocking effect was not significant at younger ages, was significant at mean age, and was significant and numerically larger at older ages, with the effect size rising from a small value at younger ages to a large one at older ages. The authors argue this descriptive pattern suggests older children showed greater susceptibility to the block than the youngest group, though they caution that the non-significant interaction means the increase cannot be treated as established in this sample.

Perception also mattered. Children completed an 80-trial identification task in which they heard English words starting with either R or L and pressed the corresponding button. Higher perceptual accuracy predicted larger baseline acoustic distinctness in production of the non-native contrast, and this relationship did not vary with age. Interestingly, younger children actually showed numerically higher mean perceptual accuracy than older children, though the difference was not statistically significant. The authors draw on the PRIMIR framework to suggest that younger children may perceive the non-native contrast through general perceptual skills without yet having well-specified phonemic targets, whereas older children, with more second-language experience, may have begun to establish higher-level representations that support more distinct productions.

Exploratory analyses added a further layer. Within the older age group, defined by a median split at 123 months, baseline production distinctness significantly predicted the magnitude of the blocking effect, and the effect of perceptual accuracy on the blocking effect strengthened as baseline distinctness increased. A Johnson-Neyman analysis indicated that perceptual accuracy predicted the blocking effect only when baseline distinctness exceeded a threshold of about −49 Hz, meaning the effect emerged among older children whose productions showed the English-speaker-like ordering of formants, or came close to it, rather than a reversed one. The authors take this as evidence that the lasting blocking effect depends on the degree of specification of children’s articulatory representations: contrasts that are better learned and more specified have more to lose when somatosensory feedback is unexpectedly impeded. They also note that the exploratory correlation between native and non-native blocking effects was weak and non-significant, so the data cannot adjudicate whether a domain-general susceptibility operates across both contrast types.

The authors are careful about limitations. The fixed task order means condition was not counterbalanced, though comparisons of first and third production attempts within the no-block task found no evidence of systematic practice or fatigue effects. The study did not collect direct measures of English exposure, did not assess age of first exposure, and did not test native English-speaking children, so the age-related pattern for the non-native contrast could reflect the articulatory challenge of the specific sounds, their non-native status, or cumulative formal instruction, and the data cannot disentangle these possibilities. The exploratory subgroup analyses rest on only 18 older children. The bite-block was present only during listening, so the authors cannot isolate whether the effect originated in perceptual perturbation that shifted motor planning or in more general sensorimotor adaptation to the block itself. Still, the study is, to the authors’ knowledge, among the first to chart children’s production of both a native and a non-native contrast within the same speakers using the same perturbation paradigm, and it suggests that perceptual sensitivity to non-native contrasts can exist without stable articulatory representations, while speakers who have more stable representations show corresponding vulnerability when feedback is disrupted.

Subject of Research: How temporary bite-block perturbation during listening affects Japanese children's production of native and non-native speech contrasts

Article Title: Bite-block effects on native s/ʃ and English r/l in child speakers of a language with/without this contrast

Article References: Woon, F. T., Seidl, A., & Earle, F. S. (2026). Bite-block effects on native s/ʃ and English r/l in child speakers of a language with/without this contrast. Frontiers in Human Neuroscience, 20, Article 1949253. https://doi.org/10.3389/fnhum.2026.1949253

Image Credits: AI Generated

DOI: 10.3389/fnhum.2026.1949253

Keywords: speech production, speech perception, bite-block, Japanese children, non-native contrast, phonological development, acoustic analysis, sensorimotor, L2 acquisition, articulatory perturbation, English /r/-/l/, developmental psychology