A landmark study published in Nature in 2025 revealed how the human brain processes speech in both native and non-native languages, uncovering a clear distinction between universal mechanisms and those shaped by linguistic experience. Researchers recorded high-density electrocorticography activity in adult Spanish, English, and Mandarin speakers as they listened to sentences in both their native language and a foreign language they could not understand. Overall, the researchers examined 883 speech-responsive electrodes from 20 monolingual respondents, and an additional group of Spanish English bilingual participants was analyzed.
The results indicate that in the superior temporal gyrus, the identical groups of neurons react to speech irrespective of language familiarity. An example electrode was shown to have a Pearson correlation of r (250) = 0.986 and p < 0.001 between native and foreign speech responses. Across all the participants, both languages had significant responses in 79.84% of the electrodes. Tuning of acoustic and phonetic features like vowel formants, classes, and amplitude envelope of consonants was highly preserved across languages. These features were strongly correlated with each other, with r (1992) = 0.86 and p < 0.001, and 93.37% of the electrodes demonstrated correspondence above chance. The above findings suggest that the superior temporal gyrus encodes common acoustic and phonetic representations shared across all spoken languages.
Although the brain may process simple low-level sound features across all languages in a similar manner, there seems to be a very strong contrast in higher levels of linguistic processing that strongly depends on language experience. The increased encoding in word boundaries, word frequency, word length, and phoneme surprisal showed enhanced responses when participants listened to their native language. Significant differences in unique variance for word boundaries, t(243) = 4.00, with P <0.001; word boundaries, t(253) = 5.18, with P <0.001; word frequency, t(123) = 3.65, with P <0.001; word length, t(29) = 3.58, with P = 0.0012; and phoneme surprisal. These effects were localized on the bilateral mid superior temporal gyrus, 34% of speech-responsive electrodes coded word-level speech features in the native speech, relative to 12% coded speech features in the foreign language listening.
Language experience also played a major role in enhancing the ability to recognize the starting and ending points of words, which is critical to speech comprehension. The neural decoders that had been trained using signals of electrocorticography could classify a word-syllable boundary as compared to a syllable-syllable boundary with an area under the curve of 0.77 ± 0.052 during the native speech condition and only 0.66 ± 0.053 during the foreign condition. Advantages in decoding native language words were higher when word boundaries were acoustically ambiguous. One-way ANOVA analysis of variance showed that decoding accuracy increased significantly as ambiguity rose, with values of F (2, 672) = 12.88 and P < 0.001 in Spanish, and F (2, 672) = 24.91 and P < 0.001 in English.
In bilingual participants, the encoding of word-level and phoneme surprisal features did not differ between their two familiar languages. Unique variance values showed a significant correlation between languages (Spearman r(45) = 0.31, p = 0.03). Across multilingual participants with diverse language backgrounds (Arabic, Russian, Korean, Mandarin, Portuguese, Spanish), English proficiency was positively associated with the accuracy of decoding English word boundaries, confirmed using a linear mixed-effects model (t(253) = 4.58, p < 0.001).
Overall, this study suggests that the superior temporal gyrus integrates universal acoustic and phonetic operations with experience-related encoding of phoneme and word sequences. This interactive process enables fluent perception of foreign languages and efficient segmentation of native languages into discrete units.
References: Bhaya-Grossman I, Leonard MK, Zhang Y, et al. Shared and language-specific phonological processing in the human temporal lobe. Nature. 2025. doi:10.1038/s41586-025-09748-8
