Abstract
This study presents the first investigation into the neural correlates of lexically-guided perceptual learning—the cognitive process by which listeners retune phoneme categories based on vocabulary context. Using Event-Related Potentials (ERPs), the researchers examined how Dutch non-native listeners of English adapt to ambiguous speech sounds. While only half of the participants demonstrated a significant behavioral shift in categorization, nearly all participants exhibited a significant reduction in the P3b ERP component. These findings suggest that the brain may undergo neural adaptation even when explicit behavioral changes are not observed, indicating a potential decoupling between internal model updating and decision-making.
Introduction
Human speech is characterized by high acoustic variability stemming from diverse accents, individual vocal characteristics, and environmental noise. To maintain comprehension, listeners employ "lexically-guided perceptual learning." For instance, when encountering an ambiguous sound between /f/ and /s/ at the end of the word "loa_", the brain utilizes lexical knowledge of the word "loaf" to categorize the sound as /f/ and subsequently adjusts its internal phonetic map for that specific speaker.
While behavioral evidence for this phenomenon is well-established, the underlying neural mechanisms have remained largely unexplored. This study aims to identify the neural signatures of this adaptation, specifically focusing on how listeners handle ambiguous phonemes in a second language.
Methods and Materials
The researchers utilized a pretest-exposure-posttest experimental design coupled with continuous EEG recording.
- Pretest: Participants categorized ambiguous sounds from an 11-step acoustic continuum between [f] and [s].
- Exposure: Participants listened to a narrative where all /f/ sounds were replaced with an ambiguous [f/s] token. The lexical context (e.g., sheriff, belief) necessitated an /f/ interpretation.
- Posttest: The categorization task was repeated to measure shifts in the phonetic boundary.
The primary neural metric was the P3b component, a positive-going ERP waveform occurring 250–550ms post-stimulus, associated with attention, stimulus evaluation, and categorization difficulty.
Stimuli
Target words (20 ending in /f/, 20 in /s/) were selected from the SUBTLEX-US corpus. Audio was recorded by a native speaker of American English, and ambiguous tokens were generated using the STRAIGHT algorithm in MATLAB to ensure a precise acoustic continuum.
Results
Behavioral Divergence
A notable finding was the "learner split": only 50% of the 36 participants (n=18) showed a statistically significant behavioral shift toward categorizing ambiguous sounds as /f/ following exposure. This group showed a 22.9% decrease in /s/ responses, effectively shifting their category boundary.
Neural Adaptation
In contrast to the behavioral results, a reliable reduction in the mean amplitude of the P3b component (approximately 1 μV, $p = 0.02$) was observed across the entire participant pool. This shift occurred in the 250–550ms window over medial central channels (Fz, Cz, Pz), suggesting the brain's processing of these sounds was fundamentally altered by the story exposure regardless of explicit behavioral change.
Brain-Behavior Decoupling
The study revealed a surprising lack of correlation ($r = 0.22$, non-significant) between the magnitude of neural change and behavioral shift. Even "non-learners" exhibited the P3b reduction, suggesting that the brain may update its internal models "under the hood" without necessarily reaching the threshold required to alter explicit categorization decisions.
Discussion
The reduction in P3b amplitude likely reflects a change in the ease of categorization; as the brain adapts to the speaker's idiosyncratic sounds, the ambiguous tokens become easier to process. However, several factors warrant further investigation. The observed neural changes could partially reflect habituation, although the P3b is generally resistant to such effects. Furthermore, the participants' status as non-native (Dutch) listeners of English may have introduced existing phonemic biases that influenced the learning process.
Future research should focus on the exposure phase itself to capture neural changes as they occur in real-time, providing a more granular view of the transition from lexical processing to phonetic adaptation.
Summary updated