Some adults learn new words much more easily than others. This study asks whether that difference can already be seen in the resting brain, before the learning task even begins. More than one billion people are learning a foreign language each year, yet foreign language acquisition, especially in adulthood, is characterized by considerable variability in learning rate and ultimate attainment.

For some people, learning a foreign language is time-consuming and challenging, while for others, high levels of language proficiency can be attained with relative ease and little time investment. Understanding the neural mechanisms responsible for this heterogeneity among late language learners is crucial for elucidating the nature of learning ability and further enhancing learning efficiency through precise neuromodulation.

A resting-state functional connectome measures the temporal synchronization of spontaneous neural activity between anatomically separated brain regions. Functional connectome profiles can act as a unique fingerprint used to predict inter-individual differences in behavior and cognition, making resting-state connectivity promising for studying language-learning differences.

Most earlier work used functional magnetic resonance imaging, which has high spatial resolution but measures neural activity indirectly through slow hemodynamic changes. Electroencephalography has much higher temporal resolution, in milliseconds, and reflects real-time neural processes based on cross-region coupling of fast oscillations.

The study focused on novel word learning, an essential component of language acquisition that is crucial to developing a large vocabulary and literacy skills. Unlike native-language word acquisition among children, learning novel words tends to be less efficient and highly variable among adults.

Adults vary in their ability to both learn and retain novel words, yet the origin of this inter-individual word-learning variability remains elusive. Word-learning ability is generally understood as the capability to establish connections between novel words and their semantic referents.

It remained largely unknown whether individual differences in word-learning ability may depend partly on differences in spontaneous brain connectome patterns at rest. The study investigated whether and how pre-task brain connectome measured by EEG was related to the ability to acquire novel words in adults.

The expectation was that individual resting-state EEG alpha coherence between occipital and frontal regions could predict inter-individual variability in word-learning ability. A total of 106 healthy, right-handed undergraduate students participated, including 50 males, with a mean age of 21.41 years and a standard deviation of 2.20 years.

All participants were native Chinese speakers who had received classroom-based English education for approximately 10 years, with an age of second-language acquisition of 8.23 years and a standard deviation of 2.15 years. They reported normal or corrected-to-normal vision and no history of neurological or language-related disorders, and all participants signed informed consent before the experiment.

Before the eyes-closed resting-state EEG recording, participants completed a language background questionnaire and a demographic survey. Afterward, they completed the paired-associate novel-word learning task and the LLAMA B task in sequence.

The novel-word learning task consisted of two stages: learning and testing, with participants instructed to learn and memorize novel words and their Chinese meanings presented on a computer screen. Participants studied 60 pairs of novel words over 6 rounds, with the pairs in each round presented only once in random order.

After learning, participants completed a distractor task and then a paper-and-pencil cued-recall test, writing the Chinese equivalent for each pseudo-English cue within 10 minutes. The number of words correctly recalled in the final cued-recall test was used as an index of word-learning ability.

Figure one lays out the study sequence: a five-minute eyes-closed resting-state EEG recording, followed by the novel-word learning task, and then the ten-minute LLAMA B task. The learning task includes a twenty-minute learning phase, a three-minute delay with a distractor, and a ten-minute testing phase; the lower panels illustrate sample learning and testing trials.

This timeline matters because it shows how resting brain activity and later language-task performance were measured within the same session. Continuous EEG was recorded with a 32-channel active electrode system while participants stayed awake with their eyes closed. The signals were bandpass filtered between 0.05 and 100 hertz and sampled at 500 hertz.

The resting-state EEG recording lasted 5 minutes and was implemented before the behavioral tasks. Functional connectivity between distinct brain regions was estimated using EEG coherence. After excluding two mastoid electrodes and FCz, the remaining 29 sites were used for the EEG coherence analysis.

Artifact-free data were transformed from the time domain into the frequency domain using Fast Fourier Transform. Coherence between all possible electrode pairs, 406 pairs, was calculated in delta, theta, alpha, beta, and gamma frequency bands. Participants showed robust inter-individual differences in novel-word learning performance, correctly recalling an average of 39.97 words, with a range from 6 to 60.

The LLAMA B scores also showed large inter-individual differences, with a mean score of 56.93 and a range from 10 to 100. Novel-word learning performance had a significant positive correlation with LLAMA B scores, so participants who performed better on the novel-word task also scored higher on the aptitude test.

The results suggest that word-learning ability is a relatively stable trait. Figure two maps the electrode analysis and its alpha-band connectivity results. Panel B highlights cortical electrode pairs whose coherence correlated significantly with word-learning performance after false-discovery-rate correction, while panel C shows the corresponding correlations across twenty-nine by twenty-nine channel pairs.

Panel D illustrates three examples—O1-FC1, Oz-FC5, and Oz-FC1—with positive relationships and reported Pearson correlations of r equals zero point three four eight, zero point three four zero, and zero point three three six, respectively. Significant positive correlations surviving correction for multiple comparisons were found only in the alpha frequencies, from 8 to 12 hertz, across the whole scalp sites.

Strong correlations were mainly observed between alpha-band long-range electrode connections over occipital regions O1 and Oz and frontal regions FC1, FC2, FC5, and FC6. No correlation reached significance after correction in the delta, theta, beta, or gamma bands, suggesting that the association was specific to the alpha band.

Alpha-band coherence between occipital and frontal electrode pairs was more strongly correlated with word-learning performance than connectivity in other regions, with O1-FC1, Oz-FC5, and Oz-FC1 showing the highest correlations. Regional analysis found that individual resting-state coherence between occipital and frontal regions significantly correlated with differential word-learning performance, with a correlation of 0.328 and a p value of 0.001.

Occipito-frontal coherence also differed significantly between learning-ability groups, and connectivity in the high-performance group was significantly stronger than in the low-performance group. Hierarchical regression showed that alpha-band occipito-frontal coherence contributed significantly beyond age and gender, with a beta value of 0.335 and a p value below 0.001.

Actual and predicted word-learning scores were significantly correlated, with a correlation of 0.341 and a root mean square error of 14.611, indicating significant prediction power for word-learning scores. Figure three links resting alpha-band EEG coherence between occipital and frontal electrodes with novel-word learning.

Panel A shows a significant positive association in the main sample, while panel B compares coherence across low, intermediate, and high learning-ability groups. Panel C relates actual and predicted learning scores, and panel D reproduces the coherence association in a thirty-five-participant replication study.

Panel E shows corresponding correlations separately for the left and right hemispheres, supporting an occipito-frontal network relationship with learning ability. The main findings of the EEG regional analysis were replicated in an independent validation study.

In that study, individual resting-state alpha-band coherence averaged between occipital and frontal regions significantly correlated with word-learning performance, with a correlation of 0.427 and a p value of 0.012. The results showed that higher alpha-band functional connectivity between occipital and frontal areas was associated with better word-learning performance in this study.

Resting-state functional connectivity within the occipito-frontal network could predict word-learning scores significantly beyond the contribution of age and gender. The study used pseudo-English words, so it remains unclear whether the findings generalize to real-word learning.

The study used only the number of correctly recalled words as an index of word-learning ability, while future studies may use measures such as learning speed or duration. Resting-state EEG cannot study multiple cognitive functions involved in word acquisition, such as attention and memory, and the cross-sectional design cannot establish causal relationships.

The limited spatial resolution of EEG also restrained conclusions about which brain regions are involved in the network, while simultaneous EEG-fMRI might improve temporal and spatial resolution. Further studies using non-invasive neuromodulation are needed to determine whether increased alpha-band connectivity could enhance word-learning performance.

Stronger resting-state alpha-band connectivity between occipital and frontal regions was associated with better novel-word learning and replicated in an independent group, but the cross-sectional design cannot establish causation.