Il Sonniloquio come via d’accesso all’elaborazione cognitiva in sonno: uno studio elettrofisiologico sulle produzioni verbali notturne

INTRODUCTION Sleep Talking (ST) is defined as the utterance of speech during sleep. The available literature cannot address ST as a REM or NREM parasomnias. Rarely studied as an isolated phenomenon, we currently have no definite evidence of its neural correlates. Recently, psycholinguistic features of verbal production in ST has been investigated, pointing to coherence with formal features of the language in wakefulness. This evidence, within the hypothesis of an involvement of parasomnias in sleep-related cognitive processing, suggests the importance of understanding the neural mechanism underlying ST. We aimed to investigate EEG correlates predictive of verbal activation (Verbal ST), with a comparison with Non-verbal ST (moaning, laughing, crying, etc...) with the general hypothesis of shared mechanisms with neural correlates of language processing and production. METHOD Six highly frequent ST (3M, 3F, age 19-27, mean 23.83±3.60) recruited through an online survey (general health assessment, PSQI for the self-declared quality of sleep, MUPS for self-declared presence and frequency of the phenomenon). Presence and frequency of ST, together with the compliance in maintaining a regular sleep schedule, has been further assessed through one week of home sleep/dream-logs and audio-activated recorder. The suitable participants have been recorded in the laboratory for at least 2-consecutive nights of video-PSG. We pursued the conditions of Verbal and Non-verbal ST, obtaining a total of 21 Verbal ST and 21 Non-verbal ST in Stage 2 NREM. Control comparison has been performed between Vocalizations (N=42 vocal activations, combining Verbal and Non-verbal) and Baseline (N=42 equivalent sleep interval, 2 mins preceding each ST), to assess time-locking and specificity of the observed EEG pattern, in relation with vocal production. Artifacts were off-line rejected for the 20 seconds EEG preceding each Verbal and Non-verbal ST and Baseline, on a 4-seconds basis. EEG power spectra have been obtained throughout a Fast Fourier Transform (FFT) routine. The power spectra have been then averaged to obtain the canonical sleep EEG bands: delta (0.5-4.5Hz), theta (4.75-7.75Hz), alpha (8-11.75Hz), sigma (12-15.75Hz) and beta (16-24.75Hz), and subsequently log-transformed. RESULTS Statistical comparisons (t-Test) show a general decrement in power spectra for Verbal ST vs. Non-verbal ST for the theta and alpha EEG bands. This effect is strongly lateralized to the left hemisphere and specifically localized on centro-parietal-occipitals channels. A single left parietal channel (P7; theta t=-4.48, p=0.0002; alpha t=-3.29, p=0.0037) was significant also after the Bonferroni correction. T-test comparisons (t=≥2.96; p≤0.0051) for Vocalizations vs. Baseline show a general increment in power spectra for delta on frontal, central and temporal channels, almost on the entire scalp for the alpha band; on Fc2 and C4 for the beta band. CONCLUSIONS Our results suggest shared neural mechanisms between Verbal ST and language programming during wakefulness. Specifically, the selective decrement for the theta band on the left parietal sites is coherent with the literature about linguistic planning in wakefulness, suggesting a possible functional overlapping. Moreover, the phenomenon seems time-locked to the interval of sleep preceding ST, as demonstrated throughout a comparison with a preceding sleep interval