Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Low-frequency neural dynamics are sensitive to the timescales of multi-word phrases during speech production planning
Poster Session E, Friday, October 2, 11:00 am - 1:00 pm, Wangari Maathai
Jordi Martorell1,2, Daria Goriachun1,2, Liina Pylkkännen3, Kristof Strijkers1,2,4; 1Aix-Marseille Université (AMU) & Laboratoire Parole et Langage (LPL), Aix-en-Provence, 2Institute for Language, Communication and the Brain (ILCB), Marseille, 3New York University (NYU), 4CNRS
Electrophysiological evidence indicates that low-frequency (< 4 Hz) neural dynamics support speech comprehension. This is often observed as neural synchronization to the timescales of multi-word phrases, corresponding to the duration of their acoustic and abstract features. A few speech production studies have reported analogous effects before articulation, suggesting similar low-frequency synchronization to linguistic timescales in production. However, these studies have not directly manipulated the timescales of the to-be-produced speech and thus remain inconclusive about how low-frequency neural dynamics support speech production planning. In this electroencephalography (EEG) experiment, we investigated the sensitivity of low-frequency neural dynamics to linguistic timescales in multi-word phrase production. We focused on French 2-word phrases (noun + color, e.g., “château vert”, castle green) that can be easily produced from simple pictures while minimizing articulation-related EEG artifacts. We manipulated the relative number of syllables in each word (1 syllable = short vs. 2 syllables = long), resulting in two conditions with reversed word durations (short/long noun + long/short color). This manipulation thus selectively varied the internal timescales of phrases but matched their number of words and syllables. French-speaking participants (n = 26) performed a picture-naming task with colored objects representing such phrases (224 trials divided into 2 blocks) while recording their produced speech and EEG signals. We selected correct trials with comparable speech onset times (0.5 - 1.1 s from picture presentation, i.e., low-frequency timescales), and EEG signals were further time-locked to single-trial speech onset at negative lags (-0.5 s before articulation). To assess neural synchronization, we computed inter-trial phase coherence (ITPC) at low-frequency (< 4 Hz) neural activity (obtained from low-pass filtering and Hilbert transform methods). Behavioral results showed between-condition speech onset differences (short-long > long-short) in both blocks. Crucially, one particular condition-block contrast (short-long in block 2 vs. long-short in block 1) displayed no speech onset differences. ITPC results revealed distinct low-frequency dynamics between conditions at centro-posterior electrodes, with earlier peaks (maximum ITPC) for short-long (~ -0.25 s) than long-short (~ -0.2 s) phrases across blocks. These latency effects (short-long < long-short) were also observed even without speech onset differences (critical condition-block contrast), along with other dissociations from behavior. We further replicated peak latency patterns (short-long < long-short) with broadband event-related potentials (ERPs) at centro-posterior slow dynamics (maximum amplitude) for most statistical contrasts, possibly ruling out methodological confounds from the filtering method. Our results provide evidence that the timescales of multi-word phrases impact speech production planning. The brain-behavior asymmetry between conditions (earlier EEG peak, later speech onset) together with certain brain-behavior dissociations suggests that neural latency effects probably cannot index motor preparation for speech articulation. Instead, our EEG results are more consistent with that low-frequency neural dynamics reflect the speech planning of multi-word phrases containing distinct timescales. Importantly, this selective sensitivity to the internal timescales of phrases (peaking earlier/later depending on the short/long duration of their words) aligns with previous findings from speech comprehension. Our results thus extend the link between low-frequency neural dynamics and linguistic timescales to speech production planning.
Topic Areas: Language Production, Syntax and Combinatorial Semantics