Search Abstracts | Symposia | Slide Sessions | Poster Sessions
Levels of Inner Speech – An Investigation of Oscillatory Dynamics
Poster Session F, Friday, October 2, 2:45 - 4:45 pm, Wangari Maathai
Viktoria Schmitzer1, Jutta L. Mueller1; 1University of Vienna
Inner speech has been hypothesized to exist on different representational levels corresponding to the processing steps of overt speech production. Such levels are broadly a semantic level, a phonological level, and an articulatory level. Yet, these levels have not been sufficiently supported by neurophysiological evidence. The LISP project aims to induce these different forms of internal verbalization by means of a working memory paradigm and uses time-frequency data obtained through EEG to differentiate between them. Furthermore, interindividual differences are investigated through questionnaires about participants’ propensity for inner speech and visual imagery. Here, we present data from two out of three planned EEG experiments, targeting the phonological (experiment 1) and semantic (experiment 2) levels of inner speech. In each experiment, participants view four objects placed in a 6x6 grid and have to memorize either the objects themselves for the verbal conditions or the position of the objects for the visual control conditions. After a short maintenance phase, another object is presented, and participants perform a rhyme judgment (experiment 1) or a semantic judgment (experiment 2). In the visual control condition (experiments 1 & 2), the maintenance phase is followed by a grid with “X” placed in four positions and participants judge the accuracy of these four positions. Crucially, the exact same stimulus material was shown in the encoding phase across these conditions. A second visual control task (experiments 1 & 2) using diffeomorphic images in place of the objects was used to further control for the influence of object recognition and potential automatic implicit labeling during the visual task. The conditions contained 40 trials each. EEG was recorded from 35 adults using 62 scalp electrodes in experiment 1 and 37 adults using 64 electrodes in experiment 2. Time-frequency power values were obtained for the encoding and maintenance phase of each trial through wavelet analysis. Cluster-based permutation tests and linear mixed models were used for statistical analysis. Both phonological level and semantic level inner speech are characterized by higher power in the alpha and beta bands compared to the visual control condition during the encoding and maintenance of the objects or their positions. Furthermore, this power difference appears to be modulated by the general working memory capacity as well as the propensity for inner speech and visual imagery, as measured through the VISQ and IRQ questionnaires. While both verbal conditions show significant differences from the visual control condition in the alpha and beta frequency bands, preliminary analysis suggests that the distribution of the significant clusters varies between phonological and semantic level inner speech. Such differences suggest that when participants use inner speech in a working memory task, they can employ different levels of linguistic encoding to best fit the demands of the task.
Topic Areas: Language Production, Control, Selection, and Executive Processes