Presentation
Search Abstracts | Symposia | Slide Sessions | Poster Sessions | Lightning Talks
Speech Recognition from a Familiar Speaker Engages the Person Identity Network
Poster E74 in Poster Session E, Thursday, October 26, 10:15 am - 12:00 pm CEST, Espace Vieux-Port
Gaël Cordero1, Jazmin Paredes1, Katharina von Kriegstein2, Begoña Díaz1; 1Department of Basic Sciences, Faculty of Medicine and Health Sciences, Universitat Internacional de Catalunya, Barcelona, Spain, 2Faculty of Psychology, Technische Universität Dresden, Dresden, Germany
Numerous studies have found that phoneme perception is dependent on spectro-temporal features characteristic of the speaker’s voice. Furthermore, voice familiarity facilitates speech recognition in acoustically challenging scenarios. Such interactions between voice and phoneme processes are unexpected when considering their anatomical and functional underpinnings. Voice sensitive regions are primarily located along the right temporal lobe and are functional constituents of the person identity network. Phoneme recognition predominantly engages regions in the left temporal lobe which are part of the language network. Previous neuroimaging studies have identified two neurofunctional mechanisms which might support the interaction between voice and phoneme processes. Firstly, several studies have identified regions along the temporal cortices and in the right temporoparietal junction that are sensitive to both voice and phonetic information. Secondly, studies have found interhemispheric functional connectivity between right voice sensitive regions and left phoneme sensitive regions when recognizing speech from multiple speakers. Whether the same neural mechanisms enable the use of voice priors during speech perception has not yet been investigated. By means of fMRI, we investigated if recognizing speech from a familiar speaker elicits neurofunctional connectivity and activity distinct from perceiving speech from an unfamiliar speaker. Eighteen right-handed adult participants performed a phoneme recognition task on non-words enunciated by a familiar and an unfamiliar speaker. We employed an independent functional localizer to define as the voice sensitive area a cluster situated in the right Temporal Pole which extended into the right anterior Superior Temporal Sulcus, and a cluster situated in the left posterior Superior Temporal Sulcus as the phoneme sensitive area. Our findings revealed the functional engagement of an extended network when processing speech from a familiar voice, rather than an increase in functional connectivity between phoneme and voice sensitive regions. The voice sensitive area exhibited greater functional connectivity with regions involved in person identity recognition, the anterior Cingulate Gyrus, and the left Frontal Pole. Furthermore, voice familiarity led to an increase in connectivity between the voice sensitive area and the right Supramarginal Gyrus, an area which is sensitive to phoneme predictability and to the typicality of the pronunciations of speakers. Better task performance with the familiar speaker correlated with greater functional anticorrelations between the voice sensitive area and two regions of the person identity network: the right Frontal Pole and left Superior Parietal Lobe. Finally, activity analysis revealed no differences between the familiar and the unfamiliar voices, but a task learning effect in the phoneme sensitive area evidenced by an increase of the BOLD activity as the experiment advanced. It has been shown before that voice-sensitive regions are recruited during speech recognition, but here we show for the first time that the extended person identity network is recruited for speech recognition as well. We propose that during speech perception, the person identity network builds predictions about the incoming sensory input and that the neural underpinning of said predictions is a network of higher-order regions which not only encode speaker voice properties, but semantic speaker knowledge.
Topic Areas: Speech Perception, Phonology