Neural representation of prosody
Tamar I Regev1, Niharika Jhingan1, Hee So Kim2, Hope Kean1, Colton Casto1, Evelina Fedorenko1; 1Massachusetts Institute of Technology, 2Wellseley College
Supra-segmental prosody refers to acoustic features of spoken language beyond the phonetic information, and is sometimes described as “the melody of language”. Prosodic cues, which include pitch, loudness, and duration, can convey linguistic, emotional, and other socially relevant information. Does the brain contain areas specialized for processing prosody or is prosodic information processed by some combination of areas that process pitch, speech, language, or some aspects of social interaction? Previous neuroimaging studies have reported sensitivity to various aspects of prosodic information in a number of brain regions, but have typically only included a limited number of conditions, making it difficult to establish functional selectivity for prosodic processing and thus to infer the underlying neural computations. Here, we designed a new robust fMRI paradigm for localizing regions sensitive to prosody in individual participants, and also included several other extensively validated ‘localizer’ tasks for auditory, linguistic, and social brain areas. Our critical experiment included six conditions: A. Content+prosody+: spoken sentences (extracted from spoken-language corpora) pronounced with an expressive prosody by a native English speaker; B. Content+prosody-: the same sentences as in A, but with reduced/distorted prosody (each word was recorded separately and then re-combined); C. Content-prosody+: sentences made up of phonologically matched pseudo-words and recorded with a matched prosody to the sentences in A; D. Content-prosody-: the same pseudo-word sentences as in C, but recorded one word at a time and re-combined, as in B; E. invertedC: the stimuli from condition C temporally inverted; and F. invertedD: the stimuli from condition D temporally inverted. Conditions A-D constituted a 2x2 design crossing linguistic content and prosody, whereas E-F served as controls for low-level acoustic differences between the prosody+ and prosody- conditions. In addition, each participant completed a localizer for pitch perception areas (Norman-Haignere et al., 2013), speech-selective areas (Overath et al., 2015), high-level language areas (Fedorenko et al., 2010), social perception areas (Pitcher et al., 2011), and domain-general multiple-demand areas (Duncan, 2010; Fedorenko et al., 2013). Two main results emerged: First, the left-lateralized language-processing areas (Fedorenko et al., 2010) and their right-hemisphere homotopes were both sensitive to prosody but only in the absence of linguistic content, as evident by a significant effect for C>D and no effect for A>B, in both hemispheres. This result suggests that previous reports of the right-hemisphere’s role in prosody could not be arising in the right homotopes of the language areas. And second, a set of brain areas in frontal and temporal cortex showed functional specialization for prosody, in materials with and without linguistic content (A>B and C>D), relative to pitch perception, speech perception, language comprehension, social perception, and general cognitive demands. Our results elucidate the neural mechanisms that support prosodic processing, including critically demonstrating the existence of regions that are functionally selective for prosodic processing and distinct from several other well-established functional areas. This work lays a critical foundation for further investigations of prosodic processing and its neural basis.
Topic Areas: Prosody, Perception: Speech Perception and Audiovisual Integration