Slide Slam S5 Sandbox Series
Decoding the scope of planning in sentence production
Laura Giglio1,2, Markus Ostarek1,2, Peter Hagoort1,2; 1Max Planck Institute for Psycholinguistics, 2Donders Institute for Brain, Cognition and Behaviour
The neurobiology of sentence production has been understudied compared to sentence comprehension. A recent study on the differences and similarities between comprehension and production found substantial overlap in the networks engaged by each modality, but the BOLD time course for the production and comprehension of the same sentences was different (Giglio et al., in revision). In particular, the BOLD response was found to peak earlier for more complex structures relative to simpler structures in production, whereas the opposite pattern was observed in comprehension. This dissociation likely reflects earlier event conceptualization in production than comprehension, in line with eye-tracking evidence showing an initial focus on event apprehension in picture descriptions and action encoding before speech onset (Griffin & Bock, 2000; Konopka, 2019). Picture-word interference tasks instead show that the verb is not planned before speech onset unless the object is said first (Momma et al., 2016). In the present fMRI study, we aim to provide new insight into the processes and representations engaged in different levels of sentence planning, from event conceptualization to grammatical and phonological encoding, using multi-voxel pattern analysis. We will run a sentence production study with a sentence recall paradigm. Participants read sentences word-by-word at a very fast pace (150 ms per word) and repeat the sentence aloud after a distraction task. This paradigm was previously shown to lead to the reconstruction of the sentence from the conceptual meaning rather than by verbatim memory (Potter & Lombardi, 1990, van den Velde & Meyer, 2014). The use of sentence recall allows for getting closer to production elicited from a conceptual thought held in memory as in natural speech, whilst maintaining high experimental control. The stimuli are sentences in active or passive voice. Agents and patients refer to people either playing sports or music, and performing either actions requiring contact (e.g. kicking, hitting) or involving perception (e.g. seeing, noticing: “the musician sees the athlete”). We will perform classification of subject, object and verb semantic categories in the time windows before speech onset to determine to what extent these concepts are processed before sentence initiation. In addition, we will investigate how specifically the thematic role of the subject is encoded before speech onset. This will be done for broad thematic roles across actions (subject as agent), as well as for the narrower noun-verb role conjunctions (e.g., subject as kicker), which were found to be distinct in a previous comprehension study (Frankland and Greene, 2020). This will allow us to learn more about compositional processing in production. Finally, we will zoom in on sentence initiation to gain insight into phonological, morphological, and syntactic encoding during speech. In particular, we will run representational similarity analysis to characterize the brain regions involved in representing different levels of sentence production without the use of task manipulations as usually performed in univariate fMRI designs.