You are viewing the SNL 2017 Archive Website. For the latest information, see the Current Website.

Poster B23, Wednesday, November 8, 3:00 – 4:15 pm, Harborview and Loch Raven Ballrooms

Cortical tracking of linguistic structures: the role of covert prosody

Anastasia Glushko1,2, David Poeppel3,4, Max Wolpert1,2, Toivo Glatz5, Karsten Steinhauer1,2;1McGill University, 2The Centre for Research on Brain, Language and Music, 3New York University, 4Max Planck Institute for Empirical Aesthetics, 5University of Groningen

Using magnetoencephalography (MEG), Ding, Melloni, Zhang, Tian, and Poeppel (2016) have shown that neural oscillations track structural characteristics of spoken language. Because MEG power peaks appeared at the same frequencies as the syntactic phrase boundaries and in the absence of overt prosodic phrase boundary cues, the authors concluded that their results reflected processing of hierarchical syntactic structures. We tested the validity of this claim by investigating the relative role of syntactic and covert prosodic phrasing in the increase of oscillatory power at phrasal rate. We hypothesized that effects reported by Ding and colleagues (2016) were mediated by covert prosodic/rhythmic chunking (see e.g., Steinhauer & Friederici, 2001; Nozaradan, Peretz, Missal, and Mouraux, 2011). Our participants' electroencephalography (EEG) was recorded while they were listening to German four-word sentences or sequences of four nonwords of the same duration (1 second in total; 250 ms/(non-)word), which differed in their prosodic/rhythmic structure. Sentences of two types of syntactic phrasing were used: e.g., “Tom | goes to Bonn” (“1+3” syntactic grouping: word 1 forms a noun phrase; words 2-4 form a verb phrase; note that the sentences were presented in German) and “Your song | sounds good” (“2+2” syntactic grouping). Each sentence was presented in two versions of prosodic grouping (“1+3” and “2+2”), both of which sounded natural, but only one of which corresponded to the syntactic grouping of the respective sentence. After listening to the sentences containing overt prosodic cues, German native speakers (N=17) were presented with the same sentences with all prosodic cues eliminated (i.e., sentences with flattened pitch and controlled sound intensity contours). Participants were asked to mentally map the intonation of the overt-prosody sentences onto the de-prosodized stimuli. We found that at least in a subset of experimental conditions with no overt prosodic cues, frequency tagging analysis reflected processing of covert (mentalized) prosodic grouping rather than formal syntactic structure of the speech stream (e.g., a 1Hz but not a 2Hz EEG power peak for sentences with 2+2 syntactic and 1+3 prosodic grouping). Additionally, we tested the domain-general nature of this cognitive chunking mechanism in the same frequency tagging experimental setting by presenting participants with nonwords grouped either based on vowel quality (1+3: “kö-bü-gü-tü” vs. 2+2: “kö-bö-gü-tü”) or on music-like pitch differences. Neural tracking of rhythmic structure in these conditions was similar to that for sentences with overt and covert prosodic grouping. In the case of nonword sequences, the effects were independent of the specific cues used to mark chunk boundaries and detectable in the absence of sound intensity modulations in the auditory stimulus. Our preliminary results, therefore, suggest that the emergence of EEG power peaks corresponding to the rate of syntactic phrasal boundaries is mediated by the processing of covert prosodic/rhythmic grouping of the sentences. Given the absence of a one-to-one correspondence between prosodic and syntactic grouping in natural language, caution should be taken when using the frequency tagging technique as a marker of syntactic processing/grammar learning in neurolinguistic research.

Topic Area: Grammar: Syntax

Back to Poster Schedule