Sequences of Intonation Units form a ~ 1 Hz rhythm

Abstract

Studies of speech processing investigate the relationship between temporal structure in speech stimuli and neural activity. Despite clear evidence that the brain tracks speech at low frequencies (~ 1 Hz), it is not well understood what linguistic information gives rise to this rhythm. In this study, we harness linguistic theory to draw attention to Intonation Units (IUs), a fundamental prosodic unit of human language, and characterize their temporal structure as captured in the speech envelope, an acoustic representation relevant to the neural processing of speech. IUs are defined by a specific pattern of syllable delivery, together with resets in pitch and articulatory force. Linguistic studies of spontaneous speech indicate that this prosodic segmentation paces new information in language use across diverse languages. Therefore, IUs provide a universal structural cue for the cognitive dynamics of speech production and comprehension. We study the relation between IUs and periodicities in the speech envelope, applying methods from investigations of neural synchronization. Our sample includes recordings from every-day speech contexts of over 100 speakers and six languages. We find that sequences of IUs form a consistent low-frequency rhythm and constitute a significant periodic cue within the speech envelope. Our findings allow to predict that IUs are utilized by the neural system when tracking speech. The methods we introduce here facilitate testing this prediction in the future (i.e., with physiological data).

Rhythm Comes, Rhythm Goes: Short-Term Periodicity of Prosodic Phrasing

“…Initial support suggests that syllable frequency across languages is confined to a narrow band between 4 and 8 Hertz (Ding et al, 2017;Coupé et al, 2019;Tilsen & Arvaniti, 2013), consistent with the previously proposed role of theta-band synchronization in syllable processing (Luo & Poeppel, 2007;Peelle et al, 2013;Doelling et al, 2014;Howard & Poeppel, 2012). Above the syllabic rate, a recent study found that the onset of intonation units that are analogous to ToBI's ip (Chafe, 1979;Du Bois, 1992) associates with a consistent phase angle of the speech envelope at a frequency of ∼1 Hertz (Inbar et al, 2020). This suggests that amplitude modulations at prosodic boundaries may indeed be sufficiently regular for synchronizing a periodic neuronal sampling mechanism.…”

“…While we studied a symbolic annotation of ips instead of their acoustics, ip boundaries are defined by multi-dimensional modulations, second in ToBI break index strength only to IP boundaries (Silverman et al, 1992;Pierrehumbert, 1980). In line with this, a subset of these cues, amplitude modulations at the boundaries of intonation units analogous to ips, are periodic at 1 Hertz (Inbar et al, 2020;Chafe, 1979;Du Bois, 1992). We extend this work here in two important ways.…”

“…In turn, this showed that ip period exhibits a range rather than being fixed to a strict one Hertz. The dependence on IP duration could even explain the cross-linguistic variance in periodicity observed by Inbar et al (2020), which should be tested by future work.…”

“…We extend this work here in two important ways. While Inbar et al (2020) analyzed periodicity within fixed windows of two seconds starting from the onset of an intonation unit, the hierarchical annotation scheme of ToBI allowed us to use IPs as windows. This revealed that periodicity outlasts time windows of two seconds (i.e., two intonation units), continuing for up to five ips (Figure 2B).…”

“…In general, listeners employ prosodic boundaries to segment verbal stimuli (Bower & Springston, 1970;Farrell & Lelièvre, 2012;Sturges & Martin, 1974;Ghitza, 2017); in particular, behavioral and electrophysiological data indicate that prosodic boundaries trigger the termination of syntactic units (Lehiste et al, 1976;Price et al, 2015;Wightman et al, 1992;Snedeker & Casserly, 2010;Steinhauer et al, 1999). It is often proposed that the sampling of the words contained within an IP allows the listener to decode a single coherent piece of information (Inbar et al, 2020;Wahl, 2015;Speer & Ito, 2009). Why is speech subdivided into prosodic units, rather than providing an uninterrupted stream of information?…”