TL;DR
This paper investigates how the pitch range of Japanese boundary pitch movements (BPMs) is controlled, a topic previously under-explored. The authors found that while peaks of BPMs do lower with preceding accents (peak downstep), the pitch range itself doesn't consistently shrink, suggesting a different control mechanism than previously assumed for Japanese intonation.
What Problem Does This Paper Address?
In Japanese, boundary pitch movements (BPMs) mark the ends of prosodic phrases and convey pragmatic information like questioning or continuation. While researchers have studied what BPMs exist and what they do, how their pitch range is controlled has been largely overlooked. A key concept in Japanese intonation is "downstep," where preceding accents can lower the pitch of subsequent phrases. The central question here is whether this downstep also affects the pitch range of BPMs, or if their pitch range is controlled independently. If BPMs are indeed independent, it challenges current models of Japanese intonation.
Methods
The researchers analyzed a large corpus of spontaneous Japanese monologues. They focused on two main prosodic units: accentual phrases (APs) and intonational phrases (IPs). To quantify downstep, they measured pitch characteristics (peaks, bottoms, and pitch ranges) of both the "AP body" (the part of an AP excluding the BPM) and the BPMs themselves. They analyzed these characteristics as a function of the number of preceding accents within the same IP, a metric termed "#PrecAcc." Statistical analyses, including one-way ANOVAs, were used to determine the significance of observed trends.
Datasets
The study utilized a subset of the Corpus of Spontaneous Japanese (CSJ) Core. This included approximately 90 hours of "academic presentation speech" and 20 hours of "simulated public speech," all of which were monologues.
Key Insights and Ideas
- BPMs are not simply passive recipients of downstep: Unlike the main body of accentual phrases (AP bodies), where pitch range consistently shrinks with preceding accents (range downstep), BPMs show a more nuanced response.
- Peak downstep is present, but range downstep is limited: All BPM types examined showed a lowering of their peak pitch as the number of preceding accents increased (peak downstep). However, consistent range downstep (compression of pitch range) was largely absent, except for the falling portion of the HL% BPM type.
- Rising BPMs are predictable from AP bodies: For H% and LH% BPMs, and the rising part of HL%, the pitch range remained constant. The lowering of their peaks and bottoms was found to be directly related to the downstep occurring in the preceding AP body. This suggests their pitch range is not independently controlled but rather follows the contour of the AP body.
- Falling BPMs show range compression: The falling part of the HL% BPM type, however, did exhibit range compression as preceding accents increased, behaving more like AP bodies in this regard.
- Current intonation models may be insufficient: The findings suggest that existing frameworks for Japanese intonation, often built on the analysis of AP bodies, may not adequately capture the specific pitch range control mechanisms of BPMs.
Key Results
- AP bodies: Confirmed range downstep, with both peaks and bottoms decreasing as #PrecAcc increased, and pitch ranges being compressed.
- H% BPMs: Showed peak downstep but no range downstep; pitch ranges remained constant. The lowering of peaks and bottoms was attributed to downstep in the AP body.
- HL% BPMs: Exhibited peak downstep. The rising part showed no range downstep, similar to H%. However, the falling part showed a tendency towards range compression (fall range became smaller) as #PrecAcc increased.
- LH% BPMs: Similar to H%, showed peak downstep but no range downstep. However, the authors note that the sample size for LH% was smaller, requiring cautious interpretation.
Limitations and Future Work
The authors acknowledge that the sample size for LH% BPMs was significantly smaller than for other types, making the results for this type less robust. They also note that the study focused on spontaneous monologues, and further research is needed to understand how these findings generalize to other speech styles or dialogues. The paper strongly suggests that a new theoretical framework is necessary to fully account for the observed pitch range control of BPMs, incorporating these findings into a more comprehensive model of Japanese intonation.
Summary updated