Phonetic realization of second occurrence focus in Japanese

Satoshi Nambu, Yong-cheol LeeInterspeech 2010

TL;DR

This paper investigates how "second occurrence focus" (SOF) – a repeated piece of information that's still emphasized – is pronounced in Japanese. They found that SOF in Japanese has a distinct pitch pattern, being less prominent than main focus but more prominent than neutral speech, and importantly, it experiences both pre- and post-focus pitch compression, offering a unique insight into cross-linguistic focus realization.

What Problem Does This Paper Address?

When we emphasize a word or phrase in speech, it's called "focus." This focus is usually marked by prosodic prominence, often through higher pitch, longer duration, or increased intensity. However, there's a special type of focus called "second occurrence focus" (SOF). SOF applies to information that has already been introduced in the conversation (old information), unlike "main focus" which applies to new information. Previous research suggested that SOF might not have strong prosodic prominence, with some studies finding it indistinguishable from neutral speech in certain positions.

The key problem this paper addresses is the lack of a clear understanding of SOF's phonetic realization, particularly in comparison to neutral speech. While previous studies compared SOF to main focus and pre/post-focus positions, they often missed the crucial comparison to speech without any focus. This omission makes it difficult to pinpoint the exact prosodic status of SOF. The authors aim to clarify this by comparing SOF to neutral-focus, main-focus, and pre/post-focus in Japanese, and in doing so, shed light on cross-linguistic differences in how focus is signaled.

Methods

The researchers conducted an experiment with native Japanese speakers. They created four sets of sentences to elicit different focus conditions: neutral-focus, main focus, second occurrence focus (SOF), and pre/post-focus. To ensure naturalness, SOF and main focus sentences were embedded in discourse contexts, while pre/post-focus conditions were elicited using prompt questions.

The core of their methodology involved analyzing the fundamental frequency (F0), which is the acoustic correlate of pitch. They used a Praat script to extract F0 from the target constituents in each sentence. Crucially, they measured both the mean F0 and the maximum F0 (Max F0) for these constituents. The analysis focused on comparing these F0 measures across the four focus conditions, both when the focused element appeared in an initial position and in a final position within the sentence. Statistical analysis, specifically paired t-tests, was used to determine the significance of observed differences in F0.

Datasets

The study used a custom-built dataset of 28 target sentences, each containing 6 words. These sentences were designed to naturally elicit the four focus conditions (neutral-focus, main focus, SOF, pre/post-focus) in both initial and final positions. The focused elements were marked using Japanese focus particles like "dake" (only) and "mo" (also). No pre-existing benchmark datasets were used or introduced.

Key Insights and Ideas

  • SOF has distinct pitch prominence in Japanese: Unlike some previous findings in other languages, SOF in Japanese does exhibit noticeable pitch prominence, though it's less than main focus.
  • Pre- and Post-Focus Compression in Japanese: The study provides strong evidence for both post-focus compression (PFC) – where pitch decreases after a focused element – and, uniquely, pre-focus compression – where pitch is lowered before a focused element. This pre-focus compression is not observed in languages like Korean or English.
  • Japanese SOF's unique position: SOF in Japanese is realized with higher pitch than neutral-focus in the pre-focus position but lower than neutral-focus in the post-focus position due to PFC. This suggests SOF is "focused" but its prominence is modulated by its position relative to other focused elements.
  • Cross-linguistic implications: The discovery of pre-focus compression in Japanese highlights significant typological variation in how languages signal focus, suggesting that Japanese focus realization is more complex than previously thought.

Key Results

  • Initial Position: In the initial position, the pitch hierarchy was Main Focus > SOF > Neutral-Focus = Pre-Focus. SOF was significantly higher than neutral-focus and pre-focus.
  • Final Position: In the final position, the pitch hierarchy was Main Focus > Neutral-Focus > SOF > Post-Focus. SOF was significantly lower than neutral-focus, demonstrating the effect of post-focus compression.
  • Pre-focus Compression: A statistically significant difference was found between the pre-focus element and the neutral-focus element, indicating pre-focus compression.
  • Post-focus Compression: A statistically significant difference was also found between the neutral-focus element and the post-focus element, confirming post-focus compression.

Limitations and Future Work

The authors acknowledge that while they found evidence for SOF's prosodic prominence in Japanese, further research is needed to precisely characterize pre-focus elements. They suggest that future work should aim to ascertain more precise characteristics of pre-focus elements in Japanese. The study also implicitly suggests that further cross-linguistic comparisons, particularly focusing on languages with and without pre-focus compression, would be valuable for understanding the broader landscape of focus realization.

Summary updated

BibTeX
@inproceedings{nambu2010phonetic,
  author = {Satoshi Nambu and Yong-cheol Lee},
  title = {{Phonetic realization of second occurrence focus in Japanese}},
  booktitle = {Interspeech},
  year = {2010},
}
Figure