Effects of Syntactic Structures on Intonational Pitch Movement in Mandarin Chinese

Abstract

Previous research on Mandarin Chinese tones and intonation has focused primarily on universal sentence pitch patterns (declination) and sentence types (declarative and interrogative). The specific impact of internal syntactic structures remains under-explored. This study presents two acoustic experiments using controlled Tone 1 (high-level) stimuli to isolate intonational “big waves” from lexical “small ripples”. Experiment 1 investigates how syntactic position (subject vs. object), relative clause type (subject-relative vs. object-relative), and word class (verb vs. noun) influence pitch contours. Experiment 2 resolves conflicting findings regarding word-class pitch by testing nouns and verbs across four sentential contexts. The results indicate that subject positions carry significantly higher pitch than object positions, reflecting an interaction between SVO word order and declination. Crucially, subject-relative (SR) clauses exhibit a falling pitch tendency, while object-relative (OR) clauses show a rising trend. These results suggest that pitch realization is a complex “algebraic sum” of universal phonological trends, syntactic hierarchy, and semantic information structure.

1. Introduction

Language is a system of intricately interwoven components. While phonology, syntax, and semantics have traditionally been studied in relative isolation, the past two decades have witnessed a growing recognition of their mutual interdependence. The emergence of prosodic grammar (Selkirk, 1984; Nespor & Vogel, 1986; see also Féry & Ishihara, 2016 for a recent overview), a field examining the interface between sound structure and grammatical organization, has brought to the fore questions concerning how syntactic configurations leave measurable traces in the speech prosody. Leveraging a quantitative acoustic methodology, the present study explores the mechanisms by which specific syntactic structures modulate intonational patterns in Mandarin Chinese.

When discussing Chinese intonation, it is essential to address the relationship between tone and intonation. Tone is defined as a phonemic feature that systematically employs pitch to distinguish lexical meaning. In contrast, the definition of intonation remains relatively ambiguous and subject to considerable debate. Synthesizing the existing literature across various languages, a consensus within the academic community can be summarized into three core dimensions (Zhang, 2016, pp. 17–19):

First, intonation is intrinsically linked to suprasegmental phonetic features. In a narrow sense, the physical parameter of intonation is restricted to pitch; in a broader sense, it encompasses intensity, duration, and other factors. The present study adopts the narrow definition, focusing exclusively on pitch characteristics. Second, intonation pertains to pitch movement at the sentential level, occurring above the lexical level—a distinction that separates intonation from lexical tone. Third, sentential pitch movement conveys both linguistic and paralinguistic information. For linguists, the primary concern is the linguistic dimension. Specifically, the structural features that intonation expresses. This study investigates pitch movement within the framework of these three consensus points.

In non-tonal languages, such as English, pitch movement is generally synonymous with intonation. In contrast, Mandarin Chinese is a tonal language where pitch trajectories simultaneously encode both lexical tone and intonation. Yuen Ren Chao posited that “the pitch movement of Chinese speech is the algebraic sum of tone and intonation” (Chao, 1956/2006, p. 598; 1968, p. 39). He metaphorically described this relationship by stating that “the relation between Chinese tone and intonation is like small ripples riding on big waves” (Chao, 1968, p. 39). In this analogy, lexical tones represent the “small ripples”. For instance, the four tones in Standard Mandarin exhibit distinct pitch trajectories: Tone 1 is high-level, Tone 2 is high-rising, Tone 3 is low-dipping (or low-falling), and Tone 4 is high-falling. The acoustic characteristics of these “small ripples” across various contexts have been extensively and systematically documented (e.g., Y. Xu, 1997; Zhang & Shi, 2020). However, the movement of the intonational “big waves” and the factors governing them are far more complex and have received less systematic attention.

A crucial theoretical constraint that has gained recognition in recent years is the Principle of F₀ Monofunctionality (Feng, 2015), which states that the fundamental frequency (F₀) contour at any given point in a tonal language cannot simultaneously encode both a lexical tone target and an independent intonational target. This principle explains why sentence-final particles (SFPs) emerge in tonal languages: they provide a mechanism for indicating sentence-level meanings (e.g., question vs. declarative) without conflicting with lexically specified tones. As will be discussed below, this principle has important implications for how freely the intonational “big wave” can be realized in Mandarin. To date, several features of the Chinese (including Mandarin and dialects such as Cantonese) “big waves” have gained academic recognition:

First is the universal phenomenon of pitch declination (e.g., Cao, 2002; Liu, 2005). Declination is considered a default intonational pattern across the world’s languages (Ladd, 1984), characterized by the progressive lowering of pitch across successive syllables. Some researchers attribute this to physiological constraints, specifically the gradual decrease in subglottal air pressure during an utterance (Cruttenden, 1986; Cao, 2002; Liu, 2005). Others argue that declination is not merely a passive physiological byproduct but carries specific linguistic significance (Ohala, 1978; Zhang, 2014, 2017).

Second is the distribution of intonation within a sentence. Fox et al. (2008) distinguished between “utterance-body” and “utterance-final” intonation. Utterance-body intonation spans the main backbone of the sentence; beyond universal declination, it involves pitch reset at phrasal boundaries, resulting in a global downward “sawtooth” pattern known as downdrift. Utterance-final intonation, often termed boundary tone, primarily affects the final syllable. In declarative sentences, the final syllable often exhibits an additional pitch drop (Vance, 1976), which typically does not neutralize the lexical tone. In general questions, however, the final syllable often presents a high-rising contour (Ma, 2007; B. R. Xu & Mok, 2011), which may even neutralize the original lexical tone. Shen (1994) proposes a more granular structural division (crown, head, nucleus, and tail) but emphasizes the importance of a high-level bipartite distinction, noting that the “intonational trunk” serves as the primary carrier of intonational information.

While the aforementioned studies have successfully identified the impact of global pitch trends (declination) and sentence modality (declarative vs. interrogative) on the “big waves”, they have largely overlooked the influence of internal syntactic structure on prosody. If lexical tones, sentential positions, and sentence types are held constant, will different grammatical structures or word classes still result in divergent pitch realizations? In other words, can we find empirical acoustic evidence for the interaction between Chinese prosody and syntax?

The present study conducts a preliminary investigation into this question through two acoustic experiments. To isolate the intonational “big waves” from the interference of varying “small ripples”, all target syllables were restricted to the same tone. Specifically, the high-level Tone 1 in Mandarin was selected as the target, as its lack of inherent contour fluctuations facilitates the clear observation of intonational effects on pitch movement. However, it is important to acknowledge a limitation of this design from the outset: level tones present minimal conflict with intonational rises and falls. As discussed further in Section 4, the patterns observed here may not fully capture the behavior of the “big wave” when it interacts with contour tones such as Tone 2 (rising) or Tone 4 (falling). This limitation, while methodologically necessary for isolating intonation from tone, means that the present findings should be interpreted as a baseline that requires verification with non-level tones in future research.

2. Experiment 1: Syntactic Position and Relative Clause Type

2.1. Experimental Design

In this experiment, we investigate whether different relative clause types (Subject-Relative, SR vs. Object-Relative, OR) appearing in different sentential positions (Subject Position, SP vs. Object Position, OP) exert distinct influences on the intonational “big waves”. Additionally, we examine whether different word classes (Verbs, V vs. Nouns, N) exhibit inherent pitch difference.

The nouns selected for the stimuli include two-character names such as Jīn Xīn, Dīng Xīng, and Yīn Xīn (specifically designed with high front vowels in the nucleus to match the phonological environment of the target verb). The verbs include gēnzōng (track) and qīngxīn (adore). Furthermore, two functional words (F), de (attributive particle) and le (aspect marker)—both realized in the neutral tone—were used to connect these verbs and nouns. Four sentence types were designed as follows:

(1)

Subject Position, Subject-Relative Clause (SPSR):

[qīngxīn	Jīn Xīn]RC	de	Dīng Xīng	gēnzōng	le	Yīnxīn.
[adore	(name)]	attributive particle	(name)	track	aspect marker	(name)
The Ding Xing who adores Jin Xin tracked Yinxin

(2)

Subject Position, Object-Relative Clause (SPOR):

[Jīn Xīn	qīngxīn]RC	de	Dīng Xīng	gēnzōng	le	Yīn Xīn.
[(name)	adore]	attributive particle	(name)	track	aspect marker	(name)
The Ding Xing whom Jin Xin adores tracked Yin Xin.

(3)

Object Position, Subject-Relative Clause (OPSR):

Yīn Xīn	gēnzōng	le	[qīngxīn	Jīn Xīn]RC	de	Dīng Xīng.
(name)	track	aspect marker	adore	(name)	attributive particle	(name)
Yin Xin tracked the Ding Xing who adores Jin Xin.

(4)

Object Position, Object-Relative Clause (OPOR):

Yīn Xīn	gēnzōng	le	[Jīn Xīn	qīngxīn]RC	de	Dīng Xīng.
(name)	track	aspect marker	(name)	adore	attributive particle	(name)
Yin Xin tracked the Ding Xing whom Jin Xin adores.

In the examples above, the bracketed portions […]_RC represent the relative clause target regions. As illustrated above, the bold regions represent our primary target areas for observation. Within these target regions, we can directly compare SR and OR clauses, as well as the staggered contrast between different word classes (verbs vs. nouns) in the same linear position. By rotating the specific names (Jīn Xīn, Yīn Xīn, Dīng Xīng) across different positions, a total of 24 experimental sentences (6 variations × 4 types) were generated. Detailed stimuli are provided in the Appendix A.

2.2. Participants and Procedure

Ten native Mandarin speakers (five male and five female, aged 20–25) recruited from the Education University of Hong Kong participated in the experiment. All participants were born and raised in Beijing. The 24 experimental sentences (see Appendix A), along with 24 filler sentences (containing various tone combinations and syntactic structures not relevant to the current analysis), were presented in fully randomized order for each participant. A different randomization was generated for each speaker to prevent order effects. Fillers were interspersed to minimize contrastive or focus reading strategies. Prior to formal recording, participants were given a practice session to familiarize themselves with the stimuli. Informal debriefing confirmed that, despite the controlled phonological environment, participants found the sentences natural and meaningful. No participant reported significant difficulty in reading the sentences fluently. Participants were instructed to read the stimulus sentences twice in a quiet indoor environment. Both tokens were included in the subsequent data analysis. Recordings were made using a Marantz PMD620 professional digital recorder in mono at a sampling rate of 44,100 Hz and a 16-bit resolution.

The recorded data were segmented, and fundamental frequency (f₀) values were extracted using Praat (Boersma & Weenink, 2018). For each target syllable, the entire final portion (including vowel nucleus and coda) was annotated. Eleven equidistant measurement points were taken across the total duration of the syllable from the onset of the vowel nucleus to the offset of the coda (if any). Initial consonants (regardless of being voiceless or sonorant) were excluded from pitch extraction. As f₀ is the primary physical correlate of tone and intonation, the raw Hertz values (f_HZ) were first converted into semitones (f_ST) to better reflect the perceptual relationship with pitch, following the formula provided by Hart et al. (1990):

$$f_{\mathrm{S}\mathrm{T}}={\displaystyle \frac{12}{{\mathrm{l}\mathrm{o}\mathrm{g}}_{10}2}}{\mathrm{l}\mathrm{o}\mathrm{g}}_{10}{\displaystyle \frac{f_{\mathrm{H}\mathrm{Z}}}{100}}$$

To account for inter-speaker variation in pitch range, the data were normalized using Z-scores (Jassem, 1971). The Z-score serves as the final parameter for comparison:

$$z_i={\displaystyle \frac{f_{{\mathrm{S}\mathrm{T}}_i}-m}{s}}$$

In this formula, z_i represents the Z-score at point i, f_STi is the semitone value at that point, m is the mean semitone value across all data points for a specific speaker, and s is the standard deviation for that speaker. The Z-score indicates a relative pitch value calibrated to an individual speaker’s range, effectively normalizing the data and facilitating cross-speaker statistical aggregation and mean calculations.

2.3. Results

2.3.1. Global Intonational Contours and Declination

The mean pitch contours for the four sentence types are illustrated in Figure 1. In this visualization, blue open icons represent verbs (V), red solid icons represent nouns (N), and semi-open black icons denote functional words (F). Despite the fact that all syllables (excluding functional words) carry the same lexical Tone 1 (meaning the “small ripples” are phonologically identical), the “big wave” across all four sentential structures exhibits a consistent three-step descending trajectory. Specifically, the first four syllables occupy the highest pitch register, followed by the medial four syllables, with the final two syllables reaching the lowest register.

This overarching “big wave” pattern can be characterized as downstep, a specific manifestation of pitch declination. Given that Mandarin is an SVO language where the subject typically precedes the object, these results provide empirical evidence for the interaction between syntax and intonation: constituents in the subject position inherently carry a higher pitch than those in the object position. This phenomenon can be viewed as a prosodic derivative of global declination mapped onto syntactic linear order.

2.3.2. Effects of Clause Type and Word Class

To further examine the impact of relative clause type and word class on the intonational “large wave”, Figure 2 juxtaposes the mean curves of SPSR and SPOR (subject position) and OPSR and OPOR (object position). Observation of the target regions, i.e., the relative clause segments, reveals distinct divergence between the contours of SPSR/SPOR (initial four syllables) and OPSR/OPOR (medial four syllables). Conversely, outside these target regions, the curves for each pair nearly overlap. This localized divergence confirms that the variation in the “large wave” is specifically driven by the internal configuration of the relative clauses.

The data from Figure 2 suggest two primary characteristics across both subject and object positions:

(1)

V-N pitch differential: When comparing nouns and verbs in identical linear positions within the sentence, verbs (V) consistently exhibit higher pitch than nouns (N). This is visually represented by the blue V-curves tracking above the red N-curves in the target regions.

(2)

Structural contours (SR vs. OR): Subject-relative (SR) and object-relative (OR) clauses demonstrate contrasting directional trends regardless of their sentential position. SR clauses follow a falling trajectory (V-high to N-low), whereas OR clauses follow a rising trajectory (N-low to V-high). This structural crossover results in a wider pitch gap at the first disyllabic word compared to the second.

2.3.3. Smoothing Spline Analysis of Variance (SSANOVA)

To further verify whether the contrasts observed within the target regions are statistically significant, we employed SSANOVA. The resulting visualizations, generated using R, are presented in Figure 3 and Figure 4. In an SSANOVA plot, the confidence intervals (standard error bands) surrounding the curves serve as the criterion for significance: if the bands of two curves do not overlap, the difference between them is statistically significant; conversely, overlapping bands indicate a lack of significant divergence. This method of statistical modeling and visualization provides an intuitive and rigorous approach to analyzing functional curve data and has been increasingly adopted in recent acoustic phonetic research (Davidson, 2006).

In Figure 3 and Figure 4, vertical black dashed lines are employed to demarcate the boundaries between successive syllables. It should be noted that significant pitch fluctuations often occur near these boundaries; however, these transitional segments are typically considered phonetic artifacts lacking linguistic significance. Therefore, the present analysis focuses exclusively on the relatively stable pitch regions within each syllable.

The SSANOVA results in both Figure 3 and Figure 4 reveal that during the first disyllabic word, the confidence intervals for the verb (V) and the noun (N) do not overlap, indicating a statistically significant difference: the V in the Subject-Relative (SR) clause is markedly higher in pitch than the N in the Object-Relative (OR) clause. In contrast, during the second disyllabic word, the two sets of curves converge and exhibit substantial overlap (particularly in Figure 3), with the V in the OR clause appearing only slightly higher than the N in the SR clause.

To further illustrate the global pitch dynamics, colored arrows have been added to the figures to indicate the overall trajectory of each clause type. In both experimental conditions (subject and object positions), the SR clauses exhibit a slight downward (falling) trend, whereas the OR clauses follow a distinct upward (rising) trend.

2.3.4. Concluding Remarks for Experiment 1

The results of Experiment 1 provide empirical evidence for the interaction between syntactic structure and the intonational “large wave” in Mandarin Chinese. Based on the acoustic data and SSANOVA modeling, several conclusions can be drawn:

First, the global pitch contour is governed by a syntactic-prosodic hierarchy. While universal declination accounts for the general downward trend of the sentence, the specific register is determined by syntactic position: constituents in the subject position (SP) consistently occupy a higher pitch range than those in the object position (OP). This suggests that in Mandarin, the “large wave” is not a linear decay but a hierarchically organized structure where the SVO framework provides the primary scaffolding for pitch resets.

Second, internal Relative Clause (RC) types significantly modulate local pitch trajectories. A clear contrast exists between the falling contour of Subject-Relative (SR) clauses (V-high to N-low) and the rising contour of Object-Relative (OR) clauses (N-low to V-high). The statistical divergence observed in the first disyllabic word of the RC indicates that the onset of a modifier is a critical site for encoding syntactic distinctions through prosodic marking.

Finally, the observation that verbs (V) generally exhibit a higher pitch than nouns (N) within these RCs raises a crucial question: is this a fixed property of the word class (lexical category), or is it a byproduct of the information structure (IS) within the clause? In the RCs tested (e.g., qīngxīn Jīn Xīn ‘adore Jin Xin’), the verb often acts as the “new” relational information, while the noun may be perceived as a secondary argument. To disentangle the effects of word class from those of information structure, we proceed to Experiment 2, which systematically tests these categories across a wider variety of sentential and pragmatic contexts.

3. Experiment 2: Word Class and Information Structure

3.1. Motivation and Theoretical Background

The findings from Experiment 1 indicated that verbs (V) consistently exhibited a higher pitch than nouns (N) within relative clause structures. However, this result appears to contradict previous findings by Li et al. (2011), who reported that nouns carried a higher pitch than verbs. To resolve this discrepancy, it is necessary to scrutinize the experimental designs of both studies and identify the underlying factors responsible for these conflicting outcomes.

In the study by Li et al. (2011, p. 389), the researchers utilized pairs of homophonous pseudo-words (e.g., N-chásòng 茶宋 vs. V-chásòng 查送) presented in isolation and in sentence-final positions within carrier phrases such as “My [N] ~ I also [V]” (Wǒ de N ~ Wǒ yě V), “This [N] ~ Don’t you [V]” (Zhège N ~ Nǐ bié V), and “A [N] ~ Cannot [V]” (Yīge N ~ Bùnéng V). We identify a potential methodological limitation in their carrier phrase design: the syllables immediately preceding the target N and V were phonologically inconsistent. Specifically, the targets were preceded by various tones—Neutral (de), Tone 3 (yě), Tone 4 (ge), and Tone 2 (bié/néng). Given that preceding tones can trigger significantly different pitch transitions and carryover effects (Y. Xu, 1997), these inconsistencies likely confounded the pitch comparison between the target nouns and verbs.

Consequently, Experiment 2 adopts and refines the methodology of Li et al. (2011). While we retain the use of homophonous pseudo-words to eliminate lexical frequency effects, we introduce stricter controls over the phonological environment. By ensuring consistent tonal contexts and systematically placing the target words in various sentential positions (initial, medial, and final), Experiment 2 aims to provide a more comprehensive account of how syntactic position and information structure influence the “big wave” of intonation.

3.2. Stimuli Design and Experimental Procedure

In Experiment 2, we utilized a pseudo-word design to eliminate potential confounding effects from word frequency. Five pairs of homophonous, heterographic pseudo-words were constructed, all consisting of two high-level (Tone 1) syllables to facilitate the observation of intonational “big waves”. The pairs included: jiājiē (V 加接 vs. N 家街), bōshān (V 拨搧 vs. N 波山), shāngshū (V 伤输 vs. N 商书), shīdēng (V 失登 vs. N 诗灯), and guāxī (V 刮吸 vs. N 瓜溪). Each pseudo-word pair consisted of two distinct orthographic forms that were homophonous but visually distinguishable. The verb meaning was cued by morphemes with action semantics (e.g., 加 ‘add’, 拨 ‘flick’, 伤 ‘injure’, 失 ‘lose’, 刮 ‘scrape’), while the noun meaning was cued by morphemes referring to entities or places (e.g., 家 ‘home’, 波 ‘wave’, 商 ‘commerce’, 诗 ‘poetry’, 瓜 ‘melon’). Native speakers reliably interpreted the intended word class based on these orthographic cues, as confirmed by a post-experiment debriefing.

To systematically examine the influence of syntactic position and context, four experimental conditions were designed:

(1)

Isolation (ISO): The target word was read as a standalone item.

(2)

Sentence-final position (FIN): “I also [V]” (Wǒ yě V) vs. “I have [N]” (Wǒ yǒu N).

(3)

Sentence-initial position (INI): “[V]-ing the leaves is much better” (V le shùyè hǎoduō le) vs. “The leaves of [N] are much better” (N de shùyè hǎoduō le).

(4)

Sentence-medial position (MED): “He wants to [V] leaves” (Tā yào V shùyè) vs. “He wants [N] leaves” (Tā yào N shùyè). In this condition, the verb and noun conditions were presented as separate blocks, with written instructions explicitly indicating the intended interpretation. For the verb condition, the instruction read: ‘He wants to PERFORM AN ACTION on leaves’; for the noun condition, ‘He wants the TYPE OF leaves named ‘pseudo−word’. Additionally, the distinct orthographic forms of the pseudo-words provided visual disambiguation. In a post-experiment debriefing, all participants confirmed that they understood the intended distinction.

Ten female native Mandarin speakers from the author’s university, all born and raised in Beijing, participated in this experiment. Each target word and sentence was produced twice by each participant, with both tokens included in the subsequent data analysis. All other experimental conditions, recording specifications, and acoustic measurement methodologies remained identical to those described in Experiment 1.

3.3. Results and Discussion

Following the data processing protocol established in Experiment 1, we utilized SSANOVA to test for statistically significant differences between verbs (V) and nouns (N) across the four experimental contexts. The SSANOVA visualizations are presented in Figure 5, where purple-blue bands represent verb data and light-red bands represent noun data.

The results indicate that in the Isolation (ISO), Sentence-Initial (INI), and Sentence-Medial (MED) contexts, the confidence intervals for verbs and nouns exhibit substantial overlap, suggesting no statistically significant difference in pitch between the two word classes. Notably, in the isolation and sentence-initial conditions, the first syllable of the verb appears marginally higher than that of the noun, though this does not reach statistical significance. In contrast, in the Sentence-Final (FIN) position, the noun is markedly and significantly higher than the verb, with no overlap between their respective standard error bands.

The findings of Experiment 2 are consistent with the results reported by Li et al. (2011) in two respects: first, in isolation, verbs and nouns show similar pitch levels with no significant divergence; second, in the sentence-final position of a carrier phrase, nouns exhibit higher pitch than verbs. Furthermore, our results extend these findings by demonstrating that N and V also show no significant differences in sentence-initial or sentence-medial positions. Synthesizing the results from Experiment 1, Experiment 2, and Li et al. (2011), it can be inferred that word class (lexical category) does not inherently dictate pitch directionality. Instead, three distinct logical possibilities emerge depending on the context:

• N > V: As seen in the sentence-final context of Experiment 2 and Li et al. (2011).
• V > N: As observed in the relative clause structures of Experiment 1.
• V ≈ N: As demonstrated in the isolation, initial, and medial contexts of Experiment 2.

The variation in results across different contexts in Experiment 2 suggests that pitch realization is heavily influenced by Information Structure (IS). As a component of grammar, information structure reflects the status of information shared between interlocutors and is closely related to referential form, morphology, word order, and prosody (Arnold et al., 2013). In particular, the distinction between new and given (old) information significantly impacts prosody; new information naturally serves as the focus, receiving prosodic prominence characterized by an expanded pitch range, increased duration, and higher intensity.

The significant pitch gap between N and V in the sentence-final position likely stems from the semantic information structure of the carrier phrases. In the phrase “Wǒ yě V” (I also V), the function word “yě” (also) indicates that the verb represents given or “old” information. Conversely, in “Wǒ yǒu N” (I have N), the “N” serves as the informational focus (“new” information). Consequently, the noun receives prosodic focus and exhibits a significantly higher pitch than the “old” information verb. Similarly, in the “My N ~ I also V” contrast used by Li et al. (2011), the N serves as the focus of new information while the V is treated as old information, leading to the same N > V pitch pattern. In the other three conditions (isolation, initial, and medial), the informational status of N and V is balanced, resulting in no significant pitch divergence.

4. General Discussions

By synthesizing the results of Experiment 1 and Experiment 2, we can now address our initial research questions through the following points of discussion:

First, this study demonstrates that the syntactic structure of Mandarin Chinese significantly modulates intonational pitch movement. In Experiment 1, regardless of sentential position (subject vs. object), the target regions containing Subject-Relative (SR) and Object-Relative (OR) clauses exhibited non-negligible pitch differences. Meanwhile, the segments where the grammatical structures remained identical showed nearly overlapping contours. This provides empirical evidence that the observed pitch variance is directly attributable to the internal syntactic configuration of the relative clauses.

Second, universal pitch declination interacts with Mandarin word order and specific grammatical structures to produce predictable prosodic patterns. As shown in Figure 1, constituents in the subject position generally exhibit higher pitch than those in the object position. Since Mandarin is an SVO language, the subject typically occupies an earlier linear position in the utterance; under the influence of global declination, it naturally receives a higher pitch register than the subsequent object.

However, our data also reveals that declination is not an absolute rule. Exceptions occur, such as in the final disyllabic words of all sentence types in Experiment 1 (Figure 1) and across the initial, medial, and final positions in Experiment 2 (Figure 5), where the second syllable consistently exhibits a higher pitch than the first. While previous literature on rising intonation often focuses on interrogative boundary tones or phrasal pitch resets, these explanations do not fully account for the current findings. These phenomena warrant further investigation into the deeper underlying mechanisms of Mandarin prosody.

Third, the influence of syntax on intonation is a multi-layered phenomenon that must be analyzed across different prosodic domains. Our results indicate that relative clause type dictates specific pitch trajectories: Subject-Relative (SR) clauses follow a downward trend, whereas Object-Relative (OR) clauses follow an upward trend. Furthermore, the relative pitch of verbs (V) and nouns (N) lacks an inherent lexical bias; rather, it is context-dependent.

Following the hierarchical models proposed by Wang et al. (2013) and Feng (2016), we can explain these findings by looking at how different levels of “stress” or “prominence” apply to Mandarin sentences:

(1)

The verb phrase level (NSR core stress): Within a verb phrase (VP), the “Nuclear Stress Rule” (NSR) typically applies. In our study, the V + O combination in SR clauses (e.g., qīngxīn Jīn Xīn ‘adore Jin Xin’) functions as a standard verb phrase. This domain follows the universal phonological default of pitch declination, resulting in a clear downward-sloping “big wave” from the verb to the noun.

(2)

The tense phrase level (focus/information stress): Conversely, the connection between a Subject and a Verb in an OR clause (e.g., Jīn Xīn qīngxīn ‘Jin Xin adores’) represents a higher-level syntactic boundary (the tense phrase or TP level). At this level, information structure becomes the dominant factor. The subject is often treated as “given” information, while the verb acts as the informational focus. This triggers a prosodic pitch reset, where the speaker raises the pitch on the verb to emphasize it, resulting in the characteristic rising trajectory observed in our OR data.

(3)

The sentential level (intonational stress): Finally, the entire sentence is governed by the “intonational trunk”. Our experiments show that local word-class distinctions (lexical register) are always subordinate to these higher-level sentential and phrasal registers.

Finally, the theoretical significance of these multi-layered interactions must be interpreted through the lens of the Principle of F₀ Monofunctionality (Feng, 2015). Precisely because Mandarin, as a tonal language, cannot freely manipulate pitch for intonational purposes without compromising lexical tone distinctions (i.e., F₀ cannot encode both a lexical tone target and an independent intonational target simultaneously), this principle forces grammatical conflicts—such as the one we observe between the downward pressure of the Nuclear Stress Rule (NSR) in SR clauses and the upward pitch reset driven by information focus in OR clauses—to be resolved locally and systematically on the pitch contour. This makes the SR/OR contrast not merely another example of syntax-prosody interaction, but a distinctive diagnostic of how a tonal language achieves prosodic expressiveness under severe phonetic constraints. The true contribution of our study, therefore, lies in the mechanism of this encoding. We demonstrate not just that a tone-constrained system expresses grammatical meaning through pitch, but precisely how it does so. Ultimately, this system achieves expressiveness by forcing hierarchical conflicts onto local pitch trajectories. In summary, and consistent with the above theoretical constraint, the divergent slopes of SR and OR clauses are not random. They reflect a conflict between two different prosodic rules: SR clauses follow the downward pressure of the Nuclear Stress Rule within the verb phrase, while OR clauses reflect an upward pitch reset driven by informational focus at the sentence level. This explains why syllables with the same lexical tone can exhibit completely opposite pitch movements depending on their position in the grammatical hierarchy.

Several limitations of the present study should be acknowledged. First, as noted in the Introduction, the exclusive use of Tone 1 (high-level) stimuli, while methodologically necessary for isolating intonation from lexical tone variation, means that the observed patterns may not fully generalize to contour tones. However, rather than treating this solely as a limitation, we emphasize it as a clear direction for future inquiry. We strongly advocate for systematic replication of Experiments 1 and 2 using Tone 2 (rising) and Tone 4 (falling) stimuli to determine whether the observed SR-falling vs. OR-rising contrast is attenuated, enhanced, or qualitatively altered when the lexical “small ripples” themselves carry contour information. Such research would directly test the boundary conditions of the Principle of F₀ Monofunctionality and further reveal how contour tones align with—or resist—the intonational “big waves” documented here. Tone 3 (low-dipping) may also be of interest, though its complex contour and sandhi behavior would require even more careful experimental design.

Second, the stimuli in Experiment 1, while natural to our participants, are relatively complex and may not represent the full range of naturally occurring sentence structures. Replication studies using naturally occurring corpus data or more varied stimulus sets would strengthen the generalizability of the findings.

Third, the Principle of F₀ Monofunctionality (Feng, 2015) suggests that intonational “big waves” cannot be realized as freely in tonal languages as in non-tonal languages. The present results demonstrate that syntactic and information-structural modulations of pitch do occur within the constraints of the tonal system, but the precise boundaries of these constraints remain to be mapped. Future work using direct comparison of the same syntactic structures across tonal and non-tonal languages (e.g., Mandarin vs. English) would help quantify the degree to which the “big wave” is curtailed by tonal requirements.

Fourth, the present study focused exclusively on pitch as the primary acoustic correlate of intonation. Duration and intensity were not systematically analyzed, as our research scope was delimited to pitch. Consequently, we cannot report any patterns for these parameters. Future research should investigate whether duration and intensity covary with the syntactic manipulations observed here.

5. Conclusions

This study provides a systematic, acoustic investigation into how syntactic architecture and information structure modulate the intonational “big wave” in Mandarin Chinese. By isolating intonational pitch from lexical tones using controlled Tone 1 stimuli, several key conclusions emerge:

First, syntactic position significantly influences pitch register. In accordance with Mandarin’s SVO word order, constituents in the subject position consistently exhibit higher pitch than those in the object position. This suggests that the universal phenomenon of pitch declination is not a linear decay but is closely coupled with the linear arrangement of grammatical constituents.

Second, internal clause structure dictates specific pitch trajectories. Subject-relative (SR) clauses exhibit a falling pitch tendency, while object-relative (OR) clauses show a rising trend. These divergent contours can be explained through a syntactic-prosodic hierarchy:

(1)

SR clauses (verb + object structure) follow the Nuclear Stress Rule (NSR) within the verb phrase, adhering to the default downward pressure of declination.

(2)

OR clauses (subject + verb structure) involve higher-level sentential boundaries (the Tense Phrase level) where the verb often serves as the informational focus, triggering a prosodic pitch reset and an upward-sloping contour.

Third, word class does not inherently possess pitch directionality. The relative pitch of verbs and nouns is a byproduct of their position within the syntactic and Information Structure (IS). When a noun serves as “new” information (e.g., in a sentence-final position), it receives prosodic focus and higher pitch; when a verb is part of the informational focus, it exhibits similar prominence. In neutral contexts, no significant pitch difference exists between the two categories.

Importantly, the observed intonational modulations operate within the constraints imposed by the Principle of F₀ Monofunctionality (Feng, 2015). Unlike non-tonal languages, where intonation can freely modulate pitch across an utterance, Mandarin intonation must co-exist with lexically specified tonal targets. The fact that we observed systematic pitch variation despite all target syllables bearing the same high-level Tone 1 suggests that the “big wave” is real, but its amplitude and shape are likely constrained when contour tones are involved. Thus, while syntax and information structure actively shape pitch movement, these effects are filtered through the tonal phonological system.

In conclusion, this research confirms that Mandarin intonation is a multi-layered system where syntactic levels, prosodic hierarchies, and semantic information structure interact dynamically. These findings underscore the importance of integrating grammatical analysis into phonetic research to fully capture the complexity of tonal languages.