Prosody as a Constructional Constraint in Four-Syllable Chinese Idioms

1960s–1990s functionalist scholarship treated “information” variably:
- Halliday 1966: equated information structure with grammatical (metrical) levels.
- Chafe 1994 & contemporaries: highlighted activation states (given/new, focus, etc.) as information architecture.
- Functionalists repeatedly note that prosody, syntax and semantics co-organise discourse, yet cognitive-linguistic frameworks rarely formalised prosody.
Construction Grammar (CxG)
- Emerged mid-1990s (Goldberg 1995, Fillmore & Kay 1999) viewing a CONSTRUCTION as a form–meaning pairing.
- Early CxG work focused almost exclusively on syntactic & morphological shape, with prosody relegated to “paralinguistic”.
- Recent calls (e.g.
- Chen & Pan 2015
- Valle & Notboom 2011) argue prosody should be modelled as a constraint inside the construction, not a parallel tier.

Can tone be treated as a bona-fide constructional constraint rather than an external performance layer?
Empirical focus: four-syllable Chinese idiomatic constructions (成语 chéngyǔ) with spatial/metaphorical semantics.
Key questions:
1. What tone-sequence patterns exist and how often do they occur?
2. Are certain lexical slots limited to particular tones?
3. How do prosodic contours correlate with directional / contrastive / metaphorical meaning types?
4. What theoretical consequences for CxG and cognitive linguistics follow if tone patterns are grammatically encoded?

Mandarin inventory (simplified):
1. Tone 1 (T1) – high level: e.g. “mā” (妈) ‘mother’ (symbolised 1)
2. Tone 2 (T2) – rising: “má” (麻) ‘hemp’ (symbolised 2)
3. Tone 3 (T3) – low dipping: “mǎ” (马) ‘horse’ (symbolised 3)
4. Tone 4 (T4) – sharp falling: “mà” (骂) ‘scold’ (symbolised 4)
Same segmental string + different tone ⇒ different lexical meanings (classic example 马 / 妈 / 麻 / 骂).

Structure: four Chinese characters, each carrying one of the four tones.
Theoretically possible tone combinations:
- $4^4 = 256$ patterns (T1–T4 in each slot)
Empirical observation: distribution is highly skewed; many patterns unattested.
Example: 东来西往 “coming from the east, going to the west” → Tone pattern 1-2-1-4; contour mirrors imagined trajectory (flat → rise → flat → fall).

Corpus: 1 703 idioms explicitly or implicitly encoding spatial/metaphorical motion.
- Explicit markers: characters meaning 南 ‘south’, 上 ‘up’, 左 ‘left’, etc.
- Implicit: idioms whose conventional meaning involves motion or spatial contrast.
Procedure:
1. Manual annotation of each syllable’s tone (1–4 numeric).
2. Extraction of global tone patterns and slot-specific frequencies.
3. Visualisation via heatmaps (5 slides in talk) → darker = higher corpus frequency.

Global frequency hot-spots (overall pattern level):
- High-frequency clusters around patterns ending in double 4 (–4-4) or double 3 (–3-3).
- Top patterns: 1-2-4-4, 1-1-2-4, 2-2-2-4.
Final rhyme preference:
- Syllables C & D show strong tendency for identical tones – esp. 4-4.
- 4-4 provides an acoustic “full stop”: throat closure + falling contour ⇒ rhetorical finality.
- Similarly frequent are 3-3, 2-2, 1-1 in final slot pair.
Metric regularity: Repetition of tones across all four slots common:
- Identical sequence examples: 1-1-1-4, 4-4-4-4 (11 tokens), 1-1-1-1 (rare but attested).
- Used for stylistic, emphatic, or nominal effects.
Zero-occurrence holes (patterns never found): examples include 1-3-0-3, 3-1-3-3, 4-3-4-3 → suggests grammatical ill-formedness, not mere sampling gap.

Idioms denoting movement / trajectory skew toward patterns starting with T1 or T2 (level or rising):
- Example: 危在旦夕 T1-?-?-? ‘disaster at dawn/dusk’ (danger approaching) → level start.
- Night-come-go 类来去 (例): pattern also starts high flat.
Rising onset evokes initiation / departure; fall at end encodes arrival / endpoint.

Idioms like 左右为难 ‘caught between left & right’ favour a T1 → T4 trajectory (flat → fall) maintaining a singular prosodic direction.
Observed rise–fall asymmetry aligns with semantic contrast (positive vs negative, upper vs lower).

Tone is not arbitrary ornamentation; it constrains which lexical items may fill which constructional slot.
Evidence for prosody-syntax interface inside Construction Grammar:
- Slot fillers licensed jointly by segmental form AND tonal contour.
- Construction = {phonemes, tones, morphosyntax, semantics, discourse function}.
Supports idea of a “prosodic tier” embedded within the construction schema.

Language teaching: idioms could be taught via prosodic profiles as mnemonic cues.
Computational modelling / NLP:
- Predictive models can exploit tone-sequence probability to flag ill-formed idioms.
- Potential for automatic idiom generation respecting tonal constraints.
Cross-linguistic expansion: method extendable to tonal languages like Yoruba, Bemba.
Diachronic studies: track how tonal templates evolve over centuries.

Prosody is small but grammatical – must be included in formal description.
Tone patterns shape constructional meaning; certain semantics prefer specific contours.
Spatial & metaphorical meaning and formal tonal pattern are co-constrained.
Construction Grammar should explicitly integrate prosodic dimensions alongside morpho-syntax.
Opens avenues for multi-modal modelling: gesture + prosody + lexis in unified construction.

Apply framework to non-Mandarin tonal languages & multimodal (gesture) constructions.
Test computationally whether neural models can predict unattested vs. attested tone sequences.
Pedagogical experiments: Does teaching idioms with tonal profiles improve acquisition?
Diachronic corpus work: When did clusters like –4-4 become dominant?

Many insights expanded in author’s (German-language) monograph (QR code in slides).
For collaboration inquiries, contact via email or QR code shared at talk.