Meta-Analysis: Changes Across Timestamped Tanka Reports

Author: ML and Cursor AI

Reports compared:
- 14 tanka (2025-03-13T19:00) — pages 27–33
- 18 tanka (2025-03-13T19:30) — pages 27–35
- 22 tanka (2025-03-13T19:54) — pages 27–37
- 50 tanka (2025-03-13T20:19) — pages 27–51

Question: How do form, function, and content change as the corpus grows?


1. Form: Structural Stability and Drift

1.1 Mora pattern (5-7-5-7-7)

Line 14 tanka 18 tanka 22 tanka 50 tanka Trend
1 4.5 4.4 4.5 4.8 Stable
2 6.6 6.7 6.6 6.5 Stable
3 4.7 4.8 4.9 4.8 Stable
4 6.7 6.7 6.7 6.5 Stable
5 6.1 6.2 6.3 6.3 Stable

Finding: The 5-7-5-7-7 pattern is stable across corpus size. Lines 2, 4, and 5 remain consistently longer than 1 and 3. The 50-tanka corpus shows line 1 slightly longer (4.8) and lines 2, 4 slightly shorter (6.5), but the overall pattern holds. Conclusion: Form is not sensitive to corpus size—the poems are structurally uniform across 14→50 tanka.

1.2 Line length (character count)

Chars 14 tanka 18 tanka 22 tanka 50 tanka Trend
5 18 23 27 67 Scales
6 21 27 35 66 Scales
4 15 17 21 57 Scales
7 8 12 15 38 Scales
3 5 8 9 17 Scales

Average: 5.26 → 5.27 → 5.29 → 5.24 characters per line.

Finding: The distribution scales proportionally—more lines are added in each category. The 5–6 character band remains dominant. The 50-tanka average (5.24) dips slightly, with more 4-char lines (57) relative to 5–6. Conclusion: No structural shift; new poems follow the same line-length distribution.

1.3 Full-poem length

Metric 14 tanka 18 tanka 22 tanka 50 tanka
Range 24–31 24–31 24–31 23–31
Mean ~26 ~26 ~26.5 ~26

Finding: The range expands slightly at 50 tanka (23–31); the mean holds around 26. Poems 31, 47–50 introduce 23–24 char minimum. Conclusion: Poem length remains largely invariant; the 50-tanka corpus adds a few shorter poems.


2. Function: Grammatical and Rhetorical Patterns

2.1 Line-ending particles (の)

Corpus の count の per line (avg)
14 tanka 9 0.129
18 tanka 11 0.122
22 tanka 16 0.145
50 tanka 40 0.160

Finding: The particle の increases monotonically in both count and per-line density. The 50-tanka corpus has the highest density (16% of lines end in の). Conclusion: The possessive/nominal chain structure (の) is strongly reinforced as the corpus grows to 50 tanka.

2.2 Other particles

Ending 14 tanka 18 tanka 22 tanka 50 tanka Trend
8 8 9 21 Rise
7 8 8 16 Rise
5 5 6 12 Rise
4 5 6 10 Rise
2 3 5 8 Rise
2 3 4 7 Rise
1 3 6 Rise
3 6 Rise
4 5 5 13 Rise

Finding: All major particles scale with corpus size. (13× at 50 tanka) rises sharply—more prohibitives and exclamations (春よ老いな, さて責むな). Conclusion: Rhetorical function continues to shift toward imperative, vocative, and prohibitive as the corpus grows to 50 tanka.

2.3 Recurring two-character endings

Phrase 14 tanka 18 tanka 22 tanka 50 tanka Trend
きし 3 3 3 4 Stable/rise
しき 4 Emerges at 50
かな 2 2 2 3 Stable
の神 2 2 3 Rise
まへ 2 3 Rise
かむ 2 3 Rise
神の 2 3 Rise
髪の 2 2 2 2 Stable
春の 2 At 50

Finding: しき (4× at 50) emerges with うつくしき (beautiful)—a recurring aesthetic judgment. の神, まへ, かむ, 神の rise to 3. Conclusion: Function continues toward sacred address and petition; aesthetic judgment (うつくしき) becomes a distinct pattern at 50 tanka.

2.4 Recurring phrases (2–4 characters)

Phrase 14 tanka 18 tanka 22 tanka 50 tanka Trend
春の 5 6 7 10 Monotonic rise
夜の 2 4 5 11 Surpasses 春の at 50
たま / たまへ 3 / 2 4 / 3 5 / 4 7 / 5 Rise
わが 4 4 5 Rise
の神 3 4 6 Rise
の御 3 4 Rise
うつくし 5 Emerges at 50

Finding: 夜の (11) surpasses 春の (10) at 50 tanka—night gains lexical dominance. うつくし (beautiful) emerges as a distinct phrase (5×). Conclusion: At 50 tanka, night overtakes spring as the top phrase; aesthetic judgment (うつくし) becomes a recurring function.


3. Content: Lexical and Thematic Evolution

3.1 Kanji frequency (top tier)

Kanji 14 tanka 18 tanka 22 tanka 50 tanka Trend
7 8 9 15 Monotonic rise
3 6 7 13 Rise
4 6 7 11 Rise
3 3 4 10 Rise
3 5 6 10 Rise
2 3 3 8 Rise
3 4 8 Rise
4 4 5 7 Rise
7 Emerges at 50
1 2 3 6 Rise
2 2 4 6 Rise
4 Emerges at 50
4 Emerges at 50

Finding: , , , continue to rise. (you/dear) emerges at 50 tanka (7×)—a major addressee. (rain), (summer) enter the lexicon. Conclusion: Content intensifies; second-person address (君) and seasonal/ elemental vocabulary (雨, 夏) expand at 50 tanka.

3.2 Thematic clusters: emergence and expansion

Theme 14 tanka 18 tanka 22 tanka 50 tanka Change
Spring (春) 7/14 8/18 9/22 15+/50 Proportion ~30%
Hair (髪) 4 6 7 11 Expansion
Night (夜) 3 6 7 13 夜の surpasses 春の
君 (you) 7 Emerges at 50
Sacred figures 3 5 6 10+ Diversify
雨 (rain) 4 Emerges at 50
夏 (summer) 4 Emerges at 50
うつくし (beautiful) 5 Emerges at 50
Place 1 4 8+ Expansion

Finding: (13) now rivals (15) in frequency; 夜の (11) surpasses 春の (10) as top phrase. (you) emerges as major addressee. , , うつくし enter. Place expands: 夜の舞殿, 泉の底, ゆふべの宮, 廬舎那仏. Conclusion: Content diversifies further; night gains lexical dominance; second-person and aesthetic registers intensify.

3.3 English word frequency (thematic proxy)

Word 14 tanka 18 tanka 22 tanka 50 tanka Trend
spring 6 7 8 15 Rise
hair 3 5 6 10 Rise
god 2 4 5 9 Rise
love 3 5 6 Rise
you 8 Emerges at 50
night 3 4 10 Rise
beautiful 2 5 Rise at 50

Finding: Spring, hair, god, love, night increase. You (8×) and beautiful (5×) emerge at 50 tanka. Conclusion: The thematic center intensifies; second-person and aesthetic registers rise.


4. Summary: Discernible Changes Across Corpus Size

Form (structure)

Aspect Change Interpretation
Mora pattern Stable 5-7-5-7-7 holds regardless of corpus size
Line length Stable Distribution scales proportionally
Poem length Stable 24–31 char range unchanged

Conclusion: Form is robust and invariant across 14→50 tanka. No structural drift.

Function (grammar, rhetoric)

Aspect Change Interpretation
の (possessive) Rises Density increases at 22 tanka
む (volitional) Rises More imperatives
よ (vocative) Rises More direct address
たまへ (please) Rises More petition
の神, 神の Emerges/rises Sacred address
の御 Emerges at 22 Honorific expansion
な (prohibitive) Rises to 13 at 50 More prohibitives/exclamations
うつくし (beautiful) Emerges at 50 Aesthetic judgment

Conclusion: Function shifts toward petition, direct address, sacred register, and aesthetic judgment. The speaker becomes more assertive; うつくし thematizes beauty.

Content (lexicon, themes)

Aspect Change Interpretation
Monotonic rise Spring remains dominant
夜, 髪, 神 Rise Sacred–erotic nexus strengthens
Rises at 22 Color/makeup intensifies
Emerges and rises Honorific body/place
Sacred figures Diversify Night god → autumn god → bodhisattvas
Place Expansion Kyoto → 奥の院 → 山ごもり → 室むつまじ
Emerges at 22 Seasonal balance
君 (you) Emerges at 50 Second-person addressee
雨, 夏 Emerge at 50 Rain, summer
夜の > 春の At 50 Night surpasses spring as top phrase

Conclusion: Content intensifies and diversifies. At 50 tanka: 夜の surpasses 春の; (you) emerges; , , うつくし enter. The honorific, sacred, and aesthetic registers expand.


5. Overall Interpretation

  1. Form is stable: The 5-7-5-7-7 pattern and line-length distribution do not change with corpus size across 14→50 tanka. The poems are structurally uniform.

  2. Function becomes more assertive: As the corpus grows, the speaker petitions more (たまへ), uses more imperatives (む), vocatives (よ), and prohibitives (な), and addresses sacred figures more directly. At 50 tanka, aesthetic judgment (うつくし) becomes a recurring function.

  3. Content intensifies and diversifies: Core themes (spring, hair, god) strengthen; new themes (autumn, bodhisattvas, rain, summer, 君) appear. At 50 tanka, 夜の surpasses 春の as the top phrase—night gains lexical dominance over spring.

  4. Lexical shift at 50 tanka: (you) emerges as a major addressee (7×); (rain), (summer) enter; うつくし (beautiful) appears 5×. The corpus gains second-person intimacy and seasonal/elemental range.

  5. The sequence drives the change: The added poems (23–50, pages 38–51) introduce 夜の舞殿, 春雨, 廬舎那仏, 君みずや, 二十の夏, 泉の底の小百合花. The shift reflects a narrative or thematic progression through the collection.

  6. Sample size: With 50 tanka (12.5% of the full 399), patterns are more robust. The 夜の > 春の shift and 君 emergence suggest genuine lexical evolution in the sequence rather than sampling noise.


Meta-analysis generated from comparison of tanka_analysis_report_2025-03-13T1900.md, T1930.md, T1954.md, and T2019.md.