@cite{futrell-gibson-2020}: Crosslinguistic Dependency Length Data

@cite{futrell-gibson-2020} @cite{zaslavsky-hu-levy-2020}

Empirical data from Table 2 of @cite{futrell-gibson-2020} "Dependency locality as an explanatory principle for word order", Language 96(2):371–412.

53 languages from Universal Dependencies corpora, measuring:

• Proportion of head-final dependencies
• Mean dependency length at sentence lengths 10, 15, 20

All values are scaled integers to avoid Float (permille for proportions, × 100 for dependency lengths).

Key Empirical Finding

Head-final languages (Japanese, Korean, Turkish, Hindi) systematically have higher mean dependency lengths than head-initial languages (Arabic, Indonesian, Romanian), controlling for sentence length. This is predicted by DLM theory: head-final order with right-branching structures creates longer dependencies.

structure Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.LanguageDLM : Type

Crosslinguistic dependency length data for a single language.

Values are scaled integers:

• propHeadFinal: × 1000 (permille), e.g., 890 = 89.0% head-final
• depLengthAt10/15/20: × 100, e.g., 245 = 2.45 mean dep length

• name : String
• isoCode : String
• family : String
• propHeadFinal : Nat
• depLengthAt10 : Nat
• depLengthAt15 : Nat
• depLengthAt20 : Nat

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instReprLanguageDLM.repr : LanguageDLM → Nat → Std.Format

Equations

• One or more equations did not get rendered due to their size.

instance Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instReprLanguageDLM : Repr LanguageDLM

Equations

• Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instReprLanguageDLM = { reprPrec := Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instReprLanguageDLM.repr }

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instDecidableEqLanguageDLM.decEq (x✝ x✝¹ : LanguageDLM) : Decidable (x✝ = x✝¹)

Equations

• One or more equations did not get rendered due to their size.

instance Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instDecidableEqLanguageDLM : DecidableEq LanguageDLM

Equations

• Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instDecidableEqLanguageDLM = Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instDecidableEqLanguageDLM.decEq

instance Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instBEqLanguageDLM : BEq LanguageDLM

Equations

• Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instBEqLanguageDLM = { beq := Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instBEqLanguageDLM.beq }

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instBEqLanguageDLM.beq : LanguageDLM → LanguageDLM → Bool

Equations

• One or more equations did not get rendered due to their size.
• Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.instBEqLanguageDLM.beq x✝¹ x✝ = false

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.LanguageDLM.isHeadFinal (l : LanguageDLM) : Bool

A language is predominantly head-final if > 50% of deps are head-final.

Equations

• l.isHeadFinal = decide (l.propHeadFinal > 500)

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.LanguageDLM.isHeadInitial (l : LanguageDLM) : Bool

A language is predominantly head-initial if ≤ 50% of deps are head-final.

Equations

• l.isHeadInitial = decide (l.propHeadFinal ≤ 500)

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.arabic : LanguageDLM

Arabic (afro-asiatic, head-initial, VSO/SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.basque : LanguageDLM

Basque (isolate, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.bulgarian : LanguageDLM

Bulgarian (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.chinese : LanguageDLM

Chinese (Sino-Tibetan, mixed)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.czech : LanguageDLM

Czech (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.danish : LanguageDLM

Danish (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.dutch : LanguageDLM

Dutch (Indo-European, mixed V2)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.english : LanguageDLM

English (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.estonian : LanguageDLM

Estonian (Uralic, mixed)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.finnish : LanguageDLM

Finnish (Uralic, head-final, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.french : LanguageDLM

French (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.german : LanguageDLM

German (Indo-European, mixed V2/SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.greek : LanguageDLM

Greek (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.hebrew : LanguageDLM

Hebrew (Afro-Asiatic, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.hindi : LanguageDLM

Hindi (Indo-European, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.hungarian : LanguageDLM

Hungarian (Uralic, head-final)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.indonesian : LanguageDLM

Indonesian (Austronesian, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.italian : LanguageDLM

Italian (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.japanese : LanguageDLM

Japanese (Japonic, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.korean : LanguageDLM

Korean (Koreanic, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.latin : LanguageDLM

Latin (Indo-European, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.norwegian : LanguageDLM

Norwegian (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.persian : LanguageDLM

Persian (Indo-European, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.polish : LanguageDLM

Polish (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.portuguese : LanguageDLM

Portuguese (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.romanian : LanguageDLM

Romanian (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.russian : LanguageDLM

Russian (Indo-European, mixed)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.spanish : LanguageDLM

Spanish (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.swedish : LanguageDLM

Swedish (Indo-European, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.tamil : LanguageDLM

Tamil (Dravidian, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.turkish : LanguageDLM

Turkish (Turkic, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.urdu : LanguageDLM

Urdu (Indo-European, head-final, SOV)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.vietnamese : LanguageDLM

Vietnamese (Austroasiatic, head-initial, SVO)

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.languages : List LanguageDLM

Representative subset of 32 languages from Table 2.

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.headFinalLanguages : List LanguageDLM

Head-final languages in the dataset.

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.headInitialLanguages : List LanguageDLM

Head-initial languages in the dataset.

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.meanDepLengthHF10 : Nat

Mean dep length at length 10 for head-final subset.

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.meanDepLengthHI10 : Nat

Mean dep length at length 10 for head-initial subset.

Equations

• One or more equations did not get rendered due to their size.

theorem Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.headFinal_higher_depLength_10 : meanDepLengthHF10 > meanDepLengthHI10

Head-final languages have higher mean dep length at sentence length 10.

This is the core empirical finding: head-final languages systematically exhibit longer dependencies, consistent with DLM theory's prediction that consistently head-final order creates longer dependencies when combined with right-branching structure.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.meanDepLengthHF20 : Nat

Same pattern at sentence length 20.

Equations

• One or more equations did not get rendered due to their size.

def Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.meanDepLengthHI20 : Nat

Equations

• One or more equations did not get rendered due to their size.

theorem Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.headFinal_higher_depLength_20 : meanDepLengthHF20 > meanDepLengthHI20

theorem Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.japanese_highest_depLength_20 : (languages.all fun (x : LanguageDLM) => decide (x.depLengthAt20 ≤ japanese.depLengthAt20)) = true

Japanese has the highest dep length at length 20 among all languages.

theorem Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.indonesian_lowest_depLength_10 : (languages.all fun (x : LanguageDLM) => decide (x.depLengthAt10 ≥ indonesian.depLengthAt10)) = true

Indonesian has the lowest dep length at length 10 among all languages.

theorem Phenomena.WordOrder.DependencyLength.FutrellEtAl2020.headFinality_gap_increases : meanDepLengthHF20 - meanDepLengthHI20 > meanDepLengthHF10 - meanDepLengthHI10

The head-finality gap increases with sentence length: the difference in mean dep length between head-final and head-initial languages is larger at length 20 than at length 10.