Maddieson 2013: WALS phonology chapters ↔ PHOIBLE 2.0 bridges #
@cite{maddieson-2013} @cite{moran-mccloy-2019} @cite{dryer-haspelmath-2013}
Cross-dataset bridge theorems verifying that PHOIBLE 2.0 inventories
are consistent with the WALS Maddieson chapter classifications, for
the 16 languages currently covered by Fragments/{Lang}/Phonology.lean.
Inventories are accessed via Fragments (not Data directly), per linglib's pattern that per-language data flows through Fragment files.
Why this study file exists #
The dissolved Phenomena/Phonology/Typology.lean carried a 16-language
hand-curated PhonProfile sample with ~80 xxx_chN grounding theorems
testing each language against WALS — vacuous round-trips (each value
matches its own WALS lookup). Dropped in the dissolution.
This file replaces the bridge work with substantive WALS↔PHOIBLE cross-dataset verification: each language's PHOIBLE consonant count is checked against the WALS Maddieson Ch 1 bin it falls into. Where PHOIBLE and WALS disagree at the bin level, the disagreement is documented (PHOIBLE often counts more segments under finer transcription).
Maddieson's WALS Ch 1 bin definitions #
WALS Ch 1 (consonant inventory size) bins, per the chapter text:
small: 6–14 consonantsmoderatelySmall: 15–18average: 19–25moderatelyLarge: 26–33large: 34+
Note: these bins come from Maddieson's UPSID counts. PHOIBLE 2.0 has finer transcription and broader source coverage, so per-language counts can differ even when both are well-formed inventories.
Featural double-coding (deferred) #
WALS Ch 7 (glottalized consonants) could derive from a PHOIBLE inventory
- a presence-of-
[+constricted glottis]predicate (Linglib/Theories/Phonology/Featural/Features.lean). Same for Ch 8 ([+lateral]), Ch 9 ([+nasal, +dorsal]), Ch 11 ([+front, +round]). Bridge theorems of that form are deferred — would require a uniform "inventory has any phoneme with feature F" helper.
Whether a consonant count falls in the WALS Ch 1 bin for a given
CInventorySize. Uses Maddieson's bin boundaries.
Returns Bool; decide resolves the comparisons concretely.
Equations
- Phenomena.Phonology.Studies.Maddieson2013.inWALS1ABin Typology.Phonology.CInventorySize.small n = decide (6 ≤ n ∧ n ≤ 14)
- Phenomena.Phonology.Studies.Maddieson2013.inWALS1ABin Typology.Phonology.CInventorySize.moderatelySmall n = decide (15 ≤ n ∧ n ≤ 18)
- Phenomena.Phonology.Studies.Maddieson2013.inWALS1ABin Typology.Phonology.CInventorySize.average n = decide (19 ≤ n ∧ n ≤ 25)
- Phenomena.Phonology.Studies.Maddieson2013.inWALS1ABin Typology.Phonology.CInventorySize.moderatelyLarge n = decide (26 ≤ n ∧ n ≤ 33)
- Phenomena.Phonology.Studies.Maddieson2013.inWALS1ABin Typology.Phonology.CInventorySize.large n = decide (34 ≤ n)
Instances For
Hindi: PHOIBLE's "hind1269" Hindi-Urdu has 71 consonants (the
breathy-aspirated × voiced/voiceless × place series produces a very
large inventory) — falls in WALS large (34+).
Yoruba: 18 consonants — exceptionally small, falls in WALS
moderatelySmall (15–18).
Maori: 10 consonants — exceptionally small (Polynesian-typical).
WALS small (6–14).
Zulu: 35 consonants (clicks + ejectives + breathy series) puts it
in WALS large (34+).
Of the 16 PHOIBLE-canonical inventories: 1 small, 1 moderatelySmall,
6 average, 5 moderatelyLarge, 3 large. The 0 moderatelySmall
finding inverts WALS Maddieson's modal-average count for the same
16 ISOs — PHOIBLE's finer transcription pulls several
moderatelySmall-in-WALS languages up into average.