Phon Tiers (%xmodsyl, %xphosyl, %xphoaln, %xphoint)

Status: Reference Last updated: 2026-06-23 07:28 EDT

The Phon extension tiers provide syllable-level phonological annotation, segmental alignment between target and actual IPA, and per-phone time intervals. They are produced by the Phon application and exported to CHAT via PhonTalk.

chatter parses and validates all four tiers as first-class CHAT tiers.

The x prefix. Phon emits these tiers with a leading x (%xmodsyl, %xphosyl, %xphoaln, %xphoint) to mark them as extension tiers. The grammar accepts both the x-prefixed names and the historical non-x names (%modsyl, %phosyl, %phoaln, %phoint); the parser and validator key off the tier kind, not the literal prefix. The canonical serialized form is the x-prefixed name.

The four tiers

%xmodsyl, %xphosyl, and %xphoaln are word-aligned to their source tier(s) with single ASCII spaces. %xphoint uses / (space-slash-space) as its word separator because single spaces already separate the phone and bullet tokens inside each word.

Tier formats

%xmodsyl / %xphosyl, syllabification

A word is one or more phone:CODE units concatenated with no internal whitespace; words are separated by single spaces. The phone is one IPA phone (IPA length is written with the modifier letter ː, U+02D0, never an ASCII colon, so the : separator is unambiguous). A leading stress marker (ˈ primary, ˌ secondary) is part of the phone it precedes.

The constituent code is one character. The legal codes are O N C L R E A D U:

The remaining Phon SyllableConstituentType mnemonics, B (boundary), S (stress), W (word boundary), T (tone), are not emitted on these tiers: boundary, stress, and tone need no per-phone marker.

*CHI:	I want three .
%mod:	aɪ wɑnt θri
%xmodsyl:	a:Dɪ:D w:Oɑ:Nn:Ct:C θ:Oɹ:Oi:N
%pho:	aɪ wɑn fwi
%xphosyl:	a:Dɪ:D w:Oɑ:Nn:C f:Ow:Oi:N

%xphoaln, phone alignment

A word is one or more comma-separated pairs; a pair is model↔actual (↔ is U+2194). Either side may be ∅ (U+2205, empty set): ∅ on the left is an epenthesis (a phone produced but not targeted); ∅ on the right is a deletion. Both sides are never ∅ at once.

*CHI:	the best .
%mod:	ðə bɛst
%pho:	ðə bɛs
%xphoaln:	ð↔ð,ə↔ə b↔b,ɛ↔ɛ,s↔s,t↔∅

The alignment lists segments (phones). Suprasegmental stress (ˈ/ˌ) that may appear on the %mod/%pho word is therefore not part of the alignment pairs; the reconstruction checks below compare modulo those stress markers.

%xphoint, per-phone intervals

%xphoint gives the time segmentation of each individual phone on %pho, effectively phone-level bullets analogous to the word-level timing on %wor. Groups (one per %pho word) are separated by /. Within a group, each phone is followed by a CLAN time-alignment bullet: the byte 0x15 (NAK), the interval start_end, then 0x15.

*CHI:	I want . •0_500•
%pho:	aɪ wɑnt
%xphoint:	aɪ •0_250• / w •250_320• ɑ •320_400• n •400_460• t •460_500•

(Bullets are shown as • above; in the file they are the 0x15 byte.)

Validation

These checks run by default. Pass --suppress xphon to silence the entire Phon %x validation surface, or suppress an individual code. (The historical --check-xphon opt-in flag is now a deprecated no-op: the checks it used to gate are on by default.)

Word-count cross-checks (each %x tier has the same number of words as the tier(s) it depends on):

• %xmodsyl ↔ %mod: E725
• %xphosyl ↔ %pho: E726
• %xphoaln ↔ %mod: E727, ↔ %pho: E728

Content checks:

See Alignment Architecture for the word-count implementation.

Parsing strategy

• %xmodsyl / %xphosyl: stored as flat word strings (talkbank-model::dependent_tier::phon::SylTier), consistent with how %pho and %mod store flat phone words. The validator tokenizes each word into typed phone:CODE units (PositionCode) to apply the content rules above; the IPA characters themselves stay verbatim for exact round-trip.
• %xphoaln: each word is parsed into a Vec<AlignmentPair>, where AlignmentPair { source, target } carries one model↔actual mapping (None is ∅).
• %xphoint: parsed into typed groups of (phone, bullet) pairs (XphointTier / XphointGroup / PhoneInterval), reusing the same 0x15 bullet machinery as %wor.

Deep phonological analysis is Phon’s domain; chatter parses the structure that validation needs and keeps the IPA content verbatim.

Phon XML source format

In Phon’s native XML format, phonological data is stored as structured elements:

<ipaTarget>
  <pho>
    <pw>
      <ph scType="onset"><base>θ</base></ph>
      <ph scType="nucleus"><base>ɹ</base></ph>
      <ph scType="nucleus"><base>i</base></ph>
    </pw>
  </pho>
</ipaTarget>

Each <pw> (phonological word) element contains <ph> elements with syllable constituent types (scType). The <alignment> element provides phone-level mappings between target and actual using index-based <pm> (phone map) entries.

Data quality notes

A small percentage of Phon corpus XML records have an orthography↔IPA word-count mismatch: the number of <pw> elements in <ipaTarget> / <ipaActual> differs from the number of <w> elements in <orthography>. This is expected in child phonology data: children may produce extra syllables, partial words, or over-productions relative to the target.

For current counts on a local CHILDES/TalkBank data tree, run:

python3 scripts/analysis/scan_phon_mismatches.py /path/to/data

The PhonTalk CHAT export handles this discrepancy inconsistently:

1. %mod/%pho are written through a OneToOne alignment path that maps IPA words to orthography words; extras are silently dropped.
2. %xmodsyl/%xphosyl/%xphoaln are written directly from the raw IPATranscript; all IPA words are included.

This produces CHAT files where %xmodsyl may have more words than %mod, triggering the E725-E728 word-count errors. This is being investigated in collaboration with the Phon team.