Scoring notes
Scoring the "th" sounds
- IPA
- /θ/ and /ð/
- ARPAbet
- TH, DH
- Manner
- interdental fricatives, voiceless and voiced
"Th" is not one sound but two separate phonemes that happen to share a spelling: the voiceless /θ/ in "think" and the voiced /ð/ in "this". Both are interdental fricatives made by placing the tongue between the teeth, both are among the last English sounds children acquire, and both are commonly replaced by /f/ and /v/ — a pattern called th-fronting that automatic phoneme scoring detects comparatively well.
Typical age of mastery
Developmental norms for the "th" sounds: commonly cited as mastered by around age 5 to 6, with /ð/ often later than /θ/.
Why this sound is hard
- The interdental placement is unusual. Most of the world's languages do not contain /θ/ or /ð/ at all, which is why they are also among the hardest sounds for adult English learners.
- The two sounds differ only in voicing, and the voiced /ð/ carries very little acoustic energy — it is short, quiet, and frequently reduced in connected speech, which makes it both harder to produce and harder to score.
- /θ/ and /f/ are acoustically similar and, in the case of /θ/ and /f/, visually confusable too. Th-fronting is so common in some English accents that it is a recognised dialect feature rather than a disorder in adult speakers of those varieties.
- English spelling gives no clue which "th" is which. "Thigh" and "thy" differ only in voicing, which matters for a scoring system because the pronunciation dictionary has to make the right choice before alignment can even begin.
Common errors
| Pattern | Example | Error type |
|---|---|---|
| Th-fronting | "think" → "fink" | substitution |
| Th-fronting (voiced) | "this" → "vis" | substitution |
| Stopping | "this" → "dis" | substitution |
| Th-alveolarisation | "think" → "sink" | substitution |
How ArticScore handles it
- Every common "th" error is a substitution with another phoneme in the inventory — /f/, /v/, /d/, /s/ — which makes "th" one of the better-behaved sounds for automatic scoring. ArticScore scores the target low and names the substitute.
- ArticScore treats /θ/ and /ð/ as the distinct phonemes they are, rather than collapsing them into a single "th", so an application can report that a child manages "think" but not "this".
- Th-fronting is one of the error patterns present in the child speakers of the corpus our benchmark is measured on, so it is among the patterns most directly represented in our published figures.
What the API returns
A single word from the response, showing the per-phoneme rows ArticScore produces for a target containing /th/.
{
"word": "think",
"score": 66.2,
"phonemes": [
{
"phone": "TH",
"phone_ipa": "\u03b8",
"score": 29.8,
"sound_most_like": "F",
"sound_most_like_ipa": "f",
"extent": [110, 235],
"flagged": true,
"error_type": "substitution",
"tier": "flag"
},
{ "phone": "IH", "phone_ipa": "\u026a", "score": 86.4, "extent": [235, 320], "flagged": false, "tier": "clean" },
{ "phone": "NG", "phone_ipa": "\u014b", "score": 90.1, "extent": [320, 415], "flagged": false, "tier": "clean" },
{ "phone": "K", "phone_ipa": "k", "score": 88.7, "extent": [415, 500], "flagged": false, "tier": "clean" }
]
}Scoring the "th" sounds — questions
Is 'th' one sound or two?
Two. English has a voiceless /θ/, as in "think" and "bath", and a voiced /ð/, as in "this" and "mother". They share a spelling but are separate phonemes, they are often acquired at different times, and a pronunciation scorer should score them separately rather than collapsing them.
At what age should a child say 'th' correctly?
Commonly cited norms place "th" mastery at around ages 5 to 6, later than most other consonants, with the voiced /ð/ often coming after the voiceless /θ/. Saying "fink" for "think" at age 4 is developmentally ordinary.
What is th-fronting?
Th-fronting is replacing /θ/ with /f/ and /ð/ with /v/ — "fink" for "think", "vis" for "this". In young children it is a normal developmental pattern. In adults it is a standard feature of several English varieties, including much of southern British English, where it is a dialect feature rather than an error.
Does ArticScore detect th-fronting?
Yes — th-fronting is a clean substitution between phonemes the model knows, which is the error type automatic scoring handles best. ArticScore scores the "th" target low and reports /f/ or /v/ as the sound heard instead. The caveat is dialectal: if your users speak a variety of English where th-fronting is standard, those flags are not errors, and you should decide deliberately whether you want them.