Diphthong · oi / oy The start is low and back and already rounded — that is the one distinction from the starts of /aɪ/ and /aʊ/, both unrounded — then it glides to /ɪ/. The usual error is splitting it into two syllables.
Three steps to /ɔɪ/
Start rounded, finish spread.
Low back start, lips already rounded
Almost the /ɔ/ start: low and back with rounded lips. That is the entire boundary with /aɪ/ and /aʊ/, whose starts are unrounded.
Glide to /ɪ/, spreading the lips
The tongue front rises and the lips go from rounded to spread. That round-to-spread change is the audible cue.
Self-check: one syllable, no pause
Say boy. The whole gesture takes about 200 ms with no stop. If it sounds like «bo-yee», the glide is broken.
Articulation (side view)
One speaker, one instant: tongue position and lip shape only. The amber circle marks what to look at; the caption says how it differs from the sound it is usually confused with.
Spelling: oi inside a word, oy at the end
A rare positional rule in English spelling — and quite reliable.
voice, choice, join, point, coin, noise. Very consistent.
★ Final position spells oy: boy, toy, joy, soy, enjoy, annoy. Position decides the spelling — English does not end words in oi, nor put oy in the middle.
royal /ˈrɔɪəl/, loyal /ˈlɔɪəl/ — the positional rule's exceptions.
Why learners replace /ɔɪ/
No L1 /ɔɪ/ means the nearest strategy is two syllables.
Split into two syllables
boy becomes «bo-yee», voice becomes «vo-ice». The highest-frequency error here, because the diphthong does not exist in the L1.
Duration is the test: /ɔɪ/ takes about 200 ms as one syllable.
Start not rounded
Start from /a/ or /ʌ/ and the whole thing becomes /aɪ/: boy sounds like buy.
Compare the drawings: /ɔɪ/ starts low back rounded, /aɪ/ low front spread. Different lip shapes entirely.
Glide dropped
Only the first half: voice → /vɔs/.
The endpoint must be /ɪ/, with the lips going round to spread.
Word list
27 words grouped by where the sound sits in the word. Click a word to hear it. These are recorded as separate words, not cut out of the sentences above — a word cut from a sentence carries its neighbours' transitions and always sounds muddier. All three groups use the same speaker, so what changes between them is position, not voice.
Word-initial9
Word-medial9
Word-final9
★ The initial group is almost entirely the oil family. Not laziness: English has very few words beginning with /ɔɪ/ beyond oil. The position really is that sparse.
Corpus and acoustic measurements
Click any word in a sentence to play from that word. The pale band over the waveform marks the measured boundary of the target word.
Per-token data
Every number below is measured directly from the audio, not hand-annotated.
| Token | Accent | Sex | Target | Word length | ɔɪ duration | F0 | Pitch range | HNR |
|---|
Where the numbers come from
Every figure on this page can be recomputed from the description below.
Corpus: 12 English sentences synthesised with four American voices (two male, two female), each containing one target word, covering initial, medial and final positions. Audio is 22.05 kHz mono 64 kbps; each clip runs from 1.2 s before to 1.5 s after the target word.
Word boundaries: taken from the engine's word-level timestamps (WordBoundary) and corrected for the fixed offset between the engine timeline and the audio start (estimated per item). The target word is shaded in the waveform.
Phoneme boundaries: measured by forced alignment with the MMS_FA acoustic model. Each character's aligned frame range is kept, then a dynamic program maps the character sequence onto the phoneme sequence (th→θ, ng→ŋ, ck→k). Boundaries fall in the middle of the transition between characters; within-word coverage is 100%. Time precision is the model frame length, 20 ms, and boundaries are taken at midpoints, so effective resolution is about 10 ms.
Pitch: Praat's autocorrelation (to_pitch), 10 ms frame step, search range narrowed to the speaker's estimated range, octave errors removed. «Pitch range» is max minus min F0 within the word.
Why there are no paired differences here: this is a single-phoneme page. Eight other pages are contrast pages (θ/ð, l/r/n, vowel length) where the numbers come from paired words read by one speaker. A single-phoneme page has nothing to compare against — pairing it with a neighbour would measure that neighbour, which belongs to a different page.