Why do 'thought' and 'taught' score differently to me?

Because they differ in exactly one sound — the consonant at the very start, /θ/ vs /t/ — and if that difference isn't registering for your ear yet, the two words are genuinely indistinguishable to you even though they're clearly different to a native listener.

30-second version: "Thought" is /θɔːt/ and "taught" is /tɔːt/. The vowel, the final consonant, everything after the first sound is identical. The entire word hinges on whether the tongue goes between the teeth for a continuous fricative (/θ/) or taps the ridge behind the teeth for a quick stop (/t/) at the very start. If your first language doesn't have /θ/ at all — which is most languages — your ear may not have learned to listen for that specific difference yet, which is a separate skill from being able to produce it.

Why this feels like a vowel problem but isn't

It's a natural mistake to assume a pair that "sounds the same" must differ somewhere subtle in the vowel, since the vowel is the longest, most audible part of both words. But here the vowel, /ɔː/, is identical in both — the entire contrast lives in a fraction of a second at the very front of the word. That's actually good news: fixing this pair means fixing one specific consonant articulation, not retraining an entire vowel system.

This connects directly to can AI really hear my th sound — the underlying issue is the same /θ/ substitution — but "thought vs taught" is worth its own look because the words are common, function almost identically in a sentence ("I thought so" / "I was taught so"), and a mix-up here changes meaning in a way that's easy for a listener to miss silently, since both readings are grammatically plausible.

How sayit tells them apart

The scoring doesn't compare "thought" against "taught" as competing words — it compares your recording's phone sequence against the canonical IPA for whichever word was the target. If you were reading "thought" and the acoustic evidence at that first slot matches /t/ rather than /θ/, the breakdown records a substitution there specifically, distinct from every other phone in the word, which stay unaffected. That's why the two words can "score differently" even when they sound identical to you — the model isn't relying on the same broad impression your ear is; it's measuring frication and tongue contact directly from the waveform.

What the fix looks like

Train the contrast before you train either sound in isolation — hearing the difference usually comes before reliably producing it. Say "thought" and "taught" back to back, slowly, and pay attention to whether your tongue tip is between your teeth (θ) or behind them (t) at the very start. A useful check: hold the first sound of "thought" for a full second before releasing into the vowel — /θ/ can be held continuously as a hiss; /t/ cannot, because it's a stop, not a fricative, so trying to hold it just produces silence followed by a burst.

Drill it

Thought vs taught is the direct pair. If it's still hard to hear the difference at all, start with a clearer contrast first — three vs tree or thin vs tin — where the rest of the word differs too, making the target sound easier to isolate by ear before you tackle a pair where everything else is identical.

Record: "I thought she taught the class well." Both target words sit three words apart, which makes a merge easy to catch on playback.

How do I know it's improving?

Test your ear separately from your mouth: have someone else (or a text-to-speech reference) say one of the pair at random and see if you can identify which one without seeing it written. If you can't yet, drilling production alone will plateau — the listening discrimination needs to catch up first, and it usually does with the same minimal-pair repetition.

Try it

Open sayit free — no install, no card — and record the sentence above. If both target words get the same verdict despite one testing /θ/ and the other /t/, that's useful information either way — it tells you whether the contrast has landed yet.