The /θ/ drills that do not survive into speech
You teach a B1 general English class of 16 adults at a private language school: nine Mandarin L1 speakers, four Vietnamese, three Arabic. Over three 90-minute lessons you have drilled /θ/ against /s/ using minimal pairs (think/sink, three/tree) with choral repetition, back-chaining and pair drills. In controlled repetition, accuracy on target words is around 90%. In a recorded discussion task at the end of lesson 3, the same learners produced "I tink so" and "tank you" 14 times out of 17 target tokens. Several learners also write "three" when they mean "tree" and spell "think" as "tink" in their journals. The school syllabus requires a pronunciation focus in every lesson but does not specify which features or how long to spend on them.
1. Before planning more /θ/ work, what exactly do you need to diagnose? Set out what the evidence in this class does and does not tell you.
Read the worked answer
The evidence tells me three separate things, and confusing them would waste the next three lessons.
- The learners can produce the sound. 90% accuracy in controlled repetition shows the articulatory target is achievable when attention is free. So this is not (mainly) an articulation problem and more drilling of the same kind will not transfer.
- The substitution is /t/, not /s/. "Tink", "tank" - so my minimal pairs against /s/ were aimed at the wrong contrast for most of the class. I need to check whether /θ/→/t/ holds in all positions (initial in think, medial in something, final in both, bath) or only word-initially.
- It breaks down under load. Accuracy collapsed when learners were managing meaning, turn-taking and formulation at once. That points to automatisation and to /θ/ being low in their processing priority, not to ignorance.
What I do not yet know, and would check in five minutes of the next lesson:
- Perception. Can they hear the difference? A quick same/different or odd-one-out task on think/tink separates a perceptual problem from a production one. If perception is fine, I stop ear-training and work on fluency.
- Intelligibility cost. Is "tink" actually causing misunderstanding, or is it a cosmetic error? That decides how much syllabus time it deserves relative to, say, their sentence stress, which matters more for being understood.
- Whether spelling is driving pronunciation. "Three/tree" confusion in writing suggests the learners are linking the two systems, and "tink" spelling suggests over-application of a sound-to-spelling rule. That is a lexis/spelling issue, not a phonology one.
2. Design a 20-minute pronunciation sequence for the next lesson that is more likely to transfer than the drills you have been using. Justify each choice.
Read the worked answer
Principles: aim at /t/ (the actual substitute), work in chunks rather than isolated words, build in dual attention so the feature is practised while meaning is processed, and recycle rather than finish.
1. Notice (4 min). Two or three high-frequency chunks on the board: I think so. Thanks a lot. Is that both of them? Learners mark where the tongue must be. Keep metalanguage to one sentence - tongue tip between the teeth, air continuous - and move on; explanation is not the bottleneck.
2. Hear it (3 min). Fast same/different on think/tink, three/tree, taught/thought, with hands up, no writing. This is a diagnostic as much as practice; if they cannot hear it, I do perception work before production.
3. Produce in chunks (6 min). Back-chain the whole phrase, not the word: -ink so → think so → I think so. Drill with a reason to say it: learners respond to quick opinion prompts ("Best day of the week?" - "I think so, yeah") so the phrase is retrieved for meaning.
4. Use it under load (5 min). A short ranking or guessing task in pairs where those chunks are the natural tool, with me monitoring and noting, not interrupting. I am checking whether the feature survives when attention is elsewhere - that is the only measure that matters.
5. Recycle (2 min). Note two or three successes aloud at the end, put the chunks on a wall list, and revisit in the next two lessons' warmers. One 20-minute slot does not fix a phoneme; spaced re-encounters do.
I would also drop the /s/ contrast from this class's plan and keep it only for learners whose substitution really is /s/.
3. What do you do about the spelling evidence ("tree" for "three", "tink" for "think")?
Read the worked answer
Keep the two systems separate in planning but connected in the learners' awareness.
- Do not let spelling drive the pronunciation syllabus. Choosing targets by spelling pattern would give me a rule-based lesson; the actual problem is retrieval speed in speech. So spelling gets its own short slot.
- Treat the high-frequency words as sight vocabulary. think, thing, thank, three, both, with are frequent enough to be learned as whole written forms with a pronunciation attached, rather than through a phonic rule that will over-generalise to "tink".
- Make the link explicit once. A two-minute noticing task: same sound, different spelling (think, both, bath); same spelling, different sound. Learners add the words to their vocabulary records with the phonemic form, not just the letters.
- Check the writing error is really phonologically motivated. If a learner writes "tree" for "three" in a composition, it may be a spelling slip under time pressure rather than evidence of a mental representation problem. I would look at several scripts before building a lesson on it.
Priority rule: intelligibility first, high-frequency vocabulary second, spelling patterns third. Anything that does not meet one of those tests waits.