Preparing for the PTE like an engineer

My first instinct going into PTE prep was the same instinct I'd bring to learning any new spoken language skill: find good material, listen and speak a lot, trust that fluency comes with volume. That instinct is basically correct for actually learning a language. It's a noticeably worse strategy for a specific exam, because I kept quietly forgetting one fact that changes the whole approach: nobody grades a PTE speaking response. A model does. And a model isn't the same thing as a person, no matter how good it's gotten — it has an actual scoring function underneath it, and a scoring function is a system, and I already know how to study a system. I'd just been studying English instead of studying the exam.

That reframe is the entire post, honestly. Everything else is just what changed once I took it seriously.

Fluency, and the trap of stopping to fix yourself

The single biggest shift was around disfluency — every "um," every restart, every self-correction mid-sentence. My instinct as a careful person, and I think this instinct is common among people who care about getting things right, is that noticing a mistake and fixing it mid-sentence is a good sign. It shows self-monitoring. It shows you know the correct form.

The scoring model doesn't read it that way at all. It reads a self-correction as a disruption to fluent, continuous speech, full stop, with no credit given for the fact that the correction was linguistically accurate. "I go to the — I went to the store yesterday" scores worse for fluency than just saying "I go to the store yesterday" and letting the tense error stand, even though the second version is factually less correct. That felt genuinely wrong to sit with at first — it's rewarding the less accurate sentence — but once I accepted it as simply how the scoring works, the practical instruction became clear and almost freeing: commit to the sentence you started. Don't backtrack. A small grammar slip costs you less than the act of stopping to fix it does.

I drilled that specifically, not general fluency. Read a paragraph aloud, and if a word came out wrong, physically make myself keep going instead of the reflexive stop-and-redo. It felt uncomfortable and slightly wrong for the first few sessions, actively fighting a habit that serves me well in every other context. It's exactly the kind of counterintuitive rule you only find by treating the exam as a system with rules to learn, rather than assuming "speaking well" and "scoring well" are simply the same target.

Templates: useful for content, dangerous for delivery

A lot of PTE prep material leans hard on templates — fixed sentence frames for the essay-writing and long-answer sections, meant to guarantee you hit every structural requirement the rubric is checking for. Templates genuinely do help there, for exactly the reason they're recommended: they guarantee complete, correctly structured content when nerves might otherwise make you ramble or skip a required part.

Where they actively hurt is speaking delivery. A memorized, rehearsed phrase read aloud has a completely different rhythm and intonation pattern than genuine spontaneous speech, and the scoring model can apparently pick up on that difference in ways that cost you on the fluency and pronunciation dimensions specifically — not the content dimension, where the template is doing its job fine. So I ended up splitting my prep cleanly in two: templates for structuring what I was going to say in the writing and long-answer sections, zero templates for how I said anything out loud. That distinction, content scaffolding versus delivery scaffolding, isn't one I'd have made on my own; it only became obvious once I stopped treating "prep material" as one undifferentiated pile of advice and started asking which specific scoring dimension each piece of advice was actually targeting.

The specific sounds worth isolating

Generic "work on your pronunciation" advice is true and almost useless, because it doesn't tell you where to actually spend limited practice time. What helped was identifying the two or three specific sounds that were both genuinely difficult for me and disproportionately common across the actual test content — not every hard sound, just the ones with real return on investment given limited prep time.

The 'th' sounds were the clear standout, both the voiced version in "this" and the unvoiced version in "think," since neither one has a real equivalent in Kinyarwanda's phoneme set, so my mouth had never been trained to make them without deliberate, isolated repetition. Rather than trying to fix pronunciation generally, I drilled minimal pairs specifically targeting that one gap: "think" versus "sink," "this" versus "dis," over and over, until the tongue placement stopped being something I had to consciously think about mid-sentence. That's a targeted fix for a diagnosed, specific gap, not a vague "practice speaking more" that spreads limited effort evenly across sounds I was already fine at.

What I actually learned

I don't think the lesson is "exam prep should always ignore what actually helps you communicate" — plenty of what improves your PTE score, genuinely fluent unbroken speech chief among them, also makes you a better communicator in real conversations, no tension there at all. The lesson is narrower: whenever a human skill gets evaluated by an automated system, the system's actual scoring function and your intuitive sense of "doing it well" can quietly diverge in specific, identifiable, learnable places — and finding exactly where they diverge is worth more prep time than generically practicing harder. I already knew that instinct from debugging: don't guess where the bug is, go find the exact line. I just hadn't thought to apply it to something as human-feeling as a speaking exam until I sat down and actually treated it as a system with rules, the same way I'd treat any other one.