Speaking: coherent, intelligible recorded responses
The Speaking domain is delivered as recorded responses to prompts. Raters judge task completion, coherence and organization, fluency, grammatical range and accuracy, vocabulary, and intelligibility. There is no interlocutor to rescue a vague answer, so structure has to be audible and immediate.
Core rules
- Use preparation time to outline three beats: claim, example, wrap. Do not script sentences; you will read them and lose fluency.
- Signpost out loud: "First", "The reason is", "For instance", "So what changed was". Raters can only credit structure they can hear.
- Answer the prompt in the first sentence. "Last spring our team had two weeks and no budget to run a normally eight-week onboarding" delivers situation and constraint at once.
- Manage pauses by paraphrasing rather than stopping: "the thing that kept blocking us - the approval step -".
- Intelligibility beats accent: slow slightly, land final consonants, stress the content word in each phrase.
Worked example
Prompt: Describe a time you solved a problem with limited resources.
Reasoning skeleton: constraint (two weeks, no budget) → two specific actions (dropped the live training day, replaced it with recorded walkthroughs; paired each new hire with one reviewer) → result with a number (onboarding cut to nine days) → reflection (constraints forced us to cut content nobody used).
The opening sentence above is strong because it names the constraint before any narration. A chronological account that begins "So, um, it was last year and we had this situation" spends the first fifteen seconds saying nothing ratable.
Traps that cost marks
- Answering with one sentence and then stopping.
- Narrating events in order with no stated point or outcome.
- Deploying a memorized template that does not fit the prompt.
- Filling pauses with "um" and repetition instead of paraphrasing.