CAEL Speaking: the 15-minute recorded interview
CAEL Speaking is a recorded, examiner-led interview of roughly 15 minutes, delivered in person or online. It does not test casual chat: it tests the kind of talk a seminar or lab demands - describing data, explaining a process, and defending a position under mild time pressure. Your responses are rated afterwards against band descriptors covering fluency and coherence, lexical resource, grammatical range and accuracy, pronunciation, and communicative effectiveness.
The three parts
- Part 1 (warm-up, 2–3 minutes): familiar questions tied to the test's overarching theme (your studies, a local issue, a habit).
- Part 2 (data task, 4–5 minutes): one minute of preparation, then you describe and interpret a chart, graph, or diagram, followed by the examiner's questions.
- Part 3 (abstract discussion, 5–6 minutes): opinions, causes, comparisons, and implications arising from Part 2.
Language that earns marks
- Signposting: "The chart compares…", "Turning to the second trend…", "Overall…".
- Approximation and hedging: "roughly a third", "just under", "this may suggest".
- Comparison structures: "twice as high as", "while", "by contrast", "the exception being".
- Extended answers built as claim + reason + example, not one-line replies.
Worked example
Prompt: a bar chart of household internet access across five provinces, ranging from 62% to 94%. A weak answer recites bars: "Ontario is 88. Quebec is 80." A stronger answer selects: "The chart compares household internet access across five provinces. Access is high everywhere - above 60% - but it ranges widely, from 62% in the lowest province to 94% in the highest. The three middle provinces cluster near 85%, while the top province stands about ten points clear of the rest. That gap may reflect differences in rural infrastructure." The reasoning: the second version frames the chart, groups the middle values instead of listing them, quantifies the outlier, and offers one hedged interpretation - precisely the selection skill and the range of structures the descriptors reward.
Traps that cost marks
- Describing every data point instead of the pattern.
- Using the preparation minute to write full sentences rather than five or six keywords.
- Delivering a memorised monologue that only half-fits the prompt.
- Answering a Part 3 question with "yes, I agree" and stopping.
- Ignoring the examiner's follow-up, which is where fluency is really tested.