Is Part 3 a lecture or a conversation?
Always a conversation between a man and a woman, one asking, one explaining. Some prep sites describe it as a monologue, that is simply wrong, expect two voices trading turns.
Part 3 plays one fact-heavy conversation where one speaker explains something the other wants to know, about 2 to 2.5 minutes, then asks 6 questions in about 6 minutes. Numbers, conditions, and corrections carry the answers, so this is the part where notes pay off most.
In Part 3 one speaker has information the other wants, a customer and a salesperson, or a new employee and an experienced one. The conversation runs 2 to 2.5 minutes, plays once, and is followed by 6 spoken questions, about 6 minutes in all. It is the most fact dense of the conversation parts, and the demo above is the exact interface.
The same computer scoring as the rest of Listening, one point per correct answer and no penalty for wrong ones. Part 3 contributes 6 of the 38 scored questions, and its dense details make notes worth more here than anywhere else.
Each of the 6 questions is right or wrong, and a blank is a certain zero. Even when a detail slipped past you, an educated guess built from your notes keeps real value on the table.
There is no separate Part 3 score. Its points join the Listening total out of 38, which converts to a CELPIP Level from 1 to 12 with difficulty equating across test versions.
The mapping is one to one, CELPIP 9 equals CLB 9, and around 33 of 38 correct is the usual bar. Part 3 sits mid test, right where strong and average listeners start to separate.
One experimental part may be mixed into the Listening test without counting, and it looks identical to the scored ones. Spend zero energy guessing which it is, full effort everywhere is the only safe play.
The introduction names the roles and the topic. Guess the shape before the audio starts, a comparison, a step by step process, or a list of tips, and draw a two column grid with room for numbers.
Prices, dates, times, and limits are exactly what the questions target. Capture each one with a single word of context, an orphaned 25 or Saturday in your notes can still win a question later.
Words like actually, but, and unless overwrite what came before, and the corrected fact is the one that gets tested. Mark corrections with an arrow, distractors quote the number from before the correction.
Questions follow the audio top to bottom, so ordered notes become a map. Hear the stem, drop two options fast, settle the last two against your notes, and click something before the timer ends.
These are the exact on screen instructions and the signpost lines that organize Part 3 information. Enumerations, corrections, and closing advice are exactly where the questions anchor.
While you replay a lost word in your head, two more facts go by. Keep listening and hold what you did understand, one gap costs one question at most, a spiral costs three.
It's twenty dollars, actually, twenty-five now. The test asks for the corrected figure, and the outdated one sits in the options waiting. Corrections are testable gold, mark every one.
Complete sentences cost you the next two facts. Point form, abbreviations, and symbols exist for exactly this part, and the test centre pen and paper reward whoever practiced with them.
The stem is spoken once and never shown. If your eyes drop to the options too early, you end up answering a question you never actually heard, listen first, then scan.
30 seconds is enough for one clean elimination pass, not for agonizing. Commit before the timer ends, a blank scores zero and a guess is free.
Always a conversation between a man and a woman, one asking, one explaining. Some prep sites describe it as a monologue, that is simply wrong, expect two voices trading turns.
In point form, and heavier than in any other part. Get every number with one word of context, number the steps of any process, and mark corrections. The centre provides pen and paper, and notes are never scored.
Keep listening and stay with what you understand. Questions come in audio order, so one lost detail risks one question, not the set. Answer it with your best guess, blanks are the only unforced error.