Claude Certification Blog
Multiple-response questions: the format that quietly costs marks
Every Claude exam tells you how many responses to select. None of the four guides says what happens if you select nearly the right set — which is why the count deserves a check of its own.
Multiple-response questions appear on all four Claude certification exams, and all four official guides say the same thing about them: the item states how many responses to select. The CCAR-F guide is the most direct, asking for exactly the stated number, no more and no fewer. What none of the four publishes is how a partially-correct set is scored — so the only safe way to treat the stated count is as a requirement, not a target.
What the guides actually say
The published item formats are multiple-choice and multiple-response. That is the whole list. No guide describes an ordering, matching or drag-and-drop format, and no guide describes a written or lab-based component. If you have seen a practice paper with those, it is somebody’s invention rather than the exam’s format.
The one piece of process every guide does commit to is the count. It is stated in the stem, on every multiple-response item, on every track. That consistency is unusual enough to be worth relying on.
What they do not say
Here is the gap that matters. No guide publishes a partial-credit rule, and no guide publishes a raw-to-scaled conversion for the 720 passing standard. Both facts are checkable and both are absent — which means anyone telling you that two out of three earns you two-thirds of a mark is guessing, and so is anyone telling you it earns nothing.
Plan against the published rule, not the unpublished one
You cannot plan around a scoring model nobody has published. You can plan around an instruction every guide states outright. Match the count exactly and the question of what a near-miss is worth never arises at all, and that half of it is entirely yours to decide.
The count is information, not just an instruction
The reframe worth taking into the exam: the number is evidence about the answer before you have read a single option. An item asking for two responses is telling you that two independent things are true about the scenario. If you have found one strong answer and are choosing between three weak ones, you have not finished the item — you have found half of it.
It cuts the other way too. If four options all look defensible and the stem asks for two, then two of them fail something, and finding out what is the actual work of the item. The count is the instruction to keep looking for a discriminator rather than to pick the two you liked most.
Three ways people lose the mark
| Failure | How it happens |
|---|---|
| Selecting too few | Two look right, the stem said three, and the third was never sought |
| Selecting too many | A defensible fourth gets added to a set the stem capped at three |
| Losing the count on review | An answer is changed late and the set silently ends up short |
None of these is a knowledge failure, which is what makes them worth a separate defence. The third is the meanest: you flag an item, come back to it, change one selection late in the session, and leave with a set that no longer matches the number the stem asked for. See the common mistakes post for the others in the same family.
They cost more time than they look
Every track gives you 120 minutes. CCAO-F, CCAR-F and CCDV-F run 60, 60 and 53 items respectively, and CCAR-P runs 63 — so the averages sit between roughly 1.9 and 2.3 minutes per item. Multiple-response questions do not consume an average share of that. Each one asks you to evaluate every option independently rather than pick a winner, and the CCAR-P guide notes separately that its longer scenario items take more time per item than single-concept ones.
The practical consequence is not to rush them, but to know that a paper with more of them will feel tighter, and to protect the review pass rather than the first pass. Our exam-day post covers the timing structure around that.
A review pass that works
Two mechanical checks, applied to every multiple-response item, in this order. First: does the number of boxes I have ticked equal the number the stem asked for? That is arithmetic, it takes a few seconds, and it needs no judgment at all. Second, only for items you flagged: is each selection independently defensible, or did one get carried along by another?
The order matters because the first check is fast and unconditional while the second is slow and selective. Doing them the other way around means running out of time before you have verified the thing you could have verified without thinking. For how much of that to practise beforehand, see how many mocks to take.
Key takeaways
- The count is always stated. All four guides confirm every multiple-response item says how many responses to select.
- Partial credit is not published. No guide describes it either way, so match the count exactly and the question never arises.
- Only two formats exist. Multiple-choice and multiple-response; nothing published describes ordering, matching or drag-and-drop.
- Read the count first. Two responses means two independent things are true, which narrows what you are hunting for.
- Most misses are not knowledge misses. Too few, too many, or a late change that quietly breaks the set.
- Count first on review, judge second. The arithmetic check is unconditional; the judgment check is only for flagged items.
This is a habit, and habits need repetitions
Reading about the count does nothing; ticking the wrong number of boxes under time pressure and catching it on review is what builds the reflex. Full-length timed papers with real multiple-response items are the only place that happens before it matters. Our claude certification study guide covers when to start them.
Try a full timed paperQuestions
Frequently asked
The follow-up questions people search next.
Do Claude certification exams tell you how many answers to select?
Yes, on every track. All four official guides state that multiple-response items say how many responses to select, and the CCAR-F guide puts it most plainly: select exactly the stated number, no more and no fewer.
Is there partial credit for getting some of a multiple-response item right?
No official guide publishes an answer either way. None of the four describes how a partially-correct set is scored, and none publishes a raw-to-scaled conversion. The practical consequence is to treat the stated count as a hard requirement rather than a target to get close to.
What proportion of the exam is multiple-response?
The guides do not publish one. They confirm both formats appear and that the count is always stated, and they stop there. Any specific proportion you see quoted is somebody’s practice allocation rather than Anthropic’s.
Are there drag-and-drop or ordering questions?
The published item formats on all four tracks are multiple-choice and multiple-response. No guide describes an ordering, matching or drag-and-drop format.
How should I use the review pass on these?
Check the count on every multiple-response item before checking anything else. It is a mechanical check that takes seconds per item and it catches the one failure mode that has nothing to do with knowing the material.
Keep reading
Related posts
Not affiliated with, or endorsed by, Anthropic or Pearson VUE. Details are summarised from publicly published program information and can change — always confirm against the official exam guide before booking.