NBME subject exam timing — how long per question is actually realistic?
I've been doing NBME practice forms and I keep running out of time on the last block. I'm averaging about 95 seconds per question on my first pass, which sounds fine on paper, but when I hit a cluster of dense clinical vignettes I lose 30-40 seconds per question and it compounds badly by the end of the block. Anyone else dealt with this and what actually fixed it?
I'm doing the Clinical Science forms right now, specifically for internal medicine. My scores on the retired forms have been in the high 60s to low 70s — 67, 71, 68 — which I think puts me around a 220-225 on the Step equivalent scale if I'm reading the conversion correctly. I want to be solidly above 230 before I schedule the real exam.
My attending told me to stop doing timed blocks and just focus on question review, but that doesn't feel right this close to the exam. I've got about 5 weeks left and I don't want to sacrifice timing practice for more passive review time. Knowing the content doesn't matter if I can't finish the block.
What's the optimal split between timed block practice and review at 5 weeks out? And is a 95-second average actually a problem or am I stressing about the wrong thing?
67-71 on the retired forms roughly maps to 220-225 as you said, but there's real variance in those conversions. A jump from 71 to 77+ on the real exam isn't uncommon if the practice forms you used were older ones — pre-2020 IM sections tend to run a bit harder than what you'll actually see.
Try doing your next 3 blocks with a hard 85-second cap enforced — skip anything you're not confident on immediately. It feels wrong at first but after a few blocks your brain learns to triage faster and the clustered vignettes stop eating your time the way they do now.
95 seconds average is fine — the issue sounds like you're spending 3+ minutes on the hard vignettes instead of flagging and moving on. Anything over 90 seconds should get flagged and skipped on first pass. Come back at the end. You'll almost always have 4-6 minutes left if you do this consistently across a full block.
Your attending's advice makes sense in a different context — early in dedicated study you should be doing untimed review to actually learn the material. But at 5 weeks out, timed blocks are exactly what you should be doing. They're probably thinking of students who use timed practice as a crutch to avoid doing real review of their wrong answers.
Honestly I almost bailed on NBME forms entirely because of this exact thing. The per question average is a lie. It's not that you're slow, it's that the time is unevenly distributed and nobody tells you that. A dense vignette with a full paragraph of labs is never going to take you 95 seconds, so stop pretending it will. What actually saved me was giving myself permission to guess and flag hard. I'd read the last line first, and if the answer didn't click in maybe 30 seconds I'd mark my best guess, flag it, and move. I didn't think it would work and I resented doing it, but the whole point is protecting the easy points at the end of the block that you're currently donating to the clock.
The compounding you're describing is real and it's the thing that fails people. One 40 second overrun isn't the problem, it's that you carry the panic into the next question and now you're rushing stuff you actually know. I kept a rough checkpoint in my head, like I want to be at question 15 with 20 minutes left, and if I wasn't I knew I had to pick up the pace right then instead of finding out at the end. It felt mechanical and honestly kind of joyless for a while. But my timing evened out over a few forms and I passed, so keep going even if it feels like it isn't clicking yet. You're closer than it feels.
I had the exact same problem and honestly what saved me was stopping my timed practice at the question level instead of the block level. I'd do a 40-question block and afterward go back and flag every question that took me over 90 seconds. Then I'd drill those vignette types specifically with flashcards until I could read the stem and know what they were testing in the first two lines. That's it. Nothing fancy.
The compounding you're describing is real but it's almost always the same two or three vignette patterns killing you. Once you identify yours it gets way more manageable. I didn't change how fast I read, I just got faster at recognizing what the question actually wanted before I finished reading it. Last block timing stopped being an issue after about two weeks of that.