September 8, 2026 · 8 min read
TL;DR: Bucket 1,948 questions by stem length and the type shifts predictably. Under 30 words: 10% definitions, 16% what-comes-first. Over 70 words: 0% definitions, 70% what-comes-first. Length is not difficulty, it is a signal of what kind of question you are about to answer, and it is readable in the second before you start.
Two questions can be worth the same mark and demand entirely different reading. Eight words asking what a milestone is. A hundred and forty words about a data-centre cutover with a failed backup.
The difference is visible before you have read either one.
Every single-answer question, bucketed by word count.
| Stem length | Count | Asks first/next | Straight definition | Longest option is keyed | Avg option length |
|---|---|---|---|---|---|
| Under 30 words | 526 | 16.2% | 10.3% | 49.2% | 54 chars |
| 30 to 49 words | 829 | 35.3% | 0.1% | 49.5% | 69 chars |
| 50 to 69 words | 444 | 42.3% | 0% | 46.8% | 84 chars |
| 70 words or more | 149 | 69.8% | 0% | 40.3% | 99 chars |
Mean stem length is 42 words. The shortest is 3, the longest 149.
Three things move together. Definitional questions vanish entirely above 30 words. Sequencing questions climb from one in six to seven in ten. And the longest-answer tell decays from 49% to 40%, because as stems grow, options grow too, and the gaps between them compress.
Because length is what a sequencing question needs.
To ask which action comes first, the item writer has to establish a state: what has happened, what has not, what is documented, who has authority, what the constraint is. Each of those facts costs a clause, and four or five of them is a seventy-word stem before any of the narrative colour.
A definitional question needs none of that. "Which of the following is a correct clarifying principle about simplicity" is complete in twelve words, because nothing in the world needs describing.
So the length is a side effect of the question type, which means it is honest signal rather than noise. And it tells you something useful before you have read a word of the content: on a long stem, the four options are near certain to be four defensible actions, and you are being asked to order them rather than judge them.
Common trap: "A long question is a hard question, so I should slow down and read it closely." Read it differently, not more slowly. A long stem is mostly narrative packaging around four or five load-bearing facts, and reading it as a story is how candidates finish the paragraph with a strong sense of the situation and no memory of whether the change was approved. On the data-centre cutover item, the keyed answer turns on one clause about the backup being unusable; everything else in the stem is stage-setting. This bank names the sentence that drove the answer rather than just marking the option wrong.
Read for state, not for story. Four questions, and you can usually answer all four on one pass:
Then read the options. On a seventy-word stem, expect all four to be reasonable actions, expect the discrimination to be about order, and expect the length heuristic to help less than usual.
It should, in one specific way: stop treating long questions as worth more. Every scored item on this exam counts identically, whether the scenario runs to eight words or a hundred and forty-nine.
The practical failure is not spending too long on a hard question. It is spending too long on a long question that is not actually hard, because the volume of text creates a feeling of consequence that the mark scheme does not share. A hundred-word scenario whose answer turns on one clause about an unsigned charter deserves the forty seconds it takes to find that clause, and no more.
The inverse is worth saying too. Short questions are where the definitional items live, and definitional items are the ones where you either know it or you do not. If you find yourself deliberating for ninety seconds over a twelve-word question, that is the signal to mark it, guess, and move on. There is no scenario to re-read, and the extra time is buying nothing.
Does the length of a PMP question tell you anything? It predicts the question type. In this bank, stems under 30 words ask what to do first or next only 16 percent of the time, and 10 percent are straight definitions. Stems over 70 words ask what to do first or next 70 percent of the time, and none are definitions.
Should I read a long scenario differently? Yes. A long stem is almost always a sequencing question, so read it for state rather than for story: what has already happened, what is documented, what has been approved, who holds the decision. The narrative detail is there to hide those facts, not to convey them.
Does the longest-answer trick still work on long questions? Less well. In this bank the longest option is keyed 49.5 percent of the time on stems of 30 to 49 words, and 40.3 percent on stems over 70 words. Options grow alongside stems, so the length differences between them compress.
Are long questions worth more marks? No. Every scored question counts the same, whether the scenario is 8 words or 149. That is worth remembering when a long one is eating your clock, because the time you spend on it is worth exactly what you would spend on a one-line definition.
PMP Practice's 2,141 questions are re-certified against PMBOK 8 and the July 2026 ECO, with every wrong answer explained, not just marked wrong, including which clause in the scenario was load-bearing. Start the free 20-question sample - no card, no signup required to try it.
Does the length of a PMP question tell you anything?
It predicts the question type. In this bank, stems under 30 words ask what to do first or next only 16 percent of the time, and 10 percent are straight definitions. Stems over 70 words ask what to do first or next 70 percent of the time, and none are definitions.
Should I read a long scenario differently?
Yes. A long stem is almost always a sequencing question, so read it for state rather than for story: what has already happened, what is documented, what has been approved, who holds the decision. The narrative detail is there to hide those facts, not to convey them.
Does the longest-answer trick still work on long questions?
Less well. In this bank the longest option is keyed 49.5 percent of the time on stems of 30 to 49 words, and 40.3 percent on stems over 70 words. Options grow alongside stems, so the length differences between them compress.
Are long questions worth more marks?
No. Every scored question counts the same, whether the scenario is 8 words or 149. That is worth remembering when a long one is eating your clock, because the time you spend on it is worth exactly what you would spend on a one-line definition.