The stem is the part of a multiple-choice item that poses the problem — the question, incomplete statement, or scenario presented before the options. The stem holds the construct being tested; the key and distractors are how the response is elicited. Haladyna, Downing and Rodriguez (2002), reviewing fifty years of item-writing research, place stem clarity at the head of every empirically supported guideline: most item-quality problems originate in stems that are vague, double-barrelled, or carry the test-taker into negative wording.
Items use one of two stem formats:
- Question form. The stem is a complete interrogative: Which of the following best describes the writer's tone? The options are full or short answers.
- Completion form. The stem is a partial statement: The writer's tone in paragraph 3 can best be described as.... The options complete the sentence. Each option must form a grammatically coherent sentence with the stem.
Question form is preferred for novice item writers because it forces an explicit task; completion form is more compact and works well when the construct is naturally a fill-in. Mixing forms within a single test section is acceptable but should be deliberate.
What goes in the stem
The stem carries four loads:
- The task. What the candidate must do — identify, infer, paraphrase, locate, evaluate.
- The reference. Where in the passage, recording, or table the candidate should look. For reading and listening items this is often a paragraph, line, or speaker reference; for context-dependent vocabulary items it is the line and word.
- The construct framing. The cognitive operation being elicited. Best describes signals interpretation; according to the passage signals literal retrieval; can be inferred signals inference.
- All shared content. Anything common to all options belongs in the stem, not repeated across options. Repetition wastes reading time and signals laziness in writing.
Empirically supported rules
Synthesising Haladyna et al. (2002) and Rodriguez's (2005) meta-analytic work:
- State the central problem clearly in the stem so candidates can answer without reading the options.
- Use positive wording. Negative stems (Which is NOT...) increase difficulty for reasons unrelated to the construct; if used, mark the negative emphatically (capitals, bold).
- Avoid window-dressing — irrelevant material that lengthens the stem without contributing to the construct. Window-dressing penalises slow readers, not weak content learners.
- Match stem and option length distribution to known patterns in the candidate population. A long, qualified key paired with short distractors is a classic test-wiseness cue.
- Include the reference if the item is passage-bound. Vague stems force candidates to reread the entire passage, inflating time pressure as a confound.
Stem types in language testing
Language items make heavy use of context-bound stems. The five most common patterns:
- Comprehension stem. What does the writer suggest about X? Operates on a defined span. Construct: comprehension at sentence, paragraph, or passage level.
- Reference stem. The word "this" in line 12 refers to.... Construct: anaphora and discourse cohesion.
- Vocabulary-in-context stem. The word "abundant" in paragraph 2 most nearly means.... Construct: lexical inference from context, not dictionary knowledge.
- Inference stem. It can be inferred from the passage that.... Construct: bridging inference; the answer is not stated.
- Function stem. The author mentions X in order to.... Construct: rhetorical purpose at the macro level.
Each pattern has its own distractor-design logic. Vocabulary-in-context distractors should include a common dictionary sense that does not fit the local context; reference-stem distractors should include earlier nouns that share number or animacy with the correct referent.
Common stem failures
| Failure | What happens |
|---|
| Double-barrelled stem | Asks two things at once; candidates can be right on one and wrong on the other |
| Unstated frame of reference | Which is best? Best for whom, in what context? |
| Lifted from passage verbatim | Tests recognition of surface form, not comprehension |
| Negative wording without emphasis | Candidates miss the not and answer the wrong question |
| Construct leakage | Stem tells the candidate what to think — Why was the policy ineffective? presupposes the answer |
References
- Haladyna, T. M., Downing, S. M., & Rodriguez, M. C. (2002). A review of multiple-choice item-writing guidelines for classroom assessment. Applied Measurement in Education, 15(3), 309–334.
- Rodriguez, M. C. (2005). Three options are optimal for multiple-choice items: A meta-analysis of 80 years of research. Educational Measurement: Issues and Practice, 24(2), 3–13.
- Hughes, A. (2003). Testing for Language Teachers (2nd ed.). Cambridge University Press.
- Alderson, J. C., Clapham, C., & Wall, D. (1995). Language Test Construction and Evaluation. Cambridge University Press.