Key (Item Writing)

The key is the correct option in a multiple-choice item — the response that scores. The other options are distractors; together with the stem they form the item. Answer key is the same word used for the full list of correct options across a test, the document used in scoring.

A key earns its place by two properties: it is unambiguously correct against the construct, and it is not betrayed by surface features that let test-wise candidates pick it without engaging the construct. The first is content; the second is craft. Most operational scoring failures trace to one or the other — a defensible alternative answer, or a giveaway in option length, syntactic form, or specificity.

Properties of a sound key

Property	Why it matters
Singular correctness	No defensible second answer under the construct or in plausible candidate readings
Surface parallelism with distractors	Same length range, same grammatical form, same level of specificity
Position rotation across the test	Keys distributed roughly evenly across A/B/C/D positions; humans favour middle positions when rushed
Construct-aligned	The reasoning needed to select it is the reasoning the test wants to measure
Defensible to a content expert	A second reviewer trained in the construct would mark it correct without hesitation

Haladyna, Downing and Rodriguez (2002) consolidate the empirical evidence: keys that are systematically longer, more qualified, or more grammatically complete than distractors are picked at higher rates by candidates who have not engaged the construct. The asymmetry is what test-wiseness training exploits.

Construct correctness vs. surface correctness

Two failure modes recur:

• Two-key items. Both an option and the intended key are defensibly correct. Surfaces in distractor analysis when two options share top endorsement and both have positive discrimination. Common in inference items where the passage supports more than one reading.
• Surface-correct keys. The key is the only option that fits the stem's syntax, length pattern, or register, regardless of content. Candidates who recognise the cue without understanding the content still score. Surfaces as anomalously high facility on items the content panel expected to discriminate.

Both are caught by pretest. Operational test programs reject items where two raters disagree on the key, and rewrite items whose key wins on surface features.

The miskey problem

A miskey is an item operationally scored against the wrong option. Diagnosed when the supposed key shows low or negative point-biserial in distractor analysis while a distractor shows positive — strong candidates are rejecting the listed key for an option the writers labelled incorrect. Standard responses: rescore against the empirically correct option, double-key (accept both), or drop the item. Miskeys are most damaging on high-stakes single-form tests, which is why operational programs run independent key-verification panels before live administration.

Position bias

Random placement of the key across positions is the textbook prescription, but Bar-Hillel (2015) and follow-up work show that human writers, left to themselves, place keys in middle positions disproportionately and that uniformly random placement reduces guessing yield. Some test programs balance positions by design — equal counts of A, B, C, D keys per section — though this can introduce its own pattern that test-wise candidates exploit. Pseudo-random placement with position-frequency monitoring is the operational compromise.

In language testing

Language items add construct-correctness wrinkles other content domains do not face:

• Inference items. The key is supported by the passage but not stated; the construct is bridging inference. The risk is that a distractor matches a reading the passage almost supports.
• Vocabulary-in-context. The key is the contextual sense; distractors include the dominant dictionary sense. The construct is contextual disambiguation, not lexical knowledge.
• Function items. The key states the rhetorical purpose; distractors state plausible alternative purposes the passage's argument structure rules out. The construct is discourse-level interpretation.

Each requires the writer to verify the key against the passage with a content-blind reviewer who has only the question and options, to catch keys that depend on knowledge the passage does not actually deliver.

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.
• Bar-Hillel, M. (2015). Position effects in choice from simultaneous displays: A conundrum solved. Perspectives on Psychological Science, 10(4), 419–433.
• 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.