Audit Reverse-Keyed Questionnaire Items

Obtain the scoring key before recoding

A reverse-keyed item is scored in the opposite numeric direction so that the resulting contribution aligns with the intended scale direction. Identify the instrument version, language, response options, and authorized scoring instructions. Do not infer the key solely from a negative correlation in the current dataset or from whether an item sounds unfavorable.

The psych package author documentation distinguishes explicitly supplied reverse keys from optional data-driven key checking. That distinction matters: automatic reversal can change a score definition in response to the sample. A scoring audit should implement a justified key and investigate unexpected results separately.

Specify the transformation and missing codes

For equally spaced numeric response codes with lower bound L and upper bound U, the usual reversal is L+U−x. Confirm that this formula matches the instrument's scoring rules; not every response scale is a simple equally spaced integer scale. Retain the raw response in a separate field and create a clearly named scored value.

For a fictional one-to-five item, the reversal is 6−x. Responses 1, 2, 3, 4, and 5 become 5, 4, 3, 2, and 1. A missing code such as 9 must first be identified as missing; mechanically applying the formula would produce −3, a meaningless score that can silently contaminate totals.

Check a complete record by hand

Suppose an invented four-item scale has responses 2, 4, 1, and 5, with items two and four reverse-keyed. The scored responses are 2, 2, 1, and 1, totaling 6. The uncorrected total would be 12. Neither number should be accepted merely because it lies inside the possible total-score range of 4 to 20.

Test all allowed response categories, not just one typical record. Reversing a valid item twice should recover its original value. Confirm that missingness remains missing and that invalid codes trigger review. These small checks are more informative than looking only at the final average score.

Review the consequences without redefining the scale

Compare item frequencies before and after reversal and inspect unexpected scored relationships. Investigate import errors, response-label mismatches, translation differences, and duplicate reversals before assuming that an item is defective. A negative item–total association can reveal several problems; it is not automatic permission to alter the scoring key.

Record decisions in the reliability planning worksheet before evaluating score consistency or using the Spearman–Brown tool for a later length scenario. Reversal does not itself prove reliability, dimensionality, or validity.

Finalize a scoring specification containing item identifiers, bounds, reverse keys, missing-item requirements, and total or mean rules. Preserve its version with the analysis output so another analyst can reproduce exactly how raw responses became the reported score.

Sources checked on 5 November 2026; all numerical examples are fictional and independently recomputed.