To reverse or to not reverse Likert-type items: That is the question.

Abstract

The suitability of using reversed items in typical response measures has been a matter of controversy for many years. While some authors recommend their use, others reject them due to their undesirable effects on tests' psychometric properties. The present research intends to analyse a third alternative based on the use of reversed items plus a procedure to control response bias effects. We analysed two forms of the same test, one with direct and reversed items and another composed only of direct items, and compared them both before and after applying a procedure to control response biases. The factorial structure and factorial reliability of both versions was almost equivalent after controlling response biases. When no effect biases were controlled, the version with both types of items exhibited less acceptable psychometric properties. The use of reversed items is not advisable without the application of a procedure to control response bias effects. When such effects are mitigated, the results are equivalent to those obtained with only direct items, but with the added value of controlling for acquiescence effects.