Atypical responses scales as the tool of non-cooperative testing behavior detection

Non-cooperative testing behavior (NCTB) during questionnaire completion is serious threat to the valid interpretations of test scores. There are a number of methods recommended for use to identify protocols with NCTB features in order to enhance the credibility of personality testing results (Karabatsos, 2003; Desimone, 2015), but not all of them are proper and accessible for practitioners. One of the earliest methods of aberrant responding detection is based on incorporation of validity scales, including those composed of response options with rare endorsement. Assumptions and considerations regarding design (without adding supplementary items) and appropriate use of “atypical responses” (AR) scales are discussed. Such scales have been designed and implemented in some known and newly designed questionnaires. Findings suggest that the AR scale cutoff that is optimally efficient for detecting NCTB in high-stake testing may be calculated on randomly generated data as MR - 2*SDR. Validation of AR cutoff scores was accomplished by means of comparison of eight empirical (N = 406-5689) and randomly generated data for six questionnaires with different number of response options. Established sensitivity (0.92-0.99) is provided evidence for acceptable discriminating power of AR scores. This figure is comparable with person-fit methods based on IRT. Overall, the results provide support for the AR scores as indexes of NCTB. Military personnel testing practice suggest that taking into account AR scores enables to enhance the credibility of testing results and to diminish risks of diagnostic failures.