Establishing supplementary response time validity indicators in the Word Memory Test (WMT) and directions for future research

Abstract

Response time (RT) measures in the Word Memory Test (WMT) offer to complement information derived from conventional accuracy measures. The current study aimed to validate the findings of Lupu, Elbaum, Wagner, and Braw in which RT variability was assessed, for the first time, in the WMT. A secondary aim was to suggest directions for the future research of RT measures in Forced-Choice Recognition Memory Performance Validity Tests (FCRM-PVTs). The study utilized a simulation research design, with participants performing the WMT’s immediate recognition (IR) subtest (N = 59). Mean RTs and a scale which combines accuracy and RT measures, but not variability in RTs, possessed adequate discrimination capacity. Enhanced discrimination capacity was found after discarding the first items’ RTs, though the IR-subtest‘s accuracy measure still showed superiority as a stand-alone validity indicator. The promise of RT measures may, therefore, lie in their ability to illuminate speed-accuracy tradeoffs among examinees with border-zone accuracy scores in FCRM-PVTs. They should, therefore, be regarded as complementary to conventional accuracy measures.

Keywords:

• Feigned cognitive impairment
• malingering
• Performance Validity Test (PVT)
• response time
• simulation
• validation
• Word Memory Test (WMT)

Acknowledgments

The research reported in this article was completed as part of T. Elbaum’s Ph.D. dissertation, Ariel University, Ariel, Israel. The study was conducted as part of the psychology undergraduate program requirements of L. Golan and. C. Salzer, and as part of the graduate program requirements of M. B. Israel. N. Laufer, and B. Richmond aided in the recruitment of the participants and in performing the experimental procedures. We thank them for their invaluable contribution to the study.

Notes

1 The WMT software currently provides mean RT data. Per item data, enabling the calculation of RT variability measures will be available in the forthcoming WMT revision (P. Green, personal communication, October 28, 2016).

2 The RT data is presented in the WMT’s results spreadsheet and produced using REPORTING, EXPORT ALL DATA commands.

3 As noted earlier, per item data will be available in the ‎forthcoming WMT revision and will enable these more advanced analyses.

4 Please note that the formula should be utilized using RT data presented in milliseconds (ms) and not in seconds.

5 The WMT was not given in its entirety. Determining pass/fail for the WMT, therefore, cannot be made in accordance with the WMT’s classification scheme (Green, 2005). Such classification necessitates, at least, data from the other primary performance validity subtest of the WMT (i.e., DR). Moreover, data from other WMT’s subtests may be used to construct a performance profile, a detection strategy that is one of the WMT’s key assets (McGuire, Crawford, & Evans, 2018). The term “pass/fail” in the context of the current study should be considered in a narrower sense and not equated with the clinical decision making that is based on additional information that was not gathered as part of the current study.

6 We also re-calculated the combined scale after discarding the first item’s RT data. The interested reader can find additional information regarding the re-calculated combined scale in Supplementary Material 2.

7 Use of multiple strategies was listed by three participants. Percentages, therefore, add to more than 100%.

8 Since effort is manifested in RT, a different source of effort for each of the two groups (i.e., controls and simulators) may explain the significant correlations found only among controls, but not among simulators (Lupu et al., 2018). Accordingly, the effort that controls may apply in trying to recall the less remembered items, led to longer RTs. Thus, the significant inverse correlation between accuracy and RT only for controls. In contrast, simulators possibly apply more effort in deception, rather than a genuine recollection of items, and, thus, the lack of significant correlation between accuracy and RT, for simulators.

9 Analysis of our previously suggested cutoffs (Lupu et al., 2018) also indicated somewhat better classification accuracies (i.e., both 2110 ms and 2410 ms cutoffs now showed adequate specificity and sensitivity).