Malingering is the intentionally fabrication or exaggeration of physical and/or psychological symptoms motivated by external incentives (American Psychiatric Association, 2022). If incentive kind cannot be determined, such behavior is labeled feigning (see Merten & Merckelbach, 2020; Rogers & Bender, 2018). A third concept related to the above is that of negative response bias (NRB; or invalid presentation/performance). NRB is the tendency to portray oneself as more mentally disturbed or problematic than one's actual level of functioning, thereby distorting psychometric test results and assessment (Giromini et al., 2022). Estimates of the prevalence of feigning/NRB vary across instruments and settings (Dandachi-FitzGerald et al., 2013; Mittenberg et al., 2002), but converge around 15% (Young, 2015).

Failure to detect NRB can result in misallocation of resources (Chafetz & Underhill, 2013), unnecessary and potentially iatrogenic interventions (van der Heide et al., 2020), denial of necessary care (Knoll & Resnick, 2006), contamination of treatment protocols (Van Egmond et al., 2005), academic research (Abeare et al., 2021; Rienstra et al., 2013), or adjudication (Soliman & Resnick, 2010). Therefore, the credibility of psychological presentations should be objectively assessed using multiple symptom validity tests (SVTs; Sherman et al., 2020; Sweet et al., 2021).

SVTs are psychometric instruments designed to assess the credibility of self-reported symptoms and to provide an estimate of the likelihood of NRB. The most common detection strategies are pseudosymptoms (deficits that patients with genuine disorders rarely/never endorse), the threshold method (implausibly high symptomatology), or combinations of symptoms that are unlikely to occur together. By design, SVTs can be free-standing or embedded (i.e., part of a broadband instrument originally developed to assess true pathology; Giromini et al., 2022).

It is recommended to assess NRB using multiple SVTs (Giromini et al., 2022). However, the clinical and practical validity of psychological tests is influenced by linguistic, ethnic, and cultural differences, so they are not automatically applicable to diverse populations (Brantuo et al., 2022; Nijdam-Jones & Rosenfeld, 2017). This is particularly true for SVTs, where idiosyncratic interpretation of item content (Ali et al., 2022) may increase the risk for both false positives (leading to the denial of care or legitimate compensation) and false negatives (misallocation of scarce resources).

SVTs in Türkiye

The country formerly known as "Turkey" has officially changed its name to Türkiye. This change was adopted by the United Nations, the United States and other organizations and countries. See, e.g., United Nations. (2022, June 3).

Determining the classification accuracy of a given instrument in a cross-cultural context is a prerequisite for its use in clinical/forensic settings. Because the actuarial evidence base of SVTs was developed in a specific population (typically native-born citizens who speak the country's official language as their first language), it is unclear whether their clinical utility is transferable to a population with different demographic characteristics (recent immigrants with limited language skills and different cultural norms). Therefore, it should not be assumed that the classification accuracy of SVTs is transferable to examinees with a different ethnic or cultural background.

Little is known about SVTs in Türkiye or Turkish SVTs, although Turkish is spoken by about 200 million people, ranking seventh in the world in terms of number of speakers and area (see Ministry of Culture and Tourism of the Republic of Türkiye, 2023). In addition, Turks make up a significant portion of the population in several Western European countries. Although some instruments have been translated into Turkish (Brockhaus & Peker, 2003; Morel & Marshman, 2008), there is little research on their use. Ardıç et al. (2019) have reported on the validation study of the Turkish version of the Structured Inventory of Malingered Symptomatology (SIMS; Smith & Burger, 1997), a widely used and researched SVT (see Shura et al., 2022). A recent review of psychological tests in Türkiye (Ayhan & Karaman, 2021) identified broadband instruments with embedded validity scales such as the Minnesota Multiphasic Personality Inventory (MMPI; Hathaway & McKinley, 1943; Savaşır, 1978) and the Miller Forensic Assessment of Symptoms Test (M- FAST; Keyvan et al., 2015; Miller, 2001) as SVTs in use and concluded that additional tests are needed (see also Akca et al., 2023).

The Inventory of Problems

To address the need for additional Turkish instruments such as SVTs, this study was designed to validate a recently developed free-standing SVT, the Inventory of Problems–29 (IOP–29; Viglione & Giromini, 2020; Viglione et al., 2017), in a Turkish community sample. The IOP–29 can be administered in paper-and-pencil format or online, takes 5–10 min to complete (Giromini et al., 2021), and has a rapidly growing evidence base (Puente-López et al., 2023; Young et al., 2020). Its item content focuses on how problems are presented, rather than their mere presence or absence, and assesses the credibility of a wide range of neuropsychiatric presentations (see Giromini et al., 2020b): Depression (Bosi et al., 2022; Ilgunaite et al., 2022), psychosis (Banovic et al., 2022; Winters et al., 2020), PTSD (see Carvalho et al., 2021), neuropsychological deficits (Gegner et al., 2021; Holcomb et al., 2022), and combinations thereof (see Giromini et al., 2020b; Pignolo et al., 2021). In addition to “True” and “False,” “Doesn't make sense” is presented as a third response option. Eliminating the traditional dichotomy of “True” and “False” allows for a more nuanced manifestation of non-credible presentations feigning strategies (rare symptom or amplification; Viglione et al., 2017).

A key difference from other SVTs is that the IOP–29 generates a False Disorder Probability Score (FDS) by comparing responses to reference data obtained from real patients and experimental feigners (Giromini et al., 2020b). The FDS is a probability estimate ranging from 0 (unlikely to feign) to 1 (likely to feign) derived from logistic regression (Winters et al., 2020). This scaling method reduces the impact of outliers in research settings, but also facilitates decision making in real-world forensic settings where findings are often expressed as percentages or “on the balance of probabilities” (Giromini et al., 2020b).

The original English version of the IOP–29 has been translated and cross-validated in several countries (Brazil, France, Italy, Lithuania, Portugal, Romania, Slovenia) and is available in several other languages (see official website: http://www.iop-test.com). Emerging data indicate that the validity of the IOP–29 is at least comparable to that of longer and more complex instruments such as the NRB indicators embedded in the MMPI and the Personality Assessment Inventory (PAI; Morey, 2007), or popular free-standing SVTs such as SIMS and the Self-Report of Symptom Inventory (SRSI; Boskovic et al., 2022; Giromini et al., 2018, 2019, 2020a; Holcomb et al., 2022; Merten et al., 2016; Pignolo et al., 2021; Roma et al., 2020). The IOP–29 is also robust to coaching (Boskovic et al., 2022; Gegner et al., 2021), a potential threat to the clinical utility of any SVT (Rogers & Bender, 2018).

The developers recommend three levels of cutoffs: a liberal FDS ≥ 0.30 for screening-only purposes, a standard FDS ≥ 0.50 cutoff, and a conservative FDS ≥ 0.65 for high-stakes contexts where specificity is of paramount importance (Viglione & Giromini, 2020). Naturally, these cutoffs offer different tradeoffs in sensitivity and specificity: 0.90 and 0.60 for FDS ≥ 0.30; 0.80 and 0.80 for FDS ≥ 0.50; 0.70 and 0.90 for FDS ≥ 0.65 (Viglione & Giromini, 2020). However, a recent qauntitative review by Giromini and Viglione (2022) reports promising weighted mean sensitivity and specificity values that exceed those reported in the test manual: 0.94 and 0.76 for FDS ≥ 0.30, 0.86 and 0.92 for FDS ≥ 0.50, and 0.76 and 0.96 for FDS ≥ 0.65.

Although most of the studies reviewed were based on a simulation design that may inflate (Rogers & Bender, 2018) or obscure (Abeare et al., 2021) the true effect size, a comparison with the criterion group study by Roma et al. (2020) yielded comparable weighted effect size (Cohen's d of 3.02 vs. 2.98), sensitivity, and specificity (Giromini & Viglione, 2022). Recently, an independent replication based on psychometrically defined criterion groups and genuine patients found that the standard cutoff (FDS ≥ 0.50) had good classification accuracy, suggesting that it “may be sufficiently specific (0.90–0.91) for routine clinical use” (Holcomb et al., 2022). A second meta-analysis confirmed the IOP–29 as an effective SVT (Puente-López et al., 2023) but cautioned against language of administration (i.e., English vs. non-English) as a potential confounding variable and emphasizing the need for further cross-cultural research.

The present study is a response to this call: it was designed to develop and validate the Turkish version of the IOP–29. In addition, we wanted to test the sensitivity of the Turkish IOP–29 for different types of feigned disorders: schizophrenia, depression, and PTSD. Finally, we wanted to assess the utility of the Random Responding Scale (RRS; Giromini et al., 2020c) in discriminating between random responding and feigning. The RRS was developed as an alternative index to detect careless, uncooperative, or inattentive responding and to distinguish this response pattern from NRB (Giromini et al., 2020b, c; Winters et al., 2020). Indeed, evidence suggests that content-unrelated distortions (CUD; Nichols et al., 1989), such as random responding, can mimic NRB (Burchett et al., 2016; Merckelbach et al., 2019).

The Current Study

The study employed a mixed design using a simulation paradigm (Rogers & Bender, 2018): the Turkish IOP–29 was presented to the same participants three times in an online format, with instructions to respond honestly (HON), at random (RND), and attempt to feign a psychiatric disorder (SIM) based on various vignettes. Individuals who respond with a careless attitude do not respond completely at random, even when prompted (see Giromini et al., 2020c). To increase ecological validity and minimize the risk of artificially inflating effect sizes with computer-generated random responses, we added the condition RND. In the SIM condition, participants were presented with three scripts instructing them to feign schizophrenia (SIM-SCZ), depression (SIM-DEP), or PTSD (SIM-PTSD). We predicted no significant difference between participants’ FDS scores in the three feigning sub-conditions, but we hypothesized that FDS scores would be significantly lower in the HON condition compared with SIM (suggesting a valid profile). In addition, we expected that mean FDS values in RND would be between those of HON and SIM but closer to the mean of SIM, as a random response set is unlikely to be credible (see Giromini et al., 2020b).

Method

Participants

Based on G*Power analysis (Faul et al., 2009) with three groups (i.e., feigning sub-conditions SIM-SCZ, SIM-DEP, and SIM-PTSD) and with parameters α (alpha) = .05, 1–β (power) = .80, and f (effect size) = 0.25, a minimum sample size of 159 participants with valid responses (i.e., meeting inclusion criteria for analysis) was determined.Footnote 2

Inclusion criteria for participation in the study were age ≥ 18 years old, able to read and understand Turkish, and provide informed consent (N = 189). Exclusion criteria were a self-reported history of psychiatric or neurological disorders (n = 12) and self-reported poor mental health during the study (n = 17). In addition, failure to pass manipulation and inattention checks resulted in exclusion from the analysis (n = 35; see “Procedure” section).

One hundred and twenty-five Turkish-speaking adults (67 or 53.6% female; see Table 1) completed the study. The majority (97.6%) reported Turkish as their native language and rated their language proficiency highly on a five-point scale (1 = “very poor”; 5 = “very good”). For the entire sample (N = 125), the mean score for Turkish language proficiency was M = 4.72 (SD = .55; range: 2–5). Mean age was 27.8 (SD = 8.2; range: 19–53). Participants assigned to between-subject factors (SIM-SCZ, SIM-DEP, and SIM-PTSD) did not differ on gender, age, self-reported Turkish language proficiency, level of education, relationship status, and self-reported mental health (all ps > .05; Table 1).

Table 1 Demographic Composition of the Sample (N = 125)
Full size table

ProcedureFootnote 3

Using a translation/back-translation method (e.g., Brislin, 1970; van de Vijver & Hambleton, 1996), the original English version of the IOP–29 was first translated into Turkish by two native speakers who were not involved in the study and then back-translated into English by a third person who was blind to the original English version. After institutional research ethics approval, the study was advertised on social media and recruitment was promoted through snowballing.

Participants were compensated for their time with the opportunity to enter a raffle to win an Amazon gift card worth 250 Turkish Liras (TL; about $13 U.S. Dollar [USD]). After provided their demographic information, participants were asked to complete the IOP-29 three times under different instructions: HON, RND, and SIM – the within-subjects factor. For the between-subjects factor, participants were randomly assigned to one of the three vignettes in the SIM condition (SIM-SCZ, SIM-DEP, and SIM-PTSD). The Turkish versions of the scenarios are available on the open science framework (OSF) platform (https://osf.io/6xksf/). To facilitate feigning, participants were presented with a vignette containing characteristic symptoms for each of the three psychiatric disorders. Participants were also warned not to “overdo it” to avoid being detected as feigners.Footnote 4 They were informed that two participants would be rewarded with a 100 TL (approximately 5 USD) Amazon gift card if they managed to successfully feign the assigned disorder (operationalized as IOP–29 FDS < .50).

The design and vignettes have been used in previous publications (e.g., Giromini et al., 2020b; Pignolo et al., 2021). To neutralize potential order effects, the sequence of the main conditions (i.e., HON, RND, and SIM) was counterbalanced, in addition to random assignment to the feigning sub-condition (SIM-SCZ, SIM-DEP, and SIM-PTSD). For each condition, pretest (understanding of the instructions) and posttest manipulation checks (execution of the task) were employed. Following methodological recommendations (e.g., Meade & Craig, 2012; Ziegler, 2015), an item was added to the Turkish IOP-29 items to check for inattentive responding (HON condition only).

Data Analysis

First, we compared FDS scores between the three feigning conditions with a between-subjects comparison (one-way ANOVA) to determine whether there were significant Turkish IOP–29 false disorder probability score (FDS) differences for different types of feigning presentations (schizophrenia feigning condition [SIM-SCZ], depression feigning condition [SIM-DEP], and PTSD feigning condition [SIM-PTSD]), followed by a Tukey-corrected post hoc contrast. The repeated measures ANOVA was followed by Bonferroni-corrected post hoc contrasts. Effect size estimates were expressed in partial eta squared (η2p) and Hedges’ g. Next, we computed classification accuracy [area under the curve (AUC), sensitivity, specificity, positive predictive power (PPP), negative predictive power (NPP), and overall correct classification (OCC)] with honest responders (HON) versus feigners of psychiatric disorders (SIM) as criterion groups across commonly used FDS cutoffs (≥ .30, ≥ .50, and ≥ .65) and alternatives. We repeated the process for the random responding scale (RRS). The above analyzes were performed using IBM SPSS 25. Data can be retrieved from the OSF platform (https://osf.io/6xksf/).

Results

Fig. 1 shows the distribution of Turkish IOP–29 FDS scores for the three feigning sub-conditions (SIM-SCZ, SIM-DEP, and SIM-PTSD). One-way ANOVA revealed no difference in FDS score: F(2, 122) = 0.052, p = .949. This was further tested by a Tukey-corrected post hoc comparison, which also showed no independent significant difference (ps > .05; gs between .03 and .07). Therefore, the three sub-conditions were combined into one main feigning condition (SIM). The conditions HON, RND, and SIM formed the within-subjects factor. Table 2 shows the mean FDS scores for each main condition and the feigning sub-conditions along with the results of two previous studies that used the same design.

Fig. 1
figure 1

Graphical representation of Turkish Inventroy of Problems–29 (IOP–29) false disorder probability scores (FDS) across the three feigning sub-conditions (between-subjects factor)

Full size image
Table 2 Descriptive statistics for Turkish IOP-29 FDS across all conditions, compared to Giromini et al. (2020b) and Winters et al. (2020)
Full size table

Comparison of FDS values with repeated measures ANOVA between HON, RND, and SIM conditions was significant: F(2, 248) = 265.0, p < .001, η2p = .68 (extremely large). FDS scores were higher in SIM (M = .80, SD = .19) compared to HON (M = .25, SD = .22; g = 2.68, very large effect) and RND (M = .66, SD = .21, g = 0.70, medium-large effect). The mean FDS score was higher for RND than for HON (g = 1.91, very large effect; Table 3).

Table 3 One-way repeated measures ANOVA outcomes for the main within-subject factor conditions HON, RND, and SIM (N = 125)
Full size table

The FDS achieved an excellent AUC of .95 (SE = .01) in discriminating between HON and SIM. The standard cutoff (FDS ≥ .50) correctly classified 88.0% of HON (specificity) and 91.2% of SIM (sensitivity). For the feigning sub-conditions (i.e., SIM-SCZ, SIM-DEP, and SIM-PTSD), sensitivity ranged from 90.2% to 92.9%. Use of the liberal cutoff (FDS ≥ .30) resulted in the predictable tradeoff between decreased specificity (71.2%) and improved sensitivity (97.6%). Conversely, the more conservative cutoff (FDS ≥ .65) achieved high specificity (93.6%) at the expense of sensitivity (81.6%; Table 4). The sequence of the experimental conditions was not related to the FDS. Table 5 shows the positive predictive power (PPP), negative predictive power (NPP), and overall correct classification (OCC) at different cutoffs and base rates for the Turkish IOP–29 FDS.

Table 4 Classification accuracy of the Turkish IOP–29 for the main conditions HON and SIM, and the feigning sub-conditions (N = 125)
Full size table
Table 5 Positive predictive power (PPP) and negative predictive power (NPP) at different cutoffs and base rates for the Turkish IOP-29 FDS
Full size table

Classification Accuracy of the RRS

RRS scores across feigning sub-conditions did not differ significantly: F(2, 122) = 2.33, p = .101 (Fig. 2). This was further tested by a Tukey-corrected post hoc comparison, which also showed no independent significant difference (ps > .05; gs between 0.23 and 0.43). However, a very large main effect was found for the main conditions (HON, RND, and SIM): F(1.854, 229.875) = 74.7, p < .001, η2p = .38.

Fig. 2
figure 2

Graphical representation of Turkish Inventroy of Problems–29 (IOP–29) random responding scores (RRS) across the three feigning sub-conditions (between-subjects factor)

Full size image

The highest RRS scores were obtained in the RND condition (M = 67.2, SD = 9.7), followed by SIM (M = 55.3, SD = 11.2) and HON (M = 54.0, SD = 8.0). As expected, random responding (RND condition) was associated with higher scores relative to honest responders (HON condition; g = 1.48, very large) and SIM (g = 1.13 large). The contrast between HON and SIM conditions was not significant (p = .837).

The recommended RRS cutoff (T ≥ 61; Giromini et al., 2020c) correctly identified 79.2% of HON (specificity) and 71.2% of SIM (specificity). For the RND condition, the sensitivity was 80.8%. Because there was no difference between HON and SIM in RRS scores, we combined the data from HON and SIM and applied the recommended cutoff T ≥ 61. This resulted in a specificity of 75.2%. Comparison of RND with HON, SIM, and the combination of these two conditions (i.e., non-random responding) resulted in an AUC of .85 (SE = .03), .80 (SE = .03), and .83 (SE = .02), respectively.

The cutoff value T ≥ 61 was recommended to balance false positives and false negatives. However, we explored alternative cutoff values based on the commonly used specificity standards of 90% and 95%. As with the recommended cutoff value (T ≥ 61), we combined the data to examine these alternative cutoff values. Examination of the AUC results showed that T ≥ 66.5 and T ≥ 71.0 for the HON and SIM conditions yielded a specificity of 90% and 95%, respectively. In the present sample, the specificity was 90.4% at T ≥ 66.5 and 96.0% at T ≥ 71.0. Table 6 shows the PPP, NPP, and OCC at different cutoffs and base rates for the Turkish IOP–29 RRS.

Table 6 Positive predictive power (PPP) and negative predictive power (NPP) at different cutoffs and base rates for the Turkish IOP-29 RRS
Full size table

FDS and RRS were positively correlated in the RND condition [r(123) = .43, p < .001], but not the HON and SIM conditions (r = .16 and r = -.10, respectively; p > .05).

Discussion

There is a growing consensus that NRB can undermine the validity of psychological assessments and therefore should be carefully monitored in both forensic and clinical contexts (e.g., Sherman et al., 2020; Sweet et al., 2021). The credibility of clinical presentation should be assessed by combining multiple sources of information, which include psychometric testing (Bush et al., 2014; Schutte et al., 2015; Sweet et al., 2021). In addition, it is important that SVTs are cross-validated in the target population in which they are used (e.g., Nijdam-Jones & Rosenfeld, 2017). The availability of SVTs in non-English speaking populations in general (Crișan, 2023) and in Turkish in particular (Ayhan & Karaman, 2021) is remarkably limited.

To fill this knowledge gap, this study reports the results of the initial validation of the Turkish version of the IOP–29, a free-standing SVT. The main objective was to investigate the cross-cultural validation of the IOP–29 (e.g., Viglione & Giromini, 2020) for Turkish speakers. We compared the response sets of 125 participants who were asked to complete the IOP–29 three times according to different instructions: Standard (HON), random responding (RND), and feigning a psychiatric disorder (SIM). In the SIM condition, participants were randomly assigned to SIM-SCZ, SIM-DEP, or SIM-PTSD. This design allowed us to compare the sensitivity of the Turkish IOP–29 to feigning different psychiatric disorders. Notably, the IOP–29. was designed to be applicable to a wide range of symptom presentations rather than symptoms belonging to a single, specific diagnosis (e.g., schizophrenia), which promotes ecological validity and relevance in a real-world setting (forensic and/or clinical) where engagement in NRB is typically associated with a variety of symptom presentations or combinations thereof (Giromini et al., 2020b; Viglione et al., 2019). This design therefore also allowed us to test the cross-cultural validity of this specific feature of IOP–29. Finally, we tested the classification accuracy of the RRS – an index specifically designed to distinguish random responding from feigned or honest presentations.

The results converge on a number of conclusions. As predicted, the Turkish IOP–29 was equally sensitive to the feigning of different psychiatric disorders: The standard cutoff (FDS ≥ .50) was uniformly sensitive (90.2% to 92.9%). This result is consistent with previous studies using the same design (e.g., Giromini et al., 2020b), suggesting that the classification accuracy of the IOP–29 is stable across different languages of administration, feigned clinical constructs, and geographic regions.

Consistent with previous reports (Giromini et al., 2020b; Šömen et al., 2021) and our a priori prediction, the mean FDS was lower for HON than for SIM (g = 2.68, extremely large effect). Compared with Winters et al. (2020), the FDS for the HON condition was higher (.14 versus .25; g = .60, medium effect) and more variable (SD of .14 versus .22), whereas the FDS for SIM was similar (M of .80 versus .82; SD of .18 versus .19), suggesting that the presentation of feigned psychiatric disorders may be culturally invariant despite significant differences in honest responding (HON). Classification accuracy (88.0% specificity at 91.2% sensitivity) of the standard FDS cutoff (≥ .50) was similar to values reported in previous studies with similar designs (Banovic et al., 2022; Carvalho et al., 2021; Giromini & Viglione, 2022; Ilgunaite et al., 2022).

As hypothesized, the average FDS in RND was intermediate between the values of the HON and SIM conditions but closer to the latter (Table 2), consistent with the reports of Giromini et al. (2020b), despite the differences in experimental conditions. We instructed participants to respond randomly without considering item content, whereas Giromini et al. (2020b) asked their participants to feign a psychiatric/neuropsychological disorder and to respond randomly (i.e., randomly/uncooperatively) and compared them with an additional sample instructed to respond completely at random (similar to the present study). The latter group yielded a mean FDS of .50 (g = .63, medium effect), which was lower and less variable than our results. It has long been known that people have difficulty producing truly random responses (Towse & Neil, 1998; Wagenaar, 1972); the present results suggest additional cross-cultural variability in the construct of randomness. From a practical perspective, the IOP–29 is likely to flag random response sets as noncredible, even though such profiles contain weaker psychometric evidence of invalid responding than individuals who intentionally present themselves as having a serious mental disorder.

The RRS, a novel alternative outcome measure to distinguish between random responders, feigners and honest responders, performed well in this sample (g: 1.13–1.48, large effects). However, the proposed cutoff of T ≥ 61 to balance false positives and false negatives (Giromini et al., 2020c) yielded lower specificity than Winters et al. (2020): 79.2% vs. 96.7% for HON and 71.2% vs. 84.1% for SIM. This discrepancy further highlights possible cultural differences in the conceptualization of random responding. Raising the cutoff to T ≥ 66.5 resulted in a specificity of 90.4% for the combined sample of HON and SIM, with a sensitivity of 61.6%. This classification accuracy resonates with the Larrabee limit – the seemingly inescapable tradeoff in validity testing that states that if specificity is set at 90%, sensitivity tends to hover around 50% (Crişan et al., 2021; Lichtenstein et al., 2019). Setting an even more conservative cutoff value (T ≥ 71) resulted in a further improvement in specificity (96%), but at the expense of sensitivity (30.4%). The fact that FDS and RRS were positively correlated in the RND condition, but not HON and SIM, provides preliminary evidence of the convergent and divergent validity of RSS. Although the target construct of RSS (distortions unrelated to item content) is clinically relevant, further research is needed to clarify its differential sensitivity to specific confounding factors (e.g., difficulty in comprehension, distraction, boredom, fatigue, etc.).

Limitations

Inevitably, the study has a number of limitations. First, the results of simulation designs have low external validity (Rai et al., 2019; Rogers & Bender, 2018). Examinees in the real world have much stronger incentives to successfully feign and may use different strategies (Abeare et al., 2022; An et al., 2017). Nevertheless, our results are consistent with previous research in Italy and the United Kingdom based on the same research design (Giromini et al., 2020b; Winters et al., 2020), suggesting that the Turkish, Italian, and English versions of the IOP–29 have similar classification accuracy when tested under the same conditions. Second, exclusion of participants with psychiatric/neurological disorders was based on self-report that could not be independently verified. Additional screening measures for specific disorders would have improved the diagnostic purity of the sample (but also increased the time required of participants). Third, although we employed manipulation checks to verify compliance with the instructions, the true fidelity with which instructions were executed could not be determined with certainty (Abeare et al., 2021). Finally, the vignettes used in this study likely did not elicit the motivation that is typical of real malingerers, which weakens the ecological validity of the experimental design.

Conclusion and Future Directions

The results suggest that the Turkish version of the IOP-29 is effective in discriminating between credible and non-credible presentations, that it is comparatively sensitive to a variety of feigned psychiatric disorders, and that the RRS has incremental validity in discriminating between random responses and feigned as well as honest responses. However, replication in psychometrically defined criterion groups based on genuine patients with a wider range of diagnoses (including neuropsychological disorders) is needed to establish its ecological validity. Incorporating the newly developed memory module (IOP–M; Giromini et al., 2020d; see also Erdodi et al., 2023) in future studies could further expand the clinical utility of the instrument among Turkish-speaking examinees (e.g., Brockhaus & Pecker, 2003), as there is ample evidence of the efficient utility of combining SVTs with performance validity tests (PVTs; see Holcomb et al., 2022).

As mentioned above, one of the main limitations of our study is the lack of a clinical sample and the use of a simulation design. Regarding the lack of a clinical sample, the use of nonclinical controls as opposed to real patients has the effect of likely overestimating the specificity results. The use of a simulation design may have the effect of maximizing internal validity, but external validity is questionable. In other words, there is no guarantee that subjects would actually respond similarly to our experimental participants in real assessments. Because of the impact of simulation designs on classification accuracy, it is of utmost importance for Turkish IOP research to switch to criterion group sampling (or known group sampling) rather than simulation designs in the future.

Overall, the combination of its brevity, ease of administration/scoring, and the stability of its classification accuracy across different countries and language communities makes the IOP–29 a cost-effective and empirically supported choice of SVTs.