In this study, we explored whether the GIBSO can detect underreporting when individuals are simply exposed to minimal, indirect prompts, while also including corresponding clinical scales to account for the broad psychological, somatic, and behavioural domains the instrument gauges. All participants were told to respond honestly, but a randomly subset also received subtle cues suggesting that their answers would be used to evaluate personal qualities such as parenting ability or risk of harmful behaviour. We included four experimental conditions (non-prime, caretaker prime, safety prime, and addiction prime) and administered questionnaires that tap into mental (depression, anxiety, and stress) and physical (somatic) health, and into antisocial behaviour (drug use and aggression). This design allowed us to assess whether the GIBSO is sensitive to both spontaneous and subtly prompted underreporting. We then compared group scores on the GIBSO and clinical scales, anticipating that prime groups would exhibit significantly higher GIBSO scores than unprimed participants. Furthermore, underreporting tendencies were expected to manifest as negative correlations between the GIBSO and these scales. As a subsidiary aim, as our participants could complete the study in English or in Dutch, we examined whether the English and Dutch versions of the GIBSO generated comparable results.

Initially, 392 student participants and members of the general public joined the study, but records of participants who 1) did not provide informed consent (n = 4), or permission to use their data (n = 11), 2) provided incomplete reports (n = 55), and/or 3) failed the attention checks (n = 17, see Materials and Measures section) were excluded. The final sample consisted of 305 participants, of whom the majority were women (n = 227, 74.4%) in their mid-twenties (M = 26.17, SD = 11.91, range = 18-72). The sample was collected primarily in the Netherlands but also across different countries. The most frequently reported nationalities were Dutch (n = 208, 66%), German (n = 23, 8%), Spanish (n = 17, 6%), Turkish, Polish, and Indian (< 3%). International participants responded in English (n = 196), and rated their English proficiency as relatively high (M = 3.94, SD = .76, range = 2-5). The rest of the sample participated in their native language, Dutch (n = 109). All participants rated their overall physical and mental health using a 5-point scale (higher values being indicative of better self-reported health), and their ratings were indicative of moderately good health (M_physical = 3.68, SD_physical = .91; M_mental = 3.36, SD _mental = .93; range = 1-5).

Participants were randomly allocated to one of four conditions: non-prime (n = 76), caretaker prime (n = 75), safety prime (n = 77), and addiction prime (n = 77). Groups did not differ in terms of age, nor in (mental and physical) health ratings, Fs < .83, ps > .48, η² < .008, nor in education, H(3) = 4.62, p = .202.

Questionnaire order was randomized. Overall, the measures were selected to tap into psychological (DASS-21) and physical (Adapted PHQ-15) health, antisocial behaviour (BPAQ and DAST-10), and into underreporting tendencies (GIBSO). Attention checks were embedded in all measures except the GIBSO (which contains built-in validity checks) and were tailored to match each scale’s response format (e.g., instructing participants to select a little or sometimes at specific places). Depending on their preference, participants either received English or Dutch versions of the measures.

The study was conducted online, using Qualtrics, during the period from March until July 2025. Non-students were recruited through snowballing globally, while students were recruited by a student research platform at our institution. After they had followed a link, participants were asked to provide informed consent, respond to demographic questions, and rate their health. Participants were then randomly allocated to one of four conditions and they received the corresponding instructions. Following this, they were administered the questionnaires in a randomized order. Afterward, participants were asked to evaluate the study and their participation. Finally, they were debriefed and asked to provide permission to use their responses. Participants were either rewarded with research credits (when students) or were entered in a raffle for a 10-euro prize. The study was approved by the standing ethics committee of Erasmus University Rotterdam, the Netherlands.

The data were analysed using the SPSS statical package (version 30.0). We first ran descriptive analyses. Then, we examined potential group differences on all measures using Multivariate Analyses of Variance (MANOVAs) and Bonferroni post-hoc pairwise comparisons. Eta squared (η²) and Cohen’s d were used as measures of effect size, and interpreted according to Cohen’s (1988) guidelines (η² ≈ .01 small, .06 medium, ≥ .14 large; d ≈ 0.20 small, 0.50 medium, 0.80 large). Pearson’s r coefficients were employed to compute correlations between inventories. Beyond correlations, the influence of demographics was evaluated using Welch’s t-tests, and differences between the English and Dutch GIBSO versions were assessed using independent-samples t-tests. We also conducted exploratory ROC analyses evaluating the diagnostic utility of the GIBSO cutoff score using low-scoring participants across most symptom inventories as a criterion group. However, as these analyses were purely exploratory, they were excluded from this manuscript to maintain focus (see Supplemental file). These results, along with the raw data and analytic outputs, are available on the Open Science Framework (OSF) at: https://osf.io/sg7pq/overview.

Participants rated their motivation, clarity of the instructions and questions, and difficulty of their task using a 5-point scale (range from one to five). On average, participants’ (N = 305) motivation was moderate (M = 3.55, SD = 0.87), and clarity of the instructions (M = 4.48, SD = 0.66) and items (M = 4.31, SD = 0.78) was relatively high, while the task was perceived as easy (M = 1.52, SD = 1.00). There were no significant differences between four groups with regard to any of these ratings, F(4, 296) = 0.44, p = .95, η² = .006.

Overall, participants obtained relatively low scores on the Depression and Anxiety subscales of the DASS-21 and moderate scores on its Stress subscale. Participants also self-reported low levels of somatic complaints on the PHQ-15. The groups did not significantly differ with regard to DASS-21 or PHQ-15 ratings, λ = .97, F(12, 788.72) = 0.72, p = .73, η² = .01. Participants reported relatively low use of substances on DAST-10, but moderate level of aggression on BPAQ. Again, groups did not significantly differ with respect to the ratings on these two tests, λ = .97, F(6, 598) = 1.20, p = .304, η² = .012, suggesting that primes did not promote underreporting. For descriptives and univariate results see Table 1.

Whether or not participants had been exposed to subtle primes did not affect their GIBSO scores; this was true for both full scale scores and the subscale scores, λ = .98, F(6, 600) = 1.03, p = .403, η² = .010. Univariate tests confirmed that there were no significant differences between groups (see Table 1). When applying the GIBSO cutoff score of > 20, 28% of participants exhibited underreporting. A similar rate was found when applying the General Underreporting cutoff point of > 17 (29%). There were no significant differences in these prevalences across groups.

Considering the lack of group differences, we collapsed the groups and, using the total sample (N = 305), calculated Pearson product-moment correlations of GIBSO full and subscale scores, symptom inventories, self-reported drug use and aggression, and self-reported health (see Table 2).

Both the Total and the GU GIBSO scores correlated significantly and negatively with scores on other inventories, and positively with self-reported health ratings (see Table 2). The C subscale showed only one significant (medium and positive) correlation, with self-reported aggression. The observed correlations were replicated when the analyses were conducted for the English and Dutch subsamples, separately (see Supplemental Table 1).

We examined potential demographic influences by correlating GIBSO scores with participants’ age (see Table 2), and evaluating gender differences. Participants’ age was modestly, positively associated with GIBSO Total and GU scores, whereas its correlation with the C scale was not significant. These findings show that older participants endorsed more items on GIBSO. Furthermore, gender differences were evaluated using Welch’s t-tests due to unequal group sizes (n_women = 227, n_men = 72). Although men scored descriptively higher across all scales, these differences were not statistically significant for the GU, t(106.70) = 1.45, p = .15, Cohen’s d = .21, or C scales, t(117.47) = 1.53, p = .13, Cohen’s d = 0.21. A similar pattern emerged for the Total score, yielding a marginal p value and a small-to-medium effect size, t(107.24) = 1.93, p = .06, Cohen’s d = 0.28. These findings suggest that older men in our sample might have shown slightly stronger tendency towards underreporting, which warrants further investigation.

We examined whether GIBSO scores differed between the two language groups (English vs. Dutch). The international sample and the Dutch sample obtained similar GIBSO Total, t(303) = .19, p = .85, Cohen’s d = 0.02, and GU scores, t(303) = 1.37, p = .17, Cohen’s d = 0.17; see Table 3. The international participants did, however, report more resistance to help-seeking and disclosure than the Dutch participants on the C subscale, t(303) = 3.27, p = .001, Cohen’s d = 0.39.

Given the different group sizes, we repeated the analyses using the Mann-Whitney U test, and the results did not change. It is also worth noting that the Dutch participants (M = 29.80, SD = 14.24, range: 18-72) were significantly older than those in the international sample (M = 24.15, SD = 9.87, range: 18-63; t(303) = 3.67, p < .001, Cohen’s d = .48). Therefore, we repeated the analyses using GIBSO scores while controlling for age, but this did not change the pattern, with the difference on the C scale remaining the only significant one (p < .001).

A methodological strength of this study lies in our attempt to employ more ecologically valid instructions rather than the usual simulation scenarios. However, our subtle manipulation may have been “too” subtle, as the primed and non-primed groups exhibited similar scores across all measures. Two plausible explanations exist for this outcome. First, participants may have engaged in spontaneous underreporting regardless of their assigned condition, a notion consistent with the elevated minimization tendencies already observed in the sample. Alternatively, the lack of group differences may reflect limitations in the GIBSO’s sensitivity to subtle shifts in underreporting, rather than an ineffective prime. To resolve this ambiguity, future studies should consider experimental manipulations that are highly self-relevant and evoke a stronger, explicit motivation to underreport.

A second strength is the composition of our sample; while students comprised the majority, we also included members of the general public to enhance generalisability. Nonetheless, this heterogeneous composition may have introduced systematic noise. Previous research indicates that students are typically more inclined to disclose health issues (e.g., 13% in Bošković et al., 2024; 20% in Bošković & Rowlands, 2025). Consequently, the higher underreporting rates observed in the current sample might reflect the inclusion of these general population participants. Future validation efforts would benefit from testing samples composed exclusively of general-public participants to isolate these baseline reporting styles.

Furthermore, because the GIBSO is one of only a few scales developed specifically to capture underreporting, empirical investigations into its psychometric boundaries are exceptionally valuable. Although the significant negative correlations with symptom measures replicate previous findings and testify to the construct validity of the GIBSO, this is the first study to compare different language versions of the instrument. While we observed no mean differences between the English and Dutch versions, this absence of significant differences does not in itself demonstrate measurement equivalence. Because our study was not statistically powered or designed for such analyses, formal tests of measurement invariance (e.g., Putnick & Bornstein, 2018; Wang et al., 2021) remain an essential next step across different linguistic, cultural, and demographic groups. Importantly, language mastery also varied within our sample: English-language respondents differed in their proficiency, whereas all Dutch-language respondents were native speakers. Further research should evaluate language equivalence using groups that include both native and non-native speakers of the same target language. This is a critical consideration for researchers implementing other active language versions of the GIBSO beyond those tested here, such as the Italian, French, German, Spanish, and Serbian translations.

Ultimately, future work should employ designs that allow for direct comparison between the GIBSO and other well-established instruments. By incorporating additional underreporting measures or external behavioural criteria, future research can establish the GIBSO’s incremental utility when administered alongside established validity scales, such as those embedded in the MMPI-2-RF (Paulhus, 2002; Picard et al., 2023). Still, our study provides provisional data on the English and Dutch versions of the GIBSO across non-clinical populations. This is particularly important given the limited number of tools available to practitioners screening for underreporting.

Our subtle priming instructions were too weak to evoke significantly different response styles among participants. Still, the negative associations between GIBSO scores and self-reported symptoms, drug use, and aggression underscore that the GIBSO captures a general tendency toward minimization across multiple domains. The results of this study offer preliminary but encouraging evidence for GIBSO’s utility, particularly for its shorter (GU) version, which may prove to be more robust across different cultural contexts. However, further systematic investigation remains necessary to establish the instrument’s psychometric precision while simultaneously clarifying the conceptual nature of underreporting across diverse cultural and demographic populations. Finally, it is important to note that the GIBSO is not designed to be the sole source of information for determining the presence of a positive response bias.