We conducted the study in German language and online, hosted on the platform SoScisurvey (www.soscisurvey.de; 18) from February 2 to March 3, 2022. The study design and methods were preregistered, and the protocol is available at https://osf.io/er43j. Study procedures were approved by the ethics committee of the medical faculty Mannheim at Heidelberg University (protocol no. 2021-524). All participants provided informed consent prior to their participation.
Participants
We advertised the study as investigating the association between personality traits and decision-making. As we wanted to assess a wide range of maladaptive trait levels, including participants with high maladaptive trait levels, we recruited through Facebook groups, Instagram pages, and online forums dedicated to personality disorders and mental health topics in general. Additionally, we sent invitation emails to individuals that participated in previous studies on borderline personality disorder and childhood maltreatment within our clinic. Additionally, we recruited participants via student mailing lists at the University of Mannheim. Based on a-priori power-analyses, we pre-registered a requited sample size of N = 390 (for details, see pre-registration protocol at https://osf.io/er43j). Since we pre-registered a number of data exclusion criteria (e.g., repeated participation, careless participation) to ensure data quality, we planned to oversample by approximately 15%. Eventually, a total of 460 individuals participated in the online study. Based on pre-registered exclusion criteria, we excluded nine participants (1.96% of the sample) because they spent less than two seconds on answering each item of the PID-5 questionnaire, four participants (0.87%) because they indicated that we should not use their data due to bad quality, and two participants (0.43%) because of repeated participation.
After applying the exclusion criteria, the final sample comprised 445 participants. Of these, 253 participants (56.85%) were recruited via advertisements in Facebook groups, online forums, Instagram channels, and websites dedicated to mental health topics. In addition, 81 participants (18.20%) who had participated in previous studies of borderline personality disorder and childhood maltreatment in our clinic were recruited via mailing lists. The remaining 111 participants (24.94%) were recruited via advertising in non-clinical settings (i.e., University of Mannheim mailing lists and snowball method on social media). The age of participants in the final sample ranged from 18 to 73 years (M = 33.79, SD = 13.17). The diversity of the sample was limited in terms of gender, nationality, and migration background, as most participants were women (n = 335, 75.28%), had a German nationality (n = 427, 95.96%), and had no migration background (n = 374, 84.04%). Furthermore, most participants had a university entrance qualification or university degree (n = 289, 64.99%) and were currently employed (n = 180, 40.45%) or studying (n = 112, 25.27%). Almost all participants were native German speakers (n = 431, 96.85%). Participants indicated an average monthly net income of 1509.00€ (SD = 1344.94), ranging from 0€ to 9500€. A total of 54 participants (12.13%) identified as belonging to a minority group (in detail, 12 participants (2.70%) due to belonging to the queer community, 9 participants (2.02%) due to mental disorders, 5 participants (1.12%) due to religion and 28 participants (6.29%) due to other reasons). Detailed demographic data are shown in Table 1.
Regarding clinical characteristics, 202 participants (45.39%) indicated that they had been diagnosed with at least one mental disorder in the past. The most frequently self-reported diagnoses were mood disorders (n = 120, 26.97%) and personality disorders (n = 91, 20.45%). Details of these self-reported clinical characteristics are shown in Table 2.
Procedure
After providing demographic information, participants were asked to complete the German version of the PID‐5 (19) and the Level of Personality Functioning Scale-Brief Form (LPFS-BF, 20; the LPFS-BF data will be used as pilot data for another project. Next, they participated in two economic games – either the dictator game or the faith game, in randomized order. We implemented five versions of the faith game with varying sure choice pay-offs, and each participant played one version that was randomly selected. After completing the first game, participants rated a series of positive and negative interpersonal statements that served as a filler for the gap between the two games and that will be used in a different project. Following this, participants completed the second game. If participants played the faith game first, they were not informed about the payoff of the faith choice (if they chose it) until after they had completed the dictator game, to avoid spillover effects.
Both games were partially incentivized. This means that, for some participants, we actually paid out the money they earned in the dictator game and the faith game at the end of the study via an amazon voucher. The voucher was automatically displayed on the last study page, so complete anonymity was guaranteed to participants (i.e., they did not have to contact the experimenter or leave any identifying information to receive the voucher). Participants were informed that “every 10th player” would win their actual outcome. Randomly, 47 participants (10.22%) were selected to receive their actual payoff. This payoff could vary between 2€ (if participants only kept their initial endowment in the dictator game and received 0€ in the faith game, when choosing the faith choice) and 18€ (if they took all the money in the dictator game, thus earning 10€ there, and if they received 8€ via the faith choice in the faith game).
Materials
Personality Inventory for DSM-5: We assessed maladaptive personality traits using the German version of the PID‐5 (19). In detail, we used the translation of the 100-item short form (PID-5-SF, 21), which covers 25 maladaptive personality trait facets nested within the five DSM-5 maladaptive trait domains negative affectivity, detachment, antagonism, disinhibition, and psychoticism. To test the hypotheses described above, we used the domain score for antagonism, which is a mean score calculated for all items within this dimension (based on 24 items). As the 100-item short form includes only four items for the PID-5 facet suspiciousness, we added three additional items for suspiciousness taken from the 220-item long form of the PID-5 and calculated the mean score for suspiciousness including all seven items. Internal consistencies for the PID-5-SF dimension antagonism (α = 0.90) and PID-5 facet suspiciousness (α = 0.89) were high in this sample.
Dictator Game: In the dictator game (10), participants (in the role of the allocator, player A) played a hypothetical game for money with an unknown interaction partner (player B). Each participant was informed that a subsequent participant in the same study would assume the role of player B, and that the exchange would be completely anonymous. In the present study, we used a modified dictator game (i.e., the instructions differed somewhat from the classic implementation of the game as described by Forsythe et al., 10), previously used by List (11). Participants were informed that they would play the role of the allocator (player A). They received 2€ by the experimenter as an initial budget, and we informed them that the other person (player B) received 8€. Participants were then told that they could take as much as they wanted from player B’s money stash and keep it for themselves (in addition to their own 2€). They were also informed that player B would have no way to react to their decision and that both player A and B would remain completely anonymous. In the role of player A, they then decided which amount of money they wanted to subtract from player B’s money stash (0€-8€). The amount of money participants deducted from player B’s money stash was used as an indicator of prosocial behavior (inverted, i.e., higher amounts taken away indicated higher levels of selfish behavior / lower prosociality). The maximum amount of money participants could earn in this game was 10€ (if they subtracted the entire 8€ from the recipient), and the minimum amount was 2€ (if they took nothing from the recipient and kept only their initial endowment). The dictator game did not include any within- or between-person conditions.
Faith Game: In the faith game (17), participants were assigned the role of player B. They were informed that a previous participant had been in the role of player A and left money in an envelope for player B. Participants (in the role of player B) did not know the amount of money that player A had left for them (the amount could be between 0€ and 8€). Participants were asked to decide whether they wanted to receive the amount that player A had left for them (faith choice) or whether they wanted to receive a fixed amount of money from the experimenter (sure choice). These fixed amounts were randomly varied at five levels (2€, 2.5€, 3€, 3.5€, 4€) as a between-participants factor. The resulting outcome variable is dichotomous and indicates whether the participant chose the sure choice or the faith choice. Choosing the sure choice indicates less trust because of lower trust in the prosociality of others. The maximum amount of money participants could receive was 8€ (if they chose the faith choice and player A had left the entire 8€ to them), the minimum amount of money they could receive was 0€ (if they chose the faith choice and player A had left them 0€). For exploratory analyses, participants were asked to guess the amount of money in the envelope of the faith choice (amount of money that a previous participant left for them).
Data Analysis
All data and analysis code is available at https://osf.io/2rvbg/. Data were analyzed using R (22) and all analyses were pre-registered (https://osf.io/er43j). To test our hypotheses, we conducted a linear multiple regression and a logistic regression model. In all models, we adjusted for the pre-registered covariate gender (dummy-coded, 0 for women [trans and cis] and 1 for men [trans and cis]), because meta-analytic results suggest cooperative behavior may vary as a function of gender (23). All analyses were calculated on a sample of n = 428 participants, because the remaining 17 participants did not indicate their gender identity or reported a gender identity other than cis/trans male/female[1].
To test H1, we entered the PID-5-SF dimension antagonism (mean score) and the covariate gender into a linear regression model to predict prosocial behavior in the dictator game (i.e., the amount of money taken). To test H2, H3, and H4, we conducted a logistic regression model with the likelihood of choosing the sure choice (versus the faith choice) as outcome. Choice was coded as 1 for the sure choice and 0 for the faith choice. As predictors, the model included the PID-5 facet suspiciousness (centered on the grand mean) and the amount of money offered as the sure choice (varying between-participants: 2€, 2.50€, 3€, 3.50€, 4€, centered on the grand mean), as well as their interaction, and the covariate gender. The main effect of suspiciousness constitutes the test of H2. The main effect of the amount of money offered as the sure choice constitutes the test of H3. The interaction term constitutes the test of H4.
[1] See Table 1 for detailed data on gender identities. Due to the small number of participants identifying with a gender other than cis/trans male/female, we decided not to add an additional dummy for this group. However, when including these participants and adding a second dummy variable to the model, results replicated (see supplemental Tables S1-S4).