Dysfunctional expectations have been suggested as core features in the development and maintenance of mental disorders. Thus, preventing development and promoting modification of dysfunctional expectations through intervention might improve clinical treatment. While there are well-established experimental procedures to investigate the acquisition and modification of dysfunctional performance expectations in major depression, paradigms for investigating other important types of dysfunctional expectations (e.g. social rejection expectations) are currently lacking. We introduce an innovative associative learning paradigm, which can be used to investigate the development, maintenance, and modification of social rejection expectations.
A pilot sample of 28 healthy participants experienced manipulated social feedback after answering personal questions in supposed webcam conferences. While participants repeatedly received social rejection feedback in a first phase, differential feedback was given in a second phase (social rejection vs. social appreciation). In a third phase, explicit social feedback was omitted.
Participants developed social rejection expectations in the first phase. For the second phase, we found an interaction effect of experimental condition; i.e. participants adjusted their expectations according to the differential social feedback. In the third phase, learned social expectations remained stable in accordance to the social feedback in the second phase.
Results indicate that the paradigm can be used to investigate the development, maintenance, and modification of social rejection expectations in healthy participants. This offers broad applications to explore the differential acquisition and modification of social rejection expectations in healthy vs. clinical samples. Further, the paradigm might be used to investigate therapeutic strategies to facilitate expectation change.
This paradigm can be used to induce and modify social rejection expectations. This allows to investigate differences in expectation acquisition, maintenance, and modification between clinical vs. healthy samples. Further, this paradigm enables research on interventions promoting expectation modification.
Recent developments in clinical psychology propose dysfunctional expectations (i.e. future-directed ‘if-X-then-Y’-predictions,
Dysfunctional expectations have been shown to play a crucial role in mental health as they negatively impact future behaviour (e.g. excessive avoidance,
Further, dysfunctional expectations have been shown to impede important clinical outcomes (e.g. treatment success,
However, acquisition, maintenance, and modification of dysfunctional expectations is still little understood (
Since especially (dysfunctional) expectations of social rejection (e.g. ‘When I open myself to others, they will refuse me!”) have serious implications for mental health (e.g.
The aim of the current study was to develop an experimental social rejection expectation paradigm (No1LikesU!), which can be used to investigate the acquisition, maintenance and modification of social rejection expectations within a highly standardised and ecologically valid procedure. In contrast to existing paradigms on social exclusion (for an overview, see
This model proposes that when entering concrete situations, individuals form situation-specific predictions about these situations (drawn from more generalised expectations) which become either (a) confirmed or (b) disconfirmed by experience. While repeated expectation confirmations should stabilise or reinforce the original situation-specific prediction (or respectively, the underlying generalised expectation), repeated expectation ‘violations’ should entail its modification (
Following the predications of the model, we hypothesise that (1) repeatedly exposing healthy individuals to situation-specific experiences of social rejection will increase levels of social rejection expectations over time. Consistent with the ViolEx-Model, we further hypothesise that (2) repeatedly exposing healthy individuals with increased levels of social rejection expectations to situation-specific experiences of social rejection (‘Stabilisation’) vs. appreciation (‘Modification’) will lead to differential changes (i.e. to an increase vs. stabilisation) in social rejection expectation levels over time.
No1LikesU! is an ecologically valid and highly standardised associative learning paradigm created to model the development, maintenance, and modification of social rejection expectations. Like the O-Cam paradigm (
The local ethics committee (reference number 2018-36k) approved the study. All participants gave written informed consent before they started the experiment. This study was part of a parent study, which additionally investigates interventions for promoting the modification of dysfunctional expectations. In the present work, we focus on the effects of the paradigm on the development, maintenance, and modification of social rejection expectations in healthy participants.
We recruited participants via e-mail lists, flyers, and the research participation system of our university. Inclusion criteria were: (a) A minimum age of 18 years, (b) sufficient German language skills, (c) no severe visual impairment, (d) no serious physical illness, (e) no current psychological stress, (f) not in psychotherapeutic treatment, and (g) a sum score in Beck’s Depression Inventory II (BDI-II;
Until now, a pilot sample of 31 healthy participants could be included in the study, which provides sufficient power to investigate our hypotheses (
Participants received credit points as compensation for their participation. Alternatively, they got the chance to win gift vouchers for different online shops.
Testing sessions started with participants reading the study information and signing informed consent (see
Research assistants checked age as well as BDI-II cut-off scores. Participants who failed the inclusion criteria received partial compensation and were fully debriefed. Participants who met the inclusion criteria received study information incorporating the cover story. To allay concerns about the authenticity of the webcam conference, participants were told that their listeners (who were announced as ‘students from an experimental internship at the university’) were instructed ‘not to talk’ during conferences for ‘methodological reasons’. Afterwards, research assistants started the paradigm on the computer and left the experimental room. The participants were fully randomised into two independent experimental conditions (group ‘Stabilisation’ vs. group ‘Modification’) and followed instructions presented on the computer screen, which guided through the paradigm.
To model key processes of the ‘ViolEx-Model’, No1LikesU! encompasses multiple trials (30) which are divided into three different experimental phases (acquisition phase, stabilisation vs. modification phase, test phase, see
These phases are structurally based on fear conditioning paradigms (
After completing the paradigm, research assistants entered the experimental room and provided paper-pencil post-questionnaires to check for suspiciousness about the cover story and emotional distress due to participation. Participants were then fully debriefed about the true purposes of the study and the deceptions within the experimental manipulation. Testing sessions lasted between 1.0 and 1.5 hours.
Note that we applied additional questionnaires to address further research questions in the parent study, which we do not describe here.
We assed situation-specific social expectations using a one-item 7-point bipolar Likert scale (social expectation rating: ‘Please indicate to what extent you expect your next listener to be interested or disinterested in you!’) ranging from -3 (maximal disinterest) to +3 (maximal interest) before each trial. Thus, lower values indicate higher social rejection expectations.
To examine how participants actually perceived a passed webcam conference, we used a one-item 7-point bipolar Likert scale (social experience rating: ‘Please indicate to what extent you experienced interest or disinterest from your last listener!’) ranging from -3 (maximal disinterest) to +3 (maximal interest) after each trial. Thus, lower values indicate higher social rejection experiences.
We assessed depressive symptoms using the Beck Depression Inventory II (BDI-II;
We used a brief self-report questionnaire in order to assess demographic variables like sex, age, nationality, relationship status, educational level, employment status, and living situation.
We assessed emotional distress due to participation by asking whether participants felt impaired due to the experimental procedures (‘Do you feel impaired due to our investigation?’). Further, we applied a one-item 5-point bipolar Likert scale (‘Please indicate to what extent you feel positive or negative in this moment!’) ranging from -2 (very negative) to +2 (very positive) to assess emotional distress. Higher values indicate lower emotional distress due to participation.
In order to assess the credibility of the cover story, the video stimuli and the experimental manipulation, we asked participants whether 1) they knew any of their ‘webcam partners’, 2) what they believed was the aim and purpose of the study, and 3) how they experienced the experimental procedure. Responses were rated on a 3-point Likert scale ranging from 0 ("not suspicious at all") to 2 ("doubted the authenticity of the webcam conferences").
Participants were seated in front of a computer with an external microphone and a webcam connected to the computer. The paradigm, including instructions, video stimuli, and social feedback, was presented on the computer screen. Participants used a mouse to interact with the computer. Video stimuli were pre-recorded with 30 volunteers (15 male, 15 female, age: 25 – 35). Volunteers were instructed to express nonverbal cues of either social rejection or social appreciation (see
Before conducting the analyses, we checked for outliers to exclude influential data points. For each expectation rating, we calculated the Mahalanobis distance which we checked against a χ2-cut-off of α = .001. We found no influential data points.
All analyses were computed using
Subsequently, we analysed the phases individually to estimate effect sizes for each phase effect. We used linear models to investigate the relationship between social expectations and group affiliation. We entered Group as fixed effect. We inspected the residual plot to check homoscedasticity and normality. Again, all plots showed patterns as expected. For all analyses, we applied sum contrasts to calculate intercepts and slopes.
Participants were predominantly young (
Variable | Stabilisation |
Modification |
Difference between experimental conditions |
---|---|---|---|
Age in years, |
21.79 (3.02) | 25.00 (8.57) | |
Sex, |
χ2 = 2.19, |
||
Male | 4 (28.57) | 1 (7.14) | |
Female | 10 (71.43) | 13 (92.86) | |
Nationality, |
χ2 = 0.37, |
||
German | 13 (92.86) | 12 (85.71) | |
Other | 1 (7.14) | 2 (14.29) | |
Romantic relationship, |
χ2 = 1.29, |
||
Yes | 5 (35.71) | 8 (57.14) | |
No | 9 (64.29) | 6 (42.86) | |
Living situation, |
χ2 = 0.01, |
||
Living alone | 2 (14.29) | 2 (15.38) | |
Living with others | 12 (85.71) | 11 (84.62) | |
Educational level, |
χ2 = 1.71, |
||
University degree | 2 (14.29) | 5 (35.71) | |
No university degree | 12 (85.71) | 9 (64.29) | |
Employment status, |
χ2 = 1.47, |
||
Employed | 6 (42.86) | 3 (21.43) | |
Not employed | 8 (57.14) | 11 (78.57) | |
BDI-II sum-score, |
4.86 (3.44) | 4.21 (3.17) | |
MSER before first trial, |
3.86 (1.29) | 4.21 (1.12 | |
Emotional distress after participation, |
3.21 (0.70) | 3.57 (0.65) |
aOne missing data point.
We investigated whether the nonverbal social feedback (rejection vs. appreciation) displayed in the videos affected the situation-specific social experience ratings of the supposed webcam conferences. Participants provided these ratings after each conference and before receiving written social feedback.
First, we performed a mixed ANOVA using a linear model of the mean social experience ratings as a function of Group (between factor) and Time (within factor) using Greenhouse-Geisser correction. We found a significant interaction of Group and Time (
Next, we performed post-hoc analyses and pairwise comparisons to further analyse the significant interaction effect.
The Bonferroni adjusted
Regarding the main effect of Time, the Bonferroni adjusted
First, we included all experimental phases in one statistical model and investigated changes in social expectation ratings across the course of the experiment. Therefore, we performed a multilevel mixed effect multinomial linear regression on the social expectation ratings as a function of Group and Time (i.e. the contrast matrix of individual social expectation ratings nested in each phase). Time therefore consists of three variables each representing an experimental phase (acquisition phase, modification vs. stabilisation phase, test phase). Unless otherwise stated, we used the standard bound optimisation by quadratic approximation (BOBYQA) optimisation for the models. We calculated the linear regression of the social expectation ratings as a function of Time (Level 1). We then subsequently added the next-level effects until arriving at the full model including Time (Level 1), random intercept for participant (Level 2), and Group with interaction term for Time (Level 3). We compared mixed-effects models using likelihood ratio tests. Here, we will describe the results of the Level-3-model, the results for the Level-1- and Level-2-models can be found in the
The Level-3-model revealed no significant Group x Acquisition Phase interaction (β = -.00,
Model | AIC | χ2 | χ |
|
---|---|---|---|---|
Level 2 (random effect for participant) | 2249.7 | – | – | – |
Level 3 (fixed effect for group) | 2243.8 | 13.87 | 4 | .007 |
Next, we used MANOVA tests to investigate the effect of Group on the social expectation ratings for each phase individually to investigate the effect sizes of the changes.
We constructed a linear model of the social expectation ratings (as outcome matrix for ratings 1 to 10) as a function of Group and Baseline Social Expectation Rating (with interaction term) to exclude differential learning for the groups and to account for inter-individual influences of baseline ratings on expectation rating during acquisition. We calculated a Type-II-MANOVA using Pillai’s test statistic for the linear model. As expected, we found no significant interaction between Group and Baseline Social Expectation Rating,
Following the significant interaction of Group x Stabilisation vs. Modification Phase in the main analyses, we constructed a linear model of the social expectation ratings (as outcome matrix for ratings 11 to 20) predicted by experimental condition to further investigate the main effect of Group. The Type-II-MANOVA revealed a marginally significant main effect for Group (
To test whether the social expectation ratings would remain consistent during test phase, we analysed a linear model of the social expectation ratings (outcome matrix for ratings 21 to 30) as a function of Group. As expected, the Type-II-MANOVA did not reveal a significant main effect for Group,
Additionally, we analysed suspiciousness of the cover story. Seven participants reported doubts about the authenticity of the webcam conferences, six reported that they felt something ‘was off’ while 15 participants found nothing wrong with the webcam conferences. Further, three participants knew some of their ‘webcam partners’. However, a sensitivity analysis excluding all suspicious participants did not reveal significant differences in the result patterns. Therefore, we based our results on the whole sample.
While social rejection expectations play a crucial role in mental health, experimental research on the processes of how these expectations develop, maintain, and change is currently lacking. Our study addresses this gap by providing an ecologically valid and highly standardised experimental paradigm to investigate the acquisition, maintenance, and modification of situation-specific social rejection expectations in healthy samples. Results indicate, that this paradigm can be used to successfully induce (Hypothesis 1) as well as differentially change (Hypothesis 2) situation-specific social rejection expectations in healthy participants as a function of social feedback (social rejection vs. social appreciation). Altogether these results are consistent with the predictions drawn from the ‘ViolEx-Model’, which assumes modification of expectations after experiencing disconfirming results (e.g. positive social feedback after negative social feedback) as well as stabilisation of expectations after experiencing confirming results (e.g.
Moreover, our results resemble basic result patterns found in fear conditioning paradigms concerning the acquisition and modification of fear (
Despite incorporating naturalistic stimuli, No1LikesU! does not provide dynamic social interactions. While the pre-scripted video stimuli ensure standardised experimental manipulation, these stimuli do not adapt to individual expressions of participants, threatening its external validity. Moreover, the paradigm only focuses on one specific social situation, i.e. self-disclosure in front of a stranger. Thus, investigating the generalisation of social rejection expectations to other social situations might be difficult within this paradigm. Additionally, a substantial amount of our participants seemed to be suspicious about the ‘webcam conferences’ and the social feedback we provided within No1LikesU!. While this issue could be solved at the expense of standardisation (for example by using real time interactions with confederates), problems with suspiciousness should not be overestimated within the actual procedure. Firstly, post-hoc questionnaires about the ‘aims and purposes’ of a study demand for suspiciousness by construction and therefore potentially overestimate actual suspiciousness of individuals during participation. Secondly, research on social exclusion shows that experiences of social exclusion stay impactful even if participants know that social feedback is simulated (e.g.
Further, while we incorporated general suggestions on fear conditioning paradigms, there are no clear instructions on how to set certain parameters in associative learning procedures (e.g. reinforcement rate or trial number). Thus, changing these parameters might also influence the effects of the paradigm.
Also, while we focused on contingency learning of outcome expectations, we did not include valence ratings for social rejection and social appreciation. Meta-analyses clearly show negative valence for social rejection (
Finally, while we incorporated the concept of ‘expectation violation’ (
No1LikesU! provides options for broad applications to investigate the acquisition, maintenance, and modification of social rejection expectations within a highly standardised and ecologically valid experimental procedure. It is adaptable to various research attempts. Future research should use No1LikesU! to identify differences in the development, maintenance, and modification of social rejection expectations between healthy and clinical samples (with special regards to patients with borderline personality disorder, social anxiety or depression). To test whether clinical samples show to be differentially more sensitive to social rejection experiences during acquisition than healthy controls and show to be less responsive to social appreciation experiences during modification, has important implications for etiological considerations and clinical treatment. On the one hand, this could call for the development of expectation-focused etiological models (with special emphasise on dysfunctional social rejection expectations as connecting link) like
From an ethical point of view, screening for and treatment of emotional distress produced by the paradigm should be enhanced when investigating clinical samples but also healthy controls. Researchers should provide extended debriefing and emotional aftercare by trained psychotherapists in order to prevent clinical subjects from transferring negative social experiences from the paradigm to their real life. Further, they should integrate phases of repeated positive social experiences at the end of their experiments by default in order to compensate for negative social experiences.
No1LikesU! is an ecologically valid and highly standardised experimental paradigm to investigate the development, maintenance, and modification of social rejection expectations. Participants pass multiple short ‘webcam-conferences’ (video stimuli) in which they answer personal questions to different ‘listeners’ (confederates). Afterwards, they receive manipulated social feedback on their self-presentation. Our results suggest that researcher can use No1LikesU! to induce and alter social rejection expectations in healthy participants. Future research should focus on differences in the acquisition, maintenance, and modification of social rejection expectations between healthy and clinical samples. Additionally, incorporating interventions on expectation violation processing might improve the modification of social rejection expectations with implications for clinical treatment.
The supplementary material contains an overview of the 30 questions used in the No1LikesU! paradigm (Appendix A). Questions were adapted from various dating websites to promote positive self-disclosure. Appendix B provides the results of the Level-1- and Level-2-mixed effects models within the multilevel mixed effect multinomial linear regression (for access, see
Winfried Rief is Editor-in-Chief of Clinical Psychology in Europe but played no editorial role for this particular article. Apart from that, the authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.
We thank René Herbstreit for providing technical support with programming the paradigm.
The authors have no funding to report.