Participants and recruitment
Sample size was calculated based on the estimated effect size of the change in interpretation bias before and after a CBM-I intervention (d = .43) according to a recent meta-analysis on the field . Following G*Power calculations , the minimum sample size (α set at 0.05, power at 0.95) to find a difference in a repeated-measures multivariate analysis of variance with one within-subjects factor (two time points) and one between-subjects factor (two groups) was 73 participants. 20% more participants will be recruited due to expected attrition .
Participants will be volunteer students from the Complutense University of Madrid who will be recruited via social media channels associated to the university and posters at the Faculty of Psychology. They will be offered course academic credits in exchange for their participation. Exclusion criteria will include any form of visual and/or auditory disability that makes participants unable to follow online sessions and lack of internet access. No restrictions will be placed regarding concomitant treatment during the study, although this information will be monitored during the assessment sessions.
This study is a randomised superiority trial with two parallel groups and a 1:1 allocation ratio. Participants will be randomly allocated (using a macro in excel) to the experimental group (cognitive bias intervention) or a control waiting list group. This control group was chosen due to the novelty of the intervention, for which a waiting list is recommended to get a first impression of its effects . The procedure has been approved by the university ethics committee (Ref. 2018/19 − 017) and has been registered (ClinicalTrials.gov NCT03987477). Moreover, it follows the recommendations for a clinical trial protocol .
Intervention: ‘Relearning how to think’ program
This is a brief online intervention designed to modify negative emotional interpretation biases. The intervention is composed of four different sessions in audiovisual format with psychoeducational content, open-answer questions and exercises to be completed by users. Different cognitive biases, such as jumping to conclusions, mental filter, overgeneralisation and negative attributions are targeted in each session following classical descriptions of biases [7, 42]-see Fig. 1. The organisation of the sessions content was based on the Cognitive Error Rating Scales (CERS) , a manual created for therapists to evaluate cognitive errors during clinical sessions, and the CBM-errors , a clinical strategy to promote more benign interpretations following Beck’s theory .
Each session of ‘Relearning how to think’ is composed of four different parts (see Fig. 2). In part 1, participants receive information about specific interpretation biases and are given examples in video format. Some of these videos are daily life scenes where professional actors represent examples of cognitive biases (following clinical vignettes described elsewhere). During each video (before the resolution of the scene), and in order to increase active involvement, participants have to complete an open-ended question about what could happen in those ambiguous situations. In the second part, users are informed about the risks of using negative interpretations and it is followed by the explanation of the strategies to avoid them (part 3). These strategies are based on “the 4 questions technique” , widely used in clinical practice. This technique involves 4 steps to re-evaluate the negative interpretation of a given situation: finding evidence for the negative thoughts, uncovering the cognitive bias present in the situation, identifying the negative consequences and creating alternative ways of thinking. Finally, during part 4, participants have the opportunity to practice the strategies in an exercise composed of imagery training  followed by negative scenarios in audio format aimed to be reinterpreted. Figure 3 shows the steps of the exercise. It starts with an imaginary training  aimed to make scenarios more vivid to users. Participants are presented a screen saying ‘Close your eyes. Imagine.’ for 1 second followed by a black screen during which a negative scenario is played in audio format with a female voice (e.g. Your partner travels to work by car and normally arrives home promptly every day. Today you notice that they are over an hour late. Your first thought is that there must have been a crash). Audio scenarios are daily life situations where negative interpretations arise, and participants are asked to imagine themselves in those situations. A beep is played for users to open their eyes and start with the exercise questions. First, they have to rate their mood (sadness, happiness, anxiety and anger) on a 10-point VAS scale based on the most frequent emotions experienced in daily life [47, 48]. Then, they are guided to apply the 4 questions technique to each scenario. Finally, users evaluate the degree to which they believe in the new alternative thoughts and emotions generated by the new scenario (sadness, happiness, anxiety and anger).
Cognitive bias measures
Ambiguous Scenarios Test for Depression-II. The AST-D-II  is a self-report measure of interpretation biases. It consists of 15 ambiguous scenarios which participants have to rate on a scale from − 5 (very unpleasant) to 5 (very pleasant). Participants are asked to imagine each scenario as vividly as possible and as if it was happening to them (e.g. As you enter the room, the commission welcomes you and begins with the oral examination. After just a few minutes you know intuitively how the examination will go). Two parallel versions are used in counterbalanced order at baseline, post-treatment and in the follow-ups. Internal consistency has been shown to be good (α = 0.87) for the general scale and also for its two short versions A (α = 0.77) and B (α = 0.78).
Mouse-based (gaze) Contingent Attention Task (MCAT) . A variant of the original Scrambled Sentence Test (SST)  is used to measure interpretation bias while monitoring attention towards emotional stimuli, based on the same principles as in the Eye-gaze Contingent Attention Training (ECAT) . At the beginning of the task, participants have to click a fixation cross at the left side of the screen to elicit natural left-to-right reading patterns. Then, the reading screen is presented with a six-word emotional scrambled sentence (e.g., “am winner born loser a I”) where each word is hidden with a blank mask. Participants are instructed to move the mouse cursor over each mask to read the hidden word and mentally form a grammatically correct sentence using five of them. They are given a time limit of 14 seconds per sentence. This procedure is used to objectively measure attention biases toward emotional words (negative or positive). Then, the answer screen is presented with the six words unmasked for participants to click the order of the sentence they had mentally formed. In this section, participants are given a time limit of 7 seconds. The resulting ratio of correctly negative unscrambled sentences and correctly emotional unscrambled sentences is considered to be the index of automatic negative interpretation bias. To maximise the appearance of biases, at the beginning of the task participants are presented a six-digit number for 5 seconds and told to keep that number in mind during the entire task as they will be asked to retrieve it at the end of the task.
SST-Memory task. As a measure of memory biases, participants are given 5 minutes to remember the sentences they constructed during the SST. Following Everaert et al.’s procedure , the ratio of correctly recalled negative/positive scrambled sentences is a measure of memory bias.
Computerised beads task . The beads task is a measure of probabilistic reasoning which was initially designed to measure jumping to conclusion (JTC) bias in schizophrenic patients . The adapted version used in this study has two parts. The first part consists of presenting two jars with beads of two different colours in different ratios (e.g. 60 orange/40 purple, and 60 purple/40 orange). Participants are told that the program selects one of the jars to take beads randomly out of it and then return them. The instruction is to decide which jar is being used, based on the number of beads of each colour. The second part follows the same procedure with the difference being that the beads are all in white but present two different ratios of positive and negative adjectives (60 positive/40 negative, and 60 negative/40 positive). The number of beads viewed before reaching a decision is an index of jumping to conclusion bias.
Depression, Anxiety and Stress Scale-21. The DASS  is a 21-item self-report questionnaire measuring symptoms of depression, anxiety and stress. Each of the three subscales contains 7 items and provide a different score for each construct. This questionnaire has shown good reliability with the following α values for the Depression, Anxiety and Stress scales, respectively: 0.84, 0.70 and 0.82 .
Patient Health Questionnaire-9. The PHQ-9  is a 9-item self-report questionnaire to assess any present episodes of depression according to the DSM-IV diagnostic criteria. Each item is rated in frequency on a four-point scale from 1 (Not at all) to 4 (Nearly every day). This questionnaire has shown good reliability with a Cronbach's α of 0.89 . An adapted PHQ-9 will also be used to measure past episodes of depression. In this study, the standard diagnostic cutoff score of PHQ-9 ≥ 10  will be used to create groups based on present and past episodes of depression.
GAD-7  is a 7-item self-report questionnaire to assess any present episodes of anxiety according to the DSM-IV diagnostic criteria. Each item is rated on a four-point scale from 0 (Not at all) and 3 (Nearly every day), with the final score being between 0–21. An adapted version of this questionnaire was also used to measure past episodes of anxiety. The cutoff score used in this study to consider present or past episodes of anxiety was GAD-7 ≥ 10, following the severity scale: minimal (0–4), mild (5–9), moderate (10–14) and serious (14–20) .
Pemberton Happiness Index. The PHI  is an 11-item self-report questionnaire measuring general, eudaimonic, hedonic and social well-being. It has shown to have a very good reliability (α = .92).
Dysfunctional Attitudes Scale. The DAS  is a scale of 40 sentences reflecting dysfunctional cognitive schemas. Participants have to rate each sentence from 0 (not applicable to me) to 3 (highly applicable to me). The sum of the scores is an index of stable dysfunctional attitudes (i.e., a measure of cognitive structures). The DAS has been found to be a predictor of major depression  and it has shown to have good reliability α = .70 .
Ruminative Responses Scale. The RRS  (Nolen-Hoeksema & Morrow, 1991) is composed of two subscales measuring rumination cognitive style. For the present study, only the 5-item brooding subscale will be used in order to measure the tendency to ruminate about negative events. The scale has shown good reliability (α = .93) .
The scale for mood assessment-EVEA  is a measure of current mood that participants take immediately before and immediately after each of the sessions of the program. It is included to reflect some possible reactions to the cognitive training procedure. Participants have to rate, from 0 to 10, their current level of anger, happiness, anxiety, depression and boredom. Scores of each subscale (4 items each) are summed up providing an index of emotional change during the session.
Credibility and expectancy questionnaire. The CEQ  is a 6-item measure used to assess the expectancy and rationale credibility of participants regarding the online program they are offered before they start it. It consists of two subscales that measure credibility based on cognition (what you think) and treatment expectancy based on affect (what you feel). Both subscales have shown to have good internal consistency (α = .86 for credibility, α = .90 for expectancy).
The Working Alliance Inventory for Internet interventions (WAI-I)  is a self-report measure to assess alliance in internet interventions. In this study, only the 8-item subscale of task and goal agreement with program was used in order to measure the level of concordance of the program with participants’ interests. This measure was used at the end of the program to know if participants were satisfied with the result. An example of an item is ‘Through the online program I have become clearer about the things I need to do to help improve my situation’. Cronbach’s α for this subscale has been shown to be good (α = .84).
Stressful Events Questionnaire (SE)  is a self-report scale to measure stressful situations that happen to participants between the second assessment and the follow-up (2 weeks and 3 months). The scale includes positive and negative ratings of high impact events as well as daily events related to different contexts (social, emotional, academic/occupational and ‘other’).
Figure 4 shows the schedule of enrolment, intervention and assessment following the recommendations for clinical trials . The main investigator will run a macro in excel for the randomisation of participants to the experimental and control groups. This method assigns a sequence of numbers with a different codification for each group. In any case, participants will be aware of the codification meaning. Participants will receive the questionnaires for the first assessment the day before the first assessment session in the lab, this will be given along with the consent form and an information file about the online program. During this session, the main investigator will explain the rationale of the intervention ‘Relearning how to think’ and will provide participants with the instructions to complete the program. They will need to log into an online platform created for the purpose of this study and create their own account and password to access the material. Information will be coded with the participant’s number and only the main researchers will have access to the data. This data will be monitored by the main investigator as adherence control. When the participants first access the platform, they will find the first session available and only 24 hours later they will be allowed to complete the second one. This intermission between sessions aims to increase participants learning and avoid cognitive overload and boredom.
After a briefing on the intervention, participants will sign the informed consent (explaining their right to discontinue the study at any moment and with no consequences) and complete some demographic information along with the three experimental tasks (MCAT, computerised beads task and SST recall task). At the end of this first assessment, participants will be notified again after approximately 10 days for a second assessment session and the experimental group will be receiving, via email, the link to start the intervention online. The control group will be given the opportunity to complete the intervention right after the second assessment. To improve adherence, both groups will receive a reminder for the second assessment session the day before the date. Finally, those participants who agree will be sent some selected questionnaires for the follow-ups (after 2 weeks and after 3 months). To increase adherence during the follow-ups, participants will be sent up to a maximum of three reminders to complete questionnaires. Both groups will have the opportunity to complete a ‘feedback question’ to give their opinion about the intervention. This question will be included in the post-assessment for the experimental group and at the follow-ups for the control group.
Demographic data and pre-treatment measures will be analysed to test for group differences with analysis of variance and chi squared test for nominal variables.
Complete case analyses will be conducted for those participants who complete all 4 online sessions and attend pre- and post-assessment evaluation sessions. A series of 2 (group: Experimental, Control) x 3 (symptom group: never, present, past) x 2 (time: pre-training, post-training) analyses of variance will be performed to evaluate the change between groups. The symptom group will be created based on present and past symptoms of depression and/or anxiety to explore their influence in the results. Intention-to-treat (ITT) analyses will be conducted with all participants, regardless of session or outcome measure completion. ITT mixed models (restricted maximum likelihood (REML) estimation) will be used to account for missing data . Binary logistic regression will be used to evaluate the assumption that data is Missing at Random (MAR). Exploratory analyses will be conducted to study the interplay between the different cognitive bias scores. Finally, follow-up assessments will include a series of analyses of variance to evaluate group differences in time. All analyses will be performed in SPSS Statistics 20 with an α level of 0.05.