Authors: Aida Mallorquí, Montse Pauta, Gemma Cardona, Claudia Alvarez-Martin, Fernanda Paz-y-Miño, Laura Nogué, Xavier Segú, Victoria Ardiles-Ruesjas, Antoni Rodriguez-Fornells, Antoni Borrell
Categories: Article, Counterfactual thinking, Early pregnancy loss, Trauma, Rumination, IES, RRS-10, Psychology, Human behaviour
Source: Scientific Reports
Authors: Aida Mallorquí, Montse Pauta, Gemma Cardona, Claudia Alvarez-Martin, Fernanda Paz-y-Miño, Laura Nogué, Xavier Segú, Victoria Ardiles-Ruesjas, Antoni Rodriguez-Fornells, Antoni Borrell
Counterfactual thinking (CT), the tendency to consider how events might have been different, is a common cognitive process following negative life events. This longitudinal study examines the prevalence and time-course evolution of CT immediately after early pregnancy loss. A sample of 119 women who experienced early pregnancy loss completed an online psychological assessment measuring traumatic impact, trait rumination, and CT presence, frequency, and content. The survey was administered at one week, one month, and four months post-loss. CT was highly prevalent (72%) and decreased over time. 99% of CT had an upward focus, imagining a better outcome than reality. Moreover, 91.6% were also self-referential, perceived as dependent on one’s own behaviors and emotions, frequently involving a heightened sense of personal responsibility for the loss. Rumination and traumatic impact predicted counterfactual thinking frequency over time, identifying psychological risk patterns. Findings support the idea that CT may serve an adaptive function by helping to process the loss and support future goal setting, including subsequent pregnancy. However, in individuals with high traumatic impact and rumination, CT may contribute to prolonged distress. Integrating psychological care alongside physical healthcare is essential to promoting long-term well-being after early pregnancy loss.
The online version contains supplementary material available at 10.1038/s41598-026-42934-w.
Early Pregnancy Loss (EPL) affects nearly one in every 10 women in their lifetime^1^. Recent research shows an increase in anxiety, depression, and post-traumatic stress disorder (PTSD) following a spontaneous loss^2^. Almost half of the women experiencing a pregnancy loss present adaptive grief reaction with distress declining during the first year^3^. Moreover, nonadaptive patterns of grief have also been identified, including trauma and complicated grief reactions^4^. These mental health consequences may be even overlooked in many societal contexts where preconceptions about pregnancy loss persist, often viewing it as either trivial or shameful (Quenby et al., 2021). Not surprisingly, it has been characterized as unrecognized, undervalued, and stigmatized, and therefore as a form of disenfranchised grief, with a recent mixed-methods study demonstrating the interactive process of disenfranchisement of perinatal grief^5^. Interestingly, the author also shows how shared social discourses and values constrain emotional expression, including the belief that lower gestational age corresponds to less intense grief, a belief that directly affects women who have experienced an EPL. Despite these emotional and social aspects being increasingly documented, little is known about the cognitive processes that support psychological adjustment after an EPL.
Among the cognitive processes that may influence psychological adjustment following an EPL, CT remains largely unexplored. CT refers to the cognitive process by which individuals reconstruct past events by mentally simulating alternative scenarios—imagining how outcomes might have turned out differently^6,7^. Counterfactuals are typically categorized as either upward, when one envisions that an outcome could have been better than actuality, or downward, imagining how an outcome could have been worse. It has been shown that the direction of the comparison can be used differentially to upregulate affective states. While upward CT is usually triggered by negative or traumatic events^8,9^, downward CT or imagining more negative scenarios lead to more positive appraisal of reality and is sometimes used as a strategy for mood regulation.
In the context of loss, studies have shown that upward CT is associated with prolonged grief and depressive symptoms longitudinally, whereas downward CT shows the opposite pattern of effects among participants who experienced a loss within the previous three years^10^ Another well studied aspect of CT is focus (role of the actor in the CT), i.e. whether the mental simulation refers to personal (in)actions (e.g., “if I had eaten more healthy…, then..” ) or another person’s^11^ (e.g., “if only the doctor had detected the problem sooner …, then..”). After bereavement, CT can also be non-referent, with the antecedent referring to a nonspecific target and simply describing ways in which the event could have turned out differently. Self-focused CT has been shown to increase self-blame and has been demonstrated to be a long-term maladaptive coping strategy, exacerbating distress, prolonged grief, and depression^10^.
CT is, in most cases, automatically triggered and intensified after negative emotional or traumatic life events^6,12^. In an early study, Davis et al., (1995) found that nearly 48% of participants involved in a motor vehicle accident in which they lost a child or a spouse engaged in CT, mentally undoing aspects of the event that might have changed its outcome. Importantly, these imagined scenarios might partially attempt to undo the negative outcome by constructing alternative realities (e.g., “what if…" thoughts) where the event could have been avoided. Considering this perspective and its ubiquity in humans, CT is regarded as an important component of problem-solving and learning, as it plays a role in high-level control functions that regulate and monitor behavior to facilitate learning from the past, prevent future failures, and support goal attainment^13–16^. However, the functional value of CT remains controversial, as both beneficial and disadvantageous aspects have been identified^17^ . On the adaptive side, reconstructing scenarios after negative outcomes may support the discovery of antecedent causes, fostering an automatic search for causal attribution and representing an important form of learning. However, when events are irreversible (e.g., the loss of a partner in a car accident), CT may lead to rumination states, repetitive and persistent thoughts that interfere with natural adaptive forms of coping.
The complexity of a negative life event also determines the feasibility of identifying its causes. While simple events (e.g., failing an exam) allow for clear attributions—such as lack of concentration or preparation—multifactorial events like car accidents or an EPL often lack a single identifiable cause. This ambiguity may prompt individuals to attribute responsibility to self-focused and controllable factors (e.g., inadequate health care or nutrition in the case of EPL) rather than to external, uncontrollable ones, thereby increasing feelings of self-blame, regret and guilt but possibly retaining a sense of control over events otherwise beyond one’s influence^10,18,19^. In such cases, the reconstructive process may become prolonged and ruminative, intensifying regret, self-blame , and guilt^10^. Research has shown that persistent rumination on how outcomes “might have been better” is linked to higher risk of negative affect, depression, anxiety disorders, and PTSD^20–23^. CT has been consistently observed in trauma victims, with greater frequency associated with persistent PTSD symptoms. Conversely, CT fluency, the ability to generate multiple counterfactuals, has been linked to adaptive outcomes, such as behavioral planning and goal-setting^24^.
In the context of perinatal losses, CT has only been studied once by Callander et al. (2007) in a sample of 62 women who had experienced at least two miscarriages within the first 24 weeks of gestation^25^. These authors investigated whether CT was linked to psychological distress and if it provided an adaptive cognitive advantage by enhancing future planning and meaning-making. CT fluency (availability of different CTs) and frequency were associated with higher anxiety levels. Moreover, upward and self-referential CTs were the most commonly reported, both contributing to increased anxiety. Conversely, neither meaning-making nor future planning correlated with CT presence, suggesting no adaptive value for this cognitive strategy. An important caveat of this study is that the evaluation of CT was not assessed immediately post-loss and therefore being non-informative on CT presence at the early stage of EPL and later development in time.
Overall, considering the lack of research of CT in pregnancy loss and previous association encountered between CT and mental health^23,26,27^, particularly its persistent prevalence in trauma victims, the present longitudinal research aims to study, for the first time to our knowledge, the activation and time-course evolution of CT starting immediately in the aftermath of an EPL and continuing over a four-month period. By focusing on the psychological traumatic impact and the role of trait rumination (examined separately through its two reflection associated with agency and problem-solving and brooding characterized by passive and negative emotional dwelling) our study aims to provide an overview of how these thoughts are activated and maintained over time after a traumatic event. We hypothesize that higher levels of trauma and rumination, specifically brooding, will predict a higher presence and frequency of CT over the four-month follow-up period.
119 consecutive women were recruited in Clinic Barcelona Maternity Hospital after presenting with an EPL between 5 and 13 weeks of gestation. After providing informed consent, they were contacted via email to complete an online survey containing a standardized psychological assessment intended to measure the impact of the EPL, and three questions concerning CT. Further sociodemographic and reproductive-history data were collected, including age, whether the current pregnancy loss occurred following Assisted Reproductive Techniques (ART), whether participants had previous children, and whether they had experienced previous pregnancy losses. These variables were included due to their known impact on psychological distress and mental health^1,28–32^ and to explore whether they contributed to increased engagement in CT. The present study was part of a broader research addressing the psychological impact of experiencing a EPL and the effects of receiving a cytogenic diagnosis. Surveys were completed 1 week after experiencing the EPL (T0, baseline), one month (T1), and four months after the loss (T2). The study received ethical approval from the Ethics Committee of Hospital Clínic prior to sample recruitment (HCB/2020/0128). Women were excluded from participation if they did not have sufficient proficiency in Spanish or Catalan to read and understand the questionnaires, no other exclusion criteria were applied. All procedures involving participants were conducted in accordance with the ethical standards of the Ethics Committee for Drug Research of Hospital Clínic and with the 1964 Helsinki Declaration and its later amendments.
This study examined three aspects of CT: (1) presence, (2) content, and (3) frequency. To track CT over the first four months post-loss, we assessed these variables at three time points.
At baseline, T1, and T2, participants answered a yes/no question about having experienced “if only…” thoughts related to their pregnancy loss. The question was phrased as follows (English; original Spanish in italics):“Have you ever found yourself ‘If only I hadn’t done X, I wouldn’t have experienced this pregnancy loss?’”«¿Alguna vez se ha encontrado Si no hubiera hecho X, no habría sufrido este aborto/pérdida gestacional?»
Participants who responded “yes” were asked to list the specific “if only…” thoughts they had experienced. The question was phrased as follows (English; original Spanish in italics):“Could you describe these thoughts in more detail?”«¿Podría describir sus pensamientos al respecto?»
Participants rated how often they had such thoughts on a 5-point Likert scale (1 = never to 5 = all the time). At baseline, the rating referred to the period since the loss; at T1 and T2, to the past month. The question was phrased as follows (English; original Spanish in italics):“How frequently have you had these thoughts since your loss / in the past month?”«¿Con qué frecuencia ha tenido estos pensamientos desde su pérdida / en el último mes?»
The IES-R (Impact of the Event Scale-Revised) is a 22-item self-reported measure intended to assess the psychological burden of a traumatic event. It evaluates different symptoms such as intrusion, avoidance and hyperarousal experienced during the past week, associated with PTSD diagnosis. Respondents are asked to identify a specific stressful event and indicate the intensity of the psychological distress experienced according to the statement of each item. In our study, participants were directed to recall the event of the pregnancy loss. IES-R scores range from 0 to 88 with higher scores indicating higher psychological traumatic impact. Subscale scores can also be calculated. Each item is answered using a 5-point Lickert scale. The scale has high internal consistency in postpartum women, with a Chronbach’s alfa of 0.88,^33^. The IES-R was administered at the three time points of the study in order to assess the traumatic impact of the EPL and observe its evolution along the 4-month period.
The Ruminative Response Scale (RRS-10) is a 10 item self-report shorter version questionnaire developed by Nolen-Hoeksema (1991). The RRS-10 comprises 10 items that capture the frequency and intensity of rumination as a personal enduring trait. The scale asks respondents to rate their level of agreement with statements related to their usual cognitive and behavioral responses in sad or upsetting situations. It has two Brooding and Reflection. The first refers to dwelling on the negative consequences distress (e.g. Think “Why do I have problems other people don’t have?”), whereas reflection is a purposeful examination aimed to understand or solve distress (e.g. Analyze recent events to try to understand why you are feeling this way). Each item is answered by using a 4-point Likert scale that measures the frequency and intensity of each of the 10 statements. Each subscale has a range from 5 to 20, with higher scores indicating higher intensity of the trait. A total score can be calculated by summing both subscores, resulting in a total score ranging from 10 to 40. The scale, when administered to Spanish populations, shows appropriate internal consistency (Chrobach alfa of 0.86)^34^. Rumination is considered a significant cognitive style underlying depression and post-partum depression^35^. For the present study, the RRS-10 was administered solely at baseline to estimate trait rumination which was used as a predictive measure of CT.
Regarding the content of the CTs elicited by participants, descriptive analyses were conducted to characterize both the direction of the CTs (upward vs. downward) and their focus (self-referent, i.e., the self is the agent of the imagined alternative scenario; other-referent, i.e., someone else is the agent of the alternative scenario; or neutral, i.e., the counterfactual does not clearly attribute agency to the self or others). In addition, the fluency of CTs generated by each participant across the three time points was analyzed using a repeated-measures ANOVA.
Analyses were conducted using generalized linear models (GLMs), cumulative link models (CLMs), and their mixed-effects counterparts when repeated measures were involved. Continuous predictors were scaled to avoid convergence issues, and backward elimination was applied in all models using likelihood ratio tests (LRTs).
To select predictors and avoid multicollinearity, we first examined correlations between the IES subscales (avoidance, intrusion, hyperarousal), RRS brooding, and RRS reflection. Due to high correlations among IES subscales, we used the total IES score (sum of all items) for modeling. RRS brooding and reflection were retained as separate predictors, given their moderate correlation and potential opposite associations with CT. We also removed one outlier based on age, resulting in a final sample of 118 participants.
Our first aim was to assess whether the presence of CT at T0 (i.e., one-week post-loss) could be predicted by demographic and individual difference Children (yes/no), Previous Losses (yes/no), ART (yes/no), Age, IES, RRS reflection, and RRS brooding. A GLM was performed with a binomial response variable (presence of CT: yes/no). All predictors were included as main effects. The full model Presence CT (yes/no) = Children (yes/no) + Previous Losses (yes/no) + ART (yes/no) + Age + IES + RRS reflection + RRS brooding. LRTs were used to assess the significance of predictors in the final model.
Next, we evaluated whether the frequency of CT at T0 could be predicted by the same variables. Since frequency was ordinal with five levels (“Never” to “All the time”), an ordinal regression was conducted using a cumulative link model. The full model Frequency CT = Children (yes/no) + Previous Losses (yes/no) + ART (yes/no) + Age + IES + RRS reflection + RRS brooding.
We then examined whether the presence of CT changed over time. After removing one outlier from the subsample who completed all three assessments, the final sample for this analysis was 63 participants. A generalized linear mixed model (GLMM) was fitted with a binomial outcome and the following Presence CT (yes/no) = Session + (Session| Subject). Session (T0, T1, T2) was treated as a fixed effect, with random slope for Session and random intercept for Subject to account for individual variation.
We similarly tested whether frequency of CT varied over time using a cumulative link mixed Frequency CT = Session + (Session| Subject).
Finally, we explored whether demographics and individual differences influenced changes in frequency of CT across time. A cumulative link mixed model was specified with double interactions between Session and each predictor (excluding ART, which had shown no effect previously): Frequency CT = Children + Previous Loses + Age + IES + RRS reflection + RRS brooding + Session + ChildrenSession + Previous LossesSession + AgeSession + IESSession + RRS reflectionSession + RRS broodingSession + (Session| Subject).
To assess the robustness of the reported estimates given the limited sample size, we applied bootstrapping to compute confidence intervals for all model parameters. For each model, nonparametric bootstrapping was conducted by resampling the dataset with replacement and refitting the model for each resample. For the GLMs and CLMs, 1,500 bootstrap iterations were used to estimate 95% confidence intervals. For the CLMMs and GLMMs, resampling was performed at the subject level to preserve the random-effects structure. Because some resampled datasets led to convergence issues in mixed-effects models, only successfully converged iterations were retained for the computation of confidence intervals. The number of valid iterations used is reported alongside each confidence interval to facilitate interpretation.
All analyses were performed using R^36^ version 4.4.2 and models were fitted using stats, lme4^37^, and ordinal^38^ packages. Bootstrapping procedures were implemented using the boot package^39^.
A total of 119 women (mean age = 36.6 ± 4.6 years; 45% had experienced previous pregnancy losses; 43% had used ART; and 26% already had children) completed the survey one-week post-loss. Eight participants reported common mental health disorders, such as mild depression or anxiety. Of the total sample, 84 completed the second survey at one-month post-loss. Finally, 63 participants completed all three time-point assessments. The final sample, who completed assessments at all three time points, had a mean age of 36.7 ± 4.4 years. Among them, 40.6% had experienced previous pregnancy losses, 40.6% had used ART, and 25% had children. Five participants who reported common mental health disorders also completed the survey at all three time points.
All CTs provided by participants were upward across the three time-point assessments, with the exception of one participant. Considering the total CT responses generated across the three time points, 91.6% were self-referent (e.g., “If I had paid more attention to my diet and lifestyle, things might have turned out differently,” “If I hadn’t been so stressed during those days,…” or “If I hadn’t worked so hard, the situation might have evolved differently”), 3.1% were other-referent (e.g., “If someone had informed me about the appropriate diet during pregnancy, things might have turned out differently,” or “If the medical team had decided to postpone the transfer,…”), and 18.2% were neutral, mostly referring to general health conditions (e.g., “If I hadn’t had a severe toothache that week, things might have progressed differently,” or “If I hadn’t had COVID previously,…”). These categories were not mutually exclusive, as most participants generated more than one CT, with some CTs belonging to more than one referent type.
Participants provided an average of 2.05 CTs, with a maximum of 8. Among participants who completed all three time points, a statistically significant effect of time was identified, indicating that the number of CTs elicited decreased significantly across assessments, F(2, 62) = 7.28, p = 0.001 (see Fig. 1).Fig. 1Frequency of counterfactuals over time. One N = 118; one N = 84; three N = 63.
At one week post-loss, 72% of participants reported experiencing CT. Examples of CT expressed one week after EPL “I think I should have paid more attention to my diet and lifestyle”, “I think I shouldn’t have been stressed during those days”, and “I think I shouldn’t have worked so hard.” The frequency of CT was distributed as Never (24%), Sometimes (27%), Frequently (23%), Very often (20%), and All the time (6%) (see Fig. 1). When considering only the 63 participants who completed all three time-point assessments, the distribution Never (25.4%), Sometimes (24%), Frequently (20.6%), Very often (22.2%), and All the time (8%; see Fig. S1 for the distribution in the sample without attrition).
At one month post-lost, 56% of the respondents reported to experience CT. The frequency of these thoughts was distributed Never (42%), Sometimes (25%), Frequently (19%), Very often (11%), and All the time (3%; see Fig. 1). Among participants who completed all three assessments, responses Never (43%), Sometimes (22%), Frequently (21%), Very often (13%), and All the time (1%; see Fig. S1).
At three months post-lost, 63% of the women reported the presence of CT. The frequency of these thoughts was distributed Never (40%), Sometimes (37%), Frequently (17%), and Very often (6%; see Fig. 1). Among participants who completed all three assessments, responses Never (40%), Sometimes (37%), Frequently (17%), Very often (6%; see Fig. S1).
To evaluate the presence of CT at one-week post-loss, we followed the backward elimination method for variable selection. We started with the full model introducing sociodemographic and psychological variables [Presence CT = Children + Previous Loses + ART + Age + IES + RRS reflection + RRS brooding]. The final GLM [Presence CT = IES + Previous Loses, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2}
\usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 16.96, p = 3.82 × 10^–5^), indicating that, the more the EPL was experienced as traumatic, the higher the probability of engaging in CT at one week post-loss. In addition, the model showed that women that had already experienced previous loses were significantly more prone to engage in CT (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 4.80, p = 0.028, 95% bootstrap CI [-0.47, 2.13], (see final model effects coefficients in Supplementary Table 1.1). However, the bootstrap analysis indicated that this effect was not robust, as the confidence interval included zero. ### Predicting the frequency of CT at T0 one week post loss When assessing CT frequency, the backward elimination method was used to select the simplest ordinal regression model. We started with the full model [Frequency CT = Children + Previous Loses + ART + Age + IES + RRS reflection + RRS brooding]. The final model [Frequency CT = IES + Age + RRS reflection, \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(3) = 66.39, p = 2.53 × 10^–14^, AIC = 307.32, LL = -146.66; model selection steps in Supplementary Table 1.2] showed a significant effect of IES (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 1.274, p < 0.001, 95% boostrap CI [\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 0.83, 1.76 $$\end{document}]), so that the higher traumatic impact of the EPL, the more frequent CT. Age also had a significant effect (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 0.491, p = 0.006, 95% boostrap CI [\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 0.132, 0.858 $$\end{document}]), indicating that the older the women are, the higher the frequency of CT. Finally, a significant effect of RRS reflection was found (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 0.521, p = 0.013, 95% boostrap CI [\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ 0.059, 1.05 $$\end{document}] (see Fig. 2 and final model effects coefficients in Supplementary Table 1.2), showing that the greater the tendency to use thought to purposeful examine and solve distress, the most frequent CT is.Fig. 2Impact of the event scale, age, and RRS reflection significantly influenced the frequency of CT at one-week post-loss. Note that for the analyses we used the continuous measure of all three variables. For the plot, in order to facilitate the interpretation of the effects, we divided the sample in two groups (i.e., Low and High) using the median-split. This was performed for each of the three variables. In addition, we collapsed Sometimes + Frequently and Very Often + All the Time. ### Assessing the evolution of CT presence over time For the following analyses, only the data from the participants who completed the three surveys (N = 63) was taken into account. To evaluate whether the presence of CT significantly varied over time, we fitted a generalized linear mixed model including Session (i.e. ordered factor with three T0, T1 or T2) as a main effect as well as Session and Participant as random slope and intercept, respectively [Presence CT = Session + (Session|Participant), \documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(2) = 6.46, p = 0.0396, AIC = 164.90, LL = − 73.45]. The GLMM revealed a significant effect of Session (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(1) = 17.95, p = 0.0001), indicating that, indeed, the presence of CT changed over time. ### Predicting the evolution of CT frequency over time First, we assessed whether the frequency of CT varied over time. To that aim, we created a cumulative link mixed model including Session as a main effect as well as Session and Participant as random slope and intercept, respectively [Frequency CT = Session + (Session|Participant), AIC = 460.95, LL = − 218.47]. The model revealed that there was a significant effect of Session (\documentclass[12pt]{minimal} \usepackage{amsmath} \usepackage{wasysym} \usepackage{amsfonts} \usepackage{amssymb} \usepackage{amsbsy} \usepackage{mathrsfs} \usepackage{upgreek} \setlength{\oddsidemargin}{-69pt} \begin{document}$$ {x}^{2} $$\end{document}(2) = − 1.77, p = 0.001), so the more the time passed, the less frequent the occurrence of CT (see Fig. 1). Finally, we tested whether any of the measured variables affected on how the frequency of CT evolved over time. To do so, we fitted a cumulative link mixed model considering the interaction of all our variables with Session and used backward elimination to select the final model. Thus, we started with the full model [Frequency CT = Children*Session + Previous Loses*Session + Age*Session + IES*Session + RRS reflection*Session + RRS brooding*Session + (Session|Participant)]. The final model [Frequency of CT = IES + RRS Reflection + RRS Brooding + Age × Session, AIC = 442.66, LL = − 191.33; model selection steps in Supplementary Table 1.3] showed a significant interaction between Age and Session (χ^2^(1) = 4.21, p = 0.041). However, the bootstrap analysis indicated that this effect was not robust, as the 95% bootstrap CI included zero (95% bootstrap CI [− 1.52, 0.29]. In addition, the model showed significant main effects of IES (χ^2^(1) = 11.50, p < 0.001; 95% bootstrap CI [0.61, 2.23]), RRS Reflection (χ^2^(1) = 4.97, p = 0.039; 95% bootstrap CI [0.30, 1.71]), and RRS Brooding (χ^2^(1) = 4.96, p = 0.030; 95% bootstrap CI [0.24, 1.60]) (see final model coefficients in Supplementary Table 1.3). These results suggest that higher levels of trauma-related distress, reflective rumination, and brooding are all associated with more frequent CT. Importantly, these effects were stable over time, indicating that they consistently influenced CT frequency across sessions (see Fig. 3).Fig. 3Evolution of IES scores of the participants reporting high frequency of CT (very often + all the time) at T0. Green line represents the 33% of the sample with the highest IES scores, in blue the 33% middle IES range and in pink participants with the lowest IES scores at T0. ## Discussion ### General findings Cognitive processes occurring in the context of perinatal loss remain understudied. In the present longitudinal study, we showed for the first time to our knowledge, that the majority of women (72%) engage in CT as early as one week after experiencing an early pregnancy loss (EPL). Interestingly, our data reveal a diminishing trend in the occurrence, frequency and fluency of CT over the first four months, as demonstrated by two statistically significant predictive models and a repeated-measures ANOVA (see Fig. 1, and Fig. S1). Regarding CT types, our sample overwhelmingly revealed upward CT: 99% of participants imagined a better alternative outcome, except for one woman who, after her pregnancy loss, was diagnosed with an arteriovenous malformation and believed the loss may have saved her life because she could have died during delivery. In contrast, all other participants envisioned a more desirable scenario in which the EPL could have been avoided. Our finding of predominantly upward CT in our sample aligns with prior studies indicating that upward CT constitute most daily CT. This tendency is thought to arise from their greater cognitive availability^11,40–43^ and their relative ease of mental simulation^44^. CT focus showed a similar pattern. In our sample, 91.6% of CTs were self-referent, reflecting how deeply women linked their pregnancies to their own actions and mental states. Common statements *“If I hadn’t exercised so much…”*, *“If I had felt happier…”*, *“If I hadn’t been worried and stressed all the time…”*, *“If I hadn’t drunk coffee…”*, or *“If I had tried to conceive earlier…”*. Notably, emotional states were also treated as controllable behaviors, suggesting that participants perceived agency even over their internal experiences. Many participants expressed the belief that they should have felt happier during the pregnancy, and they recalled their negative emotional state as linked to the outcome. One participant *“I think I shouldn’t have gotten upset and nervous, and felt so much anger, which would have prevented the miscarriage.”* Negative feelings were sometimes perceived as potentially harmful for the fetus, comparable to an unhealthy diet or other detrimental habits. The belief that a mother’s psychological state can affect the fetus, as well as the fear of developing a peripartum depression that might have consequences for the child, is widespread and commonly observed in perinatal psychological consultations. Interestingly, even IVF patients expressed upward, self-focused CTs, regretting not requesting additional tests, adjusting medication dosage, or being more careful with self-administered treatments, again predominantly attributing responsibility to themselves. A smaller proportion of CTs (18.2%) were neutral, referring mainly to health or personal conditions (e.g., age, overweight, infections, or hematologic problems) without clear self-blame. Taken together, these findings are in agreement with previous research showing that self-referential CT related to loss events (i.e., attempts to “undo” the death by imagining different personal actions) predicts prolonged grief and depression over time, especially among individuals who experienced a loss within the previous three years. Overall, these results and previous studies suggest the need to explore in the future in more detail the psychological effects of both direction and referent in the formulation of CT^10^. ### CT activation after an EPL In light of these findings, CT activation could be viewed as an adaptive cognitive mechanism that helps individuals cope with negative outcomes during a certain period. Specifically, CT may assist women in processing and integrating this unexpected and impactful experience by prompting them to evaluate their behavior in relation to the circumstances and final negative outcome. Questioning *what could have been done differently*, and focusing specifically on one’s own behavior, may facilitate learning, leading to a revision of life goals and, in this context, even preparation for a future prospective pregnancy. In fact, evidence suggests that 85% of women will conceive after a first pregnancy loss, while 75% will achieve pregnancy following two or more losses^45^. Despite CT’s focus on revisiting past events and actions, CT has been shown to play a role in goal-directed behavior, particularly when life goals are obstructed, creating a sense of distance from an ideal goal state. Through this process, CT may help reduce the discrepancy between one’s current and ideal state^46,47^, suggesting that its activation could serve as an adaptive response to help individuals reorient themselves after significant life obstacles. Although this general perspective aligns with our findings, CT in our study exhibited a more complex and dynamic evolution, interacting with relevant psychological aspects. During the first week following an EPL, the presence of CT was associated with higher trauma intensity, consistent with previous research indicating that CT in trauma survivors is strongly linked to increased psychological distress and higher levels of PTSD symptoms^26^. Furthermore, given the temporal proximity between our measurements and the EPL, our results suggest that the emotional intensity of the trauma itself may be a key element in triggering the counterfactual thought process. Interestingly, most existing trauma and CT studies—primarily cross-sectional—have examined CT in trauma survivors’ years after the traumatic event, limiting insights into its initial function as an early trigger of CT, progression, and resolution over time^26,48^. Our study represents a step forward in this regard, as it is the first to examine the activation and longitudinal evolution of CT in the immediate aftermath of a naturalistic traumatic event. Additionally, the study emphasizes the relevance of exploring the dynamics of CT. Our findings indicate a nuanced process in which CT may function adaptively at first for some individuals—particularly those characterized by high reflection who showed an increased frequency of CT. While both brooding and reflection are components of trait rumination, reflection has been identified as a more adaptive mechanism, fostering insight and promoting more constructive responses to stress. Unlike brooding, reflection alone is not predictive of depression or other negative health outcomes and its deficits are associated with an increased risk of suicidality in individuals with major depression^49–51^. Therefore, the higher frequency of CT in individuals with high reflection following EPL may further support its functional role in reevaluating life goals and behavior after a major life-disrupting event such as an EPL. ### Time-course evolution of CT and implications for mental health Another critical aspect of our study is the longitudinal evolution of CT over the first four months. Our findings indicate that higher traumatic impact, high trait rumination, and both brooding and reflection, are significant predictors of sustained CT frequency over time. Previous research converges with these findings, demonstrating that repetitive CT is predictive of general distress, prolonged grief and depression^10,17^ and is associated with intrusive thoughts and behavioral avoidance among individuals with PTSD^52^. In our study, the intensity of psychological burden of a traumatic event (high IES scores) was clearly associated with higher CT frequency over time, further reinforcing the connection between trauma severity and persistent CT. Taken together, these findings suggest an intricate interaction between frequent CT repetition and trauma symptoms (see Fig. 3), where both processes may mutually reinforce each other, increasing the risk of a maladaptive psychological state. This self-perpetuating negative loop may ultimately lead to prolonged grief or other mental health disorders, such as PTSD. However, the present study design does not allow us to draw such conclusions, and future research should address this putative recurrent process using, for example, cross-lagged panel models. Indeed, CT hyperactivation has been consistently linked to psychopathology and persistent feelings of regret, indicating that the constant repetition of negative thoughts could represent a transdiagnostic signature observable across anxiety and affective disorders, and one that can be effectively targeted through current psychological interventions^10^. In this context, also overall rumination increases the risk of persistent CT. Higher reflection and brooding scores predict greater CT frequency, suggesting that together they may render CT maladaptive. In fact, among currently depressed individuals, the two subtraits did not load as distinct factors but instead merged into a single trait^53^. ### Study limitations This study has several limitations that should be considered when interpreting the findings. First, the study duration was limited to four months, restricting our ability to assess the long-term trajectory of CT and its potential mental health consequences. Second, sample attrition reduced the sample size to 63 participants completing all three time points, which may have affected statistical power and generalizability. In the present study, the generation of CTs was completed by participants on their own, without the supervision of a researcher. Although we believe this approach helped capture more naturalistic thought production, with some fragmented thoughts likely resembling inner-speech formulations^54^, certain examples were not entirely correct or complete in terms of the supposed structure of a CT. For instance, statements such as *“Each time (I got pregnant), I try to do everything I can to prevent a miscarriage, and I keep thinking about what more I could do”* (referring to each pregnancy) were close to a CT but did not fully meet the theoretical criteria and were therefore excluded from analysis. In this regard, eliciting CTs in the presence of a researcher could have facilitated the disambiguation and more precise formulation of certain CTs. Moreover, structure of CTs, defined as the if the antecedent of the CT (*what if* statement) adds something (*if only I had studied…*) or removes something (*If only I hadn’t gone out…*), has been demonstrated, together with upward comparison, to be more impactful in terms of increased persuasiveness, plausibility and likelihood to change future behavior^11^. Future research should consider this relevant aspect in the analysis of CTs in EPL. Additionally, we lack data on how CT is ultimately resolved, how it coexists with individuals over time, and the precise mechanisms through which it contributes to prolonged grief or further mental health disorders. While our findings suggest that CT hyperactivation may accumulate and pose a risk to an individual’s well-being, our data do not allow us to determine exactly when and how this becomes problematic. Consequently, we cannot precisely establish when intervention is necessary or at what point the presence of CT should be considered a concern. Future research should aim to clarify these aspects, identifying when CT hyperactivation transitions from an adaptive response to a psychological risk factor, and defining optimal intervention points to better inform clinical guidelines and therapeutic strategies. ## Conclusions The present study showed that CT is highly prevalent in the aftermath of an EPL, with its presence, fluency, and frequency showing a diminishing trend over the first four months. Our results support an adaptive function of CT—potentially helping individuals process their loss and work toward future life goals, including, for some, the prospect of another pregnancy. However, this function may vary over time, particularly in individuals with higher traumatic impact and a stronger tendency toward rumination. These factors, both of which are significant predictors of CT frequency during the first four months, may also increase the risk of prolonged distress, potentially leading to prolonged grief or further mental health disorders such as PTSD. Given these findings, we believe that CT should be considered in the healthcare counseling and support offered to women experiencing an EPL. Addressing both physical and psychological aspects is essential to helping women navigate this impactful experience and to guiding them toward timely, appropriate care. ## Supplementary Information Below is the link to the electronic supplementary material. Supplementary Material 1.