The research team, led by Ines Culina of the University Hospital Centre and University of Lausanne, together with colleagues from Lausanne, Bordeaux, and Windsor, used a technique called Ecological Momentary Assessment, or EMA. Rather than asking participants to reconstruct their feelings in retrospect — a method notoriously vulnerable to memory bias — EMA pings them on a smartphone at fixed intervals throughout the day, capturing emotional states as they happen. In this study, participants were prompted four times a day for seven consecutive days, yielding a dense time series of each person’s inner weather.

The comparison design is what makes the study particularly valuable. Most previous EMA research on borderline personality disorder had only compared patients with healthy controls, leaving open the question of whether the observed emotional turbulence was truly distinctive of the disorder or simply a residue of any mood-related illness. To address this, the team recruited 23 participants with borderline personality disorder, 18 with remitted major depressive disorder, 18 with remitted bipolar disorder, and 269 individuals drawn from the general population with no mood disorder. The inclusion of remitted mood-disorder groups is critical, because it tests whether the emotional signature of borderline personality disorder persists even when the comparison conditions are clinically stable.

The researchers examined two distinct statistical properties of emotional experience. The first is intensity, operationalized as the mean level of a given state across all assessments — essentially, how strongly someone typically feels sad, anxious, or irritable. The second is instability, the tendency of a state to fluctuate sharply from one moment to the next. Instability was quantified using two complementary indices: within-person variance, which captures the overall spread of a person’s scores around their own average, and the mean square successive difference, which is sensitive specifically to rapid changes between consecutive prompts. Together, these measures distinguish someone who is uniformly miserable from someone whose emotions swing wildly, a distinction that retrospective questionnaires routinely blur.

Statistically, the team compared the borderline group with each of the three other groups using linear mixed models for intensity, an approach that properly accounts for the fact that repeated observations from the same person are correlated. For the instability indices, which are often skewed and bounded at zero, they turned to generalized linear models with a Tweedie distribution and a log link — a flexible family of models well suited to this kind of data. Multiple comparisons between the borderline group and the three comparison groups were adjusted using a single-step method within the framework of generalized linear hypothesis testing, guarding against spurious findings that can arise when many tests are run simultaneously.

The results were striking. Participants with borderline personality disorder reported significantly higher intensity and higher instability of sadness and anhedonia than participants with remitted major depressive disorder, those with remitted bipolar disorder, and the general population controls. In other words, even when depression and bipolar illness are in remission, the emotional lows of borderline personality disorder remain both deeper and more erratic. This is a meaningful clinical signal: it suggests that persistent sadness and pleasurelessness, churning on a timescale of hours, may be a core feature of the disorder rather than a byproduct of comorbid mood illness.

Anxiety told a partially similar story. The borderline group showed higher intensity of anxious states across comparisons, and higher instability as well, although the instability differences were less consistent across the three comparison groups. This nuance matters. It implies that while anxiety is clearly elevated in borderline personality disorder, its moment-to-moment volatility may overlap more with that seen in remitted mood disorders, whereas the volatility of sadness and anhedonia appears more specific. For clinicians, this could eventually help sharpen differential diagnosis, a notoriously difficult task when borderline personality disorder and mood disorders co-occur, as they frequently do.

The study also found that participants with borderline personality disorder displayed higher instability — though not consistently higher intensity — in irritability and agitation relative to population-based controls. Irritability and agitation are states with obvious interpersonal consequences, and their rapid fluctuation may underlie some of the turbulent relationships and impulsive conflicts that characterize the condition. The fact that these differences emerged against general-population controls but were less pronounced against the remitted mood-disorder groups hints at some shared emotional dysregulation across conditions, with borderline personality disorder occupying the extreme end of a spectrum.

Technically, the study demonstrates how modern ambulatory assessment and sophisticated statistical modeling can be combined to dissect psychopathology in real time. The Tweedie distribution, rarely seen in mainstream psychology reporting, is designed for data that mix continuous values with a spike at zero — exactly the pattern often produced by instability indices computed over short time series. By pre-registering a focused set of affective, cognitive, physiological, and psychomotor-activity states and adjusting for multiplicity, the authors aimed to keep the analysis exploratory yet disciplined. The work was embedded in long-running Swiss cohort efforts, including the CoLaus|PsyCoLaus population study and a prospective follow-up of borderline personality disorder supported by the Lausanne University Hospital and the Swiss National Science Foundation, giving the sample unusual depth for a study of this kind.

The broader implications are twofold. Scientifically, the findings bolster the view that borderline personality disorder is fundamentally a disorder of emotional dynamics, not just of emotional level — the signature lies in how feelings rise and fall across hours, a dimension invisible to standard diagnostic interviews. Clinically, EMA-based measurement could eventually move from the research lab into practice, helping therapists track a patient’s emotional volatility week by week and tailor interventions such as dialectical behavior therapy or schema therapy to the moments when instability peaks. The authors caution that their sample sizes for the clinical groups were modest, and that the cognitive and physiological measures were exploratory, so replication in larger cohorts will be essential. Still, the message is clear: to understand borderline personality disorder, science must watch emotions as they live, not ask about them after the fact. This study shows precisely how to do that, and what it reveals is a mind in which sadness and joylessness do not merely run deeper — they move faster, and that movement may be the disorder’s true fingerprint.

Subject of Research: Ecological Momentary Assessment of affective intensity and instability in borderline personality disorder

Article Title: Intensity and instability of affective and cognitive states in borderline personality disorder compared to remitted mood disorders and population-based controls: an Ecological Momentary Assessment study

Article References: Culina, I., Ranjbar, S., Strippoli, M.-P. F., Husky, M. M., Kolly, S., Conus, P., Berney, S., Preisig, M., Glaus, J., & Kramer, U. (2026). Intensity and instability of affective and cognitive states in borderline personality disorder compared to remitted mood disorders and population-based controls: an Ecological Momentary Assessment study. BMC Psychiatry. https://doi.org/10.1186/s12888-026-08688-5

Image Credits: AI Generated

DOI: 10.1186/s12888-026-08688-5

Keywords: borderline personality disorder, Ecological Momentary Assessment, emotional instability, anhedonia, remitted depression, bipolar disorder, anxiety, irritability, within-person variance, mean square successive difference, psychiatric epidemiology, emotion dynamics

The research team approached emotional experience not as a series of isolated ratings but as a dynamic system. In this framework, each distinct emotional state — such as calm, sadness, anxiety, or positive engagement — is treated as a node, and the everyday movement of a person from one state to another becomes a directed connection between nodes. When these connections are aggregated across repeated experience sampling, they form what researchers call an emotion-transition network: a personalized map of how that individual’s feelings tend to follow one another in real life.

The central question of the study was whether the shape and behavior of these transition networks relate to a well-established property of the brain: network variability. The brain does not maintain a fixed pattern of connectivity. Instead, the strength of coupling between large-scale networks — such as the default mode network, the frontoparietal control network, and the salience network — fluctuates over seconds and minutes. This moment-to-moment variability is thought to reflect the brain’s dynamic repertoire, its capacity to reconfigure in response to changing internal and external demands.

Participants in the study contributed two complementary streams of data. In the laboratory or scanner, their brain activity was recorded and analyzed to quantify how much their functional network connectivity varied over time. In daily life, they repeatedly reported their current emotional state, allowing the researchers to reconstruct the probability that one emotion would transition into another. By combining these streams, the investigators could ask a question that neither dataset could answer alone: does a brain that reconfigures more — or less — give rise to a distinctive emotional dynamics?

The findings point to a meaningful correspondence. Individuals whose brains showed particular patterns of network variability tended to display characteristic emotion-transition structures. In people with greater affective symptoms, the transition networks showed signs of rigidity or dysregulation: certain negative emotional states acted as strong attractors, with transitions more likely to remain within negative territory rather than returning to neutral or positive states. In contrast, healthier emotional dynamics were characterized by smoother, more balanced flow between states, with negative emotions less likely to dominate the sequence of daily experience.

This perspective reframes affective symptoms in a technically precise way. Rather than asking simply how intensely a person feels sad or anxious, the transition-network approach asks how emotions are organized in time. Depression, on this view, may involve not only elevated negative affect but a change in the topology of emotional life — a network in which sadness and anxiety are densely interconnected and difficult to exit, while positive states become peripheral nodes visited rarely and briefly. Anxiety may show a related but distinguishable signature, with heightened vigilance states forming tightly coupled clusters that capture the flow of experience.

The brain-side of the equation is equally important. Network variability in functional connectivity has previously been linked to cognitive flexibility, and both unusually high and unusually low variability have been associated with psychopathology, depending on the brain systems involved. The new results suggest that this neural property does not remain confined to the scanner: it appears to shape the statistical structure of emotional transitions in everyday life. A brain whose networks reconfigure adaptively may support an emotional system that can enter and exit states fluidly, whereas constrained or excessive variability may bias the system toward maladaptive sequences.

Methodologically, the study illustrates the growing power of network science in psychiatry. Tools originally developed to analyze social networks, transportation systems, and the internet — measures of node centrality, clustering, and transition entropy — can be applied to the sequence of human feelings. Transition entropy, for example, quantifies how predictable a person’s next emotional state is given their current one. Elevated predictability within negative states, meaning the system keeps returning to the same unpleasant nodes, may be a computationally tractable marker of rumination or emotional inertia, phenomena long described clinically but difficult to measure objectively.

The implications extend toward assessment and intervention. If emotion-transition networks can be estimated reliably from intensive longitudinal data — now feasible with smartphone-based experience sampling — they could serve as digital phenotypes that complement traditional symptom questionnaires. Clinicians might one day track whether an intervention is working not only by asking whether negative feelings have decreased in intensity, but by observing whether the connectivity of the emotional network itself is loosening: whether sadness, for instance, is becoming less likely to trigger further negative states and more likely to give way to neutral or positive ones.

The authors are careful to note that the study establishes associations rather than causal mechanisms. Brain network variability and emotion-transition dynamics were measured in relation to each other, and the direction of influence — whether neural flexibility shapes emotional flow, whether chronic emotional patterns sculpt neural dynamics, or whether both reflect a third factor — remains an open question. Longitudinal designs, and eventually interventions that perturb one side of the system, will be needed to disentangle these possibilities. Replication across larger and more diverse samples will also be essential, as experience-sampling studies are demanding for participants and sample sizes are often modest.

Even with those caveats, the work represents a convergence of two powerful ideas: that mental disorders can be understood as network phenomena, and that the brain’s intrinsic variability is a meaningful individual difference rather than measurement noise. By linking the two, the study offers a bridge from milliseconds of neural reconfiguration to the hours and days over which moods unfold. It suggests that the difference between a mind that moves freely through its feelings and one that circles within them may be visible, quantitatively, both in the statistics of daily emotional transitions and in the ever-shifting choreography of brain networks — a dual signature that could ultimately sharpen how researchers detect, understand, and treat affective illness.

Subject of Research: The relationship between dynamic emotion-transition networks, brain functional network variability, and affective symptoms such as depression and anxiety.

Article Title: Dynamics of emotion-transition networks link brain network variability and affective symptoms

Article References: Geng, L., Tang, S., Tie, B., Wang, X., Wang, Y., Jia, H., Feng, Q., Sun, J., Qiu, J., & Li, Y. (2026). Dynamics of emotion-transition networks link brain network variability and affective symptoms. Communications Psychology. https://doi.org/10.1038/s44271-026-00532-6

Image Credits: AI Generated

DOI: 10.1038/s44271-026-00532-6

Keywords: emotion dynamics, brain network variability, affective symptoms, depression, anxiety, functional connectivity, experience sampling, network neuroscience, emotion-transition networks, mental health, computational psychiatry, emotional flexibility