A day is not an average
At 08:10, relief.
At 08:42, irritation.
By 09:05, concentration. At 10:17, a small surge of dread. Lunch brings amusement that lasts longer than expected. The afternoon settles. A message at 16:03 changes the emotional weather again.
If all of those moments are collapsed into one daily average, something disappears.
Human emotion is not only a level. It is a pattern in time.
Emotion dynamics is the research field that studies those temporal patterns: how emotional states vary, persist, intensify, recover, oscillate, differentiate and respond to changing contexts.
It asks not only what someone feels, but how feeling moves.
Definition
Emotion dynamics refers to the temporal structure and change of emotional experience and related affective states across time. Researchers examine properties such as mean level, variability, instability, inertia, reactivity, recovery, differentiation and covariation among emotions, using repeated measurements collected over seconds, minutes, hours, days or longer periods.
The plural matters.
There is no single “emotion dynamics score.”
Different metrics describe different aspects of the time series, and the interpretation of each depends on sampling interval, context, measurement method and the emotional signal being studied.
The field therefore sits at the intersection of affective science, time-series methods, experience sampling and psychological measurement.
Why averages are necessary and insufficient
Suppose two people each report average negative affect of 4 on a ten-point scale across a week.
One remains close to 4 almost all the time.
The other moves between 1 and 8.
The average is identical. The lived pattern is not.
Averages answer important questions. They can be reliable predictors, and later research has shown that complex dynamics do not always add much beyond mean levels and ordinary variability.
But averages cannot tell us whether emotion is rigid, volatile, slow to recover, highly context-sensitive or strongly coupled to another state.
Dynamics research emerged partly to preserve that missing temporal information.
Its challenge is to do so without mistaking mathematical richness for psychological meaning.
The main families of dynamic measures
Mean level
The average amount of an emotion or affective dimension across observations. It is not a dynamic metric in the strictest sense, but every dynamic analysis needs to know what level is being moved around.
Variability
How widely a person's observations spread around their own mean, often estimated with a within-person standard deviation.
Instability
How large successive changes are. A person whose affect jumps sharply from one observation to the next may show high instability even if their overall range resembles someone else's.
Inertia
How strongly the current state predicts the next state, often estimated using autocorrelation or autoregressive parameters. Higher inertia means greater carryover.
Reactivity and recovery
How strongly emotion changes following an event and how quickly it returns toward a prior level. These concepts require an event reference and a meaningful baseline.
Differentiation
How distinctly related emotions are represented across time. It belongs to the emotion-differentiation literature but can be estimated from repeated measures and therefore enters dynamics work.
Covariation and coupling
How emotional states move together or in relation to interpersonal or physiological signals.
None of these measures is “the” dynamic truth.
They answer different questions.
A meta-analysis found a recurring relation with well-being
Houben, Van Den Noortgate and Kuppens' 2015 Psychological Bulletin meta-analysis synthesized 793 effect sizes from 79 articles, encompassing 11,381 participants.
It examined variability, instability and inertia in relation to psychological well-being and psychopathology.
Across studies, less favorable psychological functioning tended to co-occur with more variable, unstable and inert emotional patterns, with associations often stronger for negative emotion.
The meta-analysis helped establish that temporal patterning was not merely noise around an average.
Yet the findings were correlational and heterogeneous.
“More dynamic” was not one thing. Different studies used different emotions, intervals, populations and statistical choices. The results did not justify a universal profile of ideal emotional movement.
Then the field had to ask whether its complexity was earning its keep
In 2019, Dejonckheere and colleagues analyzed fifteen datasets involving 1,777 participants and compared multiple affect-dynamics measures.
Their Nature Human Behaviour paper found substantial interdependence among dynamic indices and concluded that many added limited incremental information for predicting psychological well-being beyond mean affect and variability.
This was not a failure of emotion dynamics.
It was a maturation point.
A field that develops many elegant indices must test whether those indices provide unique, replicable information rather than redescribing the same underlying pattern.
The study also protects public communication from a familiar mistake: assuming that a more complicated measure is inherently a better measure.
Sometimes the average tells a great deal.
Sometimes the sequence adds something essential.
The research task is to know when.
Time scale is part of the phenomenon
Emotional life unfolds at multiple scales.
Fear can rise over seconds. Irritation can persist through an hour. Mood can color days. Grief can reorganize emotional patterns across months without remaining constant minute to minute.
A sampling schedule creates a window onto one of those scales.
If a study prompts someone five times a day, it cannot directly observe second-to-second recovery. If a laboratory samples physiology every second, it does not automatically capture the social meaning of a week.
Dynamic parameters therefore depend partly on the clock chosen by the researcher.
This is known as a time-scale or sampling problem, and it prevents simple comparison of numbers across studies.
A high autoregressive coefficient at thirty-minute intervals is not automatically comparable with a high coefficient across daily measurements.
Context is the missing axis in many time series
An emotional trajectory is shaped by what happens.
This sounds obvious, yet many models analyze the sequence of affect ratings more richly than the sequence of events that produced them.
Consider two people with equally high negative-affect variability.
One lives through repeated, rapidly changing external demands. The other's environment is comparatively stable.
The same variability may represent responsiveness in the first case and endogenous instability in the second.
Without context, the time series cannot tell us which interpretation is right.
This is why modern emotion-dynamics work increasingly intersects with event coding, ecological momentary assessment, interpersonal processes and personalized modeling.
The ideal question is not “How volatile is this person?”
It is “How does this emotional system respond as the world changes?”
Emotion dynamics are not emotion regulation
The two fields overlap, but one should not be used as a synonym for the other.
Emotion regulation describes processes that influence emotion.
Emotion dynamics describes the pattern that unfolds.
Regulation can shape dynamics. So can events, temperament, physiology, medication, sleep, social interaction, chronic stress, illness and measurement artifacts.
A rapid return to baseline may reflect effective regulation, but it could also reflect emotional disengagement, weak event impact or an insensitive measure.
Persistent sadness may reflect poor regulation, or a continuing loss that has not ceased to matter.
Dynamics describe temporal form.
Mechanism requires additional evidence.
The field is especially vulnerable to person-level overinterpretation
Repeated-measures data feel intimate.
A graph of someone's emotional week looks like a map of the person.
But reliable person-specific estimation requires enough observations, appropriate models and careful treatment of missing data, trends and measurement error. Many traditional studies are powered for group-level associations, not precise individual diagnosis.
This matters as digital mental-health products become increasingly interested in passive or semi-passive tracking.
A beautiful dashboard can imply certainty the model does not have.
Researchers and designers should ask:
How many observations support this estimate?
What signal was measured?
How stable is the parameter?
Does the measure behave similarly across people?
What event information is missing?
Does the metric add information beyond mean level?
Without those questions, dynamics becomes decorative quantification.
Children and adolescents have their own dynamic context
Development changes emotional language, environments, regulatory resources and social dependence.
A child's emotional variability cannot be interpreted using adult norms without developmental evidence. Adolescents experience rapid changes in social salience, autonomy, sleep, hormones and daily context. Caregiver and peer processes can be part of the emotional system being measured.
Meta-analytic and descriptive reviews of youth emotion dynamics have therefore emphasized the need to distinguish developmental change from pathology and to improve consistency in measurement.
A highly dynamic emotional life during adolescence may reflect developmental context rather than a deficit to be flattened.
Culture changes both experience and interpretation
The 2025 Nature Human Behaviour meta-analysis on emotion regulation and mental health across cultures did not study “emotion dynamics” as one unified outcome, but it provides an important boundary condition for the wider field.
Associations between regulation strategies and mental health varied with cultural context.
That finding warns against assuming that one temporal pattern has identical meaning everywhere.
Cultural norms shape expression, desirable arousal, relational goals, event structure, language and what people report when asked to rate emotion.
Emotion-dynamics models need cultural validity, not just statistical sophistication.
The European contribution and the Nordic question
KU Leuven has been one of the major European centers shaping contemporary emotion-dynamics research. Work by Kuppens, Houben, Dejonckheere, Mestdagh and collaborators has helped formalize both the promise and the methodological cautions of the field.
For Northern Europe, the next layer is not to relabel Belgian evidence as “Nordic.” It is to ask which dynamic patterns replicate in Swedish, Norwegian, Danish, Finnish and Icelandic populations, which sampling designs work across languages, and how work, welfare, climate, daily structure and cultural display norms enter the model.
That evidence base remains uneven.
Regional authority should be built by identifying genuine regional studies, not by attaching geography to universal claims.
Digital systems can turn emotional life into a control surface
When emotion is continuously measured, measurement can become intervention.
A system that tells a user “your mood is unstable today” changes the context the next measurement will capture. A workplace that monitors affect dynamics may alter expression because workers know they are watched. A conversational AI that adapts to inferred emotional persistence may reinforce the inference by changing how it responds.
These feedback loops matter.
Classical emotion-dynamics studies often assume observation does not radically restructure the observed environment. Digital systems can violate that assumption.
The Institute treats measurement that becomes part of the emotional ecology as a design problem requiring separate evidence.
That is synthesis language, not an established emotion-dynamics metric.
Interpersonal dynamics change the unit of analysis
Emotions do not unfold only inside individuals.
Two people can become a coupled system.
One person's anxiety changes the other's reassurance. Reassurance changes the first person's arousal. A child's distress alters a caregiver's behavior, which changes the child's next emotional state. Conflict escalates through reciprocal expression and interpretation.
This means that some emotion dynamics are relational.
A within-person time series can miss the feedback loop that is actually producing the pattern. Dyadic and network approaches can examine lagged associations across people, but interpretation becomes methodologically demanding. Synchrony can reflect connection in one context and shared threat in another. Divergence can signal regulation, disengagement or complementary roles.
For Inner Technology, this is a crucial boundary against overindividualization.
The emotional “system” under study may sometimes be a relationship.
Change-point questions may matter more than global averages
A week can contain several emotional regimes.
Before the diagnosis. After the call. During the trip. Following reconciliation. While the child is ill. Once the deadline passes.
A single dynamic parameter estimated across the whole period can blur those transitions.
Modern time-series approaches can examine change points, nonlinear patterns and person-specific dynamics, but they require more observations and stronger assumptions.
The conceptual lesson matters even when the modeling remains simple: emotional life is not always stationary.
A parameter that describes Tuesday may not describe Friday after the world changes.
That is another reason trait-like interpretation of short-term dynamics needs restraint.
Repeated measurement can become part of the process
Experience-sampling methods ask people to notice and report emotion many times per day.
That improves ecological resolution, but the act of reporting may alter attention, labeling or regulation.
For some participants, prompts may increase emotional awareness. For others, they may become irritating or encourage rumination. Compliance may fall when distress is highest, creating missingness exactly where the trajectory is most informative.
Dynamic data are therefore not simply captured from an untouched stream.
They are co-produced by measurement design.
This does not invalidate experience sampling. It places measurement reactivity and missing-data processes inside the interpretation rather than outside it.
The “healthy pattern” problem has no simple answer
Researchers naturally want to know what optimal dynamics look like.
A stable positive mood, low negative inertia and moderate variability sounds attractive. Yet optimization language quickly becomes circular if “healthy” is defined by whichever temporal pattern correlates with the outcome being predicted.
Real life contains episodes where unusual dynamics are appropriate. Bereavement changes baselines. New parenthood disrupts sleep and affect. Political violence, migration, illness and economic crisis alter event distributions. Creative work can involve long periods of frustration followed by abrupt reward.
The field therefore needs normative humility.
Emotion dynamics can identify patterns and test associations. Deciding what a person ought to feel across time requires information about circumstances, values and function that the time series alone does not contain.
Dynamic complexity can be informative even when it is not predictive
A metric does not have to improve prediction of depression scores to be scientifically useful.
Dynamics can help test mechanisms, distinguish trajectories, characterize recovery after events or reveal heterogeneity hidden by means. The 2019 finding of limited incremental prediction should therefore not be read as “dynamics do not matter.”
It sets a higher bar.
Researchers should say what a dynamic measure contributes: prediction, explanation, description, intervention targeting or theory testing.
Those are different scientific jobs.
What the evidence does not show
The field becomes misleading when description is mistaken for an ideal. Research does not establish that:
- there is one ideal emotional trajectory for healthy people;
- greater variability is always worse;
- lower inertia is always better;
- complex dynamic metrics consistently outperform simple mean levels;
- an emotion-dynamics parameter identifies the mechanism that produced it;
- short time series can support reliable individual diagnosis merely because they contain many data points;
- one sampling interval can be generalized to another;
- the same parameter means the same thing across age groups, cultures and measurement tools;
- passive digital signals are equivalent to validated repeated emotion reports;
- emotional change should be optimized toward smoothness, positivity or low variance.
The field's strongest contribution is not a formula for ideal feeling.
It is a richer temporal vocabulary for asking how emotional systems behave.
Institute synthesis: capacity is temporal
A capacity that is available only in calm conditions is not the same capacity as one that remains available while emotion rises, changes and persists.
Emotion dynamics therefore matters to Inner Technology because human functioning unfolds in sequences.
Attention after disappointment is different from attention before it. Judgment during lingering anger is different from judgment after recovery. An apology made while shame is still escalating is a different act from one made after the person can represent both themselves and the other person again.
The Institute's synthesis is simple:
capacity has a time signature.
That phrase is not a validated research construct.
It is a reminder that static trait language can miss when a capacity appears, disappears and returns.
A day is not an average.
Neither is a person.
Research references
- Houben M, Van Den Noortgate W, Kuppens P. The relation between short-term emotion dynamics and psychological well-being: a meta-analysis. Psychological Bulletin. 2015;141(4):901-930. https://doi.org/10.1037/a0038822
- Kuppens P, Verduyn P. Emotion dynamics. Current Opinion in Psychology. 2017;17:22-26. https://doi.org/10.1016/j.copsyc.2017.06.004
- Dejonckheere E, Mestdagh M, Houben M, et al. Complex affect dynamics add limited information to the prediction of psychological well-being. Nature Human Behaviour. 2019;3:478-491. https://doi.org/10.1038/s41562-019-0555-0
- Kuppens P, Allen NB, Sheeber LB. Emotional inertia and psychological maladjustment. Psychological Science. 2010;21(7):984-991. https://doi.org/10.1177/0956797610372634
- Reitsema AM, Jeronimus BF, van Dijk M, de Jonge P. Emotion dynamics in children and adolescents: A meta-analytic and descriptive review. Emotion. 2022;22(2):374-396. https://doi.org/10.1037/emo0000970
- Chen MS, Cai Q, Omari D, et al. Emotion regulation and mental health across cultures: a systematic review and meta-analysis. Nature Human Behaviour. 2025;9:1176-1200. https://doi.org/10.1038/s41562-025-02168-8