Is violence always visible the moment it starts? Science suggests otherwise, and understanding why could change how we prevent it.
When we think about gender-based violence, we tend to picture a clear, identifiable moment: a sort of line crossed. But research in cognitive psychology is challenging this assumption in a profound way. Violence, it turns out, is rarely a sudden rupture. More often, it is the slow accumulation of something our brains have been quietly trying to explain away.
This article explores a framework rooted in predictive processing, a leading theory in contemporary neuroscience, to understand why gender-based violence so often goes unrecognized in its early stages, and why that delay is not a personal failing. It is, in many ways, a feature of how the human mind works.
The predicting brain: a brief science primer
To understand violence differently, we first need to understand the brain differently.
Modern cognitive science no longer views the brain as a passive receiver of information from the outside world. Instead, it operates as an active prediction machine. At every moment, the brain is generating hypotheses about what is likely to happen next based on past experience, social learning, and cultural context. When reality matches the prediction, everything feels smooth, safe, and coherent. When it doesn’t, a signal fires: a prediction error.
This predictive system evolved to help us navigate complex environments efficiently. In social life, it does the same thing: we unconsciously build models of how people are likely to behave, what emotional responses to expect, what physical boundaries will be respected, and what constitutes “normal” interaction.
These models are mostly invisible to us. They form the background architecture of social trust.
When “something feels off” but we can’t name it yet
Anyone who has experienced the early stages of a controlling or abusive relationship knows the sensation: something feels wrong, but you can’t quite articulate what.
This is not confusion. It is a prediction error signal.
The brain has detected a mismatch between what was expected and what is actually happening: an unexpected shift in tone, an uncomfortable moment of control, a boundary approached in a way that doesn’t feel right. But before this signal becomes a conscious recognition of harm, the mind does something characteristic: it tries to resolve the discrepancy with the least disruptive explanation available (like“I’m probably overreacting.” or “This is just how they are sometimes.”)
In predictive processing terms, the system is choosing to absorb the error rather than update the model. The prediction (that this person, this relationship, this dynamic is safe) is preserved. The discomfort is reinterpreted or minimized.
This is not weakness. It is the brain doing exactly what it is designed to do: seek stability, reduce uncertainty, and maintain coherent social models with the least possible cost.
How normalization happens: the drift of accumulated errors
The clinical and psychological literature on intimate partner violence consistently shows that abuse rarely begins with dramatic or unambiguous acts. It builds gradually through what researchers and advocates now recognize as escalation patterns, but what predictive processing theory helps us understand at a deeper mechanistic level.
Each unresolved prediction error adjusts the baseline slightly. What triggered strong discomfort the first time triggers a little less the second time. The threshold for recognizing harm shifts imperceptibly, because the model of “normal” has quietly expanded to accommodate it.
Over time, a significant gap can open between what is actually happening in the relationship and what the person perceives as within the range of acceptable behavior.
This is the architecture of normalization: the predictable outcome of a brain trying to maintain stability in the face of repeated, unresolved misalignment.
Violence, understood this way, is often not a sudden rupture. It is the end point of a long drift that was happening below the threshold of recognition.
The role of gender norms in shaping what we predict
Social predictions are never neutral. They are built from cultural material and in societies shaped by gender inequality, that material contains significant distortions.
Consider how gendered norms create predictive bias:
- Male assertiveness is culturally scripted as expected, even desirable. When it crosses into aggression, the prediction error may be smaller and therefore easier to dismiss.
- Female compliance is often anticipated in relationship dynamics, making resistance or boundary-setting more cognitively “surprising” and, perversely, more visible and more likely to be framed as a problem.
- Possessiveness and control in romantic relationships are frequently framed, culturally, as expressions of care. The prediction, in other words, is that jealousy signals love which means the brain may not register controlling behavior as a mismatch at all.
This is not a metaphor. Gender norms literally shape the predictive models through which behavior is interpreted. They act as filters on prediction error, determining which signals are amplified and which are suppressed.
This is why prevention work that addresses cultural norms is not supplementary to anti-violence efforts, but it is central to them. Norms that encode harmful dynamics into people’s expectations make it structurally harder to recognize harm early.
Violence as predictive enforcement: understanding perpetrator behavior
The predictive processing framework also offers insight into aggression from the perpetrator’s perspective, not to excuse it, but to understand its internal logic.
Research on coercive control consistently finds that abusive behavior escalates in response to perceived unpredictability in a partner: independence, resistance, emotional autonomy. When a perpetrator’s expectations (about compliance, deference, emotional availability) are not met, violence can function as what we might call predictive enforcement: an attempt to impose a controlled, predictable structure on the relationship through dominance.
This explains why escalation often follows moments of assertion or independence by the person being abused. It is not coincidental. It is structurally consistent with a psychology that uses power to eliminate social uncertainty.
Understanding this pattern has direct implications for safety planning, risk assessment, and the design of intervention programs.
The moment of recognition: when the threshold finally breaks
There is a paradox at the heart of this framework: violence often becomes nameable only after it has been present for some time.
The threshold for recognition, the point at which unresolved prediction errors can no longer be absorbed, is reached when several factors converge:
- Discrepancies accumulate beyond what reinterpretation can contain
- Emotional and physical symptoms intensify
- An external perspective (a friend, a therapist, a support resource) introduces a model that challenges the internal one
- The cognitive and emotional cost of maintaining the current framework becomes unsustainable
At this point, a restructuring occurs: what the person may describe as “suddenly seeing clearly,” even though the pattern has been building for months or years. This is not sudden clarity. It is the delayed output of a long predictive struggle.
Recognizing this temporal dynamic is critical for support services and disclosure contexts. Survivors who took time to name their experience are not unreliable witnesses. They are individuals whose cognitive systems were working exactly as designed, within a cultural context that made harm harder to detect.
What this framework means for prevention and education
Reframing gender-based violence as a failure of social prediction, rather than simply a behavioral event, has significant implications for how we design prevention, education, and support.
1. Early warning signs are not “minor”: they are early signals of misalignment. Programs should help people understand that discomfort, confusion, and the feeling that “something is off” are meaningful data, not overreactions to be suppressed.
2. Somatic and emotional responses deserve clinical legitimacy. The body often registers prediction error before the mind names it. Supporting people in trusting those signals (rather than explaining them away) is a practical prevention tool.
3. Cultural norm change is a neurological intervention. Challenging gender norms is not simply a social justice objective. It is a direct mechanism for recalibrating the predictive models through which people interpret and assess relational safety.
4. Education must address how we normalize, not just what we tolerate. Teaching people what violence looks like is not sufficient if it doesn’t also explain the cognitive processes by which harm gets absorbed and reinterpreted over time.
Conclusion: the most important moment is the one before recognition
Violence is not only what happens. It is also what the mind does with what happens: the interpretations it applies, the expectations it preserves, and the discomfort it learns to accommodate.
Understanding gender-based violence through the lens of predictive processing does not soften its severity. On the contrary, it reveals why it is so effective, so insidious, and so difficult to escape: it exploits the brain’s own drive toward coherence and stability.
The most critical moment in prevention is not when violence becomes undeniable. It is the earlier moment, when something feels off, when a small error has been registered and is waiting to be taken seriously.
That is the moment this framework asks us to pay attention to, because that is the moment that makes everything else possible or preventable.